TRANSPORT INFRASTRUCTURE AND SYSTEMS

TIS2017_Book.indb i 2/8/2017 8:00:27 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 PROCEEDINGS OF THE AIIT INTERNATIONAL CONGRESS ON TRANSPORT
INFRASTRUCTURE AND SYSTEMS (TIS 2017), ROME, ITALY, 10–12 APRIL 2017

Transport Infrastructure and Systems

Editors
Gianluca Dell’Acqua
University of Napoli Federico II, Napoli, Italy

Fred Wegman
Delft University of Technology, Delft, The Netherlands

TIS2017_Book.indb iii 2/8/2017 8:00:28 PM

 On the book cover: Concrete relief entitled “Fari, pistoni e semafori” (1959) located in the building of
the ACI directorate.
Author: Imre Tóth (Fehérvárcsurgó, Austria-Hungary 1909; Rome, Italy 1984).

CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2017 Taylor & Francis Group, London, UK

Typeset by V Publishing Solutions Pvt Ltd., Chennai, India

Printed and bound in Great Britain by CPI Group (UK) Ltd, Croydon, CR0 4YY

All rights reserved. No part of this publication or the information contained herein may be reproduced,
stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical,
by photocopying, recording or otherwise, without written prior permission from the publisher.

Although all care is taken to ensure integrity and the quality of this publication and the information
herein, no responsibility is assumed by the publishers nor the author for any damage to the property or
persons as a result of operation or use of this publication and/or the information contained herein.

Published by: CRC Press/Balkema

P.O. Box 11320, 2301 EH Leiden, The Netherlands
e-mail: [email protected]
www.crcpress.com – www.taylorandfrancis.com

ISBN: 978-1-138-03009-1 (Hbk + USB-card)

ISBN: 978-1-315-28189-6 (eBook)

TIS2017_Book.indb iv 2/8/2017 8:00:28 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Table of contents

Preface xiii
Committees xv

Transport infrastructure
Deformation and strength of asphalt concrete under static and step loadings 3
A.I. Iskakbayev, B.B. Teltayev & C. Oliviero Rossi
Roundabout design guidelines: Case study of Croatia 9
H. Pilko
High modulus asphalt concrete for Ljubljana airport apron 17
A. Ljubič
Assessment of technologies for roadway energy harvesting 21
S. Colagrande & M. Patermo
Loss of life risk due to impacts of boulders on vehicles traveling
along a very busy road 27
P. Budetta, G. Forte & M. Nappi
A big data approach to assess the influence of road pavement condition on truck fleet
fuel consumption 33
F. Perrotta, T. Parry & L. Neves
Transparency and good governance as success factors in public private partnerships 39
C. Queiroz & P. Reddel
SUP&R ITN: An international training network on sustainable
pavements and railways 47
D. Lo Presti, G. Airey, M.C. Rubio & P. Marsac
Climate resilient slope stabilization for transport infrastructures 55
A. Faiz, B.H. Shah & A. Faiz
Monitoring of railway track with light high efficiency systems 63
S. Cafiso, B. Capace, C. D’Agostino, E. Delfino & A. Di Graziano
Reference trajectories of vehicles for road alignment design 71
G. Cantisani & G. Loprencipe
Decoupling of wheel-rail lateral contact forces from wayside measurements 79
D. Cortis, M. Bruner & G. Malavasi
Design and maintenance of high-speed rail tracks: A comparison between ballasted
and ballast-less solutions based on life cycle cost analysis 87
M. Giunta & F.G. Praticò
Instability exposure and risk assessment of strategic road corridors in a geomorphologically
complex territory 95
R. Pellicani, I. Argentiero & G. Spilotro
Safety in the III Valico tunnels 103
A. Focaracci

TIS2017_Book.indb v 2/8/2017 8:00:28 PM

 AllBack2Pave: Towards a sustainable recycling of asphalt in wearing courses 109
D. Lo Presti, G. Airey, M. Di Liberto, S. Noto, G. Di Mino, A. Blasl, G. Canon Falla & F. Wellner
Feasibility and preliminary design of a new railway line in the Dolomites area of Veneto Region 119
M. Pasetto, G. Giacomello, E. Pasquini & A. Baliello
Investigation of the causes of runway excursions 127
N. Distefano & S. Leonardi
A review of sulfur extended asphalt modifiers: Feasibility and limitations 135
S.E. Zoorob, C. Sangiorgi & S. Eskandersefat
In-plant production of warm recycled mixtures produced with SBS modified bitumen:
A case study 143
A. Stimilli, F. Frigio, F. Canestrari & S. Sciolette
Environmental and engineering performance assessment of biofilters and retention
systems for pavement stormwater 153
M. Bassani, L. Tefa, E. Comino, M. Rosso, F. Giurca, A. Garcia Perez,
R. Ricci, F. Bertola & F. Canonico
Slope hazards and risk engineering in the Canadian railway network through the Cordillera 163
J.L. Rodriguez, R. Macciotta, M. Hendry, T. Edwards & T. Evans
Computational model to conform aerodrome geometry to ICAO design standards 169
E.J. Silva & N.D.F. Gualda
A speed model for curves of two-lane rural highways based on continuous speed data 177
R. Almeida, L. Vasconcelos & A. Bastos Silva
Rheological characterization of cold bituminous mastics produced with different
mineral additions 185
C. Godenzoni, M. Bocci & A. Graziani
A comprehensive methodology for the analysis of highway sight distance 193
M. Castro, C. De Santos-Berbel & L. Iglesias
Performance improvements of asphalt mixtures by dry addition of polymeric additives 201
M. Ranieri, C. Celauro & L. Venturini
An application of the parallel gradation technique to railway ballast 209
M. Esposito, F.S. Capaldo, S. Andrisani & B. Festa
Rolling resistance impact on a road pavement life cycle carbon footprint analysis 217
L. Trupia, T. Parry, L. Neves & D. Lo Presti
Intermediate stage traffic technical solution of prince Branimir Street in Zagreb 225
M. Šimun, S. Mihalinac & K. Rehlicki
Research and analysis of dolomite aggregate for resistance to fragmentation 233
L. Šneideraitienė & D. Žilionienė
Design vehicles and roundabout safety—review of Croatian design guidelines 237
Š. Bezina, I. Stančerić & S. Ahac
Laboratory investigation into the modification obtained of paving bitumen with
different plastomeric polymers 245
E. Saroufim, C. Celauro & N. Khalil
Long-term monitoring of half warm mix recycled asphalt containing up to 100% RAP 253
J.M. Lizárraga, A. Ramírez, P. Díaz, M. Martín & F. Guisado
Fiscal commitments to encourage PPP projects in transport infrastructure 261
N. Nose, C. Queiroz & M. Nose
Mix design and volumetric analysis of hot recycled bituminous mixtures using a bio-additive 267
E. Bocci, F. Cardone & A. Grilli

vi

TIS2017_Book.indb vi 2/8/2017 8:00:28 PM

 A water curtain-based wayside protection system for tunnel fire safety 275
S. Terribile, G. Malavasi, P. Firmi & A. Pranno
3D control of obstacles in airport location studies 283
D. Gavran, S. Fric, V. Ilić, F. Trpčevski & S. Vranjevac
Environmental and economic assessment of pavement construction and management
practices for enhancing pavement sustainability 291
J. Santos, V. Cerezo, G. Flintsch & A. Ferreira
Road route planning for transporting wind turbines in Europe 301
F. Autelitano, E. Garilli & F. Giuliani
The method of the friction diagram: New developments and possible applications 309
P. Colonna, N. Berloco, P. Intini & V. Ranieri
The relationships between familiarity and road accidents: Some case studies 317
P. Intini, P. Colonna, N. Berloco, V. Ranieri & E. Ryeng
Geotechnical asset management for Italian transport agencies: Implementation principles
and concepts 325
P. Mazzanti, P.D. Thompson, D.L. Beckstrand & D.A. Stanley
Evaluation of workability of warm mix asphalt through CDI
parameter and air voids 335
D.M. Mocelin, L.A.T. Brito, M.G. Johnston, V.S. Alves, G.B. Colpo & J.A.P. Ceratti
A laser profilometer prototype for applications in road pavement management system 343
G. Cerni & A. Corradini
Speed management in single carriageway roads: Speed limit setting through expert-based system 351
N. Gregório, A. Bastos Silva & A. Seco
Effects of mineral fillers on bitumen mastic chemistry and rheology 359
R.M. Alfaqawi, G. Airey, D. Lo Presti & J. Grenfell
Functional and durability properties evaluation of open graded asphalt mixes 365
V.M. Kolodziej, G. Trichês, J.S. Ledezma, G.C. Carlesso, L.M. Jardín & R.M. Knabben
Re-use in asphalt pavements of fillers from natural stone sawmilling sludge 371
E. Santagata, O. Baglieri, L. Tsantilis, G. Chiappinelli & P.P. Riviera
Risk compensation in a changing road environment 379
A. Oikonomou, P. Tafidis, P. Kyriakidis, S. Basbas & I. Politis
A PSO algorithm for designing 3D highway alignments adopting polynomial solutions 385
G. Bosurgi, O. Pellegrino & G. Sollazzo
QEMS. ANAS quality evaluation management system 395
E. Cesolini, S. Oddone & S. Gregori
Innovative approaches to implement road infrastructure concession through
Public-Private Partnership (PPP) initiatives: A case study 403
M. Arata, M. Petrangeli & F. Longo
Highway functional classification in CIS countries 411
V.V. Silyanov & J.I. Sodikov
Efficient practices in railway ballast maintenance and quality assessment using GPR 419
L.B. Ciampoli, A. Calvi, A. Benedetto, F. Tosti & A.M. Alani
BIM experience in infrastructural large projects: Doha Metro—Al Jadeda
Station al Matar B case study 425
A. Ingletti, M. Scala & L. Chiacchiari
The pavement management system for improving airports and roads infrastructures 435
G. Battiato, L. Cosimi & C. Stratakos

vii

TIS2017_Book.indb vii 2/8/2017 8:00:28 PM

 The use of porous asphalt for the improvement of the grading plan geometry and drainage
of pavement surfaces on urban roads 443
V. Ilić, M. Orešković, D. Gavran & I. Pančić
Urban tracks for Light Rail Transit (LRT). Dublin experience 451
M. Corsi
Study on paving factors of stone path in campus based on characteristics
of pedestrian walking 461
Y. Yanqun, H. Qinwei, X. Meiling & Y. Sheng
Performance evaluation of regenerated HMA with very high RAP content 469
N.A. Alvarado Patiño, J. Martinez & E. Lopez
Finite element modelling of flexible pavement reinforced with geogrid 477
G. Leonardi & R. Palamara
Critical success factors for PPP infrastructure projects in India 485
R.P. Pradhan, A. Verma, S. Dash, R.P. Maradana & K. Gaurav
An on-ramp metering and speed control application on O1 freeway in Istanbul 493
H.G. Demir & Y.K. Demir
Transport infrastructure capacity calculations using green travelling planner 499
I. Celiński, M. Staniek & G. Sierpiński
The influence of road marking, shape of central island, and truck apron on
total and truck accidents at roundabouts 509
J. Kamla, T. Parry & A. Dawson
A runway veer-off risk assessment based on frequency model: Part I. probability analysis 515
L. Moretti, G. Cantisani, P. Di Mascio, S. Nichele & S. Caro
A runway veer-off risk assessment based on frequency model: Part II. risk analysis 523
L. Moretti, G. Cantisani, P. Di Mascio, S. Nichele & S. Caro
Risk factors that increased accident severity at US railroad crossings from 2005 to 2015 529
E. Dabbour
Development of a road asset management system in Kazakhstan 537
G. Bonin, S. Polizzotti, G. Loprencipe, N. Folino, C. Oliviero Rossi & B.B. Teltayev
Building Information Modeling (BIM): Prospects for the development of railway
infrastructure industry 547
M. Leone, A. D’Andrea, G. Loprencipe, G. Malavasi & L. Bernardini
Dissipated energy approach for the fatigue evaluation of RAP asphalt mixtures 555
M. Pasetto & N. Baldo
Correlation between risk level and operating speed of road section 563
D. Bellini, M.C. Iaconis & M. Rossi

Transport systems
Implementing innovative traffic simulation models with aerial traffic survey 571
A. Marella, A. Bonfanti, G. Bortolaso & D. Herman
Social exclusion and high-speed rail: Some evidence from three European countries 579
F. Pagliara, L. Biggiero & F. Menicocci
Modeling airport noise using artificial neural networks and non-linear analysis 587
M. De Luca, D.C. Festa & G. Guido
A new safety performance index for speed-related crashes 593
G. Guido, V. Astarita, A. Vitale, V. Gallelli & F.F. Saccomanno

viii

TIS2017_Book.indb viii 2/8/2017 8:00:28 PM

 Monitoring the transport sector within the sustainable energy action plan: First results
and lessons learnt 601
I. Delponte
A model to evaluate heavy vehicle effect in motorway traffic flow 609
R. Mauro, S. Cattani & M. Guerrieri
Quality and energy evaluation of rail-road terminals by microsimulation 617
A. Carboni & F. Deflorio
New features of Tritone for the evaluation of traffic safety performances 625
V. Astarita, V.P. Giofrè, G. Guido, A. Vitale, D.C. Festa, R. Vaiana, T. Iuele, D. Mongelli,
D. Rogano & V. Gallelli
Genetic algorithm-based calibration of microscopic traffic simulation model
for single-lane roundabouts 633
O. Giuffrè, A. Granà, M.L. Tumminello & A. Sferlazza
More accessible bus stops: Results from the 3iBS research project 641
M.V. Corazza, A. Musso & M.A. Karlsson
Shaping the future of buses in Europe: Outcomes and directions from the EBSF
and EBSF_2 projects 651
M.V. Corazza, A. Musso, U. Guida & M. Tozzi
Waterfront and sustainable mobility. The case study of Genoa 661
P. Ugolini, F. Pirlone, I. Spadaro & S. Candia
New forms of mobility for an alternative territorial fruition: The rediscovery of tourist footpaths 669
R. Papa & R.A. La Rocca
Terminal operator liability 677
F. D’Orsi
Information provision in public transport: Indicators and benchmarking across Europe 681
M. Pirra, M. Diana & A. Castro
Public transport resilience during emergency: A simulated case in Torino 689
F. Deflorio, H.D. Gonzalez Zapata & M. Diana
The “form” of railways: Between construction and regeneration of contemporary landscape 697
F. Viola & C. Barbieri
Modelling ambulance and traffic behaviour using microsimulation:
The LIFE project application 705
F. Galatioto, V. Parisi, E. McCormick & C. Goves
Development and application of national transport model of Croatia 713
G. Pretnar, U. Reiter, I. Majstorović & A. Olmeda Clemares
The effects of the different uses of mobile phones on drivers: Experimental tests
on an extra-urban road 723
S. Zampino
Back-propagation neural networks and generalized linear mixed models to investigate vehicular
flow and weather data relationships with crash severity in urban road segments 731
L. Mussone, M. Bassani & P. Masci
The influence of pedestrian crossings features on driving behavior and road safety 741
A. Bichicchi, F. Mazzotta, C. Lantieri, V. Vignali, A. Simone, G. Dondi & M. Costa
The relationship between driving risk levels and drivers’ personality traits,
physical and emotional conditions 747
L. Eboli, G. Mazzulla & G. Pungillo

ix

TIS2017_Book.indb ix 2/8/2017 8:00:28 PM

 The impact of a bus rapid transit line on spatial accessibility and transport equity:
The case of Catania 753
N. Giuffrida, G. Inturri, S. Caprì, S. Spica & M. Ignaccolo
Shared infrastructures: Technique and method for an inclusive social valuation 759
S. Miccoli, F. Finucci & R. Murro
Automatic calibration of microscopic simulation models for the analysis of urban intersections 769
S. Oliveira, L. Vasconcelos & A. Bastos Silva
On-street parking management: Bridging the gap between theory and practice 777
C. Piccioni, M. Valtorta & A. Musso
Personality and driver behaviour questionnaire: Correlational exploratory study 787
J.F. Dourado, A.T. Pereira, V. Nogueira, A.M.C. Bastos Silva & A.J.M. Seco
Motorways of the sea: An outlook of technological, operational and
economic tools 795
R.G. Di Meglio, G. Mainardi, S. Ferrini & I. Toni
Assessing sustainability of road tolling technologies 803
D. Glavić, M. Milenković, A. Trpković, M. Vidas & M.N. Mladenović
Traffic impact of automated truck platoons in highway exits 811
R. Mesa-Arango & A. Fabregas
The potential of energy-efficient driving profiles on railway consumption:
A parametric approach 819
M. Gallo, F. Simonelli & G. De Luca
The importance of railways for developing countries. The feasibility study for the rehabilitation
of the Mchinji to Nkaya rail line in Malawi 825
A. Placido, M. Marchesini & V. Cerreta
Identifying driving behaviour profiles by using multiple correspondence analysis
and cluster analysis 835
D.S. Usami, L. Persia, M. Picardi, M.R. Saporito & I. Corazziari
Solution to a capacity problem using non-conventional intersections 843
M. Vsetecka, M. Novak & T. Apeltauer
A cycling mobility study case: The European cycling challenge 2015 in Naples 851
C. Aveta & M. Moraca
Secondary or no longer used rail infrastructures: A new tool for the regional railway
and land planning 859
A. Cappelli, G. Malavasi & S. Sperati
Delay estimation on a railway-line with smart use of micro-simulation 867
F. Cerreto, S. Harrod & O.A. Nielsen
Smart cyber systems incorporating human-in-the-loop towards ergonomic and sustainable
transport systems 875
F. Galatioto, C. Salvadori, M. Petracca, J. Garcia Herrero, M.J. Santofimia & A. Pollini
Development, analysis and prediction of efficiency of bulk carriers in the world market 883
S. Galić, T. Stanivuk & A. Marušić
Enhanced financial analysis to evaluate mass transit proposals in terms of contribution
on resilience increasing of urban systems 891
A. Spinosa
A reading key of motorisation trend in Italy 899
E. Garilli, F. Autelitano, F. Giuliani, A. Guga & G. Maternini
Competition between European Mediterranean containers transhipments ports 905
E. Twrdy & M. Batista

TIS2017_Book.indb x 2/8/2017 8:00:28 PM

 Using the DataFromSky system to monitor emissions from traffic 913
V. Adamec, B. Schullerova, A. Babinec, D. Herman & J. Pospisil
Structuring transport decision-making problems through stakeholder engagement: The case
of Catania metro accessibility 919
M. Ignaccolo, G. Inturri, N. Giuffrida, M. Le Pira & V. Torrisi
A demand based route generation model for transit network design 927
N. Oliker & S. Bekhor
Multi-objective network design problem considering system time minimization
and road safety maximization 931
I. Haas & S. Bekhor
The impact of road networks in urban distribution spatial 939
L.M. Jardín, C. Loch & V.M. Kolodziej
Validation of pedestrian behaviour scale in Belgrade 947
B. Antić, D. Pešić, N. Milutinović & M. Maslać
The architecture of an information system for the power management system on ship 953
M. Krčum & A. Gudelj
Multimodal accessibility and the interest towards inter-urban carsharing services:
A behavioural approach 959
S. de Luca, R. Di Pace, L. Elia & F. Martire
Sustainable transport—parking policy in travel demand management 969
D. Brčić, M. Šoštarić & K. Vidović
Traffic flows merging cooperation analysis 975
D. Nemchinov, A. Mikhailov, P. Pospelov, A. Kostsov & D. Matiyahin
The relaunch of branch lines as a territory project for the inner areas 979
F. Alberti
Total management tool oriented to carbon footprint reduction in terminals of containers 987
F.E. Santarremigia, G.D. Molero, M.D. Esclapez & S. Awad-Núñez
Pedestrian paths: Safety and accessibility for all users 997
E. Pagliari & S. Balestrieri
Designing principles of the efficient and safe urban cycling infrastructure 1005
Y.V. Trofimenko, E.V. Shashina & I.V. Markin
Analysis of the maritime traffic in the central part of the Adriatic 1013
Z. Lušić, D. Pušić & D. Medić
Study on distribution of bicycle sharing system facility based on Markov chain model 1021
R. Liu, J.F. Dai, J.X. Lin, Q. Shi, Z. Sun, F. Yang & J.C. Huang
A study on the influence of parking policy mix on travel model choice behavior within
the scope of heterogeneous travel purposes 1029
H. Yan, M. Ran & M. Li
Classification and recognition model for the severity of road traffic accidents 1037
J. Xi, L. Wang, T. Ding & W. Sun
Research on the method of multi-resource traffic data fusion for the calculation of urban
road traffic operation index 1045
S. Zeng & H. Xiang
The optimed project: A new Mediterranean hub-based Ro-Ro network 1051
P. Fadda, G. Fancello, C. Pani & P. Serra
Evaluating the capability of deterministic and stochastic simulation tools to model
oversaturated freeway operations 1059
D. Jolovic, P. Martin, A. Stevanovic & S. Sajjadi

xi

TIS2017_Book.indb xi 2/8/2017 8:00:28 PM

 Evaluation method of urban intersection safety under mixed traffic 1065
T. Chen, M. Li, L. Sun, B. Chen & Y. Gao
Sustainable development of the transport system through rationalization of transport
tasks using a specialised travel planner 1071
I. Celiński, G. Sierpiński & M. Staniek
Tram and light rail transit in modern urban mobility systems 1081
R. Emili
A methodology for evaluating the competitiveness of Ro-Ro services 1089
V. Marzano, A. Papola & F. Simonelli
Airport ground access logit choice models for fixed track systems 1099
I. Politis, P. Papaioannou & G. Georgiadis
A simulation model for managing port operations 1109
S. Gori & M. Petrelli
Air transport hinterland in Adriatic and Ionian region: Equity and connectivity matters 1115
O. Čokorilo
Which role for tramways in the next years 1123
G. Mantovani

Author index 1131

xii

TIS2017_Book.indb xii 2/8/2017 8:00:28 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Preface

The Transport Infrastructure and Systems TIS International Congress aims to promote and discuss effi-
cient planning, design, construction and maintenance of transport infrastructure and systems by address-
ing important issues related to roads, railways, airports, maritime and intermodal systems.
The 1st TIS International Congress organized by the Italian Association for Traffic and Transport
Engineering AIIT will be held April 10–12, 2017 in Rome. The TIS 2017 International Congress is focused
on emerging technologies to enable smarter, greener and more efficient movement of people and goods
around the world.
The congress aims to present and discuss current knowledge of the ever-changing challenges to sci-
entists, engineers, managers, and professionals from around the world who are involved in sustainable
development and maintenance of transport infrastructure and systems.
In many countries, transport infrastructure is currently strained to a state of limited functionality.
Increased ridership and depleting funds have caused many Agencies to become concerned with the future
of facilities. Much maintenance is required on infrastructure, and their capacity is in desperate need of
expansion.
Along with the declining state of transport infrastructure is a growing concern for the need to recognize
and mitigate mankind’s impact on the natural environment.
A more sustainable solution to transport infrastructure and systems problems must be assessed. Infra-
structure can have a large negative impact on surrounding ecosystems and overall environmental quality.
The next step in infrastructure’s advancement needs to include practices that reduce their effect on the
natural environment, increase capacity, and benefit society beyond the ability of current infrastructure.
This can be achieved by thinking to transport infrastructure that mitigate the negative impact on the envi-
ronment, include more sustainable practices than modern construction techniques, and consist of maxi-
mizing the lifetime of an infrastructure. Sustainable construction techniques include the use of recycled
materials, ecosystem management, energy reduction, increasing the water quality of storm water runoff,
and maximizing overall societal benefits. Dynamic and liveable cities rely on efficient mobility systems,
and road safety plays a large part in this.
To meet the transport infrastructure and systems needs of future generations, new goals must be set.
Research will have to be given more attention, and a new generation of infrastructure designs will have
to be developed. The TIS 2017 will provide such a forum for new concepts and innovative solutions. The
meeting program will cover all transportation modes, with more than 150 presentations in over 20 sessions
and workshops, addressing topics of interest to policy makers, administrators, practitioners, researchers,
and representatives of government, industry, and academic institutions.
This proceedings book includes submissions to the congress in the areas of Asset management in trans-
port infrastructure, financial viability of transport engineering projects/Life cycle Cost Analysis, Life-Cycle
Assessment and Sustainability Assessment of transport infrastructure/Infrastructures financing and pric-
ing with equity appraisal, operation optimization and energy management/Low-Volume roads: planning,
maintenance, operations, environmental and social issues/Public-Private Partnership (PPP) experience in
transport infrastructure in different countries and economic conditions/Airport Pavement Management
Systems, runway design and maintenance/Port maintenance and development issues, technology relat-
ing to cargo handling, landside access, cruise operations/Infrastructure Building Information Modelling
(I-BIM)/Pavement design and innovative bituminous materials/Recycling and re-use in road pavements,
environmentally sustainable technologies/Stone pavements, ancient roads and historic railways/Cemen-
titious stabilization of materials used in the rehabilitation of transportation infrastructure/Sustainable
transport and the environment protection including green vehicles/Urban transport, land use develop-
ment, spatial and transport planning/Bicycling, bike, bike-sharing systems, cycling mobility/Human fac-
tor in transport systems/Intelligent Mobility: emerging technologies to enable the smarter movement of
people and goods/Airport landside: access roads, parking facilities, terminal facilities, aircraft apron and

xiii

TIS2017_Book.indb xiii 2/8/2017 8:00:28 PM

 the azdjacent taxiway/Transportation policy, planning and design, modelling and decision making/Trans-
port economics, finance and pricing issues, optimization problems, equity appraisal/Road safety impact
assessments, road safety audits, the management of road network safety and safety inspections/Tunnels
and underground structures: preventing incidents-accidents mitigating their effects for both people and
goods/Traffic flow characteristics, traffic control devices, work zone traffic control, highway capacity and
quality of service/Track-vehicle interactions in railway systems, capacity analysis of railway networks/
Risk assessment and safety in air and railway transport, reliability aspects/Maritime transport and inland
waterways transport research/Intermodal freight transport: terminals and logistics.
At least two, but often three reviewers, including members of the Scientific Committee, subjected all
submitted contributions to an exhaustive refereed peer review procedure. Based on the reviewers’ recom-
mendations, those contributions which best suited the conference goals and objectives were chosen for
inclusion in the proceedings.
The Editors would like to thank the Scientific Committee members and individual reviewers for their
dedication and contributions of their time and efforts to ensure the high technical quality of the accepted
papers. In addition, sincere thanks are extended to Salvatore Antonio Biancardo for ensuring that final
manuscripts were in accordance with the publication format requirements.
The guidance and continuing input from the AIIT Committee members were essential in planning of
this congress, and highly appreciated. Finally, we would like to gratefully acknowledge the Organizing
Committee members for their help, suggestions and contributions to the management of the Congress
affairs.

April 10, 2017

Rome

Gianluca Dell’Acqua and Fred Wegman

Chairs of the TIS International Congress 2017

xiv

TIS2017_Book.indb xiv 2/8/2017 8:00:28 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Committees

SCIENTIFIC COMMITTEE

Gianluca Dell’Acqua, Chairman

Fred Wegman, co-Chairman

(by alphabetical order)

Andrus Aavic, Tallinn University of Technology, Estonia

Vladimír Adamec, Brno University of Technology, Czech Republic
Tomas Apeltauer, Brno University of Technology, Czech Republic
John Andrews, The University of Nottingham, UK
Orazio Baglieri, Polytechnic of Torino, Italy
Marco Bassani, Polytechnic of Torino, Italy
Ana Maria César Bastos Silva, University of Coimbra, Portugal
Shlomo Bekhor, Technion—Haifa, Israel
Salvatore Antonio Biancardo, Federico II University of Napoli, Italy
Werner Brilon, Ruhr-University Bochum, Germany
Paolo Budetta, Federico II University of Napoli, Italy
Francesco Canestrari, Polytechnic University of Marche, Italy
Giuseppe Cantisani, Sapienza University of Roma, Italy
Armando Cartenì, Federtico II University of Napoli, Italy
Maria Castro Malpica, Technical University of Madrid, Spain
Christine Chaloupka-Risser, FACTUM, Austria
Vicent de Esteban Chapapría, University of Valencia, Spain
Jack C.P. Cheng, The Hong Kong University of Science and Technology, Hong Kong
Pasquale Colonna, Polytechnic of Bari, Italy
Olja Čokorilo, University of Belgrade, Serbia
Mauro Cozzolino, University of Salerno, Italy
Kevin Cullinane, University of Gothenburg, Sweden
Antonio D’Andrea, Sapienza University of Roma, Italy
Essam Dabbour, Abu Dhabi University, United Arab Emirates
Mario De Luca, Federico II University of Napoli, Italy
Gianluca Dell’Acqua, Federico II University of Napoli, Italy
Miguel Angel del Val Melus, Tech. University of Madrid, Spain
Paola Di Mascio, Sapienza University of Roma, Italy
Coenraad Esveld, ECS BV, The Netherlands
Nicolau Dionísio Fares Gualda, University of São Paulo, Brasil
Bruna Festa, Federico II University of Napoli, Italy
Demetrio Carmine Festa, Calabria University, Italy
Gaetano Fusco, Sapienza University of Roma, Italy
Fabio Galatioto, Transport Systems Catapult, UK
Felice Giuliani, University of Parma, Italy
Hercules Haralambides, Erasmus University Rotterdam, The Netherlands
Imine Hocine, Inst. of Science and Tech. for Transport, France
Felix Huber, Wuppertal University, Germany
Matteo Ignaccolo, University of Catania, Italy

xv

TIS2017_Book.indb xv 2/8/2017 8:00:28 PM

 Renato Lamberti, Federico II University of Napoli, Italy
Davide Lo Presti, The University of Nottingham, UK
Giuseppe Loprencipe, Sapienza University of Roma, Italy
Roberto Maja, Polytechnic of Milano, Italy
Gabriele Malavasi, Sapienza University of Roma, Italy
Giulio Maternini, University of Brescia, Italy
Raffaele Mauro, University of Trento, Italy
Kazuaki Miyamoto, Tokyo City University, Japan
Goran Mladenovic, University of Belgrade, Serbia
Charles Musselwhite, Swansea University, UK
G.P. Jayaprakash, Transportation Research Board, USA
David Jones, University of California Davis, USA
Maja Krcum, University of Split, Croatia
Raimundas Junevicius, Vilnius Gediminas Tech. University, Lithuania
Robert Parsons, University of Kansas, USA
Hrvoje Pilko, University of Zagreb, Croatia
Ioannis Politis, Aristotle University of Thessaloniki, Greece
Carlo Prato, The University of Queensland, Australia
Thanos Pallis, University of the Aegean, Greece
Marco Pasetto, University of Padova, Italy
Luca Persia, Sapienza University of Roma, Italy
Anand J. Puppala, The University of Texas at Arlington, USA
Cesar Queiroz, former World Bank Highways Adviser, USA
Thomas Kjær Rasmussen, Technical University of Denmark, Denmark
Danijel Rebolj, University of Maribor, Slovenia
Stefano Ricci, Sapienza University of Roma, Italy
Ralf Risser, FACTUM, Austria
Guo Runhua, Tsinghua University, China
Francesca Russo, Federico II University of Napoli, Italy
Giuseppe Salvo, University of Palermo, Italy
Ezio Santagata, Polytechnic of Torino, Italy
Piotr Sawicki, Poznan University of Technology, Poland
Kungrabat Sharipov, TT Polytechnic University, Uzbekistan
Valentin Silyanov, State Technical University, Russia
Andrea Simone, University of Bologna, Italy
Wayne K. Talley, Old Dominion University, USA
Bagdat Teltayev, Highway Research Institute, Kazakhstan
Susan L. Tighe, University of Waterloo, Canada
Tomaz Tollazzi, University of Maribor, Slovenia
Elen Twrdy, University of Ljubljana, Slovenia
András Várhelyi, Lund University, Sweden
Ashish Verma, Indian Institute of Science, India
Alex Visser, University of Pretoria, South Africa
Zhong Wang, Dalian University of Technology, China
Fred Wegman, Delft University of Technology, The Netherlands
Alfred Weninger-Vycudil, Viagroup, Switzerland
Jean François Wounba, Université Libre de Bruxelles, Belgium
Thomas J. Yager, NASA Langley Research Center, USA
Stefano Zampino, Road and Transport Engineer, Italy
Daiva Zilioniene, Vilnius Gediminas Tech. University, Lithuania

xvi

TIS2017_Book.indb xvi 2/8/2017 8:00:28 PM

 ORGANIZING COMMITTEE

(by alphabetical order)

Stefania Balestrieri
Sonia Briglia
Lauragrazia Daidone
Gianluca Dell’Acqua
Mario Magnanelli
Edoardo Mazzia
Alessandro Ruperto
Rocco Sorropago

ORGANIZER

Associazione Italiana
per l’Ingegneria del Traffico e dei Trasporti

HOSTING INSTITUTION

Automobile Club d’Italia

UNDER THE AUSPICES OF

Transportation Research Board

Università degli Studi di Napoli Federico II
Universidade Federal do Rio Grande do Sul
Società Italiana Infrastrutture Viarie
Nottingham Transportation Engineering Centre
Transportation Research Group of India
ANAS, SITEB, CIFI, FASTIGI, FINCO, WISTA Italia
Sapienza Università di Roma
University of Pretoria

REVIEWERS

Andrus Aavic Milan Batista Werner Brilon

Francesco Abbondati Gianfranco Battiato Lelio Brito
Vladimír Adamec Michelle Beiler Massimiliano Bruner
Maja Ahac Shlomo Bekhor James Bryce
Gordon Airey Francesco Bella Luciano Bucharles
Francesco Alberti Dario Bellini Paolo Budetta
Rami M. Alfaqawi Andrea Benedetto Mariarosaria Busiello
Sharid Amiri Luca Bernardini Daniel Jorge Caetano
Scott Anderson Šime Bezina Salvatore Cafiso
Boris Antic Salvatore Antonio Biancardo Francesco Canestrari
Tomas Apeltauer Arianna Bichicchi Giuseppe Cantisani
Daniela Aresu Luigi Biggiero Francesco Saverio Capaldo
Federico Autelitano Edoardo Bocci Agostino Cappelli
Orazio Baglieri Guido Bonin Angela Carboni
Nicola Baldo Gerardo Botas Fabrizio Cardone
Edita Baltrenaite Marilisa Botte Armando Cartenì
Marco Bassani Davor Brcic Maria Castro
Ana Bastos Sonia Briglia Alberto Castro Fernandez

xvii

TIS2017_Book.indb xvii 2/8/2017 8:00:29 PM

 Selene Cattani Luís Fernandes Rosa Anna La Rocca
Clara Celauro Jose Leomar Fernandes Jr. Claudio Lantieri
Gianluca Cerni Silvia Ferrini Michela Le Pira
Eleonora Cesolini Gilda Ferrotti Matheus Lemos
Christine Chaloupka-Risse Paola Firmi Nogueira
T. Donna Chen Gerardo Flintsch Giovanni Leonardi
Tao Chen Alessandro Focaracci Salvatore Leonardi
Jack Cheng Anna Formisano Michele Leone
Laura Chiacchiari Giovanni Forte Riccardo Licciardello
Giuseppe Chiappinelli Mateus Fratoni Souza Jose Manuel Lizarraga
Sandro Colagrande Sanja Fric Aleksander Ljubic
Pasquale Colonna Francesca Frigio Davide Lo Presti
Andrea Conca Stanisław Gaca Simone Lopes
Giovanni Coraggio Fabio Galatioto Adelino Jorge Lopes
Maria Vittoria Corazza Stipe Galić Ferreira
Alessandro Corradini Vincenzo Gallelli Giuseppe Loprencipe
Marcello Corsi Mariano Gallo Lucia Losasso
Daniele Cortis Alfredo García García Jurate Kumpieneuniversity
Marco Costa Lilli Gargiulo Lulea
Kevin Cullinane Andrea Giaccherini Renato Macciotta Pulisci
Luca D’Acierno Giovanni Giacomello Igor Majstorović
Pierpaolo D’Agostino Nadia Giuffrida Gabriele Malavasi
Essam Dabbour Orazio Giuffrè Francesca Maltinti
Andrew Dawson Tullio Giuffrè Laura Mancini
Shirley Ddamba Felice Giuliani Sadko Mandžuka
Vicent de Esteban Marinella Giunta Mohammad Manghrour
Chapapria Christos Gkartzonikas Giovanni Mantovani
Francesco De Florio Carlotta Godenzoni Edoardo Marcucci
Maria Luisa De Guglielmo Marcelo Gonzalez Andrea Marella
Mario De Luca Hernan Dario Gonzalez Alessandro Marradi
Stefano de Luca Zapata Ana Marušić
Gianluca Dell’Acqua Anna Granà Vittorio Marzano
Camilla De Micheli Andrea Graziani Giulio Maternini
Vittorio De Riso di James Grenfell Raffaele Mauro
Carpinone Andrea Grilli Federico Mazzetta
César De Santos-Berbel Marco Guerrieri Giorgia Mazzoni
Ilaria Del Ponte Inbal Haas Gabriella Mazzulla
Miguel Angel Del Val Melus Michael Hendry Silvio Memoli
Roberto Devoto Theuns.F.P Henning Rodrigo Mesa-Arango
Paola Di Mascio Imine Hocine Nenad Milutinovic
Roberta Di Pace Felix Huber Kazuaki Miyamoto
Marco Diana Luis Iglesias Goran Mladenovic
Nuria Diaz Maroto Matteo Ignaccolo Douglas Mocelin
Llorente Paolo Intini Laura Moretti
Natalia Distefano Giuseppe Inturri Miso Mudric
Marilyn Dodson Alibay Iskakbayev Steve Muench
Lorenzo Domenichini G.P. Jayaprakash Charles Musselwhite
Laura Eboli Marlova Johnston Lorenzo Mussone
Manuela Esposito David Jones Dmitry Nemchinov
Coenraad Esveld Raimundas Junevicius Dusan Nicolic
Massimiliano Fabbricino Jwan Kamla Rubina Normahomed
Aldo Fabregas Marianne Karlsson Manabu Nose
Asif Faiz Nariman Khalil Nozomi Nose
Gianfranco Fancello Paraskevi Kladaki Nurit Oliker
Nicolau Dionísio Fares Katerina Konsta Sara Guerra-Oliveira
Gualda Maja Krcum Lucas Ortiz
Javed Farhan Srećko Krile Masafumi Ota
Enrico Fattorini Przemysław Kupka Rudra P. Pradhan

xviii

TIS2017_Book.indb xviii 2/8/2017 8:00:29 PM

 Francesca Pagliara Marko Rencelj Ivano Toni
Enrico Pagliari Oskar Rexfelt Alex Torday
Athanasios Pallis Stefano Ricci Adam Torok
Roberto Palmarini Ralf Risser Enza Torrisi
Aimilia A. Papachristou Keith Robinson Riccardo Troisi
Panagiotis Papaioannou Jorge Luis Rodriguez Filip Trpcevski
Andrea Papola Navarro Laura Trupia
Luigi Pariota Matteo Rossi Lucia Tsantillis
Tony Parry Francesca Russo Marialuisa Tummiello
Robert Parsons Emanuele Sacchi Marjan Tušar
Stefania Pascale Nicola Sacco Elen Twrdy
Marco Pasetto Frank Saccomanno George K. Vaggelas
Emiliano Pasquini Giuseppe Salvo Rosolino Vaiana
Ujwalkumar Patil Jose San Martin Andrius Vaitkus
Aravind Pedarla Xiomara Sanchez-Castillo Luís Vasconcelos
Orazio Pellegrino Meg Sangimino Flavio Vaz
Jorge Augusto Pereira Cesare Sangiorgi Loretta Venturini
Ceratti Ezio Santagata Ashish Verma
Federico Perrotta João Santos Carlos Videla
Dalibor Pesic Piotr Sawicki Valeria Vignali
Marco Petrelli Francesco Sdao Tejo Vikash Bheemasetti
Massimiliano Petri Ramoel Serafini Carolina Villa
Cristiana Piccioni Antonino Sferlazza Francesco Viola
Hrvoje Pilko Manuel Silvestri Alex Visser
Francesco Pirozzi Valentin Silyanov Alessandro Vitale
Antonio Placido Andrea Simone Antonino Vitetta
Maria Jose Poblaciones Zuzanna Simonova Francesco Viti
Ioannis Politis Giuseppe Sollazzo Martin Vsetecka
Silvia Portas Everton Cesar Souza András Várhelyi
Annalisa Pranno Andrea Spinosa Ting Wang
Filippo Praticò Tatjana Stanivuk Zhong Wang
Carlo Prato Arianna Stimilli Fred Wegman
Pavel Pribyl Roger Surdahl Hongbo Wu
Christos Pyrgidis Yusak Susilo Hai Yan
Cesar Queiroz Luca Tefa Stefano Zampino
Vittorio Ranieri Bagdat Teltayev Marina Zanne
Thomas Kjær Rasmussen Stefano Terribile Nataša Zavrtanik
Danijel Rebolj Andrej Tibaut Olja Čokorilo
Francesco Rebora Tomaz Tollazzi
Dv Reddy Mateusz Tomczak

xix

TIS2017_Book.indb xix 2/8/2017 8:00:29 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport infrastructure

TIS2017_Book.indb 1 2/8/2017 8:00:32 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Deformation and strength of asphalt concrete under static

and step loadings

A.I. Iskakbayev & B.B. Teltayev

Kazakhstan Highway Research Institute, Almaty, Kazakhstan

C. Oliviero Rossi
University of Calabria, Rende, Italy

ABSTRACT: Deformation and strength of fine-grained asphalt concrete were investigated under static
and step loadings experimentally in the paper. Temperature varied within the range of 20 ± 2°C. The
stress varied within the range of 0.055 MPa and 0.311 MPa during creep test. The results of experiments
showed that creep curves have three sites within specified ranges of stress—site of unstabilized creep, site
of stabilized creep and site of accelerating creep. The long-term strength curve is described successfully
by exponential function. The stress affects greatly the long-term strength of asphalt concrete: the increase
of stress for single order reduces failure time for four orders. The sequence of impacts also affects greatly
the long-term strength of asphalt concrete.

1 INTRODUCTION 0.311 MPa. Creep curves under different loads and

long-term strength curve of the asphalt concrete
Asphalt concrete is one of the main materials have been constructed. Three characteristic sites
for highway pavements. Mechanical properties of creep curves—the unstabilized, stabilized and
of an asphalt concrete are highly depending on accelerating creep sites are shown.
temperature, value, duration, and rate of load-
ing (MS-4 2008, Papagiannakis & Masad 2008,
Yoder & Witczak 1975). In real road conditions 2 MODELS FOR PREDICTION
temperature in points of asphalt concrete layers of OF ASPHALT CONCRETE DAMAGE
pavement structures due to variations of ambient
temperature, track wheels load values, their action The vehicles, running along the highways, have dif-
duration and rate varies within wide limits. There- ferent number of axles, the loads on which varies
fore, determination of mechanical behavior of an within the wide range. It is natural that the axles
asphalt concrete taking into account the variation with different values of loads cause different dam-
of the above mentioned factors has important age level for pavement. One of the prediction
practical value. models for materials and structures damage, con-
It is known that the basic methods for evalua- sidering load impact of different values, is well-
tion of mechanical behavior of viscoelastic mate- known Miner’s law (Miner 1945). Miner’s law is
rials are tests on creep and relaxation (Cristensen also known as the principle of linear summation
1971, Ferry 1980, Tschoegl 1989). Technically, of damages (Talreja & Sing 2012) and found wide
realization of creep test is easier. It is possible to application in engineering calculations. For exam-
construct creep curves and long-term strength with ple, in the USA it is used for prediction of fatigue
using its results. Relaxation curves can be obtained life for asphalt concrete pavement of highways
from the creep curves by using known methods (ARA, ERES 2004). In this regard fatigue damage
(Tschoegl 1989, Hopkins & Hamming 1957). The is calculated under the following equation:
long-term strength curves enable to determine
service life of a road asphalt concrete pavement. T
ni
In this paper test results of hot fine-grained D=∑ (1)
asphalt concrete samples on creep are presented. i =1 Ni
Creep tests were carried out by the direct tensile
scheme until complete fracture of the asphalt where D = damage; T = total number of periods;
concrete samples. Test temperature was 20 ± 2°C. ni = actual traffic period i; and Ni = traffic allowed
The applied stress was changed from 0.055 to under conditions prevailing in i.

TIS2017_Book.indb 3 2/8/2017 8:00:32 PM

 Vehicles run along the highways with different 3 MATERIALS
speed, i.e. they affect the pavement structure with
different duration. As under real road conditions 3.1 Bitumen
the speed of the vehicles varies within the wide
In this paper bitumen of grade 100–130 has been
ranges, axle load duration differs greatly. Unfortu-
used which meets the requirements of the Kaza-
nately, equation (1), based on Miner’s law, does not
khstan standard (ST RK 1373 2013). The bitumen
consider load duration, which can be a source of
grade on Superpave is PG 64–40 (Superpave series
systematic large inaccuracies during determination
No.1 2003). Basic standard indicators of the bitu-
of fatigue life for asphalt concrete pavement of the
men are shown in Table 1. Bitumen has been pro-
highway.
duced by Pavlodar processing plant from crude oil
At present, the so-called Bailey’s criterion is
of Western Siberia (Russia) by the direct oxidation
well-known in science and engineering practice
method.
(Bailey 1939), which can be described in the fol-
lowing form: 3.2 Asphalt concrete
tp Hot dense asphalt concrete of type B that meets
dt
∫ τ [σ (t )] = 1
0
(2) the requirements of the Kazakhstan Standard (ST
RK 1225 2013) was prepared with use of aggregate
fractions of 5–10 mm (20%), 10–15 mm (13%),
where tp = failure time; σ(t) = stress, varied in time; 15–20 mm (10%) from Novo-Alekseevsk rock pit
and τ[σ(t)] = dependence of material failure time (Almaty region), sand of fraction 0–5 mm (50%)
on stress. from the plant “Asphaltconcrete-1” (Almaty city)
Contrary to Miner’s law (1), Bailey’s criterion and activated mineral powder (7%) from Kordai
(2) considers load duration, i.e. the speed of the rock pit (Zhambyl region).
vehicle along the highway. Dependence τ[σ(t)] Bitumen content of grade 100–130 in the asphalt
of Bailey’s criterion represents by itself the ana- concrete is 4,8% by weight of dry mineral material.
lytical equation for the so-called curve of long- Basic standard indicators of the aggregate and the
term strength, which is constructed based on the asphalt concrete are shown in Tables 2 and 3 respec-
test results of the material according to the creep tively. Granulometric composition curve for mineral
scheme (Kachanov 1986, Rabotnov 1987). part of the asphalt concrete is shown in Figure 1.

Table 1. Basic standard indicators of the bitumen.

Indicator Measurement unit Requirements of ST RK 1373 Value

Penetration, 25°C, 100 gr, 5 s 0.1 mm 101–130 104

Penetration Index PI – −1.0… +1.0 −0.34
Tensility at temperature: cm
25°C ≥ 90 140
0°C ≥ 4,0 5.7
Softening point °C ≥ 43 46.0
Fraas point °C ≤ −22 −25.9
Dynamic viscosity, 60°C Pa⋅s ≥ 120 175.0
Kinematic viscosity mm2/s ≥ 180 398.0

Table 2. Basic standard indicators of the crushed stone.

Value
Requirements of
Indicator Measurement unit ST RK 1284 fraction 5–10 mm fraction 10–20 mm

Average density g/cm3 – 2.55 2.62

Elongated particle content % ≤ 25 13 9
Clay particle content % ≤ 1,0 0.3 0.2
Bitumen adhesion – – satisf. satisf.
Water absorption % – 1.93 0.90

TIS2017_Book.indb 4 2/8/2017 8:00:32 PM

 Table 3. Basic standard indicators of the asphalt concrete.

Indicator Measurement unit Requirements of ST RK 1225 Value

3
Average density g/cm – 2.39
Water saturation % 1.5–4.0 2.3
Voids in mineral aggregate % ≤ 19 14
Air void content in asphalt concrete % 2.5–5.0 3.8
Compression strength at temperature: MPa
0°C ≤ 13.0 7.0
20°C 3.4
50°C ≥ 1.3 1.4
Water stability – ≥ 0.85 0.92
Shear stability MPa ≥ 0.38 0.39
Crack stability MPa 4.0–6.5 4.1

means of two clock typed indicators while data

was recorded in a video camera.

5 EXPERIMENT

5.1 Creep curves

Our previous works (Iskakbayev et al. 2016a, b, Tel-
tayev et al. 2016) showed that an asphalt concrete
creep curve as the most of viscoelastic materials has
three characteristic sites: the site I of unstabilized
creep with decreasing rate, the site II of stabilized
creep with constant (minimum) rate and the site
Figure 1. Granulometric curve of mineral part of the III of accelerating creep with increasing rate which
asphalt concrete.
precedes failure. The above works show test results
of asphalt concrete for creep with relatively narrow
ranges of stress variation. This work includes test
4 TEST METHODS
results for seven values of stress from 0.055 MPa
to 0.311 MPa. Practically five samples of asphalt
4.1 Sample preparation
concrete were tested for each value of stress.
Samples of the hot asphalt concrete in form of a Figures 2 and 3 show creep curves for asphalt
rectangular prism with length of 150 mm, width concrete with stresses 0.055 MPa and 0.260 MPa
of 50 mm and height of 50 mm were manufactured respectively. It is clearly seen that at small stress
in the following way. Firstly the asphalt concrete (0.055 MPa), as well as at high stress (0.260 MPa)
samples were prepared in form of a square slab by creep curves have all three sites. It is important
means of the Cooper compactor (UK, model CRT- to remember that since the beginning of loading
RC2S) according to the standard (EN 12697-33 to the failure moment (tf) asphalt concrete passes
2003). Then the samples were cut from the asphalt three sites of deformation.
concrete slabs in form of a prism. Deviations in Figure 4 shows creep curves for five samples of
sizes of the beams didn’t exceed 2 mm. asphalt concrete with stress 0.188 MPa. This figure
shows well the impact of structural non-homogeneity
of asphalt concrete on its deformability and durabil-
4.2 Test
ity: values of long-term strength and also deforma-
Tests of hot asphalt concrete samples in a form of tion characteristics have static dispersion.
rectangular prism on creep were carried out accord-
ing to the direct tensile scheme until a complete
5.2 Long-term strength
failure. The test temperature was equal to 20 ± 2 °C,
stress was changing from 0.055 to 0.311 MPa⋅s. The dependence of failure time on stress is called
The tests were carried out in a special assembled long-term strength of a material (Kachanov 1986,
installation. The sample strain was measured by Rabotnov 1987). The long-term strength curve of

TIS2017_Book.indb 5 2/8/2017 8:00:32 PM

 Figure 2. The asphalt concrete creep curve at stress Figure 5. The long-term strength curve of the asphalt
0,055 MPa. concrete.

Figure 3. The asphalt concrete creep curve at stress Figure 6. Dependence of failure strain on stress.
0,260 MPa.
Equation shows that the stress impacts greatly
on long-term strength of asphalt concrete: stress
increase for single-order reduces long-term strength
for four orders.

5.3 Failure strain

It is known that one of the first criteria for strength
in mechanics of materials is connected with limit
value of strain. Due to the above it is important
to evaluate the values of failure strain at different
stresses. Figure 6 shows the dependence of failure
strain for asphalt concrete on stress, where you can
see that failure strain value is not constant, i.e. it
cannot be the criterion of strength. In addition,
Figure 4. The asphalt concrete creep curves at stress this dependence is of complicated character and
0,188 MPa.
described by quadric polynomial: failure strain is
increased monotonously up to the stress of 0.12–
the asphalt concrete constructed by results of the 0.13 MPa and reduces with further stress increase.
test on creep is shown in Figure 5. As it can be
seen, the long-term strength curve is successfully
described by exponential function: 5.4 Step loading
As it was said previously, vehicles with different
t f = 1, 4072 σ −4,063 (3) axle load run along the highways. Even loads on
certain axles of one vehicle differ significantly.
Tf – failure time, s; and σ = stress, MPa. The sequence of impacts for these axle loads on

TIS2017_Book.indb 6 2/8/2017 8:00:33 PM

 and step loading at temperature 20 ± 2°C and at
stresses from 0.055 MPa to 0.311 MPa showed that:

− creep curves have three sites—unstabilized

creep, stabilized creep and accelerating creep. It
is important to know that since the beginning
of loading to failure point the deformation of
Figure 7. The first scheme of loading for the sample of
the asphalt concrete.
asphalt concrete increases with different rate,
which is characteristic for certain segments of
the creep curve;
− the long-term strength curve for asphalt concrete
is successfully described by exponential function.
The stress impacts greatly on long-term strength
of asphalt concrete: stress increase for single-
order reduces failure time for four orders;
− the sequence of impacts affects greatly the
strength of asphalt concrete. It requires the
Figure 8. The second scheme of loading for the sample development of a new strength criterion for
of the asphalt concrete. asphalt concretes;
− due to structural non-homogeneity of asphalt
pavement, differing in value, is of changeable concrete its long-term strength and charac-
character. Therefore, to evaluate the sequence of teristics of deformability have essential static
impacts on asphalt concrete strength, the follow- dispersion.
ing tests were carried out under the scheme of step
loading.
The first five samples were loaded by stress equal REFERENCES
to 2.29 kg/cm2 which had not changed for 180 sec-
onds. Then the stress was reduced till 1.14 kg/cm2 ARA ERES. 2004. Guide for mechanistic-empirical design
which was constant till the sample failure. The of new and rehabilitated pavement structures. Final
samples average failure time was 6409seconds report. Part 3. Design analysis. Chapter 3. Design of
(Figure 7). new and reconstructed flexible pavements. Washington:
The next five samples were loaded firstly by Transportation Research Board.
stress equal to 1.14 kg/cm2 which was constant Bailey, J. 1939. An attempt to correlate some tensile strength
during 180 seconds. Then the stress was increased measurements on glass: III. Glass Industry 20: 95–99.
Cristensen, R.M. 1971. Theory of viscoelasticity: An
till 2.29 kg/cm2 which was constant till the sam-
introduction. New York: Academic Press.
ple failed. The samples average failure time was EN 12697-33. 2003. Bituminous Mixtures. Test Methods
937 seconds (Figure 8). for Hot Mix Asphalt. Part 33: Specimen prepared by
There was established that changes in small and roller compactor. Brussels: European Committee for
big tension stresses sequences eventually influence Standardization.
on the asphalt concrete failure time. In the case Ferry, J.D. 1980. Viscoelastic Properties of Polymers. New
when the sample was firstly loaded by big stresses York: John Wiley & Sons, Inc.
and then by small ones the failure time increases Hopkins, I.L. & Hamming, R.W. 1957. On creep and
almost for seven times than the sample which was relaxation. Journal of Applied Physics 28: 906–909.
Iskakbayev, A., Teltayev, B. & Alexandrov, S. 2016a. Deter-
loaded firstly by small stresses and then by big ones.
mination of the creep parameters of linear viscoelastic
This fact can be explained by the fact that in the first materials. Journal of Applied Mathematics 2016: 1–6.
case the duration of acting of big stress is essentially Iskakbayev, A., Teltayev, B., Andriadi, F., Estayev, K.,
less than duration of acting of small stresses. Suppes, E. & Iskakbayeva, A., 2016b. Experimental
As the sequence of load impact affects greatly research of creep, recovery and fracture processes of
the failure time of asphalt concrete, it is required asphalt concrete under tension. Proc. 24th intern. con-
to develop a new strength criterion, which differs gress on theoretical and applied mechanics, Montreal,
from Miner’s law and Bailey’s criterion, mentioned 21–26 August 2016.
above. Kachanov, L. 1986. Introduction to continuum damage
mechanics. Dordrecht: Martinus Nijhoff Publishers.
Miner, M.A.1945. Cumulative damage in fatigue. Journal
of Applied Mechanics. 12: A159-A164.
6 CONCLUSION MS-4. 2008. The Asphalt Handbook. 7th Edition. Lexing-
ton: Asphalt Institute. Papagiannakis, A.T. & Masad,
The results of experimental investigation for strain E.A. 2008. Pavement Design and Materials. New Jer-
and strength of fine-grained asphalt concrete at static sey: John Wiley & Sons, Inc.

TIS2017_Book.indb 7 2/8/2017 8:00:34 PM

 Rabotnov, Yu.N. 1987. Introduction to fracture mechan- Teltayev, B.B., Iskakbayev, A.I. & Rossi, C.O. 2016.
ics. Moscow: Nauka. Regularities of creep and long-term strength of hot
ST RK 1225. 2013. Hot mix asphalt for roads and air- asphalt concrete under tensile. In Sandra et al. (eds.),
fields. Technical specifications. Astana. Functional pavement design; Proc. 4th Chinese-Euro-
ST RK 1284. 2004. Crushed stone and gravel of dense pean workshop, Delft, 29 June-1 July 2016. London:
rock for construction works. Technical specifications. Taylor & Francis Group.
Astana. Tschoegl, N.W. 1989. The Phenomenological Theory of
ST RK 1373. 2013. Bitumens and bitumen binders. Oil road Linear Viscoelastic Behavior. An Introduction. Berlin:
viscous bitumens. Technical specifications. Astana. Springer-Verlag.
Superpave series No. 1. 2003. Performance graded asphalt Yoder, E.J. & Witczak, M.W. 1975. Principles of Pave-
binder specification and testing. Lexington: Asphalt ment Design. New Jersey: John Wiley & Sons, Inc.
Institute.
Talreja, R. & Sing, C.V. 2012. Damage and failure of com-
posite materials. Cambridge: Cambridge University
Press.

TIS2017_Book.indb 8 2/8/2017 8:00:34 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Roundabout design guidelines: Case study of Croatia

H. Pilko
Department of Road Transport, Faculty of Transport and Traffic Science, University of Zagreb, Zagreb, Croatia

ABSTRACT: The popularity of roundabout application around the world is evident. Due to the inex-
perience of construction companies and the lack of proper national guidelines, distinctiveness in design
is noticeable. In some intersections this led to reduction of Traffic (operational) Efficiency (TE). The pur-
pose of this paper is to analyze: 1) the current state of roundabouts in Croatia; (2) known approaches to
using geometry elements of roundabouts to predict TE; (3) overview and comparison of selected design
guidelines; and (4) to present and comment the latest Croatian Roundabout Design Guidelines on State
Roads 2014 and show examples of good practice. Research results will serve to disseminate the knowledge
for proper application and implementation of national roundabouts in order to compare it with interna-
tional design practice and standards.

1 INTRODUCTION It also gives a brief overview of the literature

on roundabout design and TE in Croatia. Sec-
The popularity of roundabouts around the world tion 3 gives a brief overview and comparison of
has driven substantial efforts to optimize their plan- European, American and Croatian guidelines.
ning and modeling. Analyses of how geometric ele- In Section 4 recent Croatian roundabout geom-
ments, traffic flow movements, driver behavior and etry design practice is shown. Section 5 reports the
other factors influence roundabout Traffic (opera- conclusions.
tional) Efficiency (TE) have led to the development
of numerous simulation-based computational mod-
els. In addition, they use primarily empirical and
2 BACKGROUND
gap-acceptance mathematical models to determine
TE (mainly focused on capacity and delay).
2.1 Roundabout geometry and efficiency
Croatia has slightly more than 200 roundabouts,
of which more than 60% lie within or on the edge Numerous models for determining roundabout
of urban areas, and many of them deviate substan- capacity under mixed-traffic conditions suggest that
tially from international standards for roundabout it is strongly affected by geometric elements (Dahl &
planning, design and modeling, which compro- Lee 2012). Studies of roundabouts in various coun-
mises their TE and traffic safety (Pilko 2014). tries, particularly of single-lane roundabouts in
These deviations are due primarily to the inex- urban areas, have shown that proper design and
perience of construction companies and the use of modeling can significantly improve Traffic (opera-
outdated design guidelines. tional) Efficiency (TE) (Vasconcelos et al. 2013;
The country lacks a system for monitoring and Mauro & Cattani 2012). The most important geo-
analyzing TE, and other parameters, though the metric elements influencing TE (i.e. entry capacity)
government has called for the building and recon- are entry width, entry radius, flare length, entry
struction of roundabouts as part of its National angle, Inscribed Circle Diameter (ICD) and number
Traffic Safety Plan 2011–2020 (Pilko 2014). of entry lanes (Kimber 1980; Al-Omari et al. 2004;
The present research aims to describe and Dahl & Lee 2012; Yap et al. 2013; Barić et al. 2016).
analyze (1) the current state of roundabouts in Geometry considerations have also proven use-
Croatia; (2) known approaches to using geometry ful for analyzing driver behavior in roundabout sit-
elements of roundabouts to predict TE; (3) over- uations and implications for traffic safety (Muffert
view and comparison of selected design guidelines; et al., 2013; Wang et al., 2002). Methods for predict-
and (4) to present and comment the latest Croatian ing traffic accidents have been described based on
Roundabout Design Guidelines on State Roads 2014 geometric elements (Maycock & Hall 1984), sight
and show examples of good practice. distance (Turner et al., 2009; Zirkel et al., 2013), as
The reminder of the paper is organized as fol- well as traffic dynamics and driver behavior when
lows. Section 2 contains the literature review on passing through the „potential conflict“ zone of
roundabout geometry and TE. the intersection (Mauro & Cattani, 2004).

TIS2017_Book.indb 9 2/8/2017 8:00:36 PM

 Studies of roundabout TE, conducted primarily between vehicle trajectory design speeds through the
in Western Europe and Australia, have led to sev- roundabout and observed vehicle speeds.
eral computational mathematical models that have
been integrated into various roundabout software
engineering simulation tools (e.g. ARCADY and
3 OVERVIEW AND COMPARISON
PTV VISSIM). These mathematical models can
OF DESIGN GUIDELINES
be classified as (1) empirical, (2) gap acceptance,
and (3) microsimulation. Each category has its
3.1 International design guidelines
disadvantages (Yap et al. 2013). Generally coun-
tries with updated roundabout design guidelines Worldwide positive experiences with modern
apply Highway Capacity Manual (HCM) models single-lane roundabouts contributed to further
for analyzing roundabout capacity (Chodur 2005). research, development and implementation of
These models take into account empirical and/or other types of roundabouts. Substantial practical
gap-acceptance models, but they do not address experiences initiated the formation of roundabout
the Level of Service (LOS). The capacity formula- design guidelines. The recent achievements in this
tion for urban single-lane roundabouts developed field for European countries like Austria, Ger-
in HCM is based on the mathematical formula- many, Switzerland, Poland, UK and the American
tion of geometry parameters developed by Kimber are presented here in a brief overview.
(1980) and formulation of queue length developed In general, types of roundabout are defined by
by Ning Wu (Wu 2001). Differences among these spatial limitation, location and traffic capacity. Mini
models in how data are collected and analyzed, as roundabouts are characterized by a small external
well as deviations between predicted and actual diameter and traversable central island for large
driver behavior, make it difficult to identify the most vehicles. They are commonly used in urban envi-
suitable ones for given conditions (Mauro 2010). ronments with average operating speeds of 40 km/h
Planners and designers should be aware of the spe- or less. Single-lane roundabouts represent a stand-
cific limitations of these models and the selected ard solution and they are characterized by single
model(s) must be calibrated against field data or entry lane, exist lane and circulatory lane. There is
other validated models to ensure accuracy. Unfor- a non-traversable central island and they are used
tunately, this calibration step is often neglected. In both in urban and rural environments. Multi-lane
situations where the model is not even developed, it roundabouts have two or more entry/exit and circu-
may be advisable to analyze roundabout TE using latory lanes. Due to a possibility of path ovelap at
various approaches, and HCM models may be the the entry and the exit as well as higher speeds these
most appropriate for such work (Mauro, 2010). types of roundabouts are less safe in comparison
with mini and single-lane roundabouts.
According to Austrian guidelines (Osterreich-
2.2 Roundabouts in Croatia
ische Forschungsgemeinschaft Strasse und Verkehr
The in situ work by Legac et al. (2008) on 30 round- 2010) the geometry elements in mini rounda-
abouts in Zagreb examined relationships between bouts are determined on the basis of curve of the
main geometric elements and the numbers and course of design vehicles. It is recommended that
types of traffic accidents, traffic flow demand, and single-lane roundabouts have the external diameter
numerous other determinants of capacity. That ≥ 26 m (35–40 m) and that multi-lane roundabouts
study confirmed that unsignalized roundabouts in have the external diameter ≥ 40 m (50 to 60 m).
Croatia are safer than classical intersections and it German guidelines (Forschungsgesellschaft fur
showed that geometric elements strongly influence Strassen und Verkherswesen 2006) have determined
safety. that mini roundabouts have a diameter of 13 to 22 m
Other authors have focused on applying rounda- and capacity of 18000 veh/day. A central island is
bout models from outside Croatia to the Croatian traversable and has truck apron raised by 4 to 5 cm.
situation. Their results suggest that imported mod- The width of circulatory lane is beween 4 and 6 m
els can work well, as long as they are calibrated with transversal inclination of 2.5% outwards. The
for local conditions (Otković Ištoka 2008; Ištoka entry/exit radii is from 8 to 10 m. Small single-lane
Otković & Dadić 2009; Šubić et al. 2012). For exam- roundabouts have the capacity of 25000 veh/day.
ple, Ištoka Otković et al. (2013) used neural net- The external diameter is from 26 to 45 m, the entry
works to calibrate a traffic microsimulation model radius is from 10 to 16 m and the exit radius is from
for two urban single-lane roundabouts in Osijek. 12 to 18 m. These dimensions provide great traffic
Šurdonja et al. (2013) optimized geometric elements safety. Circulatory lane being 6.5 to 9 m wide con-
such as inscribed circle radii, entry/exit radii, entry/ sists of driving and traversable part used by large
exit approach width, and vehicle path trajectory, vehicles. Small double-lane roundabouts have one
while Pilko et al. (2014) examined the relationship or two lane entry with the entry radii of 12 to16 m,

10

TIS2017_Book.indb 10 2/8/2017 8:00:36 PM

 depending on the traffic load and one lane exit with Urban multiple (two)-lane roundabouts are sup-
the exit radii of 12 to 18 m. In the circulating area posed to have external diameter from 37.5 m to
there are two traffic lanes with the total width of 8 over 55 m and from 40 m to 65 m for rural areas.
to 10 m, but they are not marked with the horizon- Single-lane roundabouts are the solutions which,
tal signalization. Diameter varies from 40 to 60 m in Polish conditions, ensure a very high level of
with the maximum capacity of 32000 veh/day. Their safety to the users and high capacity. Furthermore,
diameter is more than 60 m and they have two or two-lane roundabouts are not entirely optimal
more lanes in entry, exit and circulatory lanes (For- solutions since substantial number of drivers use
schungsgesellschaft fur Strassen und Verkherswesen mainly the right lane, both at the approach entry
2006). According to the German guideliness the and on the circulatory roadway (Macioszek 2013).
designing of small double-lane roundabouts is not The UK guidelines (UK Goverment 2007) not
allowed due to the reduced safety. The solution of the only define geometry elements for mini, single and
problem of increased through traffic is found in the multi-lane roundabouts, but they also give instruc-
use of roundabouts with the spiral traffic course, tions to: traffic safety, road users specific require-
the turbo roundabouts. ments, the assessment procedure, and conspicuity.
Swiss guidelines (Vereinigung Schweizerischer The main geometry parameters for mini, single
Strassenfachleute (VSS) 1999) have determined and multi-lane roundabouts are shown in Table 2.
that mini roundabouts and single-lane rounda-
bouts are designed with the entry radii of 10 to
3.2 Croatian guidelines
12 m, while the approach radius is five times larger.
In a properly designed entry the entry angle α has The procedures of planning and designing the
to be as large as possible. The exit radius is from 12 roundabouts in the Republic of Croatia are based
to 14 m. Mini roundabouts with circulatory lane on the current national (Institut prometa i veza
width of 7 to 8 m are characterized by external 2002) and applied foreign guidelines, especially
diameter of 14 to 16 m. Single-lane roundabouts German (Forschungsgesellschaft fur Strassen und
have external diameter of 26 to 40 m. Verkherswesen 2006), Austrian (Osterreichische
USA guidelines (NCHRP-National Cooperative Forschungsgemeinschaft Strasse und Verkehr 2010)
Highway Research Program 2010) have determined and Switzerland (Vereinigung Schweizerischer
that splitter islands can be raised, traversable or Strassenfachleute 1999), positive examples of
only marked and designed according to design world practice and empirical practice of designers
vehicle. The width of circulatory lane in single- (especially from the Netherlands). One of the first
lane roundabouts varies from 4.8 to 6 m. Circu- steps in creating national regulations for rounda-
lar shape of a central island is recommanded, but bout designing were the Guidelines for the Design
oval, irregular or raindrop shapes can also be used. of Circular Intersection (Guidelines 2002) from
The entry radius is from 15 to 30 m and the exit 2002 (Institut prometa i veza 2002). The main
radius is from 15 to 60 m. The traversable portion objective of these guidelines was the standardi-
of a central island is 50 to 75 mm raised. Multi- zation of design and implementation of rounda-
lane roundabouts have at least one entry with two bouts on public roads in the country. Then, in the
or more lanes which requires a wider roadway in absence of local regulations, designers applied the
circulating part of the intersection so that at least aforementioned foreign regulations and practices,
two vehicles can travel side by side. The width of which resulted in a large number of non-compliant
circulating double lane is from 8.5 to 9.8 m and of roundabout design. Also, insufficient attention has
circulating triple lane is from 12.8 to 14.6 m. Firstly been made to the criteria of TE.
an entry is designed with a smaller radius of 20 to Roundabout Guidelines for State Roads (Guide-
35 m, and then with a radius of 45 m and more. lines 2014) from 2014 (Deluka-Tibljaš et al. 2014)
The entry lane can be moved to the left in order to represent a logical step on the path of creating
obtain increased deflection which reduces it at exit. national regulations for planning and design-
Radius of the fastest path is between 53 and 84 m ing of all roundabout types, especially in urban
which results in design speed of 40 to 50 km/h. areas. They also represent a significant upgrade
According to Polish guidelines (Generalna of Guidelines 2002 in terms of geometry design
Dyrekcja Dróg Krajowych i Autostrad 2004) the elements, TE criteria, defining, planning and
geometry elements have determined that mini implementation of turbo roundabouts, and the
roundabouts have diameter of 14 to 22 m (excep- importance and necessity for trajectory checking
tionally 25 m), fully traversable central island, of the relevant design vehicle in the design phase.
entry/exit radii of 6 to 12 m and capacity of 15000– However, the guidelines do not indicate what mod-
17000 veh/day. Urban single-lane roundabouts are els or simulation software should be used for ana-
designed with the external diameter ranging from lyzing TE. The Guidelines2014 only briefly presents
14 to 45 m and from 30 to 50 m for rural areas. possible methods and their pros and cons of

11

TIS2017_Book.indb 11 2/8/2017 8:00:36 PM

 Austrian (Osterreichische Forschungsgemeinschaft 3.3 Guideline comparison
Strasse und Verkehr 2010), British linear regres-
From the brief overview (Table 1.a.b and Table 2)
sion, Australian (by National Association of Road
we can conclude that all European guidelines
authorities) and German (Wu 2001) method that
are more less similar in terms of main geometry
can be used for analyzing TE parameters.
design elements. Only the UK geometry param-
As the process of roundabout designing is very
eters are slightly different than the other Euro-
complex and specific, it is not always possible to
pean. This is also evident when we compare the
apply the optimal geometry design elements and
concepts and dimension with the latest Croatian
meet all requirements. However, planning and
guidelines. The most evident difference is approx-
designing in accordance with Guidelines2014 ena-
imate capacity. The USA guidelines have big-
bles unification at the national level, which derives
ger values of geometry design when compared
complete and optimal solutions. By making the final
to European and Croatian guidelines. This is
legislative framework for designing roundabouts, it
because of longer design vehicle and dimensions
is necessary to gradually adopt and implement cer-
related to specific national conditions and driver
tified foreign methodologies and practices, and to
behavior.
introduce them to state regulation, with obligatory
consideration of all national features.

Table 1.a. Guideline comparison of roundabout type and main geometry design elements.

Roundabout type

Mini Single-lane
1 2 3 4
Design guidelines/Design A D CH USA CRO A1 D2 CH3 USA CRO4
Environment urban urban or rural
External diameter <26 13–22 14–16 13–27 13–25 ≥ 26 26–45 26–40 27–55 30–40
Entry lane (m) 1 1
Exit lane (m) 1 1
Circulatory lane 1 1
Central island traversable non-traversable
Entry radii (m) – 8–10 10–12 15–20 10–12 10–16 10–12 15–30 10–12
Exit radii (m) – 8–10 12–14 ≤ 30 12–14 12–25 12–18 12–14 30–60 12–14
Circulatory lane width (m) – 4–6 7–8 4–10 4.5–5.0 6.5–9.0 4.8–6.1 5.5–7.0
Maximum recommended 25–30 30–40
entry design speed (km/h)
Capacity (veh/day) 10000 18000 ≤15000 25000 20000

Table 1.b. Guideline comparison of roundabout type and main geometry design elements.

Roundabout type

Multi-lane

Design guidelines/Design A1 D2 CH3 USA CRO4

Environment urban or rural
External diameter ≥ 40 40–60 ≥ 60 46–91 50–90
Entry lane (m) ≥1
Exit lane (m) ≥1
Circulatory lane ≥2
Central island non-traversable
Entry radii (m) 10–16 12–16 20 10–14
Exit radii (m) 12–25 12–18 ≥ 20 12–16
Circulatory lane width (m) 8–10 8–10 4.3–14.6 5.5–7.7
Maximum recommended entry 50–60 40–50
design speed (km/h)
Capacity (veh/day) 30000 32000 ≤45000 ≥ 250000

1 – Austrian, 2 – German, 3 – Swiss, 4 – Croatian.

12

TIS2017_Book.indb 12 2/8/2017 8:00:36 PM

 Table 2. Guideline comparison of roundabout type and main geometry design elements.

Roundabout type

Mini Single-lane Multi-lane

Design guidelines/Design PL1 UK2 PL1 UK2 PL1 UK2

Environment urban urban or rural
External diameter 22–45* ≥ 37.5–55*
14–22(25) <28 30–50** <28 ≥ 40–65** <100
Entry lane (m) 1 ≥1
Exit lane (m) 1 ≥1
Circulatory lane 1 ≥2
Central island traversable non-traversable
Entry radii (m) 6–12 6–20 10–15 10(20) 12–15 12–16
Exit radii (m) 6–12 8–10 12–15 15–20 14–17.5 20–100
Circulatory lane width (m) 5.0 6.0 5.5–8.75 <15 9.1–10 8–10
Maximum recommended entry 25–30 30–40 40–50 >40
design speed (km/h)
Capacity (veh/day) 15000–17000 18000 20000–25000 40000

1 – Polish, 2 – United Kingdom, *urban areas, **rural areas.

4 CROATIAN EXAMPLES

Given the above in section 2.2 and 3.2, the fol-

lowing will give a brief overview of recent imple-
mented and planned forms of urban roundabouts.
The presented examples of design practice are
designed in accordance with the Guidelines2002,
Guidelines2014, and according to aforementioned
foreign regulations.
Mounted roundabouts are also presented in the
latest guidelines, and the first examples of its usage
can be seen in the City of Zagreb. Figure 1 shows
the latest implementation of mounted roundabout
in urban area.
Reconstruction of State Road S517 on a road
section through the city of Belišće is performed
using two roundabouts (Figure 2). The recon- Figure 1. Roundabout Vončinina—Voćarska, City of
struction eliminates existing dangerous small turn Zagreb, Croatia—in traffic 2016.
radius and adverse horizontal elements.
External diameter of Roundabout 1. (to the
right in Figure 3) is 38.00 m, and is the busiest and
most important intersection in the city. Rounda-
bout 2. (to the left in Figure 3) has external diam-
eter of 36.00 m. The circulatory roadway width is
6.6 meters, with an additional truck apron width
of 1.8 m. The entry/exit radii were designed using
vehicle design trajectories for a truck with a trailer
length of 18.00 m (Pilko et al. 2015).
The first turbo roundabout in Croatia was car-
ried out in the city of Osijek in 2014. (Figure 2).
Intersection is designed in accordance with the
Guidelines2002 as well as the empirical experiences
with turbo roundabouts from Netherland. The
most important design elements are: external diam- Figure 2. Turbo roundabout in the City of Osijek,
eter is 5.15 m (outer space between shifted centers Croatia—in traffic 2014.

13

TIS2017_Book.indb 13 2/8/2017 8:00:36 PM

 Figure 3. Roundabouts at State road S517 in city of Belišće, Croatia.

of circular segments); inscribed central diameter is implementation of the existing or development of

4.95 m (inner space between shifted centers of cir- a new method and its application and validation
cular segments); R1 = 15 m (radius of the inner edge for Croatia. However, planning and designing in
of the road surface); R2 = 20 m (radius of the outer accordance with Guidelines2014 enables unification
edge of the outer lane); R3 = 20.3 m (radius of the at the national level, which derives complete and
inner edge of the outer lane); R4 = 25.2 m (radius optimal solutions. By developing the final legislative
of the outer edge of the road surface). Cycling lanes framework for designing roundabouts, it is neces-
and tram lines are also designed making this inter- sary to conduct a comprehensive and systematic
section more complex for all road users. field-analytical research on existing roundabouts.
In addition, current national studies may serve as
a certain database, as well as a source of empirical
5 CONCLUSION knowledge. It is necessary to involve all stakehold-
ers and international experts from this field. Results
There is an obvious increase in the application and of this paper can be used for dissemination of the
design of all types of roundabouts in urban areas knowledge for proper application and implementa-
worldwide. From the brief overview of the selected tion of national roundabouts in order to compare it
West European and USA roundabout guidelines with international design practice and standards.
we can conclude that West European guidelines are
more less similar in terms of main geometry design
elements. This is also evident when we compare the REFERENCES
concepts and dimension with the latest Croatian
guidelines. The most evident difference is approxi- Al-Omari, B., Al-Masaeid, H. & Al-Shawabkah, Y., 2004.
mate capacity. The USA guidelines have bigger Development of a Delay Model for Roundabouts in
values of geometry design when compared to West Jordan. Journal of Transportation Engineering, 130(1):
European and Croatian guidelines. This is because 76–82.
Barić, D., Pilko, H. & Strujić, J., 2016. An Analytic
of longer design vehicle and dimensions related to Hierarchy Process Model to Evaluate Road Section
specific national conditions and driver behavior. Design. Transport, 31(3): 312–321.
Due to the lack of Croatian regulations, some Chodur, J., 2005. Capacity Models and Parameters for
studies and designers warn of the distinctiveness Unsignalized Urban Intersections in Poland. Journal
of performance, which is at some roundabouts of Transportation Engineering, 131(12): 924–930.
resulted in reduction of TE. Guidelines2014 enables Dahl, J. & Lee, C., 2012. Empirical Estimation of
better preparation of national legislation for plan- Capacity for Roundabouts Using Adjusted Gap-
ning and design of all types of roundabouts. How- Acceptance Parameters for Trucks. Transportation
ever, the guidelines do not indicate what models or Research Record, 2312: 34–45.
Deluka-Tibljaš, A. et al., 2014. Smjernice za projektiranje
simulation software should be used for analyzing kružnih raskrižja na državnim cestama, University
TE. Future studies should make comparative com- of Rijeka, Faculty of Civil Engineering, Rijeka (in
parison of capacity formulations (TE—Measure Croatian).
of Effectiveness (MOE) of presented guidelines Forschungsgesellschaft fur Strassen und Verkherswesen,
by incorporating real traffic data collected from 2006. Merkblatt fur die Anlage von Kreisverkehren,
roundabouts in Croatia. This would enable possible Koln (in German).

14

TIS2017_Book.indb 14 2/8/2017 8:00:37 PM

 Generalna Dyrekcja Dróg Krajowych i Autostrad, Otković Ištoka, I., 2008. Capacity Modelling of Rounda-
2004. Wytyczne projektowania skrzyżowań drogowych, bouts in Osijek. Tehnički vjesnik-Technical Gazzette,
Warszawa. 15(3): 41–47.
Institut prometa i veza, 2002. Smjernice za projektiranje Pilko, H., 2014. Optimization of Roundabout Design and
i opremanje raskrižja kružnog oblika—rotora, Insti- Safety Component (doctoral thesis), University of
tute of Transport and Communications, Zagreb (in Zagreb, Faculty of Transport and Traffic Sciences,
Croatian). Zagreb (in Croatian).
Ištoka Otković, I. & Dadić, I., 2009. Comparison of Pilko, H., Barišić, I. & Bošnjak, H., 2015. Urban Round-
Delays at Signal-controlled Intersection and Rounda- abouts. In VI. Croatian Congress on Roads. Opatija:
bout. PROMET—Traffic & Transportation, 21(3): Croatian Road Association Vi-vita.
157–165. Pilko, H., Brčić, D. & Šubić, N., 2014. Study of Vehicle
Ištoka Otković, I., Tollazzi, T. & Šraml, M., 2013. Cali- Speed in the Design of Roundabouts. GRAĐEVINAR,
bration of Microsimulation Traffic Model Using 66(5): 407–416.
Neural Network Approach. Expert Systems with Šubić, N., Legac, I. & Pilko, H., 2012. Analysis of Capac-
Applications, 40(15): 5965–5974. ity of Roundabouts in the City of Zagreb according
Kimber, R., 1980. The traffic capacity of roundabouts, to HCMC–2006 and Ning Wu Methods. Tehnički
Transportation and Road Research Laboratory, vjesnik—Technical Gazette, 19(2): 451–457.
Crowthorne, Berkshire. Šurdonja, S., Deluka-Tibljaš, A. & Babić, S., 2013. Opti-
Legac, I. et al., 2008. Korelacija oblikovnosti i sigurnosti u mization of Roundabout Design Elements. Tehnički
raskrižjima s kružnim tokom prometa (MZOŠ br. 135- vjesnik-Technical Gazzette, 20(3): 533–539.
0000000-3313), Ministry of Science, Education and Turner, S.A., Roozenburg, A.P. & Smith A W, 2009.
Sports, Zagreb (in Croatian). Roundabout crash prediction models, Wellington.
Macioszek, E., 2013. Analysis of the Effect of Conges- UK Goverment, 2007. Design manual for roads and
tion in the Lanes at the Inlet to the Two-Lane Round- bridges, Volume 6: Road Geometry, Section 2: Junc-
about on Traffic Capacity of the Inlet. Activities of tions, Part 2 TD 16/07; Geometric Design of Rounda-
Transport Telematics: 97–104. bouts, London.
Mauro, R., 2010. Calculation of Roundabouts—Capacity, Vasconcelos, A., Seco, A. & Silva, A., 2013. Compari-
Waiting Phenomena and Reliability, Berlin, Heidel- son of Procedures to Estimate Critical Headways at
berg: Springer Science & Business Media. Roundabouts. PROMET—Traffic & Transportation,
Mauro, R. & Cattani, M., 2012. Functional and Economic 25(1): 43–53.
Evaluations for Choosing Road Intersection Layout. Vereinigung Schweizerischer Strassenfachleute (VSS),
PROMET—Traffic & Transportation, 24(5): 441–448. 1999. Schwizer Norm 640263 - Knoten mit Kreisverke-
Mauro, R. & Cattani, M., 2004. Model to Evaluate hren, Zurich (in German).
Potential Accident Rate at Roundabouts. Journal of Wang, B., Hensher, D.A. & Ton, T., 2002. Safety in the
Transportation Engineering, 130(5): 602–609. Road Environment: A Driver Behavioural Response
Maycock, G. & Hall, R.D., 1984. Accidents at 4-arm Perspective. Transportation, 29(3): 253–270.
roundabouts, Transportation and Road Research Lab- Wu, N., 2001. A Universal Procedure for Capacity
oratory, Crowthorne, Berkshire. Determination at Unsignalized (priority-controlled)
Muffert, M., Pfeiffer, D. & Franke, U., 2013. A Stereo- Intersections. Transportation Research Part B: Meth-
vision Based Object Tracking Approach at Rounda- odological, 35(6): 593–623.
bouts. IEEE Intelligent Transportation Systems Yap, Y.H., Gibson, H.M. & Waterson, B.J., 2013. An
Magazine, 5(2): 22–32. International Review of Roundabout Capacity Mod-
NCHRP-National Cooperative Highway Research Pro- elling. Transport Reviews, 33(5): 593–616.
gram, 2010. Roundabouts: An Informational Guide, Zirkel, B. et al., 2013. Analysis of Sight Distance, Crash
Washington DC. Rate and Operating Speed Relationships for Low-
Osterreichische Forschungsgemeinschaft Strasse und Volume Single Lane Roundabouts in the United
Verkehr, 2010. RVS 03.05.14 Strassenplanung, Plan- States. Journal of Transportation Engineering, 139(6):
gleiche Knoten—Kreisverkehr, Wien (in German). 565–573.

15

TIS2017_Book.indb 15 2/8/2017 8:00:38 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

High modulus asphalt concrete for Ljubljana airport apron

A. Ljubič
IGMAT d.d. Building Materials Institute, Ljubljana, Slovenia

ABSTRACT: Central part of the main apron at the Ljubljana airport from 1978 was scheduled for
reconstruction in 2014 because of the increased traffic loads and subsequent distress of the existing
asphalt pavement. Because of the traffic situation and faster completion time the asphalt pavement was
chosen for the reconstruction and for the planned high traffic loading and severe climatic conditions a
suitable asphalt mix had to be selected. A high stiffness modulus asphalt mix was chosen for base layers
and the article describes the mix design, its testing and finally completed reconstruction of the airport
apron area.

1 INTRODUCTION 1.2 Condition before the reconstruction

Condition of asphalt surface on the Apron area
1.1 Apron area
before the reconstruction was as follows:
Central part of the Main Apron at the Ljubljana
− Extensive surface cracks mostly filled with hot
airport from 1978 was scheduled for reconstruc-
applied sealants and joint filler (Figure 2),
tion in 2014 because of the increased traffic loads
and subsequent distress of the existing asphalt
pavement. It comprises about 19.000 m2 of asphalt
pavements, built in year 1978 during last major
reconstruction of the airport.
It consisted of two asphalt layers—base layer
in 10 cm thickness and surface layer 5 cm thick—
mainly built over the concrete 26 cm thick apron
from the year 1962 and in lesser part over asphalt
pavement enlargement of the apron area from 1975.
In 1993 an antiskid surface dressing was laid
over the entire area of asphalt surface serving as
protection against oil derivatives and at the same
time improving surface friction characteristics.
Very heavy loaded parts of Main Apron at air-
plane parking positions PSN 3, 4, 5 and 6 were Figure 1. Area of Main Apron TWY E reconstruction.
processed from asphalt pavement to concrete pave-
ment in 2006, the same was done for parking pos-
tions PSN 7, 33, 34 and 35 in year 2011, because
of the high permanent deformations of asphalt
layers.
Area planned for reconstruction in year 2014
comprised:
− Longitudinal driving axis of airplanes (TWY
E2, partly E1 and E3) during their schedul-
ing for parking in the first and second parking
rows,
− Drive-ins and drive-outs to single ‘push back’
parking positions in the first row parking
(mainly parking positions with avio-bridges),
− Drive-ins and drive-outs to single ‘push back’
parking positions in the second row parking, Figure 2. Typical asphalt surface cracks.

17

TIS2017_Book.indb 17 2/8/2017 8:00:40 PM

 and SIST 1038-1 Bituminous Concrete were the
following:
− resistance to permanent deformations according
to EN 12697-22, method B in air at the tempera-
ture of 60°C—maximum allowed wheel track-
ing rate 0,10 mm/103 passes (WTSAIR 0,10) and
maximum allowed proportional rut depth of 3%
(PRDAIR 3,0 )
− stiffness, on prismatic specimens according to
EN 12697-26, 4PB-PR at the temperature of
10°C and frequency of 10 Hz should be mini-
mum of 14000 Mpa (Smin14000)
− resistance to fatigue on prismatic specimens
according to EN 12697-24, 4PB-PR at the tem-
Figure 3. Permanent deformations in form of wheel
tracks.
perature of 10°C and frequency of 10 Hz the
minimum strain corresponding to 106 loading
cycles should be ε = 130 μm/m (ε6–130).
− Crushing of the surface layer and loss of parti-
cles on some places,
− Permanent deformations in the shape of wheel 3 HIGH MODULUS ASPHALT CONCRETE
tracks, especially in airplane trajectories along
the TWY E to the first row of parking positions 3.1 Definitions and specifications
(Figure 3).
Asphalt concrete with high stiffness modulus
(HMAC) is intended for base or binder asphalt
layers for heavy traffic roads, airports and other
2 RECONSTRUCTION PROJECT trafficked surfaces where we want a higher resist-
ance to permanent deformations and to fatigue
2.1 First area compared to bituminous concrete, for example on
At the first area (over the completely asphalt pave- slow traffic lanes, road crossings, bus lanes etc.
ment) the design of new asphalt pavement was as Higher resistance to permanent deformations
follows: and to fatigue is being reached by different mix
design and functional specifications. Character-
• AC 22 base PmB PG 64-16 with high stiffness istic feature of HMAC is a relatively fine grading
modulus in 8 cm thickness and higher bitumen content which contribute to
• AC 22 bin PmB PG 64-16 with high stiffness higher fatigue resistance, the use of harder type of
modulus in 8 cm thickness bitumen and/or polymer modified bitumen at the
• SMA 11 PmB PG 70-22 in 4 cm thickness same time enables higher resistance to permanent
(or AC 16 surf PmB PG 70-22 in 6 cm thickness deformations. Higher stiffness modulus of asphalt
on the area of 8362 m2 where only substitution layers decreases tension stresses in lower layers of
of surface layer was scheduled). pavement.
This technology was developed in France under
the name of EME (Enrobé à Module Elevé) already
2.2 Second area
in the nineties of former century and characteris-
At the second area (over the concrete pavement) the tic for it was the use of very hard bitumen in high
design of new asphalt pavement was as follows: content (till 6%).
There is a high interest for the use of HMAC
• AC 22 base PmB PG 64-16 with high stiffness
all over the world, a lot of countries already made
modulus in 12 cm thickness
use of this technology, some of the countries also
• SMA 11 PmB PG 70-22 in 4 cm thickness
meaningfully adapted French specifications to their
The drive-ins to parking positions PSN 3, 4, 5, 6 needs and boundary conditions. There is however
and 7 were designed as concrete pavement: always a high content of relatively hard bitumen
and Richness factor K greater than 3,4.
In chapter A7 of Requirements for HMAC
2.3 Additional specifications
mix design and pavement design for Slovenia, as
Additional specifications for High Stiffness Modu- a result of SPENS Project, the use of five types
lus Asphalt Concrete (or HMAC) besides already of bitumen is provided (20/30, 15/25, 1020, PmB
existing specifications in standards EN 13108-1 10/40–60 and PmB 25/55-65) with a minimal

18

TIS2017_Book.indb 18 2/8/2017 8:00:42 PM

 content 4,8% by mass, nominal aggregate size of On the test section we determined the optimum
11 mm and 16 mm, air voids content from 2 to 4%, speed of the paver, number and type of rollers and
maximum proportional rut depth at 60°C 3%, stiff- compaction pattern. It turned out that minimum
ness modulus (at 4PB, 10°C and 10 Hz) minimum temperature of the mix on site should be 170°C
14000 MPa and resistance to fatigue (at 4PB, 10°C and also heavier than usual rollers were needed,
in 10 Hz) minimum ε6–130. with mass over 10 ton. We also tried laying the
HMAC in different lifts, thickness range was from
60 mm till 120 mm, with good results—com-
3.2 HMAC design
paction rate over 98%—therefore satisfying the
Additional specifications for high stiffness modulus requirements.
asphalt concrete (resistance to permanent defor- After the satisfactory results from the HMAC
mations, high stiffness and resistance to fatigue) test production and laydown the Phase 1 of recon-
together with demands for high resistance to cli- struction works on the airport Apron began on
matic conditions at Ljubljana airport area, where 30.10.2014.
continental climate with low winter temperatures Works have been carried out in late autumn,
is predominant, dictated very careful selection of with short daylight time, every morning we usually
constituent materials, especially of the bitumen. had to wait for the fog to lift up and the air tem-
After the market survey of potential suppliers perature to rise before starting the asphalt works.
and their products we decided together with the Despite not the optimal environmental conditions
asphalt manufacturer TAČ (Ljubljana) for PmB in the first phase of the project the works were
Starfalt 10/40-65 from Austria with a declared per- done technically correct and the required minimum
formance grade index according to AASHTO of compaction rate of 98% has been exceeded on all
PG 82-28. With this bitumen and stone aggregate of the first layer, on average it was at 100,3%, and
from the quarry Črnotiče and filler Stahovica in on the second layer of HMAC reached on average
the IGMAT asphalt laboratory we made a job- 101,0% of reference asphalt mix bulk density in the
mix formula for HMAC 22 PmB with optimum laboratory.
binder content of 5,0% by mass and analysed all Average thickness of the first HMAC layer was
the required parameters from additional HMAC at 93 mm, of the second layer 87 mm, together
specifications. therefore on average 180 mm.
The testing results of additional characteristics After completion of Phase 1with surface course
were as following: SMA 11 PmB 45/80-65 FR and AC 16 PmB
45/80-65 FR in year 2014, we continued Phase 2 of
− resistance to permanent deformations according
airport Main Apron reconstruction in April next
to EN 12697-22, method B in air at the tempera-
year and finished the project in the end of May
ture of 60°C—wheel tracking rate 0,04 mm/103
2015.
passes (WTSAIR 0,04) and proportional rut depth
All the quality control result for Phase 2 were
of 1,8% (PRDAIR 1,8 )
also satisfactory, thickness of the built-in HMAC
− stiffness, on prismatic specimens according to
layer was at 130 mm, with average compaction rate
EN 12697-26, 4PB-PR at the temperature of
of 102,1%, and air voids content of 2,8%.
10°C and frequency of 10 Hz of 14196 Mpa
(Smin14196)
− resistance to fatigue on prismatic specimens 3.4 HMAC quality
according to EN 12697-24, 4PB-PR at the tem-
We checked the produced and built-in HMAC
perature of 10°C and frequency of 10 Hz-
for additional criteria of the project in terms of
the strain corresponding to 106 loading cycles
resistance to permanent deformations, stiffness
ε = 183,1 μm/m (ε6–183).
and resistance to fatigue. All the tests were done
All determined additional characteristics at IGMAT Building Materials Institute laborato-
therefore satisfied the project criteria and also all ries, on the most up-to-date testing equipment for
specifications for asphalt concrete with high stiff- dynamic testing of asphalt. Testing was done on
ness modulus, some characteristics quite clearly the cut prismatic specimens from laboratory pre-
exceeded the requirements. pared slabs with roller compactor according to EN
12697-33. Asphalt layer specimens were than tested
according to EN 12697-26 and EN 12697-24 using
3.3 HMAC production and laydown
the four point bending test on prismatic specimens
Start of the HMAC 22 base/bin PmB 10/40-65 pro- (4PB-PR) method.
duction was on 21.10.2014 in TAČ asphalt plant Test results on additional characteristics
and test section for laydown on Zaloška street in of built-in HMAC on airport Apron were as
the city of Ljubljana. following:

19

TIS2017_Book.indb 19 2/8/2017 8:00:43 PM

 grade index PG 64-16 class. PG class could not
be determined because of the too high initial kin-
ematic viscosity, however the determined critical
high temperature was at 88°C and critical low-
est temperature was at −20,9°C, that exceeds the
requirements for the Ljubljana airport.

4 CONCLUSIONS

During airport surfaces reconstruction we are always

facing special conditions and requirements for asphalt
pavement, be it greater loads and traffic situation or
Figure 4. Stiffness E* and phase angle dependent on
test frequency graph.
special working conditions. This case was the first use
of high stiffness modulus asphalt concrete in Slove-
nia and went down successfully. This proves that with
a lot of knowledge and hard work, some courage
and a little bit of luck, if not elsewhere then with the
weather, it is possible to get the base course suitable
for the high traffic loading and severe climatic con-
ditions on airport Apron and not necessarily always
implementing concrete technology for that.

REFERENCES

AP-T283-14 Technical Report, 2014 High Modulus

High Fatigue Resistance Asphalt (EME2) Technology
Transfer, Austroads.
Figure 5. Resistance to fatigue graph (deformation Bankowski W. et al., 2009 Laboratory and field imple-
dependent on load cycles number). mentation of high modulus asphalt concrete. Require-
ments for HMAC mix design and pavement design,
SPENS.
− resistance to permanent deformations according Horak, E., High Modulus Asphalt (HiMA) French devel-
to EN 12697-22, method B in air at the tempera- opment, Total.
ture of 60°C—wheel tracking rate 0,04 mm/103 IGMAT d.d., 2014 Elaborat k projektu za izvedbo—
obnova dela glavne letališke ploščadi (TWY E).
passes (WTSAIR 0,04) and proportional rut depth IGMAT d.d., 275-POA-14 Odpornost proti kolesnicam
of 1,7% (PRDAIR 1,7). IGMAT d.d., 281-POA-14 PSAZ AC 22 base PmB
− stiffness, on prismatic specimens according to 10/40-65.
EN 12697-26, 4PB-PR at the temperature of IGMAT d.d., 302-POA-14 Togost.
10°C and frequency of 10 Hz of 14913 Mpa IGMAT d.d., 303-POA-14 Odpornost proti utrujanju.
(Smin14913). IGMAT d.d., 321-POA-14 Odpornost proti kolesnicam.
− resistance to fatigue on prismatic specimens IGMAT d.d., 359-POA-14 Delno poročilo o kakovosti
according to EN 12697-24, 4PB-PR at the tem- asfalta AC 22 base PmB 10/40-65.
perature of 10°C and frequency of 10 Hz- IGMAT d.d., 8-POA-15 Togost.
IGMAT d.d., 9-POA-15 Odpornost proti utrujanju
the strain corresponding to 106 loading cycles IGMAT d.d., 78-POA-15 Poročilo o kakovosti asfalta
ε = 166,8 μm/m (ε6–167). AC 22 base PmB 10/40-65.
All the above mentioned results were inside the Le Bouteiller, E. 2012 High Modulus Asphalt The French
required specifications of the project and therefore experience, Colas, AAPA Melbourne.
Ramtech d.o.o., I-2015-77 PG-10-40-65 Test Report on
proving the quality of HMAC asphalt mixes and Performance Grade.
layers on airport Main Apron. RD-JN-2014-G2-A/3/AP 2014 Obnova dela glavne
We also analysed used bitumen Starfalt PmB letališke ploščadi in izgradnja 2. faze ceste okoli
10/40-65 in company Ramtech laboratory for satis- letališkega perimetra na letališču Jožeta Pučnika
fying the AASHTO M320 criteria for performance Ljubljana, razpisna dokumentacija.

20

TIS2017_Book.indb 20 2/8/2017 8:00:43 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Assessment of technologies for roadway energy harvesting

S. Colagrande & M. Patermo

University of L’Aquila, L’Aquila, Italy

ABSTRACT: This paper deals with the electric energy production from roads, through an overview of
the existing and experimental technologies that could guarantee, in a not so distant future, a substantial
source of power. The study analyzes the performances of several possible technologies for energy gen-
eration from road infrastructures as follows: (i) piezoelectric device able to generate electricity from the
surface movement of the road platform caused by automobile traffic; (ii) photovoltaic systems achieved
through the installation of photovoltaic panels both on noise barriers placed by the roadside and on shel-
ters of parking service stations; (iii) photovoltaic panel devices distributed on the road surface that allow
to convert solar energy into electricity and sustain high loads and stresses generated by vehicular traffic.
Finally, a practical application of these technologies on a A24 highway section located in central Italy has
been evaluated and the results are compared in terms of energy production.

1 INTRODUCTION 2 TECHNOLOGIES FOR ENERGY

PRODUCTION FROM THE ROADS
Modern Italian highway network is the result of
a design and development model conceived in the 2.1 Piezoelectric production
sixties. This implies that, from the environmental
sustainability point of view, we are actually deal- An interesting answer for the production of piezo-
ing with the technologies and the road concep- electric energy from roadways comes directly from an
tion developed 55 years ago. Nowadays, on the Israeli company, Innowatech, with its research offices
contrary, the highway should not be intended located in the Institution of Technion, Haifa (Israel).
as a pollution source because, when properly The company has created a piezoelectric device suit-
designed and requalified through the integra- able for producing electricity from the traffic-related
tion of specific systems, it can become a valuable movement of the road platform (Ronchi 2014).
energetic resource for the entire civil community Using the principle of piezoelectricity, for which
(Cappelletti 2011). some materials with crystalline structure (named pie-
It is well known that, since 1966, Italian trans- zoelectric crystals) generate electric voltage from elas-
port (both commercial and non-commercial) tic deformations, researchers developed a product to
mainly has been taking place by road and has been be installed under the asphalt pavement (or inside of
noticeably increasing year by year. In particular, the railway sleepers), able to generate electricity from
the middle-range and long-range transport is prac- the transit of vehicles (or trains) and accumulate it
ticed on the highway network. in a battery system. According to the company, this
Since the energetic problem closely regards all technology is easy and fast to be implemented both
the people and Italy, as well as European Union in the case of construction of new roads and during
states, has important goals to pursue within the maintenance operations of the existing ones.
2020 in terms of energy conservation and reduc- In particular, the vertical load by vehicle tires
tion of CO2 emissions in the atmosphere (ENEA produces a compressive stress that proportionally
2014), why the possibility to transform such a decreases with depth (Vigo Majello 2013). For this
mechanic energy into electricity should not be reason, generators designed by Innowattech are
considered? Again, why not increasing the utiliza- placed in the road pavement to a depth of 5 cen-
tion of the road spaces for installing photovoltaic timeters, where stresses produced by vehicles are
plants? The answer to these questions brought to more intense (Fig. 1).
explore the topic of the sustainable road infra- This solution, when applied to a highway, pro-
structure (Diamantini et al. 2011). duces electricity as a function of the number of

21

TIS2017_Book.indb 21 2/8/2017 8:00:46 PM

 The life cycle of a piezoelectric system is about
thirty years, but actually the system is under test-
ing and it is characterized by high implementation
costs that could be reduced if a production in mass
will be promoted (Edery-Azulay 2010, Kurzweilai
2011).

2.2 Photovoltaic systems

Nowadays, photovoltaic systems are commonly
known; they can be distinguished in rooftop or
Figure 1. Piezoelectric energy generator for road
ground-mounted systems and are able to convert
applications. solar energy into electricity. Among the multiple
uses, there are very interesting photovoltaic panel
vehicles, of their weight and speed. Consequently, applications on roadways. In particular, these pan-
the greater is the traffic and the more convenient els can be integrated on noise barriers, shelters of
will be this solution. service stations parking or even the road surface
The operational steps for the installation of pie- according to the project by the American engineer
zoelectric generators are the following: Scott Brusaw.
– Cutting of the pavement surface,
2.2.1 Photovoltaic noise barriers
– Laydown of a quick-setting concrete,
Photovoltaic Noise Barriers (PVNB) can be consid-
– Positioning of the piezoelectric generators
ered not only as a noise isolation device, but also as
(30 × 30 cm) and drowning in quick-setting
a street furniture element; they are able to improve
concrete,
the visual impact (strictly linked to noise barriers
– Connecting the cables,
installation), protect the surrounding area from the
– Overlaying the generators with an asphalt layer
noise of the highway or train routes and concur-
(to provide better adhesion between concrete
rently produce clean and renewable energy. (Fig. 2).
and asphalt layer),
The energy output of each linear meter of this
– Laydown of final asphalt wearing layer.
type of barrier is on average 169 kWh per year. This
Once completed the installation, generators col- value refers to an annual yield according to Enea
lect the mechanical energy produced by vehicles Data for the province of Pisa, considering southern
and convert it into electricity, which is then trans- exposure of the panels (Redazione Ingegneri 2009).
ferred in storage systems installed along the road- In a traditional photovoltaic system costs are
way, every 500 meters. strongly related to the initial investment for its
This system provides many advantages: implementation; in particular, the construction of
supporting structures, site purchase and preparation
– It does not require the occupation of areas adja-
determine 10% of the above-mentioned costs. In this
cent to the roadway,
case, costs related to the use of the area and support-
– It is suitable to operate in any weather
ing structures are attributed to the noise barriers.
conditions,
A considerable potential saving income of total
– It requires little maintenance (replacement or
costs is obtained with the standardization of the
repair after about 30 years),
production phase. From a more general point of
– It allows the power supply for street lighting
and/or vertical road signs.
Its application is suitable for multiple urban and
suburban applications: local streets, roads char-
acterized by heavy traffic but especially highways,
railways and subways.
First tests by Innowattech dealt with the instal-
lation of piezoelectric nano-generators along a
stretch of 10 meters in a road asphalt pavement. In
this case the generators could potentially produce
about 2 kWh, which means that one lane could
generate about 200 kWh per km. This trial allowed
to experimentally verify that the system works bet-
ter when traffic is at least 600 vehicles/hour with an
average speed of about 72 km/h. Figure 2. Photovoltaic Noise Barriers (PVNB).

22

TIS2017_Book.indb 22 2/8/2017 8:00:46 PM

 view, a considerable saving for the electricity distri- able to convert solar energy into electricity and
bution network derives from the possibility to install concurrently bear load and stresses caused by road
energy production facilities near to the points where traffic (Fig. 4).
the energy is actually used, particularly if the daily The panel, designed to substitute the asphalt
production is close to the market demand. Moreo- wearing course, is composed of three layers:
ver, PVNB should be placed close to urban area also
– Surface layer made of a rough glass, anti-abra-
for a proper acoustic protection performing.
sive, self-cleaning and highly resistant, which
In Germany and Switzerland numerous PVNB
contains photovoltaic cells and led diodes; this
structures have been installed across different
layer has the main function to resist weather-
highways and railway tracks; the longest PVNB
ing and protect the electronic apparatus located
in the world is actually in Italy, along the Brenner
underneath;
highway (A22), and it consists of 3944 panels for
– Intermediate electronic layer, which contains a
an exposition surface of 5035 m2 on a length of
microprocessor for controlling and monitoring
1070 m, which generate 690 MWh of energy per
loadings and lighting;
year (Area Tecnica Autobrennero 2015).
– Bottom layer, which carries the energy collected
by the intermediate layer to various storage sys-
2.2.2 Photovoltaic shelters
tems connected to the roadway.
Photovoltaic shelters are structures covered with
photovoltaic panels (Fig. 3), designed to cover car The produced energy is carried to storage sys-
parks, bus service or storage areas (Massa 2015). In tems located near the road surface every 12 m and
particular, photovoltaic shelters are used to protect can be directed to a primary network for satisfy-
vehicles against the weather and concurrently pro- ing various energy requirements (e.g. homes, street
duce green electricity which makes them environ- lighting and road signs, service stations).
mentally responsible. Thus, these sites, which are The proceeds would be significant in terms of
generally dedicated exclusively to parking, become energy amount: it has been estimated that, for an
solar electricity generating areas and their energy average daily solar irradiation of 4 hours, each
output could be used for equipment power supply photovoltaic panel should be able to produce
(mobile phone charging, Wi-Fi access points, bill- around 7.6 kWh per day. Considering a road sec-
boards lighting or even ventilation systems for the tion 1.6 km long and 10 m large, it could be pos-
creation of refreshment areas). sible to satisfy the energy requirement of about
A photovoltaic shelter, which has a surface of 500 families (Verdi Ambiente e Società 2014).
30 m2 for 3 kW of capacity and shades a park- However, maintenance procedures for dust accu-
ing place for 2 cars, annually produces 3610 kWh mulation, the duration of photovoltaic cells and
of electricity (Far Systems 2014). In general, this the high costs still make the photovoltaic panel for
solution is easy to be installed because it does not road surfaces in need of improvements.
require any structure construction and is adaptable The Solar Roadways is currently being tested in a
to any soil. However, a project for panel installa- section of a highway (70 km long) located between
tion is necessary in order to determine the proper Coeur D’Alene and Sandpoint in Idaho. Also in
orientation and dimensioning and achieve a better France, a photovoltaic pavement has been realized
efficiency. and named with the explicit term Wattway. The lat-
ter has been constructed by the National Institute
2.2.3 Photovoltaic panels as a road surface of Solar Energy (INES) in cooperation with Colas
As previously mentioned, the photovoltaic panel
placed on the road surface is a pioneering idea by
the American engineer Scott Brusaw who, sup-
ported by his working team, realized the “Solar
Panel Road”. This device is a photovoltaic panel

Figure 3. Photovoltaic shelters. Figure 4. Photovoltaic panels as a road surface.

23

TIS2017_Book.indb 23 2/8/2017 8:00:47 PM

 (company specialized in transport infrastructure). theoretical estimation of the energy production per
This construction represents an example of a uni- single primary segment (i.e. road segment between
fied concept of photovoltaic road surface: the two tollbooths) of both lanes has been performed.
Wattway slabs include photovoltaic cells made of First, the evaluation of the average annual traffic
polycrystalline silicon incorporated in a substrate per hour in each primary segment (reference year
few millimeters thick. At the side, the system is 2014) allowed the identification of all road segments
connected to a case that contains electronic secu- with more than 600 vehicles/h (considered for the
rity components. The slabs are antiskid, resistant, energy production estimation). Only the first three
adaptable to every surface and have been designed out of fourteen primary segments resulted suitable
to bear the traffic of all types of vehicles, includ- for the installation in the pavement structure of the
ing heavy trucks. The installation of these slabs is piezoelectric device. They are comprised between
very easy: they can be directly glued on the existing the following tollbooths: Roma East—Connection
pavement surface without any further construc- A1/A24 (segment A); Connection A1/A24 – Tivoli
tions (Energie Rinnovabili 2015). (segment B); Tivoli—Castel Madama (segment C).
As a consequence, only these three segments have
been considered for the estimation of the energy
3 PRACTICAL APPLICATION production as illustrated in Table 1.
In particular, for calculating the energy produced
Bearing in mind what described so far, an exam- by the piezoelectric device, the energy production has
ple of application of such technologies has been been considered proportional to the traffic volume.
hypothesized for the Italian highway A24 (section As reported in the table, the energy production value
Roma-L’Aquila-Teramo, located in central Italy) is higher than 200 kWh for 1 km of each segment.
and the potential energy performance has been Finally, by multiplying the so calculated energy val-
evaluated (Fig. 5). ues by the length of each corresponding segment, the
potential amount of piezoelectric energy produced
in the A24 highway results equal to 7.7 MWh.
3.1 Piezoelectric system
As previously described, on the basis of tests car-
3.2 Anti-noise photovoltaic barrier
ried out by the University of Israel in cooperation
with Innowattech company, generators have the In this paragraph, an ideal anti-noise photovoltaic
maximum productivity when they work on road barrier plant is evaluated. Based on the orography
sections with a traffic higher than 600 vehicles/h of the territory where the highway is situated, only
characterized by a speed of 72 km/h. With these viaducts nearby existing towns were considered.
conditions, the energy production is estimated As already described, in central Italy each meter of
around 200 kWh in 1 km per lane. Considering those barriers facing south can produce on aver-
the traffic data provided by “Strada dei Parchi age 169 kWh per year. Hence, considering that the
SpA”, managing company of A24 Highway, a total length of the barriers is 6390 m, the valuation
of the production is given in Table 2. The total pro-
duction results about 1081 MWh/year.

3.3 Photovoltaic shelters

This section evaluates the hypothesis of an installa-
tion of photovoltaic shelters in the parking lots sit-
uated along the A24 Highway. The calculation was
based on the squared meters of parking lots that
could be covered with shelters, equal to 13583 m2
(Table 3). Considering that 30 m2 of photovoltaic
shelters can produce 3610 kWh per year, it is pos-
sible to estimate about 1634 MWh/year.

3.4 Photovoltaic panel for road pavements

The installation of photovoltaic panel for road
pavements was hypothesized in the highway seg-
ment between L’Aquila West and L’Aquila East
characterized by a total length of 4.1 km (without
Figure 5. A24 Highway. considering the length of the three tunnels). Such a

24

TIS2017_Book.indb 24 2/8/2017 8:00:48 PM

 Table 1. Energy produced by the piezoelectric device in the A24 Highway.

Traffic vol. [vehicles/h] Energy per km [kWh] Tot. Energy [MWh]

Sect. Right lane Left lane Right lane Left lane Length [km] Right lane Left lane

A 1066 944 355 315 3.2 1.1 1.0

B 881 850 294 283 1.5 0,5 0.4
C 630 624 210 208 11.2 2.4 2.3
SUM 4.0 3.7
TOTAL 7.7

Table 2. Energy produced by the anti-noise photo-

voltaic barriers in the A24 Highway.

Energy produced
Viaduct name Length [m] [MWh/year]

Licenza II 180 30
Roviano 175 30
Colle Alto 350 59
Peschieto 2 245 41
Peschieto 1 85 14
Valle Mura 270 46
Fiume Aterno 105 18
SS17 80 14 Figure 6. L’Aquila West—L’Aquila East Highway sec-
Fosse Vetoio 200 34 tion (source: Google Maps).
Pettino 485 82
San Sisto 1755 297 project will be justified by the presence of the near
San Giacomo 390 66 city of L’Aquila that will directly use the energy
Cerchiara 1280 216 produced (Fig. 6).
Villa Ilii 790 134 It was considered that the two separate carriage-
TOTAL 6390 1081 ways (one for each direction) would be covered
by photovoltaic panels; each carriageway width is
10 m (two lanes with the width of 3.5 m and one
Table 3. Summary of parking lots and service stations shoulder with the width of 3.0 m). Considering
and energy produced by the photovoltaic shelters in the that a single panel with an area of 1 m2 should be
A24 Highway. able to produce 7.6 KWh/day (based on an average
daily irradiation of 4 hours), the selected segment
Surface Tot. Energy (4100 m of total length and 20 m of total width)
Type Name [m2] [MWh/year] covered by photovoltaic panel should produce a
total of 623.20 MWh/day.
Service station Colle Tasso 450 54.15
North
Service station Colle Tasso 1042 125.33
South 4 CONCLUSION
Parking lot Roviano 1675 201.56
East From the result analysis it can be deduced that
Parking lot Roviano 3406 409.81 the use of photovoltaic systems can determine the
West production of significant amounts of energy. It is
Service station Civita North 798 96.03 clear that the piezoelectric technology is also very
Service station Civita South 1436 173 promising, even if the benefits/costs ratio must be
Service station Valle Aterno 498 59.93 considered.
West This study aimed to demonstrate the potenti-
Service station Valle Aterno 582 70.03 ality of a highway to become an energy genera-
East tor for the territory that it crosses. The proposed
Parking lot Gran Sasso 3696 444.75 hypotheses are purely theoretical and they have
TOTAL 13583 1634.59 not been deepened on a proper scale to verify their
applicability. However, the paper describes new

25

TIS2017_Book.indb 25 2/8/2017 8:00:49 PM

 technological systems for the next-generation The Highway to Innovation, Israel national roads
design of the binomial infrastructure/territory, company Ltd., 1–3 November, Tel Aviv, Israel.
with the objective to provide a different approach Enea 2014. Documento di predisposizione del Piano
and modify the timeworn concept of the highway d’Azione italiano per l’Efficienza Energetica
(PAEE2014).
as a pollution source. Energie Rinnovabili 2015. Wattway—anche in Francia
Lastly the energy produced will be used as much arriva la strada fotovoltaica. URL rinnovabili.it.
as possible during its production, otherwise must Far Systems SpA 2014. Pensilina Fotovoltaica “Boston”.
be stored in appropriate batteries. URL farsystems.it.
Kurzweilai 2011. Innowattech attach harvests mechani-
cal energy from roadways. URL kurzweilai.net.
Massa, M.F. 2015. Pensiline fotovoltaiche: come alimen-
REFERENCES tare auto elettriche, cellulari, pannelli pubblicitari con
una pensilina fotovoltaica. URL fotovoltaicosulweb.
Area Tecnica Autobrennero 2015. A22 Viaggia verso il it.
sole verso il futuro—barriera antirumore fotovoltaica. Redazione Ingegneri 2009. Barriere Antirumore Foto-
URL autobrennero.it. voltaiche. URL ingegneri.cc.
Cappelletti, G. 2011. L’autostrada come condotto ener- Ronchi, F. 2014. Da Israele l’Asfalto che ricarica le
getico: scenari per la produzione di energia rinnovabile batterie—Energia gratuita da strade e ferrovie. Sci-
lungo la pedemontana lombarda. Graduation Thesis, enza e Tecnologia. URL lacritica.org.
Faculty of Architecture, Politecnico di Milano. Verdi Ambiente e Società 2014. In USA pannelli solari al
Diamantini, C., Scaglione, G., Rizzi, C., Staniscia, S., posto delle strade asfaltate. URL vasonlus.it
Ricci, M. & Cribari, V. 2011 Reinventing A22 Eco- Vigo Majello, M.C. 2013. Sistemi innovativi per
boulevard. Verso infrastrutture osmotiche. XIV SIU l’implementazione energetica del patrimonio edifi-
Conference – 24/26 March. cato: dispositivi piezoelettrici applicati alle pavimen-
Edery-Azulay, L. 2010. Innowattech: Harvesting tazioni. PhD Thesis. Università degli studi di Napoli
Energy and Data. First International Symposium “Federico II”.

26

TIS2017_Book.indb 26 2/8/2017 8:00:50 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Loss of life risk due to impacts of boulders on vehicles traveling

along a very busy road

P. Budetta, G. Forte & M. Nappi

Department of Civil, Architectural and Environmental Engineering, University of Naples “Federico II”,
Naples, Italy

ABSTRACT: The paper is aimed to describe the used approach for calculating the risk along a road
stretch belonging to a very busy coastal road in Southern Italy. During the time span 1969–2013, 22 rock-
falls affecting this road were inventoried. On 18th February 2014 a new rockfall happened and several
boulders reached the northern lane of the road. On the basis of collected data concerning the landslide
hazard and road vulnerability, a procedure for the probability evaluation of a fatal accident—for a road
user—is presented, discussed in details and compared. The analysis is meant to allow the design of appro-
priate protection devices along the cliffs overhanging the road.

1 INTRODUCTION (RO.MA.) method by Peila & Guardini (2008),

concerning a very busy road affected by recurrent
Rockfalls on transportation corridors can cause rockfalls.
casualties mainly due to direct impact on vehicles Figure 1 presents a schematic operational
or by impact of vehicles with deposed material, as flowchart of the RO.MA. approach.
well as a large amount of economic consequences This method develops through five steps, includ-
due to traffic interruptions. The rockfall risk evalu- ing (Figure 1): identification of unstable areas and
ation affecting roads is a highly complex operation the number of rocks per year that may hit the road
requiring the assessment of the hazard (trigger- (Nr); road vulnerability assessment; event tree
ing mechanisms and the run out parameters) and analysis; risk assessment; evaluation of the effi-
the vulnerability of vehicles on the roads along ciency of rockfall protective measures and calcula-
the foothills. Vulnerability mainly depends on the tion of the residual risk (Peila & Guardini 2008;
vehicle speed and length, the available decision Mignelli et al. 2012). The number of boulders that
sight distance, the traffic volume, the length of may hit the road or that may stop upstream (Ns), is
the rockfall risk section of the route, the number obtained from field data or, alternatively, through
of occupants in a vehicle, and the type of vehicle. trajectory simulations (Jaboyedoff et al. 2005).
In quantitative terms, the annual probability of For the vulnerability evaluation it is need to have
loss of life to an individual is given by multiplying data concerning the length of the hazardous road
the annual probability of occurrence of a rockfall stretch (Lr), the average (or limit) speed of the vehi-
event (of a given magnitude) by the probability cles (Vv), the average vehicle length (Lv), and the
of a falling rock hitting the moving vehicle, and number of vehicles travelling on the road per hour
by the vulnerability of the person given a block (Nv). The complete sequence of events which may
of size m impacting the vehicle. In order to apply result from a rockfall up to the killing of a road user
this approach, the number of boulders that may (fatal accident) is evaluated by means of the event
hit the road must be obtained through trajectory tree analysis. This analysis develops along twelve
simulations by calculating the percentage of all tra- different paths (Figure 1A) and the probability of
jectories that could fall on the road or that are not occurrence of each of them can be calculated from
interfering with it. the product of each single event that constitutes
In the literature several methods concerning the the path itself. The final value of each path is given
Quantitative Risk Analysis (QRA) affecting roads by the sum of probabilities concerning paths with
have been proposed (Bunce et al. 1997; Hungr et al. the same final result (i.e. fatal accident, non-fatal
1999; Budetta 2002; Peila & Guardini 2008; Ferlisi accident, and no accident).
et al. 2012; Mignelli et al. 2012; Budetta et al., in We partially modify the approach given by Peila
press). Here, we show results obtained by means & Guardini (2008), to calculate the spatio-temporal
of application of the ROckfall risk MAnagement probability (PST) for a vehicle to be in the path of the

27

TIS2017_Book.indb 27 2/8/2017 8:00:51 PM

 Figure 1. Procedural flow chart for ROckfall risk MAnagement assessment (A); Abacus defining the threshold values
of rockfall risk. Acceptable values are lower than 10–5, the ALARP band is 10–5, unacceptable risk values are upper than
10–5 (B); histogram representing risk values versus the representative road sections (C). After Peila & Guardini (2008)
and Mignelli et al. (2012) modified.

falling boulder, when the mass falls, using both the NV x(We Lv )
PST = (1)
dimension of the falling block (We) and the length
of the vehicle (Lv) (Nicolet et al. 2015) as follows: Vv

28

TIS2017_Book.indb 28 2/8/2017 8:00:52 PM

 where PST = spatio-temporal probability; Nv =
number of vehicles travelling on the road per hour;
We = dimension of the falling block; Lv = length of
the vehicle; Vv = average speed of the vehicles.
The probability of a fatal accident due to a
direct impact of a boulder on a moving vehicle was
calculated according to Bunce et al. (1997) whereas
considering the impact of a block on the road sur-
face, the probability that a travelling vehicle has
an accident due to the damaged road surface was
evaluated as a function of the deposed volume
because generally volumes more than 0.1 m3 may
cause deep potholes (Budetta et al., in press).
Another reason for which we preferred the sug-
gested approach by Peila & Guardini (2008) is Figure 2. The harbor of Salerno and the surrounding
hills. The white rectangle shows the study area.
given by the possibility to evaluate the ability of
installed protection measures along the cliffs (e.g.
rockfall barriers, reinforced wire rope nets, mesh
drapes, etc.) to stop the falling rocks, reducing the
number of those that may hit the road. The number
of retained blocks (N’r) can be calculated on the
basis of the catching capacities of these devices, by
means of the percentage of rocks (C) that can be
stopped by the protection device. N’r is given by:

N’r = (1 – C) Nr (2)

In this way, we can compare calculated risk

values (without protection measures) with those
obtained taking into account the efficiency of
these protections, or with risk criteria defined in
the international literature. In this respect, Mignelli Figure 3. The rockfall occurred on 18th February
et al. (2012) suggested the use of an abacus defin- 2014.
ing the threshold values of acceptable and unac-
ceptable rockfall risks (Figure 1B). joints with the cross-bedded dolomitic layers result
The above-mentioned approach, which has in isolated, potentially unstable boulders. Almost
been partially modified with respect to the original vertical slopes contribute to the free fall of boul-
one by Peila & Guardini (2008), was applied to a ders on the road, whereas in the remaining cases
350 m-long road stretch belonging to a coastal road irregular rock faces, due to the presence of ridges
that links the Vietri sul Mare resort with Salerno. or benches with lower slopes, cause launching and
rebounding phenomena.
Along the entire road stretch linking Vietri sul
2 GEOMORPHOLOGICAL AND Mare with Salerno, between 1969 and 2013, 22
GEOLOGICAL SETTING rockfalls were inventoried based on the information
in the Italian Landslide Inventories (AVI and IFFI
The studied road crosses a coastal area character- Catalogues), historical documents, and newspapers.
ized by high reliefs lying near the harbor of Salerno Some rockfalls of about several cubic metres caused
(Fig. 2). Given the near vertical cliffs and very prolonged traffic interruptions. About 13 rockfalls
steep slopes, in a short distance from the coast, the occurred during the period 1993–1997, whereas
relief goes from the sea level to heights greater than only five landslide events are reported in the subse-
700 m ASL. On slopes flanking the road, heavily quent period up to 2013. Likely, this is due to pas-
fractured dolomites rocks outcrop. Some low- sive devices (wire nets) which, during that time, were
angle normal faults and several pervasive mutu- installed along cliffs above the road. Despite the
ally intersecting joints affect the rock mass. These considerable number of recorded rockfalls, there is
joints identify wedges whose lines of intersection incomplete information regarding the magnitudes
are almost normal to the cliff face. Other randomly as well trajectories and stopping points. On 18th
oriented deep tension cracks are caused by tensile February 2014, a new rockfall of about 40 m3 affected
stresses along cliffs. The intersections of these the road at the kilometer 51+300 (Figure 3).

29

TIS2017_Book.indb 29 2/8/2017 8:00:53 PM

 Figure 4. The trajectories that could stop on the road or that are going beyond it. With the black lines, we show the
real trajectories traveled by the boulders of the rockfall of 2014. With the dotted gray lines, we show trajectory simula-
tions obtained using kinematic parameters inferred by means of the back-analysis of the rockfall. Only trajectories that
may interfere with the road are shown.

Several boulders falling from the overhanging to analyze topographic spatial models by means
cliff reached the northern lane of the road (in the of deterministic and/or probabilistic (by a Monte
direction towards Vietri sul Mare) threatening an Carlo sampling technique that uses a normal dis-
adjacent petrol station and causing a new pro- tribution) approaches. In this code the topographic
longed traffic interruption. The average volume surface is generated using the Delaunay triangula-
of the deposed boulders on the road was evalu- tion method. As the program uses a lumped-mass
ated about 65 dm3. It is worth to observe that the method, possible boulder fragmentation during
collapsed cliff was already protected by wire nets the fall is not considered.
which, evidently, were unfit to arrest the blocks. Afterward, using kinematic parameters obtained
from the back-analysis, a hazard scenario concern-
ing probable trajectories arising from other poten-
3 TRAJECTORY SIMULATIONS tially unstable rockfall sources, has been prepared
for falling blocks with volumes equal to 65 dm3
On the basis of deposed block volumes as well (Figure 4).
run-out distances and cliff geometries, several The rockfall sources have been carefully chosen
trajectory simulations were performed using a 3D on the basis of the morphology of the area and
approach in order to obtain more suitable energy the starting point for boulders were conventionally
restitution (normal, RN and tangential, RT) and located at the highest point of each source (in most
rolling (friction angle, φ) coefficients as well kinetic cases corresponding with the cliff top). Each simu-
energies, velocities, and bounce-height along tra- lation consisted of releasing 10,000 blocks from
jectories travelled by boulders. In such a way, a any potential rockfall source. In order to facilitate
detailed back-analysis of the rockfall of 2014 was the reading of the map, in Figure 4 only boulder
performed. Trajectory simulations were performed trajectories that could stop on the road or that are
using a three-dimensional code (AZTECROCK going beyond it have been shown. In such a way,
10.0, by Aztec Informatica Inc. 2012) designed along the total length of the road potentially prone

30

TIS2017_Book.indb 30 2/8/2017 8:00:55 PM

 to rockfalls (about 350 m) it is possible to measure with volumes of 65 dm3, result in dangerous pot-
the lengths of sections exposed to boulder impacts holes that may cause serious damage to vehicles
and more precisely calculate the probability that (Budetta et al., in press).
a moving vehicle is located in the section covered In order to perform repeated and complex cal-
by the event when it happen (Figure 4). Exclud- culations, an Excel spreadsheet was prepared that
ing the effective length of the road section affected uses the above-mentioned input data. In such a
by blocks coming from the rockfall of 2014 (about way, it was possible to obtain probability values
5.0 m), other calculated potential lengths range concerning the 12 paths of the event tree. Finally,
between 9.0 and 3.6 metres. It is worth noting by the sum of values of identical outcomes, proba-
that, at this stage, the effect of installed protec- bility values of a fatal accident, non-fatal accident,
tion devices (wire nets) flanking the road that are and no accident were calculated.
able to stop and/or alter boulder trajectories, is not As the rockfall of 2014 already proved that the
taken into account. existing wire nets were unable to retain block vol-
umes of 65 dm3, it was not considered necessary to
recalculate N’r on the basis of the percentage of rocks
4 FATAL ACCIDENT RISK (C) that can be stopped by this protection device.
CALCULATION Finally, for the entire road stretch potentially
prone to rockfalls, the annual risk of fatal accident
In order to calculate the risk, expressed as the for a road user is 4.52 × 10–3. According to the aba-
annual probability of a fatal accident for a road cus defining values of acceptable and unacceptable
user, vulnerability concerning this stretch of the rockfall risks (Figure 1C) suggested by Mignelli
road firstly was evaluated. On the basis of traffic et al. (2012), the individual risk is not acceptable,
data furnished by the Municipality of Salerno, a and some actions are required in order to lower it.
mean hourly traffic density (number of vehicles
travelling on the road per hour – Nv) of 4500 was
established. The mean velocity (Vv) and length (Lv) 5 CONCLUSIONS
of vehicles were 50 km h–1 and 4.5 m, respectively.
The mean block diameter (We) inferred from field It is useful to compare the fatal accident risk value
surveys was 0.5 m. Regarding other hazard data calculated by means of the modified RO.MA.
needed for risk calculation, on the basis of trajec- method with those concerning all car accidents
tory simulations the number of rocks hitting the resulting in loss of life in Campania (the region
road per year (Nr) was assumed equal to 1.15. where the studied area is located). During the time
According to eq. 1, the probability that a vehicle span 1996–2008, available data from the Italian
travelling on the road is hit by a falling rock (PST) is Institute for Statistics (ISTAT 2014) shows a mean
1.87 × 10–2. For simplicity, we assume that through- value of about 330 fatalities/year or 3.41 × 10–2
out the year the traffic is constant. fatalities year–1 km–1: the length of the entire Cam-
Associated with the 12 paths of the event tree pania road network (motorways, national and
(Figure 1A), the fatal accident probability value provincial roads) being about 9652 kilometres.
was calculated for the above-mentioned hazard Another comparison can be done with the fatal
scenario, using the approach by Peila & Guardini risk caused by rockfalls on the Amalfitana road
(2008). Unlike the original method, we used the since its construction, in the second half of the
calculated PST value because neglecting the dimen- 19th century (1.65 × 10–4) (Budetta et al., in press).
sion of the block or the dimension of the vehicle This state road crosses an area characterized by a
might lead to inexact results (Nicolet et al. 2015). geomorphological and geological layout similar to
According to Bunce et al. (1997), probability the studied one. The rockfall risk value calculated
values of a fatal accident due to moving vehicle/ for the study area is also above the acceptability
falling rock and moving vehicle/fallen rock inter- limit defined for ‘‘involuntary’’ risk such as rock-
actions (Figure 1A) of 0.2 and 0.1 were assumed, falls, as proposed by Geotechnical Engineering
respectively. Furthermore, the probability of a Office Hong Kong (1998). This means that the
fatal accident due to an impact of a falling rock individual risk is not acceptable, and more effec-
on a stationary vehicle was assumed equal to 0.125 tive countermeasures are required such as rockfall
(Bunce et al. 1997). Finally, the probability of seri- barriers or reinforced wire rope nets.
ous damages on the road surface and consequent The rockfall hazard evaluation and the assess-
fatal accident, due to the damaged road surface, ment of the risk performed by means of the above-
was 0.7. This value assesses the likelihood that a mentioned approach are affected by uncertainties
pothole produced by a falling and rebounding and limitations that we must bear in mind in order
block on the asphalt is more than the wheel diam- to use the results for risk mitigation and planning
eter. In other words, 70% of rebounding blocks, purposes. The quality of the rockfall hazard sce-

31

TIS2017_Book.indb 31 2/8/2017 8:00:56 PM

 nario depends on the preciseness of trajectory sim- Ferlisi, S., Cascini, L., Corominas. J. & Matano, F. 2012.
ulations and the accurate identification of rockfall Rockfall risk assessment to persons travelling in vehi-
sources. cles along a road: the case study of the Amalfi coastal
With reference to the risk analysis, major dif- road (southern Italy). Nat Hazards 62:691–721.
Geotechnical Engineering Office 1998. Landslides and
ficulties concern the assessment of more exact boulder falls from natural terrain: interim risk guide-
probability values, which must be assigned to inter- lines. GEO report no. 75. Geotechnical Engineering
actions between rocks and vehicles. Sometimes the Office, The Government of the Hong Kong Special
definition of these values was heuristic, and to Administrative Region: pp 183.
some extent arbitrary. Hungr, O., Evans, S.G. & Hazzard, J. 1999. Magnitude
Even though this quantitative risk analysis must and frequency of rock falls and rock slides along the
be improved, the applied methodology demon- main transportation corridors of south-western Brit-
strates that it is possible to perform a good assess- ish Columbia. Can Geotech J 36:224–238.
ment of the expected individual loss of life, since Italian Institute for Statistics ISTAT 2014. Anno
2013 incidenti stradali in Campania. Statistiche focus.
it is based on the essential elements concerning the http://www.istat.it. Accessed 18 June 2015 (in Italian).
rockfall hazard evaluation and road vulnerability. Jaboyedoff, M., Dudt, J.P. & Labiouse, V. 2005. An
attempt to refine rockfall hazard zoning based on the
kinetic energy, frequency and fragmentation degree.
REFERENCES Nat Hazards Earth Syst Sci 5:621–632.
Mignelli, C., Lo Russo, S. & Peila, D. 2012. Rockfall risk
Aztec Informatica Inc.: AZTECROCK v. 10.0 2010.
management assessment: the RO.MA. approach. Nat
Caduta massi. Casole Bruzio, Cosenza, Italy.
Hazards 62:1109–1123.
Budetta, P. 2002. Risk assessment from debris flows in
Nicolet, P., Jaboyedoff, M., Cloutier, C., Crosta, G.B. &
pyroclastic deposits along a motorway, Italy. Bull Eng
Lévy, S. 2015. Brief communication: on direct impact
Geol Env 61:293–301.
probability of landslides on vehicles. Nat. hazards
Budetta, P., De Luca, C. & Nappi, M. In press. Quanti-
Earth Syst Sci Discuss. DOI 10.5194/nhessd-3-7311-
tative rockfall risk assessment for an important road
2015.
by means of the rockfall risk management (RO.MA.)
Peila. D. & Guardini, C. 2008. Use of the event tree
method. Bull Eng Geol Environ. DOI 10.1007/s10064-
to assess the risk reduction obtained from rock-
015-0798-6.
fall protection devices. Nat Hazards Earth Syst Sci
Bunce, C.M., Cruden, D.M. & Morgenstern, N.R. 1997.
8:1441–1450.
Assessment of the hazard from rockfall on a highway.
Can Geotech J 34:344–356.

32

TIS2017_Book.indb 32 2/8/2017 8:00:56 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A big data approach to assess the influence of road pavement

condition on truck fleet fuel consumption

F. Perrotta, T. Parry & L. Neves

Faculty of Engineering, Nottingham Transportation Engineering Centre, University of Nottingham,
Nottingham, UK

ABSTRACT: In Europe, the road network is the most extensive and valuable infrastructure asset.
In England, for example, its value has been estimated at around £344 billion and every year the govern-
ment spends approximately £4 billion on highway maintenance (House of Commons, 2011).
Fuel efficiency depends on a wide range of factors, including vehicle characteristics, road geometry,
driving pattern and pavement condition. The latter has been addressed, in the past, by many studies show-
ing that a smoother pavement improves vehicle fuel efficiency. A recent study estimated that road rough-
ness affects around 5% of fuel consumption (Zaabar & Chatti, 2010). However, previous studies were
based on experiments using few instrumented vehicles, tested under controlled conditions (e.g. steady
speed, no gradient etc.) on selected test sections. For this reason, the impact of pavement condition on
vehicle fleet fuel economy, under real driving conditions, at network level still remains to be verified.
A 2% improvement in fuel efficiency would mean that up to about 720 million litres of fuel (∼£1 billion)
could be saved every year in the UK. It means that maintaining roads in better condition could lead to
cost savings and reduction of greenhouse gas emissions.
Modern trucks use many sensors, installed as standard, to measure data on a wide range of parameters
including fuel consumption. This data is mostly used to inform fleet managers about maintenance and
driver training requirements. In the present work, a ‘Big Data’ approach is used to estimate the impact
of road surface conditions on truck fleet fuel economy for many trucks along a motorway in England.
Assessing the impact of pavement conditions on fuel consumption at truck fleet and road network level
would be useful for road authorities, helping them prioritise maintenance and design decisions.

1 INTRODUCTION by road vehicles (Haider et al. 2011). Due to con-

tinuous reliance on fuel oil derivatives, road trans-
One of the most important challenges that the world port is an important cause of energy consumption
is facing nowadays is the reduction of Greenhouse and GHG emissions.
Gas (GHG) emissions. This problem has been high- Fuel economy is an essential part of the Life-
lighted in the road transport industry and by road Cycle Assessment (LCA) analysis of road pave-
agencies asking themselves how they can play a role ments but frequently not taken into account due to
in this. Agency decisions influence design, construc- inadequacy of existing models (Zaabar and Chatti,
tion and maintenance of roads, but since European 2010) and lack of a standard and widely accepted
road networks are fairly complete, one major impact methodology (Santero, Harvey and Horvath 2011).
of agencies is in reducing the emissions of GHGs Therefore, it is difficult to conduct a complete and
through adequate maintenance policies. reliable assessment.
In England—as in other countries across Fuel consumption is influenced by vehicle tech-
Europe—the road network is the most extensive nology, road geometry, weather conditions, vehi-
and valuable infrastructure asset. Recent estimates cle speed, and pavement condition among other
assess that in England its value is about £344 bil- factors (Beuving et al., 2004; Zaabar and Chatti,
lion and every year the government spends approx- 2010; and Haider et al. 2011). Sandberg (1990) and
imately £4 billion for maintaining highways (House Laganier and Lucas (1990), among others, have
of Commons, 2011). concluded that rougher roads lead to higher fuel
Although the transport industry is one of the consumption. In particular the impact of road
major contributors to the economic growth of a roughness on fuel consumption has been assessed
country, recent estimates assessed that a quarter of at approximately 5% (Zaabar and Chatti 2010) of
the total energy consumption in Europe is used in the whole vehicle fuel consumption. This means
transportation, with more than 80% of it consumed that well designed and well maintained roads can

33

TIS2017_Book.indb 33 2/8/2017 8:00:57 PM

 influence road vehicle fuel economy, reducing their ments have been taken under carefully controlled
impact on the environment. conditions, on flat roads, testing only a few vehi-
Looking at the latest available statistics it is cles driving on a few selected road segments. For
possible to see that about 36 billion litres of fuel this study data from 43 trucks driving for one year
are used every year by road vehicles circulating in over various segments of the M18 motorway in
England (Department for Transport, 2015), so road England are considered. 1970 data points for these
condition accounts for around £1 billion every year articulated trucks equipped with 12,419cc, euro 5
(depending on the price of fuel). This is equivalent engine, driving on M18 in 2015 at 85 (±2.5) km/h
to 25% of the total annual investment in road main- using gear 12 are available. These data were chosen
tenance on highways in England. Maintaining some for this initial study from hundreds of thousands
roads in a better condition could therefore lead to of records, to remove the influence of truck type,
environmental mitigation and overall cost savings. speed and gear from the analysis.
However, conclusions of previous studies of the This and other sensor data are collected daily
impact of road condition on fuel consumption can- and analysed to assist truck fleet managers in deci-
not be considered completely exhaustive. In fact sion making about vehicle maintenance and driver
previous studies consider just a few vehicles driven training.
on a few selected road segments, in specific road Each record contains data recorded at an ‘event’
conditions, with specific road geometry. Because during the journey of each truck (license number
of this, the impact of road conditions on vehicle and tracker ID define the specific truck). At any
fuel consumption at network level under real driv- brake, stop or anomaly in vehicle performance etc.,
ing conditions still remains unclear. or each 2 minutes (120 s) or 2 miles (∼3,219 m), an
In the era of ‘Big Data’, large amounts of data are event is generated and a data record taken.
collected every day and this includes for road vehi- Each vehicle’s performance database contains
cles. Modern trucks are equipped with a wide range hundreds of gigabytes of data but not every record
of sensors defined in SAE J1939 (SAE International is stored or is useful for the data analysis. The data
2002). These sensors collect data about vehicle per- accessed during this study are:
formance (fuel consumption, vehicle speed and
other factors) and they are usually analysed to help − the vehicle profile, identifying the vehicle,
truck fleet managers in their truck maintenance and − the tracker ID, reference for the system of sen-
driver training decision making processes. At the sors installed on the vehicle,
same time, every year, road agencies collect data on − the geographical position of the truck, given in
road condition (roughness, skid resistance, texture, the British National Grid (latitude and longitude)
and stiffness among other parameters). coordinate system (5 metres GPS precision),
The objective of this work is to compare truck − the distance travelled by the vehicle since the
fuel consumption and road condition data using previous event (metres),
a Big Data approach. This will lead to the defini- − the time spent by the vehicle to travel to the cur-
tion of more accurate models for the LCA of pave- rent position from the previous event (seconds),
ments. Using this approach, almost real time data − the total fuel consumed until the current event is
can be used and any generated model can be pro- recorded (0.001 litres precision, rounded to 0.1
gressively updated with time, following improve- litres for the purpose of reducing database size),
ments in vehicle and road technology. This paper − the air temperature (0.1ºC precision),
reports some initial results exploring the feasibility − the current gear,
of this approach. − the current engine torque percentage,
− the engine revolutions (revs/min).
HAPMS (the Highway Asset Performance Man-
2 AIM AND OBJECTIVES agement System) is a database owned by Highways
England (the agency) which contains all data col-
The main aim of this study is to investigate the fea- lected during annual surveys of the condition of
sibility of using a Big Data approach to assess the the road network. The database contains histori-
impact of road pavement conditions on truck fleet cal information about each road of the network,
fuel consumption at road network level under real including:
driving conditions.
− a road identifier code,
− a direction code,
3 DATA − the year of construction,
− the latest date of significant maintenance,
A limitation of previous studies of vehicle fuel − the construction materials,
consumption and road conditions is that measure- − roughness measurements,

34

TIS2017_Book.indb 34 2/8/2017 8:00:57 PM

 − texture measurement, steady speed, for a certain type of vehicle can be
− skid resistance measurement, separately considered and analysed.
− deflection measurement (not considered in this In this study only data for 3-axle tractor with
study). 3-axle trailer articulated trucks at 85 km/h using
gear 12, equipped with 12,419cc euro 5 engines are
Each record refers to a 10 metre segment of considered. This has been done in order to avoid
road. Data are linked to their geographical posi- the influence of speed, engine performance and
tion by coordinates (in WGS84 coordinate system). vehicle type, and results in 1970 records.
Because of the different coordinate systems used in
data collection for vehicle performance and road
pavement condition, the coordinates of vehicle per-
formance are transformed to the WGS84 coordinate 5 RESULTS
reference system in order to compare the data.
In this initial study only information about road Comparing the vehicle fuel consumption (as litres
surface condition is taken from HAPMS. This per 100 kilometres) and the road condition, a pre-
includes measurements of roughness (Longitudi- dictive model has been generated. Data about the
nal Profile Variance, LPV, at 3 and 10 metres in engine torque percentage (as a substitute for the
mm2), texture (sensor-measured texture depth in vehicle payload), the road gradient, LPV at 3 and
mm), skid resistance (SCRIM Coefficient), and 10 metres wavelength and texture measurements
road gradient (0.1% resolution). Although previ- are considered. First each variable has been related
ous studies used measurements like IRI (Interna- to fuel consumption separately in order to estimate
tional Roughness Index) and MPD (Mean Profile the correlation between each single variable and
Depth) as roughness and texture measurements fuel economy.
they are similar to the LPV and texture measure-
ments used in this study. Road surface roughness FC = 25.46 + 0.12T % (1)
and texture have been shown to have a significant FC = 26.15 + 7.18g % (2)
effect on fuel consumption in previous experimen-
tal studies (Sandberg, 1990, Laganier and Lucas, FC = 28.36 + 3.14 LPV
V 03 (3)
1990, du Plessis, Visser and Curtayne, 1990, Beuv- FC = 28.46 + 0.53LPV
V 10 m (4)
ing et al., 2004, Zaabar and Chatti, 2010, Haider
FC = 26.17 + 1.80t (5)
et al., 2011, Zaabar and Chatti, 2012).
where, FC = predicted fuel consumption
[l/100 km]; T% = engine torque percentage on
4 METHOD maximum available [%]; g% = road gradient [%];
LPV03 = Longitudinal Profile Variance at 3 m
A Big Data approach has been taken in this study. wavelength [mm2]; LPV10 = Longitudinal Profile
910,591 events have been recorded and are avail- Variance at 10 m wavelength [mm2]; t = texture
able in total from trucks driven on M18 in 2015. depth [mm].
In order to limit the variables at this initial stage, Based on the literature review the payload and
the data were filtered for truck model, vehicle the gradient are two of the most influential vari-
speed, engine type, road geometry, etc. ables (Beuving et al., 2004). However, as already
In the analysis reported here the following sub- mentioned, no data about the payload is currently
set of data was used where the records show: available for this study. Therefore the torque gener-
ated by the engine for moving the vehicle (moving
− records for which the trigger event was the
a certain payload) is used instead.
default time or distance (i.e. no other driving
The relationship between the fuel consumption
event (e.g. harsh braking or cornering) triggered
and the road gradient is shown in Figure 1.
the record),
Including only two variables, T% and g% in the
− an average speed of 85 km/h
model we obtain:
− the initial and final speed were similar to the aver-
age speed (calculated as the travelled distance
divided by the travelled time) and do not differ FC = 24.34 + 0.068T % 6 65 g % (6)
by more than 2.5 km/h (this assumption makes
Then, performing a backward analysis based on
sure that the vehicle speed is steady between the
the Aikake Information Criterion (AIC, Aikake,
two considered events),
1973) the profile variance and texture data have
− no gear change.
been included in the model. Among all the models
No data about the truck payload is currently generated the one that shows the lowest AIC coef-
available. Applying these filters, all the data at ficient is:

35

TIS2017_Book.indb 35 2/8/2017 8:00:57 PM

 Figure 1. Plot of the direct impact of road gradient [%] on truck fleet fuel consumption [l/100 km] for articulated
trucks driven at 85 km/h equipped with 12,419cc euro 5 engine.

Figure 2. Plot of the predicted value of fuel consumption [l/100 km] with the real measurements for articulated trucks
driven at 85 km/h equipped with 12,419cc euro 5 engine.

FC = 19.188 + 0.066T % 6 85 g % + 1.91LPV 10 and the measured fuel consumption is shown in

+ 2.77t (7) Figure 2.

The correlation coefficient (r) between the 6 DISCUSSION

predicted and the measured fuel consumption The generated model considers a limited number
is 0.54. The correlation between the predicted of road condition variables (avoiding overfitting)

36

TIS2017_Book.indb 36 2/8/2017 8:00:58 PM

 including those ones that previous studies consid- and consider the fuel consumption of the unladen
ered to be the most influential on fuel economy. It vehicles for the rest of their journey. It may also be
assesses the impact of road roughness on fuel con- possible to collect data records more frequently and
sumption to be up to 4.1% and up to 1.25% for the to record the fuel consumption more precisely for a
texture. Zaabar and Chatti (2010 and 2012) esti- limited number of trucks over a limited time period
mated the impact of roughness on fuel consump- but this approach will be limited by the size of the
tion at 5% and our first estimate from this initial generated data set and hence cost. Furthermore, it
work lies slightly below that. may be possible to add meteorological data (e.g.
Beuving (2004) reported that for a truck driving wind speed and direction) but because this data is
at 80 km/h the impact of other variables such as the routinely reported only over extended time periods
vehicle mass and the gradient of the road is about (e.g. average per hour), it may be of limited use.
40% of the entire fuel consumption. The model It is anticipated that including more reliable data
estimates that the impact of torque (payload) and using these approaches will improve the generated
gradient measured for this data set is about 41.6%. model, which can then be extended to further types
Compared to experimental studies this gives confi- of truck and speeds and a wider road network.
dence in this Big Data approach. These models of the influence of road condi-
Although the results match what was found tion on fuel consumption can be used to validate
previously by other authors in experimental stud- those from experimental studies and extend them
ies, they cannot be considered valid as yet. In fact, to truck fleet and road network level. This will
in this initial study only specific conditions were allow more confidence in the use of the results in
selected: only one vehicle model was considered, road maintenance decision making, for instance
driven at steady speed, 85 km/h, etc. However, to include vehicle fuel cost and GHG emissions in
previous studies include many assumptions which determining maintenance strategies.
may not be valid at network level under real driv-
ing conditions. Previous studies in fact considered
only few vehicles tested at selected steady speeds ACKNOWLEDGMENTS
on about ten flat road sections. They base their
conclusions on a significantly lower amount of The authors would like to thank Alex Tam (High-
data without considering the impact of the consid- ways England) for giving permission to use data
ered variables in real driving conditions. from the HAPMS database, Mohammad Mesgar-
The correlation coefficient between the meas- pour and Ian Dickinson (Microlise Ltd) for allow-
ured fuel consumption and the predicted value cal- ing us to use an anonymized part of their truck
culated using the generated model is 0.54. This is a telematics database and Emma Benbow, David
low level of correlation compared to the more con- Peeling and Helen Viner (TRL Ltd) for their help
trolled experimental studies in the literature but is in the interpretation of results and for their sup-
the result of analysing 1970 data records for 2 mile port in this initial part of the research.
long sections over the majority of the 42.6 km long
motorway. Models derived in this way represent This project has received funding
real driving conditions (albeit at constant speed from the European Union Horizon
and gear in this initial study) for many trucks and 2020 research and innovation pro-
over a route length. For these reasons they may be gramme under the Marie Sklodowska-Curie grant
considered more representative of real world fuel agreement No. 642453 and it is part of the Train-
consumption than the limited experimental stud- ing in Reducing Uncertainty in Structural Safety
ies. This demonstrates the strength and weakness project (TRUSS Innovative Training Network,
of the Big Data approach. Although techniques www.trussitn.eu).
for reducing the noise included in the data exist
(and will be implemented soon) the lack of preci-
sion in the model estimate is an inevitable conse-
quence of using this approach. Further research REFERENCES
using this approach is underway. A better estimate
of payload is being developed using a mechanical Akaike, H. 1973. Information theory and an extension
model based on the engine performance using fur- of the maximum likelihood principle. In Second
ther sensor data. Also, by considering single trips International Symposium on Information Theory,
ed. B. N. Petrov and F. Csaki, 267–281. Budapest:
performed by individual constantly loaded trucks, Akailseoniai–Kiudo.
it might be possible to remove the effect of payload Beuving, E., De Jonghe, T., Goos, D., Lindhal, T., and
on the fuel consumption estimates. Another poten- Stawiarski, A., 2004. Environmental Impacts and
tial solution is to identify from position and time Fuel Efficiency of Road Pavements. Industry report.
data, when and where trucks discharge their loads Eurobitume & EAPA Brussels.

37

TIS2017_Book.indb 37 2/8/2017 8:00:59 PM

 Chatti, K. & Zaabar, I., 2012. Estimating the Effects SAE International 2002. Vehicle Application Layer—
of Pavement Condition on Vehicle Operating Costs, J1939-71—Surface Vehicle Recommended Practice
National Cooperative Highway Research Program, Rev. Aug. 2002.
Report 720. Washington, DC. Sandberg, Ulf S. I. 1990. Road Macro-and Megatexture
Du Plessis, H.W., Visser, A.T., and Curtayne P.C. 1990. Influence on Fuel Consumption. ASTM STP 1031:
Fuel Consumption of Vehicles as Affected by Road- 460–479.
Surface Characteristics. ASTM STP 1031: 480–496. Santero, N.J., Harvey, J. & Horvath, A., 2011. Environ-
Haider, M., Conter, M., and Glaeser, K.P. 2011. Dis- mental policy for long-life pavements. Transportation
cussion paper what are rolling resistance and other Research Part D: Transport and Environment, 16(2):
influencing parameters on energy consumption in 129–136.
road transport, Models for Rolling Resistance in Zaabar, I. & Chatti, K., 2010. Calibration of HDM-4 mod-
Road Infrastructure Asset Management Systems els for estimating the effect of pavement roughness on
(MIRIAM), AIT, Austria. fuel consumption for U.S. conditions. Transportation
House of Commons, 2011. House of Commons Com- Research Record, 2155: 105–116.
mittee of Public Accounts Departmental Business Zaniewski, J.P., 1989. Effect of Pavement Surface Type
Planning. on Fuel Consumption. Portland Cement Associa-
Laganier R. & Lucas J. 1990. The Influence of Pavement tion—Research & Development Information. Special
Evenness and Macrotexture on Fuel Consumption. report. FHWA.
ASTM STP 1031 pp. 454–459, USA 1990. SAE Inter-
national, 2002. Surface Vehicle Recommended Prac-
tice, 4970: 724–776.

38

TIS2017_Book.indb 38 2/8/2017 8:01:00 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Transparency and good governance as success factors in public private

partnerships

C. Queiroz
Claret Consulting LLC, Washington, DC, USA

P. Reddel
Brisbane, Australia

ABSTRACT: Many governments do not have all the financial resources required to meet their country’s
infrastructure needs. Facing limited resources and searching for more efficient and effective delivery of
their infrastructure, many countries have involved the private sector under long term Public-Private Part-
nership (PPP) contractual arrangements. Typically, PPPs involve a private sector company or consortium
(the ‘concessionaire’) contracting with government to finance, build, operate, and then transfer the asset
back to the public sector at the end of a concession. On the premise that good governance can lead to
more successful PPPs, this paper reviews some key requirements for good governance in such projects,
including competitive selection of the concessionaire, public disclosure of relevant information, and regu-
latory oversight of the concession contract. Because PPPs in infrastructure tend to have monopolistic
features, good governance in managing them is essential to ensure that the private sector’s involvement
yields the optimum benefit for society.

1 INTRODUCTION asset are funded through the revenue it receives for

providing a flow of services from that asset. This
Many governments do not have all the financial revenue comes either directly from the government
resources required to expand, maintain, and oper- or through the concessionaire imposing a charge/
ate their country’s transport networks and other toll/tariff on users. This differs from the conven-
forms of infrastructure. Because of such limited tional public sector model where the government
resources and a search for more investment effi- directly funds the building of a physical asset and
ciency, in a number of countries the private sec- then employs the personnel for (or contracts out)
tor has been involved in delivering infrastructure its long-term maintenance and operation.
through concessions under PPP programs. The difference in the PPP delivery model can be
While there is no widely agreed, single defini- illustrated in Figure 1 showing the varying roles for
tion or model for PPPs, the PPIAF Public-Pri- government and the private sector (the concession-
vate Partnerships Reference Guide: Version 2.0 aire, usually through a Special Purpose Vehicle or
(PPIAF 2014) provides a useful, broad description: SPV) over the planning and preparation for a road
“A long-term contract between a private party and project, its long-term operation by the private sector
a government entity, for providing a public asset or over the concession period, and the considerations
service, in which the private party bears significant for government in the post-concession period.
risk and management responsibility, and remuner-
ation is linked to performance.”
PPPs have some key features around the differ-
ent roles for the government and private sector,
the length of the contracts, how they are financed
and the risk transfers involved. They are long-term
associations between the public and private sectors
that usually involve the private sector undertaking
investment projects that traditionally have been
executed (or at least financed) and owned by the
public sector. A central feature is that the private
sector’s investment and long-term operation of the Figure 1. Government and concessionaire roles.

39

TIS2017_Book.indb 39 2/8/2017 8:01:01 PM

 For the Pre-Concession Period: finance used to build the asset takes the form of a
service or availability payment from government.
• Government funds all expenditure on construc-
These are sometimes termed social infrastruc-
tion of road and operation and maintenance
ture PPPs because this model is typically used for
• Government starts of planning and preparation
schools, hospitals, prisons and other ‘social’ (i.e.,
for possible PPP: pre-feasibility and feasibility
non-income producing) infrastructure. The second
studies
is where the primary source of funding takes the
• PPP transaction go-ahead if viable, with selec-
form of charges paid by the users of the infrastruc-
tion of successful bid
ture, such as tolls paid by the users of a toll road.
For the Concession Period: These economic infrastructure PPPs are typically
used for roads, railways and other ‘economic’ (i.e.,
• A Special Purpose Vehicle (SPV, i.e. the Conces-
income producing) infrastructure.
sionaire) is competitively selected for delivery of
The archetypical PPP is a Build-Operate-
road project
Transfer (BOT) project where the Concessionaire
• SPV responsible for financing and delivery of
finances, builds, operates and transfers the asset
road project and ongoing maintenance and
back to the public sector at the end of the conces-
operations over the concession period
sion life. The Concessionaire sells the final service
• Ownership of assets remains with government
to the public sector or to the public (i.e., the users)
• SPV responsible for keeping road in acceptable
under a government concession. Figure 2 shows
condition, meeting Key Performance Indicators
how such a road PPP concession would typically
(KPIs) specified in the Concession Agreement
be structured where tolls were imposed on road
• Revenue to the SPV would come from payments
users.
by government (through availability payments
Key features of such a PPP include:
or shadow tolls), from users (i.e. tolls), or some
hybrid from both government and users • SPV would be competitively selected for the road
• Government role through the concession period project.
to monitor performance of Concessionaire • SPV would be responsible for constructing the
against KPIs road, ongoing maintenance and operations
• Towards end of concession period, assess per- including the collection of tolls, and would
formance of Concessionaire and decide if to retain that revenue.
extend concession or plan for handback • SPV would be responsible for keeping the road
in acceptable condition, meeting the KPIs speci-
For the Post-Concession Period:
fied in the Concession Agreement.
• Following handback, government would be • The government/Road Agency could make an
responsible for funding of all expenditure on upfront payment to the SPV and, if triggered,
road, any further upgrading and operations and any payments from a Minimum Revenue Guar-
maintenance antee (MRG), if specified in the Concession
• If government decides to extend concession, Agreement.
terms of agreement would be reset: • SPV would manage the road over the concession
period while the ownership of assets remains
– Toll or availability payment
with government/Road Agency.
– Concession period
• SPV would borrow from banks, as needed. The
– KPIs
government/Road Agency could consider pro-
Rather than directly controlling the financing, viding guarantees such as an MRG, if agreed
building, operation and maintenance of a road, and needed.
government takes responsibility for its delivery and
the overall enabling environment for PPPs while
having the private sector contribute to reducing
the overall cost of delivering infrastructure serv-
ices through increased efficiency and better man-
agement of some risks (such as construction). In
successful PPP projects, the private sector’s higher
cost of financing and the need for a return on its
investments are offset by the benefits provided by
the private participation.
There are two general types of PPPs (Utz 2013).
The first is where the main revenue stream or
source of funding that repays the private sector Figure 2. Structure of a toll road concession.

40

TIS2017_Book.indb 40 2/8/2017 8:01:02 PM

 Figure 3. Road concession structure without tolling.
Figure 4. PPI investment in road projects by region:
Alternatively, a road PPP concession could be 2011–2015.
structured so that there was no tolling of road
users, as shown in Figure 3. been in Latin America and The Caribbean with
Key features of this type of road PPP conces- 44%. Investment has grown strongly over the last
sion would include: 10 years, with a trend in more recent years of fewer
but larger deals.
• An SPV would be competitively selected for the
KPMG in its June 2015 review of emerging glo-
road project.
bal trends in PPPs notes that “A particular chal-
• SPV would be responsible for upgrading the
lenge in developing markets, such as Asia and
road and ongoing maintenance and operations
the Middle East, is uncertainty around the ease
over the concession period.
of business and the transparency of governance
• SPV would be responsible for keeping the road
arrangements.”
in acceptable condition, meeting the KPIs speci-
In the past, governance has often referred to hier-
fied in the Concession Agreement.
archical, top-down systems of control by govern-
• The government/Road Agency would make
ments and institutions (Brinkerhoff & Brinkerhoff
ongoing payments to the SPV over the conces-
2011). Increasingly under the emerging environment
sion period, based on meeting the KPIs. The
for PPPs, there is also a less hierarchical model where
government/Road Agency could also make an
governance is without government; or at least with-
upfront payment to the SPV to reduce the level
out a hierarchical control-based role for government
of ongoing payments.
(Stoker 1998): “Governance is ultimately concerned
• SPV would manage the road assets transferred
with creating the conditions for ordered rule and
to it over the concession period while the own-
collective action. The outputs of governance are not
ership of assets remains with the government/
therefore different from those of government. It is
Road Agency.
rather a matter of difference in process.”
• SPV would borrow from banks, as needed. The
On the premise that good governance can lead
government/Road Agency could consider pro-
to more successful PPP projects, this paper reviews
viding guarantees, if agreed and needed.
some key requirements for good governance in
The impact of PPP projects has been mixed. As such projects. Because PPPs in infrastructure tend
stated by McCarthy (2013), at their best, PPPs can to have monopolistic features, good governance in
provide rapid injections of cash from private finan- managing them is essential to ensure that the pri-
ciers, timely and within-budget delivery of quality vate sector’s involvement yields the maximum ben-
services, and overall cost-effectiveness the public efit for the society.
sector can’t achieve on its own. But at their worst, Crucial requirements for good governance in
PPPs can also drive up costs, under-deliver services, PPPs include (Queiroz & Izaguirre 2008): (i) com-
harm the public interest, and introduce new oppor- petitively selecting the strategic private investor,
tunities for fraud, collusion, and corruption. (ii) properly disclosing relevant information to the
The World Bank’s Private Participation in public, and (iii) having a regulatory entity appropri-
Infrastructure (PPI) Database (PPI 2016) provides ately oversee the contractual agreements over the
some useful insights into the extent of recent road life of the concession. Additionally, an appropriate
investment in developing countries. Figure 4 shows legal framework helps reduce the need for public
the regional breakdown of total PPI roads invest- sector guarantees, thus facilitating the transfer of
ment of $122 billion over the last 5 years (2011– risks to the private sector, which is a key feature of
2015). The largest regional investment share has PPPs (Queiroz & Lopez-Martinez 2013).

41

TIS2017_Book.indb 41 2/8/2017 8:01:02 PM

 2 COMPETITIVE SELECTION OF process. Pre-qualification ensures that invita-
CONCESSIONAIRES tions to bid are extended only to those who have
adequate capabilities and resources. Prequalifi-
Open and transparent competitive bidding is usu- cation shall be based entirely upon the capability
ally perceived as a prerequisite to ensuring the effi- and resources of prospective bidders to perform
cient allocation and use of scarce public resources. the particular concession contract satisfactorily,
The World Bank Procurement Guidelines recom- taking into account their (i) experience and past
mend the use of International Competitive Bidding performance on similar contracts, (ii) capabilities
(ICB) to select the concessionaire or entrepreneur with respect to personnel, equipment, and con-
under BOO (Build, Own, Operate), BOT (Build, struction, and (iii) financial position. All bidders
Operate, Transfer), BOOT (Build, Own, Operate, that pass this stage are by definition qualified
Transfer) or similar types of concessions for projects and should be considered for the next phase.
such as toll roads, tunnels, harbors, bridges (World e. Inviting pre-qualified firms/consortiums to submit
Bank 2011). The Guidelines state that the ICB pro- bids. Define the procedures for the pre-qualified
cedures may include several stages to arrive at the bidders to carry out their own due diligence of
optimal combination of evaluation criteria, such as the proposed project. In addition, a Data Room
the cost and magnitude of the financing offered, prepared by the Client should be made available
the performance specifications of the facilities to potential investors, to enable them to fully
offered, the cost charged to the user, other income assess the investment opportunity.
generated by the facility, and the period of the facil- f. Bidders’ review and comments. In order to mini-
ity’s depreciation. Competition can help to reduce mize opportunities for post-bid negotiations
prices and expand access, and should be used to the on substantive issues with the winning bidder,
maximum extent possible (Harris 2003). To facili- major transaction documents (such as con-
tate bid evaluation, it is highly recommended that cession contract, shareholders’ agreement)
a simple criterion be applied to compare the price should be circulated to the bidders for review
proposals. Examples of such criterion include min- and comments before bids are submitted. The
imum toll rate to be charged to the users, minimum clear understanding to bidders should be that
availability payments to be paid by the government the period designated for review and provid-
to the concessionaire, or maximum present value ing comments will be their only opportunity to
of the payments to be made by the concessionaire influence the terms of the bidding process.
to the government during the concession life. g. Competitive Bidding Process. Once the structure
The European Union encourages the use of the of the transaction has been approved, organize
‘competitive dialogue procedure’ for major projects a competitive bidding process to award the con-
(Freshfields 2005). Such competitive dialogue is cession contract to a strategic investor. Steps in
somewhat like the World Bank ICB procedure that the bidding process include preparation of the
involves two stages (World Bank 2011). tender documents, administering the offering
Steps in the selection process are likely to include period for bidders’ due diligence, and preparing
(Queiroz & Lopez-Martinez 2013): the bid procedures and selection criteria.
h. Bid Evaluation. Evaluate the bids received
a. Advertising. A notice requesting expressions of based on the agreed, transparent selection cri-
interest to prequalify should be published in teria, and recommend award to the best evalu-
at least one international newspaper and one ated bidder. As an example, in the case of road
of national circulation and should include the projects, typically the final selection criterion is
scheduled date for availability of prequalifica- based on: (i) the minimum toll rate proposed
tion documents. by the bidders; or (ii) the minimum public con-
b. Investor Feedback. Meeting with selected poten- tribution, or subsidies, to the construction cost
tial investors/concessionaires to solicit feedback of the project, which is required by the bidders;
on the options being analyzed as well as on the or (iii) the minimum availability fee or annuity
key parameters and assumptions underpinning proposed by the bidders.
the conclusions of financial feasibility. i. Transaction Closure. The principal parties
c. Public Information. Implement an appropri- complete and execute the concession contract,
ate program to disseminate information to the shareholders’ agreement and other documents
public on the financing and construction of the necessary for the satisfactory closing of the
proposed facility (or project). transaction.
d. Prequalification of Concessionaires. Develop j. Public disclosure of the concession agreement.
operational and financial criteria to be used in By providing a further check on corruption,
judging the suitability of prospective bidders, this can enhance the legitimacy of private sec-
and conduct a transparent pre-qualification tor involvement.

42

TIS2017_Book.indb 42 2/8/2017 8:01:02 PM

 The selection process described above follows
general international good practice. However, in
cases where competition is perceived to be relatively
limited and the country environment is prone to
collusion of prospective bidders, some innovative
approach in the procurement process may lead to
better results. A good example is provided in the
selection of concessionaires under the second phase
of the Brazilian federal road concession program.
The first phase, in the 1990s, had followed a more
traditional approach, which led to relatively high
toll rates. In the second phase, around 2008, the
selection was carried out by the Sao Paulo Stock
Exchange (BOVESPA) through an auction (Amore-
Figure 5. Toll booth, Rio de Janeiro to Juiz de Fora
lli 2009). The process consisted of the following expressway, Brazil. Good governance helps assure that
main steps: the minimum possible toll rates are charged to road users.
a. Public consultation on the draft bidding Source: Queiroz and Kerali 2010.
documents.
3 UNSOLICITED PROPOSALS
b. Preparation of the final bidding documents.
c. Adopting as the selection criterion the lowest
Unsolicited proposals, which seem attractive to
toll rate (as in the first concession phase).
some governments in their wish to accelerate the
d. BOVESPA held a simultaneous auction for
implementation of infrastructure projects in the
seven road concessions, without prequalifica-
country, tend to be so controversial (usually involv-
tion, defining the bidder offering the lowest
ing allegations of corruption), that in fact they usu-
toll rate for each project road, including foreign
ally take longer to award than an open, competitive
firms.
tender procedure. In theory, unsolicited proposals
e. The bid evaluation committee reviewed compli-
could generate beneficial ideas; in practice, there
ance of the technical offer and bid guarantee
have been several unfavorable experiences, mostly
of the best-ranked candidate for each project,
as a result of exclusive negotiations behind closed
while other technical offers were not opened.
doors. There have been cases where a contract
f. Award of each concession contract to the
signed between a government and a private com-
respective lowest bidder, without negotiations.
pany included a clause that prohibits any leakage
Compared to the first phase of road conces- of the signed contract.
sions, the innovative approach used in the sec- Several countries have adopted specific leg-
ond phase showed considerable advantages, by (i) islation to deal with such proposals, while some
reducing the time between invitation to bid and governments have simply forbidden unsolicited
contract signing, and (ii) leading to substantially proposals to reduce public sector corruption and
lower toll rates, with an average of about US$0.02/ opportunistic behavior by private sector compa-
car-km, compared to about US$0.04/car-km in the nies. The general experience with unsolicited pro-
first phase. Additionally, the sequencing of bid- posals is often negative, reflecting the fact that
ding steps (financial offer first, technical second) projects of this type have usually represented poor
limited the impact of judicial recourses—a critical value for money, and were frequently incompatible
factor in traditional tendering processes in Brazil with the actual development needs of the coun-
(Véron & Cellier 2010). A toll booth showing toll tries, and their ability to pay. They also often lead
rates for a concession bid under the first phase is to allegations of corruption. Corruption has been
given in Figure 5. shown to be associated with the lack of adequate
While this innovative approach worked well for transport infrastructure in a country, as well as a
these relatively simple projects, it is likely that the low degree of economic development (Queiroz &
traditional approach, including prequalification, Visser 2001). It is essential to eliminate or minimize
will be the most appropriate for more complex the perception, as well as the reality, of corruption
projects. in PPP programs so that such programs can best
Nevertheless, the above example illustrates contribute to a country’s economic development.
that even in a country with a well-established Some governments have adopted procedures to
legal framework it is important to keep some transform unsolicited proposals for private infra-
flexibility for applying innovative solutions to structure projects into competitively tendered
specific problems (such as collusion of potential projects. Such countries include Chile, the Republic
concessionaires). of Korea, the Philippines, and South Africa (Hodges

43

TIS2017_Book.indb 43 2/8/2017 8:01:02 PM

 2003a and 2003b). How to respond to unsolicited role of the regulator. Inherent PPP risks can be
bids so as to protect transparency in the procurement reduced if the terms, costs, and benefits of the PPP
process and recognize the initiative of the proponent, project are made more understandable and accessi-
is typically difficult. The World Bank considers that ble to governments, private parties, and consumers
unsolicited proposals should be dealt with extreme (McCarthy 2013).
caution and does not permit the use of unsolicited Nevertheless, not all concession contracts are
proposals in World Bank-funded projects. open to public scrutiny. Excuses range from a
A number of approaches on managing unsolic- claimed need for confidentiality to the cost of
ited proposals are available at the World Bank PPP photocopying (The Economist 2007).
in Infrastructure Resource Center for Contracts, In one country in Central and Eastern Europe,
Laws and Regulation (World Bank 2013). Such the main text of a concession agreement was
approaches include: published but key annexes including financial
and technical obligations of the concessionaire
a. UNCITRAL Approach. UNCITRAL sets out
were not open to the public. In a Latin Ameri-
suggested legislative language in provisions
can country, the full final draft of the concession
20 to 23 of its Model Legislative Provisions
agreements is published, but the signed version
(UNCITRAL 2004). Whenever a host author-
is kept confidential. As a result, potential last
ity receives an unsolicited bid, UNCITRAL
minute negotiations conducted behind closed
recommends that the authority first consider
doors between the successful bidder (i.e., the con-
whether the proposal is potentially in the pub-
cessionaire) and the agency responsible for the
lic interest. If so, the authority then requests
project, if inserted in the contract, are not made
further information from the proponent in
available to the public or to the other contend-
order to make a full evaluation. If the authority
ers in the competitive bidding process (Queiroz
decides to go ahead with the project, it deter-
2009). Full disclosure, in every case, increases
mines whether the project involves intellectual
accountability of both the concessionaire and the
property, trade secrets or other exclusive rights
regulator.
of the proponent. For projects that do not
Based on a review and an assessment of prac-
involve these rights, a full selection procedure
tices conducted by the World Bank Institute (WBI
is followed, with the proponent being invited to
2013), some key elements of proactive disclosure
take part in the selection. If it does involve the
should be:
proponent’s intellectual property, a full selec-
tion procedure does not need to be followed. In • The disclosure of the current PPP contract (iden-
this case, the contracting authority may request tifying any changes made since the contract was
the submission of other proposals, subject to originally signed) and relevant side agreements
any incentive or other benefit that may be given including government guarantees, with minimal
to the person who submitted the unsolicited redactions which reflect commercially confiden-
proposal. tial information
b. Chilean law approach. Chile has adopted an • The disclosure of future stream of payments and
approach whereby the project proponent is government commitments under PPP contracts
required to take part in a fully competitive ten- • The publication of a summary which provides in
der process, but is given bonus points in relation plain language the most important elements of
to the evaluation (Chile 1997). the contract and project and key information on
the rationalization of the project, selection as a
In view of the risks involved with unsolicited
PPP, and procurement
proposals, it seems essential that the legal frame-
• Information on a regular basis on the perform-
work in each country include a clear provision to
ance of the project
deal with this type of offers. The absence of com-
• A process by which information is authenticated/
petitive pressure on bidders also leads to concerns
validated
of less commercial outcomes from such proposals.
Bloomgarden and Neumann (2016) stress
the benefits of open data and increased disclo-
4 DISCLOSURE OF CONCESSION sure in leading to better PPPs through improved
AGREEMENTS user feedback, improved decisions around con-
tracts and performance (particularly on contract
Full disclosure of concession agreements, an indi- renegotiations and adjustments), and increasing
cation of good governance, helps ensure that the competition from bidders having improved infor-
users know what to expect from the facility under mation about PPPs at all stages from planning
concession, thus increasing transparency in the onwards.

44

TIS2017_Book.indb 44 2/8/2017 8:01:03 PM

 5 LONG TERM MONITORING OF It seems fair to conclude that governments will-
CONCESSION AGREEMENTS ing to implement successful PPP programs should
consider properly implementing the three key fac-
More than two centuries ago Adam Smith (1776) tors discussed in the paper.
wrote that “a high road, though entirely neglected,
does not become altogether impassable. The pro-
prietors of the tolls upon a high road, therefore, REFERENCES
might neglect altogether the repair of the road, and
yet continue to levy very nearly the same tolls.” Amorelli, L. 2009. Brazilian Federal Road Concessions:
To avoid such situations, which might occur even New challenges to the regulatory framework. Institute
today, many countries have established regulatory of Brazilian Business and Management Issues, Minerva
agencies that monitor the performance of infra- Program, George Washington University, Washing-
structure facilities under concession. For example, ton, D.C., USA. http://www.gwu.edu/∼ibi/minerva/
in 2001 Brazil established the National Agency for Spring2009/Lara.pdf
Land Transport, which, inter alia, monitors federal Bloomgarden, D. & Neumann, G. 2016. Open data +
road concessions (Brazil 2001). increased disclosure = better PPPs. Handshake, Issue
17 https://handshake.pppknowledgelab.org/features/
Roads and other infrastructure concession con-
open-data-increased-disclosure-better-ppps/
tracts typically include required standards for con- Brazil. 2001. National Agency for Land Transport.
struction, operation, maintenance, and toll (or fee) Brazil. http://www.antt.gov.br
collection. For monitoring the quality of the facil- Brinkerhoff, D. & Brinkerhoff, J. 2011. Public–Private
ity during the life of the concession, several indica- Partnerships: Perspectives on Purposes, Publicness,
tors of condition are usual. In the case of roads, and Good Governance. Public Administration and
such indicators include roughness, skid resistance, Development, Volume 31, Issue 1, February 2011,
luminescence of pavement markings, and the pres- pp 2–14 http://onlinelibrary.wiley.com/doi/10.1002/
ence and condition of signs, lighting, and other pad.584/full
Chile. 1997. Chile Concession Regulation No.956. Min-
safety features.
isterio de Obras Publicas. Santiago, Chile. http://
Monitoring indicators that fall outside the siteresources.worldbank.org/INTINFANDLAW/
boundaries of acceptability may lead to penal- Resources/Chileanconcessionregulations.pdf
ties for the concessionaire. Enforcing such stand- Clayton Utz. 2013. Improving the Outcomes of Public
ards helps the government and the users to reap Private Partnerships. https://www.claytonutz.com/
maximum benefits of PPP projects. The long-term ArticleDocuments/178/Clayton-Utz-Improving-The-
monitoring of PPP contracts by government was Outcomes-Of-Public-Private-Partnerships−2013.pdf.
identified in Figure 1 above as part of the changing aspx?Embed = Y
role for government under such arrangements. Economist, The. 2007. Who Benefits from the Minerals?
Vol 384, Number 8547, September 22, page 62. Lon-
Sensible negotiation of disputes that may occur
don, UK.
during the life of a PPP project should assure the Freshfields Bruckhaus Deringer. 2005. Competitive dia-
continuation of services and prevent the collapse logue: the EU’s new procurement procedure. http://
of projects and consequent public waste (UNECE www.freshfields.com/publications/pdfs/2005/13494.
2008). pdf
Harris, C. 2003. Private Participation in Infrastructure
in Developing Countries: Trends, Impacts, and Policy
6 SUMMARY AND CONCLUSIONS Lessons. World Bank Working Paper No. 5. The World
Bank. Washington, D.C., USA. Page 25. http://rru.
worldbank.org/Documents/PapersLinks/1481.pdf
This paper presented a review of some key require-
Hodges, J. 2003a. Unsolicited Proposals—Competitive
ments for good governance in PPP projects. Because Solutions for Private Infrastructure. Public Policy for the
such infrastructure projects tend to have monopo- Private Sector, Note No 258. The World Bank. Washing-
listic features, transparency and good governance ton, D.C., USA. http://rru.worldbank.org/Documents/
in managing them is essential to ensure that uncer- PublicPolicyJournal/258Hodge-031103.pdf
tainty is reduced, encouraging the private sector’s Hodges, J. 2003b. Unsolicited Proposals—The Issues for
involvement to yield the maximum benefit for the Private Infrastructure Projects. Public Policy for the
public. Good governance in this case requires, inter Private Sector, Note No 257. The World Bank. Wash-
alia: (i) competitively selecting the strategic private ington, D.C., USA. http://rru.worldbank.org/Docu-
ments/PublicPolicyJournal/257Hodge-031103.pdf
investor, including an adequate treatment of unso-
KPMG. 2015. Public Private Partnerships: Emerging Glo-
licited proposals; (ii) properly disclosing relevant bal Trends and the implications for future infrastructure
information to the public; and (iii) having a regula- development in Australia. https://home.kpmg.com/
tory entity oversee the contractual agreements over au/en/home/insights/2015/06/public-private-partner-
the life of the concession. ships-global-trends.html

45

TIS2017_Book.indb 45 2/8/2017 8:01:03 PM

 McCarthy, L. 2013. Fixing Fraud in Public-Private hn.psu.edu/faculty/jmanis/adam-smith/Wealth-
Projects. Voices: Perspectives on development. The Nations.pdf
World Bank, Washington, D.C., USA. http://blogs. Stoker, G. 1998. Governance as theory: five propositions.
worldbank.org/voices/fixing-fraud-in-public-private- International Social Science Journal, Volume 50, Issue
projects 155, March 1998, pp 17–28 https://www.researchgate.
Public-Private Infrastructure Advisory Facility-PPIAF. net/profile/Gerry_Stoker/publication/227980445_
2014. Public-Private Partnerships Reference Guide: G ov e r n a n c e _ a s _ t h e o r y _ f i v e _ p ro p o s i t i o n s /
Version 2.0. Washington, D.C., USA. https://ppp. links/53f5bc4f0cf2888a7491d638.pdf
worldbank.org/public-private-partnership/library/ UNCITRAL. 2004. Model Legislative Provisions on
public-private-partnerships-reference-guide- Privately Financed Infrastructure Projects. United
version-20 Nations, NY, USA. http://www.uncitral.org/pdf/eng-
Queiroz, C. & Izaguirre. A.K. 2008. Worldwide Trends lish/texts/procurem/pfip/model/03–90621_Ebook.pdf
in Private Participation in Roads: Growing Activity, UNECE. 2008. Guidebook on Promoting Good Govern-
Growing Government Support. Gridlines series, no. ance in Public-Private Partnerships. United Nations
37, Public-Private Infrastructure Advisory Facility- Economic Commission for Europe, United Nations,
PPIAF, Washington, D.C. USA. https://www.ppiaf. New York and Geneva. http://www.unece.org/filead-
org/sites/ppiaf.org/files/publication/Gridlines-37- min/DAM/ceci/publications/ppp.pdf
Worldwide%20Trends%20in%20Private%20-%20 Véron, A, & Cellier, J. 2010. Private Participation in
CQueiroz%20AIzaguirre.pdf the Road Sector in Brazil: Recent Evolution and
Queiroz, C. & Kerali, H. 2010. A Review of Institutional Next Steps. World Bank Transport Paper No. TP-30.
Arrangements for Road Asset Management: Lessons Washington, D.C., USA. http://go.worldbank.org/
for the Developing World. World Bank Transport 6HDCYBMRT0
Paper No. TP-32. Washington, D.C., USA. http:// World Bank Institute. 2013. Disclosure of Project and
go.worldbank.org/6HDCYBMRT0 Contract Information in Public-Private Partnerships.
Queiroz, C. & Lopez Martinez, A. 2013. Legal Framework World Bank, Washington, D.C., USA. http://wbi.
for Successful Public-Private Partnerships. In “The worldbank.org/wbi/document/disclosure-project-and-
Routledge Companion to Public-Private Partnerships,” contract-information-public-private-partnerships
edited by de Vries, P. & Yehoue, E.B. pp 75–94. http:// World Bank. 2004. Reducing the ‘Economic Distance’
www.routledge.com/books/details/9780415781992/ to Market: A Framework for the Development of the
Queiroz, C. & Visser, A. 2001. Corruption, Transport Transport System in South East Europe. Washington,
Infrastructure Stock and Economic Development. D.C., USA. http://wbln0018.worldbank.org/ECA/
Infrastructure and Poverty Briefing for the World Transport.nsf/ECADocByLink/BEF3FC761FF49D0
Bank Infrastructure Forum, CD-ROM. World Mar- 785256FB200508860?Opendocument
kets Research Centre Ltd. The World Bank. Washing- World Bank. 2011. Guidelines: Procurement of Goods,
ton, D.C., USA. http://www2.udec.cl/∼provial/expo/ Works, and Non-Consulting Services under IBRD
WB%20Infrast%20Forum%20May%202001%20 Loans and IDA Credits & Grants. The World Bank.
TranspInfra&Corrup%20CQ%20AV.pdf Washington, D.C., USA. http://go.worldbank.org/
Queiroz, C. 2009. Financing of Road Infrastructure. 1KKD1KNT40
Proceedings of the 5th Symposium on Strait Crossings. World Bank. 2013. PPP in Infrastructure Resource
June 21–24, 2009. ISBN 978-82-92506-69-1. pp 45–57. Center for Contracts, Laws and Regulation (PPPIRC).
Trondheim, Norway. http://www.straitcrossings.com Washington, D.C., USA. http://www.worldbank.org/
Smith, A. 1776. An Inquiry into the Nature and Causes of pppiresource
the Wealth of Nations. Pennsylvania State University, World Bank. 2016. “Private Participation in Infrastruc-
Electronic Classics Series, Jim Manis, Faculty Edi- ture Database” Washington, D.C., USA. http://ppi.
tor, Hazleton, PA USA, 2005, page 593. http://www2. worldbank.org/

46

TIS2017_Book.indb 46 2/8/2017 8:01:03 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

SUP&R ITN: An international training network on sustainable

pavements and railways

D. Lo Presti & G. Airey

University of Nottingham, Nottingham, UK

M.C. Rubio
Universidad de Granada, Granada, Spain

P. Marsac
IFSTTAR, Nantes, France

ABSTRACT: The Sustainable Pavements & Railways Initial Training Network (www.superitn.eu) is
a training-through-research programme that is empowering Europe by forming a new generation of
multi-disciplinary professionals capable of conceiving, planning and executing sustainable road and
railway infrastructures. The SUP&R ITN, started at the beginning of October 2013, it’s a 4 million €
effort entirely funded by the Euroepan Commission through its Marie Curie Actions 2012 of the FP7
programme 2006–2013. The University of Nottingham is leading this effort which is the first of its kind
and involves 29 partners between universities, research centres and companies/industries, from five EU
countries (UK, Italy, France, Ireland and Spain) and USA. This paper aims to be a reference for those
who plan to structure similar programmes and will present the overall idea, the research framework, the
training platform and a summary of the main scientific results obtained so far.

1 INTRODUCTION To setup a multidisciplinary and multi-sectorial

network in order to form a new generation of engi-
The design, construction, maintenance, use and end- neers versed in sustainable technologies and to provide,
of-life management of road pavements and railways to both academia and industry, design procedures and
is associated with a number of important impacts sustainability assessment methodologies to certify the
on the environment; namely the consequences of sustainability of the studied technologies to the ben-
energy consumption, unsustainable use of materials/ efit of the European community.
resources, waste generation and release of hazard- To overcome these barriers, the Nottingham
ous substances into the environment. Whilst envi- Transportation Engineering Centre (NTEC)
ronmental sustainability legislation requires road shaped the programme and it’s in charge of coordi-
and rail infrastructure to deliver more sustainable nating a consortium of 29 partners between univer-
solutions, progress to date has been limited. This is sities, research centres and companies/industries,
the result of a lack of well-founded scientific tools from five EU countries (UK, Italy, France, Ireland
to measure sustainable solutions, the lack of suit- and Spain) and USA. The SUP&R ITN started at
ably trained personnel to design and use such solu- the beginning of October 2013 and will empower
tions, and the fragmented approach generally taken Europe by forming a new generation of multi-
by researchers and industrial stakeholders across disciplinary researchers and professionals capable
Europe. Industry and the private sector are well of conceiving, planning and executing sustainable
aware of this problem and have been highly support- road and railway infrastructures, and deliver long-
ive of a coherent research and training approach term benefits in terms of:
that can address and rectify this situation.
SUP&R (Sustainable Pavement & Railways) 1. Eco-designed road and rail infrastructure that
ITN, through a coherent research and training maximizes the recycling of waste materials and
approach involving close collaboration between ensures best performance characteristics to suit
research institutions and industrial stakeholders the diverse set of European environments;
across Europe, will allow this step change in the 2. Reduced installation, maintenance and oper-
sustainability of road and rail infrastructure to be ating costs as well as long term sustainable
addressed by targeting the following overall aim: solutions;

47

TIS2017_Book.indb 47 2/8/2017 8:01:05 PM

 3. A bespoke sustainability assessment tool, tai- Table 2. SUP&R ITN associate partners.
lored to needs of product development in the
road pavement and rail infrastructure sector. Eiffage ETP Ecopneus ECOPN
Travaux
This paper will present the overall programme, Public
the planned training and dissemination activities, Network NR Euro Rail EIM
together with the details of each of the main sci- Rail Infrastructure
entific Work Packages as well as a summary of the Managers
main results. More details in the website: www. Nynas NYNAS Forum of FEHRL
superitn.eu European
National
Highway
Research
2 SUP&R ITN PROJECT Laboratories
Universidad UPM Ceequal CEEQUAL
2.1 Project consortium and formation strategy politécnica
de Madrid
The SUP&R International Training Network
Slovenian ZAG GreenRoads GR
(ITN) will offer a training-through-research pro- National
gramme to selected young researchers across Building
Europe. The project includes a high level four year and Civil
training programme which will be international, Engineering
multi-disciplinary and multi-sectors. Institute
The network of SUP&R ITN is comprised Aggregate AI University of ILL
of partners (Tables 1 and 2) who, in addition to Industries Illinois
having outstanding expertise and state-of the-art Urbana-
facilities, complement each other to bring together Champaign
the necessary scientific and professional power Phoenix PI University of UW
Industries Washington
needed to provide high quality training and cut-
Administrador ADIF Virginia Tech VTTI
ting edge research in designing sustainable trans- de Infrae- Transporta-
port infrastructures. structuras tion Institute
The industry partners cover the whole stake- Ferroviarias
holders’ chain of road pavements and railways Euro Asphalt EAPA
from transport infrastructure managers (IRAIL, Pavement
ADIF, NR), engineering designers (URS), con- Association
tractors (SACYR, ETP) and manufacturers
(REPSOL, NYNAS, AI, PI). This consortium is
complemented by partners ranging from research New partners have been introduced to add new
and training bodies (TUD, UW, VTTI), Profes- complementary skills and expertise to the existing
sional European Associations (CEDR, EAPA, structure. Instead, the choice of the private sec-
ECOPN, EIM), experts in sustainability (GR, tor full partners has been strategically planned in
CEEQUAL), online training of sustainability order to cover the whole stakeholder chain of road
(ILL) and communication in transport engineer- pavements and railways within transport infra-
ing field (FEHRL). The choice of collaborators structure. Table 3 summaries the exploitation of
from academic sector was driven by excellence synergies and complementary expertise within the
and based on a selection of specific expertise of whole SUP&R ITN consortium.
each partners and successful past collaborations.

Table 1. SUP&R ITN full partners.

2.2 Project structure
To achieve the overall research and training objec-
University of Nottingham NTEC— UNOTT tives, the SUP&R ITN has been implemented with
(Coordinator) 6 Work Packages (WPs). Training-through-research
Universita’ degli Studi di Palermo DICAM UNIPA
projects will be delivered within three scientific
University College Dublin CSEE UCD
research work packages (WP1-WP3), while three
IFSTTAR IFSTTAR
additional WPs will handle respectively; Network-
Universidad de Granada LABIC UGR
wide Training (WP4), Management (WP5) and the
Universidad de Huelva CFEL UHU
external and internal Communication (WP6).
AECOM—URS AECOM
IrishRail IRAIL
The Coordination at The University of Notting-
Sacyr Vallehermoso SACYR
ham is in charge of the management of the SUP&R
REPSOL REPSOL ITN (WP5) as well as shaping the international, mul-
ti-disciplinary and multi-sectoral training platform

48

TIS2017_Book.indb 48 2/8/2017 8:01:05 PM

 Table 3. SUP&R ITN expertise matrix.

Figure 1. SUP&R ITN project structure.

which consists in: a local and network-wide training

activities (WP4) and scientific dissemination and
public outreach activities of each fellow and the con-
sortium. (WP6). The next sections will provide more Figure 2. SUP&R ITN research framework.
details on the contents and quality of the research
framework and the training platform, as well as a
and railway transport infrastructure (standard
summary of the main results obtained so far.
railway infrastructure or high speed railways).
• Providing advanced characterisation of recycled
and reused materials generated from road and
3 RESEARCH FRAMEWORK
railway infrastructure and/or other production
processes for use in road and railwaya.
SUP&R ITN will drive the researchers on their
• Developing detailed material modelling and
multidisciplinary training-through-research path-
design approaches for immediate uptake by
way by closely investigating promising pavements
industrial stakeholders working in the area of
and railway technologies with practical methods,
transport infrastructure.
models and tools for estimating their sustainabil-
• Developing and refining estimation tools so that
ity. This will be carried out through a systematic
industry can assess where it currently stands in
integration of sustainability aspects at a very early
key sustainability indicators and determine how
stage in the technology design. The SUP&R ITN
much it can improve using low-energy and more
programme will achieve these results by developing
recycled materials. Such estimation tools will
inter-linked research projects aiming at:
allow industry, road and railway agencies to rec-
• Developing new sustainable technologies and ognize the sustainability of different technologies
materials for use in road pavements (highways) and methodologies already at the design stage.

49

TIS2017_Book.indb 49 2/8/2017 8:01:05 PM

 3.1 Characterization and design of transport • Project ESR10: Settlement monitoring and pre-
infrastructures technologies (WP1 and WP2) diction in railway tracks
• Project ESR11: Optimisation techniques for geo-
WP1 and WP2 will be carried out predominantly by
physical assessment of rail support structures
the ESRs through the development of 12 research
• Project ESR12: The use of waste materials in
projects with the common goal of closely investigat-
railways
ing promising sustainable technologies, respectively
for road pavements and railways. The selected road
technologies will be investigated for specific applica- 3.2 SUP&R tool: Development of the sustainability
tion in highways, while the railway technologies will assessment methodology (WP3)
be designed to be applied either on standard rail-
way infrastructure or high-speed tracks. Perform- In WP3, results arising from both previous phases
ance prediction will be validated through simulative will allow researchers and partners involved in
tests in laboratories and, for some research projects, the development of the projects, to understand
through possible validation results from field trials. the possibility of widespread use of the proposed
Thanks to the joint contribution of the universities sustainable technologies. Once the methodology is
and industrial partners involved in each project, developed and tested, the assessment system will
the ESRs will characterise and design the pro- be incorporated in a methodology, to be included
posed technologies and at the same time will also in software tool, for a wider group of users within
contribute to building a Lifecycle Inventory to be the industry. This will assist product develop-
used in WP3. This data, indeed, will represent the ers in industry to assess and compare alternative
most critical step to accurate building the LifeCycle products and approaches. While similar tools are
Inventories and for the validation of the sustainabil- being developed for the operational phase, this
ity assessment methodology developed in WP3. tool will be focused on the product development
Sustainable Pavements (WP1). This series of stage. Unlike existing infrastructure sustainability
projects will focus on investigating several sustaina- assessment tools, it will be limited to the road and
ble technologies for road pavements. The academic rail product sector and this will allow it to be tai-
leader is IFSTTAR (P. Marsac) with the Indus- lored specifically to the assessment needs of these
trial deputy leader being REPSOL (E. Moreno). products, and include well defined methodologies
WP1 includes the following projects that will be all which will assist in quantification and comparison
carried out within the timeframe 2014–2017: of results. Its use will assist in early assessment of
emerging products and practices and in this way
• Project ESR1: Pavement design for cold in-situ help in prioritizing the most promising technolo-
recycled materials gies at an early stage of development. WP3 has
• Project ESR2: Design and characterization of UNOTT (D. Lo Presti & T. Parry) as the academic
bituminous mixes manufactured with biomass leader with IFSTTAR (A. Julien) as the deputy
• Project ESR3: Long term performance of low- leader and will include the following tasks:
temperature asphalt mixes containing reclaimed
asphalt • Project ER1: Definition of sustainability assess-
• Project ESR4: Rubberised binder and asphalt ment factors and current state-of-the-art in
mixes for wearing course sustainable practices—2014–2016
• Project ESR5: Binder design for low-temperature • Project ER2: Sustainability assessment of Railways
asphalt mixes and SUP&R ITN technologies—2016–2017
• Project ESR6: Half Warm Mixes Asphalt Recy- • Project ER3: Life cycle impact assessment of
cling Asphalt in urban roads Road and Railway Practices—methodology and
tools—2016–2017
Sustainable Railways (WP2). This series of
projects will focus on investigating several sustain-
able technologies for railway infrastructure with
4 TRAINING PLATFORM
academic leadership by UGR (MC. Rubio) and
Industrial leadership by IRAIL (C. Bowe). WP2
The overarching aim of SUP&R ITN is to pro-
will include the following projects that will be all
vide high quality interdisciplinary, inter-sectors
carried out within the timeframe 2014–2017:
and international training to young researchers,
• Project ESR7: Optimization of trackbed design thereby contributing to the development of a new
and maintenance generation of multidisciplinary engineers, fully able
• Project ESR8: Characterisation of rubberised to work within the field of sustainable road and
asphalt for railways sub-ballasts rail technologies. These objectives will be achieved
• Project ESR9: Modelling and Design of rubber- through the following a structured programme of
ised asphalt for railways sub-ballast training and dissemination activities.

50

TIS2017_Book.indb 50 2/8/2017 8:01:08 PM

 4.1 Training activities and outreach activities are organised every year.
This training offers the opportunity to fellow to
Supervision arrangements and Personal Career Devel-
team up and to provide them with a great expo-
opment Plans: Once the fellows have officially started
sure to the industrial partners. A specific training
the outcome of their initial evaluation and their
programme is shaped for “complentary skills”
career options and own ideas will lead to the design
and focuses on the skills that fellows will need to
of a Personal Training and Career Development
develop at different stages in the development of
Plan (PTCDP) detailing the knowledge gaps and the
the network and their PhDs and will involve the
need for both local and network-wide training. The
consultancy and industrial partners raising an
team leaders then guide the fellows locally, but they
awareness of the demands of organizations outside
also receive a network-wide supervision directly from
academia for the fellows. Furthermore, to comple-
the Training Responsible at the coordination. How-
ment the expertise of the consortium world-side
ever, The overall responsibility for the planning and
experts are invited as visiting lecturer is some of
implementation of the training programme lies with
the training weeks; amongst those: Dr. A. Scar-
the network through the Supervisory Board (SB).
pas of Delft University; Prof. Steve Muench, from
SB is composed by the representative of the partners
UW and GR; Prof. G. Flinch, VTTI, USA and
and shall make sure that there is an adequate bal-
Dr. J. Tomkin, ILL, USA.
ance in the training activities through personalised
Network-wide training structured activities are
research and complementary training such that use-
coordinated by the University of Nottingham with
ful skills for career development and contribution to
the collaborations of the various local partner.
society are achieved at the end of the project.
In fact, the structured training takes place at the
Local training activities: The SUP&R ITN fel-
different institutions of the consortium. Here the
lows will conduct their research primarily at their
calendar and details of the on-going training pro-
host institutions where the local training features a
gramme which is constantly updated and generally
number of important elements, including:
opened also to external participants:
• Each ESR/ER has individual scientific and
• 2014 Sep—Nottingham (UK)—“Scientific Intro-
on-the-job training provided by academic and
duction on Sustainable roads and railways”
industry partners and tailored to his/her specific
• 2015 Mar—Dublin (IE)—“Railway engineering”
research project.
• 2015 Jan and Mar—Nottingham (UK)—
• The ESR/ER are encouraged to interact with
“Outreach and Complementary Skills”
the local research group at their host institu-
• 2015 Jun—Delft (NL)—Advanced Scientific
tion. This will include research forums, national
training with Dr. Scarpas of TU Delft
symposiums and meetings. This approach will
• 2015 Sep—Palermo (IT)—“Sustainability Assess-
ensure that the ESR/ER achieves a true perspec-
ment of Transport Infrastrctures”
tive of the research being conducted at the local
• 2016 Feb—Granada (ES)—“From research to
level. Similarly, perspectives acquired during
Sustainable Pavement and Railway to imple-
this secondment can be brought back to the host
mentation in Practical engineering Projects”
institution and shared with fellow researchers.
• 2016 Oct—Madrid/Huelva (ES)—“Bituminous
• Most of the partners are located at universities
Products for Sustainable Asphalt Technologies”
with strong post-graduate programmes with
• 2017 Jan/Feb—Nantes (FR)—Still to be planned
modules on various topics related to managing
transport infrastructure. The ESRs have access
to these modules, which include scientific lec- 4.2 Dissemination and outreach activities
tures, but also generic research methods, project SUP&R ITN involves a series of outreach
management, etc. activities for public audiences around Europe.
• At the very initial stage of the programme, if These activities, are carried out by the SUP&R ITN
needed all the fellows are provided language fellows and supported by coordination at UNOTT.
courses from the host institutions. Outreach programme: coordinates the commu-
nication between the scientific community and the
Network-wide training activities: This training
general public to convey the importance of and
will be offered network-wide and will focus on the
increase the awareness of sustainability in road
skills that fellows will need to develop during and
and rail infrastructure. Here is a list of the out-
after the development of the their PhDs. The train-
reach activities (updated regularly):
ing programme started with a kick-off meeting to
introduce the fellows to SUP&R ITN, to build up • Outreach Training: To prepare for outreach
team spirit and to deliver the basic scientific train- activities, all fellows attended the course Public
ing. Since then an average of two network-wide Understanding of Science—Communication
events between workshops, spring/summer schools Skills for Researchers

51

TIS2017_Book.indb 51 2/8/2017 8:01:08 PM

 • Workshop Day: The fellows showed their project study is needed for fatigue criteria as failure mode
directly at schools to kids with an age of 15–17 for the cold recycled material is distinct from usual
• SUP&R ITN Open days: During training weeks hot materials.
for the fellows, the public and other entities (like The objective of ESR2 is to promote the use of
industry) are invited to an open day. These have binders obtained from renewable resources (bio-
been held already in Italy and Spain binders) through undertaking their mechanical
• Multimedia Releases: SUP&R ITN fellows have characterization and design optimization for mix-
produced videos to introduce their research and ture with Reclaimed Asphalt (RA). Bio-binders
created a Youtube channel to distribute them. are characterized and their interactions with RA
• SUP&R ITN movie. The project will end with a binder and their rejuvenating effect are studied
final event will be an occasion to share the achieve- and compared to standard and polymer modified
ments of the entire consortium to the scientific binders for 50% RA mixtures. Rheological char-
community and to the general public. For that acterization reveals that all the bio-binders studied
event, the consortium will premier al illustrated produced a rejuvenating effect. Fatigue (G*sinϕ)
film that will present the sustainability aspects of and rutting (G*/sinϕ) parameters suggests that
the studied technologies for a wide audience. This bio-binders can enhance fatigue resistance without
movie will summaries the learned principles of compromising rutting behavior.
sustainable pavements and railways and will rep- The evaluation of the long term performance
resent an educational resource to everyone from of low-temperature asphalt containing reclaimed
teachers to business to community leaders. asphalt is the objective of ESR3. An experimental
plan combining different recycling rate and ageing
Scientific Dissemination: It is also important that step on a reference hot mix asphalt and warm mixes
SUP&R ITN fellows disseminate their research to (additives process and foaming process) was com-
the scientific community. This is done as follows: pleted. Binders’ ageing was characterized through
• ESR specific dissemination: the coordination infrared spectroscopy. For each combination the
provides support with advices on interesting sci- rheological properties of the binder were evaluated
entific conferences and symposiums as well as and the modulus and fatigue resistance were meas-
other industry-related meetings, so that each fel- ured on the mixes. Compared to the hot process
low could have the chance to participate. the warm processes induce a lower ageing but this
• Annual dissemination event: the programme has tend to be compensated by a stronger long-term
an annual dissemination event which is a group ageing. For all processes, modulus and fatigue per-
talk where the whole project is presented by the formances tend to improve with RA addition and
fellows in several venues where they can increase ageing. Furthermore, a new method to estimate
their exposure. The final outreach event will be the apparent molecular weight distribution of a
the SUP&R ITN Symposium in 2017. binder was validated to relate mixes and binders
properties (Perez-Martinez et al. 2015).
The main objective of the ESR4 research is to
investigate the design of a hybrid system consist-
5 SUMMARY OF RESULTS UP-TO DATE
ing of crumb rubber and polymers to form a new
binder. Extensive laboratory tests were carried out
5.1 WP1 sustainable pavements
with the same crumb rubber content but for different
The WP1 is composed by 6 complementary indi- bituminous base, polymer type and polymer con-
vidual projects attempting at improving sustain- centration in order to select the blending according
ability of road pavements by means of actions that to storage stability parameters. The physico-chem-
can be grouped in 4 broad categories: materials, ical, thermo-mechanical and rheological properties
manufacturing process, structure design and dura- of these selected blending were characterized. The
bility improvement. The aim of ESR1 is to inte- selected hybrid system will be then used to manu-
grate materials cold recycled with foamed asphalt facture three type of hot asphalt mixes for wearing
and cement in pavement design methods. As few courses and results will be compared with at least
empirical data are available for this type of mate- one control for each type of mix.
rial, a mechanistic approach with a layered elastic The task of ESR5 is to design new binders for
model was chosen to relate stress, strain and deflec- low-temperature asphalt mixes incorporating high
tion field within the pavement to load conditions. amount of RA (Yuliestyan et al. 2015). As the ratio
The following relevant input material parameters binder to aggregate need to stay almost constant in
were identified: stiffness, fatigue and permanent an asphalt mixture, high recycling rate implies low
deformation. Adapted tests and procedures were amount of new binder added. To insure a decent
evaluated and notably an Indirect Tensile Fatigue coating, a new emulsion process (bitumen in the
Test associated with a dissipated energy approach. form of micro droplets) was selected. A Kraft
The experimental data obtained show that further lignin bio-polymer (an abundant by-product of the

52

TIS2017_Book.indb 52 2/8/2017 8:01:08 PM

 paper pulp industry) was modified by amination in three WP2 projects aimed to optimise the bitumi-
order to obtain a suitable emulsifier to be used with nous subballast design by introducing sustainable
tailored soft bitumen. Taking advantage of these techniques which include the use of waste such
innovative constituents, a pioneering emulsion was as crumb rubber (wet and dry processes), and the
designed and successfully used to produce 100% reduction of the asphalt mixes manufacturing
cold recycled asphalt mixes. The firsts mechanical temperature (WMA). These have being carried
tests conducted on samples of the produced mixes out through the projects: “ESR8 Characterisation
show very promising results. of rubberised asphalt for railways sub-ballasts”,
Half-warm mix can be considered as an interme- “ESR9 Modelling and Design of rubberised asphalt
diate step between cold and warm mix potentially for railways sub-ballast”, “ESR12 The use of waste
associating the reduced manufacturing energy of the materials in railways”.
first and the mechanical properties of the second. To In this sense a WMA was designed to be used
rule on uncertainties that tend to prevent their wide- as bituminous subballast, and tested under dif-
spread use, ESR6 investigate their actual perform- ferent railway loads and temperature conditions.
ances in laboratory and in the field when they are Results have proved an adequate performance to
combined with high RA content. Mixtures with 70% be used in the track. In the same way, two types
RA and 100% RA and emulsion were compared of bituminous subballast (wet and dry processes)
to a reference hot mix. Results show discrepancies containing crumb rubber have been designed and
between tests on laboratory and in the field, notably are being tested. For this porpoise the adaptation
as regard to the stiffness modulus, probably linked to of Superpave asphalt mix design was needed, once
different laboratory compaction methods (gyratory Superpave considers traffic loads while railway
and Marshal). The three mixes tested showed equiva- loads were the ones to consider in this case. Also,
lent in field performances. Research is carried out to a model bases on the fractional calculus is being
further evaluate the long-term performances in field. developed to predict the mechanical behaviour of
the crumb rubber asphalt subballast.
On the other hand, WP2 includes projects focused
5.2 WP2 sustainable railways
on improving prediction techniques to obtain more
Sustainable Railway Projects aimed to optimize accurate information about the deterioration of the
materials, design and manufacturing techniques to track. The project “ESR10 Settlement monitoring
improve problems associated with railway tracks and prediction in railway tracks” develops and tests
such as settlement, bearing capacity reduction, methods whereby sensors in a train can be used to
vibrations and noise among others; from a sustain- detect tract settlement and bridge damage; numeri-
able approach. For this purpose innovative meth- cal models of the dynamic interaction of a train
ods and tests have been carried out, including field with track, sleepers, ballast and bridge are used to
measurements in real conditions. analyse vehicle response to detect damages. This
Four of the six projects that compose WP2 deal is the first time optimisation techniques have been
with new materials to be used in the granular layers. used to process vehicle measurements to identify
Some of these materials are: bitumen emulsions, low track stiffness, track profile and bridge damage.
temperature bitumen sub-ballast, and crumb rubber Finally, the project “ESR11 Optimisation tech-
to design bitumen sub-ballast. As part of the project: niques for geophysical assessment of rail support
“ESR7: Optimization of trackbed design and mainte- structures” seeks to employ non-destructive, vibra-
nance” the study of bitumen emulsion to be used as tion based monitoring of railway substructure. One
a ballast stabilisation method for newly constructed of the challenges consists in detecting weak spots
or exiting track has been carried out (D’Angelo et al. in the substructure in transversal and longitudinal
2016). For this purpose, different emulsions were direction of railway track. By now, numerical stud-
tried and tests at different scales reproducing rail ies and field work have been carried out. Results
loads where performed. Results obtained from both obtained shows that impact hammer testing could
half and full-scale box tests highlighted the potential be used to assess the substructure condition.
for this solution to reduce ballast problems associ-
ated with its unbound nature such as settlement and
5.3 WP3 SUP&R sustainability assessment
particle deterioration, limiting therefore the need
methodology and tool
for maintenance and increasing trackbed durability.
Results have shown that the application of bitumen SUP&R ITN sustainability assessment tool is
emulsion during track initial construction can lead being built by three complementary projects: ER1
to a clear decrease in ballast permanent deformation project began with a state-of-the-art review of sus-
(–25%) and deformation rate (–30%), indicating its tainability assessment and rating tools. This review
effectiveness for both short and long-term behavior. identified key criteria and subject areas that are
Considering that bituminous subballast has found in sustainability assessment tools for civil
shown to be an alternative to granular subballast, engineering infrastructure and critically reviewed

53

TIS2017_Book.indb 53 2/8/2017 8:01:08 PM

 • Fellows are clearly experiencing something dif-
ferent from conventional PhD students. They
conceive research already as a collaborative,
multi-disciplinary and international effort.
• Secondments are a key part of this type of train-
ing and allow strengthening relation between
research institutions as well as exploiting collab-
orations between academic and industrial part-
ners. Furthermore, they provide fellows with
great exposure to several sectors and in some
Figure 3. SUP&R Multi-Criteria Decision Analysis cases their research products are already being
tool.
up taken from industrial partners.
the information across the tools. The results helped • Network-wide activities are another fundamen-
to identify limitations and areas of contributions tal part that allow personal and professional
for the project, and in combination with a review development of all the participants as well as
of the literature, they supported the development fostering collegiality amongst fellows, advisors
of the conceptual framework for the SUP&R and research groups as a whole.
ITN tool. The state-of-the-art review was also the • SUP&R ITN is highlighting the importance of
first step in developing a set of indicators for the the introduction of lifecycle approach and sus-
SUP&R ITN tool (Brice et al 2016). By focusing tainability concepts into transport infrastruc-
specifically on pavement-related indicators used ture engineering. This is significantly changing
in existing tools, we have developed a shortlist of the research and engineering practices of the
indicators for the SUP&R ITN tool. It is important partners involved. As example, it has inspired a
to note that there are no assessment tools specific new master course on Sustainable Highways at
to railway pavement or construction. As a result, University of Nottingham and it has also cre-
Project 3.2 (ER2) is developing a comprehensive ated a solid group whom core as already ensured
library of best practices in sustainable construc- the funding for another training programme of
tion for railways and this will assist the identifica- this type: SMARTI ETN 2017–2021
tion of indicators specific to railways. Both railway
and roadway indicators are organized into higher
level domains, which the conceptual framework ACKNOWLEDGEMENTS
terms “means objectives,” or objectives critical for
comprehensive sustainability. The means objectives This project has received funding from the Euro-
will be weighted in the tool and indicators classified pean Union’s Seventh Framework Programme for
under them will receive that weighting. This is being research, technological development and demon-
developed within project 3.3 (ER3) that is also pro- stration under grant agreement n. 607524.
gramming and coding the tool to calculate weight-
ings using two processes, the Analytical Hierarchy REFERENCES
Process and the PROMETHEE method. The tool
will also have a default weighting that will be estab- Bryce J., Brodie S., Parry T., Lo Presti D. A Systematic
lished based on a survey of industry professionals. Assessment of Road Pavement Sustainability through
Finally, through Multi-Criteria Decision Analy- a Review of Rating Tools, Resources, Conservation
sis (MCDA), the SUP&R ITN WP3 will produce and Recycling, Elsevier, (accepted 2016).
a tool that will be freely available and will support Colinas N., Di Paola M., Pinnola F.P., Fractional charac-
teristic times and dissipated energy in fractional linear
decision making for sustainable pavements at least
viscoelasticity, Communications in Nonlinear Sci-
at the design stage. At last, once the tool is defined, ence and Numerical Simulation, Vol. 37, 2016. DOI:
WP3 will then use it to perform the sustainabil- 10.1016/j.cnsns.2016.01.003
ity assessment of the investigated technologies in D’Angelo, G., N. Thom, D. Lo Presti, Bitumen stabilised
WP1 and WP2. ballast: A potential solution for railway track-bed, Con-
struction and Building Materials 124(2016):118–126.
July 2016. doi: 10.1016/j.conbuildmat.2016.07.067
6 CONCLUSIONS Perez-Martinez, M., P. Marsac, T. Gabet, M. Lopes, S.
Pouget, F. Hammoum (2015). Durability Analysis Of
SUP&R ITN programme is the first international Different Warm Mix Asphalt Containing Reclaimed
training programme at PhD level on Sustainable Asphalt Pavement. Proceedings of the XXVth Road
Transport Infrastructures and one of the firsts ITNs World Congress, Seoul.
dedicated to transport infrastructure engineering. Yuliestyan, A., Cuadri, A. A., García-Morales, M., &
Partal, P. (2016). Influence of polymer melting point
The programme is entering its last year and as a
and Melt Flow Index on the performance of ethylene-
result of this experience the following points play in vinyl-acetate modified bitumen for reduced-tempera-
favour of this kind of educational programme: ture application. Materials & Design, 96, 180–188.
54

TIS2017_Book.indb 54 2/8/2017 8:01:08 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Climate resilient slope stabilization for transport infrastructures

A. Faiz
Consultant, World Bank, Washington DC, USA

B.H. Shah
Consultant, National Highway Authority and Punjab Forest Department, Islamabad, Pakistan

A. Faiz
Analyst, Faiz and Associates, LLC, Arlington VA, USA

ABSTRACT: In regions subject to climate stress from increased storm intensity and frequency, soil
bioengineering coupled with appropriate water management techniques can help prevent slope failures in
transport infrastructure facilities. Soil bioengineering uses locally available plant and vegetative materials
to deter slope failure and treat slope instability. In soil bioengineering systems, grasses and plants, espe-
cially deep-rooted species, are an important structural component to reduce the risk of soil erosion and
to stabilize slopes. Where technically feasible, use of soil bioengineering alternatives produces equal or
better economic and environmental results than the sole application of traditional geotechnical solutions.
The paper includes a case study to show soil bioengineering and biotechnical applications in remediating
failed slopes on an expressway in Pakistan. The underlying bioengineering principles and techniques for
erosion control and slope stabilization are equally applicable to railways, airports, ports and other physical
infrastructure.

1 INTRODUCTION improving water infiltration rates, soil porosity,

and permeability (Fay et al. 2012).
Soil bioengineering combined with appropriate Different types of vegetation serve different
water management can be the key to preventing functions. Grasses and other herbaceous cover
slope failures in transport infrastructure facili- protect sloped surfaces from rain and wind erosion.
ties, especially in regions subject to climate stress Shrubs, trees, and other vegetation with deeper
from increased storm intensity and frequency. Soil roots are effective at preventing shallow soil failures,
bioengineering is a technique that uses plants and as they provide mechanical reinforcement with the
vegetative material alone, whereas biotechnical roots and stems and modify the slope hydrology by
techniques use plants in conjunction with more root uptake and foliage interception (Schor & Gray
traditional engineering measures and structures 2007). The main function of the structural elements
to stabilize slopes and control erosion of stream is to help vegetation to become established and
banks. Moreover, soil bioengineering and biotech- take over the role of slope stabilization; these inert
nical techniques contribute to sustainable devel- structural components eventually deteriorate or
opment practices by reducing the environmental disintegrate. Appropriately designed and installed
impacts of highway construction, operations and vegetative systems can be self-repairing, with only
maintenance (Shah 2008, Fay et al. 2012) Vegeta- minor maintenance to maintain healthy and vigor-
tive or inert structures can be used alone or jointly ous vegetation (USDA 1992).
to stabilize engineered or natural slopes and reduce As with all engineering works, it is most impor-
the risk of landslides. Stabilization measures uti- tant that the techniques selected are correct for the
lizing soil bioengineering are designed to aid or site to be treated, and that the work is carried out
enhance the reestablishment of vegetation. with all due care and attention (Howell 1999a).
Planting trees and vegetative growth alone can In general, combined slope protection systems
play an important role in stabilizing slopes by inter- have proven to be more cost effective than the use
cepting and absorbing water, retaining soil below of vegetative treatments or structural solutions
ground with roots and above ground with stems, alone. The overall cost of the soil bioengineering
controlling runoff by providing a break in the path or biotechnical components in slope stabilization
of the water and increasing surface roughness, and projects is often little more than one percent of the

55

TIS2017_Book.indb 55 2/8/2017 8:01:10 PM

 total project budget (Fay et al. 2012). Where tech- faction; and (iv) eroded stony surfaces, stripped of
nically feasible, use of soil bioengineering alterna- vegetation and trees, resulting from uncontrolled
tives tends to produce more sustainable outcomes logging, wildfires, animal grazing, foraging for
than the sole application of traditional geotechni- fuelwood, solid waste disposal and construction
cal solutions (Faiz et al. 2015). activities (Shah 2008).

2 CLIMATE RESILIENT SLOPE 2.3 Ensuring climate-resilient slopes

STABILIZATION
A basic premise for obtaining stable and climate
resilient slopes in transport infrastructure is to
2.1 Slope stabilization
address slope protection and drainage (both sur-
Slope stabilization is the science and art of manag- face runoff and groundwater) issues during loca-
ing slopes (both natural and manmade) that pose tion, design and construction—and not as an
a hazard to the built environment including roads, afterthought. Slope instability has its genesis for
railways and other physical infrastructure. the most part in shoddy engineering or cost cut-
Slope instability is generally a mass-wasting ting design and construction shortcuts. Climate
process, while erosion is typically a surficial process. resilience is enhanced by using a combination of
Stabilization solutions address both processes and drainage, slope protection and stabilization meas-
can range from allowing grass to reestablish on a ures, including both inert (engineered) and vegeta-
disturbed slope to building an engineered retaining tive components. Appropriate technologies include
structure. The treatment depends on the affected retaining structures and earthworks (e.g. benched
area, technical feasibility, and cost. The ‘one-size- slopes), mechanical stabilization (e.g. mechanically
fits-all’ approach does not work as site conditions stabilized earth walls), vegetative erosion control
and constraints vary greatly. (e.g. hydroseeding), soil bioengineering, and bio-
technical stabilization.
2.2 Causes and characteristics of slope instability
Most landslides and slope stability problems 3 SOIL BIOENGINEERING
involve water-related factors–presence of natural
springs, seepage from water channels and septic Soil bioengineering uses locally available plants
tanks, improper disposal of black water, intense and plant material and a minimum of equipment
rainfall (cloudbursts), rapid snowmelt, floods, to inhibit slope failure and treat unstable critical
rapid changes in water table and soil pore pres- slopes. It attempts to mimic nature (Lewis et al.
sures, as well as stream erosion. Other isolated 2001). In bioengineering applications, it is neces-
causes include earthquakes, volcanic eruptions, but sary to distinguish between slope protection and
above all human activity, including improper road slope stabilization, the former is aimed at control-
construction techniques, e.g. uncontrolled clearing ling slope erosion and shallow slope failure, while
and grubbing, blockage of natural drainage paths, the latter is aimed at remediating failures associ-
side-casting of excavated earth and rock, poor ated with deeper slope movement. Bioengineering
placement of fill material, uncontrolled blasting, applications combine elements of protection and
and vibrations from passing vehicles. stabilization to prevent long-term, permanent slope
The shape of slope is often a defining factor in deterioration. They are most effective in combat-
slope stability. Natural slopes are generally con- ing erosion (both rill and gully), soil flowage (up
cave (most stable and least susceptible to erosion). to a depth of 150 mm) and localized translational
Man-made slopes, on the other hand, are linear (in shear failure of soil slopes (TRL/ODA 1997).
many cases, these slopes erode until they assume Deep rooted grasses (e.g. Vetiver) and hydrophilic
a concave shape). Slope linearity is often broken trees (such as poplar and acacias) have shown to be
by benching and terracing, but poorly designed effective in soil fixing and protecting slopes up to a
benches can cause severe gully erosion at water depth of about 2–3 meters.
outfalls (Schor & Gray 2007, Shah 2008).
Other typical features of unstable slopes include:
3.1 Vegetative stabilization mechanisms
(i) excessive slope angle or height; (ii) increased slope
loading often due to slope toe cutting by streams Soil bioengineering utilizes six stabilization mech-
and rivers and accumulated material/man-made anisms, conceptually similar to how a soil mass is
loadings at crest; (ii) unstable geological strata and fixed by a deep-rooted tree as shown in Figure 1.
presence of joints, faults, disconformities, creep, These six functions serve: (i) to catch eroded mate-
and weak foundations; (iii) poor drainage, seepage, rials with physical barriers (e.g. vegetated walls,
natural springs, high soil moisture and soil lique- plants, trees); (ii) to armor the slope from erosion

56

TIS2017_Book.indb 56 2/8/2017 8:01:10 PM

 Table 1. Comparison of geotechnical engineering
and soil bioengineering systems relative to stabilization
function.

Geotechnical
engineering Soil bioengineering
Function system system

Catch Gabion wire Contour lines (grass

bolster lines; brush layers)
Check dams Live check dam
Catch walls Shrubs and bamboo
(many stems)
Armor Revetment walls Mixed plant stories,
Stone pitching giving complete
cover
Grass carpet (dense,
fibrous roots)
Reinforce Reinforced earth Densely-rooting
Figure 1. Stabilization mechanisms involved in soil Soil nailing grasses, shrubs,
mass fixation by deep-rooted vegetation. and trees
Most vegetation
structures
caused by runoff or rain splash using vegetative Anchor Rock anchors Deeply-rooting trees
cover, partial armoring using lines of vegetation; by bolting and shrubs (long
string roots)
(iii) to reinforce soil physically with plant roots; (iv)
Support Retaining walls Large trees and
to anchor surface material to deeper layers using
Breast walls bamboo
large vegetation with deep roots; (v) to support soil Prop walls (deep and dense
by buttressing with vegetated walls or large vegeta- root system)
tion; and (vi) to drain excess water from the slope Drain Surface drains Down slope and
through evapotranspiration and the use of drains French drains diagonal vegetation
(Howell 1999a, Schor & Gray 2007). lines
A comparison of geotechnical and soil bioengi- Angle fascines,
neering techniques and their combined use in rela- pole drains
tion to the six stabilization functions is summarized
in Table 1. The choice of stabilization techniques Source: Shah 2008, Howell 1999.
(both geotechnical and bioengineering) depends
on an identification of the functions needed to sta-
bilize and protect the slope. A detailed description
of these techniques and their application is pro-
vided by Howell 1999b.

3.2 Soil bioengineering components

and structures
There is a wide range of techniques associated with
bioengineering systems and their management as
reported by Coppin & Richards 1990, Gray & Sotir
1996, Schiechtl & Stem 1996, Howell 1999b, Shah
2008, and Fay et al. 2012. Grasses, shrubs and trees
are the building blocks of soil bioengineering. As Figure 2. Soil bioengineering components and associ-
ated engineered structures and devices.
shown in Figure 2, they can be used directly for
slope stabilization or transformed into vegetative The soil bioengineering components used in
structural components with hedge and brush layer- typical engineered structures for slope protection
ing, branch packing, stakes, palisades, fascines, and and stabilization are summarized in Table 2.
so on. More advanced components include veg- The selection of a soil bioengineering technique
etated walls and drains, and live fences, sills, check for protecting or stabilizing a particular slope
dams. The effectiveness of soil bioengineering sys- depends on many site-related factors. In order
tems is significantly enhanced when they are used in to select the most relevant and appropriate tech-
conjunction with engineered structures and devices. nique, it is necessary to consider these factors in as

57

TIS2017_Book.indb 57 2/8/2017 8:01:10 PM

 Table 2. Soil bioengineering applications in slope stabi- 3.3 Advantages and disadvantages of soil
lization structures. bioengineering
Structure Bioengineering component Soil bioengineering provides a sustainable means
of protecting and stabilizing slopes for reasons of
Retaining walls Vegetated soft gabions, timber availability, relatively low cost, appropriateness of
crib walls and used tire installation techniques and compatibility with rural
walls; Live brushwood environment. It is particularly appropriate in situa-
Diversion channels Sodding treatment and stone tions where large areas of slope are affected, a com-
pitching
mon situation on road cuttings and over unstable
Surface drains Sodding treatment and stone
pitching with live sills
mountain slopes. The enhancement of roadside
Check dams Vegetated poles, live brush vegetation also has a positive aesthetic effect both
wood, palisades, vegetated visually and in terms of plant diversity (TRL/ODA
brush wood 1997). Soil bioengineering applications are labor
Barriers for fixing Brush wattles, brush layering, intensive and a source of job creation, especially
loose debris brush fences, semi-dead for rural women, a distinct advantage in labor
fences with live hedges, abundant economies. As native plant material is
seeding and turfing used, it can be sourced locally without the risk of
Retaining walls Vegetated soft gabions, introducing invasive plant species. Technically it is
timber crib walls and used simple and does not require sophisticated engineer-
tire walls; Live brushwood ing designs. The bioengineered structures are long
lasting and get stronger with passage of time. Once
mature, they become a productive asset providing
timber, fuelwood and fodder, and may also provide
a habitat for wildlife. Soil bioengineering is envi-
ronmentally and ecologically friendly and could
have a significant role in reducing GHG emissions.
Soil bioengineering, however, has some disadvan-
tages. It is a season bound activity and cannot be car-
ried year round. Moreover, it is affected by high and
low temperatures as well as dry and wet spells. The
bioengineered structures can be adversely affected
by droughts and intense storms (damage from wind,
rain, hail and snow) especially during the first 2–3
years. The timing of implementation also becomes a
critical factor, with planting mostly confined to the
dormant season (late fall or early spring in temper-
ate zones, and just before the rainy season in tropi-
cal zones). Its implementation requires on-the-job
Figure 3. Soil bioengineering treatment of a landslide training and skilled oversight, as well and support
in Northern Pakistan. and vigilance of local communities to protect newly
A) Treatment plan for landslide remediation. B) Actual established vegetation and structures. Plants need
application of remediation measures. C) Before bioengi- protective fencing from foraging and grazing ani-
neering treatment. D) After bioengineering treatment. mals for 4–5 years, and from humans seeking fodder,
firewood, and forage. The design and implementa-
tion of bioengineering programs requires interdis-
much detail as time and skill permit. Assessments
ciplinary skills and interagency coordination (e.g.
rely mainly on experience as it not practicable to
between forestry and road departments) and may be
quantify the many variables involved. Although
difficult to obtain in institutionally weak settings.
vegetation cannot be designed or built to engineer-
ing specification in the conventional sense, it can
be selected and arranged on the slope to perform a 4 CASE STUDY: ISLAMABAD-MURREE
specific engineering function. This function should EXPRESSWAY (PAKISTAN)
be identified as part of the process of bioengineer-
ing design (TRL/ODA 1997). Figure 3 illustrates The 43 km Islamabad—Murree Expressway (E75)
the treatment plan and application of soil bioen- connects Islamabad, the country’s capital, with
gineering techniques for landslide remediation in Lower Topa (Murree District) in Pakistan. This
northern Pakistan, as well as ‘before’ and ‘after’ four-lane divided expressway was constructed by
condition of the treated site. the National Highway Authority (NHA) over the

58

TIS2017_Book.indb 58 2/8/2017 8:01:11 PM

 period 2004–2012. The E75 traverses the Murree
Formation, a geological zone of young, steep and
fragile mountains (Bossart & Ottiger 1989), con-
sisting primarily of cross bedded sandstone and
claystone/shale strata, with extensive springs and
seepage zones. The rock structure varies from fine
to coarse grain, with parallel laminations, amalga-
mation, and calcitic veins (Khan & Niazi 2010).
Unstable geology, a wet climate (torrential mon- Figure 5. Massive reinforced concrete walls used as
soon rains in summer), and human activity (defor- resistance structures to contain landslides.
estation, unstable and poorly drained terraces for
cultivation and housing, and improper disposal of The monsoon seasons since 2010 have been char-
sewage) are some of the causal factors responsible acterized by more intense and frequent rainfall, with
for the landslide hazard along the route. The geo- severe weather events, resulting in landslides at some
graphical location of E75 is shown in Figure 4. 70 locations along an 18 km section of the express-
The E75 has an average roadway formation width way. The roadway has subsided at several sites
of about 25 m, which required extensive clearing of resulting in cracking and collapse of the massive
natural vegetative cover and deep linear cuts (with retaining walls and closure of one carriageway in a
side casting of waste material) in steep (30–40% few sections. Remedial treatment utilizing a combi-
gradients) natural slopes, making them unstable and nation of geotechnical and biotechnical measures
triggering huge landslides even during the construc- has helped to control the landslides and stabilize
tion period. Other landslides resulted from bank the disturbed slopes at most of the affected loca-
cutting by mountain streams due to inadequate tions. NHA, however, continues to struggle with the
drainage and lack of slope protection measures E75 slope instability problem. It owes its genesis to
such as provision of cut-off drains and check dams. poor route location, design and construction prac-
Based only on rudimentary geotechnical investiga- tices, and will now require continuous monitoring
tion and design, resistance structures (massive rein- and intensive routine and emergency maintenance.
forced concrete retaining walls) were constructed at
points where the landslides were triggered during
4.1 Landslide stabilization project
construction. The height, length, and width of these
walls were dimensioned as a function of the severity In January 2014, NHA in collaboration with the
and extent of the landslide (Figure 5). provincial Punjab Forest Department initiated a
Control structures such as diversion channels, project for the stabilization and follow-up main-
and surface and subsurface drains were not used tenance of active landsides with biotechnical, soil
despite their critical role in landslide management bioengineering and biological methods. In total,
and control. Measures to control rill and gully about 70 active landslides between Km 26 and Km
erosion of exposed slopes were also absent dur- 44 (18 km long section) were treated during 2014,
ing expressway construction. As a result landslides 2015 and spring of 2016. In addition gabion and
remained active along the expressway during and vegetated gabion check dams were installed on
after construction and many reinforced concrete mountain streams endangering the expressway’s
walls failed with increased water content and pore right of way through bank cutting. The project
pressures in the backfill during the wet season. was allocated a budget of PKR 42.5 million (USD
453,000), at an average cost of about $75,000 per
landslide treatment location.

4.2 Measures used for slope stabilization

The slope stabilization measures utilized by the
project included engineering, biotechnical, soil
bioengineering and biological (vegetative) applica-
tions. The primary focus was on soil bioengineer-
ing applications.

4.2.1 Engineering measures

These consisted of gabion check dams on 15 mountain
Figure 4. Physical location of Islamabad—Murree streams to reduce bank cutting and common gabion
Expressway, E75, with concentration of unstable slopes retaining walls to further strengthen the Reinforced
between Km 26 and 44. Concrete (RCC) retaining structures. Each stream

59

TIS2017_Book.indb 59 2/8/2017 8:01:11 PM

 was treated with 3–5 check dams, with the height
ranging from 2–3 m. The main purpose of the gabion
walls, placed in one or two layers, was to prevent
landslide debris from overtopping the RCC walls.

4.2.2 Biotechnical measures

When living plants are incorporated in engineer-
ing structures they are called biotechnical struc-
tures. Such biotechnical structures were used in the
project for retaining walls, check dams, diversion
channels and surface drains.
Retaining walls. For constructing a vegetated
dry masonry wall brush and hedge layering was
applied at every 0.5 m height of the dry masonry
Figure 7. Vegetated check dams. A) Pole construction
wall. Brushwood of poplar, willow, mulberry and before sprouting and B) after sprouting. C) Soft gabion.
seedlings of Robinea pseudoacacia were used in D) Dry masonry.
the brush layering treatment. Although used vehi-
cle tires have been used for constructing retaining
walls in different parts of the world, the use of a channels and surface drains both were treated with
vegetated used tire wall was pioneered under this stone pitching and supported with willow stakes
project for the first time in 2015. Used tires were to help retain the stones in place. The stakes after
placed in staggered layers with stakes and seed- sprouting help stabilize the channels and the drains.
lings planted in the earth-filled tire. In addition to
the plantings, brush layering treatment was also 4.2.3 Soil bioengineering measures
applied to each layer of tires (Figure 6). A variety of soil bioengineering structures were
Vegetated check dams. Both gabions and dry put to use as retaining walls, drainage structures,
masonry were utilized for construction of check and mechanical barriers to fix the landslide debris
dams. Vegetated gabion check dams were con- before planting.
structed on streams running laterally to the dual Retaining Walls. The structural elements for these
carriageway formation. These check dams, one walls consisted of soft gabions made from used
meter in height, consisted of three layers of gabion empty cement bags filled with landslide debris or
wire boxes. Each gabion layer was covered by soil timber crib walls. The vegetated soft gabion walls
and brushwood of poplar willow Ipomea with the were used on most of the landslides because they
basal ends of brushwood extending beyond the are an economical alternative to standard wire gabi-
check dam wall. The check dams were back filled ons filled with stone, especially where stones are not
with soil for rooting of the brush wood (Figure 7). available on site or it is cumbersome to haul stones up
Vegetated dry masonry check dams were con- steep eroded slopes. The bags were filled with debris
structed in the gullies within the landslide-affected at the toe of the landslides and used as the founda-
slopes and proved effective in stabilizing them. tion for constructing the retaining wall with brush
Drains. Biotechnical treatment was applied and hedge layering. After sprouting, the brushwood
to diversion channels to divert runoff water from provides a strong and durable retaining structure.
entering the slide area and to surface drains to The bags eventually rot and plant roots, trunks and
remove rainwater from the slide area. The diversion branches serve as a vegetated wall (Figure 8).
Although vegetated timber crib walls are not
economical in Pakistan they were used at two land-
slide locations in this project as there was no other
suitable technical alternative. The timber cribs were
constructed using brushwood of native poplar and
willow species (Figure 9).
Drainage structures. Soil bioengineering was
used for construction of diversion channels, sur-
face and pole drains and check dams. Extensive
diversion channels were excavated at ten project
sites in the undisturbed area above the active
landslide. Some excavated channels were lined
with stone pitching but others were treated with
sodding and supported with vegetated soft gabion
Figure 6. Vegetated used tire wall at time of construc- walls on downhill side. Sodding is quite effective as
tion and after sprouting of vegetation. the grass cover protects the channel from scoring

60

TIS2017_Book.indb 60 2/8/2017 8:01:12 PM

 of plant types was used to provide a range of rooting
depths. As far as possible, local species rather than
imported material were used because native plants
are better adapted to grow in the hostile conditions
found on bare sites and are resistant to local diseases.
Fast growing broad leave tree species such as Robinea
pseudoacacia, Ailanthus altisimma, Poplar (Populus
detoides) and Willow (Salix tetrasperma) were planted
on landslide debris mass to provide quick vegetation
cover. It was also necessary to plant local conifer tree
species, Chir pine (Pinus roxburghi) and at higher alti-
tude Deodar (Cedrus deodara), in combination with
broad leave tree species to encourage ultimate colo-
nization by the indigenous varieties. Stakes were pre-
Figure 8. Vegetated soft gabion wall (one seasonal cycle). pared by cutting fresh branches of willow and poplar
(Populus deltoides and Populus ciliata) to supplement
the planting of seedlings. To supplement these plant-
ings, the seeds of indigenous trees and bushy plants
were also sown, namely Chir pine, Acacia modesta,
Wendlandia exserta (Ukan), Leucania leucocephalla
(Ipel ipel) and Dodenia viscosa (Sanatha).

Figure 9. Timber crib walls used in combination with 4.3 Results of slope stabilization project
gabions.
The E75 Slope Stabilization Project treated 70 land-
and helps stabilize the channel quickly. Excavated slides ranging from earth slumps to rock and debris
surface drains were sodded and supported by soft slides to earth subsidence: 51 in 2014, 17 in 2015
gabion walls at five sites to drain runoff produced and 2 new ones in 2016. The remedial work was car-
on the landslide. On steep slopes surface drains are ried out in spring to permit the vegetative elements
not stable and pole drains were utilized for draining to root before the monsoons. The typical treatment
surface seepage water. Pole drains were installed at (other than diversional channels, drains and check
four sites by excavating a trench and placing bun- dams as needed) consisted of vegetated dry stone
dles of live brushwood in the trench and covering it masonry and soft gabion walls, brush layering and
with soil, with one fifth portion of the brush wood planting, staking, and sowing of indigenous tree
bundle left uncovered for sprouting. Four types species (Figure 10). This basic soil bioengineering
of soil bioengineering check dams were used for treatment helped stabilize all the treated landslides
stabilizing gullies on the bigger landslides, namely, except at three locations, despite unusually heavy
palisades, vegetated soft gabion, vegetated pole, and rains both during the two ensuing monsoons as
vegetated brushwood check dams. well as in the winter of 2015 and spring of 2016
Mechanical barriers (contour walls). It is neces- (which triggered several new slides).
sary to fix the landslide mass debris before plant- Vegetation recovery was impressive at all the
ing; otherwise seedlings planted in the upper three failed locations. However, the presence of
portion of the landslide are uprooted while plant-
ings on lower side are buried under the debris. The
normative engineering solution for fixation of
debris is the construction of stone contour bunds
(contour walls). The soil bioengineering techniques
for creating mechanical barriers are brush layer-
ing, bush hedge layering, semi-dead fences with live
hedges and sodding. The use of these live mechani-
cal barriers in the project also helped to inhibit rill
and gully formation.

4.2.4 Biological (vegetative) measures

Purely biological applications for erosion control
and slope stabilization used in the project included
planting of bare rooted seedlings and tube plants,
staking with live willow and poplar branches, and Figure 10. Treatment of landslide at kilometer 29 + 250
sowing seeds of indigenous plant species. A mixture on Route E75 below a damaged house.

61

TIS2017_Book.indb 61 2/8/2017 8:01:13 PM

 a spring in the middle of the landslide and seep- iv. any slope segment in which civil engineering
age from the septic tanks of two houses above the structures are planned or have been built, and
landslide caused the soil bioengineering treatment the surface remains bare;
to fail in one location. The second failure was due v. any area that has failed and needs to be restored,
to the landowner not permitting the proper instal- other than rock slopes; and
lation of surface drains on his property. And the vi. any area, such as tipping and quarry sites, or
third failure was due to water seepage into the camp compounds, that requires rehabilitation.
slide area from the drainage runoff from roofs of
houses located above the landslide; the applica-
tion of a pole drain, however, proved effective in ACKNOWLEDGEMENT
draining the seepage water and the landslide was
eventually stabilized with vegetation. At six loca- Maryam Faiz prepared all the art work for the paper.
tions, cattle grazing could not be controlled and Her technical support is gratefully acknowledged.
vegetation recovery was not satisfactory.
REFERENCES
5 CONCLUSIONS AND Bossart, P. & Ottiger, R. 1989. Rocks of the Murree For-
RECOMMENDATIONS mation in Northern Pakistan: Indicators of a descend-
ing foreland basin of late Paleocene to middle Eocene
With increasing rainfall intensity and shorter age. Eclogae Geologicae Helvetiae 82(1):133−165.
return periods of climate-induced extreme weather Coppin, N.J. & Richards, I.G. 1990. Use of vegetation in
events, unstable slopes and landslides pose a seri- civil engineering. London: Butterworths.
ous risk to transport infrastructure assets, espe- Faiz, A. et al. 2015. Prevention is better than cure: bio-
cially in mountainous areas. There is an urgent engineering applications for climate resilient slope sta-
need to deliver cost-effective and sustainable solu- bilization of transport infrastructure assets. Proc. of
tions to mitigate the landslide risk. First International Conference on Surface Transporta-
tion System Resilience to Climate Change and Extreme
As shown by the case study included in this Weather Events. 16–18 September, 2015. Washington,
paper, barring catastrophic events like earthquakes DC. Transportation Research Board.
or floods, most landslide problems affecting trans- Fay, L. et al. 2012. Cost-Effective and Sustainable Road
portation facilities have their origin in faulty route Slope Stabilization and Erosion Control. NCHRP
location, design that is not sensitive to environ- Synthesis 430. Washington, DC. Transportation
mental and climate change considerations or con- Research Board.
struction and maintenance practices that worsen Gray, D.H. & Sotir, R.B.1996. Biotechnical and soil bioen-
the vulnerability to the landslide hazard, especially gineering slope stabilisation. A practical guide for ero-
in mountainous terrain. The study shows that a sion control. New York: Wiley & Sons.
Howell, J. 1999a. Bioengineering in Bhutan: Interim
majority of slope instability problems can be recti- manual for the road sector on bioengineering tech-
fied by the application of fairly simple and envi- niques for slope protection and stabilization. Thimpu.
ronment friendly slope stabilization measures such Department of Roads.
as vegetative erosion control, soil bioengineering, Howell, J. 1999b. Roadside Bio-engineering: Site Hand-
and biotechnical stabilization. Climate resilience is book, Kathmandu. His Majesty’s Government of
significantly enhanced by using a combination of Nepal (Department of Roads).
drainage, slope protection and stabilization meas- Khan, I. & Niazi, M.J. 2010. Geological Report of Hazara
ures, including both inert (engineered retaining Division. Abbottabad, Pakistan. COMSATS Institute
structures and earthworks and mechanical stabili- of Information Technology.
Shah, B.H. 2008. Field Manual on Slope Stabilization,
zation) and vegetative components. Islamabad, United Nations Development Program.
Most landslides and slope stability problems Schiechtl, H.M. & Stern, R. 1996. Ground bioengineer-
involve water-related factors. Soil bioengineering ing techniques for slope protection and erosion control.
combined with appropriate water management can Oxford: Blackwell Science.
be the key to preventing slope failures in transport Schor, B. & Gray, D.H. 2007. Landforming: An Environ-
infrastructure facilities, especially in regions sub- mental Approach to Hillside Development, Mine Rec-
ject to climate stress from increased storm inten- lamation and Watershed Restoration. Hoboken, NJ.
sity and frequency. As recommended by Howell John Wiley & Sons. J.
(1999a), soil bio-engineering techniques for stabi- TRL/ODA 1997. Principles of low cost road engineer-
ing in mountainous regions. Overseas Road Note 16.
lizing slopes should be used on: Crowthorne, Berkshire. Transportation Research
i. all areas of bare soil on cut face and embank- Laboratory.
ment slopes; USDA.1992, Soil Bioengineering for Upland Slope Pro-
ii. wherever there is a risk of gullying; tection and Erosion Reduction, National Engineering
Handbook, Part 650, Engineering Field Handbook,
iii. all slopes where there is a risk of shallow slumps Chapter 18. Washington DC. Natural Resources Con-
or planar slips of less than 500 mm depth; servation Service, US Department of Agriculture.

62

TIS2017_Book.indb 62 2/8/2017 8:01:15 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Monitoring of railway track with light high efficiency systems

S. Cafiso, B. Capace, C. D’Agostino, E. Delfino & A. Di Graziano

Department of Civil Engineering and Architecture, via Santa Sofia, University of Catania, Catania, Italy

ABSTRACT: Rail Track performance is highly dependent upon the magnitude and variation of dif-
ferential track geometry. Consequently, there is an increasing interest in obtaining a consistent measure
of trackbed stiffness and track geometry in the physical and time restrictions applicable on live track for
assessing potential maintenance requirement and subsequent design of remedial measures. Furthermore,
a lack of a systematic monitoring brings to the impossibility to produce an effective long term track
management system, by allocating budget where emergencies come. In this framework, in the case of local
railway track with reduced gauge, the use of the traditional high speed track monitoring systems is not
feasible. This paper details a site investigation comprising trial Ground Penetrating Radar (GPR), Light
Falling Weight (LWD) and Laser Measurement System (LCMS) testing. These systems are currently used
in road pavement maintenance where have shown their reliability and effectiveness. Application of such
Non Destructive Tests in railway maintenance are promising but in the early stage of investigation. In the
paper literature review and trial site testing are used to identify Strengths, Weaknesses, Opportunities and
Threats (SWOT analysis) of the application of GPR, LWD and LCMS for the assessment of trackbed
stiffness and geometry.

1 INTRODUCTION speeds and higher capacity (more trains on the

tracks and heavier trains). Nowadays more trains
Railway lines are investments with very long life. occupy the track and the competition with other
Today many tracks are over 100 years old. Of means of transportation becomes harder. To
course components and mainly rolling stock have face the new circumstances, more effort has to
been exchanged during the years, but parts of the be put on track maintenance to ensure the issues
track might remain the same—especially the sub- of safety, comfort, serviceability and economy.
structure. Typical lifetimes of for rails are 30–60 Moreover, there is also a trend towards decreased
years and turnouts 20–30 years (H. Sundquist, time for maintenance and decreased funds for
Byggande, 2000). However, to ensure this long life maintenance.
a large amount of maintenance is necessary. There Theoretically, the number one solution for
are several, quite obvious reasons for maintenance, optimal maintenance is to do the right measure at
for example: the right time to fulfil the requirements of safety,
comfort and serviceability in terms of LCC. This
• Safety—Probability for accidents needs to be
task is practically impossible, since it requires
low.
complete knowledge about the current condi-
• Comfort—Comfort is important, both for pas-
tion of the track and what effect different kinds
sengers and freight as well as for the environ-
of maintenance, or no maintenance, will have on
ment in terms of noise and vibration.
the track. Instead, the goal of condition based
• Serviceability—With lots of failures and speed
maintenance is to come as close to this optimum
restrictions due to safety etc. the serviceability
as possible. This could be done with the help of
of the track will be low.
measurements of important parameters which are
• Economy—A track with low quality is cost driv-
analyzed to give knowledge about the condition
ing, since the deterioration of both track and
of the track. After that, regulations, budget con-
trains will be higher.
straints and knowledge about the deterioration
At the same time maintenance is expensive. (from models and/or skilled engineers) are used
Optimization and Life Cycle Cost (LCC) plan- to make decisions about maintenance. Unfortu-
ning is needed. Since many railway tracks are nately, again this kind of approach require a con-
quite old, the demands put on the tracks today trol of the network, or of those parameters able
are different from the ones when the tracks to represent the real condition of the track and
were built. There is a clear trend towards higher substructure.

63

TIS2017_Book.indb 63 2/8/2017 8:01:17 PM

 2 CONTROL TEST IN RAILWAY as the economic possibility for a systematic global
survey. The result is that the maintenance is carried
Track failures is one of the main problems that rail- out to avoid track failures with visual and manual
roads have faced since the earliest days is the preven- inspection of the track based on the experience of
tion of service. The North American railroads have the technicians.
been inspecting their costliest infrastructure asset, the Local railway tracks, usually with reduced gauge
rail, since the late 1920’s; and, with increased traffic have not the potentiality to develop their own high
and more frequent failures on railroads, rail inspec- efficiency monitoring systems. Unfortunately, the
tion is more important today than it has ever been lack of a systematic control and the economic dif-
(C. NDT, 2013). Although the focus of the inspec- ficulties, especially in local railway track, lead to the
tion seems like a fairly well-defined piece of steel, impossibility of producing a long-term effective
the testing variables present are significant and make track management system. Specifically, in presence
the inspection process challenging. To keep railroads of local tracks with reduced gauge, the use of tradi-
safe and prevent any high maintenance costs caused tional monitoring systems track at high speed is not
by failures on the railroads, scheduled inspections feasible. In this framework the preventive monitor-
must be performed on rail tracks, soil, and bridges. ing of the track with alternative Non-Destructive
There are several types of track inspections such as Techniques (NDT), such as GPR, LWD and LCMS
soil inspection, railroad bridge inspection, and rail- are promising for soil and rail inspection and to
road inspection, and each track inspection type has plan a preventive maintenance from the early stage
their subcategories shown in the following Figure 1. of investigations. In recent decades, these devices
Soil inspection investigates the Ballast, Sub- have already proven their effectiveness in the field
grade, and Roadway (BSR) component, which of road pavement engineering and prospects are the
includes all earthen materials on the track structure, same in the rail sector which is increasingly growing
tracks, and embankments (Uzarski et al., 1993). (Cafiso et al. 2016a, Cafiso et al. 2016b).
It focuses on the thickness of the ballast, subsoil
material and geotechnical properties of subgrade.
In addition, the inspection is executed mainly by 3 SWOT ANALYSIS
digging trenches at evenly spaced intervals and in
locations of special interest (Hugenschmid, 1999). Because of the interest in the introduction of new
Soil inspection also investigates the plants on the NDT for the investigation of railway track, a SWOT
railroads. For instance, Eriksen et al. (2004) tried analysis will be performed basing on the literature
to improve the reliability of soil inspection by add- review and trial tests of GPR, LWD and LCMS.
ing the investigation of plant growth in the inspec- Originated by Albert S. Humphrey in the 1960s,
tion process. Rail inspection investigates the rail SWOT analysis is a basic, straightforward model that
heads, switch blades, bolt holes, foot of the rail, assesses what an organization can and cannot do as
rail gauge, thermite welds, etc. (H. NDT, 2014). well as its potential opportunities and threats. It is an
Latest technology can improve the rail inspection extremely useful tool for understanding and decision-
with higher cost of the equipment (Cerniglia et al., making for all sorts of situations and disciplines in
2006). This technology is mainly developed directly project planning, management and business for
from the railway Agency to monitor their network. investigating problems from a strategic perspective.
The great advantage is related to the speed and SWOT is an acronym for Strengths, Weaknesses,
precision of the measurement, avoiding to inter- Opportunities and Threats that, specifically, takes
fere with the normal use of the infrastructure. information from an environmental analysis and
Generally local railways have not the most
advanced equipment to survey the network as well

Figure 1. Types of track inspections in a railroad. Figure 2. SWOT analysis general schema.

64

TIS2017_Book.indb 64 2/8/2017 8:01:17 PM

 separates it into internal strengths and weaknesses,
as well as its external opportunities and threats.
SWOT Analyses are often arranged as a 2 by
2 matrices with the lists of strength and weaknesses
in the first two boxes in the first row and the lists
of opportunities and threats in the second row. By
arranging the analysis this fashion, the lists are sepa-
rated into internal factors that can affect a project on
the first row and external factors on the second row.
In addition, the first column consists of the positive
factors (strengths and opportunities) and the second
column consists of negative actors (weaknesses and
threats.). This method provides a simple framework
to keep lists organized and conceptualize how the Figure 3. GPR principles (Saarenketo, 2006).
lists are related. SWOT analysis is presented in the
present research work, for the application of GPR, – Mapping soil, rock or fill layers in geological and
LWD and LCMS to survey a railway track. geotechnical investigations, or for foundation
The analysis was conducted for each single design.
equipment after a trial on a real railway environ-
ment and post processing of the results. The track substructure, consisting of the ballast,
SWOT analysis resulted particularly interesting sub ballast, and subgrade layers, has a profound
to evaluate the feasibility of application of those influence on track performance. The substructure
equipment in a different context than their usual performance is significantly affected by moisture
test environment. accumulation and thickness of the roadbed layers
(Selig & Waters, 1994).
Accurate knowledge of the substructure condi-
tion is important in effectively assessing the poten-
4 GROUND PENETRATING RADAR (GPR)
tial for service interruptions and the need for slow
orders. A significant part of a railroad’s track
The ground penetrating radar is a geophysical
maintenance budget is allocated to correct track
radar system with two antennas and receivers
roughness that is caused by movements in the sub-
used to perform non-destructive investigations of
structure under repeated train loading. In this field,
underground characteristics with high resolution
the GPR method seems to be a good alternative to
and in depth (up to 3.2 m from the surface).
traditional core inspection techniques. Methods of
GPR operation principle is based on electromag-
applying GPR to railways are being developed to
netic theory, its functioning consists in sending short
provide a continuous evaluation of the track sub-
electromagnetic pulses into a medium and when pulses
structure conditions relative to subsurface layering,
achieve an interface they are reflected back partially
material type, moisture content and density.
and collected by the receiving antenna (Figure 3). The
GPR application to railways, during construc-
reflected energy is displayed in waveforms and the
tion and monitoring phase, is relatively recent.
greatest amplitudes represent the interfaces between
In Germany, Saarenketo (2006) referred that
layers with distinct dielectric characteristics (Daniels,
Göbel et al. (1994) performed GPR tests for deter-
2004;). Therefore, GPR measures the travel time
mining: ballast thickness, layers interfaces, ballast
between the transmission of the energy pulses and its
pockets and mudholes location.
reception. Transmission and reception of radar pulses
Jack and Jackson (1999) studied ballast layer
are performed from one or more antennas that are
along a track using two antennas with different
moved on the investigated medium (Sussmann et al.,
frequencies, they found: clearer ballast interfaces
2003). The collected data are processed and saved on
indicating a clean ballast and thickness variations,
a control unit, that is also used to generate the neces-
affirmed GPR as a useful tool for identifying track
sary pulses for the operation of the antennas.
sections with urgent necessity of rehabilitation.
As mentioned GPR can provide a fast, nonde-
Gallagher et al. (1999) researches found posi-
structive measurement technique for evaluating
tive results for survey of ballast/subgrade interface,
railway track substructure condition (Olhoeft &
such as anomalies detection. Hugenschmidt (2000)
Selig, 2002; Selig et al. 2003).
reports a study developed on different alignments,
Main application of GPR in railway track are:
evaluating the ballast thickness, fouled zones and
– monitoring the condition of railway ballast, ballast/subgrade interface depth. Their conclusion
and detect zones of clay fouling leading to track was that radar survey is useful combined with tra-
instability ditional inspection methods. Recently, to determine

65

TIS2017_Book.indb 65 2/8/2017 8:01:18 PM

 the correlation between water content or fouling of
a railroad track and GPR signals, a full-scale rail-
way track model was designed and constructed at
the University of Massachusetts Amherst (Hamed
et al. 2016). Different models were tested with mois-
ture content conditions of dry, saturated and two
points between these extremes. 450 MHz and 2 GHz
frequency antennas were used to evaluate the differ-
ent conditions. The results shown that the dielectric
permittivity and frequency spectrum can be used
as an indicator of fouling percentage and moisture
Figure 4. Ground Penetrating Radar. content in a track. In addition, a linear correlation
is observed between the fouling percentage and
moisture content under saturation conditions.

4.1 GPR in field trials

The system is composed by two antennas of
600 MHz and 2 GHz frequency. The frame was
adapted for the test by modifying the wheel system
and the Distance Measurement Instrument (DMI)
configuration. Wheels are built with cone shape and
made with resin for running on the railway track
and avoiding electrical interferences (Figure 4).
Figure 5 shows the radar cans obtained from the
two antennas. As expected 2 GHz gives better res-
olution for the upper layers’ interface (e.g. 0.3 m)
complemented by the 600 kHz for layers’ interface
at lower depth (e.g. 1.0 m).
Basing on the literature review and the practi-
cal in field trial, the SWOT analysis for GPR was
developed and it is reported in Table 1.

5 LIGHT WEIGHT DEFLECTOMETER

Figure 5. Radar scans of the trial test with 600 kHz (a)
and 2 GHz (b). The Light Weight Deflectometer (LWD) (Fig-
ure 6), is a hand portable device that was developed
in Germany to measure the soil in situ dynamic

Table 1. SWOT analysis for GPR System.

STRENGHT WEAKNESSES
(internal, positive factor) (internal, negative factor)

• Reliable and established technology • Radar analysis subject to interpretation of the operator
• NDT testing • Still provisional models
• Speed of execution • Need of technologically and advanced equipment
• Reduced number of operators • Need of calibration tests for data deduction
• More issues investigated • Interference of magnetic materials
• Potential data on the whole structure
OPPORTUNITIES THREATS
(external, positive factor) (external, negative factor)

• Development of new models for extrapolation • Results distorted by unexpected and/or unforeseen
of results. conditions
• Approach to a field of engineering technique • Difficulties in the interpretation of the data obtained
that is not “saturated’’ • Ignorance and mistrust of the Public Administrations in the
• Possibility of agreements with public authorities or use of these devices for NDT on railways
private companies that need of an advanced know-
how for the development of its own procedures

66

TIS2017_Book.indb 66 2/8/2017 8:01:18 PM

 modulus. The LWD consists of a circular plate track geometric position caused by the pushing of
(150, 200, 300 mm diameter) loaded by a falling fine-grained soil from the subballast into the ballast
mass (10–15 – 20 kg). bed (so-called pumping effect) and by the missing
The load resulting from the drop is dampen by base layer between the ballast and the subgrade.
a series of buffers, producing a transient force, The exploitation potential of the Lightweight
dynamic in nature, simulating the effect of the Falling Deflectometer is still increasing, and the
vehicles moving on the road surface. At the same results are aptly combined with other analyti-
time the resulting deflection is measured directly cal and mathematical methods (Fernandes, et al.,
under the plate. The LWD, used for the trial test, 2012; Burrow et al, 2007). In some cases, the
had one geophone positioned in the centre of the impact device is used for a specific assessment of
plate and 2 additional geophones that can be used spots with problems resulting from the unstable
for specific measures outside the plate. geometric track position (Sharpe & Govan, 2014).
The LWD is ideal for Quality Assurance/Quality
Control on subgrade, subbase and thin flexible
pavement constructions to verify that specifica- 5.1 In field trials
tions are met. It can also be used to identify weak- LWD was used to test the response in terms of
nesses, leading to further tests using FWDs and deflection in railway track with different character-
other material analysis techniques. istics of ballast and sleepers and with the loading
When applied directly on the subgrade, data plate on the ballast or on the sleeper. Particularly
processing from LWD test is carried out accord- the trial was conducted on four different sites with
ing to Boussinesq theory for the estimation of the the same LWD configuration (plate 150 mm, weight
Elastic modulus E by the formula: 10 kg, drop height of about 68 cm–27 inches):
• Site 1: concrete sleepers on ballast with partial
f (1 − v 2 )σ 0 ⋅ a
E= renovation and hilling.
d0 • Site 2: concrete and wood sleepers on ballast in
bad conditions.
where (f) is the plate rigidity factor, (a) is the load • Site 3: concrete sleepers on just renewed ballast;
plate radius, (d0) is the measured deflection under
the center of the plate, ν is the Poisson’s ratio. Results, reported in Figure 7, show a clear vari-
The influence of soil depth of LWD tests is con- ability in the measured deflection among the sites
sidered to be 1÷1.5 diameters of the plate. with different ballast conditions.
Neupane et al. (2016) from the university of Site with ballast of good quality (site 3) reports
Kansas estimated with LWD the resilient modulus lower deflections and higher uniformity.
of the substructure considering different combi- The SWOT analysis for LWD is reported in
nation of fouled ballast (10–40% by weight) and Table 2.
moisture content (1–10%). The test was conducted
by reproducing in laboratory the real condition.
6 LASER CRACK MEASUREMENT
Considering the various combination, they con-
SYSTEM (LCMS)
cluded that moisture content has the highest influ-
ence in reducing the bearing capacity. The LCMS system employs high speed cameras, cus-
Hornícek et al. (2014) used the LWD for the tom optics, and laser line projectors to acquire 2D
long-term evaluation of the condition of trial sec- images and high resolution transversal profiles of
tions with the application of under-ballast geocom- road surfaces 4 meters wide. LCMS was originally
posites useful to avoid long-term problems of the for automatic detection of cracks, rut depth, macro-
texture and other road pavement distress. The LCMS
can be operated at speeds of up to 100 km/h.

Figure 6. Light Weight Deflectometer. Figure 7. Light Weight Deflectometer.

67

TIS2017_Book.indb 67 2/8/2017 8:01:20 PM

 Table 2. SWOT analysis for LWD System.

STRENGHT WEAKNESSES
(internal, positive factor) (internal, negative factor)

• Reliable technology • Techniques of analysis often subject to

• NDT testing comparison
• Reduced number of operators • Variability of the results
• Possibility to investigate the various components • Loads insufficient for the optimal evaluation
of the infrastructure of the ballast bearing capacity
• Application of mathematical and analytical models • Need of calibration tests for the
data acquisition
• Speed of execution
OPPORTUNITIES THREATS
(external, positive factor) (external, negative factor)

• Development of new models for the interpretation • Results distorted by unexpected and/or
of the results unforeseen conditions and on-board effects
• Approach to a field of engineering technique that cannot be assessed
is not “saturated’’ • Returning of untruthful data about the
• Possibility of agreements with public authorities bearing capacity of the superstructure
or private companies that need an advanced know-how • Ignorance and mistrust of the Public
for development of its own procedures Administrations in the use of these devices
for NDT on railways

The only experimental trials of LCMS on rail-

way tracks were carried out by Pavemetrics®. The
surveys were conducted by using a multifunction
vehicle adapted for running on a railway track with
standard gauge (Figure 8).
That configuration minimize the needs for
sensors placement representing the optimum in
terms of sensor distances from the ground DMI
synchronization.
There are not yet other experiences reported in
literature. Most probably it is related to two key
factors, (i) the technology of LCMS was developed
only in the last decade and (ii) the high cost of
Figure 8. Multifunctional vehicles used by Pavemetrics®
equipment does not allow a widespread diffusion to test the LCMS performance on surveying railway tracks.
in a such short time.

6.1 In field trials

LCMS are actually installed on the Automatic
Road Analyzer of the Department of Civil Engi-
neering & Architecture of University of Catania.
The test was conducted putting ARAN on ad hoc
prepared railway wagon (Figure 9). This configura-
tion is not the optimum can be achieved because of
the increased distance of the laser from the track, Figure 9. LCMS configuration and DMI installed on
the rail wagon wheel.
but, at this stage of the research, that was the only
configuration available to survey the reduced rail-
way gauge which does not allow ARAN to travel calibrated for taking into account the new con-
with its own wheel on the railroad track. figuration (LCMS height and DMI resolution).
Another issue to be solved, was the coupling of As mentioned above, basic LCMS output are
LCMS with the DMI which control the frequency images and cross section profiles. High resolution
of acquisition. images can be used to replace a visual rail inspection
A medium resolution DMI (2048 pulse/round) (e.g. sleepers, fast, ballast). Cross section profile
was installed and connected with the acquisition allows to detect gauge values and track geometry.
unit in ARAN. Before the trial the system was re- A sample output of is reported in Figure 10.

68

TIS2017_Book.indb 68 2/8/2017 8:01:21 PM

 Figure 11. 3D rendering of railroad track (courtesy of
Figure 10. Output from LCMS survey. Pavemetrics®).

Table 3. SWOT analysis for LCMS system.

STRENGHT WEAKNESSES
(internal, positive factor) (internal, negative factor)

• Established technology with increasing diffusion • Technologically advanced equipment (expen-

• NDT testing sive and complex)
• High performance in surveys and post-analysis • Require high operator training for survey and
• Reduced number of operators post elaboration
• Chance to investigate more features of the infrastructure • difficulty adapting configurations used in road
• Data or information about the most important track issues surveys to railway tracks
• Higher precision of data • Not suitable for adverse climatic conditions

OPPORTUNITIES THREATS
(external, positive factor) (external, negative factor)

• Development of new models for the extrapolation of indirect • Ignorance and mistrust of the Public Admin-
index from surveys data istrations in the use of these devices for NDT
• Approach to a field of engineering technique that is not “saturated’’ on railways
• Possibility of agreements with public authorities or private companies • Partial lack of interest of railway companies to
that need of an advanced know-how for the development of its own use new technologies
procedures

Due to the laser height and DMI resolution the work, monitoring is still challenging in terms of
resolution of output was of 3 mm in transversal pro- costs and maintenance needs. Furthermore, a lack
file with a longitudinal acquisition step of 5 mm. of a systematic monitoring brings to the impos-
Coupling the system with an Inertial Measure- sibility to produce an effective long term track
ment Unit (IMU) it will be possible to correct the management system, by allocating budget where
undesired wagon motion (roll, pitch, yaw) joining emergencies come. The present paper has explored
the successive section to obtain a complete 3D new and promising solutions to overcome this limi-
scan of the track. tation by using the equipment well known for road
This data can be combined with the digital pavement monitoring, but not yet diffused in the
images with surprising results in terms of resolu- railway management. Particularly the paper focused
tion and details (Figure 11). on the adaptation needed to use GPR, LWD and
The SWOT analysis for LCMS is reported in LCMS on a railway track to test the bearing capac-
Table 3. ity and quality of ballast and track geometry. The
results are presented in terms of a SWOT analysis
based literature review and on site trials.
7 CONCLUSIONS More specifically, the field tests were carried out
on a local railway with reduced gauge of 900 mm.
The main railway lines are by now monitored That condition gave the opportunity to test the
with high efficiency equipment based on the most system in an environment open to the introduction
advanced technologies in the field. In the case of of such system, but posed also specific limitations
local railway, which are not part of the main net- for the use of the systems.

69

TIS2017_Book.indb 69 2/8/2017 8:01:22 PM

 GPR and LWD are not new in such application railway tracks, In Fifth International Conferention on
and confirmed the suitability of the equipment for Ground Penetrating Radar.
testing the quality of ballast. More specifically, Sundquist, H. Byggande. 2000. Drift och Underhåll av
LWD, despite of the limited load (6 kN in the trial Järnvägsbanor, TRITA-BKN, Rapport 57, KTH, Com-
pendium in Swedish, Stockholm.
test, while an optimum configuration required Hamed F. Kashani, Carlton L. Ho, William P. Clement,
heavier loads) applied directly on the sleeper, was & Charles Oden, 2016. Evaluating The Correlation
able to detect defects in the bearing capacity of the Between Geotechnical Index and Electromagnetic Prop-
Sleeper/Ballast system. erties of Fouled Ballasted Track By Full Scale Labora-
More challenging is the introduction of LCMS tory Model, TRB 2016 Annual Meeting.
that showed high potentialities, but also equip- Horníček, L. & Břešťovský, P. 2014. Using the Lightweight
ment costs and installation issues that need further Falling Deflectometer for Monitoring Trial Railway Sec-
investigations. tions with Under-Ballast Geocomposites, In Railway Con-
dition Monitoring (RCM 2014), 6th IET Conference on.
Hugenschmid, J. 1999. Railway track inspection using
ACKNOWLEDGEMENT GPR. Journal of Applied Geophysics: 147–155.
Hugenschmidt, J., 2000. Railway track inspection using
The authors wish to express their gratitude to Pav- GPR. J. Appl. Geophys. 43:147–155.
Humphrey, A. December 2005. SWOT Analysis for Man-
emetrics for the support in the development of the agement Consulting. SRI Alumni Newsletter. SRI Inter-
trial test with LCMS. ARAN, LWD and GPR were national: 7–8.
acquired by the University of Catania in the frame- Jack, R. & Jackson, P. 1999. Imaging attributes of railway
work of the Regional RESET project, funded by the track formation and ballast using ground probing radar.
Sicily Region Authority. The authors wish to express NDT&E International, No. 32:457–462.
their gratitude also to “Ferrovia Circumetnea” and NDT, C. 2013. Rail Inspection, Retrieved from NDT Resource
Ventura Company for them in field assistance during Center: http://www.ndted.org/AboutNDT/SelectedAppli-
the surveys and to make available the railway sites. cations/RailInspection/RailInspection.htm
NDT, H. 2014. A document outlining the emergence of the
eddy current NDT inspection method as an important
REFERENCES part of rail maintenance and safety.
Neupane, M., Parsons, R.L. & Han, J. 2016. Rapid estima-
Burrow, M.P.N., Chan, A.H.C. & Shein, A. 2007. Deflec- tion of fouled ballast material properties. Transportation
tometer based analysis of ballasted railway tracks, In Research Board of the National Accademy, Washington
Proceedings of The Institution of Civil Engineers geotech- DC, Annual Meeting 2016.
nical Engineering, 160(3): 169–177. Olhoeft, G. R., & Selig, E. T., 2002. Ground penetrating
Cafiso, S., Capace, B., D’Agostino, C., Delfino, M., & radar evaluation of railroad track substructure condi-
Di Graziano A. 2016a. Application of NDT to railway tions. Proc. Of the 9th Int’l Conf. on Ground Penetrating
track inspection. International Conference on Traffic and Radar, Santa Barbara, CA, April, S. K. Koppenjan and
Transport Engineering, 24th – 25th November 2016, Bel- H. Lee, eds., Proc. of SPIE, vol. 4758: 48–53.
grade (Serbia). Saaranketo, T. 2006. Electrical properties of road materials
Cafiso, S., D’Agostino, C., Capace, B., Motta, E., & Capill- and subgrade soils and the use of ground penetrating
eri, P. 2016b. Comparison of in situ devices for the radar in traffic infrastructure surveys (PhD Thesis). Uni-
assessment of pavement subgrade stiffness. 1st IMEKO versity of Oulu.
TC4 International Workshop on Metrology for Geotech- Selig, E. T. & Waters, J. M. 1994. Track Geotechnology and
nics, March 17–18, 2016, Benevento, Italy. Substructure Management. Thomas Telford Services
Cerniglia, D., Garcia G., Kalay, S. & Prior, F. 2006. Appli- Ltd., London.
cation of Laser Induced Ultrasound for Rail Inspection. Selig, E. T., Hyslip, J. P., Olhoeft, G. R., & Smith, Stan,
Railway Research Center. 2003. Ground penetrating radar for track substruc-
Daniels, D.J. 2004. Ground Penetrating Radar, 2nd Edition. ture condition assessment. Proc. Of Implementation of
IET. Heavy Haul Technology for Network Efficiency, Dallas,
Eriksen, A., Gascoyne, J., & Al-Nuaimy, W. 2004. Improved TX, May, pp. 6.27–6.33.
Productivity & Reliability of Ballast Inspection using Sharpe, P.C. & Govan, R. 2014. The use of Falling Weight
Road-Rail Multi-Channel GPR. Proceedings of Railway Deflectometer to Assess the Suitability of Routes for
Engineering. London, UK: IEEE. Upgrading”, In Proceedings of the Second International
Fernandes, J., Paixão, S. Fontul, & Fortunato. E. 2012. Conference on Railway Technology: Research, Develop-
The Falling Weight Deflectometer: Application to Rail- ment and Maintenance, Civil-Comp Press, Paper 134,
way Substructure Evaluation, In Proceedings of the Stirlingshire, UK.
First International Conference on Railway Technology: Sussmann, T.R., Selig, E.T. & Hyslip, J.P. 2003. Railway
Research, Development and Maintenance, Paper 130, track condition indicators from ground penetrating
Civil-Comp Press, Stirlingshire, UK. radar. NDT E Int. 36:157–167.
Gallagher, G.P., Leiper, Q., Williamson, R., Clark, M.R., & Uzarski, D.R., Brown, D.G., Harris, R.W. & Plotkin, D.E.
Forde, M.C., 1999. The application of time domain 1993. Maintenance Management of U.S. Army Rail-
ground penetrating radar to evaluate railway track bal- road Networks-the RAILER System: Detailed Track
last. NDT E Int. 32: 463–468. Inspection Manual. Champaign, IL: US Army Corps
Göbel, C., Hellmann, R., & Petzold, H., 1994. Georadar- of Engineers; Construction Engineering Research
model and in-situ investigations for inspection of Laboratories.

70

TIS2017_Book.indb 70 2/8/2017 8:01:24 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Reference trajectories of vehicles for road alignment design

G. Cantisani & G. Loprencipe

Dipartimento di Ingegneria Civile, Edile e Ambientale, Sapienza, University of Rome, Rome, Italy

ABSTRACT: The geometric design principles traditionally hypothesize that trajectories of vehicles can
be assumed corresponding to the road longitudinal axle of the road or, more correctly, to the median
line of each allowed lane. In real conditions, instead, vehicles always travel along trajectories each other
different; the variability is due to dynamic actions affecting the motion and, in a great measure, because the
control of vehicle trajectories, performed by users, is not perfect. In order to consider if theoretical models
can be effective for safety and comfort verifications in design process, it is important to evaluate how a
reference trajectory can statistically represent the whole population of road users. In fact, the difference
between a real trajectory of a generic vehicle and the theoretical one can emphasize the safety problems
related to geometric characteristics of roads. To deal with these problems, it appears interesting to analyse
the dispersion of trajectories in various road sections; in this way, in fact, the “reference trajectory” along
a road alignment can be recognized by means of a statistical approach. Starting from surveys on real road
elements, the paper presents a method aimed to obtain trajectories that have formal geometric expression
and that can correctly represent the scattering of vehicles’ position, because reference lines are defined
after a statistical analysis of collected data.

1 INTRODUCTION et al, 1988; Cantisani & Di Vito, 2012) provide

speed values and speed profiles as the result of the
Vehicle trajectories, in real conditions, are generally analysis of road alignments. In these models the
scattered about the geometric reference alignments dynamic equilibrium of vehicles refers to circular
of the given road section; in order to use design curves corresponding to the reference road axis.
models that can better represent vehicle trajecto- In other research papers the same basic con-
ries of real vehicles in motion, it is important to cepts are used for calculating and evaluating speed
evaluate how reference trajectories can statistically profiles (Messer, 1980; Fitzpatrick et al, 2000, Has-
represent the whole population of road users. san, 2004, Cafiso et al, 2005), in order to ensure
Traditionally, in fact, the design theories assume consistency and homogeneity of road alignments
that the median line of each allowed lane repre- (Gibreel et al, 1999).
sents the reference curve for an “ideal” trajectory. With regard to safety issues, vehicle stability
As a consequence, considering that the lanes are along theoretical trajectories is noted as a critical
usually parallel, the centre of the carriageway— problem, because conventional models are some-
that is equicentric to the median lines of all allowed times deemed too approximate. In particular, the
lanes—can be assumed as the design reference line effect of a vertical curve or grade has not been
or, more briefly, as the road axis. In this assump- considered (Furtado et al, 2002). Other consid-
tion, it is implicitly established that the geometry erations are referred to the analysis of horizon-
of design line has to be consistent with vehicle tal curves (Reinfurt et al, 1991) that can increase
dynamics (in particular speed and speed rate) and speed reduction and edgeline encroachments on
various possible manoeuvres along road section. the inside lane.
Most of road design guidelines in the world Some studies underline that the width of lanes
(AASHTO 2001; VSS 1991; IT-DM 1991; TAC cannot be neglected with regard to the influence
1999), are based on this design hypothesis. on speed and vehicle lateral placement (Neuhardt
In addition, theoretical studies on road geo- et al, 1971; McLean, 1974); comparisons between
metric design generally correspond to the above real road conditions and simulator environment
presented basic principles. For example, the works confirm these outcomes (Blana & Golias, 2002).
focused on the Design Speed, Operating Speed and Moreover, the effects of lane widths on road safety
Design Criteria (Leisch & Leisch, 1997; Lamm are not so easy to recognize: Hauer (2005) affirms

71

TIS2017_Book.indb 71 2/8/2017 8:01:25 PM

 that lane width plays a different role in single and how the real position of vehicles in a traffic stream
multi-lane roads, because for single-lane roads it differs from the ideal one, because in this way a
has a bigger influence on driver behaviour, in terms statistical evaluation regarding the percentage of
of trajectory and selected lateral position. vehicles that have a risk exposure greater that the
Considering various phenomena that can occur theoretical one can be obtained. At the same time,
where vehicles travel along a road, real trajectories considering that the transversal dimension of road
are always different to the “ideal” one (Glennon elements and the distances between vehicles and
& Weaver, 1972); in fact, the dynamic actions on lateral obstacles can influence some important
vehicle body, tires and mechanical parts (Bonneson functional parameters, like the average of vehicu-
et al, 2007), in addition to the intrinsic variability lar speeds and the capacity of the road sections,
due to imperfect user control of steering and speed the statistical distribution of the lateral displace-
(Rosey & Auberlet, 2012; Cantisani et al., 2013), ment is significant in order to consider the func-
significantly affect the vehicle’s motion or the tional performances of a road section.
user’s positioning of their vehicle in the available In particular, the scattering of vehicles’ posi-
cross-section or both. As a result, designers need to tions in each cross section, belonging to a road
be cautious when using current design theories to alignment, has to be investigated with the aim of
improve safety and comfort of road sections. defining a reference line as the result of statistical
In particular, a probabilistic approach could be analyses on collected data. For these purposes, var-
incorporated in the road design process to account ious researches focused on the collection of experi-
for vehicle positioning and trajectory, which in mental data about operating conditions in traffic
turn impacts vehicle speed and safety. Accord- flows and/or single vehicles motion (Yu et al, 2001;
ing to (Hirsh et al, 1986), designers should note Harlow & Peng, 2001; Coifman et al, 1998).
that «current practice is based on a deterministic Important tools for performing surveys over
approach whereas the factors involved in the geo- vehicular traffic are now available. It is possi-
metric design process (e.g, speed, friction, reaction ble to obtain measurements and observations by
time) are stochastic in nature and vary among road means of automatic detection sensors like micro-
users». In this way it would be possible to evalu- wave radar and infrared detectors (Cantisani et al,
ate how a reference trajectory can statistically rep- 2012), or video cameras, and to store and process
resent the road user population; in other studies the data by a number of electronic resources. In
(Blana & Golias, 2002), it is recommended that particular, the AVIP—Automatic Video Image
designers explore a more “rational style”, rather Processing (Semertzidis, 2010; Song & Tai, 2007;
than a “pragmatic” one, for road safety manage- Laparmonpinyo & Chitsobhuk, 2010)—are inte-
ment, including a review of the criteria aimed to grated systems that allow to continuously detect
obtain a correct design and assessment of road the traffic scenes on a section of road and obtain
geometric characteristics. traffic parameters and relevant lane positioning
To fulfil the above recommendations, the knowl- information.
edge of users’ behaviour and the study of their The data provided by traffic monitoring systems
vehicle trajectories are essential in order to deter- can be presented as a set of points (like scattered
mine the driver’s likely behaviour on the actual seeds) that represents the sequence of positions
project being analysed. of a representative part of each vehicle that is fol-
lowing its particular path. Linking these points
(Figure 1), after a coordinates transformation
2 STUDY, METHODS AND OBJECTIVES process (Wei et al, 2005; García & Romero, 2009),
it is possible to obtain curves that give a continu-
To deal with the above presented problems, it is ous representation of trajectory: in this way, a
interesting to analyse the dispersion of trajectories geometric approximation of the position of the
at various road sections. The “reference trajecto- vehicle, between two points recorded by the moni-
ries”, in fact, can be assumed as a statistical rep- toring system, is produced, but the error results
resentations of the actual dispersed ones, but it is not significant for the usual frequency sampling of
necessary to know the limits and the accuracy of the recording devices.
this assumption. It is also possible to obtain the distribution of
In other terms, many important design verifi- lateral vehicle displacements (Figure 2), at each
cations, directly related to the safety of the infra- point along the road element, as the intersection
structure, like the dynamic equilibrium in a curve of each vehicle’s trajectory with cross section lines.
and the sight distance assessment, are developed The distribution can be analysed by means of sta-
starting from the assumption that the position tistical techniques, so obtaining a representation of
of any vehicle along a road section agrees to an the way in which the road user (driver) population
ideal trajectory. Therefore, it is important to know travels along the road infrastructure.

72

TIS2017_Book.indb 72 2/8/2017 8:01:26 PM

 3 DATA ANALYSIS

Statistical analyses were performed on real vehicle

trajectory data; the data came from video camera
surveys on various road elements, collected at criti-
cal points (curves, ramps, intersections, …) along
a road, and processed with an original method. In
particular, in various cross sections along the exam-
ined roads, the distribution of lateral displacement
was extracted as previously described.
The methodology for successive treatments was
previously developed and presented by the same
Authors (Cantisani & Loprencipe, 2013); it is inter-
esting to consider, here, what is the theoretical and
practical interpretation of the obtained results.
In particular, the statistical meaning of data dis-
tribution can be expressed as following: if the lat-
eral displacements present a normal distribution, it
is possible to individuate the mean (μ) as the value
that corresponds to a point, in the road cross sec-
tion, where the probability a vehicle exactly passes
there reaches the maximum. Other statistical param-
eters (like standard deviation σ, etc.) can allow to
establish where are the points related to some char-
acteristic percentiles, with reference to the prob-
ability that vehicles, passing through the analysed
sections, fall into the space included between these
points. For example, it is possible to consider the
characteristic points defining the intervals where a
certain percentile of users is included: this is the so-
called 68-95-99.7 empirical rule (see Figure 3): for a
sample of data normally distributed, about 68% of
values fall into the interval [μ-σ, μ+σ], about 95%
into the interval [μ-2σ, μ+2σ] and about 99.7% into
the interval [μ-3σ, μ+3σ]. The statistical parameters
can be used also for safety and functional design
evaluations (Figure 4).
Figure 1. Example of vehicle trajectory, obtained as a To obtain trajectories having a formal geomet-
continuous line that interpolates a points’ seeding cor- ric expression (but also representing the scatter-
responding to the left rear tyre of vehicle. ing of vehicles’ positions) the means of statistical
distributions in selected sections can be linked by

Figure 2. Distribution of vehicle displacement obtained Figure 3. Percentiles of population falling into the char-
by crossing the observed trajectories with section lines. acteristic intervals for the 68-95-99.7 rule.

73

TIS2017_Book.indb 73 2/8/2017 8:01:26 PM

 Figure 4. Examples of design evaluation of sight condi-
tions and lateral clearance for the mean and for the 95th
percentile of users (Cantisani & Loprencipe, 2013).

a curve that interpolates these points. The curve

should be derivable and should maintain a conti-
nuity of order of at least equal to 2, in order to sat-
isfy the real variation of the kinematic properties
related to the vehicle motion. Furthermore, for a
generic couple of points in sequence, it is required
that no sign change has to occur, in the value of the
curvature ρ = 1/R.
In this research the best results, as regard to the Figure 5. The diagram shows, in the sequence: the cubic
spline interpolation of the maximum probability points
geometric representation, were obtained by inter-
(“reference trajectory”), the curvature diagram and the
polating the points by means of polynomial curves dispersion of real trajectories.
“in pieces” (Hermite Parametric Cubic Spline, see
Figure 5).
The meaning of the described treatments is the road sections. For example, the vehicle trajectories
following: the resulting lines join the points having in the section no.6 are more scattered than in the
the maximum probability of vehicle passages, so, section no.3 where the vehicles adopt a trajectory
in a statistical sense, they are the most representa- cutting the curve in order to minimize the centrip-
tive lines that one can assume as “reference trajec- etal acceleration forces and to maximize the speed.
tories” (Figure 6). In addition, the large scattering of trajectories in
In the Figure 6 is possible to note that the normal the section no.6 probably is due to the presence
distributions are not the same for all the considered of a ramp entrance where two traffic flows are in

74

TIS2017_Book.indb 74 2/8/2017 8:01:28 PM

 are reported: they are calculated starting by the
trajectories of about 200 vehicles recorded on a
road ramp. The average values are calculated with
reference to different points in the cross sections
because the dimension of lane is different along
the road. In fact, the zero reference point used for
calculating the displacement in each section is the
edge of the internal line delimiting the ramp.

4 DISCUSSION OF RESULTS

The above presented results, obtained by statisti-

cal treatments on collected experimental data, deal
with some remarks about the actual behaviour of
drivers and the way they interact with the road
infrastructure. In particular, two main observa-
tions can be presented, that result in good agree-
ment with other studies presented in the literature
review (see previous chapter 1).
First of all, we can see that if we link the maxi-
mum probability points, in each cross section
along a road alignment, the obtained line can
significantly differ from the theoretical one. This
circumstance frequently occurs and in particular,
referring to the presented examples, it can be noted
in the case of the ramp terminal belonging to an
intersection.
In this road section, in fact, the radius (R2) of
the osculating circle obtained by linking the point
of maximum probability in the cross sections
no.2, no.3 and no.4, is larger than the one (R1)
considered for the geometric design (or theoreti-
cal assessment) of the road element (see Figure 7,
R2 >> R1). This circumstance entails that the
actual speed vehicles can reach by travelling along
the ramp terminal, even if they maintain a safe
Figure 6. Example of reference trajectory for the case condition with regard to the dynamic equilibrium,
of a ramp terminal in a grade-separated intersection. can result significantly higher than the one result-
ing by theoretical analyses. The underestimation
Table 1. Average values and the standard deviations of of speed, in this particular road section (ramp ter-
the six normal distributions of vehicle trajectories. minal in a grade-separated intersection), can also
produce a not negligible influence in the merging
Average of Standard deviation process between primary and secondary traffic
vehicle positions of vehicle positions flow. Thus, in the design analyses, the problem
Section μ (m) σ (m) should be taken in account in order to prevent
possible conflicts in the traffic manoeuvres. As a
1 2.032 0.450
2 1.968 0.448
suggestion, we can deduce that in similar cases the
3 2.068 0.341
design could consider a set of possible input data
4 2.188 0.427 (referring to geometrical and cinematic parameters
5 2.640 0.573 of vehicle’s motion), ensuring that also the most
6 3.003 0.597 critical ones will be compatible with road infra-
structure and traffic conditions.
In addition, another remark emerges by consid-
conflict and there are different opportunities to ering the variability of standard deviation values
occupy the main lane for merging vehicles. in various sections. In the sections no.1 and no.2,
In the Table 1 the average values and the stand- in fact, the standard deviation is higher than in
ard deviations of the six normal distributions the sections in the middle of the curve (no.4 and,

75

TIS2017_Book.indb 75 2/8/2017 8:01:29 PM

 5 CONCLUSIONS

The vehicular trajectories, along a road section, are

important because their configuration can influence
the development of design safety and functional
verifications (dynamic equilibrium, sight distance,
road capacity or average speed of a traffic stream).
However, it is important to consider that a theoreti-
cal trajectory can only have a statistical meaning,
because in real conditions the vehicles can move
through the road section also passing in a different
position (in comparison to the “ideal” one).
Therefore the analyses of vehicle paths, along
road sections, can be used to identify the “refer-
ence trajectories” for the design of road alignments,
based on statistical analysis of experimental data.
In fact, it is possible to observe the distribution of lat-
eral displacement in each cross section along a road
element. After a geometric treatment of the data,
the reference trajectories can be obtained as the lines
that continuously interpolate the points where the
probability to find a vehicle, when it passes through
each cross section line, reaches the maximum. In the
same examined points, other statistical properties
can be observed, like the shape of distribution and
its parameters (standard deviation, etc.).
The possible uses of these reference lines deal
with the safety or functional problems, related to
road design procedures. In fact, a reference trajec-
tory can be assumed as the median line of each
allowed lane, but often it would be also necessary
to examine other conditions, for the design verifi-
cations accomplishment, so considering the scat-
tering of vehicles’ positions in the real trajectories.
As a practical deployment, it is possible to observe
Figure 7. Comparison between the reference trajectory in the Figure 4 how the design verifications, regard-
(in red) and the theoretical one (in blue) in the case of a ing sight conditions and lateral clearance respectively
ramp terminal in a grade-separated intersection. for the mean and for the 95th percentile of users, are
different; so, the knowledge of trajectories distribu-
tion allows to consider and evaluate the risk expo-
especially, no.3). This depends essentially by the sure conditions for the whole population of users.
more effective trajectory control that a sharper In addition, based on the observed data, some
curvature radius can determine in comparison to remarks regarding main implications for the design
a larger one. Moreover, also the lane width repre- and the checks for ramps are presented in the paper,
sents an essential factor, because the displacement in agreement with other literature studies.
typically presents a larger variance when the lane More general, the application of presented prin-
is larger. Consequently, we can assume that the ciples allows improving the knowledge of real road
lane width should be correctly assessed in the geo- operations; this opportunity is showed in the paper,
metric design process, also considering the special considering various case studies, especially referred
characteristics or the function of the road element to some critical points (curves, ramps, intersections
(a ramp terminal, in this case), in order to ensure a …) along examined road elements.
more effective control of trajectories.
Different considerations have to be reserved to
REFERENCES
the standard deviation in the sections no.5 and
no.6, where the high values are primarily due to AASHTO (2001). Policy on Geometric Design of High-
the entrance manoeuvre accomplishment; there- ways and Streets. Washington, DC: American Associa-
fore, they seem quite independent respect to road tion of State Highway and Transportation Officials, 1:
geometry. 990, 2001.

76

TIS2017_Book.indb 76 2/8/2017 8:01:31 PM

 Blana, E., & Golias, J. (2002). Differences Between Vehicle Hirsh, M., & Prashker, J.N., and Ben-Akiva, M. (1986).
Lateral Displacement on the Road and in a Fixed-Base New Approach to Geometric Design of Highways. In
Simulator. In Human Factors: The Journal of the Human Transportation Research Record: Journal of the Trans-
Factors and Ergonomics Society, 44(2), pp 303–313. portation Research Board, No. 1100, pp 50–57.
Bonneson, J. A., Pratt, M., Miles, J., & Carlson, P. (2007). IT, D.M. (2001). Norme Funzionali e Geometriche per
Development of Guidelines for Establishing Effective la Costruzione delle Strade. (Italian Standard for Geo-
Curve Advisory Speeds. Texas Transportation Institute, metric and Functional Road Design)—D.M. 05/11/2001.
Texas A&M University System. Rome: Ministry of Infrastructures and Transportation.
Cafiso, S., Di Graziano, A., & La Cava, G. (2005). Actual Lamm, R., Choueiri, E.M., Hayward, J.C., & Paluri, A.
Driving Data Analysis for Design Consistency Evalua- (1988). Possible Design Procedure to Promote Design
tion. In Transportation Research Record: Journal of the Consistency in Highway Geometric Design on Two-
Transportation Research Board, No. 1912, pp 19–30. lane Rural Roads. In Transportation Research Record:
Cantisani, G., & Di Vito, M. (2012). CCV: A New Model Journal of the Transportation Research Board, No. 1195,
for S85 Prediction. In Procedia-Social and Behavioral pp 111–122.
Sciences, No 53, pp 765–776. Laparmonpinyo, P., & Chitsobhuk, O. (2010). A Video-
Cantisani, G., Di Vito, M., & Luteri, P. (2012). VPL based Traffic Monitoring System Based on the Novel
Project’09: An Integrated Station for Vehicles’ Operat- Gradient-edge and Detection Window Techniques. The
ing Conditions Survey. In Procedia-Social and Behavio- 2nd International Conference. In Computer and Automa-
ral Sciences, 53, pp 777–788. tion Engineering (ICCAE), 3, pp 30–34.
Cantisani, G., & Loprencipe, G. (2013). A statistics based Leisch, J.E., & Leisch, J.P. (1977) New Concepts in Design-
approach for defining reference trajectories on road sec- Speed Application. In Transportation Research Record:
tions. In Modern Applied Science 7 (9), pp. 32–46. Journal of the Transportation Research Board, No. 631,
Cantisani, G., Loprencipe, G., & Primieri, F. (2011). The pp. 4–14.
integrated design of urban road intersections: A case McLean, J.R. (1974) Driver Behaviour on Curves—A
study. In The International Conference on Sustainable Review. In Australian Road Research Board (ARRB),
Design and Construction (pp. 722–728). Conference, Adelaide. Vol. 7.
Coifman, B., Beymer, D., McLauchlan, P., & Malik, J. Messer, C.J. (1980) Methodology for Evaluating Geometric
(1998). A Real-time Computer Vision System for Vehi- Design Consistency. In Transportation Research Record:
cle Tracking and Traffic Surveillance. In Transporta- Journal of the Transportation Research Board, No. 757,
tion Research Part C: Emerging Technologies, 6(4), pp. 7–14.
pp 271–288. Neuhardt, J.B., Herrin, G.D., & Rockwell, T.H. (1971).
Fitzpatrick, K., Wooldridge, M.D., Tsimhoni, O., Collins, Demonstration of a test-driver technique to assess the
J.M., Green, P., Bauer, K. Parma, K.D., Koppa, R., effects of roadway geometrics and development on speed
Harwood, D.W., Anderson, I., Krammes, R.A., & Pog- selection. Federal Highway Administration, Study No
gioli, B. (2000). Alternative Design Consistency Rating Hpr-1(4).
Methods for Two-lane Rural Highways. In FHWA-RD- Reinfurt, D.W., Zegeer, C.V., Shelton, B.J., & Neuman, T.R.
99-172, pp 1–154. (1991). Analysis of Vehicle Operations on Horizontal
Furtado, G.A., Easa, S.M., & Abd El Halim, A.O. (2002). Curves. In Transportation Research Record: Journal of
A Vehicle Stability on Combined Horizontal and Verti- the Transportation Research Board, No. 1318, pp 43–50.
cal Alignments. Carleton University. Annual Conference Rosey, F. & Auberlet, J-.M. (2012) Trajectory variability:
of the Canadian Society for Civil Engineering. Montreal, Road geometry difficulty indicator. In Safety Science,
Quebec, Canada, June 5−8. 50(9), pp. 1818–1828.
García, A., & Romero, M. (2009). Discussion of ‘Video- Semertzidis, T., Dimitropoulos, K., Koutsia, A., & Gram-
Capture-Based Approach to Extract Multiple Vehicu- malidis, N. (2010). Video Sensor Network for Real-time
lar Trajectory Data for Traffic Modeling’ by Heng Wei, Traffic Monitoring and Surveillance. In Intelligent
Chuen Feng, Eric Meyer, and Joe Lee. In Journal of Transport Systems, IET Vol. 4 Issue 2, pp 103–112.
Transportation Engineering, 135(3), pp 149–150. Song, K.T., & Tai, J.C. (2007). Image-based Traffic Moni-
Gibreel, G.M., Easa, S.M., Hassan, Y., & El-Dimeery, I.A. toring with Shadow Suppression. In Proceedings of the
(1999). State of the Art of Highway Geometric Design IEEE, 95(2), pp. 413–426.
Consistency. In Journal of Transportation Engineering, TAC. Geometric Design Guide for Canadian Roads. Ottawa,
125(4), pp 305–313. Canada: Transportation Association of Canada. 1999.
Glennon, J.C., & Weaver, G.D. (1972). Highway Curve VSS. (1991) Schweizer Norm SN 640-080-b, Projektierung
Design for Safe Vehicle Operations. In Highway Research Grundlagen–Geschwindigkeit Als Projektierungselement.
Record, No. 371. Vereinigung Schweizerrischer Strassenfachleute, Zurich,
Harlow, C., & Peng, S. (2001). Automatic Vehicle Clas- Switzerland.
sification System with Range Sensors. In Transporta- Wei, H., Feng, C., Meyer, E., & Lee, J. (2005) Video-
tion Research Part C: Emerging Technologies, 9(4), capture-based Approach to Extract Multiple Vehicu-
pp 231–247. lar Trajectory Data for Traffic Modeling. In Journal of
Hassan, Y. (2004). Highway Design Consistency: Refining Transportation Engineering, 131(7), pp 496–505.
the State of Knowledge and Practice. In Transportation Yu, X., Sulijoadikusumo, G., Li, H., & Prevedouros, P.
Research Record: Journal of the Transportation Research (2011). Reliability of Automatic Traffic Monitoring
Board No. 1881.1, pp 63–71. with Non-Intrusive Sensors. In Proceedings of the 11th
Hauer, E. (2005). The Road Ahead. In Journal of Transpor- International Conference of Chinese Transportation Pro-
tation Engineering, 131(5), pp 333–339. fessionals. Nanjing, China, August 14–17.

77

TIS2017_Book.indb 77 2/8/2017 8:01:32 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Decoupling of wheel-rail lateral contact forces from wayside

measurements

D. Cortis, M. Bruner & G. Malavasi

Department of Civil, Building and Environmental Engineering, Sapienza University of Rome, Rome, Italy

ABSTRACT: Running safety and stability of railway vehicles are provided by respect of track geometry
and by observance of the limits of wheel-rail contact forces. The ratio between lateral and vertical forces
has an effect on the guiding forces provided by the rail, which prevent the wheel flange climbing and the
derailment. This highlights the importance of the development of measurement methods of wheel-rail
contact forces. In this paper, we present an experimental method to estimate the lateral contact force
starting from the measurement of the strains on the rail foot surface. A suitable combination of the
recorded strains allows to reproduce the same continuous signal of the applied lateral force, decoupling
the effects of the vertical one. These studies, based on finite element simulations and laboratory tests,
show how to find a constant ratio between the applied lateral load and the recorded strains on the rail
foot surface.

1 INTRODUCTION attracted the interest of the railways infrastructure

managers. The identification in real-time of the
Running safety and stability of railway vehicles magnitude of the forces exchanged between the
are provided by respect of track geometry and wheel and the rail allows the monitoring of the rail
by observance of the limits of wheel-rail contact traffic and the reporting of any non-compliance
forces. The ratio between lateral and vertical forces to the respective railways companies. The main
has an effect on the guiding forces provided by issue of this type of continuous measurements is
the rail, which prevent the wheel flange climbing the decoupling of the effects of the wheel-rail con-
and the derailment. The safety against derailment tact forces. The simultaneous presence of bending
depends moreover by the wheel-rail contact condi- moments and torques produce on the rail surfaces
tion as the contact area, the shape of the wheels a complex strain state that makes difficult identi-
and rail’s profile, the wear of the contact surfaces, fying the effects produced by the corresponding
the friction factor and the influence of the envi- forces.
ronmental conditions. All these physical quantities One of the strategies to decouple the wheel-rail
change the distribution and the magnitude of the contact forces and evaluate their magnitude is to
contact forces. The evaluation of the wheel-rail con- discover, on the rail surface, specific areas in which
tact forces plays also a key role on the maintenance to measure the strains. These specific areas could
of the track and rolling stock. The knowledge of be identified through the analysis and the study of
the history of the contact forces allows the devel- the structural properties of the rail. Regarding the
opment of mathematical models to predict the pro- measurement of the vertical force (Q), there are sev-
gressive degradation of the track and to improve eral well-developed technical solutions. For exam-
the maintenance planning. The measurement of ple, the Italian Infrastructure Manager (RFI, Rete
the wheel-rail contact forces is still required for the Ferroviaria Italiana) is experiencing and testing
homologation and the acceptance of the new vehi- the SMCV system (Accattatis et al. 2014) to moni-
cles. The design of rolling stocks in fact needs the tor the vertical load. The evaluation of the lateral
knowledge of the loads exchanged by the wheels force (Y) is, instead, a less developed field and there
and the rail during all types of the running con- are less known solutions. In the last decades, some
ditions (Malavasi, 2014). The European Standard authors have deal with these subjects, as Ahlbeck
EN14363 regulates the experimental tests for the & Harrison (1977, 1980, 1981) and Moreau (1987).
homologation. Recently, there have been some studies concern the
These considerations highlight the importance evaluation of lateral forces through different exper-
of the development of measurement methods of imental approaches. Milkovic et al. (2013) have
wheel-rail contact forces. In the last years, the way- tried to decouple vertical and lateral forces using a
side monitoring systems of track loads have indeed method based on independent component analysis

79

TIS2017_Book.indb 79 2/8/2017 8:01:35 PM

 of recorded strain signals on the rail surfaces. Del-
prete et al. (2009), Bracciali et al. (2001, 2004) and
Di Benedetto et al. (2010) have studied a sensor to
measure at the same time vertical and the lateral
contact forces applying transducers inside the rail
web close to the barycentre axes. These and other
techniques use strain gauges placed on the rail sur-
face, in order to measure shear or bending effects.
In this paper, we present an experimental method
to estimate the lateral contact force starting from
the measure of the strains on the rail foot surface.
The experimental measurement system (SMCT)
is composed by four shear strain gauges sym-
metrically arranged on the rail foot and connected
together through a full Wheatstone bridge. The full Figure 1. V-shaped strain gauges.
bridge configuration, with an opportune combina-
tion of the recorded strains, is able to reproduce
the same continuous signal of the applied lateral
force decoupling the effects of the vertical one. The
present research draws on the study of Yifan et al.
(2001) and the investigations made by the research
group in railway engineering of the Department
of Civil, Building and Environmental Engineer-
ing, Sapienza University of Rome (Bruner et al.
2015). These studies show a methodology to find
a constant ratio between the applied lateral load
and the recorded strains on the rail foot (Bruner
et al. 2016). They also highlight the robustness
of this experimental procedure, representing with
finite element simulations the not influence of the
boundary conditions (thermal gradient, inclina-
Figure 2. V-shaped strain gauges: measuring grids.
tion of the rail, status of the rail profile, stiffness
of fastenings and ballast) on the measurements.

2 MEASUREMENT SYSTEM

The monitoring system for lateral loads (SMCT)

consists of four V-shaped strain gauges, with two
measuring grids arranged at an angle of about 45°,
placed on the rail foot surface. Typical applica-
tions for these sensors include measurements of
shear stresses. The strain gauges are symmetrically
disposed on the rail foot and they are connected
together by a Wheatstone circuit.
Considering one-meter long rail segment, the
sensors are symmetrically located respect the
y-axis of the rail at 120 mm from each other and
at 20 mm from the outer edge of the rail (Fig. 1).
An opportune combination of the measuring grids Figure 3. Wheatstone bridge circuit configuration.
inside a full Wheatstone bridge circuit generates as
output a strain having the same signal of the lateral the output voltage, RA, RB, RC, RD, RA’, RB’, RC’ and
force (Y) decoupling the effects of the vertical one RD’ are the measuring grids of the four V-shaped
(Q). In fact, the symmetrical position of the meas- strain gauges.
uring grids removes the contribution of the vertical The following expressions show the relation of
force on the recorded strains (Fig. 2). the full Wheatstone bridge for this specific applica-
Figure 3 represents the full Wheatstone bridge tion, where f is the gauge factor of the sensor and
circuit configuration: VA is the supply voltage, VU is ε the recorded strain.

80

TIS2017_Book.indb 80 2/8/2017 8:01:35 PM

 urement system. One-meter long rail segment (type
VU 1 ⎛ ΔRAB ΔRA′ B ′ ΔRC ′D ′ ΔRCD ⎞
= ⋅ − + − (1) UIC60/60E1) was reproduced with a 3D model.
VA 4 ⎜⎝ RAB RA′ B ′ RC ′D ′ RCD ⎟⎠ The model was meshed using hexahedral elements
ΔR that changes their size along the rail longitudinal
= fε (2) axis.
R The smaller size of the volume mesh is 5 mm,
corresponding to the location of the four V-shaped
For a single V-shaped strain gauge, the imbal- strain gauges. The mesh was refined in this area in
ance is: order to improve the numerical solutions. Moreover,
in this rail section the hexahedral elements were
Δ AB ⎛ ΔRA ΔRB ⎞ ⎛ RA f ε A + RB f ε B ⎞ covered and glued with 2D shell elements (Fig. 4).
=⎜ ⎟⎠ = ⎜⎝ ⎟⎠
RAB ⎝ RAB RA RB This finite element modelling technique was used to
Rf ( A + B ) f compare the rail foot strains with the experimental
= = ( A + B) (3) results of the four V-shaped strain gauges.
2RR 2 Concerning structural constrain conditions, the
model was considered fixed between two sleepers
Considering the other strain gauges: (spaced 600 mm from each other). The bottom side
of the rail foot was connected to the ground with
ΔR
RA′ B ′ f
= ( A
A′ + B′ ) (4) spring-damper elements (mesh size 1 mm). These
RA′ B ′ 2 elements generate a vertical displacement of the rail
proportionally to the stiffness of elastic fastenings
ΔR
RC ′D ′ f
= ( C′ + D
D′ ) (5) and rail pad (300 kN/mm). Moreover, the rail foot
RC ′D ′ 2 surface was constrained for the cross displacement.
ΔR
RCD f Finally, in order to simulate a wheel-rail contact in
= ( C + D ) (6) curving, the vertical (Q) and lateral (Y) forces were
RCD 2 applied on the railhead in the mid-section of the
model, 20 mm from the z-axis and 14 mm under
Replacing the previous expressions into the the rolling plane (Fig. 5). The simulations were
equation (1): performed with a static analysis using ANSYS
Code. The static analysis has been chosen because
VU f
= ( A+ B − A − B + C + D − C − D ) of, at the beginning, the SMCT measurement sys-
VA 8 tem is design to work in areas where rolling stocks
f transit at very reduced speed (below 30 km/h), like
= εttot (7) stations, depots and marshalling yards. The finite
8
element model of the rail segment has 67,349 ele-
Because of the SMCT measurement system ments and 75,669.
produces as output a total strain (εtot) that has the
same signal of the applied lateral force (Y), it is 3.2 Results of simulations
possible find a constant K, define as follow:
Simulations results, which consider a distance (d)
Y between the strain gauges of 120 mm (Fig. 4), are
K= (8)
εttot

This constant allows to evaluate the lateral force

(Y) applied by the wheel on the rail in a generic
situation, starting from the measurement of the
rail foot strains (9).

Y K ⋅ εttot (9)

3 FINITE ELEMENT SIMULATION

3.1 Finite element model

The finite element simulations were performed to
check and validate the results of the SMCT meas- Figure 4. Finite element model.

81

TIS2017_Book.indb 81 2/8/2017 8:01:36 PM

 Y/Q ratio of 0.1 and ended with 1.4. The highest
value allowed by the European Standard EN 14363
is 1.2.
Other tests are carried out by changing the dis-
tance (d) between the V-shaped strain gauges. The
value of constant K changes modifying this dis-
tance (Fig. 6). The results prove that K depends
by the distance of strain gauges, remaining almost
constant. Increasing the distance (d), the absolute
value of the rail foot strains raises up and the con-
stant K decreases proportionally.
In order to evaluate the influence of the vertical
force on the strain measures, Figure 7 reports the
results of simulations considering applied on the
railhead only the vertical force (Q). Two different
schemes were considered: the first with the force
applied directly on the z-axis of the rail (Fig. 7,
solid line) and the second with the force applied at
a distance of 20 mm from the z-axis (Fig. 7, dot-
ted line).
The first condition shows the SMCT method
removing the effects of the vertical force from the
recorded strains combination. The recorded strain
(εtot) is about zero with any distance (d) of the strain
gauges. The second condition, typical scenario for
the vertical force in a generic wheel-rail contact,

Figure 5. Application of contact forces.

Table 1. Results of simulations (d = 120 mm).

Vertical Lateral Ratio Total Constant

Force Q Force Y Y/Q strain εtot K = Y/εtot
kN kN – με kN/με

100 10 0.1 40.5 0.247

100 20 0.2 88.0 0.227
100 30 0.3 135.5 0.221
100 40 0.4 183.0 0.219 Figure 6. Constant K with a different distance of strain
100 50 0.5 230.6 0.218 gauges.
100 60 0.6 278.1 0.216
100 70 0.7 325.6 0.215
100 80 0.8 373.1 0.214
100 90 0.9 420.7 0.214
100 100 1.0 468.2 0.214
100 110 1.1 515.7 0.213
100 120 1.2 563.2 0.213
100 130 1.3 610.7 0.213
100 140 1.4 658.3 0.213

reported in Table 1. The ratio between the lateral

force (Y) and the recorded strains (εtot) is almost
constant changing the load conditions. The aver- Figure 7. Influence of the vertical force on the strain
age value is 0.218 kN/με. Test runs started with a measures.

82

TIS2017_Book.indb 82 2/8/2017 8:01:38 PM

 Figure 8. Error curve for the evaluation of the lateral Figure 9. Wheelset and test bench.
force.

displays that when the distance (d) between the

strain gauges increases, the recorded strain (εtot)
increases too. This means that by increasing the
distance between strain gauges proportionally
increases the measurement error. Therefore, the
distance (d) should be chosen minimizing the
measurement error and maximizing the accuracy
of strain gauges measurement location.
Taking into account the above considerations,
we fixed the distance (d) equal to 120 mm and
assessed the percentage error (Fig. 8) for the evalu-
ation of the lateral force (Y). The diagram shows Figure 10. Technical scheme of the test bench.
that the percentage error grows exponentially for
low values of Y/Q, whereas for the others (≥0.3)
vertical reaction (c2) and another one for the lat-
remains below 4%.
eral force (cH). The rail vertical reaction (Q1) was
This trend depends by the low influence that
derived from the Q2. To properly evaluate the rail
the lateral force (Y) has on rail foot strains for low
reaction (Q1 and Y) it was considered the standard
values of Y/Q. In this case, the twist strain on the
inclination of the rail (1/20). In this conditions, the
rail foot, produced by the vertical force, is greater
structural system is isostatic: at point A, where the
than the one generated by the lateral force and the
wheel-rail contact point occurs, we could assume
SMCT measurement system could not properly
a theoretical revolute joint, whereas at point B a
decouple lateral and vertical forces.
roller joint.

4 EXPERIMENTAL TESTS 4.2 Experimental results

The results of laboratory tests are reported in
4.1 Test bench configuration
Figure 11. The chart represents a generic load
Experimental results were reached on the test cycle. The highest vertical and lateral load applied
bench (Fig. 9). As the finite element simulations, by the test bench is about 45 kN. For each load
one-meter long rail segment (type UIC60/60E1) curves, the signal is reported in function of time.
was used, fixed between two sleepers spaced of Moreover, in this chart the rail foot strain and the
600 mm from each other. value of constant K are shown together thanks to
The rail was connected to the support by four two different reference axes.
indirect elastic fastenings (type Vossloh), two for Analysing the results, the strain curve (εtot) has the
each sleeper. The distance (d) between the V-shaped same shape of the lateral load curve (H). This con-
strain gauges was set to 120 mm. firms the existence of an experimental constant (K),
The technical scheme of the test bench (Fig. 10) as done in the theoretical approach. Besides, when
shows that the vertical (Qb) and lateral loads (H) the lateral load (H) is applied, the vertical reaction
are applied to the rail segment by a wheelset. (Q2) decreases because of the isostatic equilibrium
These forces were recorded by four loads cell, two of the structure increases the rail reaction (Q1), tilt-
to measure the vertical force (cb1, cb2), one for the ing the wheelset. Also, at the beginning of the load

83

TIS2017_Book.indb 83 2/8/2017 8:01:39 PM

 Figure 13. Position of rail between the two fastenings.

Table 2. Positions of strain gauges.

Figure 11. Results of laboratory test.
Distance Distance Distance Constant
Rail d m n = (d/2) − m K = Y/εtot
position mm mm mm kN/με

P0 120 0 + 60 0.219
P20 120 + 20 + 40 0.219
P80 120 + 80 − 20 0.195
P140 120 + 140 − 80 0.182

leaving the application point of vertical and lateral

forces in the mid-section between the fastenings.
The load cycle (Fig. 11) was repeated for different
positions of the rail with the contact forces in the
area between the V-shaped strain gauges (Table 2:
Figure 12. Experimental and finite element error position P0 and P20) and with the contact forces
curve. externally (Table 2: position P80 and P140).
The results of laboratory test (Figs. 14–16) con-
cycle, there is no lateral load (H) and the strain curve firm the previous outcomes: the same behaviour
became negative (εtot): this is the effect of the action for the signal of the recorded strain (εtot) and the
of the vertical load (Qb) on the measure which rep- lateral load curve (H).
resents a part of the experimental error, as seen in However, in this case the value of constant K
the finite element simulations (Fig. 7). is different for the forces out of the area between
Considering the average value of constant K the V-shaped strain gauges (Table 2: position P80
(0.219 kN/με), which is quite the same value of and P140). This condition shows that it is not pos-
simulations, the experimental and the finite element sible to obtain the magnitude of the lateral force
error curve has the same trend. Unfortunately, the (Y) using only one value of constant K when con-
technical specifications of the test bench have lim- tact forces are applied out of the area between the
ited the Y/Q ratio to 0.65. strain gauges.
In order to investigate this experience, addi-
tional finite element simulations, based on the
4.3 Extension of results
previous experimental tests, were conducted. The
The structure of the test bench did not allow to contour plots of the rail foot surface (Figs. 17–20)
perform experimental tests by moving the vertical show the extension and compression areas for the
and lateral load along the rail. Despite this limita- rail positions P80 and P140 under a vertical (Q)
tion, in order to extend the results and determine and a lateral (Y) load of 100 kN (Y/Q = 1.0). The
the strain effects of contact forces on the entire extension and compression areas are calculated
length of the rail segment, the rail was shifted along the main axes of the two measuring grids of
through the two fastenings (Fig. 13). the four V-shaped strain gages (ξ and η).
The term d is still the distance of the strain When contact forces are completely external to
gauges, while the distance m is the traslation of the the strain gauges’ area (Figs. 19–20), the experi-
rail along the x-axis and the term n is the distance mental results (Table 2) have shown that the combi-
of the strain gauges from the application point of nation of the measuring grids produces a different
contact forces. Thanks to this procedure, the strains value of constant K. This effect occurs when the
were measured in different sections of the rail, strain gauges are placed in compressed rail section

84

TIS2017_Book.indb 84 2/8/2017 8:01:39 PM

 Figure 18. Contours plot of the rail foot surface along
η axis (Rail position P80).
Figure 14. Results of laboratory test (Rail position P20).

Figure 19. Contours plot of the rail foot surface along

ξ axis (Rail position P140).

Figure 15. Results of laboratory test (Rail position P80).

Figure 20. Contours plot of the rail foot surface along

η axis (Rail position P140).

instead of stretched and vice versa: there is no

longer the symmetrical arrangement of the meas-
uring grids like for the rail positions P0 and P20.
Figure 16. Results of laboratory test (Rail position P140). This situation occurs even if the forces are just
outside the area of the strain gauges (Figs. 17–18)
because of an uneven strains distribution.
This complementary finite element investiga-
tion confirms the experimental evidence: it is pos-
sible evaluate the lateral force (Y), starting from
the measurement of the rail foot strains (εtot) and
using the constant K, only when contact forces are
applied in the area between the strain gauges.

5 CONCLUSION

Figure 17. Contours plot of the rail foot surface along In this paper, it was presented an experimental
ξ axis (Rail position P80). method to estimate the wheel-rail lateral contact

85

TIS2017_Book.indb 85 2/8/2017 8:01:40 PM

 force starting from the measure of strains on the rail Ahlbeck, D.R. & Harrison, H.D. 1981. Development and
foot surface. The experimental measurement sys- evaluation of wayside wheel/rail load measurement
tem (SMCT) is composed by four V-shaped strain techniques. Proceedings of the International Conference
gauges, symmetrically arranged on the rail foot. on Wheel/Rail Load and Displacement Measurement
Techniques, 19–20 January 1981. Cambridge Mass.,
A suitable combination of the recorded strains USA.
allows to reproduce the same continuous signal of Bracciali, A. et al. 2001. Progetto e validazione di un
the applied lateral force (Y), decoupling the effects sensore estensimetrico multifunzione per il binario
of the vertical one (Q). The study, based on finite ferroviario. XXX Convegno AIAS, 12–15 September
element simulations and laboratory tests, show how 2001. Alghero, Italy.
to find a constant ratio (K) between the applied lat- Bracciali, A. & Folgarait, P. 2004. New Sensor for Lateral
eral load (Y) and the recorded strains (εtot) on the & Vertical Wheel-Rail Forces Measurements. Railway
rail foot surface. Thanks to this constant ratio and Engineering Conference, 6–7 July 2004. London,
the low influence of the vertical force, it is possi- England.
Bruner, M. et al. 2015. Tecniche di misura sperimentali
ble to evaluate the wheel-rail lateral contact force per la determinazione delle forze laterali di contatto
with a percentage error below 4% for values of Y/Q ruota-rotaia. IV Convegno Nazionale Sicurezza ed
greater than 0.3. Moreover, the experimental and Esercizio Ferroviario: Soluzioni e Strategie per lo Svi-
numerical investigations have shown the feasibility luppo del Trasporto Ferroviario, 2 October 2015. Roma,
of the measurement method only when wheel-rail Italy.
contact forces are applied externally to the strain Bruner, M. et al. 2016. The Rail Strain under Different
gauges’ area. Loads and Conditions as Source of Information for
In order to make operative a prototype of the Operation—Proceedings of the Third International
measurement systems, further experimental inves- Conference on Railway Technology: Research, Devel-
opment and Maintenance, J. Pombo, Civil-Comp Press,
tigations, both in laboratory and on a rail line, Stirlingshire, 5–8 April 2016. Cagliari, Italy.
have been scheduled. The first aim of these analy- Delprete, C. & Rosso, C. 2009. An easy instrument and
sis is to validate the experimental prototype of the a methodology for the monitoring and the diagnosis
SMCT measurement system applied on a longer of a rail-Mechanical System and Signal Processing 23:
rail. In conclusion, the research confirms that the 940–956.
rail can be used as a useful measurement system to Di Benedetto, L. et al. 2010. Diagnosi delle condizioni
ensure the safety of railways traffic and as a mean di marcia di un rotabile con un sistema di misura
for the development of new track maintenance delle forze laterali e verticali al contatto ruota-rotaia.
management. XXXIX Convegno AIAS, 7–9 September 2010.
Maratea, Italy.
Malavasi, G. 2014. Contact Forces and Running Stability
of Railway Vehicles. International Journal of Railway
REFERENCES Technology 3(1): 121–132.
Milkovic, D. et al. 2013. Wayside system for wheel–rail
Accattatis, F.M.D. et al. 2014. Measurement of the verti- contact forces measurements. Measurement, Journal
cal loads transferred to the rail. Ingegneria Ferroviaria of the International Measurement Confederation 46:
11: 1001–1041. 3308–3318.
Ahlbeck, D.R. & Harrison, H.D. 1977. Technique for Moreau, A. 1987. La verification del la securite contre
Measuring Wheel/Rail Forces with Trackside Instru- le deraillement. Revue Générale des Chemins de Fer 4:
mentation. ASME Winter Annual Meeting, 27 Novem- 25–32.
ber – 2 December 1977. Atlanta, USA. Yifan, L. et al. 2011. Wheel-Rail Lateral Force Con-
Ahlbeck, D.R. & Harrison, H.D. 1980. Techniques for tinuous Measurement Based on Rail Web Bending
Measurement of Wheel-Rail Forces. The Shock and Moment Difference Method. Applied Mechanics and
Vibration Digest 12. Materials 105–107: 755–759.

86

TIS2017_Book.indb 86 2/8/2017 8:01:41 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Design and maintenance of high-speed rail tracks: A comparison between

ballasted and ballast-less solutions based on life cycle cost analysis

M. Giunta & F.G. Praticò

University Mediterranea, Reggio Calabria, Italy

ABSTRACT: The increase of train speed and axle load in the European rail network is an essential
goal to make the railway transport more and more competitive for passengers and freights. High speed
trains call for a better structural and geometrical stability of the track. To this aim it is crucial to apply
innovative track design and materials. In the last decades, various types of ballast-less track systems have
been developed and put in service around the world. These systems seem to perform better than ballasted
solutions especially when high-speed passenger trains share the track with freight trains. The main advan-
tages of innovative slab systems are the following: low maintenance needs/costs, higher availability,
increased service life (50–60 years), higher lateral stability, reduction of weight and height of the track,
easier and more economic vegetation control. Weaknesses of slab tracks against ballasted tracks are as
follows: higher construction cost, higher noise radiation. In the light of the above considerations, in the
study presented in this paper a life cycle cost assessment of two competing track solutions (ballasted
and ballast-less) has been carried out, considering short- and long-term perspectives. Cost analysis uses
the present value of agency (construction, inspection, maintenance and renewal), environmental (CO2
emission), and user (delays-related etc.) costs. The analysis of the trend of agency, user, and externality
costs of the alternatives over the entire life cycle of the infrastructure allows recognizing the most sustain-
able option. Results show that solutions that are more affordable in the short term can yield maintenance
and renewal processes which are unfavorable or less sustainable in the long term. Furthermore, in the long
term, the difference between the (total) present values of the two solutions becomes too small to yield
sound conclusions in favor of the ballast-less solution with respect to the ballasted one.

1 INTRODUCTION Ballast-less tracks were developed to achieve the

following objectives:
1.1 Background
• Increase speed;
Track design and maintenance are nowadays • Increase capacity;
affected by the increasing axle loads and train • Reduce the number of track maintenance opera-
speed in railway lines. Higher loads and speed tions and thereby the costs for maintenance.
require greater structural and geometrical stabil-
ity of the track. It appears crucial to withstand the Many types of ballast-less systems have been
stresses in the rails, fasteners, sleepers/slabs, ballast developed in the last years: Shinkansen, Rheda,
and subgrade due to: i) the static mass of the vehi- Sonneville-LVT, Züblin, Stedef and Infundo-
cles; ii) the dynamic actions, such as lateral cen- Edilon, IPA.
trifugal forces on curves, longitudinal acceleration In general, ballast-less track is a continuous slab
and braking forces; iii) vertical inertial forces from of concrete in which the rails are usually supported
the motion of the wheel-set and its suspension; directly on the upper surface (of the slab) by using
iv) the vibrational forces induced from imperfec- resilient pads.
tions in the rail surface (corrugations, joints, welds, Ballast-less track exhibits the following benefits
defects) and in the wheels (flats and shells); v) the in comparison with the traditional ballasted track:
dynamic response of the track components to the low maintenance costs (approximately 20–30%
above actions (Tzanakakis, 2013). less), higher availability, increased service life
Based on the abovementioned considerations, (50–60 years), higher lateral stability, reduction
accuracy in design, application of enhanced main- of weight and height of the track, and easier and
tenance concepts and new or improved construc- cheaper vegetation control.
tion methods are needed (Esveld, 1999; Esveld, On the other hand, the main weaknesses of slab
2001; Gautier, 2015). tracks are: higher construction cost, higher noise

87

TIS2017_Book.indb 87 2/8/2017 8:01:43 PM

 radiation due to the lack of noise absorption of the WLC include a wider range of externalities or
ballast bed (Bilow & Randich, 2000; Esveld, 2001; non construction costs, such as finance costs,
Darr & Fiebig, 2006; Lichtberger, 2005). The miti- business costs and income streams.
gation of noise and vibration may further increase
Best practice LCCA calls for including not only
the costs of the slab track construction (Di Mino,
direct agency expenditures (for example, construc-
et al. 2009).
tion or maintenance activities) but also user costs.
In the light of the above considerations, an
User costs are costs to the public resulting from
analysis based on long time frames is needed. The
work zone activities, including lost time and vehi-
Life Cycle Cost (LCC) Analysis is an appropriate
cle expenses. Another important class of costs to
tool to evaluate the suitability of different track
consider is the externality costs, namely the costs
solutions considering the level of traffic in the line,
of the environmental impact. Usually, these costs
the maximum speed allowed and other boundary
are not considered in the LCC, because their esti-
conditions.
mate is very difficult. However it is very important
to take into account also these intangible costs
1.2 Objective of the work because they can significantly affect the total cost
(Stalder, 2001; Zoeteman, 2001, Lee et al., 2008).
The objective of the study described in this paper is to
The model proposed in this work, includes this
set up a LCCA-based method to compare two com-
class of costs.
peting rail track solutions (ballasted and ballast-less).
LCCA Methodology include the following main
Short-term and long-terms perspectives are consid-
steps:
ered. An LCCA-based model able to evaluate the
total cost of competing track solutions is proposed • Establish alternative design strategies.
and applied. By means of this model, traditional • Determine activity timing.
tracks (made up of ballast, sleepers, fastenings, base- • Estimate agency costs.
plates, rail) and innovative tracks (pre-stressed con- • Estimate user costs
crete slabs, fastenings, pad and rails) are compared. • Estimate externality costs
The present value of agency (construction, • Determine life-cycle cost.
inspection, maintenance and renewal), environmen-
tal (CO2 emission), and user (delays-related etc.) The implementation of the LCCA has involved
costs is considered. two main phases:
• The Life Cycle Inventory (LCI) refers to data
collection. It is a portion of LCCA. It entails the
2 METHODOLOGY detailed tracking of all the flows in and out of the
system. Raw resources or materials, construction
2.1 Basic concepts of LCCA processes, energy, emissions to air (CO2e), unit
According to the ISO 15686-5, “Life Cycle Cost- costs are included. It is crucial to take into account
ing (LCC) is a valuable technique which is used for that if the scope is too wide then the LCI may
predicting and assessing the cost performance of become quite useless to help in decision-making.
constructed assets”. On the contrary, if the scope is too narrow then
LCC includes construction, operation, mainte- the results may be biased and the output partisan;
nance, end of life, Environment (including energy • The Life Cycle Cost Model (LCCM), which
and utilities) costs (ISO 15686-5). focuses on cost estimation and discounting, and
The following definitions apply: allows determining the total cost pertaining to a
given solution.
i. External Costs: costs associated with an asset
which are not necessarily reflected in the trans-
action costs between provider and consumer 2.2 The proposed LCCA model
(e.g. business staffing and productivity, user
A model to evaluate the suitability in long-term
costs, etc). Collectively these elements are
perspective of different alternatives for high speed
referred to as externalities.
rail tracks has been set up (Praticò & Giunta 2016,
ii. Environmental cost impacts: …costs (or savings
Praticò & Giunta 2016a). Three main classes of cost
via rebates) to LCC depending on the effects.
have been associated to a given track solution:
on the environment. Examples could include
cost premiums for the use of non-renewable − the agency costs (Cag), which refer to the expen-
resources or for green house gas emissions. ditures for construction, maintenance and reha-
Where these costs are external to the constructed bilitation. Consequently, these costs include the
asset they will form part of a WLC analysis. initial cost and the running costs. The latter is
iii. Whole Life Cost (WLC) elements. Typically affected by the service life of the track structure
the difference to LCC is that the elements of and its components;

88

TIS2017_Book.indb 88 2/8/2017 8:01:43 PM

 − the user costs (Cus), mainly related to the delays, Consequently, in a long term perspective, the
in operational phase, caused by the maintenance equilibrium between the sum of agency and user
(slowdowns) and/or rehabilitation (re-routing) costs (tangible costs) and externality cost (intangi-
works. ble costs) is crucial. To this aim a balancing factor
− the externality costs (Cext) connected to the (ν) has been defined as the minimum ratio of tan-
impacts on the environment (resources deple- gible to intangible costs (see equation 7).
tion, noise, air pollution, etc.) due to the In order to make all the discussed and defined
construction/maintenance/renewal processes. costs comparable, they are discounted to a base
year considering the interest (r) and inflation (i)
It follows: rates. The present values (PV) for the quoted costs
can be calculated by means of the equations (4)
Cag Ccons + Cmain Crenw (1)
to (6):
Emain
i
where Ccons is the cost for track construction Cmain ⎛ 1+ i ⎞
the costs for maintenance and Crenw the costs of Vag = Ccons
PV ∑Cmain ⋅ ⎜⎝ 1 + r ⎟⎠
renewal. Erenw
⎛ 1+ i ⎞
Regarding the user costs, they can be calculated + ∑ Crenw ⋅ ⎜ ⎟ (4)
starting from the costs of delays for maintenance ⎝1+ r⎠
and renewals. According to Lovett et al., 2015, user Emain
i
⎛ 1+ i ⎞
costs can be calculated as: Vus = ∑ CD
PV Dmain ⋅ ⎜
⎝ 1 + r ⎟⎠
Erenw
Cus CDmain CDrenw
CD ⎛ 1+ i ⎞
= ((CO + CF ) ⋅ NL + CW ) ⋅ TD
Dmain + ∑ CD
Drenw ⋅ ⎜ ⎟ (5)
⎝1+ r⎠
+ ((CO + CF ) ⋅ NL + CW ) ⋅ TDrenw (2)
Ek
⎛ 1+ i ⎞
where, CDmain and CDrenw are the costs (euro/train- Vext = CEX
PV CEX ∑k CCEX
EX k ⋅ ⎜
⎝ 1 + r ⎟⎠
(6)
hour) of delays caused by maintenance and renewal
activities respectively, CO is the locomotive operat- where PVag is the present value agency costs, PVus
ing cost (euro/locomotive-hour), CF is the fuel cost is the present value user costs, PVext is the present
(euro/locomotive-hour), NL is the average number value externality costs, i is the inflation rate, r is the
of locomotives, CW is the crew cost (euro/train- interest rate, CEX0 and CEXk refer to the external-
hour), TDmain is the average length (hours) of delays ity costs at construction stage and at k-th mainte-
for maintenance activity and TDrenw is the average nance/rehabilitation phase respectively, Emain is the
length (hours) of delays for renewal activity. expected life for maintenance activity, Erenw is the
The externality costs refer to environmental expected life for renewal, Ek is the expected life for
impacts produced during the construction and the the k-th activity of maintenance or renewal.
life cycle of the track (CEXkj). Each j-th impact The total present value (TPV) of a given alter-
produced by the k-th process (Qkj) can be associ- native is the sum of the present values of all costs
ated to a unit cost (UPkj). Having in mind the sym- related to them, as already defined. In order to take
bols already defined, the externality costs can be into account the issue related to the CO2 cost fluc-
calculated by means of this equation: tuation, which affects the externality costs, a bal-
ancing factor has been defined as follows:
Cext ∑ ∑CEX
K
C
j
kj =
k
∑Q
j
kj UPkj
UP (3)
PVaag j + PVuc
u j
ν = min (7)
The quantification of these costs is a difficult j = , ,.. k PVeext j
task; in this work the quantity of CO2 equivalent
corresponding to the given process and material where j is the j-th alternative and k is the number of
has been considered. the alternatives in comparison.
For a given mixture and amount of greenhouse Having in mind this factor, the TPV for a given
gas, the carbon dioxide equivalent refers to the j-th alternative can be derived from the equation:
amount of CO2 that would have the same Global
Warming Potential (GWP) when measured over a V j = PV
Vag j + PV
TPV uc j +
Vuc ⋅ PV
Vextjt (8)
specified timescale (generally, 100 years). Regard-
ing the cost of a ton of CO2e, it should be noted
that it is extremely variable (Ian et al. 2009, Praticò This allows determining the differential (gain)
et al, 2011) and this variability is a critical factor between two competing solutions:
because it may imply different incidence of exter-
nality costs on total cost. Gtot = TPV
VS − TPV
VB (9)

89

TIS2017_Book.indb 89 2/8/2017 8:01:43 PM

 where TPVS and TPVB are the discounted total reinforced pre-stressed concrete slabs measuring
cost pertaining respectively to ballasted and slab 4.93 m × 2.34 m × 0.19 m (4.95 m × 2.34 m × 0.16 m
tracks. in tunnels) and asphalt cement mortar injected
In this way the gain produced by one alternative under and between the slabs. The slabs weigh
(i.e. Ballast) with respect another (i.e. Slab) can be approximately 5 tonnes (Esveld, 2001).
evaluated in a short-term or/and long-term time In order to estimate the costs, according to the
frame. formulas set up in the previous section, a prelimi-
nary inventory phase has been carried out. The
data gathered to estimate each cost are shown in
3 APPLICATION AND RESULTS Table 2.
The sources of data have been Italian executed
3.1 Data acquisition and ongoing projects and the scientific literature.
It should be noted that maintenance encom-
The proposed LCCA-based model was applied to
passes all those minor activities aiming at repairing
two track systems: a traditional ballasted track and
(corrective maintenance) or preventing (preventive
the Shinkansen type (Figure 1). The main charac-
maintenance) rails and sleepers damage, including
teristics of the track components and their average
tamping, track stabilization, ballast injection.
expected service life (Milford & Allwood, 2010)
Maintenance costs are affected by the traf-
are reported in Table 1. A double track line with a
fic (typically Millions of Gross Tons, MGT) and
length of 1000 m has been considered.
speed: the higher the traffic the higher the main-
For ballasted track, the main components are:
tenance costs. Furthermore, higher speeds corre-
rails, sleepers, fastenings, ballast and sub-ballast.
spond to higher maintenance costs (Baumgartner,
The Shinkansen slab track consists of a sublayer
2001; Silavong et al., 2014; Thompson, 1986).
stabilized using cement, cylindrical “stoppers”
The maintenance costs in this work have been
to prevent lateral and longitudinal movement,
considered as annually running (Jimenez-Redondo

Table 2. Life cycle inventory.

Costs Inventory

Ccons Components track

Quantities
Unit Prices
Construction processes and costs
Cmain Traffic (Million of Goss Tons)
Speed
Service life of components
Scheduling of maintenance
Materials and components
Unit price
Maintenance processes and costs
Crenw Traffic (Million of Goss Tons)
Speed
Service life of track
Figure 1. Slab track [Source: Esveld 2001].
Scheduling of renewal
Components track
Table 1. Track components and expected service life. Unit price
Disposal process and costs
Service life Reconstruction processes and costs
Component [Years] Cus Length of maintenance renewal activities
Operation costs of train
Rail—60 UNI 28 Train composition
Sleepers—Pre-stressed mono-block 40 Fuel cost
Fastenings—Elastic type Vossloh W14 AV 40 Crew cost
Ballast—Crushed stones, 500 mm. depth 40 Length of delays
Subballast/Concrete road-bed—Cement 40 Cext Materials
treated layer 200 mm depth Processes
Slab—Pre-stressed concrete 60 CO2e emissions
with cylindrical bollard Cost of a ton of CO2

90

TIS2017_Book.indb 90 2/8/2017 8:01:44 PM

 et al., 2012) and have been calculated according
to the following formula, which was set up and
calibrated on the basis of data gathered from the
Italian industry:

(V −100)
⎡
Cmain = ⎢2.2 ⋅
(V − ) + 4⎤ ⋅ GTK
T ⎣
⎡
⎢ −0.05⋅
200
⎤
+ 0.63⎥
⎦ (10)
⎥
⎣ 200 ⎦

where GTK stands for Gross Tons × Kilometer,

and V [Km/h] stands for speed.
Regarding the user costs, having in mind equa-
tion 2, the average delay for maintenance and
renewal has been derived on the basis of operat-
ing costs (locomotive operating costs, crew cost for
train, fuel/energy cost). Data were gathered from
RFI (Italian Agency Owner of Railway network)
and Trenitalia.
On the other side, the externality costs have been Figure 3. Trends of PVs for slab track.
estimated considering the CO2e emission associated
to material production and processes (Milford &
Allwood, 2010) and considering the unit price of
the CO2e.

3.2 Results of the application

The results of the study are reported in the follow-
ing Figures 2–7.
Figures 2 and 3 illustrate the value the initial
costs (construction), running costs (maintenance/
renewal) user costs, and externality costs for the
two track alternatives under comparison, and for
a time frame of 0, 30, 60 and 120 years.
The comparison of the two figures highlights
the higher construction and externality costs of
the slab track at year 0 (construction stage) with
respect to the ballast track. For the construc-
tion costs, the results comply with the literature Figure 4. Trend of the PVs of agency costs.

Figure 2. Trends of PVs for ballasted track. Figure 5. Trend of differential of PVs user costs.

91

TIS2017_Book.indb 91 2/8/2017 8:01:45 PM

 As for the gain of the agency costs (Figure 4), it
can be observed that in the initial period the bal-
lasted track alternative appears more convenient
than the slab track.
However, the gain of this alternative decreases
over time and reaches the value 0 after 16 years.
The discontinuities along the curve are due to
the concentration of expenditures for the given
solution (maintenance and/or renewal).
The gain of user costs (Figure 5) is negative dur-
ing the reference period. This indicates the higher
appropriateness of slab tracks with respect to bal-
lasted tracks. This derives from the appreciable
availability and service life of slab tracks.
Externality costs follow a reverse trend. In this
case, the ballasted solution appears more “environ-
Figure 6. Trend of the PVs of externality costs. mental friendly”. Note that the embedded carbon
of a slab track is 0.3–0.4 Kg CO2/Kg. In contrast,
for ballast and sleepers it is 0.005 Kg CO2/Kg and
0.27–0.28 Kg CO2/Kg, respectively (Milford, 2010,
Praticò et al. 2013).
By referring to the total cost of the two solutions
(Figure 7), the tendency of the gain may appear sim-
ilar to the one observed for agency costs (Figure 4).
Anyhow, some differences have to be considered.
In the short term the ballasted track is eco-
nomically convenient, but this advantage becomes
smaller and smaller over time.
After 36 years the break-even point is achieved.
In the long term, due to the negative value of the
gain, the slab solution emerges as convenient choice,
even if the gain is somewhat small and constant.
A sensitivity analysis has been carried out in
order to evaluate how the variability of the user and
Figure 7. Trend of differential of PVs total cost. externality costs can affect the results. It has been
found that the higher the user costs, the higher the
(Schilder & Diederich, 2007; Pichler & Fenske, difference of the present value (i.e. gain) between
2013; Gautier, 2015). High externality costs are the ballasted and the slab track. User costs affect
due to the great carbon footprint associated to the also the externality cost EX’ due to the facts that
production of the cement. different values of user costs result in different val-
Conversely, the slab track exhibits lower main- ues of the parameter ν (see eq. 7). By referring to
tenance/renewal and user costs, due to the higher the environmental costs, it is noted that thanks to
service life of the entire structure and of its com- equation 7, variation in CO2e cost do not impact
ponents (i.e. concrete slab). EX’ value. For example if CO2e cost is 11 €/ton it
The trend of the user costs follows the one of turns out EX’ = 1.6 M€ (ballasted) or 1.9 M€ (slab).
running costs because the user costs, as already In contrast if Co2e cost is multiplied by ten it still
explained, depend on the delays due to the activi- results: EX’ = 1.6 M€ (ballasted) or 1.9 M€ (slab).
ties carried out on the track during the service life.
The following Figures 4–7 illustrate the trend
over time (until 120 years) of the gain, evaluated 4 CONCLUSIONS
for each considered cost.
The gain is calculated as the difference between In the present paper the comparison between bal-
the discounted costs (Present Value, PV) pertain- lasted and ballast-less track solutions has been car-
ing to the Slab track (S) and the one pertaining to ried out based on Life Cycle Cost Analysis.
the Ballasted track (B). A method to evaluate the costs of the two alter-
With this assumption, a positive value of the natives has been proposed and applied.
gain indicates that the ballasted track is preferable The method aims at offering to practition-
to the slab track and vice versa. ers and researches a tool for evaluating the best

92

TIS2017_Book.indb 92 2/8/2017 8:01:46 PM

 strategy for the design and management of a rail- Jimenez-Redondo, J., N. Bosso, N., Zeni, L., Minardo,
way track over time (maintenance and renewal). A., Schubert, F., Heinicke, F., and Simroth, A., 2012.
Long-term perspective, tangible and intangible “Automated and cost effective maintenance for railway
costs are considered. (acem-rail)”. Transport Research Arena 48:1058–1067.
doi: 10.1016/j.sbspro.2012.06.1082.
For each solution, technical features, traffic, Lee, C.K., Lee, J.Y. & Kim, Y.K., 2008. Comparison of
speed of the railway line, costs of construction, environmental loads with rail track systems using sim-
maintenance and renewal, environmental impacts plified life cycle assessment (LCA). WIT Transactions
produced by the construction/maintenance/renewal on the Built Environment 101:367–372.
activities, greenhouse emissions and related costs Lichtberger B. Track compendium. First edition, Eurail
have been considered. Press. 2005.
The trend of agency, user, and externality costs Lovett, A.H., C.T. Dick, C.J. Ruppert, Jr. & C.P.L. Bar-
of the alternatives over the entire life cycle of the kan, (2015). Cost and delay of railroad timber and
infrastructure allows recognizing the most sustain- concrete crosstie maintenance and replacement. Trans-
portation Research Record: Journal of the Transporta-
able solution from different standpoints. tion Research Board Vol. 2476, pp. 37–44.
Based on the data gathered, it can be reasonably Milford R.L. Allwood, J.M, 2010. Assessing the CO2
observed what follows: impact of current and future rail track in the UK.
− solutions that are more affordable in the short Transportation Research Part D 15: 61–72.
Pichler, D.; Fenske, J., 2013. Ballastless track systems
time can yield maintenance and renewal proc- experiences gained in Austria and Germany. Proc. of
esses which are unfavorable or less sustainable in AREMA Annual Conference, Indiana Convention
the long term; Center Indianapolis.
− the externality costs play an important role in Praticò, F.G.; Vaiana, R.; Giunta, M.; Iuele, T.; Moro, A.,
the evaluation of the two alternatives; 2013. Recycling PEMs back to TLPAs: Is that possible
− in the long term, despite the change of sign notwithstanding RAP variability? Applied Mechanics
in total gains, the difference between the total and Materials Vols. 253–255 (2013) pp 376–384.
present values appears quite negligible; Praticò, F.G.; Giunta, M., 2016. Assessing the sustainability of
− the above observations imply that further analy- design and maintenance strategies for rail track by means
life cycle cost analysis. In Proceedings of COMPRAIL
ses and studies are needed to yield rigorous con- 2016 15th International Conference on Railway Engineering
clusions in favor of the ballast-less solution with Design and Operation, July 19–21 Madrid, Spain.
respect to the ballasted one or vice versa. Praticò, F.G.; Giunta, M., 2016a. Issues and perspec-
tives in railway management from a sustainability
standpoint. In Proceedings of International Conference
REFERENCES on Transportation Infrastructure and Materials, July
16–18, Xian, China.
Baumgartner, J.P., 2001. Prices and costs in the railway Praticò F.G., Vaiana R., and Giunta M., Sustainable
sector. École Polytechnique fédérale de Lausanne. rehabilitation of porous European mixes, ICSDC
Laboratoire d’Intermodalité de Transports et de 2011: Integrating Sustainability Practices in the Con-
Planification. struction Industry—Proceedings of the International
Bilow, D.N., Randich, P.E., 2000. Slab track for the next Conference on Sustainable Design and Construction
100 years. Portland Cement Association. Skokie, IL. 2011, pp. 535–541.
Darr E. & Fiebig W. 2006. Feste Fahrbahn: Konstrktion Schilder, R.; Diederich, D., 2007. Installation Quality of
und Bauarten für Eisenbahn und Strassenbahn, Second Slab Track—A Decisive Factor for Maintenance. 2007
edition. Eurailpress. Germany. RTR Special 76–78.
Di Mino G., Giunta M. & Di Liberto M.C., 2009. Assess- Silavong, C.; Guiraud, L.; Brunel, J., 2014. Estimating
ing the open trenches in screening railway ground- the marginal cost of operation and maintenance for
borne vibrations by means of artificial neural network. French railway network. Proc. ITEA Conference–
Int. Journ. Advances on Acoustics and Vibration, Article Toulouse, France.
ID 942787, 12 pages ISSN: 1687–6261. Stalder, O., 2001. The life cycle costs (LCC) of entire rail
Esveld C., Slab track: A Competitive Solution, 1999. networks: An international comparison, Rail Inter-
Esveld C., Modern Railway Track, Second Edition. Delft national, International Railway Congress Association,
University of Technology 2001. Vol. 32, No. 4, pp. 26–31, ISSN 0020-8442.
Gautier, P.E., 2015. Slab track: Review of existing systems Thompson, L.S., 1986. High-Speed Rail. Technology
and optimization potentials including very high speed. Review, v. 89, pp: 32–43, 70.
Construction and Building Materials 92: 9–15. Tzanakakis K. 2013. The Railway Track and Its Long Term
Ian, W.H., Parry, I.W. H & Timilsina, G.R., 2009. Pricing Behaviour. A Handbook for a Railway Track of High
Externalities from Passenger Transportation in Mexico Quality, Volume 2 of the series 2013 Springer Tracts on
City, Policy Research Working Paper 5071, The World Transportation and Traffic pp 279–292.
Bank, Development Research Group. Environment Zoeteman, A., 2001. Life cycle cost analysis for manag-
and Energy Team. ing rail infrastructure: Concept of a decision support
ISO 15686–5:2008 Buildings and constructed assets— system for railway design and maintenance EJTIR, 1,
Service-life planning—Part 5: Life-cycle costing. no. 4 pp. 391–413.

93

TIS2017_Book.indb 93 2/8/2017 8:01:46 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Instability exposure and risk assessment of strategic road corridors

in a geomorphologically complex territory

R. Pellicani, I. Argentiero & G. Spilotro

Department of European and Mediterranean Cultures, University of Basilicata, Matera, Italy

ABSTRACT: Risk and exposure of the road corridors of Matera Province (Basilicata Region, Southern
Italy) to landslide phenomena was assessed. The provincial road network (1,320 km length) represents the
main connection network among thirty-one urban centers due to the lack of an efficient integrated (road,
railway and aerial) transportation system through the whole regional territory. The strategic importance
of these roads consists in their uniqueness in connecting every urban center with the socio-economic
surrounding context. The exposure was evaluated in terms of amount of traffic, as a function of
population of each centers. The vulnerability was assessed in function of the presence of criticalities along
roads. The exposure and vulnerability to landslides were combined in order to evaluate and map the risk.
The classification of the road sections in terms of risk levels represent a support for decision making and
allows to identify the priorities for designing appropriate landslide mitigation plans.

1 INTRODUCTION ing mainly steep and rocky slopes. For these rea-
sons, landslides have greater impact on the fixed
Landslides affecting transportation corridors can element at risk (road and traffic). Nevertheless, the
cause direct and indirect consequences, respec- assessment of landslide risk along road corridors is
tively, in terms of traffic disruption and impact poorly treated in literature.
with vehicles. In this paper, a procedure for assessing and map-
In general, quantifying, in mathematical terms, ping the landslide exposure and risk along the road
the landslide risk can be very complicated, due to corridors of Matera Province (Basilicata region,
several aspects, related to the complexity in assess- Southern Italy) is presented.
ing the temporal probability of a specific landslide The road exposure was evaluated consider-
event with given intensity (hazard) and the prob- ing the amount of vehicular traffic on each road
ability of damaging a given element at risk, i.e. stretch. This estimation was carried out by using
vulnerability (Glade 2003, Pellicani et al. 2014a). population data and by ranking in different orders
Assessing the landslide risk with regards to a the roads connecting several urban areas (nodes)
mobile elements at risk can be even more difficult according to the type of connection and number
(Pellicani et al. 2016). of linked nodes. Subsequently, the assessment of
In the recent literature, quantitative risk assess- landslide risk was carried out by using a qualita-
ment procedures, attempted to estimate the risk in tive matrix approach. This procedure consists in
terms of annual probability of direct impact, in overlaying the consequences and hazard maps
terms of life loss of occupants of a vehicles, espe- and by combining in a matrix the relative classes.
cially with regards to rockfall phenomena, have been The consequences were derived by combining the
developed (Corominas et al. 2005; Pantelidis 2011; vulnerability and exposure maps; while the hazard
Ferlisi et al. 2012; Budetta et al. 2015; Nicolet et al. was evaluated in function of susceptibility and
2016). Among the different typologies of landslide landslide intensity, depending on size and velocity
(Varnes 1984), rockfalls are generally characterized of instability phenomena.
by small size, but by relatively high magnitude due
to the high falling velocity of blocks and accord-
ingly by greater impact and damaging potential on 2 MATERA PROVINCIAL ROAD
elements at risk, especially those mobile (vehicles). NETWORK
Other typologies of landslides, such as flows and
slides, affecting road corridors are more wide- The road corridors of the Matera Province, in Basil-
spread on the territory than rockfalls, as they affect icata region (Southern Italy), extend for 1,324 km,
hillslopes with small slope angle and characterized connecting 31 municipalities (Figure 1). This road
by different lithologies, unlike the rockfalls affect- network exerts an important role for the entire

95

TIS2017_Book.indb 95 2/8/2017 8:01:48 PM

 transportation system of the provincial territory,
of which represents the main connection network
between among the urban centers, due to the lack
of other type of effective infrastructures (railway,
motorway, aerial, etc.). The strategic importance
of these roads consists in their uniqueness in con-
necting every urban center with the socio-economic
surrounding context. Consequently, without these
roads, the urban centers would be isolated.
These road corridors and their relative vehicular
traffic are continuously exposed to landslide proc-
esses (Figure 2a). In the last years, about the 44%
(584 km) of the total length of roads was affected
by damages and, subsequently, by reparation works
(Figure 2b). These landslide phenomena are char-
acterized both by high intensity and low frequency
and by low intensity and high frequency. This last
typology is particularly hazardous for the roads
since it is widely distributed along the transporta-
Figure 1. Study area: Matera provincial road network tion network and its occurrence (depending by the
(Basilicata region, Southern Italy). return time) is connected to rainfall events.

3 METHODOLOGY: FROM EXPOSURE

TO RISK

Landslide risk was defined by Varnes (1984) as the

expected number of lives lost, persons injured, dam-
age to property, or disruption of economic activity
due to a particular damaging natural phenomenon for
a given area and period of time. For a given category
of elements at risk, the specific risk can be quanti-
fied as the product of hazard, i.e. the probability
of occurrence of a specific hazard scenario with a
given return period in a given area, vulnerability,
intended as degree of loss to element at risk, and
exposure, in terms of amount or economic value
of elements at risk (Van Westen et al. 2006). In the
following, the method for assessing landslide risk
along the road network is explained, focusing on
the procedure for exposure estimation.

3.1 Road exposure estimation

The exposure of the road corridors was evaluated
in terms of amount of potential traffic. This choice
was derived from the assumption that the economic
value and the reconstruction costs of the entire
road network are constant (exposure in monetary
terms) and performing a probabilistic analysis of
vehicles and passengers distribution along the road
network is not possible, due to the lack of detailed
input data in relation to the analysis scale.
A relatively simple model to calculate traffic vol-
ume on road segments (links) connecting several
Figure 2. Landslides affecting provincial roads (a) and towns (nodes) is the Gravity model (McNally 2007,
road segments subjected to landslide damages and repa- Hong & Jung 2016). It is based on Newton’s Law
ration works. and assumes that the trips produced at an origin

96

TIS2017_Book.indb 96 2/8/2017 8:01:48 PM

 and attracted to a destination are directly propor- The qualitative estimation of potential traffic
tional to the total trip productions at the origin and volume on these three types of roads was carried
the total attractions at the destination. In particular, out by calculating, for the 1st order roads, the arith-
the vehicular flow is assessed by means of a Four metic average of population of the interconnected
Step model, consisting in trip generation (depend- urban centers, for the 2nd order roads, the average
ing on the activity system, represented by socio- of traffic values of the connected 1st order roads,
economic and demographic data), trip distribution, and, finally, for the 3rd order road, the average of
mode choice and route choice (depending on the traffic values of the connected 2nd order roads.
transportation system, represented by road network
typology and characteristics). The Gravity model
3.2 Risk assessment along road network
application is not straight forward because depends
on several factors complex to be determined, The spatial distribution of landslide risk along the
unless using survey data, such as total number of road corridors of Matera Province was assessed
households and employees, trip purposes, traveler and mapped using a qualitative matrix approach
behavior, etc. Another factors is a calibrating term (Chowdhury and Flentjie 2003, AGS 2007), in
representing the reluctance or impedance of persons which risk is obtained by combining in a two-
to make trips of various duration or distances. dimensional table or matrix a set of hazard catego-
For all these reasons it was choose to to priori- ries with a set of consequence categories (Figure 3).
tize and rank the links from busiest to least busy, The first matrix regards the landslide hazard, i.e.
without necessarily knowing the exact daily traffic the probability of a landslide of certain magnitude
volume. Therefore, the traffic along each road sec- to occur, is a function of the return time, which
tion, connecting two or more towns, was qualita- depends on landslide typology and intensity. For
tively assessed as a function of population of each a given type of landslide mechanism, the inten-
of them. The road classification was carried out as sity changes according to the areal extension and
following: velocity of landslide. Hazard matrix was obtained
qualitatively through the following steps:
• first order road: direct link connecting only two
towns; • Assessing the velocity range of landslide phe-
• second order road: link connecting first order nomena, by associating a velocity category to
road segments; each type of landslide mechanism.
• third order road: link connecting second order • Evaluating the areal extension of landslide phe-
road segments. nomena and subdivision into four classes.

Figure 3. Flow chart synthesizing the procedure for landslide risk assessment.

97

TIS2017_Book.indb 97 2/8/2017 8:01:50 PM

 • Determining the landslide intensity by combin- (E1 = 0.1–0.25), medium (E2 = 0.25–0.4), moderate
ing in a matrix the velocities and the areas with (E3 = 0.4–0.5) and high (E4 = 0.5–1). The conse-
relative classes. quences map was obtained by overlying in GIS
• Assessing the landslide susceptibility of the the vulnerability map and exposure map and by
entire Matera provincial area through a polyno- combining in a matrix the five vulnerability classes
mial heuristic-bivariate statitical model (Pellicani with the five exposure classes and ranking the com-
et al. 2014b). A set of thematic maps related to binations into five classes: insignificant, minor,
the predisposing factors was prepared and cor- medium, major and catastrophic. Finally, the risk
related with the landslide inventory map in order matrix was produced by combining among them
to obtain through a bivariate procedure (Van the hazard and consequence classes and by associ-
Westen 1993) the weights representing their ating to each combination the following three risk
importance on the instability process; finally, classes: low, medium and high. The risk map was
the factors were weighted and combined among obtained by overlaying in GIS the hazard and con-
them in a polynomial function. sequence maps and associating the corresponding
• Combining the intensity classes with the sus- risk classes.
ceptibility classes in a matrix and classifying
these combinations in terms of five hazard
classes: very low, low, medium, high. A column 4 RESULTS AND DISCUSSION
was added to the hazard matrix to consider the
area not affected by existing landslides, in which The classification of provincial road segments was
future landslides could occur. carried out considering the criteria defined in the
paragraph 3.1 and some special conditions. In
The landslide hazard map was obtained by over- particular, border roads, which link directly towns
lying in GIS the susceptibility map and the land- of Basilicata with towns of Apulia (for examples
slide intensity map. Matera and Gravina di Puglia) or connect urban
Consequences are generally defined as the out- centers, belonging to Matera province, to major
come or potential outcome to an element at risk roads (state highway, for example SS106 Ionica and
arising from the occurrence of a landslide of cer- SS407 Basentana), were considered as first order
tain magnitude (Glade and Crozier 2005). There- roads. While road segments connecting towns to
fore, consequences are a function of the amount industrial areas, small villages and abandoned or
of elements at risk and the vulnerability of the relocated urban centers as damaged by landslides
affected elements. (for example, Craco and Alianello) or in presence
The vulnerability of provincial road corridors of a better alternative roadway were classified as
crossing the Matera territory was assessed consid- second order roads. Finally, road connecting sec-
ering the previous criticalities and repair works car- ondary roads outside region or unpopulated sites
ried out along road network. In particular, about were assumed of 3rd order. The resulting road clas-
the 44% (584 km) of roads was affected, in the last sification is shown in Figure 4a. Out of the total
years, by instability and different types of works length of road network, 1st order roads are about
were carried out in order to repair it (Figure 2b). the 45% (600 km), 31% (407 km) and 24% (317 km)
As the vulnerability is intended as the degree of are, respectively, 2nd and 3rd order roads. By cal-
damage to a given element at risk caused by an culating for each road the amount of traffic, as
instability phenomenon, the road repair works, indicated in the paragraph 3.1, the exposure values
representative of the type and severity of damages, were achieved. These were normalized from 0 to
were ranked in five classes, to which a score ranging 1 in order to obtain the exposure map (Figure 4b).
from 0 (no works) to 1 (severe damages and struc- Then the same were subdivided into five classes.
tural consolidation works) was assigned. As the The exposure zonation has highlighted that the
road stretches affected by damages were repaired 15% of roads is characterized by moderate (E3)
by performing different typology of work, the amount of traffic, the 17% by both very low (E0)
vulnerability values were derived by summing, for and medium (E2) exposure, the 20% by high expo-
each road section, the several scores associated at sure (E4) and the 31% by low exposure (E1).
the work categories and by normalizing the result- The landslide risk analysis was carried out
ing values from 0 to 1. Five vulnerability classes exclusively along the road corridors. For this rea-
were associated to the following score ranges: null son, the spatial data contained in the input maps
(V0) to zero value, low (V1) to 0–0.25, medium (V2) were made uniform, in order to obtain raster maps
to 0.25–0.45, moderate (V3) to 0.45–0.75 and high with the same spatial resolution, i.e. 20 × 20 m.
(V4) to 0.75–1. At the same way, the exposure val- The landslide inventory map was transformed
ues, previously assessed, were normalized and sub- into a raster map, classified in terms of landslide
divided into five classes: very low (E0 = 0–0.1), low mechanism type, and then reclassified in terms of

98

TIS2017_Book.indb 98 2/8/2017 8:01:50 PM

 Figure 4. (a) Classification of road network into three orders and (b) exposure map.

Figure 5. (a) Landslide hazard map obtained by overlying susceptibility and intensity maps; (b) Consequence map
obtained by overlying vulnerability and exposure maps.

landslide velocity and area to obtain the intensity is affected by high hazard level, 3% is character-
map. While, the vulnerability and exposure raster ized by medium hazard and 40% by low or very
maps were obtained by rasterizing the road vectors low hazard.
and reclassified them, respectively, in terms of road The vulnerability zonation along road corridors
repair work category and amount of traffic. has revealed that about the 56% of roads is free by
By overlying the landslide intensity map on the damage, the 8% by low vulnerability, the 16% by
susceptibility map and by combining the corre- both medium and moderate vulnerability and 4%
sponding classes into the hazard matrix, the haz- by high vulnerability. By comparing the vulnerabil-
ard zonation was obtained (Figure 5a). Based on ity and exposure maps (Figure 5b) it can be noted
the hazard zonation, the 57% of the road corridors that road stretches affected by highest levels of vul-

99

TIS2017_Book.indb 99 2/8/2017 8:01:50 PM

 of risk corresponds to the following combinations:
very low hazard with major and catastrophic conse-
quences, low hazard with medium and high conse-
quences, medium hazard with minor consequences
and high hazard with insignificant and minor con-
sequences. The graph in Figure 6 shows the percent-
ages of the overall road network and of the stretches
crossed by landslide bodies included in each of the
three risk classes. The road sections subjected to low
landslide risk are 22% (290 km) on the total length,
while the remaining 48% (642 km) and 30% (392 km)
of the examined road corridors is affected, respec-
tively, by medium and high risk levels. The compari-
son between the risk map and the landslide inventory
recognized along roads has also revealed that about
the 10% (136 km) of roads are crossed by instability
phenomena. In particular, on the total of instability
phenomena, the 49.5% of landslides affected sections
Figure 6. Landslide risk map of Matera provincial road where the risk was evaluated high and 41% sections
network and graph showing the percentages of the entire where the risk was assessed medium. The remaining
road network and of the segments affected by landslides 9.5% of landslides crossed road stretches where the
included in each risk class. risk was mapped as low.

nerability (i.e. V4 and V3) are conversely character- 5 CONCLUSIONS

ized by low amount of traffic. Indeed the 78% of
roads free by damage (V0) has moderate and high A procedure for assessing and mapping the land-
exposure levels. The distribution of consequence slide exposure and risk along the Matera provin-
levels along roads (Figure 5b) was achieved by cial road network (Basilicata region, Southern
overlapping of vulnerability and exposure maps Italy) was explained. The main aim of this study
among them and from the combination of the cor- is provide a landslide risk map as support to local
responding classes into the consequence matrix. In governments for identifying the priorities for the
particular, the assignment of consequence levels to design of appropriate mitigation plans. This study
each combination of vulnerability and exposure was carried out for the provincial roads as they
classes was carried out by assuming more preva- represent the main network connecting the urban
lent the influence of exposure than vulnerability, centers of the Matera Province and is affected for
since this last does not derive from a probabilistic the 10% of its total length by landslide instability
analysis of potential damage degree but from the phenomena.
evaluation of distribution of past damages on the Road exposure to landslides was estimated in
roads. Based on consequences matrix, the 29% of terms of amount of traffic, qualitatively obtained
roads is affected by insignificant consequences, the by ranking the road segments in three differ-
19% by minor, 36% by medium, 9% by major and ent order and by calculating, for 1st order roads
7% by catastrophic consequences. (direct links among towns), the arithmetic aver-
Hazard and consequence maps were overlaid, age of population of interconnected urban cent-
the corresponding classes were combined in the ers and, for 2nd and 3rd order roads (connecting,
risk matrix and the risk classes (low, medium, high) respectively, 1st and 2nd order roads), the average
were associated to each combination, up to obtain of traffic amount on connected road stretches. The
the final landslide risk map along the 1,324 km of landslide risk was evaluated, in absence of detailed
Matera provincial roads (Figure 6). Road sections input data due to the scale of analysis, by a quali-
affected by very low and low hazard and by insignifi- tative matrix-based approach. The proposed pro-
cant and minor consequences, or by medium hazard cedure has allowed to assess and map landslide
and insignificant consequences or medium conse- risk, through the following three sequential steps:
quences and very low hazard are characterized by (i) producing landslide velocity and areal exten-
low risk. Conversely, medium and high hazard levels sion maps from landslide inventory map, accord-
with major and catastrophic consequences, medium ing to the movement type, and combining them
consequences with both medium and high hazard for obtaining the landslide intensity map; (ii) com-
levels or catastrophic consequences with low haz- bining, pairwise, intensity and susceptibility maps
ard involve a high risk level. Finally, a medium level and vulnerability and exposure maps for obtaining

100

TIS2017_Book.indb 100 2/8/2017 8:01:53 PM

 the hazard and consequences maps, respectively; Chowdhury, R.N. & Flentje, P. 2003. Role of slope reli-
(iii) overlying the hazard and consequences maps ability analysis in landslide risk management. Bulletin
and combining the corresponding classes in a of Engineering Geology and Environment 62: 41–46.
matrix in order to obtain the final risk map of the Corominas, J., Copons, R., Moya, J., Vilaplana, J.M.,
Altimir, J., Amigó, J. 2005. Quantitative assessment
roads, subdivided in low, medium and high risk of the residual risk in a rockfall protected area. Land-
levels. slides, 2: 343–357.
The comparison between the risk map and Ferlisi, S., Cascini, L., Corominas, J., Matano, F. 2012.
the distribution of landslides crossing the road Rockfall risk assessment to persons travelling in vehi-
stretches, has showed the majority of instability cles along a road: the case study of the Amalfi coastal
phenomena (about 91%) are located on road sec- road (southern Italy). Natural Hazard, 62: 691–721.
tions classified at medium and high risk. This is a Glade, T. 2003. Vulnerability assessment in landslide risk
significant result, as it shows that it is possible to analysis. Die Erde 134, 121–138.
use, for a regional scale landslide risk assessment, Glade, T. & Crozier, M. 2005. The nature of landslide
hazard impact. In: Landslide hazard and risk, edited
a simplified-qualitative procedure, based on a few by: Gladem T., Anderson M., and Crozierm M.,
data, easy to find and manage, and reliable in rela- Wiley, Chichester, 43–74.
tion to the scale of analysis. Although the pro- Hong, I. & Jung W.-S. 2016. Application of the gravity
posed methodology for the risk assessment did not model on the Korean urban bus network. Physica A:
consider the prediction of landslide run-out and Statistical Mechanics and its Applications 462: 48–55.
temporal probability, the results obtained allow to McNally, M.G. 2007. The Four Step Model. Chapter 3 in
rank the road stretches in terms of risk levels and, Hensher and Button (eds). Handbook of Transport
consequently, to identify the priorities for designing Modeling, Pergamon.
detailed field surveys and appropriate landslide risk Nicolet, P., Jaboyedoff, M., Cloutier, C., Crosta, G.B.,
Lévy, S. 2016. Brief communication: on direct impact
mitigation plans. probability of landslides on vehicles. Natural Hazards
and Earth System Sciences 16: 995–1004.
Pantelidis, L. 2011. A critical review of highway slope
ACKNOWLEDGMENTS instability risk assessment systems. Bulletin of Engi-
neering Geology and the Environment 70: 395–400.
The Authors wish to point out that Dr. R. Pellicani Pellicani, R., Van Westen, C.J., Spilotro, G. 2014a. Assess-
has been responsible for the research methodologi- ing landslide exposure in areas with limited landslide
cal approach and Prof. G. Spilotro for research information. Landslides 11 (3): 463–480. DOI 10.1007/
coordination. s10346-013-0386-4
Pellicani, R., Frattini, P., Spilotro, G. 2014b. Landslide
susceptibility assessment in Apulian Southern Apen-
nine: heuristic vs. statistical methods. Environmen-
REFERENCES tal Earth Sciences 72 (4): 1097–1108. DOI: 10.1007/
s12665-013-3026-3
AGS 2007. Pratice note guidelines for landslide risk man- Pellicani, R., Spilotro, G., Van Westen, C.J. 2016. Rockfall
agement 2007. Extraxt from Australian Geomechan- trajectory modeling combined with heuristic analysis
ics Journal and News of the Australian Geomechanics for assessing the rockfall hazard along the Maratea
Society 42 (1): 63–114. SS18 coastal road (Basilicata, Southern Italy). Land-
Budetta, P., De Luca, C., Nappi, M. 2015. Quantita- slides: 1–19. DOI 10.1007/s10346-015-0665-3
tive rockfall risk assessment for an important road
by means of the rockfall risk management (RO.MA)
method. Bulletin of Engineering Geology and the Envi-
ronment: 1–21 DOI 10.1007/s10064-015-0798-6.

101

TIS2017_Book.indb 101 2/8/2017 8:01:53 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Safety in the III Valico tunnels

A. Focaracci
Foundation Fastigi President, CEO, Prometeoengineering.it srl, Rome, Italy

ABSTRACT: The Genoa-Milan High-Speed Railway Line (III Valico di Giovi) is characterized by the
presence of several tunnel sections, including the double-barreled Valico tunnel with a length of 27 km
(not including interconnections, making it one of the longest galleries in the national territory), and the
Serravalle tunnel with a length of 7 km. In recent years the issue of tunnel safety has been the subject of
specific laws, both at the national level, with Ministerial Decree 28.10.2005 on Safety of Railway Tunnels,
and at the European level through Council Decision 2008/163/EC of 20 December 2007 on the technical
specifications of interoperability relating to Safety of Railway Tunnels in the trans-European conven-
tional rail system and to high-speed referred to as Technical Specifications for Interoperability or TSI.
The III Valico di Giovi is part of the European Genoa-Rotterdam corridor and, therefore, falls within the
before mentioned TSI introduced following the approval of the final design in 2005. The safety design,
based on in-depth analyses carried out by means of fire and exodus models and probabilistic risk analysis,
has led to the definition of the functional layout and performance specifications of safety systems, and
the development of a final plan that is among the most advanced in Europe from an operational safety
point of view.

1 INTRODUCTION tunnels, making the project particularly demand-

ing. The type of the tunnels planned are consist-
The Genoa-Milan High-Speed Railway Line (“III ent with the latest security standards, including
Valico di Giovi”) is characterized by the presence the construction of two side-by-side, single-track
of numerous tunnel sections, including the double- tunnels with cross-connections, which enable each
barreled Valico tunnel (not including interconnec- tunnel to be safe place from any events occurring
tions, making it one of the longest galleries in the in the other. The proposed route will start about
national territory), and the Serravalle tunnel with 800 m before Bivio Fegino on the line originat-
a length of 7 km. As part of the final design of ing from Genova Piazza Principe. The line passes
the railway line, which was built in 2004–2005, the under the Liguria Apennines with a tunnel,
safety of the tunnels was addressed according to the approximately 27 km in length, that opens up into
relevant laws and RFI standards that were required the municipality of Arquata Scrivia where the rail-
at the time. In light of recent changes in the regula- way line is expected to connect with the Libarna
tory framework, which do not entirely overturn the track, running along the Plain of Novi and, then,
security principles already used in the (III Valico) passing under the Serravalle Scrivia territory with
final design, but rather refine and integrate some of a tunnel about 7 km in length. The route ends in
the concepts verifying the design choices through a Tortona where a grade-level connection with the
security design methodology based on the risk anal- Piacenza/Milano line is planned. The plan for the
ysis, it has been necessary to re-evaluate the design Valico Tunnel, which constitutes the most signifi-
choices concerning safety of the III Valico galler- cant infrastructure work of the project, includes
ies. Present document describes the safety measures the construction of four windows, including two
(active and passive) adopted in the III Valico rail- exploratory narrow tunnels, which were par-
way tunnels and the final project completion added tially completed in 1996–98, for further develop-
in 2005 in order to adequalte the project to the new ment. Upon their completion these tunnels will
safety regulations for railway tunnels. constitute the Castagnola adit (Municipality of
Fraconalto) and the Val Lemme adit (Municipal-
ity of Voltaggio). Following the general diagram
2 THE PROJECT of the route with the works planned for the new
rail link and the associated plant equipment and a
The III Valico Railway runs along a route of about summary table of the salient features of individual
53 km characterized by the presence of several long tunnels are reported.

103

TIS2017_Book.indb 103 2/8/2017 8:01:55 PM

 by an adit through which a conflagrant train can be
driven and which will allow the controlled exodus
of travelers and the intervention of rescue teams.
The Val Lemme safety area consists of two evacu-
ation tunnels, which extend 750 m from the axis of
the adit, parallel to the tunnel axis, located 35 m
between the even and odd track, respectively. The
evacuation tunnels are accessible by the platform,
through branches, placed at a center-to-center dis-
tance of 50 m and are connected by a walkway,
placed over the two barrels, at the Val Lemme
Figure 1. Layout route.
tunnel access connection. Access from outside the
safety area takes place through the Val Lemme
3 SAFETY DESIGN adit, 1592 m long; at the tunnel entrance and at the
kp 0+700.00 two ventilation chambers are placed.
The method adopted for the safety design is the The tunnel safety area will be equipped with the
IRAM-RT methodology, already employed in the following systems to efficiently and effectively con-
safety design of the recently opened to traffic tun- trast the tunnel emergencies:
nels on the Bologna-Florence High-Speed railway
− ventilation system/smoke control;
line. The safety design of a railway tunnel foresees
− fire water system;
the following operating steps:
− automatic extinguishing foam system;
− analysis of infrastructure vulnerability, starting − hazardous liquid collection system.
from the acquisition of geometric, structural
The ventilation system/smoke control is designed
and equipment installation characteristics of the
according to the engineering approach to fire
work, traffic and accident data;
safety with reference to the international standards
− identification, structural and plant design of
as NFPA 92B and NFPA 130 and analysing simi-
safety requirements that may prove to be neces-
lar systems as those designed for the Turin–Lyon
sary after the vulnerability analysis (Ministerial
railway.
Decree of 10/28/05);
Using a probabilistic approach to safety design,
− risk analysis for the verification of safety targets
the flow rate of the ventilation system was assumed
achievements (Ministerial Decree 28/10/05);
to range from 200 m3/s to 400 m3/s. The design of
− operating procedures and in particular the
the system was carried out with reference to the
preparation of emergency management plans
maximum capacity and, considering the uncertain-
(Ministerial Decree 10/28/05).
ties related to possible dysfunctions, the complex
geometry and the behavior of airflow circuits in
the presence of hot smoke, a variability of about
4 UTILITIES SYSTEM EQUIPMENT
100 m3/s was estimated.
The system features a distributed fume extrac-
The key works for which the new regulations have
tion design whereby extraction is carried out by
required the design are:
a set of extraction points located along the safety
− the construction of a safety area inside the Valico area, placed at the tunnel access and inside of six
tunnel provided with suitable exit routes, smoke distributed bypasses. The scheme above described,
extraction and automatic shutdown system; allows optimizing the extraction of fumes in rela-
− the construction an open-air safety area between tion to the most likely locations where the fire may
the Valico tunnel and the Serravalle tunnels occur within the safety area and to geometric char-
equipped with an automatic shutdown system; acteristics of the same area. Fumes, once drawn
− the construction of new by-passes; and channeled, are conveyed into a false ceiling
− the upgrading of the system services of the tun- inside the Val Lemme tunnel access to be expelled
nel access; through the shaft foreseen in the project.
− the construction of new ventilation shafts and The ventilation control unit is located in a tunnel
adaptation of fire safety equipment. made ad-hoc before the shaft, designed to accom-
modate four two-stage axial fans able to extract up
to 120 m3/s each. The following figures show the
4.1 Val Lemme safety area
ventilation control unit at Val Lemme, the location
In compliance with STI requirements, the construc- of the by-passes used for the fumes extraction, and
tion of a safety area for passengers and freight trains an illustrative section showing the path fumes.
is planned at a mid-point within the Valico gallery The bypass system connecting the barrels of the
(about 27 km). This safety area will be accessible train tunnel with the safety area leading to the exodus

104

TIS2017_Book.indb 104 2/8/2017 8:01:55 PM

 guishing of flammable liquids and combustible
fuels the design proposes fire protection by means
of monitors to AFFF foam additives (Aqueous
Film Forming Foam), cooling agent and the forma-
tion of a protective film on any liquid fuel (B class).
The protection system provides a high foam flow of
up to 3000 l/min directly to the fire location, inhib-
iting combustion on the surfaces and subsequently
cooling them. The use of foam allows better cover-
age of the wet surfaces. In case of spillage and fire
of hazardous liquids the AFFF additive will cause
the rapid formation of an impermeable liquid film
on the surface of the spilled liquid. The designed
system, thanks to the high flows and possibility to
concentrate their action at the fire location, allow
for significant mitigation of the force of the fire,
Figure 2. Layout to the ventilation control unit at Val such higher as earlier the system is activated. The
Lemme. presence of an automatic extinguishing system at
the Val Lemme area reduces the uncertainties of
risk management, by permitting the reduction of
the fire power to less than 100 MW, resulting in a
significant improvement of tunnel system perform-
ance in terms of emergency management.
Along the entire length of the safety area there
will be a collection system of potentially hazardous
liquids. The spilled liquids and waters discharged
Figure 3. Layout odd platform for the fumes extraction. from the automatic shutdown will be channeled
into a tank located in the lower point of the safety
area where the flammable liquids will be separated.

4.2 Arquata Libarna safety area

The definition of new safety standards has neces-
sitated the creation of an external safety area, with
a length of 1166 m, located in the vicinity of the
Figure 4. By-passes used for the fumes extraction. Arquata Libarna PC. The above mentioned area
is accessible by emergency vehicles through a spe-
shaft is equipped with pressurization system, which cial road. The safety area contains zones equipped
will create an excess of pressure in the safety area so with a Triage area, a technological building, a res-
as to prevent the entry of fumes from the compro- cue helicopter pitch, and razed pathway for posi-
mised barrel. This is accomplished by means of a pair tioning the bimodal tack mechanism.
of fans (one backup) capable of preventing fumes The area’s external security systems are:
from entering the safety area. Fresh air is drawn into
the safety area from outside through a vent in the − fire water system;
false ceiling along the adit, passing through a control − automatic foam extinguishing system;
unit located outside the tunnel access. In the case of − hazardous liquid collection system.
malfunctioning fans, this control unit is also able to The fire water system, in accordance with the
provide a minimum of excess pressure to the safety provisions of the Ministerial Decree 28.10.2005 and
area. Moreover, the design of the external control reference TSIs, has been designed based on what is
unit includes a Saccardo ventilation installation, already reported in the 2005 project considering the
capable of pressurizing the entire tunnel access. In increase of pump capacity from 600 to 800 l/min.
conclusion, the design includes an extraction system The automatic extinguishing system is the Monitors
in the rescue vehicles parking area that picks up the type similar to the one designed for the Val Lemme
fumes directly from vehicle exhausts. area but with monitors spacing equal to 50 m
The fire water system, in accordance with the which can be active in groups of 3–6. The system
provisions of the Ministerial Decree 28.10.2005 and facilitates the extinguishing operations and is able
reference TSIs, will be composed of two pressuriza- to handle high flow rates of foam. It is also effective
tion control units, a storage tank and a network of against dangerous liquid fires with an AFFF foam
fire hydrants equidistant to 125 m. For fires extin- additive able to extinguish class B fires. In addition,

105

TIS2017_Book.indb 105 2/8/2017 8:01:56 PM

 Figure 5. Layout of the Arquata Libarna outdoor
safety area.

there will be a potentially hazardous liquid collec-

tion system. The Figure 5 illustrates the layout of
the Arquata Libarna outdoor safety area.

4.3 Cross connections

The escape system of the main line consists of a series
of by-pass link connections between the two single-
track railway tunnels (odd and even) every 500 m
approximately, in both the III Valico and Serravalle
tunnels. By-passes are used for people escaping from
one railway tunnel to a parallel railway tunnel; each
bypass is compartmentalized into both galleries. The Figure 6. By-pass link connections.
conceptual basis for the safety analysis is the consid-
eration of safe place of the intact (unaffected) tun-
nel. The ventilation system foreseen (pressurization
4.4 Access tunnels (adits)
of the bypass links road) allow keeping the exodus
ways free from the fumes produced in the compro- The exodus system with the adit allows for a widened
mised tunnel, with the following basic criteria: area at the end of each side tunnel, which forms a
space that is intended to allow for the reverse gear of
− ensuring effective pressure in the link road with
rescue vehicles and to accommodate the beginning
respect to the compromised tunnel both when
of the passenger flow from the tunnel towards the
the access doors (to the compromised tunnel
outside. Each of these areas, referred to as “transi-
and the unaffected tunnel) are open, or closed;
tion chambers”, is equipped with a series of doors
− guarantee, even in minimum load conditions,
(on the side railway tunnel and the on the side adit)
a suitable flow rate of air replacement to the
and a ventilation system capable of keeping the
considerable possible presence of persons inside
same chamber in slight overpressure with respect to
the bypass;
the tunnel. Furthermore, there is a second filter area,
− determine the air velocity in the exodus zones with
between the two rails tunnels (crossing the tracks)
values compatible with the emergency situation of
also equipped with a ventilation system, able to keep
the passengers, hit by substantial flow of air;
the same area in slight overpressure with respect to
− reduced start-up times of the fans (less than 30 sec)
the tunnel. In case of tunnel fire, the ventilation sys-
in order to reach, in the shortest possible time
tem prevents fumes from entering the exodus adits,
(about 35 sec), the standard overpressure expected
allowing the passengers to evacuate in safety. Finally,
for the volumes involved. The system will still keep
there is an extraction system designed for the adits
the by-passes free from any fumes present in the
in the emergency vehicle parking area, similar to
line gallery where the accident occurred.
that of the Val Lemme safety area, which directly
The exodus system for the Interconnection lines picks up the fumes from the vehicle exhausts.
consisted of two pedestrian by-pass crossing that
are connected one to the other. Since at the starting
4.5 Ventilation shafts
point the crossings are rather long and of reduced
section, a filter chamber beside the rail tunnel was In line with the ventilation strategies adopted for
created. This chamber is pressurized in a way simi- Italian railway tunnels and with the provisions of
lar to the “transition chamber” in the adit, with Annex II of the Ministerial Decree of 28/10/2005,
the same considerations in terms of safety analysis the points of transition from a twin-tube tunnel
(intact/compromised tunnel) and the conditions with a single barrel tunnel will be designed with
assumed in calculations. measures preventing the circulation of fumes from

106

TIS2017_Book.indb 106 2/8/2017 8:01:56 PM

 the compromised tube to the unaffected tube, by possible to reduce the uncertainties associated with
means of ventilations shafts. Intake grids posi- the transportation system efficiency, in the occur-
tioned on top of the tunnel will suck fumes into a rence of chaotic nature hazardous events such as
circular segment section plenum. As a results of the fires. As an example, passenger exodus simulations
design specifications based on the analysis of train and thermo-fluid dynamic simulations of smoke
fire scenarios and risks (where a thermal power extraction carried out for the Val Lemme safety
of 10 MW was used for potential passenger train area are reported in the following sections.
fires and 50 MW for freight train fires), the adjust-
ment intervention of the III Valico final design
5.1 Val Lemme safety areas—thermo fluid
includes the addition of new ventilation shafts and
dynamic smoke extraction simulations
the modification of the extraction flow rate for
those already planned. The ventilation shafts have The analysis of accident scenarios, using the three-
been designed based on the thermos-fluid dynamic dimensional Fire Dynamics Simulator calculation
simulations results to allow flow rates extraction of code, was performed by means of the simulation
approximately 200 m3/s. of the smokes spread generated by a stationary
train at the Val Lemme safety area. The simula-
tions target is the functionality and performance
5 RISK ANALYSIS of the smoke extraction system verification.
The proposed smoke extraction system involves
The Valico tunnel system, considered together with the construction of n. 6 extraction points to be
the Compasso tunnel and the Voltri Interconnec- assessed based on the maximum distance between
tion as one single system, is greater than 9000 m in the ventilation outlets. The identified scenarios are
length. Therefore, in accordance with Annex III of summarized below:
the Ministerial Decree of 28/10/2005 the expanded
− passenger train with a potential thermal power of
risk analysis of the system was performed. The
20 MW and suction capacity of about 200 m3/s;
analyses conducted using the IRAM RT method
− freight train with a thermal power of 50 MW
have shown how the safety measures in the tunnel
and a suction capacity of about 400 m3/s.
design allow for a level of risk which falls within
the attention zone, primarily due to the high vol- The results of the simulations, expressed in
ume of freight trains predicted, as exemplified in terms of the distribution of the fumes in the tun-
the next figure. nel, profiles of temperature, concentration of
The adoption of a training program aimed to carbon monoxide and visibility for different time
limit the contemporaneity between freight trains intervals, showed that:
and passenger trains would determine further risk
− in both the analysed scenarios the ventilation
reduction. In order to verify the functionality of
system ensures environmental conditions at the
the designed works in terms of both smoke man-
dock compatible with the timelines, taken as ref-
agement and exodus, numerous three-dimensional
erence for the exodus of the passengers and the
simulations for representative scenarios were car-
train crew;
ried out. The analysis made it possible to calibrate
− the optimum configuration for the extraction
the statistical models adopted for the risk calcu-
points has been detected in a mixed combina-
lation, review the emergency management time-
tion, which provides for a distribution of vents
lines, supporting the choices made with regard to
every 100 m with the three control units every
safety systems such as power shutdown and ven-
50 m. The vents gathered centrally, where it is
tilation. The execution of the simulations made it
easier to happen the train on fire, nevertheless
ensure the protection of 450 meters of tunnel
that includes an entire passenger train.
In the case of freight train, even if fire occurs at
the train end, the simulations have shown that the
vents at 100 m distance are sufficient to ensure that
the ventilation system guarantees the protection of
drivers. In the case of freight train fire and in the
phases of Firefighter intervention, the opening of
the extraction vent localized on the graft of the adit
has been designed, in order to allow the extraction
to the maximum system capacity. The simulations
conducted show that the security plan as a whole,
Figure 7. Global FN curve. thanks to the presence of the combined distributed

107

TIS2017_Book.indb 107 2/8/2017 8:01:58 PM

 was characterized by a generated maximum ther-
mal power close to 20 MW with a gradual increase
in a time equal to 10 min.
The simulations carried out on the exodus proc-
ess of 500 persons aboard a passenger train stopped
at the Val Lemme safety area, show that, the com-
bined fire simulation with the exodus simulation
localized at the dock, in presence of train accident,
all the passengers leave the train and enter in the
connecting branches with a total evacuation time
of about 3 minutes.
Figure 8. Temperature simulation.
5.3 Conclusions
The safety of travelers is a major issue, which in
recent years has become increasingly important
also because of a significant change in the regula-
tory framework (Ministerial Decree of 28/10/2005
and Technical Specifications for Interoperability).
Based on these recent changes, the union of the
primary and most competent engineering compa-
nies in Italy has allowed the creation of a consor-
tium of design, which by increasingly advanced
and cutting-edge technologies, has contributed
to the current project of the III Valico railway
Figure 9. Exodus simulation. design. The approach adopted in the design led
to the definition of a performance benchmark for
extract ventilation system, guarantees a minimum facilities that allow for emergency management in
level of security largely compatible with the current the most likely scenarios supporting the risk anal-
regulations at national and European level. The ysis results provided by the Ministerial Decree of
following figures summarize the three-dimensional 28/10/2005.
analyses performed for the verification procedure The obtained results at the design level discussed
of the tunnel access extraction system. in this article, backed up by detailed analyses con-
The simulation results have shown that the fume ducted by means of fire and exodus models and the
extraction zone is characterized by flows leading analysis of probabilistic risk, have led to the defini-
to high load losses in justification of the high fan tion of the functional layout, and of the perform-
performance necessary for the system operation ance specifications of security installations, and to
phase. the development of a final project that is among
the most advanced in Europe from the point of
view of safety in the operation phase.
5.2 Val Lemme safety areas—exodus simulations
The analyses of accident scenarios was performed
using the three-dimensional Fire-Dynamics Simu- REFERENCES
lator calculation code by the simulation of the
Focaracci A., Lunardi. P. Silva C. (2001): “The Bologna
smoke propagation generated by a passenger train to Florence Hight Speed Railways line: 92 km through
stopped at the Val Lemme safety area, coupled the Appenines”, Progress in Tunneling after 2000 –
with the simulation of the exodus process, con- Bologna 2001.
ducted through the EVAC code. Focaracci A., (2007): “Designing Safety”—Italian Risk
The targets of the simulations consist of: Analysis Method—Le Strade n°4–2007.
Focaracci A., (2007): “Commission Tunnels Safety to
− simultaneous verification of the functionality route and railways”—Conference Safety in Tunnels—
and performance of the safety systems; Genova, 27–28 March 2007.
− verification of the escape times of the passen- Focaracci A., (2010): “Italian Risk Analysis method
gers in accident conditions; IRAM” 11th International Conference Underground
Construction Prague 2010—Transport and City Tun-
The scenario analyses, with the primary purpose nels. 837–845.
of verifying the emergency management, is char- Focaracci A., (2010): “Safety in Tunnels: innovation
acterized by a greater verisimilitude. The assump- and tradition” 11th International Conference Under-
tion adopted for the definition of reference fire, to ground Construction Prague 2010—Transport and
check safety conditions of the exposed population, City Tunnels. 846–851.

108

TIS2017_Book.indb 108 2/8/2017 8:01:58 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

AllBack2Pave: Towards a sustainable recycling of asphalt in

wearing courses

D. Lo Presti & G. Airey

University of Nottingham, Nottingham, UK

M. Di Liberto, S. Noto & G. Di Mino

Universita’ degli Studi di Palermo, Palermo, Italy

A. Blasl, G. Canon Falla & F. Wellner

Technische Universität Dresden, Dresden, Germany

ABSTRACT: Nowadays, asphalt plant technologies allow producing asphalt mixtures incorporating
up to 100% reclaimed asphalt. Unfortunately policies are still behind technology and in order to sug-
gest guidelines for a widespread use for surface courses, road managers feel the need of having a deeper
understanding of optimised design strategies, information related to the handling in asphalt plants and
on the performance of these mixes. This paper provides the summary of the main results and the details
of the main idea behind “AllBack2Pave 2013–2015” a two-years, 500 K€ project funded by the CEDR
Transnational Road Research project that evaluated the feasibility of going towards 100% recycling of
asphalt pavements into surface courses. The project, coordinated by the Technische Universität Dresden
in Germany, together with the University of Nottingham and University of Palermo, was structured so to
provide an overview of the European panorama by involving 3 EU countries from the South (Italy), Cen-
tre (Germany) and North (UK). In order to facilitate the deployment of lean concepts and lean produc-
tion practices, the investigation was implemented in close collaboration with the private sector, including
asphalt mixing plants, chemical additives producers and waste material managers.

1 INTRODUCTION in response to the common research needs of its

National Road Administration members. This
Nowadays, the amount of recycling of Re-claimed Transnational Research Programme called “Road
Asphalt (RA) in new asphalt pavements has grown construction in a post-fossil fuel Societyî had as
to the point that it is no longer a mere alternative overall aim to develop new concepts for road con-
when virgin materials are not easily available (i.e. struction especially for pavements in the context of
The Netherlands), but a common practice in almost the finite nature of fossil fuels and other resources.
all of Europe. However, the dismantling and end Economic and environmental benefits are the driv-
of life strategies for these pavements in European ing forces behind research. In turns, the required
countries are very divergent relating to the amount solutions needed to be feasible, valid and cost-ef-
of RA recycled in new pavement layers, although, fective through increased recycling and consequent
due to mainly policy’s constrains, over the whole minimisation of the impact in the society.
Europe the share of recycling of RA in new asphalt “AllBack2Pave: towards 100% recycling of
courses remains rather lower than it could be tech- reclaimed asphalt in wearing courses” was one
nically. In fact, some of the most recent European of the selected proposal. The project started in
collaborative research projects, such Direct-MAT, November 2013, lasted 2 years, and was led by
2011 and Re-Road, 2012, have shown that the the Technische Universität Dresden in Germany,
durability of hot asphalt mixes with RA proved to together with the University of Nottingham in the
be satisfactory, however the complete reuse of the UK and University of Palermo in Italy. The inves-
reclaimed construction materials requires a precise tigation was implemented in close collaboration
assessment of the properties of the components. with the private sector, including asphalt mixing
On this wave, in 2012 the Technical Group plants and chemical additives producers and main
Research of CEDR, the Conference of Euro- idea behind it was evaluating the feasibility of going
pean Directors of Roads, initiated a programme towards 100% recycling of asphalt pavements into

109

TIS2017_Book.indb 109 2/8/2017 8:02:01 PM

 surface courses by following the structure shown in 2 LABORATORY-BASED DESIGN
Fig. 1 and aiming at the following objectives:
Within this phase the partners aimed at designing
− establishing, through laboratory tests on binders
two commonly used asphalt mixtures for surface
and asphalt mixes, whether the use of high rates
courses, SMA and AC, incorporating the highest
of RA is feasible in developing mixes with a high
amount possible of RA (Fig. 2). These mixtures are
level of durability.
characterized by a strong coarse aggregate skeleton
− developing the so-called “AllBack2Pave end-
that gives good resistance to permanent deforma-
user manual” with best practices to produce
tion, in case of the SMA obtained also thanks to
cost-effective asphalt mixes with high RA
a relatively high bitumen content (>6% and often
content.
a PmB), and providing enhanced resistance to
− assessing the sustainability performance of the
fatigue. Furthermore, the advantages in terms of
technologies, at first through an evaluation of
longer life and improved performance make these
the design life of the mixtures and then with
mixes economically viable, even if the SMA ini-
an holistic approach based on Life-Cycle based
tial costs is around 20% higher than conventional
techniques such as Life-Cycle Assessment (LCA)
asphalt concrete mixes. Therefore, being successful
and Life-Cycle Cost Analysis LCCA.
with maintaining these high standards, by replacing
This paper illustrates the outcomes of the a high percentage of virgin materials with reclaimed
project and each of the following paragraph will asphalt, will then result in a more economic and
summaries the main results of each of the follow- potentially environment-friendly technology.
ing technical Work Packages: Within this laboratory exercise several key points
were discovered in both case studies. Good results
• In Work Package 2 “Laboratory Design” differ-
as well as discovering limitations have led the
ent reclaimed asphalts and virgin materials were
partners to highlight some key areas that deserves
collected and characterized in Germany and
attention before undertaking the design of high-
Italy. Mix design of currently used asphalts for
content reclaimed asphalts for wearing courses:
wearing courses in these countries were under-
taken considering 0%, 30%, 60% and the closest
feasible amount to 100% of RA. 2.1 Sampling of RA
• Work package 3 “Plant scale manufacturing and
There is a common concern that the high quality
end-user manual” aimed at validating the labora-
requirements for asphalt wearing courses are not
tory design and led to recognize and solve issues
met if the major volumetric component of the mix
related to the plant manufacturing.
comes from a recycled material. The background
• Work Package 4 “Performance prediction”
of this fear is the inherit heterogeneity of the RA,
aimed at assessing the durability of the high-
which properties depend in great extent on factors
content RA Warm asphalt mixes in terms of
such as the technique used to reclaim the asphalt, the
mechanical properties and evaluating design life
of the mixes in typical traffic scenarios.
• Work Package 5 “Sustainability Assessment”
involved carrying out a life cycle assessment, life
cycle cost analysis in the context of a broader
sustainability assessment of the investigated
technologies. Additionally a state of the art
review on sustainability rating systems was pro-
vided together with the development of a bespo-
ken sustainability assessment methodology.

Figure 1. AllBack2 Pave project structure. Figure 2. SMA11S and AC16 case studies.

110

TIS2017_Book.indb 110 2/8/2017 8:02:01 PM

 maintenance history of the road, the storage con-
ditions, etc., which are not considered in standard
mix design procedures. In fact, high quality aggre-
gates with high resistance to wear/abrasion (polish-
ing) are needed for wearing courses. Thus, properly
milled (i.e. layer by layer) and stockpiled RA is a
mandatory prerequisite in order to produce durable
asphalt wearing courses with high content of RA.
This is a major concern because in the majority
of the cases the RA is produced and stockpiled by
non-selective methods in which valuable high qual-
ity aggregates of wearing courses are milled/stock-
piled together with lower quality materials of deeper
layers. That means that the aggregates of the RA
will automatically be downgraded losing economic
value. One possible solution is to allow the use of
high percentage of RA in wearing courses only if
Figure 3. Grading of RA aggregates (white-curve) for
the RA material proceeds from the same location the SMA and AC with limits respectively from TL09 and
and layer where the new mix will be placed. ANAS case study.

2.2 Characterization of the RA towards designing

asphalt with the highest feasible content
indicated that the maximum feasible percentage of
The RA needs to be characterized before the actual RA was 70% for the SMA11S case, while for the
mix design. This is because with ageing and oxida- AC16 case was over 90%.
tion some changes may occur in the mix: for the Skid resistance: The aggregates ability to resist
binder, this includes hardening and loss of ductility skidding is crucial when designing wearing course
and for the aggregates the gradation may change mixes. The Polished Stone Value (PSV) is a com-
due to degradation caused by traffic loads and mon evaluation method used for measuring skid
environmental conditions. It is therefore important resistance. Two skid tests were performed to deter-
to characterize the RA with the same rigor used mine the PSV of the RA aggregates by using the
for the virgin aggregates except for the fact that British pendulum. As example, in the SMA case
family of materials is a complex agglomeration of study, the tests showed that the RA aggregates have
stones that retain visco-elastic properties. Within an average PSV value of 51, which means that the
the project the RAs were subjected to binder recov- polishing characteristics of the RA aggregates are
ery procedures to obtain the so-called white RA. good enough to be used for wearing courses of fed-
Then white RA was characterized according to the eral high volume roads in Germany (min. required
standard used for aggregates for road pavement. PSV value of 51, acc. to TL Asphalt-StB 07).
Gradation: In particular, the gradation of the
RA aggregates is one of the most important fac-
2.3 Preliminary binder blend design
tors associated with the control of asphalt mixes
with high RA content. At very high recycling The binder selection for a specific mix design is
rate, the gradation of the RA affects the overall usually done based on traffic and loading require-
performance of the final mix including stiffness, ments expected in the wearing courses of the
fatigue resistance, rutting and moisture damage. selected road. For instance, According to the
Aiming towards 100% recycling, the gradation of German bitumen specification, for a federal high
the selected RA must be as close as possible to the volume road with a surface layer made of a SMA
bands, but also the distribution of particle sizes in material, it is required to use modified bitumen of
the aggregates of the RA must have just the right the type 25/55–55. The contract specifications of
density so that the final mix will contain the opti- this type of bitumen are presented in Figure 4.
mum amount of asphalt binder and air voids. Then, in order to obtain a high-content RA mix-
In case the grain size distribution of the RA does ture comparable to a more standard mix, it is neces-
not meet the control points of the target mix, as for sary that the binder blend of bitumen contained on
the SMA11S in Fig. 3, it is necessary to add virgin the RA and the other additives/rejuvenators complies
aggregates in order to balance the grading. Using with the above-mentioned standard. In order to do
an iterative algorithm that decreases the amount that it is therefore necessary to perform an accurate
of RA and increases the amount of virgin aggre- binder blend design, which is specific to the selected
gates can do this. In our investigation, the results RA, virgin bitumen and other components.

111

TIS2017_Book.indb 111 2/8/2017 8:02:02 PM

 • 3) The batch plants with additional parallel drum
(maximum RA content up to 100% for counter-
flow heating drum, ∼60% for parallel RA heating
drum): where RA and the virgin constituents are
added in separate material flows to the mixer;
this process allows the gentle heating of the RA.
Then continuous drum mixing plants (maximum
Figure 4. Contractual requirements for a PmB 25/55–55 RA content ∼25%) exist and this is where the RA
(TL Bitumen-StB 07, 2007). is heated in the same drum mixer as the aggre-
gates but this may provoke an intensive aging of
the RA binder.
Within the project two very different RAs were According to this classification, two plants were
selected: the RA used for the SMA in Germany was selected to produce the mixtures and with two dif-
short-term aged (pen22) while the one used in Italy ferent aims: 1) understanding production issue
for the AC16 case study, was extremely aged (pen8). in a batch plant with a parallel drum capable of
To determine suitable virgin bitumen, in both cases manufacturing 100% RA mixtures (SMA11S);
a comprehensive blend study was performed at the 2) Evaluating strategies to maximize the amount
laboratory of UNOTT. This study allowed defin- of RA in a simpler batch plant with RA heating
ing a procedure to assess whether the selected vir- drum (AC16).
gin bitumen and additives/rejuvenators would allow
obtaining a final binder with the desired target prop-
erties even considering the variability of degree of 3.2 SMA and AC mixes manufacturing
blending (DoB) and binder content of the RA. The SMA11S mixtures were produced at the mixing
targeted properties were based on both conventional plant of Richard Schulz Tiefbau GmbH & Co
and rheological properties. A detailed explanation that is located in Gilching, near Munich, Germany
of this procedure has been published elsewhere (Lo (Fig. 5). The mixing temperature were established
Presti et al. 2016, Del Barco Carrion et al. 2015). As at the design stage and were measured straight
a result, for the SMA11S case study the properties of after the production from the silo (Table 1).
the virgin bitumen from the RA did not differ much The AC16 mixtures were instead produced in
from the properties of the target binder, therefore, it Italy with a simpler batch plant with coaxial cylin-
was decided to use as virgin binder a PmB 25/55–55 der RA heating plant of Ferrara & Accardi near
(i.e. the same type and grade as the target). Instead
for the Italian case study, the RA was so much aged
that each of the mixes (30%, 60% and 90%) needed
the addition of a rejuvenator that was also supposed
to act as warm-mix additive.

3 PLANT SCALE MANUFACTURING AND

END-USER MANUAL

3.1 Selection of the asphalt mixing plant

Generally three types of batch plants are in use
for which further modular machinery may allow
higher recycling rates:
• 1) The batch plant (most widely used, charge Figure 5. Scheme of the batch plant with a parallel
addition—maximum RA content ∼30%) where drum used for the SAM11S case study.
RA is added cold directly into the mixer and
heated up indirectly through contact with super- Table 1. Mixing Temperatures in German Plant
heated stone fractions into the mixer This proc- (SMA11S).
esses may provoke an intense aging of the RA
binder. Mix Mixing temperature [°C]
• 2) The batch plant with RA heating drum (maxi- 0% RA 170
mum RA content of ∼ 60% or counter-flow RA 30% RA 170
heating drum with maximum RA content of up 60& RA 170
to 100%): allows gentle heating of RA into the 90% RA + Add. 165
drum to avoid excessive binder aging.

112

TIS2017_Book.indb 112 2/8/2017 8:02:02 PM

 Catania (Figure 6). Due to the ageing of the RA, 3.3 End-user manual methodology
ando also to try maximizing the RA amount, all the
The scope is to define the guidelines of drafting of
mixes were produced with the addition of the addi-
an End-Users Manual addressed to all stakehold-
tive which was fed directly into the mixer. Despite
ers such as road agencies, production plants and
several trials to try producing the 90% RA moisture,
construction enterprises. Two main technical goals
this plant allowed producing only the asphalt with
can be summarised: 1) the consensus between mix-
up to 60% of RA. In fact, in order not to have an
tures produced in laboratory and in plant; 2) the
efficient filtering of harmful fumes, the maximum
achievement of the pavement performances tar-
percentage of RA permitted in the production of
geted with the mix design.
high-content RA mixes is directly proportional to
Therefore the main concept behind the idea of
the amount of virgin aggregate content inside the
the End-Users Manual lies on trying to maximize
drum. For this reason, the mixture with 60% of RA
both the RA use within new road pavement with-
was produced by feeding the recycling ring of the
out compromising the performance of the super-
drum with only 50% of the total weight of required
structure. Since the overall RA recycling process,
RA, while the remaining cold part was directly sent
from the dismantling of old pavements to the reuse
to the mixer, once the virgin aggregate and hot RA
within new pavements, is complex and several users
were already inside. With regards to the mixture with
have specific roles and objectives, it seemed most
90% RA, because of the limits of the mixing plant
practical to use tailored check-lists for each user that
system, most of the RA needed to be fed directly
summarise all actions to be implemented to achieve
cold into the mixer which was not able to achieve
the purposes. The structure of the checklist provides
a mixing temperature above 110°C. The too small
for each operation, the Basic Rule, required in order
amounts of virgin aggregate inside the drum did not
to perform well each step of the process. The type of
allow a sufficient heat exchange to ensure the attain-
rule can be identified by a regulatory requirement
ment of temperatures necessary for the blending;
(Regulation, R), a prescription contract (Specifica-
in addition, the shortage of virgin material within
tion, S), and a good practice (Best Practices, BP).
the drum did increase the temperature of the fumes,
In order to evaluate the quality of the generic
leading it abruptly to the threshold value permitted
process path, a basic rule can be weighted depend-
by the system. Thus all mixes were produced in plant
ing on the category (Regulation, R; Specification,
except the one with 90% RA. The latter was then
S; Best Practices, BP) to which it belongs. The
manufactured in the laboratory for the experimen-
weights may be set conventionally 3, 2, 1, by set-
tal survey. The mixing temperatures are reported in
ting a descending hierarchy that goes from R to
Table 2 provided for each of the blends produced.
BP. To assess the reliability of the generic process,
the REliability Measure parameter (REM) was
defined as follows:

n
REM
E ∑WV[
i =1
YES ]
BRi Y

where:
• n is the total number of the basic rules of the
given path;
• WV [BRi = yes] the value, varying from 1 to 3, of
the ith basic rule put in place.
Figure 6. Scheme of the batch plant with coaxial RA The reliability of the generic process is calculated
heating drum used for the AC16 case study. by dividing REM to the maximum achievable REM,
i.e. when the checklist is populated with all its posi-
Table 2. Mixing Temperatures in the Italian Plant
tive answers. It then defines the coefficient of reliabil-
(AC16).
ity “Reliability Ratio” (RR) varying from 0 to 1, as:
Mix Mixing temperature [°C]
∑ WVV[
n
i =1 YES ]
BRi Y
0% RA 160 RR =
∑ WVV[
n
30% RA + Add. 163
AllBRi =YES ]
i =1
60& RA + Add. 165
90% RA + Add. –
(170 produced in lab)
where: WV [AllBRi = yes] is the maximum value
of the REM.

113

TIS2017_Book.indb 113 2/8/2017 8:02:03 PM

 Table 3. Check List applied to the Italian case (AC16 ).

Operation Basic rule (Code) Description Done (Y/N)

Single Source Separate Storage Area BP Minimum area should be not less Y
Stockpiles (4S-SA1) than 1500 m2. Area should be
sloped (six degree is ideal)
Single Source Separate Storage Area BP Treatment of the surface area Y
Stockpiles (4S-SA2) (no water, no clay)
Single Source Separate Storage Area BP Permeable to air roof with N
Stockpiles (4S-SA3) permeable to air membrane
Single Source Separate Stockpiling BP Stockpile must have conical shape Y
Stockpiles (4S-S1) with maximum height of 6 m
Single Source Separate Stockpiling R Searching for deleterious N
Stockpiles (4S-S2) materials (EN 12697-42)
Sampling according to EN 932-1
Single Source Separate Stockpiling R Determining aggregate grading N
Stockpiles (4S-S3) (EN-13043) Sampling according
to EN 932-1
Single Source Separate Stockpiling R Determining binder content Y
Stockpiles (4S-S4) (EN 12697-1) Sampling
according to EN 932-1
Crushing & Fractioning Preliminary Screening BP Preliminary screening of the finer N
(CF-PS1) particles by a suitable sieve
(3/16 ASTM series or equivalent)
Crushing & Fractioning Before & After analysis BP Gradation control before & after
of RAP gradation the in ine crusher to determine the N
CF-BAG1) RAP aggregate size (within 1B&A
analysis 2 samples for RAP source)
Crushing & Fractioning Number of Screening BP The Plant must have 3 screening N
(Sieves) Unit unit at least, typically 3/4, 3/8, 3/16
(CF-NSU1) (ASTM series or equiv.)
Batch Plant Drying RAP BP The plant must have a rotary Y
(BP-D2) drum dryer with recycling ring
(T = 110 ÷ 130°C)
Warm Recycling Mixing Time BP Verifying the mixing time is within YES
Control (WR-MTC1) the range 25 ÷ 90 s
Warm Recycling Emission Control R Verifying the pollutant emission YES
(WR-EC1) according to UE Directive 75/2010

In order to evaluate the technical processes of tigated according to the following plan: (i) the stiff-
both Italian plant and German one, the methodol- ness modulus by both Four Point Bending Beam
ogy described above, has been applied. According (4PBB) test and Indirect Tensile stiffness Test
to the responses of the checklist, the “Reliability (ITT); (ii) the resistance to fatigue also by 4PBB
Ratio is equal to 0.52 and 0.85 respectively. Here test and ITT; (iii) the resistance to permanent
the checklist for Italian plant is reported as exam- deformation through Wheel Tracking Test (WTT)
ple in Table 3 and Uniaxial Compression Test (UCT) and (iv) the
More details of the results of this section are resistance to moisture damage by means of Indi-
reported elsewhere (Allback2Pave D3.1, 2016). rect Tensile strength Test.
In addition, the characterisation of the bind-
ers extracted from the final mixes has been carried
4 MECHANICAL CHARACTERISATION out in order to understand whether performance-
AND PREDICTED PERFORMANCE related tests on binders compare with mixture’s
results also to validating the laboratory mix
On the basis of both the scientific literature and design undertaken in WP2. This was done by
the availability of test devices in the laboratories assessing (i) rutting resistance by Multiple Stress
involved, the mechanical properties of the eight Creep Recovery (MSCR); (ii) fatigue resistance by
produced wearing course asphalt mixes were inves- Time Sweep tests and analysis of (iii) the thermal

114

TIS2017_Book.indb 114 2/8/2017 8:02:04 PM

 cracking resistance plus the determination of criti-
cal temperatures. A part of the stiffness and fatigue
related asphalt test results have been used as input
data for design life calculations.

4.1 Performance-related properties of binders

Given the results of all the tests, different con-
clusions can be drawn: for each case study, with
regards to the use of the additive and as a result of
comparison with the design stage. In the SMA11S
case, RA binder and virgin bitumen were known
to have not very different properties. In this sense,
results for high RA content binders, even with-
out using rejuvenators, show that binder recov-
ered from high-content RA mixes have similar Figure 8. Suggestion for a performance-based CEDR
properties to recovered virgin binder. Specifically, sustainability assessment methodology.
as the percentage of RA increases in the bind-
ers (30–60%), rutting resistance improved, while
4.2 Performance-related properties of mixes
fatigue and low-temperature behaviour did not
change in comparison to the scenario without RA. • Stiffness: The results of the stiffness related indi-
However, when the rejuvenator was added, rutting rect tensile tests and four point bending beam
resistance slightly decreases compared to the virgin tests confirm a comparable dependence of the
one, but fatigue and thermal cracking resistance material behaviour of the amount of RA within
improved, obtaining binders with good predicted- the investigated eight asphalt mixes. Regard-
performance (Figs. 7, 8). ing the SMA mixes it can be concluded that, as
In the AC16 case, RA binder and virgin bitumen expected, the stiffness increases with an increas-
were known to be extremely different; therefore, ing amount of reclaimed asphalt. However, the
some issues could be expected when high RA per- asphalt mixture with 60% RA + Additive shows
centages were to be used. However, thanks to the a lower stiffness due to its higher binder content
use of rejuvenator, and in accordance to the design and maybe also due to the presence of additives.
carried out in previous deliverables, rutting resist- The fourth SMA mix has been designed as an
ance and low-temperature behaviour improved experimental asphalt mixture, where originally
with higher RA content. Regarding fatigue results, the allowance of additives was not required by
binders with “30% and 60% of RA + Additives” the results of the laboratory tests on recovered
showed very similar behaviour to that of the “0% binders in line with the mix design procedure. For
RA” content, and only in the case of “90% RA + comparison, the AC mixes show almost no dif-
Additive“ fatigue life would be shorter, but still ferences between the temperatures related stiff-
within acceptable limits. With these results we can ness. It seems that the additive has completely
confirm that the additive generally allows obtain- compensated the influence of the RA what con-
ing a good fatigue behaviour of the binders up to firm an optimal blend design procedure.
60% RA. Some details are reported in Figure 7. • Rutting: the results of the wheel tracking tests and
The whole investigation is detailed elsewhere such the modified uniaxial compression tests do not
the deliverable 4.1 of the project (Blasl et al. 2016, lead to a consistent ranking of the asphalt mix-
Allback2Pave D4.1, 2016). tures. Regarding the SMA materials the plastic
deformation results of the two different labora-
tory tests can be qualitatively compared, whereas
the AC mixtures results show completely different
rankings. Regarding the SMA 8 S mixes it can be
concluded that the amount of reclaimed asphalt
has no significant influence on the plastic defor-
mation. The mixture with 60% RA and additives
has a higher binder content in comparison to the
other mixes, which leads to higher plastic strains.
Therefore, a relation between additive and the
development of higher plastic strains at the mixes
Figure 7. Performance related tests on recovered binder with 60% RA cannot be concluded. In terms
SMA11S case study: critical temperatures (top). of the AC16 materials the accumulated plastic

115

TIS2017_Book.indb 115 2/8/2017 8:02:04 PM

 strains considerable decrease with an increasing developed by this project and proposed in details
amount of RA. The influence of temperature for ease of use by CEDR members.
and stress level on the permanent deformation Each of these sections provided specific inputs
behaviour is significant. Therefore, the evalua- such as an overview of sustainability assessment
tion of the plastic deformations should take into system, a screening on freely available tools to
account a wide range of the testing conditions. perform LCA and LCCA, practical example of
More details are reported in the deliverable environmental and economical impact assessment
• Fatigue: The results of the fatigue related tests for road pavements and finally also a bespoke
provide the conformation that the stiffer an sustainability rating system for asphalt pavement
asphalt material is the higher is the number of which adapts to Europe the existing BE2ST tool
load cycles until failure. The outcomes of the developed by The University of Winsconsin and
fatigue related indirect tensile tests and four make uses of the ECORCE software developed
point bending beam tests partially show a dif- from IFSTTAR (Dauvergne et al. 2014). A sum-
ferent dependence of the material behaviour of mary of the forecasted advised methodology
the amount of RA within the investigated eight for a sustainable decision related to new design,
asphalt mixes. Clear conclusions and compari- maintenance and rehabilitation of EU road pave-
son of the results obtained by both test results ments is indicated in the figure below and will
very difficult to undertake. be detailed elsewhere (Allback2pave D5.1, D5.2,
• Moisture damage: Finally, with regards to the D5.3, 2016).
moisture damage resistance, all asphalt mixes
show no substantial sensitivity to damage
moisture. 6 CONCLUSIONS
Details will be reported elsewhere (Allback2-
Pave D4.1, 2016). AllBack2Pave is a 2 years project that involved a
big efforts from three partners to provide CEDR
with strategies, methodologies and tools to design,
5 SUSTAINABILITY ASSESSMENT manufacture high-content RA asphalt mixes and
assess the sustainability of their use within Europe.
AllBack2Pave supports the philosophy that sus- Here is a summary of the main findings:
tainability principles should be already consid- RA is a family of materials, therefore a specific
ered at the design stage of a road pavement, as preliminary characterization of the RA is neces-
well as any other structures, so that a sustainable sary. Furthermore, any mix design containing RA
transport infrastructure could be defined as: a con- should first have a preliminary binder blend design
struction that maximizes the recycling of waste/sec- (such as in WP2) that take into account concepts
ondary materials within its structure and minimizes such as degree of blending of the RA, as well as
the environmental impacts, through the reduction of allowing the use of additives. In WP3, the project
energy consumption, natural resources and associ- provided also an End User manual to produce the
ated emissions while meeting performance condi- investigated technologies. This provide best prac-
tions, specification requirements and social needs tices at the asphalt plants to maximize RA use
throughout its entire life cycle. With this concept within new road pavement without compromis-
as driver, within the WP5 led by the University of ing its performance. Within WP4 plant-produced
Nottingham, the consortium closely collaborated asphalt mixes have been investigated, as well as
with industrial partners and sustainability profes- recovered binders, and indication on the validity
sionals to collect data and sustainable practices for of the undertaken mix design are given. At last,
providing road authorities with tools to be used for the WP5 provided CEDR an answer to the ques-
further designs and to performing a broad sustain- tion: Is maximizing recycling of asphalt into pave-
ability assessment of the investigated technologies. ment a sustainable practice? It was proved that
This work package produced the following reports: maximizing reclaimed asphalt can be a sustainable
D5.1: A state of the art review of existing sustain- option, however in some of the case studies other
ability assessment tools of the impact of road variables were more important than recycling.
pavement infrastructures; D5.2: Evaluation of the These were maximizing durability of the asphalt
environmental impact (through LCA technique) mixes and decreasing transport distances during
and economic impact (through LCCA techniques) construction and maintenance (100 Km is a good
of the defined technologies taking into account the reference).
European level of the project and adapted to real The authors are aware that most of the results
case studies; D5.3: Sustainability assessment of the not included here. This is because approval from
AllBack2Pave technologies adapted to real case CEDR is still pending. Results will be soon avail-
studies at European level, through a methodology able in the project website and at the conference.

116

TIS2017_Book.indb 116 2/8/2017 8:02:04 PM

 ACKNOWLEDGEMENTS Direct-MAT. 2011. Dismantling and recycling tech-
niques for road materials. Available at: http://www.
The research presented in this paper was carried direct-mat.eu
out as part of the CEDR Transnational Road Jiménez del Barco Carrión, A. Lo Presti, D. Airey, G.,
2015 “Binder design of high RAP content hot and
research Programme Call 2012. The funding for the warm asphalt mixture wearing courses” - Road Mate-
research is provided by the national road adminis- rials and Pavement Design, Taylor&Francis,, Special
trations of Denmark, Finland, Germany, Ireland, Issue: EATA 2015
Netherlands and Norway. Lo Presti, D. Jiménez del Barco Carrión, A. Airey, G.
Hajj, E. 2016. Towards 100% recycling of reclaimed
asphalt in road surface courses: binder design method-
REFERENCES ology and case studies: Journal of Cleaner Production.
131, 43–51
Allback2Pave D2.1; D3.1; http://allback2pave.fehrl.org Re-Road. 2012. End of life strategies of asphalt pave-
(accessed Sep 16). Report D4.x and 5.x will be avail- ments. Available at: http://re-road.fehrl.org
able soon. RDO Asphalt 09. 2009. Richtlinien für die rechnerische
Blasl, A. Kraft, J. Lo Presti, D. Di Mino, G. Wellner, F. Di
2016. Performance of asphalt mixes with high recy- TL AG-StB 09. 2009. Technische Lieferbe-dingungen für
cling rates for wearing layers, EE congress 2016 Asphaltgranulat. Köln. FGSV Verlag GmbH.
Dauvergne, M., Jullien, A., Proust, C., Tamagny, P., TL Asphalt-StB 07. 2007 Technische Lieferbedingun-
Ventura, A., Coelho, C., et al. ECORCE M User’s gen für Asphaltmischgut für den Bau von Verkehrs-
Manual. Nantes, FR: French Institute of Science and flächenbefestigungen. Köln. FGSV Verlag GmbH.
Technology for Transport Development and Networks
(IFSTTAR, 2014).

117

TIS2017_Book.indb 117 2/8/2017 8:02:04 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Feasibility and preliminary design of a new railway line in the

Dolomites area of Veneto Region

M. Pasetto, G. Giacomello, E. Pasquini & A. Baliello

Department of Civil, Environmental and Architectural Engineering, University of Padua, Padua, Italy

ABSTRACT: The current conformation of the railway line situated in the Dolomites area of Veneto
Region does not represent an efficient network to fully connect the territory and serve the global mobil-
ity demand. In fact, the existing railway line lies along the Piave valley and connects cities on the Veneto
plain with Belluno and the Cadore area, whereas public road transport links the train stations with the
other towns in Belluno province. Given this background, a new railway line (from Calalzo to Cortina
d’Ampezzo passing through Auronzo) could meet the province mobility needs, leading to a potential
increase in traffic and thus improving economic and tourism development in the area. Indeed, historical
as well as natural and tourist (ski) areas would benefit. Moreover, a new railway line is an attractive and
valid alternative to the development of the existing road network, allowing a proper integration between
rail and road transportation systems. In this sense, a preliminary design has shown noise levels at least
halved with respect to a road context eliminating any level crossing, thus also proving to be a convincing
solution under the environmental point of view. However, a more specific and detailed assessment of
the environmental impact of the new railway line will be necessary taking into account that the region
involved is one of the most scenic Dolomite areas (UNESCO Heritage Site), with complex morphol-
ogy and geology. A 45 km long single-track railway layout has been designed hypothesizing two pas-
senger trains per hour and assuming a minimum speed of 80 km/h, a minimum bend radius of 300 m, a
maximum longitudinal slope of 20‰ and gauge of 1435 mm. A total cost of 500 million euros has been
estimated for the realization of the new infrastructure.

1 INTRODUCTION Alto Adige and Friuli Venezia Giulia Regions and

is a prevalently mountain territory. Although it is
The territory of the province of Belluno has always the largest province in Veneto (3678 km2), it is the
been penalized by its distance from the industrial- one with the fewest inhabitants (206,856) (ISTAT
ized Veneto plain because of the area’s orographic 2013). Compared with other not easily reached
and geomorphological characteristics. In particular, mountain areas, however, the area is fairly well
the dolomitic territory of Cadore and the Ampezzo populated and characterized by large municipali-
Valley, in the central-northern part of the province, ties uniting sparse hamlets.
enjoys a limited north-south accessibility. In the The two most populated dolomitic areas of the
other directions, it has a system of links and a viabil- province and the most important for tourism are
ity typical of mountain areas (mainly via valleys). Cadore and the Ampezzo Valley. The former is a
The increase in traffic that has been registered all well-defined morphological and historical unit cor-
over Italy in recent years has also affected Cadore responding to the upper basin of the river Piave
and the Ampezzo Valley, where there has been a upstream of the village of Longarone. The latter
rise in tourist numbers, an increase in the mobility corresponds to the valley of the Boite torrent and
of the residents for regular (home, work, school) historically has always been linked to Cadore.
and sporadic travel (free time) and the movement The area of Cadore and the Ampezzo Valley
of goods. The road network carries steady flows of lies in the Dolomites, characterized by two associ-
traffic along the valleys, which gradually decrease ated types of rocks: dolomitic and volcanic, which
leaving the main valleys for the peripheral areas. derive from totally different processes. The dolo-
mitic rock is more resistant to atmospheric agents
than the rocks of volcanic origin, which erode
1.1 The territorial setting: the province of
easily: the results are high dolomitic peaks over-
Belluno, Cadore and the Ampezzo Valley
looking deep valleys. Because of this morphology,
The province of Belluno is in northern Italy, in the in many parts of this area and along the valleys
north of the Veneto Region, between the Trentino (especially that of the Boite) there are phenomena

119

TIS2017_Book.indb 119 2/8/2017 8:02:07 PM

 of instability of the mountainsides and surface characteristics and dimensions that allow them to
subsidence. The distinctiveness of these mountain fully integrate into the territory with the possibil-
areas has led to their classification as “Sites of ity of offering places and landscapes from a differ-
Community Importance” and “Special Protection ent perspective, in which the natural and historical
Areas”. background that has led to the formation of this
environment is clear.
1.2 The economic-social context of the
Dolomites area 1.3 The mountain railway in the context
The principal municipalities in the area are: Cala- Although major international railways lines exist
lzo di Cadore (approx. 2,250 inhabitants, 43 km2 in the north-east of Italy (like the Brennero line
in area), Domegge di Cadore (2,441 inhabitants, and the Pontebbana railway) that connect the plain
50 km2 in area), Lozzo di Cadore (1,383 inhabit- areas and the Adriatic with the Alpine area and the
ants, approx. 30 km2 in area), Auronzo di Cadore regions of central Europe, the Dolomite areas of
(approx. 3,350 inhabitants, 220 km2 in area), San Cadore and the Ampezzo Valley are not directly
Vito di Cadore (1,857 inhabitants, approx. 61 km2 crossed by these lines.
in area) and Cortina d’Ampezzo (5,907 inhabit- The province of Belluno is currently served by
ants, approx. 252 km2 in area) (ISTAT 2013). two lines that connect Belluno with Padova and
The main economic activities in the area are Venice. The stretch of railway Belluno—Calalzo,
livestock rearing, forestry and wood processing, built for military purposes in 1914, has the charac-
but the most important sources of wealth are the teristics of a mountain line: single-track, not elec-
many companies that produce glasses and the tour- trified, 25‰ maximum slope, bend radii of 200 m
ist sector. Indeed, tourism (both summer and win- (Agostini 2014).
ter) is of enormous economic importance in these In the past (1921–1964) a railway line existed
areas. According to the 2012 data (ISTAT 2013), (known as the “Dolomites railway—Ferrovia
the province of Belluno accounts for approxi- delle Dolomiti”), which connected Calalzo di
mately 86% of tourist presences in the mountain Cadore with Cortina d’Ampezzo and terminated
zone of Veneto Region. In terms of tourist num- at Dobbiaco. Nowadays, the stretch Dobbiaco—
bers, Auronzo di Cadore accounts for about 70% Cortina d’Ampezzo is a footpath, while the stretch
of the entire Cadore and about 7% of that of the Cortina—Calalzo has been almost entirely con-
province of Belluno. Cortina d’Ampezzo instead verted into a cycle path. The Dolomites railway
attracts around 25% of the annual tourist pres- involved a total rise of 810 m and was 65 km long,
ences in the province of Belluno. Cadore alone had a maximum slope of 35‰, a minimum bend
accounts for 42.3% of the entire mountain tourist radius of 60 m, a narrow gauge of 950 mm and
sector of the Veneto Region. electrification at 3000 V (Gaspari 2005).
The large volume of tourists who visit the Dolo- In this context, the general rediscovery of rail
mites travel in their own cars. It emerged from an transport in the last decades has also stimulated, in
ISTAT study that the majority of travel in the area the municipalities of the Dolomites, the idea that
is by private transport, with negative effects on the the train could be the ideal solution to solve the
environment due to air and noise pollution and on problems of mobility and contemporarily enrich
traffic in general. the tourist offering.
There is a growing need to upgrade the mobil- This rediscovery is supported by examples of
ity of the territory, with a modern public transport transport systems integrated in the landscape
system to complement and integrate use of the pri- and still functioning, such as: the Stubaital rail-
vate car. However, this is a situation for which, at way (1904) that unites Innsbruck with the district
the moment, there is no easy solution in the short- of the valleys of the Stubai Alps in Austria; the
medium term. This fact will probably be strongly Bernina Express (1880) that covers the stretch
dependent on a series of dynamics, already under- between Tirano, Saint Moritz and Chur (Swit-
way, but that will increasingly affect the situation zerland) in an itinerary of outstanding land-
in the territory of the province of Belluno. The scape with slopes of up to 70‰ (Wall 1987); the
demographic decline, ageing of the resident popu- Trento–Malè–Marilleva railway (1909) in Trentino
lation, depopulation of the more marginal areas Alto Adige (Forni 1975, Serra 1989); the Zillertal
and concentration of the population in the most railway (1902) that crosses the valley of the Ziller
important centres, the rise in tourist flows and the uniting Jenbach with Mayrhofen in the Austrian
increasingly difficult economic management of the Tyrol; the Merano-Malles railway (1906) along
public services. the Val Venosta (Marseiler 2006), a railway closed
In order to avoid the increase of these prob- through lack of use, then reopened and now grow-
lems, the transport infrastructures must assume ing strongly thanks to efficient management.

120

TIS2017_Book.indb 120 2/8/2017 8:02:07 PM

 These examples demonstrate that a railway has Ansiei passing through Auronzo di Cadore. In the
the potential to provide an efficient and sustain- vicinity of the Somadida forest the line enters a
able response to the mobility needs in the area of tunnel, exiting at Chiapuzza (hamlet of San Vito
the present study. di Cadore) and, traversing the valley of the Boite
torrent, ends at Cortina d’Ampezzo (Fig. 1).
The fundamental elements that characterize a rail-
2 FEASIBILITY STUDY OF THE way route are essentially the minimum radius of the
PLANNED ROUTE bends and maximum slope of the gradients. These
elements are set according to the maximum speed
2.1 Project hypothesis and characteristics of the trains that will use the line.
Were therefore defined: the stops and stations
The hypothesized railway route runs for about
(Table 1), minimum bend radius (300 m) and maxi-
45 km in an area with complicated morphology that
mum slope (not above 18‰).
implies very high construction costs (for tunnels
and bridges). Otherwise, costs may be compensated
in a relatively short period of time if an increased Table 1. Stations and stops along the designed railway
tourist flow and the scenic beauty of the places tra- line.
versed by the infrastructure are considered.
Station or Partial
The tourism interest in this project is even
Station name stop distance (m)
clearer if a possible extension of the line to Misu-
rina is considered, where the parking places for Calalzo di Cadore station 0
visitors to the Tre Cime di Lavaredo have become Domegge di Cadore stop 3,400
insufficient, or further north towards Dobbiaco, in Lozzo di Cadore station 3,000
Trentino Alto Adige. Cima Gogna stop 4,700
It is hypothesized that the railway will be single- Auronzo Centro station 4,500
track, with trains travelling in opposite directions Auronzo Impianti stop 2,000
passing one another in the stations. Somprade stop 3,400
Given the orographic characteristics of the San Marco station 6,000
areas to be crossed, a railway line was designed, Briglie-Somadida stop 3,400
with numerous tunnels and bridges. It begins from San Vito di Cadore station 6,600
the existing station at Calalzo di Cadore at an alti- Acquabona stop 4,600
tude of approximately 740 m. The route runs along Cortina d’Ampezzo station 3,700
the valley of the river Piave and then of the river

Figure 1. Designed railway layout: project (white), prolongation of the railway line towards Dobbiaco (grey) and
prolongation of the railway line towards Misurina (dotted grey).

121

TIS2017_Book.indb 121 2/8/2017 8:02:07 PM

 The maximum superelevation allowed on a after the municipality of Lozzo, onto the left bank
bend is 160 mm for speeds less than 160 km/h. over a bridge).
For bend radii, greater than those corresponding The route will then return to the right bank just
to the maximum superelevation, the non-compen- before the confluence of the river Ansiei. The route
sated acceleration reduces in proportion to the lies parallel to the valley of the river Ansiei along
superelevation. the left bank. The line runs to the west of Cima
The planned railway line, according to the Gogna (hamlet of the municipality of Auronzo
schedule of the Italian Railway Network (Rete Fer- di Cadore), with a stop planned at the end of the
roviaria Italiana - RFI), is of “type B” (secondary hamlet (easily reached and close to the intersection
lines). The route has 43 straight stretches and 41 with the road to Val Comelico).
round bends linked by transition bends on entering After this the railway line, over a succession of
and leaving. stretches in the open and through tunnels, reaches
The railway line leaves from the existing station the municipality of Auronzo di Cadore. Two sta-
at Calalzo with a single not electrified track. Its tions are planned at Auronzo: one in the town cen-
route is planned first along the right bank of the tre (“Auronzo Centro”) and one next to the ski lifts
river Piave and then of the river Ansiei. The rail- (“Auronzo Impianti”). The two stations are on the
way axis has been designed with the intention of right shore of the Santa Caterina Lake and, in the
maintaining an adequate distance between infra- project, are designed to be in tunnels: the “Auronzo
structure and buildings and between infrastructure Impianti” station is at an altitude of 850 m a.s.l.,
and the right bank of the Piave. the same as that of the ski lifts.
The station at Calalzo could also be used as ter- From the station “Auronzo Impianti”, the line
minus by the trains, because it already has at least passes over a bridge on the Val da Rin and another
three tracks to allow trains travelling in different on the left bank of the river Ansiei and then arrives
directions to pass one another. From Calalzo, the at the station of Somprade (Fig. 2). After this,
line crosses the river Molinà over a bridge and con- the route passes over a small viaduct, an ample
tinues in the open along the right bank of the Piave stretch in the open following the valley of the river
remaining at some distance from the hamlet of Ansiei and a long tunnel, after which it arrives at
Vallesella (for which no stop is planned). The stop in the station of San Marco. The route then returns
the municipality of Domegge di Cadore is planned to the right bank of the river Ansiei over a bridge,
at some distance from its centre, so access will entering the Somadida forest, where a station at
have to be arranged. Between the municipalities of Briglie—Somadida is planned. This station, con-
Domegge and Lozzo di Cadore the railway line runs ceived according to a model at low environmental
parallel to the Piave through a tunnel. The station in impact, will serve tourists visiting the reserve situ-
the municipality of Lozzo is planned slightly before ated in a splendid area of the Dolomites. The route
the built-up area and, having two tracks, will allow then enters the longest tunnel (about 5 km) and
the passing of trains travelling in opposite direc- exits in the valley of the Boite at Chiapuzza (hamlet
tions. Immediately after the station at Lozzo there of the municipality of San Vito di Cadore), where
are two critical points: the intersection with the a station is planned. From here the line arrives at
state road “Alemagna” (which will be raised to the station of Acquabona, in the lower part of the
allow the line to pass underneath and maintain valley. The route then crosses the industrial zone of
the design slopes) and the Ruoiba landslide, on the Zuel in a tunnel and, remaining on the left bank of
right-hand side of the river Piave. In that point the the Boite torrent, terminates at the station of Cor-
valley becomes very narrow and enclosed, forming tina d’Ampezzo, situated next to the cross-country
a “V” (the railway route will be shifted, immediately skiing trails in the south-east of the town.

Figure 2. Crossing of the Val da Rin and river Ansiei, arriving at the Somprade station (gallery in black and bridge
in dotted white).

122

TIS2017_Book.indb 122 2/8/2017 8:02:08 PM

 Figure 3. Detail of the longitudinal profile between Calalzo station and Auronzo Impianti stop.

Table 2. Steady gradients of designed railway line. To reduce the transversal dimensions of a sec-
tion (in embankment or cutting), shallow slopes
Steady gradients Slope (‰) Length (km) supporting the soil with scarp or counter scarp
walls can be built (in masonry or simple or rein-
1 0.0 0.1
2 3.0 1.7
forced concrete). For the bridges, it can be planned
3 5.5 2.9
that the crossing over a waterway, road or railway,
4 0.5 2.2 will be at an angle of 90°, but the railway route
5 6.0 2.0 often does not allow this so the bridge must be
6 13.0 6.2 built diagonally. Bridges also exist on bends in the
7 10.5 2.6 route. The choice of bridge structure to be built is
8 18.0 14.2 directly linked to each individual situation. In the
9 0.9 5.0 case of tunnels a profile limit is stipulated within
10 12.0 1.5 which every vehicle including transported loads
11 18.0 4.1 must remain.
12 12.0 2.9
2.2 The rolling stock and infrastructure
The rolling stock was chosen in order to meet serv-
From the altimetric point of view, the slopes ice needs (times) and to define the design parameters
assigned to the route are lower than the maximum of the infrastructure. The requisites identified are:
prescribed value (18‰). Figure 3 shows a part of travel times (maximum speed and high acceleration
the longitudinal profile whereas Table 2 reports all and braking performances), journey comfort (noise,
the slopes utilized and the lengths of the gradients. vibrations, air-conditioning, passenger information
More specifically, between Calalzo and Auronzo and entertainment), accessibility (flush access, large
slopes are very gentle (the steepest is 13‰), while doors for easy and rapid mounting and dismount-
between Auronzo and Cortina the slope reaches ing of the passengers, multifunctional spaces for
18‰ (in the tunnel between Somadida forest and the transport of bicycles or skis), environmental
Chiapuzza it remains below 1‰). impact.
The construction of embankments, cuttings, Among the many types of trains that meet
bridges and tunnels is planned. The gradient the above requisites, the powered railway car-
assigned to the slopes depends on the angle of fric- riages produced by the Swiss Stadler Rail “GTW
tion of the soils but is generally set at 2/3 on an 2/6” and “GTW 4/12” were chosen. The first is a
embankment and 1/1 in a cutting. light articulated railway carriage useable for local

123

TIS2017_Book.indb 123 2/8/2017 8:02:09 PM

 transport, with two driving axles (located in the 2.3 The timetable and transport service
central section) on a total of six axles, and offers
The train timetable was drawn up on the basis
111 seats. The second, with four driving axles
of the localities to be served and the other data
(located in the central section) on a total of twelve
hypothesized for the project. The timetable defines
axles, instead offers 245 seats. The GTW 4/12
the route, travel times, performances of the services
trains are two GTW 2/6 permanently coupled via
(spatial accessibility, temporal accessibility and
an intercommunicating passage accessible also to
commercial speed), the attractiveness of the serv-
the passengers.
ices, the capacity of the system to be competitive
Once the type of train had been decided, it was
with other methods of transport, and the stops.
possible to classify the railway line (the category
The presence of an optimal number of stops/
determines the characteristics of the superstruc-
stations:
ture) on the basis of “maximum weight per axle”
and “maximum load per metre allowed” (i.e. the – offers better cover of the territory with a greater
ratio between the total weight of the loaded vehi- number of departures and destinations;
cle and its length including buffers). The planned – increases the attractiveness of rail transport also
railway line can be classified in category B2, with a for travel from, to and between secondary desti-
maximum weight per axle of 18 tons and a weight nations, guaranteeing higher profitability of the
per unit length of 6.4 tons per metre. service (fuller trains), also in small places that
According to the RFI regulations (Mayer lack attractors of “point-to-point” traffic;
2004), the line can also be classified depending on – allows a larger swathe of the population to have
the intensity of traffic measured by the dummy access to the comfort of rail transport.
load, expressed in gross tons hauled daily [gth/d].
The presence of a railway stop signifies, espe-
RFI has divided the lines on the basis of the
cially in rural and mountain ambits, promotion of
dummy load starting with group 1 (with a load
the territory, as it greatly modifies the temporal
above 102,000 gth/d) up to group 9 (with a load
geography, i.e. the time necessary for journeys.
less than 1,000 gth/d). This line can be classified
In this study, an estimate was made of the travel
in group 9.
time necessary to reach the station at Cortina
The characteristics of the infrastructure (Fig. 4)
d’Ampezzo leaving from Calalzo di Cadore and
were hypothesized as being those of a secondary line
also an estimate of the number of trains that can
according to RFI specifications (Mayer 2004): rails
transit in an hour in both directions.
type UIC 60, with type R200 steel (type R260 on iso-
The inputs used in the calculation were: place of
lated bends with radius less than 600 m), gauge of
departure (Calalzo di Cadore), destination (Cortina
1,435 m, pre-compressed reinforced concrete sleep-
d’Ampezzo) and vice versa, distance travelled between
ers (module 0.66 m, weight 2,500 g, length 2.3 m and
every stop/station, travel time of these distances, stop-
trapezoidal section variable in height, with maximum
ping times at the single stops/stations (4 minutes in
dimensions at the heads and reduced in the centre),
the stations with more than one track, to allow trains
indirect joints between sleepers and rails, staggered
coming from opposite directions to pass, and 2 min-
joints between the rails, stone ballast (basalt, por-
utes in the other stops), average speed.
phyry or granite rocks, with sharp edges and almost
The travel time between Calalzo di Cadore and
uniform size 25–30 mm in diameter, ballast depth
Cortina d’Ampezzo results as being 62 minutes.
0.35 m), settlement plan (i.e. the support for the bal-
Moreover, the time necessary for two trains to pass
last that follows the planimetric trend of the route),
one another in opposite directions at the stations of
useful life of the superstructure of 30 years.
Lozzo di Cadore and San Vito di Cadore is about
The arrangement of the substructure for the
50 minutes whereas at Auronzo Centro and San
construction of an embankment will consist of the
Marco is about 35 minutes. Therefore, in 1 hour,
prior removal of the vegetated layer of soil, to a
two trains can transit in opposite directions, with a
depth of at least 20 cm. Analyses of the soils will
time separation of 28 minutes (Fig. 5).
be conducted to learn the carrying capacity and
It was hypothesized that the transport service
mechanical characteristics.
could be divided on the basis of working and holi-
day periods: on weekdays, a service would connect
all the stops (from Calalzo to Cortina and vice
versa, focused on local people and tourists, with
the above timetable), while at weekends and on
holidays there would be a service to Venice (with a
maximum of 3 trains per day that connect Cortina
directly with Venice). This would create a tourist
service for the passengers of cruise ships that arrive
in Venice, offering the opportunity of a daytrip to
Figure 4. Cross section of the designed line. Cadore and the Ampezzo Valley.
124

TIS2017_Book.indb 124 2/8/2017 8:02:10 PM

 2.4 The estimate of the costs struction characteristics of the line, the localities
to be served were decided first, then the number
The advantages of a railway line with respect
of tracks, type of line, type of rolling stock, type
to a road are the possibility of providing more
of station (simple stop or station with double the
transportation, the smaller surface area necessary
number of tracks for trains to pass one another).
(for the same amount of traffic), the greater com-
The overall cost depends on the values of maxi-
fort, less congestion of the infrastructure due to
mum slope, length of the route, number of struc-
traffic and less influence of atmospheric agents.
tures, and in particular tunnels, that lead to the
Against these advantages it is necessary to take
cost per kilometre. The running costs were calcu-
into account the economic and social magnitude of
lated taking into account the capacity of the line
the investments and therefore conduct an evaluation
(number of trains that can circulate in the unit of
of the choices with a careful analysis of the costs,
time and train speed).
impacts and benefits. The construction costs of a
The following construction cost was estimated:
railway are always high, as are the fixed costs that
9.5 million euro for each bridge (with a fixed cost
must be sustained independently of the amount of
equal to 160 million euro), 8.95 million euro for
traffic. It follows that a railway line can only be justi-
each tunnel, 0.1 million euro per km for train
fied economically when traffic is sufficiently intense.
control system and other technologies, 4 million
It was decided to insert in this feasibility study
euro for each railway vehicle (there are a total of 5
a comparison between the construction costs
trains), 1 million euro for each station/stop (there
(Table 3), running costs and revenues that can be
are 12 stations/stops).
expected, also within the perspective of a wider
Given its insertion in a delicate area, the total
territorial plan or growth of international and
construction cost was recalculated utilizing a cor-
intermodal traffic.
rective factor of 1.4 (Law n. 443 of 2001) and the
Costs and revenues are difficult to determine
cost of the 12 stations (12 million euro), train con-
in a uniform way as the former depend on the
trol system (4.54 million euro) and trains (20 mil-
construction characteristics and the latter on the
lion euro) were added, thus obtaining a total
type of service. As the costs depend on the con-
construction cost of the entire infrastructure of
509.09 million euro.
The maintenance costs of the railway line were
estimated utilizing a standard value of 7,000 euro/
km that, for the total length of 46 km, is 322,000
euro/year. For maintenance of the trains a cost of
625,000 euro/train was hypothesized, which for 5
trains becomes a cost of 3.13 million euro per year.
To calculate the earnings a service was hypothe-
sized, as indicated in the previous section, composed
of: 3 trains from Venice to Cortina d’Ampezzo
(and vice versa) on weekends and holidays, and 20
trains per day from Calalzo di Cadore to Cortina
d’Ampezzo (and vice versa) on weekdays. Were also
estimated: the average capacity (seats plus stand-
ing room) of a GTW 4/12 train equal to approxi-
Figure 5. Timetable of designed railway line.

Table 3. Construction costs.

L* E/T** B*** Tu**** Total

Railway line stretch km million euro

Calalzo di Cadore –
Auronzo di Cadore 17.7 10.5 € 49.9 € 78.5 € 138.9 €
Auronzo di Cadore –
Briglie-Somadida 14.2 8.1 € 36.1 € 58.1 € 102.3 €
Briglie-Somadida –
San Vito di Cadore 5.0 0.0 € 0.0 € 44.7 € 44.7 €
San Vito di Cadore –
Cortina d’Ampezzo 8.5 6.3 € 11.8 € 33.5 € 51.6 €
Total 45.4 24.9 € 97.8 € 214.8 € 337.5 €

*Length, **Embankment/Cutting, ***Bridge, ****Tunnel.

125

TIS2017_Book.indb 125 2/8/2017 8:02:10 PM

 mately 300 people, a one-way ticket from Venice to – a single-track railway line requires an overall
Cortina of about 25 euro per person and a one-way width of around 4 metres with an impact com-
ticket from Calalzo to Cortina of about 5 euro. Dif- parable more to a cycle path than to a road;
ferentiating the income by summer season, winter – lower level of noise generated, less than that of a
and during autumn/spring, an annual income was vehicle on a road and concentrated in the moments
calculated for the passenger service of approxi- when a train passes 2 or 4 times per hour;
mately 13 million euro. Subtracting annual cost for – no level crossings are planned as the railway line
the maintenance of the line and trains from this, a and the existing road never intersect.
value is obtained from which the number of years
Further studies are anyway required so that the
to amortize the initial construction costs of the line
railway can become an integral and sustainable part
is derived, which in this case is about 50.
of the territorial context. Following the approach
of other researchers (Keshkamat et al. 2009, De
Luca et al. 2012), it will also be useful to conduct a
3 CONCLUSIONS study utilizing GIS and multi-criteria analysis.
The creation of the new railway line would be
The railway connection between Calalzo di Cadore, further stimulated if it was decided to build a rail-
Auronzo di Cadore and Cortina d’Ampezzo would way connection between Cortina d’Ampezzo and
respond to some needs, such as: Misurina or else to extend the planned railway
– improving connections between the municipali- line from Cortina d’Ampezzo to Dobbiaco (where
ties of Cadore and the Ampezzo Valley; the Ferrovia delle Dolomiti existed). In this way
– activating new tourist circuits to integrate with the number of passengers per kilometre could
those already existing and that spread to nearby markedly increase, encouraging the development
areas; of new accommodation and sports facilities, with
– creating an alternative to investments in the the creation of new jobs and an environment more
road sector in the Dolomites area, increasingly conducive to demographic growth.
opposed by the local populations;
– encouraging a multimodal system of road/rail
traffic for both people and goods. REFERENCES
This new stretch of railway would interest a Agostini, A. 2014. 1914-2014, Belluno—Calalzo di
pool of users corresponding to the resident pop- Cadore: Una ferrovia nel cuore delle Dolomiti. Belluno:
ulation and a tourist sector that is busy in both Momenti AICS.
summer and winter, which would constitute the De Luca, M., Dell’Acqua, G. & Lamberti, R. 2012. High-
basic prerequisite for the development of tour- speed rail track design using GIS and multi-criteria
ism and production in the areas involved. The analysis. Procedia—Social and Behavioral Sciences.
railway would also encourage a reorganization of 54: 608–617.
Forni, M. 1975. Piccola storia della ferrovia Trento-Malè.
the existing network of infrastructures that would
Ingegneria Ferroviaria 30(11): 70–75.
allow the improvement and enhancement of the Gaspari, E. 2005. La ferrovia delle Dolomiti: Calalzo,
historical centres, natural areas and ensure greater Cortina d’Ampezzo, Dobbiaco, 1921−1964. Bolzano:
functionality of the new residential and produc- Athesia.
tive areas. ISTAT. 2013. Veneto. Roma: ISTAT.
The development of a railway line that connects Keshkamat, S.S., Looijen, J.M. & Zuidgeest, M.H.P.
Venice with Cadore and Cortina also aims to cre- 2009. The formulation and evaluation of transport
ate a wider offering for the tourists who arrive in route planning alternatives: a spatial decision support
Venice, giving them the opportunity to make a system for the Via Baltica project, Poland. Journal of
Transport Geography 17:54–64.
daytrip to Cadore and the Ampezzo Valley.
Marseiler, S. 2006. Cent’anni di ferrovia in val Venosta:
It must anyway be taken into account that 1906–2006. Bolzano: Assessorato al Turismo ed alla
the zone involved in the study is one of the most Mobilità.
striking areas of the Dolomites from the land- Mayer, L. 2004. Impianti Ferroviari. Roma: CIFI.
scape-environmental point of view, characterized Serra, M. 1989. Il prolungamento della ferrovia elet-
by a highly complex morphology and geology. trica Trento-Malè. Ingegneria Ferroviaria 44(1–2):
The environmental impact caused by a railway 54–58.
line is certainly less than that of a new fast road Wall, H. 1987. Bernina Express: in treno attraverso le
because: Alpi. Salò: ETR.

126

TIS2017_Book.indb 126 2/8/2017 8:02:10 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Investigation of the causes of runway excursions

N. Distefano & S. Leonardi

Department of Civil Engineering and Architecture, University of Catania, Catania, Italy

ABSTRACT: The risk of runway excursion dependent on multiple factors related to operating condi-
tions. These include Runway Contamination, Adverse Weather Conditions, Mechanical Failure, Human
Error.
A multivariate analysis of historical data on accidents on runways carried out in order to quantify the
effect that various factors have upon runway excursions (landing and takeoff overruns and landing and
takeoff veer-offs).
In this paper, an in-depth data study was conducted of all runway excursion accidents over the period
spanning 2006–2015 in 8 geographical regions to investigate the causes of runway excursion accidents. All
data in this study are from the Aviation Safety Database, published by Aviation Safety Network (ASN) by
Flight Safety Foundation, and have been augmented by appropriate investigative reports when available.
The technique deployed in this work is the multiple logistic regression model, often used in recent lit-
erature and proved suitable to examine and quantify the effect of various factors on accidents risk. This
technique revealed interesting relationships among the variables both for landing accidents and for take-
off accidents. The goal of a logistic regression analysis is to predict correctly the outcome for individual
cases using the parsimonious or least complex model.
The results of this paper show that the main cause of landing veer-off is the mechanical failure in most
of the regions for any type of aircraft; the leading cause of landing overrun is human error in all regions
especially for small aircrafts: the takeoff overrun mainly are caused by mechanical failure, especially for
very large aircrafts. The weather conditions assume a dominant role, especially for small aircraft and for
overrun accidents.

1 INTRODUCTION identification of recurring patterns or risk factors

that are not always apparent when individual acci-
A major initiative to improve safety is the increas- dents are investigated. More recently, proactive
ing role of Safety Management Systems (SMS), approaches to determining ways to improve safety
intended as “an organized approach to manag- have become increasingly popular. An example of
ing safety, including the necessary organizational such a proactive approach is the analysis of inci-
structures, accountabilities, policies, and proce- dent data to identify areas of increased risk that
dures” (ICAO, 2012). ICAO has championed this may lead to an accident (Oster et al. 2013)
initiative, which is now included in international A large number of air accidents occur dur-
standards for airline operations (ICAO, 2009). ing the take-off and landing phases. Most occur
The basic structure involves four “pillars”: iden- beyond the designated safety and protection areas,
tifying safety hazards; safety risk management around the runway, when an aircraft overruns the
through remedial actions to address safety risks; runway-end during take-off or landing, or when it
continuous monitoring and assessment of the undershoots the runway, with regard to the thresh-
safety level sought and achieved; and programs old, during landing.
for continuous improvement in the overall level The most common accidents that occur during
of safety. these flight phases are:
Another major contributor to the improved
safety record can be traced to the careful investi- – Landing overrun (LDOR—LanDing OverRun).
gation of past accidents to determine what led to – Landing veer-off (LDVO—LanDing Veer-Off).
the accidents and what needs to be done to pre- – Landing undershoot (LDUS—LandDing
vent such events from occurring again. This reac- UnderShoot).
tive approach to improving aviation safety has – Take-off overrun (TOOR—Take-Off OverRun).
been enhanced by the thorough analysis of data – Take-off veer-off (TOVO—Take-Off Veer-Off)
from numerous accidents, which has aided in the – Ground collision after take-off.

127

TIS2017_Book.indb 127 2/8/2017 8:02:12 PM

 Runway excursions during take-off and landing Oster jr. et al. (2013) says there is still a role for
continue to be the highest category of aircraft acci- careful accident investigation and there are still
dents and often exceed 25% of all annual commer- lessons to be learned from the few accidents that
cial air transport accidents (IATA 2011). these carriers have. But with improvements in
Runway excursions can result in loss of life and/ safety and major reductions in accidents, airline
or injury to persons either on board the aircraft or safety analysis will have to shift toward analysis
on the ground. The effect of runway excursions can of incident and operational data with the intent
result in damage to aircraft, airfield or off-airfield of identifying safety risks before accidents occur.
installations including other aircraft, buildings or There are two potential benefits from looking at
other items struck by the aircraft. these types of data. One is to address the ques-
A runway excursion accident is defined as an tion of why some sequences of events result in
accident where an aircraft on the runway surface accidents while other sequences do not result
departs the end of the runway or side of the run- in accidents. A better understanding of how
way surface during take-off or landing (IATA potential accidents were avoided in some situ-
2011). ations may lead to more such avoidances in the
It consists of two types of events: future. A second potential benefit is to identify
trends in or the emergence of potentially hazard-
– Veer Off: A runway excursion in which an air-
ous sequences of events before they result in an
craft departs the side of a runway
accident. Here again, by identifying such trends,
– Overrun: A runway excursion in which an air-
it may be possible to take corrective action before
craft departs the end of a runway
an accident occurs.
It excludes both accidents where the aircraft did Distefano et al. (2014) proposes a risk assess-
not initially land on a runway surface, and take- ment procedure in which the probability of each
off excursions that did not start on a runway (e.g., accident is proportional to the cumulative prob-
inadvertent take-offs from taxiways). ability of the causes identified for the accident.
The current definitions of runway overrun do Čokorilo et al. (2014) has produced a model
not specify any distance for arrivals and departures in order to estimate the frequency of aviation
where an incident can not be considered to be an accidents in different environmental conditions
overrun or undershoot, because it is too far out and was based on the physical characteristics of
from the runway. For example, the current defini- the aircraft. The data were processed and then
tions used by Eurocontrol are: aggregated into groups, using cluster analysis
based on an algorithm of partition binary ‘Hard
– Overrun on Take-Off: ‘‘A departing aircraft fails
c means.’
to become airborne or successfully reject the
Ayres jr. et al. (2012) shows that location models
take-off before reaching the end of the runway”.
could be improved if greater attention was paid to
– Overrun on Landing: ‘‘A landing aircraft is
causality but data difficulties exist, for example, on
unable to stop before the end of the runway is
meteorological influences on overrun distance and
reached”.
that this is often mis-recorded in accident dockets
The common taxonomy group of the ICAO so lateral deviations are more difficult to model;
commercial aviation safety team refers to: Runway while in consequence modelling the variation in
excursion as veer off or overrun off the runway aircraft type, wingspan and speed ought to be
surface. This definition is only applicable during included as well as pavement type variations that
either the take-off or landing phase. The excursion will affect deceleration.
may be intentional or unintentional (for example, Wagner et al. (2014) focuses his work on predict-
deliberate veer off to avoid a collision, brought ing if the excursion will generate fatalities: human
about by a runway incursion). This classification errors are the strongest associated feature with
applies in all cases where the aircraft left the run- fatal excursions, another feature strongly associ-
way regardless of whether the excursion was the ated with fatal excursions is adverse weather con-
consequence of another event or not. ditions, fatal excursions occur more frequently on
Arnaldo Valdés et al. (2011) developed probabil- commercial flights than other categories of aircraft
istic models of runway overruns and proposed risk operation, and overruns are the most fatal category
models supported by historical data; they noted of runway excursions.
that runway overrun risk depends on multiple fac- The risk of runway excursion dependent on
tors in relation to operating conditions including multiple factors is related to operating condi-
wind, runway surface conditions, landing or take- tions. The goal of this paper is to identify risk
off distance required, presence of obstacles, run- factors for type of accident of runway excursion
way distance available, existence and dimensions in relation to certain elements such as the aircraft
of runway safety areas. type and geographical region. The low accident

128

TIS2017_Book.indb 128 2/8/2017 8:02:12 PM

 rate of aviation means that no particular airport
has sufficient accident occurrences in the recent
past to support an accident frequency model
with reasonable statistical confidence (Piers,
1994; Hale, 2001). Therefore, this study is based
on a large database of relevant accident cases.
The results of the current study can be used by
a broad range of civil aviation organizations
for risk assessment and cost-benefit studies of
actions improvements.

2 RUNWAY EXCURSION DATA

The primary data source used in this work is a

database created by Aviation Safety Network
Figure 1. Distribution of events by type of accident.
(ASN). The Aviation Safety Network is a private,
independent initiative founded in 1996. On line
since January 1996, the Aviation Safety Network 4.1 Data preparation
covers accidents and safety issues with regards to
In order to make statistically significant data to be
airliners, military transport planes and corporate
analysed fields suitable to regroup together the ele-
jets.
ments of some records of the database have been
The ASN Safety Database contains detailed
created.
descriptions of some 15.800 incidents, hijack-
The following are the criteria used for the defini-
ings and accidents to airliner, military transport
tion of new fields.
category aircraft and corporate jet aircraft safety
occurrences since 1921. Most of the information
2.1.1 Geographical regions
are from official sources (civil aviation authori-
The airport’s countries where there was the runway
ties and safety boards), including aircraft produc-
excursion were grouped as recited in the IATA clas-
tion lists, ICAO ADREPs, and country’s accident
sification of the geographic regions (IATA 2011):
investigation boards.
For the purposes of this paper, it created a data- – AFI: Africa,
base containing solely runway excursion accidents, – ASPAC: Asia Pacific,
in a period between 2006 and 2015, for all catego- – CIS: Commonwealth of Independent States,
ries of aircraft, and in all world regions. – EUR: Europe,
Runway excursions are categorized into the fol- – LATAM: Latin America & the Caribbean,
lowing categories: Landing Overrun (LDOR), land- – MENA: Middle East & North Africa,
ing Veer Off (LDVO), Take-Off Overrun (TOOR), – NAM: North American, and
and Take-Off Veer Off (TOVO). The four runway – NASIA: North Asia.
excursion categories are mutually exclusive.
Table 1 shows the number of events by type of
In this database for every event are recorded
accident in each geographic regions.
information about:
– date 2.1.2 Risk factors
– airport The following four classes of risk factors have been
– airport’s country defined which includes all the causes of the events
– accident type recorded in the database:
– phase of flight
– Aircraft System Faults:
– potential cause
– Engines
– n° of fatalities
– Brake (wheel brakes, spoilers or
– n° of occupants
– reversers)
– aircraft type
– Hydraulic
– nature of flight
– Electric
– aircraft damages and
– Main gear
– dynamic event.
– Tire
Figure 1 shows the distribution of events con- – Other
tained in the database as a function of the type of – Human errors:
accident. – Incorrect Flight Planning

129

TIS2017_Book.indb 129 2/8/2017 8:02:12 PM

 Table 1. Distribution of events by type of accident in Table 2. Distribution of events by type of accident and
geographic regions. by risk factor.

LDVO LDOR TOVO TOOR LDVO LDOR TOVO TOOR

AFI 18 17 3 6 Aircraft System Faults 53 25 5 13

ASPAC 43 21 5 2 Human errors 49 42 12 8
CIS 8 5 3 6 Weather Conditions 38 15 6 4
EUR 25 16 6 3 Runway Conditions 12 11 1 5
LATAM 21 31 4 8 Unknown 30 44 10 8
MENA 21 12 1 3
NAM 44 35 12 10 – General Aviation aircraft (GA): Typically these
NASIA 2 0 0 0 aircraft con have one (single engine) or two
engines (twin engine). Their maximum gross
weight is usually below 14.000 lb.
– Communication/Coordination – Corporate Aircraft (CA): Typically these air-
– Pilot error craft can have one or two turboprop driven or
– Visual Illusion jet engines (sometimes three). Their maximum
– Fatigue gross mass is up to 90.000 lb.
– Excessive speed – Commuter Aircraft (Com A): Usually twin
– Loss of control engine aircraft with a few exceptions such as
– Other the De Havilland DHC-/which has four engines.
– Weather Conditions: Their maximum gross mass is below 70.000 lb.
– Low Visibility – Transport Aircraft (TA):
– Rain – Short-Range (S-R): Their maximum gross
– Wind Shear mass is usually below 150.000 lb.
– Tailwind – Medium-Range (M-R): These are transport
– Crosswind aircraft employed to fly routes of less than
– Ice 3.000 nm (typical). Their maximum gross
– Low Ceiling mass is usually below 350.000 lb.
– Strong Wind – Long-Range (L-R): These are transport air-
– Turbulence craft employed to fly routes of less than 3.000
– Freezing Rain nm (typical). Their maximum gross mass usu-
– Other ally is above 350.000 lb.
– Runway Conditions:
Table 3 shows the number of events by type of
– Wet
accident in relation to the aircraft class.
– Contaminated—Standing water
– Contaminated—Rubber
2.1.4 Nature of flight
– Contaminated—Oil
Nature of flight was reported in ASN database;
– Contaminated—Ice
the following are the items used for the analysis:
– Contaminated—Slush
– Contaminated—Snow – Ambulance,
– FOD – Cargo,
– Wildlife Hazards – Domestic Non Scheduled Passenger,
– Down Slope – Domestic Scheduled Passenger,
– Executive,
When the accident report does not mention the
– Ferry/positioning,
potential cause, this was referred to as unknown.
– International Scheduled Passenger,
Table 2 shows the number of events by type of
– International Non Scheduled Passenger,
accident in relation to the class of risk factors.
– Private,
– Other (Agricultural, Training, Test, Parachuting),
2.1.3 Aircraft classes
and
The accidents were analysed according to the dif-
– Unknown
ferent classes of aircraft operations represented
in: General Aviation (GA), Corporate Aircraft Scheduled services are flights scheduled and per-
(CA), Commuter Aircraft (Com A) and Transport formed for remuneration according to a published
Aircraft (TA)—to address these different opera- timetable, or so regular or frequent as to constitute
tional perspectives. The aircraft operation classes a recognizably systematic series, which are open to
are defined as: direct booking by members of the public,

130

TIS2017_Book.indb 130 2/8/2017 8:02:13 PM

 Table 3. Distribution of events by type of accident and scale to a series of predictor variables. Ordered
aircraft class. probability models are derived by defining a latent
variable z as a basis for modelling ordinal ranking
LDVO LDOR TOVO TOOR data. This unobserved variable can be specified as
GA 29 25 4 7 a linear function for each observation such that:
CA 39 31 14 14
Com A 48 33 6 10 z = βX + ε (1)
S-R 45 34 8 3
M-R 14 11 2 1 where X is a vector of variables determining the
L-R 7 3 0 3 discrete ordering for observation n, β is a vector
of estimable parameters, and ε is a random distur-
bance. Using this equation, observed ordinal data,
Table 4. Distribution of events by type of accident and y, for each observation are defined as:
nature of flight.

LDVO LDOR TOVO TOOR

y = 1 if z ≤ μ0
y = 2 if μ0 < z ≤ μ1
Ambulance 3 2 2 2
Cargo 27 15 4 5 y = 3 if μ1 < z ≤ μ2
Domestic Non 15 5 2 0 y=…
Scheduled
Passenger y = I if z ≥ μi-1,
Domestic Scheduled 72 48 10 14
Passenger where the μ are estimable parameters (referred to
Executive 11 14 3 4 as thresholds) that define y, which corresponds
Ferry/positioning 7 11 2 5 to integer ordering, and I is the highest integer
International 19 13 1 1 ordered response. Note that during estimation,
Scheduled non-numerical orderings such as never, sometimes,
Passenger
and frequently are converted to integers (for exam-
Military 7 9 4 2
ple, 1, 2, and 3) without loss of generality.
Private 11 11 2 2
The μ are parameters that are estimated jointly
Other 3 6 3 1
Unknown 7 3 1 2
with the model parameters (β). The estimation
problem then becomes one of determining the
probability of I specific ordered responses for each
A non-scheduled air service is a commercial air observation n. This determination is accomplished
transport service performed as other than a sched- by making an assumption on the distribution of ε.
uled air service. A charter flight is a non-scheduled If ε is assumed to be normally distributed across
operation using a chartered aircraft. observations with mean = 0 and variance = 1, an
Table 4 shows the number of events by type of ordered probit model results with ordered selection
accident in relation to the nature of flight. probabilities as follows:

P (y = i) = Φ (μi – βX) – Φ (μi+1 – βX),

3 STATISTICAL APPROACH FOR
EXCURSION DATA ANALYSIS where μi and μi+1 represent the upper and lower
thresholds for response category i and Φ (…) is the
In many transportation applications discrete data cumulative normal distribution.
are ordered, for example to analyse categorical Estimation is done using maximum likelihood
frequency data (e.g., accident, serious incident, methods. For this study, XLSTAT software was
incident). used to estimate separate ordered probit models
Although these data are discrete, application of in order to examine the risk factors of the runway
the standard or nested multinomial discrete mod- excursions depending on the flight nature and in
els does not account for the ordinal nature of the relation to different geographic regions of occur-
discrete data and thus the ordering information rence and the aircraft class involved. Runway
is lost. To address the problem of ordered dis- excursions risk factors are assessed on a five-point
crete data, ordered probability models have been scale.
developed. When interpreting the results of the ordered
Ordered probability models (both probit and probit models, a positive value for particular β
logit) are a form of discrete outcome models that implies that if the condition corresponding to
relate dependent variables on an ordered discrete that parameter is true (e.g., the runway excursions

131

TIS2017_Book.indb 131 2/8/2017 8:02:13 PM

 category is LDVO) the probability of the highest Table 5. Ordered Probit Model (normalized
ordered discrete category (e.g., risk factor: aircraft coefficients).
system faults) will increase and the probability of
the lowest category (e.g., risk factor: runway con- Std. 95% Confidence
Variable Coeff. Error interval
ditions) will decrease.
Assessing the impacts on intermediate catego- Event
ries is more difficult because these results are con- LDVO 0.000 0,000
ditional based upon the threshold values (μ). As LDOR −0.201 0,061 (−0.320, −0.081)
such, to examine the effects of a specific variable TOVO −0.117 0,059 (−0.232, −0.002)
on one of the interior categories, marginal effects TOOR −0.015 0,060 (−0.134, +0.103)
are computed as the change in the estimated
probabilities when a specific indicator variable is Airport’s country
changed from zero to one with all other variables NAM 0.000 0,000
held equal to their means. These marginal effects EUR 0.023 0,066 (−0.106, +0.151)
can be interpreted as the change in the probability LATAM −0.076 0,069 (−0.211, +0.058)
of a particular preference category, P(y = i), given AFI −0.190 0,067 (−0.321, −0.058)
a change in the respective variable. ASPAC −0.192 0,070 (−0.329, −0.055)
Ordered probability models were developed in CIS −0.116 0,062 (−0.237, +0.006)
order to gain greater insight as to how differences MENA −0.126 0,067 (−0.256, +0.005)
in geographic regions, aircraft type, and other fac- NASIA −0.083 0,058 (−0.197, +0.030)
tors influenced the cause of the runway excursion.
Aircraft class
GA 0.000 0,000
4 RESULTS AND DISCUSSION CA −0.040 0,076 (−0.189, +0.109)
Com A −0.026 0,082 (−0.187, +0.135)
Table 5 presents the parameters of the ordered S-R 0.028 0,084 (−0.137, +0.194)
probit model that has been developed for the esti- M-R 0.034 0,071 (−0.105, +0.172)
mation of the probabilities of the causes that gen- L-R 0.115 0,067 (−0.016, +0.247)
erate the runway excursions. Coefficient estimates
are provided for this model, along with standard Nature of flight
errors, as well as confidence intervals for each vari- Ambulance 0.000 0,000
able. The modelling results allowed to infer inter- Cargo −0.125 0,135 (−0.390, +0.139)
esting considerations, both general and in relation Domestic Non −0,101 0,098 (−0.294, +0.092)
to individual types of accidents. Scheduled Passenger
The comments relating to each type of accidents Domestic Scheduled −0.105 0,181 (−0.461, +0.250)
Passenger
will be performed with reference to a subdivision
Executive −0.118 0,110 (−0.333, +0.096)
of accidents in relation to the different geographic
regions of occurrence. Ferry/positioning −0.011 0,102 (−0.211, +0.189)
For each type of runway excursion in each geo- International −0.064 0,120 (−0.299, +0.171)
Scheduled Passenger
graphic region has been identified the cause that
Military −0.102 0,101 (−0.300, +0.096)
has the highest probability of generating the event.
Private −0.043 0,103 (−0.244, +0.158)
Unknown −0.117 0,083 (−0.281, +0.046)
4.1 Generality Other −0.010 0,085 (−0.177, +0.156)
The weather conditions play an important role,
although not predominant, for the overrun acci-
dents (both in take-off and landing). In LATAM and AFI the probability that the
In NAM and EUR the probability that an over- various types of accidents are caused by unknown
run is due to bad weather conditions is about 16% causes assumes a higher percentage than in countries
(with 18% values for smaller aircraft); while the of NAM and EUR, especially for small aviation.
probability that a veer-off is due to unfavourable This consideration is more evident in the countries
weather conditions is about 13% (with values 8% AFI even more than in those of LATAM.
for larger aircraft). In ASPAC the percentage of accidents involving
In LATAM, AFI and ASPAC the probability large aircraft (M-R and L-R) is 5.6%. For large air-
that bad weather conditions cause all kinds of acci- craft (L-R), the most likely cause is the mechani-
dents is on average 17% for all types of aircraft. cal failure for all types of accident; human error
Only for a L-R aircraft such probability is reduced is manifested by a reduced rate compared to other
to 10% in LATAM and ASPAC. types of aircraft.

132

TIS2017_Book.indb 132 2/8/2017 8:02:13 PM

 In CIS occurred only 5.6% of total accidents; error and mechanical failure in the case of small
the predominant cause for all incident types and aviation (GA and CA); for larger aircraft (com-A
for all types of aircraft is human error (from 20 and S-R), the probability that the human error
to 30%); the probability that accidents are caused is the main cause grows slightly, while for large
by unknown factors is about 25%. The mechanical aircraft (M-R and L-R) this type of accident is
failure is relevant in the case of LDVO, while the infrequent.
weather conditions are the potential cause of 18% In LATAM the main cause of TOVO is human
of accidents on average for all types of aircraft. error (around 27%), although the percentage
In NASIA there were only 2 LDVO accidents; of unknown causes is very high for all types of
for them the unknown cause is the most likely aircraft.
(43%). In AFI and ASPAC, accidents during take-off
involve only small aviation (CA, Com A); for such
aircraft, the most likely cause of TOVO is human
4.2 LDVO (Landing Veer-Off)
error (20%) followed by the bad weather condi-
In EUR and NAM, the main causes of this type tions (17%). Also in this case the unknown causes
of accident are those related to mechanical failure, are in high proportion (45%).
regardless of the type of aircraft and the flight
nature. For example, the LDVO for GA aircraft
4.5 TOVO (Take-Off Overrun)
and “Private” flight nature, has a 42% probabil-
ity of being caused by mechanical failure, 31% to In NAM and EUR accidents of this type are quite
have as a triggering event a human error, 13% to be rare (9.9% in NAM and 6% in EUR); the mechani-
caused by weather conditions and only 5% to be cal failure is certainly the most likely cause, espe-
caused by the runway conditions. cially for very large aircraft.
In LATAM, AFI, ASPAC and MENA the lead- Also in AFI and ASPAC the TOOR type acci-
ing cause of LDVO for small aircraft is human dents are not very frequent (13,63% in AFI and
error, while for large aircraft in LATAM, MENA 12.4% in ASPAC) and involve only small aircraft
and ASPAC is the mechanical failure and in AFI is (CA, Com A); for these aircraft the most likely
always human error. cause of such accidents is human error (about 26%)
followed by the bad weather conditions (18%). The
probability of unknown causes is high in percent-
4.3 LDOR (Landing Overrun)
age (30%).
In EUR and NAM, this accident has a higher In CIS the incidents of this type are quite fre-
probability of occurrence for causes related to quent (27,27%) and the likely cause is human error
human error (30%), regardless of the type of (25–30%) in the case of small and medium-sized
aircraft. It should however be pointed out that, aircraft. The large aircraft have not recorded any
as regards the nature of the flight, in the case such incident.
of “Other” (Test, training, parachuting), the In LATAM the frequency of these accidents is
mechanical failure is the main cause of this acci- very low (2.8%) and the most likely cause is the
dent type (over 30%). mechanical failure (about 35%) for small aircrafts,
In LATAM, human error is the main cause and even 62% for aircraft type L-R. The probabil-
(around 27%) of LDOR for small aircraft and ity that the TOOR is due to unknown causes is
for aircraft of larger size (L-R) the main cause is instead quite low.
mechanical failure (40%).
In AFI and ASPAC, the main cause of LDOR
for small aircraft is human error (about 20%); the 5 CONCLUSION
unknown causes, however, assume a high percent-
age (between 40 and 50%) for all types of aircraft. Aviation safety analysis historically has empha-
In NEMA, the main cause of LDOR is human sized accident data. For the most part, the aviation
error (approximately 25%) for small aircraft, while industry and government regulators have used data
the unknown causes contribute in a high percent- reactively to identify the causes of aircraft acci-
age (between 30 and 40%) for all types of air- dents and to take steps to prevent these types of
craft, with the exception of L-R mainly caused by accidents from recurring. But there are still lessons
mechanical failure (33%). to be learned from the few accidents that these car-
riers have. With improvements in safety and major
reductions in accidents, airline safety analysis will
4.4 TOVO (Take-Off Veer-Off)
have to shift toward analysis of incident and oper-
In EUR and NAM, the TOVO has an equal prob- ational data with the intent of identifying safety
ability of occurring from causes related to human risks before accidents occur.

133

TIS2017_Book.indb 133 2/8/2017 8:02:13 PM

 This paper shows how the same event may be for Traffic and Transport Engineering (IJTTE) 4 (1):
caused by different risk factors in relation to differ- 62–73.
ent geographic regions or types of aircrafts. Hale, A., 2001. Regulating airport safety: The case of
This results represent a very proactive tool for Schiphol. Safety Science 37: 127–149.
IATA. 2011. Runway excursion analysis report
future aviation safety analysis. 2004–2009 (RERR 2nd Edition—Issued 2011). Mon-
Future work includes efforts on improving pre- tréal, Québec.
dictive performance with statistic methods that International Civil Aviation Organization. 2009. Safety
address imbalanced data more effectively. Fur- management manual (2nd ed.). Montreal, Quebec.
ther, it will be considered the consequences of the International Civil Aviation Organization, Co-operative
events in terms of damage to the aircraft and/or Development of Operational Safety and Continuing
persons. Airworthiness Programme. (2012). Safety Manage-
ment System (SMS)
Oster, C.V. Jr., Strong J.S. & Kurt Zorn C. 2013. Ana-
REFERENCES lyzing aviation safety: Problems, challenges, oppor-
tunities. Research in Transportation Economics 43:
Arnaldo Valdés, R.M., Gómez Comendador, F., Mijares 148–164.
Gordún, L. & Sáez Nieto, F.J. 2001. The development Piers, M., 1994. The development and application of a
of probabilistic models to estimate accident risk (due method for the assessment of third party risk due to
to runway overrun and landing undershoot) applica- aircraft accidents in the vicinity of airports. In: 19th
ble to the design and construction of runway safety International Council of Aeronautical Sciences Con-
areas. Safety Science 49: 633–650. gress, Anaheim September ICAS proceedings. ICAS
Ayres, M. Jr., Shirazi, H., Carvalho, R., Hall, J., (19): 507–518.
Speir, R. & Arambula, E. 2013. Modelling the loca- Wagner, D.C.S. & Barkerb, K. 2014. Statistical methods
tion and consequences of aircraft accidents. Safety for modeling the risk of runway excursions. Journal of
Science 51: 178–186. Risk Research 7 (17): 885–901.
Distefano, N. & Leonardi, S. 2014. Risk assessment pro- Aviation Safety Network (ASN Database), 2016. <http://
cedure for civil airport. Journal International Journal aviationsafety.net/database/> (Giugno 2016).

134

TIS2017_Book.indb 134 2/8/2017 8:02:13 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A review of sulfur extended asphalt modifiers: Feasibility and limitations

S.E. Zoorob
Energy and Building Research Centre, Kuwait Institute for Scientific Research, Kuwait

C. Sangiorgi & S. Eskandersefat

Department of Civil, Chemical, Environmental and Materials Engineering, University of Bologna, Bologna, Italy

ABSTRACT: Sulfur Extended Asphalt Modifier (SEAM) is currently being marketed as an additive
for hot asphalt mixtures. Typically 40% by mass of the binder phase in the asphalt mix can be replaced
by SEAM using conventional mix design and production techniques. A number of full scale trials using
SEAM have been reported in the literature and, except for some minor concerns regarding rutting and
moisture resistance, it appears that overall the SEAM modified mixes have the potential for improved
mechanical performance compared to conventional asphalt. During the production stages of SEAM
asphalt mixes in conventional hot mix plants, the sulfur component will exist in the liquid phase, which
requires careful thermal management to control gaseous emissions. As a consequence, this paper includes
a general review of exposure limits for sulfur dioxide and hydrogen sulfide emissions and their short and
long term health effects on healthy and asthmatic individuals.

1 INTRODUCTION handling, storage and transportation. The pellets

have melting point in the range of 93 to 104°C,
In the early 1970s, due to concern over asphalt which has been specifically designed to melt when
cement (bitumen) supply issues and an anticipated in contact with a hot asphalt mix. The additive is
overabundance of elemental sulfur, several organi- intended to act both as a bitumen extender and as
zations in the US and Canada, began to evaluate an asphalt mixture modifier. The manufacturers
the potential for sulfur to substitute for bitumen claim (backed up by a number of full scale inves-
as binder extender, this became known as Sulfur tigations) that SEAM improves the performance
Extended Asphalt (SEA). Indentation/hardness of conventional bitumens by imparting tough-
tests carried out on elemental sulfur-bitumen sys- ness, strength and rut resistance. Furthermore, it
tems, showed that a large increase in resistance to is claimed that SEAM additives can provide cost
indentation can be obtained in the range in which savings when compared to other polymer modified
the continuous sulfur phase begins to form (i.e. asphalts (Rock binders Inc 2014).
at sulfur-bitumen ratios of 1.5–2.0) (Beaudoin &
Sereda 1979).
In the late 1980s, a sharp rise in the price of
2 SEAM IN ASPHALT PAVEMENTS
sulfur in addition to the fact that the use of hot
liquid sulfur during production generated a signifi-
2.1 Composition of SEAM pellets, mix
cant amount of fumes and odors brought its use in
proportions and construction of asphalt layers
road paving to an end. However, the recent rises in
bitumen prices coupled with the production of low The key to the composition of SEAM (ThiopaveTM)
sulfur fuels has once again made sulfur a market- is that liquid sulfur is plasticized by the addition of
able product in the asphalt industry. carbon black at a concentration of at least 0.25%
To overcome the problems with hot liquid sul- (preferably 0.4 to 0.8%) and the plasticized sulfur
fur used in bituminous mixes, Sulfur Extended is further treated with amyl acetate at a concentra-
Asphalt Modifier (SEAM) in solid pellet form tion of at least 0.08% (preferably between 0.2 and
(patented process) was developed and marketed as 0.4%) to produce an even more manageable plas-
an additive for hot mix asphalt under the brand ticized sulfur additive. The carbon facilitates the
name Shell ThiopaveTM (Shell 2014, FHWA 2012, plasticization reaction with the sulfur, additionally
Tran et al. 2010). the carbon creates an ultraviolet light shield which
The SEAM pellets are in a solid, non-sticky, non- helps to prevent ultraviolet degradation of the
melting form at ambient temperatures, which eases final bitumen plus aggregate product. On the other

135

TIS2017_Book.indb 135 2/8/2017 8:02:15 PM

 hand, the amyl acetate helps to reduce unwanted Table 1. Sample identification for data of Fig. 1.
odors from the product and thereby improve its
overall handling (Bailey et al. 2003). % % Design
SEAM may be used in asphalt mixes made Mix ID Base Binder Thiopave Air Voids
with conventional bitumen in weight ratios from Control-67 PG 67-22 0 4
20%/80% to 50%/50% SEAM/bitumen. SEAM Control-76 PG 76-22 0 4
melts readily into the bitumen and becomes a part Thio-67-30-2 PG 67-22 30 2
of the binder in the asphalt mixture. Stiffness and Thio-67-30-3.5 PG 67-22 30 3.5
Marshall stability are expected to increase with Thio-67-40-2 PG 67-22 40 2
increased amounts of sulfur (Predoehl 1989). Thio-67-40-3.5 PG 67-22 40 3.5
Once a SEAM/bitumen ratio is selected, the Thio-58-30-2 PG 58-28 30 2
equivalent volume replacement of binder (by% Thio-58-30-3.5 PG 58-28 30 3.5
weight) can be determined. Since sulfur is about Thio-58-40-2 PG 58-28 40 2
twice as dense as bitumen, a given weight of sulfur Thio-58-40-3.5 PG 58-28 40 3.5
has therefore about half of the volume of the same
weight of bitumen (Mahoney et al. 1982).
The sulfur pellets at ambient temperature are
added to the pre-heated aggregate and bitumen
during the asphalt mixing process rather than
the sulfur being pre-blended with bitumen. For
health and safety reasons, the manufactures stress
that when incorporating SEAM, the asphalt mix
discharge temperatures must be carefully con-
trolled to never exceed the recommended range
135–146°C. At this temperature, the pellets melt
quickly and the shear conditions in the mixer are
high enough to disperse the sulfur into the asphalt
mix in a very short time that is compatible with
asphalt mix production. Figure 1. Average E* results at 10 Hz and 21°C for all
Delayed delivery trucks arriving at the paving mixtures after 1 and 14 days curing.
site with SEAM mix below a temperature of 115°C
shall be rejected (Rock binders Inc. 2014). The E* results comparable to those of the PG 76–22
manufacturers claim that SEAM may be used with control mix. The average 14-day E* results of the
any conventional bitumen grade, but not polymer- Thiopave mixes were comparable or greater than
modified bitumen. Warm Mix Asphalts (WMA) the respective average 14-day E* values for both
that incorporate water foaming type additives in the control mixes (Tran et al. 2010).
their compositions are also not recommended A more recent laboratory investigation has also
for Sulfur extended asphalt production (FHWA shown that sulfur modified asphalt mixtures con-
2012). tinue to increase in stiffness and reach a plateau
after approximately 30 days. This increase in stiff-
ness recorded at 30 days, amounted to about 50% of
2.2 Stiffness evolution of SEAM asphalts and
the initial measurements that were taken at 9 days
general findings from full scale SEAM trials
following compaction (Cocurullo et al. 2012).
In a recent investigation (Tran et al. 2010), 10 A large full scale investigation was conducted
asphalt mix designs were laboratory manufactured by the California DoT to determine whether the
(2 control and 8 Thiopave composites). Details incorporation of sulfur with a soft grade of bitu-
of the mix compositions are listed in Table 1. men could change the temperature-viscosity rela-
Figure 1 shows the average and% increase in tionship of the resulting binder, thereby making
dynamic modulus (E*) for each of the 10 mixes it more useful in cold and hot climates (Predoehl
tested at 10 Hz and at 21°C after 1 day and 14 days 1989). The two trial sections (a hot and a cold cli-
of curing. mate test sections) utilized the same sulfur/bitumen
When a PG 67–22 was used as the base binder in blends, i.e. 20% and 40% sulfur by weight of total
4 Thiopave mixes, the average 1-day E* results of binder, and each project used the same blending AR
these Thiopave mixes were comparable or greater 2000 bitumen (i.e. RTFOT residue with viscosity
than the respective average 1 day E* values for the of 2000 Poise at 60°C). Overall, based on the find-
PG 67–22 control mix. In addition, except for ings of this full scale study, SEAM binders were a
the 30% Thiopave mix with 2% design air voids, viable alternative to conventional bitumens. Results
the other Thiopave mixes yielded average 1-day of physical tests (Hveem system) on compacted

136

TIS2017_Book.indb 136 2/8/2017 8:02:15 PM

 briquettes and field cores revealed no significant reported excellent pavement condition index (PI)
differences in stability, cohesion and surface abra- and no wheel path depressions or cracking of a full
sion between SEAM and conventional asphalt scale trial monitored for 3 years made with a 35/65
mixes. The findings of this study raised one major (Sulfur/bitumen) hot mix asphalt.
concern, in that it appeared that SEAM blends with
over 20% sulfur by weight (e.g. at a 40%/60% sulfur/
2.3 Rutting susceptibility of SEAM mixes
bitumen ratio) should not be utilized in overlays in
colder climate areas due to an early thermal crack- In 2004, the City of Calgary undertook a series of
ing potential. This was not an issue for the hotter full-scale demonstration projects aimed at evaluat-
climate test section (Predoehl 1989). ing the potential performance enhancement offered
The US Federal Highway Administration also by several premium surfacing materials, including
completed a field study to compare the performance (Johnston et al. 2005):
of sulfur-extended asphalt pavements to conven-
1. Stone Matrix Asphalt (SMA), 12.5 mm nominal
tional asphalt control pavements. A representa-
maximum size (NMS) using 6.0% Performance
tive set of pavements from 18 States was chosen
Grade (PG) 70-31 polymer modified binder
to provide a comprehensive evaluation. The sulfur/
(PMA).
bitumen ratios investigated were (10/90, 20/80,
2. 12.5 mm NMS SMA, modified with 3% man-
25/75, 30/70, 35/65, 40/60) including surface &
ufactured shingle modifier (MSM), with PG
binder courses. The primary conclusion was that
67-37 PMA, for a total binder content of 6.0%.
there was no difference in overall performance
3. Superpave (12.5 mm NMS, Fine Graded) using
between the SEAM and control sections. A labora-
the same PG binder and content as 1.
tory study complemented the field study, and cores
4. Superpave (12.5 mm NMS, Fine Graded) using
were obtained from many of the pavements for
the same percentage of PG binder and MSM
testing. In general, the laboratory test results sup-
modification as 2.
ported the results of the field study. Overall, sulfur
5. Superpave (12.5 mm NMS, Fine Graded) incor-
did not increase or decrease most test properties,
porating 3.9% 150/200 A bitumen and 2.6%
and often it had no effect on a given test property
SEAM.
of a mixture. However, sulfur did decrease the
resistance to moisture susceptibility in the labora- An Asphalt Pavement Analyser (APA) (a modi-
tory (this is further discussed below). There were fied version of a laboratory wheel tracker) was
also minor trends indicating that with some mix- used to provide an assessment of the rutting poten-
tures, sulfur reduces the susceptibility to rutting. tial of the various mixture types. Field cores were
Additionally stress-controlled repeated load indi- acquired from the various installations within sev-
rect tensile fatigue tests were carried out on cores eral months of construction. Superpave gyratory
obtained from the trial sections and when consid- compactor specimens were also fabricated from
ering all projects there was an indication that sul- plant mixes produced for the projects. Figure 2
fur decreased the fatigue life (Stuart 1990). presents the average APA results, ranked based on
Nicholls (2012) summarised key findings from laboratory fabricated samples. The average APA
selected full scale trials covering the period 2004 to field core test results are shown for comparison.
2008 from a number of countries including: Can- The results clearly show that with respect to rut-
ada, US, Saudi Arabia, Qatar (Al-Ansary 210) and ting, the SEAM mixes were the worst performers.
China. Although it is not entirely clear from this investi-
Cores taken from a wide range of sites have gation whether the choice of bitumen grade for
shown that the incorporation of sulfur pellets does SEAM incorporation was optimal. Selecting a
not result in any impairment of the compaction, harder bitumen grade may have reduced rutting,
with the bulk density being marginally greater than but in turn this may have also narrowed down the
the control in most cases. That marginal difference
may be due to the higher density of the pellets than
the bitumen. In general, Marshall stability of the
sulfur-modified mixtures were similar to the con-
trol mixtures on day 1, however the stability values
were observed to increase with time.
Al-Mehthel (2010) also presented excellent per-
formance results from four full scale trials (using
30/70 sulfur/bitumen blends) laid in 2006 at vari-
ous locations across Saudi Arabia that were moni-
tored for signs of rutting and fatigue cracking for a Figure 2. Comparison of lab prepared and core speci-
period of up to four years. Nazarbeygi (2012) also men asphalt pavement analyzer rut depths.

137

TIS2017_Book.indb 137 2/8/2017 8:02:16 PM

 allowable temperature window for production and weakened by the moisture conditioning processes
laying operations. (Stuart 1990).
Other laboratory based rutting investigations In another recent investigation (Tran et al. 2010),
showed much more promising results. For exam- moisture susceptibility testing was performed
ple, 45°C wheel tracking trials using Iranian 60/70 on six mix designs as shown in Figure 3 (details
pen grade bitumen with sulfur replacement lev- of the mix compositions were shown earlier in
els ranging from 25 to 45% showed progressively Table 1). Testing was conducted in accordance
reduced wheel tracking depths with increased sul- with ALDOT 361-88 and AASHTO T 283-07 after
fur replacement levels (Johnston et al. 2005). the specimens were allowed to cure for 14 days at
Nicholls (2012) also collated the results of labo- room temperature.
ratory conducted rutting tests of sample extracted For the ALDOT method, the conditioned speci-
from selected the full scale trials. Overall, the results mens were vacuum saturated to the point at which
were very positive, with rut depth values of sulfur- 55% to 80% of the internal voids were filled with
modified mixes being 20 to 55% less than the cor- water.
responding control mixes. The indirect tensile strength ratio at 25 ± 0.5°C
was calculated for each set by dividing the average
tensile strength of the conditioned specimens by
2.4 Moisture susceptibility of SEAM mixtures
the average tensile strength of the unconditioned
In one large laboratory investigation carried out specimens (Tran et al. 2010).
by the Washington State DoT, it was found that For the AASHTO T 283 method, the condi-
mixture moisture susceptibility (assessed using tioned specimens were vacuum saturated so that
the Lottman moisture conditioning test) increases 70–80% of the internal voids were filled with water.
when increasing amounts of sulfur are added to These specimens were then wrapped in plastic and
the binder (i.e. increasing SEAM ratio). Samples placed in a leak-proof plastic bag with 10 mL
were subsequently investigated using a SEM, and of water prior to being placed in the freezer at
it became apparent that the sulfur crystals formed −18 ± 3°C for a minimum of 16 hours. After the
in the samples prior to moisture conditioning freezing process, the conditioned samples were
were often destroyed or broken during the con- placed in a 60 ± 1°C water bath for 24 ± 1 hours to
ditioning process. Since strength (more specifi- thaw. All samples, conditioned and unconditioned,
cally stiffness) seems to be imparted to the SEAM were brought to room temperature in a 25 ± 0.5°C
samples through establishment of a network of water bath to equilibrate the sample temperature
sulfur crystals in the voids and binder, it is likely for 2 hours just prior to testing. Calculation of the
that this strength will decrease when that network failure load, splitting tensile strength, and TSR
is damaged by the combination of moisture and value was conducted at 25 ± 0.5°C.
freeze-thawing. The investigation found that, as As shown in Figure 3, the Thiopave modified
expected, increasing amounts of added sulfur mixes had lower TSR values than the control mix-
results in increased mixtures stiffness. At the same tures tested by both the ALDOT and AASHTO
time, increasing amounts of added sulfur gener- methods. In this study, the TSR values for the
ally results in increased mixture stiffness loss fol- Thiopave mixtures were found to be lower than the
lowing the moisture and freeze-thaw conditioning commonly accepted failure threshold of 0.8 (Tran
(Mahoney et al. 1982). et al. 2010).
In a large scale US Federal Highway Adminis-
tration investigation (Stuart 1990), the susceptibil-
ity to damage by moisture of SEAM pavements
was evaluated on cores obtained from a number of
trial pavements (representing 18 States). The test-
ing procedure was in accordance with ASTM D
4867 (minor variation on the Lottman procedure)
with resilient modulus and tensile strength values
being determined before and following moisture
conditioning. It was concluded that when consid-
ering all the projects, the effect of sulfur was to
reduce both the Tensile Strength Ratios (TSR) and
the resilient Modulus Ratios (MrR), but not the
visual percent stripping. It was hypothesised that
the lower ratios were related to a loss of cohesion
rather than a loss of adhesion. Overall the results Figure 3. Summary of TSR results (after 14 days of
clearly indicated that the SEAM binders were curing).

138

TIS2017_Book.indb 138 2/8/2017 8:02:16 PM

 In yet another more recent laboratory investiga- Sulfur recovery from global petroleum refiner-
tion (Cocurullo et al. 2012), asphalt mixtures com- ies alone is predicted to reach 50 Mt/yr in 2025.
posed of 40/60 pen grade bitumen modified with Future gas production is also likely to come from
30% SEAM by mass of total binder were slab com- deeper, hotter, and more sour deposits that would
pacted and cored. Indirect tensile stiffness modu- result in even more excess sulfur production. If
lus test results on dry and moisture conditioned non-traditional uses for elemental sulfur increase
(immersed in water at 85°C) sulfur modified speci- significantly, and alternative technologies for rein-
mens showed that after 9 days of water immersion jection and long-term storage are not developed,
approximately 50% of the initial dry modulus was the oversupply situation will result in tremendous
retained. Interestingly, further tests revealed that fol- stockpiles accumulating around the world.
lowing water immersion tests, long term dry recovery The aforementioned discussion on the advan-
periods (minimum 2 weeks) caused the specimens to tages and disadvantages of incorporating SEAM
partially recover their stiffness values with time. as a partial replacement of the bitumen fraction in
No definitive explanation has been put for- asphalt mixtures is typical of the type of informa-
ward to justify the loss of mechanical property tion readily available to road engineers and road
in moisture conditioned SEAM asphalt mixtures. material specifiers. In the opinion of the authors,
The greatest impact of sulfur on bitumen proper- this type of analysis is slightly restrictive in scope,
ties and performance probably relates to air oxi- in that when a new composite equals or even out-
dation of the sulfide forms indigenous to crude performs its rivals in terms of density, voids, stiff-
oil to various sulfur oxide species. For example, ness, fatigue, or even price performance, there
conversion of aliphatic sulfides to the correspond- remains other equally important more holistic
ing sulfoxide (RS(O)R), occurs during oxidative issues to consider, in particular when the new com-
aging tests for bitumen. The change in physical posite contains a non inert bulk waste material.
properties of sulfides with oxidation is dramatic. In the following sections, an attempt is made to
For example, methyl sulfide (CH3SCH3), is a expand the discussion by exploring the exposure to
light (ρ = 0.846 g/mL), water-insoluble, low boil- sulfur dioxide and hydrogen sulfide emissions and
ing liquid (bp = 36°C); whereas methyl sulfoxide their short and long term health effects on healthy
(CH3S(O)CH3) is a heavy (ρ = 1.48), water miscible and asthmatic individuals.
liquid boiling at 189°C. With regard to bitumens,
the sulfoxide group exhibits a significant interac-
3.1 Reaction kinetics of H2S evolution
tion with most aggregates which is fairly suscepti-
ble to water stripping, in contrast to the negligible At temperatures above 130°C the sulfur ring mol-
adsorption of sulfides onto aggregate. Oxidation ecules suffer partial decomposition and polymerize
of sulfides to sulfoxides would thus be expected to in to long diradical chains. When sulfur is blended
increase asphalt hardness/viscosity, and to increase with bitumen at high temperature these sulfur
the concentration of amphoteric species. It would radicals react with some components of bitumen
also potentially increase susceptibility to water provoking two competitive reactions. These radi-
stripping and strength of asphalt-aggregate adhe- cals may either extract hydrogen from hydrocar-
sion (Green et al. 1993). bon molecules with subsequent hydrogen sulfide
formation by means of dehydrogenation reactions,
or can be incorporated to bitumen structure as
3 SEAM FROM A WIDER PERSPECTIVE carbon-sulfur bonds by means of addition reac-
tions. Which of these two reactions is predomi-
The sulfur industry is different from many other nant depends on temperature, sulfur content and
important modern mineral industries in that the heating rate. Generally, the dehydrogenation reac-
disposal of excess supplies of sulfur is becoming a tions predominate at temperatures above 150°C
more important issue than that of how to maintain (Masegosa et al. 2012).
sustainable production. Worldwide recovered sulfur H2S gas concentrations can gradually increase to
output is expected to increase significantly in the high levels in the air voids of a loose paving mixture,
future. Sulfur surpluses were expected beginning during storage in silos and during truck delivery
in 2010 with acceleration thereafter as a result of to the paving site. The “entrapped” gas is released
increased production, especially from oil sands in when the air pockets in the mixture are opened up
Canada, natural gas in the Middle East, expanded as the mixture is dumped from the delivery trucks
oil and gas operations in Kazakhstan, and heavy- or as the mixture is subjected to mechanical mix-
oil processors in Venezuela. Additional increases ing. High paver screed temperatures can also cause
are expected to come from Russia’s growth in sul- unacceptable emissions (Saylak & Tex 1988).
fur recovery from natural gas and Asia’s improved The SEAM Materials Safety Data Sheet states
sulfur recovery at oil refineries. that momentary levels of H2S from 1 to 20 ppm

139

TIS2017_Book.indb 139 2/8/2017 8:02:16 PM

 have been recorded in the paver hopper as a new very low concentrations and the threshold for
mix was being dumped, and at the paver screed perception is between 0.02 to 0.13 ppm. Nonethe-
level while it was being preheated. The same docu- less, odor perception is unreliable as a warning
ment also states that at processing temperatures of high exposures, olfactory fatigue may develop
below 150°C the sulfur-bitumen mix gives off very at concentrations of ≥ 100 ppm (Costigan 2003,
little H2S, but that at temperatures above 150°C, Simonton 2007). During vigorous exercise, low
the rate of H2S evolution increases significantly level exposures (5 to 10 ppm) cause a shift to
and at very high temperatures emissions may not anaerobic respiration, leading to increased lactic
dissipate readily and are therefore considered haz- acid formation and irritation to eyes, nose and
ardous (Rock binders Inc 2014). throat, but evidence on eye irritation is inconsist-
In one recent investigation (Masegosa et al. ent. The balance of evidence suggests that eye
2012), blends of 40pen bitumen containing 5% irritation is only likely to occur at high exposure
and 10% sulfur by mass of bitumen were analyzed concentrations, in region of hundreds of ppm
at elevated temperatures in order to determine (Costigan 2003, Simonton 2007).
the amount of volatile fractions evolved from the In asthmatic individuals, levels as low as 2 ppm
blends. The mass loss rates at 130°C and 140°C can cause bronchial constriction as well as sponta-
(composed of H2S + evaporated S), averaged over neous abortion. Between 50 to 200 ppm symptoms
the first 7 days of thermal exposure are shown in may include, severe respiratory tract irritation, eye
Table 2 (Masegosa et al. 2012). The results indi- irritation/acute conjunctivitis, shock, convulsions,
cated that the percentage of sulfur evolved by the coma and death in severe cases (Simonton 2007).
blends depended primarily on the storage tempera- Exposures for a few hours to concentrations of 500
ture. No significant differences were found in the ppm and above are likely to be fatal. Exposure for
rate of volatiles evolved by the blends as a function a few minutes to concentrations of 1000 ppm and
of the blended sulfur content. It has been observed above are likely to cause rapid unconsciousness
that the volatile compounds produced by the and death (within minutes).
blends contain both sulfur gas as well as hydrogen These findings may be of relevance to personnel
sulfide. In this study, the authors concluded that it who are continuously involved with the produc-
would be best to maintain the temperature as close tion and laying of SEAM mixtures for a number
as possible to 130°C during the stages of mixing, of years.
handling and storage of sulfur bitumen mixtures,
in order to control emissions.
3.3 Other SEAM gaseous emissions to consider
3.2 Toxicological profile of hydrogen sulfide Assuming a typical SEAM/Bitumen ratio of
40%/60%, with a sulfur density twice that of bitu-
Hydrogen Sulfide (H2S) is a colorless, flammable men. For one tonne of asphalt mix manufactured
gas. H2S is rapidly absorbed from the lungs fol- at a presumed binder content of 6% (by mass of
lowing inhalation exposure. Absorption of the gas mix), the amount of sulfur encapsulated in the
through the skin is minimal. Following absorp- asphalt mix would be approximately 51 kg. In
tion, H2S is widely distributed around the body, view of the substantial amounts of sulfur that can
primarily as undissociated H2S or as HS—ions. potentially be used, the authors decided to conduct
H2S binds reversibly to metalloenzymes, includ- a small laboratory trial to investigate emissions.
ing those involved in aerobic cellular respiration Several 10 kg samples (40%/60% SEAM/bitu-
such as cytochrome oxidase. H2S is a metabolic men) asphalt mixes were manufactured. Different
poison, it blocks the electron transport chain in hot batches were kept in the loose state in metallic
mitochondria, inhibiting cellular utilisation of trays in laboratory ovens preset at various temper-
oxygen (Costigan 2003). H2S has a pungent odor atures varying from 130°C to 210°C. The tem-
reminiscent of rotten eggs, the rotten egg odor is perature of each mix and the gaseous emissions
recognizable up to 30 ppm. It has a sweet odor were continuously monitored for up to 6 hours.
at 30 ppm to 100 ppm. The odor is detectable at A Dräger X-am 5000 hand held gas detector with
H2S and SO2 sensors was especially purchased for
Table 2. Rate of mass loss through gaseous emissions the trial. Although not an accurate or conclusive
of sulfur stored at high temperatures.
investigation by any means, nonetheless, the fol-
Sulfur in blend Storage at Storage at lowing trends were very prevalent:
(%mass of bit.) 130°C 140°C • at 150°C and below, there was very little detecta-
ble H2S emissions (all measured values were well
5% 0.52%/day 0.66%/day
below a concentration of 5 ppm). It was very
10% 0.31%/day 0.60%/day
clear that the H2S suppressant was functioning

140

TIS2017_Book.indb 140 2/8/2017 8:02:16 PM

 in line with the manufacturer’s expectations. potential for improved mechanical performance
Above about 160°C, H2S emissions can become (stability, stiffness, cohesion, surface abrasion)
a concern, and as expected, higher temperatures compared to asphalts manufactured with conven-
further intensify emissions. tional bitumens.
• in the case of SO2 emissions, things were quite When sufur and bitumen are heated and com-
different. At 150°C, concentrations were cer- bined the sulfur can react chemically with the
tainly higher than 10 ppm, with some readings bitumen causing dehydrogenation which results in
even approaching 20 ppm. hydrogen sulphide gas being formed, causing also
significant changes in the rheological properties of
The aforementioned results prompted the
the bitumen.
authors to investigate further the effects of SO2 on
Research on the reactions of elemental sulfur
human health and acceptable exposure limits. While
with various bitumens and residues has confirmed
health effects are documented at various concentra-
that H2S evolution starts at temperatures as low as
tions by different researchers and organizations, a
about 130–135°C. However, the activation energy
sampling of thresholds for health effects is provided
and the reaction order are different for various
here. Adverse effect on pulmonary function by an
types of bitumens. These findings may be of rel-
exposure at low concentrations (0.75 ppm) is revers-
evance to personnel who are continuously involved
ible. At higher concentrations, airways can become
with the production and laying of SEAM mixtures
severely obstructed and the effects can involve the
for a number of years.
lower airways. At 5 ppm, dryness of the nose and
In the UK, for instance, control of inhalation
throat can be observed and resistance to bronchial
exposure to H2S in the workplace must comply
airflow significantly increases. At 6–8 ppm tidal
with occupational exposure standards of 5 ppm
respiratory volume may noticeably decrease. At
(8 hour timed weighted average) and 10 ppm (short
10 ppm, sneezing, coughing, and wheezing may be
term exposure limit). With respect to SO2 emis-
observed, possibly accompanied by eye, nose and
sions, the US national ambient air quality stand-
throat irritations. Nosebleeds may also be seen. At
ards for SO2 recommends that repeated exposures
this level, asthmatics are likely to experience asth-
to 5-minute peak SO2 levels (≥ 0.60 ppm) could
matic paroxysm, lasting possibly for several days. At
pose an immediate significant health risk for a
20 ppm, bronchospasms tend to begin and eye irri-
substantial proportion of asthmatic individuals at
tation is very likely. At 50 ppm discomfort becomes
elevated ventilation rates.
extreme, but permanent injury is unlikely if expo-
The aim of this paper was to highlight vari-
sure is less than 30 minutes duration. Above 50 ppm,
ous examples of short and long term aspects
reflex closure of the glottis can take place and last
that highway materials engineers must consider
for a period of minutes. Exposure to sulfur dioxide
when assessing the viability of encapsulating
at a concentration of 400 ppm will likely constitute
waste materials, including SEAM or other sul-
an immediate danger to life (Miller 2004).
fur based composites, in roads. We therefore all
have a duty to pass on to our future generations
road networks that also function as potential “lin-
4 CONCLUSIONS ear quarries” and not as “linear waste dumps”.
The decision making process must weigh up all
Sulfur Extended Asphalt Modifier (SEAM) in the short and long term factors on a case by case
solid pellet form (primarily composed of sulfur basis and should principally be based on sound
that is plasticized by the addition of carbon black engineering principles.
and amyl acetate) was marketed as an additive for
hot mix asphalts. A typical SEAM/bitumen ratio
recommended for most mixes is 40%/60% (sulfur/ REFERENCES
bitumen) by mass of total binder.
SEAM may be used with any conventional bitu- Al-Ansary, M. 2010. Innovative solutions for sulphur in
men grade, but not polymer-modified bitumens Qatar. The Sulphur Institute’s (TSI) Sulphur World
and the use of free lime is not recommended. Con- Symposium, Doha, Qatar.
ventional mix designs and production techniques Al-Mehthel, M., Al-Abdul, W., Al-Idi, S. & Baig M.G.
are entirely suitable for SEAM mixes. 2010. Sulfur extended asphalt as a major outlet for
A substantial number of laboratory and full sulfur that outperformed other asphalt mixes in the
Gulf. Sulphur Institute’s (TSI) Sulphur World Sympo-
scale trials using SEAM have been reported in the sium, Doha, Qatar.
literature, and except for some minor concerns Bailey, W.R., Rock Binders Inc. & Shell Canada Energy.
regarding rutting and “more than minor” concerns 2013. International Patent Application no. WO
with respect to moisture resistance, it appears 03/014231 A1, PCT/US02/25333, Sulfur additives for
that overall the SEAM modified mixes have the paving binders and manufacturing methods.

141

TIS2017_Book.indb 141 2/8/2017 8:02:16 PM

 Beaudoin, J.J. & Sereda, J. 1979. A two-continuous-phase Academies by The Center for Research Information,
sulfur-asphalt composite-development and character- Inc., Contract No. IOM-2794-04-001, pp. 1–58.
ization. CAN. J. CIV. ENG. Vol. 6, pp. 406–412. Nazarbeygi, A.E. & Moeini, A.R. 2012. Sulfur Extended
Cocurullo, A., Grenfell, J., Yusoff, N.I. and Airey, G.D. Asphalt Investigation, Laboratory and Field Trial.
2012. Effect of moisture conditioning on fatigue prop- 5th Eurasphalt & Eurobitume Congress, 13–15th June
erties of sulfur modified asphalt mixtures. 7th Rilem 2012, Istanbul.
International Conference on Cracking in Pavements, Nicholls, J.C. 2012. Sulphur Extended Asphalt Modifier.
Vol. 2, Edited by Scarpas A. et. al., pp. 793–803. 5th Eurasphalt & Eurobitume Congress, 13–15th June
Costigan, M.G. 2003. Hydrogen Sulfide: UK Occupa- 2012, Istanbul.
tional Exposure Limits. Occup. Environ Med 2003. 60: Predoehl, N.H. 1989. An evaluation of sulfur extended
pp. 308–312. asphalt (SEA) pavements in cold and hot climates.
FHWA-HIF-12–087 Tech Brief. 2012. An Alternative Office of Transportation Laboratory, State of Califor-
Asphalt Binder, Sulfur-Extended Asphalt (SEA), US nia Dept. of Transportation, Final Report No. FHWA/
Dept. of Transportation, Federal Highway Administra- CA/TL-89/01.
tion, Office of Pavement Technology. Rock Binders Inc., Material safety data sheet http//wrbai-
Green, J., Yu, S., Pearson, C. & Reynolds, J. 1993. Analy- ley.com/rockbinders/index.html. Accessed May 2014.
sis of Sulfur Compound Types in Asphalt. Strategic Saylak, D. & Tex B. 1988. Sulfur-Coated Asphalt Pellets.
Highway Research Program, SHRP-A-667, Sept. 1993. US Patent No. 4,769,288, Sept. 6.
Johnston, A.G., Yeung, K. & Tannahill D. 2005. Use Shell ThiopaveTM. www.shell.com/sulphur/thiopave. Accessed
of Asphalt Pavement Analyzer Testing for Evaluat- May 2014.
ing Premium Surfacing Asphalt Mixtures for Urban Simonton, S., 2007. Human Health Effects from Expo-
Roadways. Annual Conference of the Transportation sure to Low-Level Concentrations of Hydrogen
Association of Canada, Calgary, Alberta, pp. 1–16. Sulfide. article published online in the Oct. 2007 issue
Mahoney, J.P., Lary, J.A., Balgunaim, F. & Lee, T.C. 1982. of Occupational Health & Safety. http://ohsonline.
Sulfur Extended Asphalt Laboratory Investigation, com/home.aspx
Mixture Characterization. Washington State Depart- Stuart, K.D. 1990. Performance evaluation of sulfur-
ment of Transportation, Final Report WA-RD 53.2. extended asphalt pavements—laboratory evalua-
Masegosa, R.M., Canamero, P., Cabezudo, M.S., Vinas, tion. Office of Engineering and Highway Operations
T., Salom, C., Prolongo, M.G., Paez, M. Ayala. 2012. R & D, US Dept. of Transportation, Federal Highway
Thermal behaviour of bitumen modified by sul- Administration, Publication No. FHWA-RD-90–110.
phur addition. 5th Eurasphalt Eurobitume Congress, Tran, N., Taylor, A., Timm, D., Robbins, M., Powell, B. &
13–15th June 2012, Istanbul. Dongre, R.. 2010. Evaluation of Mixture Perform-
Miller, V. 2004. Health Effects of Project SHAD Chemi- ance and Structural Capacity of Pavements Using
cal Agent: Sulfur Dioxide, Prepared for the National Shell Thiopave. NCAT Report 10−05.

142

TIS2017_Book.indb 142 2/8/2017 8:02:16 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

In-plant production of warm recycled mixtures produced with SBS

modified bitumen: A case study

A. Stimilli, F. Frigio & F. Canestrari

Università Politecnica delle Marche, Ancona (AN), Italy

S. Sciolette
Laboratorio Dottor Sciolette, Arriccia (RM), Italy

ABSTRACT: Asphalt mixtures produced at reduced temperatures through Warm Mix Additives
(WMA) allow reductions of fuel consumption and harmful emissions, ensuring economic and environ-
mental benefits. Considering that nowadays mixtures include more and more often reclaimed aggregates,
the combination of recycling and WMA technologies represents a major challenge in road construction
and needs further investigations for identifying drawbacks/advantages. Several concerns derive from regu-
lar in-plant productions since warm recycled mixtures have been mainly optimized through laboratory
studies without evaluating possible issues of large-scale in-plant productions and lay-down processes. So
far, few field constructions were realized, limited to small trial sections and without considering the use
of modified bitumens. Given this background, the paper describes in-plant productions of warm recycled
mixtures prepared with three chemical WMA additives for the construction of an extensive motorway
segment. Dense-graded mixtures for binder and base courses as well as open-graded mixtures for wearing
courses were produced. Three full-scale trial sections included warm mixtures whereas a further section,
used as reference for comparison purposes, comprised analogous mixtures realized through hot recycling
according to the current practice. The paper describes the construction steps and the controls carried out
to verify technical standard requirements in terms of volumetric properties, compactability and Indirect
Tensile Strength. The main objective was to attest the feasibility of large-scale productions to adequately
reproduce the mix design previously implemented through laboratory studies when WMA technologies
and recycling techniques are concurrently involved. Moreover, gas emissions monitoring at the asphalt
plant during the real scale productions was conducted through Continuous Emissions Monitoring Sys-
tems in order to quantify the potential benefits in terms of pollution. Results demonstrate the suitability
of WMA chemical additives to produce at low temperature mixtures with adequate performance, concur-
rently recording a significant reduction in pollutants without needing mix design modifications or imple-
mentation of expensive new technologies.

1 INTRODUCTION rather than field experimental projects (D’Angelo

et al. 2008). The consequent lack of in situ experi-
1.1 Warm mix asphalt ence determines low reliability on these production
techniques and risk to compromise the complete
Technologies to produce Warm Mix Asphalt
exploitation of WMA technologies potential. This
(WMA) are gaining growing interest in the field of
matter is even more accentuated by the more and
road construction due to the number of economic
more urgent need of combining warm technologies
and environmental advantages deriving from the
and recycling techniques, so dealing with warm
significant reduction in the production tempera-
mixtures incorporating Reclaimed Asphalt Pave-
tures compared to traditional Hot Mix Asphalts
ment (RAP) material. Such a challenge introduces
(HMA) (D’Angelo et al. 2008, Frank & Prowell
further variables that could determine significant
2014). Nowadays, several WMA additives based on
variations on mixtures behavior (Dinis-Almeida &
different operational mechanisms are available on
Lopes Afonso 2015, Dinis-Almeida et al. 2016, Lee
the market (i.e. organic waxes, chemical additives,
et al. 2009). Additional concerns are related to the
water-based products) (Rubio et al. 2012). Due
use of Polymer Modified Bitumens (PMB), whose
to the recent implementation of such technolo-
use is currently increasing worldwide for the pro-
gies, over the last years their effectiveness has been
duction of road mixtures. Polymers significantly
mainly investigated through laboratory studies

143

TIS2017_Book.indb 143 2/8/2017 8:02:18 PM

 improves bitumen properties, but usually require Moreover, the study allowed the evaluation of pos-
higher working temperatures to gain acceptable sible issues related to the asphalt plant operation
mixture workability. Therefore, the use of warm when producing at reduced temperatures in large
technologies could result not suitable when com- scale amounts, as well as advantages and drawbacks
bined with PMBs. This aspect further complicates during lay-down and compaction. Finally, thanks
the overall analysis and WMA technology appli- to the large quantity of material involved for the
cability. In this context, full-scale trial sections are construction of the trial sections, the asphalt plant
essential to provide reliable performance data and worked at reduced temperatures for long opera-
appropriately assess warm mixtures suitability, tional windows, so giving the opportunity to have
concurrently optimizing possible concerns related appropriate durations for adequately monitoring
to the operations of the asphalt plant during pro- pollutant emissions, key aspect to define the over-
duction and potential in-situ problems during lay- all convenience of WMA technologies.
down.

1.2 Full-scale trial project 2 EXPERIMENTAL PROGRAM

The research project presented in this paper, realized
2.1 Materials
in April 2016, consists in the construction of four
full-scale trial sections located along a stretch of an In each trial section, the pilot project involved the
Italian motorway pavement (800 m along the motor- construction of three bituminous courses prepared
way A1, segment Orte—Fiano Romano, sud direc- as follows:
tion, km 513+000 – km 513+800). The A1 motorway
– wearing course: open graded mixture (4 cm
is a double carriageway high traffic volume road
thick) with 15% by aggregate weight of selected
with three lanes per direction. The project involved
RAP (fraction 8/16 mm, deriving only from old
the re-construction of the slow lane after milling of
motorway porous courses), virgin basalt aggre-
all existing bituminous courses. Three sections (each
gate fractions combined with 70% cellulose-30%
200 m long) were reconstructed with bituminous
glass fibers (dosed at 0.3% by aggregates weight,
pavement courses all realized with recycled warm
added to prevent draindown issues) and 5.25%
mixtures, adopting in each case a different WMA
by aggregate weight of total bitumen content
additive. All the selected WMA products were
(virgin bitumen + reclaimed bitumen from
chemical additives since, based on previous labora-
RAP);
tory studies (Frigio & Canestrari 2016; Frigio et al.
– binder course: dense graded mixture (10 cm
2016), this kind of WMA technology was found
thick) with 25% by aggregate weight of RAP
the most promising in terms of performance when
(RAP 0/16 mm, unfractioned, deriving from old
dealing with modified bitumens and RAP aggre-
binder and base motorway courses), virgin lime-
gates (for both open and dense graded mixtures).
stone aggregate fractions and 4.8% by aggregate
A further section was built with traditional HMAs
weight of total bitumen content;
and used as reference for comparison purposes. A
– base course: dense graded mixture (15 cm
Styrene-Butadiene-Styrene (SBS) polymer modi-
thick) with 30% of RAP by aggregate weight
fied bitumen was used for preparing all produced
(RAP 0/16 mm, unfractioned, deriving from old
mixtures. Trial sections were constructed along a
binder and base motorway courses) virgin lime-
continuous stretch of the motorway in order to
stone aggregate fractions and 4.5% by aggregate
have the same boundary conditions in terms of
weight of total bitumen content.
existing materials, courses thickness, road geom-
etry (straight flat stretch), traffic volume and For all courses, the virgin bitumen used was a
weather conditions. SBS polymer modified bitumen (i.e. 3.8% of SBS
by bitumen weight), the same typically employed
for maintenance and construction activities in the
1.3 Objectives
Italian motorway network. Likewise, the aggregate
The main objective of the study was to evaluate grading curve (as well as the RAP percentage) was
the feasibility of WMA technologies for produc- equal to the one usually adopted at the asphalt
ing recycled bituminous mixtures prepared with a plant for each corresponding course in case of
PMB and the related environmental consequences. HMA productions. For determining the right
Three different chemical WMA additives were amount of virgin bitumen content, RAP fractions
selected in order to evaluate the effectiveness of were previously subjected to bitumen extraction.
different chemical compositions and operational In each trial section it was used a different type
mechanisms on mixture workability, volumetric of chemical WMA additive and, within the same
properties and main mechanical performance. section all bituminous courses were prepared with

144

TIS2017_Book.indb 144 2/8/2017 8:02:18 PM

 the same WMA product and at equal working 1.00 g/cm3 at 20°C and a flash point higher than
temperatures (i.e. 130° and 120°C in case of warm 200°C.
mixtures, 170°C and 160°C in case of hot mix- In each case, the additive was added to the
tures, for mixing and compaction, respectively). virgin bitumen by means of a volumetric pump
Each additive was dosed in accordance with the (equipped with a timer and an electronic control)
range prescribed by the producer, in quantity dif- connected with an external tank where the WMA
ferent for each course (Table 1) depending on the additive is stored and that allowed to supply in
“working” bitumen content (virgin bitumen + continuous the bitumen with the additive accord-
reactivated RAP bitumen) in order to consider the ing to the predetermined delivery capacity. Special-
different amount of RAP aggregates. This implies ized technicians carried out the calibration of the
that the exact quantity of each WMA additive was pump few days before warm mixtures productions.
determined taking into account a certain degree The WMA agent injection happened right before
of RAP bitumen reactivation that, based on pre- adding each material component (i.e. virgin bitu-
vious specific studies (Stimilli et al. 2015), can be men, virgin and RAP aggregates) to the mixing
assumed equal to 70% in case of HMAs. Due to chamber. In this way, veery limited modifications
the reduced working temperatures, this value was to the asphalt plant were applied for producing
precautionary considered lower for WMA mix- at reduced temperatures without compromising
tures and equal to 60%. the possibility to easily switch to standard HMA
WMA products were selected with different productions. Moreover, since WMA additives are
chemical composition and operational mecha- characterized by a rapid decay curve, in order to
nisms. In particular, one trial section (section 1) fully exploit WMA additives potential, it is pref-
was realized with an additive (coded as C1) mainly erable to use the bitumen immediately after the
composed of ammine substances which act as sur- addition of the chemical agent without insertion
factants and adhesion enhancers. It is a viscous of rest times, as it happened in the present investi-
liquid at 25°C with a density of about 1.0 g/cm3, a gation. Apart the addition of WMA additives, the
pour point of about −8°C and a flash point higher operational steps followed at the asphalt plant for
that 140°C. producing warm mixtures were the same usually
In the second section, a chemical additive (coded adopted for HMAs.
as C2) containing alkylates and fatty acids, which
act as viscous regulators, was added to asphalt
mixtures. At ambient temperature, it is an amber- 2.1 Working program
colored, inodorous liquid, insoluble in water. It is The construction activities lasted four days as fol-
characterized by a density of 0.86–0.90 g/cm3 at lowing described. In the first working day, the
20°C, a flash point higher than 220°C and a boil- original pavement was milled for a total depth of
ing point between 300°C and 408°C. 34 cm. The milling activities, starting from sec-
The third section was realized with a chemical tion 1, were carried out in two subsequent steps
additive (coded as C3) that, analogously to C2 in order to keep separate the RAP aggregate pro-
(even if characterized by a different chemical com- duced by milling the porous surface course from
position), acts as viscous regulator. C3 is liquid RAP aggregates of binder and base courses.
at ambient temperature with a density of about Concurrently, the warm mixture with additive C1 for
the base course (B_WC1) was prepared at the asphalt
Table 1. Main mixture characteristics per each course. plant. After completion of the milling activities in
the first section, the base mixture was laid-down
WMA and compacted. Right after, the base course of the
additive second section (B_WC2), whose mixture was mean-
Total [% by while produced at the asphalt plant (with the additive
bitumen virgin
C2) after mixture B_WC1, was constructed.
RAP [% by [% by agg. bitumen
Course agg. weight] weight] weight] The second working day started with the pro-
duction and in situ application of the base course
WEARING 15 5.25 C1 = > 0.42 for section 3 (B_WC3) and the binder course for
(OG: 4 cm) (RAP 8–16 mm) C2 = > 0.70 section 1 (BIN_WC1), both prepared at reduced
C3 = > 0.45 temperatures. Afterwards, the asphalt plant was
BINDER 25 4.80 C1 = > 0.50 switched to produce the base course of section 4
(DG: 10 cm) (RAP 0–16 mm) C2 = > 0.80 (B_H prepared at standard temperatures). The lat-
C3 = > 0.50 ter was produced as last mixture of the day in order
BASE 30 4.50 C1 = > 0.55 to optimize the working timetable and avoid long
(DG: 15 cm) (RAP 0–16 mm) C2 = > 0.90 waiting periods needed to reach the desired mixing
C3 = > 0.55 and compaction temperature when switching from

145

TIS2017_Book.indb 145 2/8/2017 8:02:18 PM

 hot to warm productions. In fact, rising the asphalt and compaction allowed the evaluation of in situ
plant operational temperature is much faster than mixture workability to confirm what observed
the opposite. Moreover, in such way the warm pro- through the laboratory evaluation.
duction was continuous with a long duration that
permitted the monitoring of pollutant emissions
2.3 Emissions monitoring
for a significant period. Concurrently, the asphalt
plant had enough time to reach a full capacity at Emissions of pollutants were monitored during in
reduced temperature, so allowing the observation of plant productions in order to evaluate the effective
potential critical issues in the mid-long term opera- environmental benefits deriving from lower working
tion. The same plan was carried out during the third temperatures. In compliance with current regula-
working day when mixtures for the binder course tions, a specialized staff carried out the monitoring
were completed in the following order: warm binder of gases and dusts emissions through direct in plant
for section 2 (BIN_WC2), warm binder for section 3 detections. An in-situ Continuous Emissions Moni-
(BIN_WC3) and hot binder for section 4 (BIN_H). toring System (CEMS) was adopted as a tool to
The last working day was dedicated to the pro- monitor flue gases (Harrison & Kemmer 1996). The
duction and subsequent construction of all wear- system consists of a sample probe, a filter, a sam-
ing courses, starting from section 1 (OG_WC1) ple line, a gas conditioning system, a calibration gas
and finishing with section 4 (OG_H) in order to system, and a series of gas analyzers which reflect
have a continuous warm production with all the the parameters being monitored. Emissions were
advantages previously described. measured using a pollutant concentration moni-
tor apparatus. A “hot dry” extractive method was
adopted: the sample is carried along a sample line
2.2 Mixtures evaluations
into a sample conditioning unit (without dilution)
After in-plant production, a certain amount of where it is dried and filtered to remove particulate
each asphalt mixture was collected and brought matter and moisture (also the latter was measured
to the laboratory located right near the plant. By through a specific control unit). Once conditioned,
means of the Superpave Gyratory Compactor the sample enters a sampling manifold. Here,
(SGC) set up with standard compaction param- individual gas analyzers extract each investigated
eters (vertical pressure 600 kPa; rotation speed: component and measure the corresponding concen-
30 revolutions/min; vertical mold inclination: tration by means of various techniques (e.g. infra-
1.25°), this material was used to compact cylindri- red and ultraviolet adsorption, chemiluminescence,
cal specimens useful to check volumetric (i.e. air fluorescence, chromatography and spectrography).
voids, Compaction Energy Index) and mechani- Through a Data Acquisition and Handling System
cal properties (i.e. Indirect Tensile Strength) of (DAHS), the signal output from each gas analyzer is
the corresponding mixture, in accordance with recorded and automatically collected into an emis-
technical standard acceptance requirements. A sions database. One advantage of this method, com-
minimum of four samples was prepared per each pared to dilution extractive methods, is the ability to
material (diameter = 100 mm, height = 63.5 mm). measure the percent amount of oxygen in the sam-
The number of gyrations was selected based on ple, which is often required in the regulatory calcula-
technical standards specifications for motorway tions for emission corrections. Measurements were
pavements (i.e. 130 and 200 for OG and DG mix- performed both during warm and hot productions
tures, respectively). It is worth noting that prior to to obtain a reliable comparison at equal boundary
compacting, each mixture was kept for one hour in conditions (e.g. wind intensity and direction, air
a force-draft oven at a temperature equal to the in- temperature, humidity, asphalt plant configura-
situ compaction temperature (i.e. 120°C for WMA tion) so avoiding the reference to historical data of
mixtures, 160°C for HMA mixtures) in order to generic HMA productions and effectively isolat-
properly simulate the aging effects underwent dur- ing the impact of working temperatures reduction.
ing lay-down and transportation. Concurrently, With the aim to obtain reliable data, the monitor-
each mixture was subjected to solvent extraction ing of emissions was extended for several hours, so
in order to determine the actual bitumen content guaranteeing that the asphalt plant was working at
and the aggregate gradation. This evaluation was full capacity. Totally, four detections were realized,
useful to verify weather in-plant produced warm one per each type of WMA additive and one for the
mixtures complied with the original laboratory reference HMA mixture. In each case, the measure-
mix design or the reduced working temperatures ment point was the same and was chosen to detect
caused problems at the asphalt plant in terms of the most significant emission rate (measurement
material composition with consequences on volu- point set in correspondence of the stack connected
metric and mechanical properties. Moreover, in with the aggregate drying drum). Table 2 lists all
situ visual inspections during and after lay-down fluid dynamic and chemical parameters monitored,

146

TIS2017_Book.indb 146 2/8/2017 8:02:18 PM

 Table 2. Fluid dynamic and chemical parameters for quent volumetric properties. Although reduced
monitoring pollutant emissions. production temperatures, the mixture must guar-
antee good compactability and an air void content
Parameters Standard within the limit range prescribed by the technical
Flow rate ISO 16911:2013 standards. In this sense, the Compaction Energy
Temperature – Index (CEI), calculated for each investigated mix-
Humidity – ture using SGC data recorded during laboratory
Dusts UNI EN 13284 specimens compaction, can provide an idea of
Oxygen Italian DM 25.8.2000 the compaction efforts needed to achieve a cer-
Carbon monoxide (CO) Italian DM 25.8.2000 tain target density (Mahmoud & Bahia 2004). For
Nitrogen oxide (NOx) Italian DM 25.8.2000 dense-graded mixtures, CEI was determined as
Sulfur oxide (SOx) Italian DM 25.8.2000 area under the densification curve from the 8th
Volatile Organic Compound UNI EN 12619 gyration (considered to simulate the effort applied
(VOCs) by the paver) to the 92% density; for open-graded
Polycy. Aromatic Hydrocar. Italian DM 25.8.2000 mixtures, CEI was calculated as the area under the
(PAHs) densification curve from the 8th gyrations until the
final gyration. Concurrently, the air voids content
(Vm) was determined for each specimen based on
the maximum (EN 12697–05) and the bulk density
specifying the corresponding technical standard
of each material (EN 12697–06). Results are sum-
used to carry out the measurement.
marized in Figure 1 as average of four replicates.
In terms of compactability, mixtures prepared
with WMA additives generally achieved good per-
3 RESULTS
formance in line with the reference HMA mixture,
so demonstrating that all WMA products were able
3.1 Bitumen percentage and aggregate gradation
to overcome the higher bitumen viscosity caused
Table 3 summarizes the results for each mixture in by reduced production temperatures.
terms of total bitumen content (Bt), expressed in In particular, the additive C2 appears the one
percent by aggregate weight, and aggregate grada- able to guarantee the highest compactability apti-
tion, both determined after solvent extraction. tude, even higher than the reference material in
Regardless of the bituminous course considered, the case of open graded mixtures. Such an out-
warm mixtures were characterized by a total bitu- come is consistent with the operational mecha-
men content close to the design value. Also in terms nism of this product. As viscous regulator, it
of gradations, WMA mixtures were able to achieve should be able to modify bitumen viscosity, so
the desired aggregate skeleton, without significant allowing acceptable workability at reduced pro-
differences compared to the reference HMA and in duction temperatures.
total agreement with the grading envelope prescribed The same principle should characterize the
by the technical standards. The limited variabil- additive C3. However, mixtures prepared with the
ity detected is mainly attributed to the production latter always achieved higher CEI values compared
process and to the inherent heterogeneity provided to the other materials, proving the important influ-
by RAP aggregates. Such results suggest that in ence of the additive chemical structure.
plant productions of WMA mixtures in large-scale The additive C1 exhibited very good perform-
amounts are feasible and reduced temperatures do ance in case of dense-graded mixtures, whereas
not alter the material composition of the resulting it showed lower potential when added to porous
mixture even without needing of significant modifi- asphalts. The different behavior dependent on the
cations to the asphalt plant. Also, they validate the type of pavement course is reasonable considering
hypotheses assumed for calculating the total bitu- that the product C1 mainly acts as surfactant: open-
men content in case of reduced temperatures (i.e. graded mixtures have a discontinuous aggregate
lower RAP bitumen reactivation). The similarity skeleton with high air voids percentage. Therefore,
in terms of material composition among mixtures the internal friction during lay-down and compac-
excludes the influence of further parameters except tion is much higher and the final aggregate skeleton
reduced temperatures on mixture behavior, so allow- configuration is achieved earlier, so preventing the
ing a significant performance comparison. additive to express its entire potential in terms of
workability. However, the contact between aggregate
particles happens in few specific points. Consider-
3.2 Compactability and volumetric properties
ing that the product C1 is also composed by adhe-
One key aspect to judge the suitability of a warm sion enhancers, it could result particularly efficient
mixture is surely its workability and the conse- in terms of performance providing higher resistance

147

TIS2017_Book.indb 147 2/8/2017 8:02:18 PM

 Table 3. Aggregate gradation and bitumen content.

WEARING course

Sieve [mm] WC1 WC2 WC3 H Design Envelope

20 100 100 100 100 100 100 100

14 87.6 89.8 88.1 89.0 90.3 85 94
10 46.9 48.4 50.3 52.0 45.8 38 53
6.3 24.3 26.2 25.0 25.2 24.9 22 32
2 15.2 16.4 15.8 16.1 16.0 14 22
0.5 9.8 10.7 10.1 10.7 9.9 9 15
0.25 7.7 9.2 8.8 8.2 7.7 7 12
0.063 5.8 6.1 7.2 6.8 4.6 4 8
Bt [% by agg. wt] 5.00 5.23 5.14 5.10 5.25 –
BINDER course

Sieve [mm] WC1 WC2 WC3 H Design Envelope

20 95.8 97.1 94.9 95.4 91.2 85 98

14 82.7 84.0 78.4 80.9 78.5 70 87
10 66.0 62.2 64.2 61.6 61.4 58 78
6.3 49.3 47.3 48.9 47.4 48.2 46 66
2 27.8 29.9 28.5 29.8 28.9 25 38
0.5 12.0 12.2 13.8 15.4 14.1 11 21
0.25 8.2 8.5 9.3 10.3 9.1 7 17
0.063 6.4 6.0 6.2 5.8 4.7 4 8
Bt [% by agg. wt] 5.02 4.93 4.85 4.79 4.80 –
BASE course

Sieve [mm] WC1 WC2 WC3 H Design Envelope

20 93.8 92.3 93.9 92.5 83 73 94

14 76.3 73.6 75.6 76.6 65.4 51 76
10 57.1 58.6 57.6 55.3 56.2 40 64
6.3 38.5 43.8 42.0 39.9 47.4 31 55
2 26.2 26.3 26.3 24.3 23.7 19 38
0.5 13.8 10.6 12.7 9.9 13.4 8 21
0.25 9.6 6.6 9.1 6.9 8.7 5 16
0.063 5.8 5.0 5.2 4.0 4.4 4 8
Bt [% by agg. wt] 4.47 4.68 4.45 4.15 4.5 –

and balancing the lower workability effects. In fact, limit prescribed by the standards (i.e. Vm ≥ 2% at 200
an improved adhesion in the very limited contact SGC gyrations) differently to the HMA mixture that
points is a key aspect for open-graded mixtures, had an air voids content in line with what required.
especially when subjected to the action of water. Considering that all base mixtures were character-
As far as the air voids content is concerned, it is ized by similar aggregate gradations, such a result
possible to notice that all open-graded and binder suggests that the WMA additive amount considered
mixtures were characterized by values within the for designing the base course could be reduced. This
limit range prescribed by the technical standards overestimation can be related to a lower RAP bitu-
(i.e. open-graded mixture: Vm ≥ 14% at 130 SGC men reactivation than what theoretically hypoth-
gyrations; binder: Vm ≥ 2% at 200 SGC gyra- esized for the calculation (i.e. 60%). Probably, due to
tions). Moreover, generally the air voids trend was the higher percentage of RAP in the base course (i.e.
consistent with the CEI values trend: the higher the 30%) compared to the other bituminous courses, the
CEI, the higher the air voids. reclaimed bitumen release at reduced working tem-
On the contrary, for the base course, all WMA peratures is even lower than 60%. Since WMA addi-
mixtures exhibited air voids contents lower than the tives were dosed based on the hypothetical effective

148

TIS2017_Book.indb 148 2/8/2017 8:02:18 PM

 needed for evaluating the acceptability of a mix-
ture was based on the Indirect Tensile Strength
(ITS) calculated as following specified:

2⋅P
T =
ITS max [GPa] (1)
π ⋅tt d

where Pmax = maximum load in kN; t = specimen

height in mm; d = specimen diameter in mm.
Results are summarized in Figure 2 (as average
of four replicates for all the investigated mixtures)
along with error bars expressed in terms of stand-
ard deviation. All specimens were conditioned in air
(dry conditioning) at the test temperature of 25°C
for a minimum of four hours prior to testing. With
the aim to verify the water susceptibility of the wear-
ing course (crucial aspect for open-graded mixtures
due to the high air voids content), the ITS test was
performed on both dry (ITS_dry) and wet (ITS_
wet) specimens. Wet specimens were conditioned in
water at 40°C for a period of 72 hours prior to dry
conditioning and testing (according to EN 12697-12
Method A). The ITS ratio (ITSR) between wet and
dry specimens was calculated to quantify the effect
of water in terms of performance (Fig. 2).
Regarding open-graded mixtures, all materials
exhibited lower ITS values compared to the refer-

Figure 1. CEI and air voids.

bitumen acting in the mixture (i.e. 60%), the additive

effects in terms of workability were too high. Bear-
ing in mind this aspect for future investigations, it is
worth noting that the lower air voids of WMA mix-
tures could be partially attributed also to the slightly
higher bitumen content detected with respect to the
reference mixture (see Table 3).
In all cases, the in-situ visual inspection during
lay-down and compaction confirmed the good
results above described in terms of workability
and volumetrics. No specific issues were detected
by the operators during construction with respect
to the reference HMA mixture.

3.3 Indirect tensile strength

In accordance with technical standards require-
ments, the preliminary performance analysis Figure 2. ITS results.

149

TIS2017_Book.indb 149 2/8/2017 8:02:18 PM

 ence HMA, both in dry and wet conditions. How- the investigated gas medium and are calculated
ever, data recorded in dry conditions were always in relation to the reference oxygen concentration
higher than the limit prescribed by the technical established by the technical standards. Therefore,
standards regardless of the type of WMA additive all measured values were corrected as follows:
considered. In wet conditions, additives C2 and C3
determined a significant loss in performance com- 21 − O
pared with the reference mixture as confirmed by E= ⋅ EM (2)
21 − OM
the ITSR trend. On the contrary, the additive C1
was able to guarantee an optimum resistance also in where E = corrected concentration; EM = meas-
wet conditions, demonstrating an ITSR ratio even ured concentration; O = reference oxygen concen-
higher than the reference material. Such considera- tration; OM = measured oxygen concentration.
tion confirms what previously anticipated about Table 4 also indicates the threshold acceptance
the potential benefits deriving from the presence of values specified for each substance by the current
adhesion enhancers within the chemical structure regulations.
of the additive C1 that helps to improve the resist- Although all emissions data were within the
ance developed at the few contact points between threshold limit regardless of the mixture produc-
aggregate particles, so enhancing overall mixture tion temperatures considered, it is worth noting
performance especially against water action. that the average values measured for warm produc-
Regarding dense-graded mixtures (both binder tions were generally lower than those recorded for
and base course), no significant differences were HMA mixtures. In details, the following remarks
detected with respect to the HMA material. can be drawn:
WMA mixtures demonstrated ITS values equal
or even higher than the reference ones, regard- – a significant reduction (around 18%) in the nitro-
less of the additive type considered. This indicates gen oxide (NOx) concentration was detected in
that reduced production temperatures did not case of WMA productions (i.e. 52.4 mg/Nmc)
negatively affect mixture performance. It is worth with respect to HMA productions (i.e. 61.8 mg/
noting that also for the base course, although Nmc). This result is direct consequence of the
the potential overestimation of WMA additive lower temperature measured at the fumes stack of
amounts previously discussed, WMA mixtures the asphalt plant for WMA productions than for
were characterized by ITS values generally higher HMA mixtures (temperature reduction around
than the reference mixture. 14 °C). Such a finding confirms that, as expected,
In conclusion, large-scale productions of warm a direct correspondance exists between NOx con-
mixtures are possible even at low production tem- centration and plant operational temperatures.
peratures without need of modifying the produc- – Also the sulfur oxide (SOx) amount was lower
tion process, the asphalt plant assembly or the for WMA mixtures than for the reference HMA.
aggregate mix design. However, attention must be Although SOx emissions depend on both the con-
paid to the additive amount and the effective bitu- tent of the volatile sulfur in the raw materials and
men content, especially in presence of high percent- the type of fuel used (equal for all mixtures in the
age of RAP aggregates. Moreover, the selection of present study), the temperature achieved during
a WMA chemical additive with a chemical compo- the combustion process is, also in this case, the
sition able to promote bitumen-aggregate adhesion main parameter to control SOx emissions.
appears fundamental in case of open-graded mix-
tures for developing good water resistance. Table 4. Emissions data in terms of corrected concen-
tration E.
3.4 Emissions Production
As discussed in paragraph §2.4, concurrently to temperatures
the assessment of mechanical/volumetric proper- Standard limit
Parameters [mg/Nmc] WMA HMA
ties and the verification of the working process
feasibility of the asphalt plant, the environmental Dusts 20 11.5 11.8
impact of productions at low temperatures was CO 1081.0 1021.0
also evaluated. NOX 500 52.4 61.8
Table 4 summarizes the values obtained for those SOX 1700 113.2 129.0
parameters considered more relevant in terms VOC – 8.4 10.4
of environmental impact for both the reference IPA 0.1 < 0.067 < 0.067
HMA mixture and WMA mixtures (WMA results
Temperature 92.7 107.0
are expressed as the average of three independent
at the stack
detections). All data are quantified in terms of [°C]
concentration of each substance identified within

150

TIS2017_Book.indb 150 2/8/2017 8:02:19 PM

 – As far as the VOC concentration is concerned, ture reduction up to 40 °C is able to successfully
the same improving trend demonstrated by warm induce significant emissions reductions avoiding
productions compared with hot productions in the expensive implementation of specific solutions
terms of NOx and SOx can be observed, cor- (e.g. asphalt plant fairing) often required to obtain
responding to a COV concentration reduction acceptable emissions.
around 23% (i.e. 8.43 mg/Nmc Vs 10.4 mg/Nmc In a context of more and more urgent environ-
for WMA and HMA mixtures, respectively). mental needs, the overall findings obtained through
– No significant changes were detected in CO and this experimental investigation represent an impor-
IPA concentrations between WMA and HMA tant reference to promote the use of WMA technol-
productions. ogies for the production of new bituminous mixtures
in combination with the use of reclaimed aggregates
The above-described data demonstrate that in- and modified bitumens. Future works will involve
plant productions at reduced working temperatures also the monitoring of emissions during lay-down
can provide relevant benefits in terms of environ- and compaction to quantify the overall environmen-
mental impact, allowing a significant reduction in tal impact of WMA mixtures. Moreover, the trial
those substances considered more dangerous for sections will be continuously monitored to check the
air pollution. It is also worth noting that the moni- performance evolution of each section and obtain a
toring was carried out only at the asphalt plant reliable evaluation of long-term performance.
during production and did not involve lay-down
and compaction phases, aspect that could further
enlarge environmental benefits, with particular
emphasis for the impact on operators’ health. REFERENCES

D’Angelo, J., Harm, E., Bartoszek, J., Baumgardner, G.,

Corrigan, M., Cowsert, J., Harman, T., Jamshidi, M.,
4 CONCLUSIONS Jones, W., Newcomb, D., Prowell, B., Sines, R. & Yea-
ton, B. 2008. Warm-Mix Asphalt: European Practice.
The present paper describes a project concerning Report FHWA, USA, American Trade Initiatives.
the in-plant production of warm recycled mix- Dinis-Almeida, M. & Lopes Afonso, M. 2015. Warm
tures through WMA chemical additives. Volumet- Mix Recycled Asphalt – a sustainable solution. J
ric and compactability properties as well as main Clean Prod 107:310–316.
mechanical performance parameters were assessed Dinis-Almeida, M., Castro-Gomes, J., Sangiorgi, C.,
Zoorob, S.E., Afonso, M.L. 2016. Performance of
in accordance with the acceptance requirements Warm Mix Recycled Asphalt containing up to 100%
prescribed by the technical standards. Labora- RAP. Const Build Mater, 112:1–6.
tory results demonstrate the feasibility of large- Frank, B. & Prowell, B. 2014. Method for calculating
scale productions when recycling techniques and Warm Mix energy saving based on stack gas measure-
WMA technologies are combined without any ments. In Proceedings of the ISAP 2014, 1:49–58.
need of modifications to the asphalt plant or to the Frigio, F. & Canestrari F. 2016. Characterisation of
production process. No alterations in the mixture warm recycled porous asphalt mixtures prepared with
composition (i.e. aggregate gradation and bitumen different WMA additives. Eur J Envir and Civil Eng,
content) were caused by lower production tem- published online.
Frigio, F., Stimilli, A., Bocci, M. & Canestrari F. 2016.
peratures. Optimum performance were achieved
Adhesion properties of warm recycled mixtures pro-
by WMA mixtures, with minimum gap compared duced with different WMA additives. In Proceedings
with the reference HMA both in terms of volumet- of the 4th CEW, Delft (The Netherlands).
rics (i.e. air voids), compactability (i.e. CEI) and Harrison, R. & Kemmer, T. 1996. Preferred and alterna-
Indirect Tensile Strength. Results also highlight the tive methods for estimating air emissions from hot-mix
importance of the WMA additive chemical struc- asphalt plants. In Final Report of Emission Inventory
ture to balance the lower production temperatures. Improvement Program (EIIP), Vol. 2 Chapter 3.
Particularly, for open-graded mixtures the use of Lee, S.J., Amirkhanian, S.N., Park, N.W., & Kim K.W.
an additive composed of surfactants and adhe- 2009. Characterization of warm mix asphalt binders
containing artificially long-term aged binders. Const
sion enhancers appears the most suitable solution
Build Mater 23(6): 2371–2379.
to guarantee low water susceptibility. Attention Mahmoud, A.F.F. & Bahia, H.U. 2004. Using the gyratory
must be paid to the additive amount considering compactor to measure mechanical stability of asphalt
the effective RAP bitumen reactivation degree. mixtures. Wisconsin Highway Research Program 05–02.
Specific study are recommended to investigate Rubio, M.C., Martìnez, G., Baena, L. & Moreno, F. 2012.
this aspect. Finally, the monitoring of pollutant Warm mix asphalt: an overview. J Clean Prod, 24:76–84.
emissions allowed an overall evaluation of WMA Stimilli, A., Virgili, A. & Canestrari, F. 2015. New
mixtures benefits. The findings in terms of pollut- method to estimate the “re-activated” binder amount
ants concentration demonstrated that a tempera- in recycled hot-mix asphalt. RMPD 16:442–459.

151

TIS2017_Book.indb 151 2/8/2017 8:02:19 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Environmental and engineering performance assessment of biofilters

and retention systems for pavement stormwater

M. Bassani, L. Tefa, E. Comino, M. Rosso & F. Giurca

Politecnico di Torino, Torino, Italy

A. Garcia Perez & R. Ricci

Biosearch Ambiente s.r.l., Torino, Italy

F. Bertola & F. Canonico

Buzzi Unicem s.p.a., Casale Monferrato (AL), Italy

ABSTRACT: This work is aimed at investigating the use of green infrastructure technologies for the
infiltration of stormwater runoff from urban road pavements into soil, and as a means to support the
evapotranspiration and harvesting processes. Porous concrete pavements and vegetated biofilters were
considered to assess pollutant removal and hydraulic performances. The investigations were at bench-
scale for single materials, and at column and box scale for composite structures. The pollutant reduction
was determined on the basis of suspended solids, heavy metals and hydrocarbon concentration, with
promising results that were found to be dependent on filter material characteristics, layer distribution,
and permeability.

1 INTRODUCTION environmentally unfriendly, and there is universal

agreement on the need to limit their impact espe-
Increasing urbanization worldwide is having a cially in urban areas. Many commenters sustain
severe impact on the environment and quality of that urban transportation infrastructures may help
life. Available data confirm that 50% of the world’s the achievement of a “Low Impact Development”
population already lives in cities, and forecasts (LID), by limiting as much as possible any damage
indicate that this will rise to 67% by 2050 in devel- to soil, vegetation, and aquatic systems (US EPA
oping countries, and 86% in developed ones. Cities 2000, Prince George’s County 1999). To alleviate
will be home to about 6.3 billion inhabitants, while their effects, existing and future urban road facili-
rural areas will not face this demographic growth ties should be more environmentally friendly and
(van Leeuwen, 2015). sustainable. One solution is to include new “Green
The increase in urban land cover causes irre- Infrastructures” (GI) such as pervious pavements,
versible environmental changes, a major one being planter boxes, bio-swales, and other solutions as
the loss in its ability to absorb rainwater. In urban part of an urban stormwater management system.
areas, rain falls on impervious surfaces where it is The development of GI in urban areas is strongly
rapidly directed into the drainage sewer system, promoted by the European Commission (Dige
often overloading it and sometimes flooding the 2011). Nevertheless, there is limited experience and
roads. Furthermore, stormwater runoff carries a lack of knowledge of their potential benefits glo-
waste material, bacteria, heavy metals, and other bally and, specifically, in Italy. To bridge this gap,
dangerous pollutants with the first flush responsi- a team of researchers from the Department of
ble for the highest level of contaminants. Environment, Land and Infrastructure Engineer-
Ranging from 65% to 95% of the total surface ing of the Politecnico di Torino, Biosearch Ambi-
of an urban area, the built cover predominates on ente s.r.l., and Buzzi Unicem s.p.a. investigated the
pervious vegetated surfaces (Ferguson 2005). Road use of alternative GI technologies. Specifically, the
pavements, sidewalks, and parking lots occupy work aimed at the assessment of their capabili-
most of the urban land destined for industrial, com- ties to infiltrate, evapotranspirate, and/or harvest
mercial, and residential use. In intensely built-up urban stormwater runoff, thus contributing to a
areas, they cover more than 50% of all the land. reduction in the water discharged into the drainage
As a result, road infrastructures are considered system.

153

TIS2017_Book.indb 153 2/8/2017 8:02:21 PM

 In the research program, two GI technologies, geotextiles are able to reduce the concentration of
porous concrete pavements, and biofilter boxes hydrocarbons in runoff.
to treat the water runoff pollutants from imper- Rahman et al. (2015) demonstrated that the use
vious pavements, were investigated under identi- of construction and demolition waste (CDW) in
cal experimental conditions to derive ecological combination with geotextiles increases pollutant
and hydraulic indicators useful to their design as removal. CDW was found to have geotechnical and
urban drainage facilities. A comparison between hydraulic properties equivalent or superior to that
the hydraulic performance and pollutant removal of conventional granular materials from quarries.
capacity of the two GI systems, can be of use to The described advantages can be eroded by
designers who wish integrate them into an urban clogging phenomena that can occur over time, as
stormwater management system. well documented in Abbott & Comino-Mateos
(2003), and Kayhanian et al. (2012). Thus, pervious
pavements need regular maintenance (Scholz &
1.1 Pervious pavements
Grabowiecki 2007).
The use of pervious pavements is well established
in the LID approach (Dietz 2007). They include
1.2 Biofilters
porous asphalt, pervious concrete, concrete blocks,
plastic or concrete grid systems (Eisenberg et al. Several studies have promoted the use of vegeta-
2015). Their application in urban areas results in a tion and soils for the interception and infiltration
reduction in stormwater volumes that would oth- of stormwater runoff from road pavements, to
erwise go directly into the storm-sewer system (US improve water quality and hydrological response
EPA 1999, Holman-Dodds et al. 2003). Aihablame of a site. While it is clear that in urban areas and
et al. (2012) reported that pervious pavements can near rural roadways, stormwater treatments are
reduce the runoff by between 50% and 93% in com- useful to mitigate the effects of pollutants, there
parison with traditional impervious pavements. An is a distinct lack of knowledge in the literature as
in situ investigation of permeable pavements by regards the integration of full-scale stormwater GI
Abbott & Comino-Mateos (2003) demonstrated a technologies in the urban environment.
reduction in peak flows and an extended duration Some authors (Dietz, 2007; Davis et al., 2009)
of outflows. Collins et al. (2006) observed a reduc- use the term “biofiltration” or “bioretention” to
tion in the peak flow rate for pervious pavements indicate soil and vegetation based systems used
compared to the peak runoff flows from usual for stormwater pollution removal. The biofiltra-
asphalt surfaces: an 80% reduction for porous con- tion process includes adsorption, plant uptake and
crete, 72–84% for permeable interlocking concrete removal of pollutants, plus biological transforma-
pavers, and 89% for concrete grid pavers. tion and degradation (Kadlec & Wallace, 2009).
Road pavements are also a source of pollut- According to Zinger et al. (2013), the most danger-
ants, which are washed off by runoff and then ous pollutants in terms of ecological toxicity are
contaminate water (Sansalone & Buchberger 1997, Cd, Pb, Ni, Fe, Mg, Cu, Zn, Cr, suspended solids
Drapper et al. 2000). However, numerous authors and hydrocarbons. They demonstrated that the
showed that porous pavements can mitigate this level of concentration of pollutants in stormwater
problem. Legret et al. (1996) compared the quality can vary greatly depending on the particular site
of runoff from porous pavements with that from and the time elapsed since the last rain event.
conventional surface drainage systems, finding a The removal efficiency depends on the charac-
reduction in pollution concentration in terms of teristics of the materials used to filter stormwater
suspended solids (64%), and heavy metals such (Lim et al. 2015). Compost had high removal effi-
as Pb (79%), Zn (72%), Cd (67%). Furthermore, ciencies for heavy metals (>90%), while potting soil
Dierkes et al. (1999), Pagotto et al. (2000), Fach and commercial mix offered the best uptake for Zn
and Geiger (2005), Gilbert and Clausen (2006) and Cd (Reddy et al. 2015). However, no single fil-
calculated the capacity of different pervious pave- tering material was capable of removing all heavy
ments in lowering heavy metal concentration. metals (Li et al. 2016). Therefore, a combination
These systems have also been recognized as being of filtering materials should be investigated for the
efficient in the retention of hydrocarbons (Pratt simultaneous removal of heavy metals. A contribu-
et al 1999). Coupe et al. (2003) tried to inoculate tion may also be provided by vegetation and the
a permeable pavement surface with specifically microbial community that can grow on geotextiles
hydrocarbon-degrading microorganisms, finding and/or inside the filtering materials (Mazer et al.
that the treatment did not improve the oil deg- 2001, Bayon et al. 2015). Most of the growth in
radation capability in comparison with the natu- biofilm responsible for degrading pollutants is con-
rally developed microbial communities. Newman centrated in the geotextile layer. This is due to its
et al. (2004) found that pervious pavements with retention capacity, which stabilises microorganisms

154

TIS2017_Book.indb 154 2/8/2017 8:02:21 PM

 and provides them with enough time to consume growth was evaluated through direct and micros-
hydrocarbons. copy observation.
Experiments on biofilters have been docu- After biofilm development, this geotextile was
mented from the laboratory to the full scale. For included in GI systems in some interfaces between
example, Zinger et al. (2013) based their lab experi- layers. Its pollutant removal properties were evalu-
ences on small columns, while Legret et al. (1996) ated in preliminary bench-scale tests, as described
and Kazemi et al. (2015) based their work on large in Section 2.3.
boxes. A combination of biofilters with rainwater
tanks was investigated by Demuzere et al. (2014),
2.2 Stormwater
while Hurley & Forman (2011) contributed to the
assessment of a detention basin. Small volumes of input stormwater were collected
from urban pavements in two steps, and then used
to run tests at the bench and column scales. The
2 MATERIAL AND METHODS differences in concentration of suspended solids,
heavy metals (Cd, Cr, Fe, Mn, Ni, Pb, Cu, Zn) and
In this investigation, the pollutant retention capac- hydrocarbons between the two stormwater sam-
ity of several combinations of biofilters and per- ples are reported in Table 1 and Table 4.
vious pavements was evaluated starting from the A larger volume of stormwater was also synthe-
bench scale (small samples of single filtering mate- tized in a tank, mixing tap water with a concen-
rial), to complete laboratory systems in which the trated solution of heavy metals and diesel fuel. The
materials were combined in different ways to form blend of pollutants included was set to be quite
columns and boxes. At the same stage, hydraulic different from the one collected from the pavement
performances were evaluated to obtain fundamen- with more hydrocarbons and fewer heavy metals.
tal parameters for design purposes. The investiga- The results for pollutant concentrations are listed
tion involved the following stages: in Tables 5, 6 and 7.
1. the development of a microbial biofilm on arti-
ficial supports; 2.3 Preliminary pollutant removal assessment at
2. the assessment of ordinary and recycled materi- the bench scale
als for the formation of pervious pavements and
A first materials assessment of the two GI technol-
biofilters;
ogies was carried out on the basis of the literature
3. the formation of different system configurations;
review, previous experience, and local availability.
4. their analysis in terms of hydraulic and pollut-
The materials considered included: (a) natural
ant removal properties.
sand, (b) organic soil, (c) compost, (d) expanded
According to Zinger et al. (2013), the laboratory clay, (e) geotextile with biofilm, and (f) subgrade
chemical test methods on pollutant removal prop- soil. All of them were submitted to a preliminary
erties carried out on inflows and outflows included: bench scale test to get information on their pollut-
(a) concentration of suspended solids (AP-AT ant reduction capability (Figure 1).
CNR IRSA 2090 B Man 29: 2003), (b) concentra- The stormwater was poured into glass bowls
tion of heavy metals (UNI EN ISO 17294-2:2005), containing the abovementioned materials and
and (c) total hydrocarbons (AP-AT CNR IRSA filtrated for 3 hours. Outflow filtered water was
5160 B2 MAN 29: 2003). collected and tested for the presence of the same
pollutants and pH.
As shown in Table 1, input stormwater had a
2.1 Biofilm
high content of hydrocarbons and suspended sol-
A microbial biofilm inoculum was prepared to ids. Conversely, metal content was very low. Fil-
introduce active microorganisms into the two GI tration through a single media slightly altered the
systems to favour organic pollutant abatement neutral pH of the outflow, and led to a significant
by biodegradation processes. A polyester needle- reduction in hydrocarbon content.
punched non-woven geotextile (400 g/m2) with no The best performing substrate was the geotextile
chemical binders, was used as support for biofilm with biofilm, which decreased the initial amount of
development. 2700 μg/l of hydrocarbons to non-detectable lev-
For biofilm development, the geotextile was els. The same level of reduction was recorded for
introduced into a plastic container and irrigated sand, compost and subgrade soil, which showed an
continuously in a water recirculating system. abatement percentage of 99%.
Deionized water was added with 10 g/l of com- The inflow amount of 518 mg/l of suspended
post, incubated at room temperature for 5 days, solids was adequately reduced by material samples:
and filtered through the same geotextile. Biofilm from the 99% abatement by subgrade soil to the

155

TIS2017_Book.indb 155 2/8/2017 8:02:21 PM

 Table 1. pH and concentration of hydrocarbons, Suspended Solids (S.S.) and heavy metals in input stormwater and
outflow water.

Pollutant
[μg/l] InSW Sand OS C EC GTX SS

pH 6.8 7.4 7.5 7.5 6.7 7.7 N/A

S.S. [mg/l] 518.0 34.0 73.0 128.0 145.0 37.0 3.0
Hydrocarbons 2700 33 110 27 930 5 29
Cadmium <0.1 <0.1 <0.1 <0.1 <0.1 <0.1 <0.1
Chrome <2.5 <2.5 <2.5 <2.5 <2.5 <2.5 <2.5
Iron <25.0 <25.0 <25.0 <27.0 <25.0 <25.0 <25.0
Manganese 8.0 20.0 2.5 2.5 13.0 2.5 2.5
Nickel 1.2 1.2 1.0 2.3 5.8 1.0 1.0
Lead <1.0 <1.0 <1.0 <1.0 <1.0 <1.0 <1.0
Copper <10.0 <10.0 <10.0 <10.0 <10.0 <10.0 <10.0
Zinc <25.0 <25.0 <25.0 <25.0 <25.0 <25.0 <25.0

Notes: InSW = input stormwater, OS = organic soil, C = compost, EC = expanded clay, GTX = geotextile, SS = sub-
grade soil.
Table 4. pH and concentration of pollutants [μg/l] in
Input Stormwater (InSW), in outflow water from col-
Table 2. Density (γ) and coefficient of permeability (k)
umns (a) and (e).
of materials.
Vegetated Pervious
Materials γ [kg/m3] k [m/s]
Pollutant InSW biofilter (a) pavement (e)
Subgrade soil 2135 8.28 ⋅ 10−9 pH 7.4 6.9 7.5(*)
Subbase material 2277 2.35 ⋅ 10−7 S.S. [mg/l] 1127.0 82.0 4.0
Gravel 5–15 mm 1449 9.07 ⋅ 10−5 Hydrocarbons 12000 67 79
Gravel 15–30 mm 1419 N/A Cadmium 0.2 0.6 <0.1
CDW aggregates 1184 N/A Chrome 3.8 8.8 <2.5
Organic soil 594 1.00 ⋅ 10−4 Iron 1100.0 2200.0 <25.0
Concrete mix 1 (4–16 mm) 1873 1.24 ⋅ 10−4 Manganese 210.0 505.0 295.0
Concrete mix 2 (10–20 mm) 1919 1.24 ⋅ 10−4 Nickel 15.0 125.0 9.5
Lead 10.4 12.5 38.0
Copper 41.5 79.5 7.7
Table 3. Total Layer Volume (LV), Water Storage Capac- Zinc 130.0 N/A N/A
ity (WSC) in the layers, Void content for single layers (V)
and Average for the whole system (AV) in the five boxes. Note: (*) in this test the concrete layer was not included
in the column (e).
Box of LV WSC V AV
Figure 3 Layer [m3] [m3] [%] [%] Table 5. pH and concentration of pollutants [μg/l] in Input
Stormwater (InSW) and outflow from biofilter (a) and (b).
(b) OS 0.106 0.029 27 33
G5–15 0.053 0.019 36 System InSW System (b)
G15–30 0.053 0.023 43
Layer – OS G15–30
(c) OS 0.106 0.031 29 44
CDW 0.106 0.062 58 Depth [cm] – 40 80
(d) OS + CDW 0.106 0.025 24 29
G15–30 0.106 0.036 34 pH 6.4 6.8 6.8
(e), (f) CL 0.240 0.084 35 28 S.S. [mg/l] 1.0 10.5 9.4
Subbase 0.240 0.055 23 Hydrocarbons 117000 860 550
Subgrade 0.480 0.134 28 Cadmium 17.5 <0.1 <0.1
Chrome 7.0 <2.5 4.0
Notes: OS = organic soil, CL = concrete layer, Iron 34.0 34.0 34.0
G = gravel. Manganese 107.0 33.0 12.0
Nickel 2.0 26.0 57.0
Lead 51.0 <1.0 <1.0
72% by expanded clay. Metal content remained Copper 67.0 14.2 36.9
very low after filtration with no significant varia- Zinc 272.0 52.0 54.0
tions between inflow and outflow values.

156

TIS2017_Book.indb 156 2/8/2017 8:02:21 PM

 Table 6. Concentration of pollutants [μg/l] in the Input 2.4 Physical characterization of materials
Stormwater (InSW) and in outflow from biofilter system
(c) and (d). In the first stage of the investigation, the organic
soil, the subgrade soil, and the geotextile were con-
System InSW System (c) System (d) sidered suitable to be employed in the formation of
the two GI technologies.
Layer – OS CDW OS+CDW G15–30 Additional granular materials were also
included to build layers. Specifically, three natural
Depth [cm] – 40 80 40 80 unbound granular materials (a granular pavement
pH 6.4 6.8 7.2 7.1 7.0
subbase, and two gravels 5–15 and 15–30 mm),
S.S. [mg/l] 1.0 10.5 12.2 10.1 7.6 and a Construction and Demolition Waste (CDW)
Hydro- 117000 860 560 550 570 were employed. They are used in road pavements,
carbons embankments, and biofilters. CDW was submitted
Cadmium 17.5 <0.1 0.1 0.1 0.1 to a leaching test in accordance with Italian stand-
Chrome 7.0 <2.5 54.0 6.0 6.0 ards (UNI 10802) to assess the heavy metal con-
Iron 34.0 34.0 32.0 45.0 32.0 tent of the eluate. Results in μg/l are: Cd < 0.25,
Manganese 107.0 33.0 44.0 13.0 4.0 Cr = 12.5, Fe = 5.7, Mn < 0.14, Ni < 1.5, Pb < 4.2,
Nickel 2.0 26.0 23.0 11.0 30.0 Cu = 4.8, Zn = 4.5. All these values fall within Ital-
Lead 51.0 <1.0 1.3 1.1 2.9 ian limits (D.Lgs 152/2006), so the CDW was con-
Copper 67.0 14.2 17.6 10.9 19.5 sidered suitable in the investigation.
Zinc 272.0 52.0 31.0 49.0 73.0 Figure 2 exhibits the particle size distribution
of the granular materials selected, with the excep-
tion of the organic soil. The subgrade soil had a
Table 7. pH and concentration of pollutants [μg/l] in well-graded particle distribution in a wide range
the Input Stormwater (InSW) in paved systems (e and f). of diameters (0–50 mm). The granular material
for subbase layers presented a well-graded parti-
Pollutant InSW System (e) System (f) cle distribution in the 0–30 mm range, while the
Sub- Sub- Sub- Sub- CDW particles and the two gravels (15–30 and
Layers – base grade base grade 5–15) exhibit a discontinuous distribution in the
same range, which is typical for materials selected
Depth [cm] – 40 90 40 90 through a sieve operation in a relatively small size
interval. Their grain dimension indicates gravels
pH 6.4 11.3 10.9 11.8 11.5 which have high permeability. Figure 2 also shows
S.S. [mg/l] 1.0 14.8 43.3 16.3 14.0 the grain size distribution of the aggregates used in
Hydro- 117000 10500 1500 740 670 the two concrete layers described later.
carbons The materials used to form layers in columns
Cadmium 17.5 0.1 0.2 <0.1 <0.1 and boxes were subjected to density and perme-
Chrome 7.0 5.0 8.0 8.0 14.0 ability analysis. Loose materials were compacted
Iron 34.0 <25.0 192.0 <25.0 39.0 into known volume moulds, weighed, and then
Manganese 107.0 3.0 17.0 3.0 4.0 tested in a permeameter under a constant hydrau-
Nickel 2.0 2.0 2.0 3.0 2.0 lic head.
Lead 51.0 <1.0 5.8 <1.0 1.9
Copper 67.0 <10.0 <10.0 <10.0 10.0
Zinc 272.0 <25.0 39.0 <25.0 28.0

Figure 1. Bench scale tests on different media. On

the left: grey coloured input stormwater and transpar-
ent filtered outcome water. On the right: water outflow
samples. Figure 2. Particle size distribution of materials.

157

TIS2017_Book.indb 157 2/8/2017 8:02:22 PM

 In the case of subgrade soil and subbase granu-
lar material, Table 2 reports the average dry density
derived from the Proctor compaction method on
samples prepared at the optimum moisture con-
tent and at this value ±2%.
Table 2 also contains the unit mass of com-
pacted samples of the two gravels, the CDW and
the organic soils. For the two concrete mixtures,
the density was measure on compacted cylindrical
samples.
The subgrade soil and the subbase material
present the lowest coefficients of permeability, on
account of their grain size distribution (Figure 2).
Very high coefficients of permeability were
observed for the other materials. In particular, the
two concrete mixtures for porous pavement layers
exhibited the same value of permeability coeffi-
cient (1.24 ⋅ 10−4 m/s), due to their high void con-
tent. The gravel with particles gradation 5–15 mm
had a permeability of 9.07 ⋅ 10−5 m/s. Due to its
very low density and high porosity, the organic
soil was the material with the highest coefficient of
permeability.

2.5 Columns and boxes Figure 3. Layered systems used in columns and boxes:
(a) vegetated column, (b) vegetated box with traditional
Figure 3 reports the stratigraphy of cylindrical col- gravel drainage layers and organic soil, (c) vegetated box
umns (20 cm in diameter) and boxes (rectangular with CDW drainage layer and organic soil, (d) vegetated
60 × 100 cm for pavements, and cylindrical 60 cm box with traditional gravel drainage layer and a mixture
diameter for vegetated biofilters). Thicknesses of of organic soil and CDW, (e) pervious pavement column
the layers were defined on the basis of literature and box with pre-treated geotextile with biofilm, (f) per-
vious pavement box with ordinary geotextile.
review and current practices. Materials were laid
to form compacted layers of uniform thickness. In
particular, granular materials and concrete mix-
tures were manually compacted in columns with a The same stratigraphy in terms of materials and
Proctor hammer, while a falling weight acting on a layer thickness was adopted for the system placed
rigid plate was used to compact layers in boxes. in a box in Figure 3b, with the distinction that in
Vegetated systems need to meet requirements to this case the geotextile was not pre-treated with
support the vegetation growth, although they were biofilm.
not planted during this first investigation. Pervi- To evaluate the ecological and economic poten-
ous pavements require structural and functional tial of vegetated systems with recycled materials,
characteristics in terms of quality of materials and two additional systems containing CDW were set
installation. In the meantime, all they need to do up. A first box (Figure 3c) was prepared with a
is satisfy water infiltration and pollution reduction bottom layer of 45 cm of CDW for water storage
principles. When making a comparison, it should purposes, and a top layer of organic soil. A second
be noted that in pervious pavements the drainage box (Figure 3d) was prepared with a bottom layer
layers are in the upper part of the system, whereas of coarse gravel (15–30 mm) and a top layer made
they are located in the lower part of vegetated up of a mixture of two thirds organic soil and one
biofilters. third CDW.
The vegetated biofilter of Figure 3a was built Pervious pavement stratigraphy was defined
in a column and is composed of a double layer according to ordinary solutions for low-volume
of coarse (15–30 mm) and fine gravel (5–15 mm) roads and parking lots. Two Portland cements,
divided by the geotextile previously treated with CEM I 52.5 R for concrete mixture 1, and CEM
the microbial biofilm. The two granular layers II/A-LL 42.5R for concrete mixture 2, were used
contribute to drainage and water retention, and to in a quantity of 350 kg/m3. To avoid the release of
promote physical processes for pollution removal. Cr VI, in the production of these two cements fer-
A 40 cm layer of organic soil was placed on top rous sulfate was added to the cement powder as a
to create the substrate for vegetation rooting. reducing agent.

158

TIS2017_Book.indb 158 2/8/2017 8:02:22 PM

 Mixture 1 was made up of aggregates of sizes 3.2 Pollutant removal assessment
4–16 mm, and with a Water/Cement (W/C) ratio
3.2.1 Columns
equal to 0.31. To improve workability, a plasticizer
Table 4 lists the concentration of hydrocarbons,
was added, while the cohesion between particles
suspended solids, and heavy metals in the input
was enhanced through the addition of a viscosity
stormwater and in the outflow water of the two
modifying admixture for self-compacting concrete.
columns (a) and (e) in Figure 3.
Aggregates of 10–20 mm were employed in mix-
These systems performed very well in hydrocar-
ture 2 with a W/C ratio equal to 0.38. Compressive
bon abatement, since they resulted in a reduction
strength on cubic samples at 28 days of curing was
of 99% in the starting concentration of 12000 μg/l.
7.1 MPa for mixture 1, and 8.4 MPa for mixture 2.
In terms of suspended solids and heavy metals, the
A column (Figure 3e) and two boxes (Figure 3e
pervious pavement outperformed the vegetated
and 3f) were built with a subgrade of 40 cm, a sub-
one. The paved column (e) decreased the con-
base layer of 20 cm, and two porous layers made
centration of Cd, Cr, Fe, Ni and Cu, while water
up of concrete mix 1 and 2 with a combined thick-
passing through the vegetated column (a) collected
ness of 10+10 cm.
huge quantities of these metals from the filter-
The pores of the upper (surface) layer were
ing materials, since their concentrations resulted
smaller to contrast the clogging produced by urban
higher than those of the input stormwater.
wastes (leaves and trash), making urban road clean-
ing operations easier. The lower concrete layer had
a higher void content that resulted in a greater 3.2.2 Boxes
direct water storage capacity. The concentration of pollutants in the vegetated
In systems (e) and (f) subbase and subgrade were box (b) in Figure 3, in comparison with the input
separated by the geotextile to avoid layer contami- stormwater is provided in Table 5. The hydro-
nation and, in the case of system (e) the geotextile carbons were strongly reduced (abatement of
was pre-treated to develop biofilm. 99–100%), as well as heavy metals, with the excep-
In order to prepare and clean cylinders and tion of Fe that did not change its concentration,
boxes, they were firstly saturated with tap water, and Ni that seemed to be released from the system.
allowed to dry for 3 days before being saturated Water passing through the filter obviously col-
with polluted stormwater for 3 hours. Outflow lected a huge amount of suspended solids, espe-
samples were collected in some interfaces between cially in the organic soil layer.
layers and at the bottom. The released water was A comparison of the results for the two layers
stored in glass bottles and then analysed. provides evidence that the gravel in the biofilter is
able to reduce S.S., hydrocarbons and Mn. It does
not significantly affect the pH and the concentra-
3 RESULTS AND DISCUSSION
tions of Fe, Cd, Cr, Pb and Zn, while it produces
an increase in Ni and Cu.
3.1 Hydraulic assessment
The concentration of pollutants in outflow water
The hydraulic capacity of boxes was evaluated from the biofilters (c) and (d) is shown in Table 6.
in terms of Water Storage Volume (WSV), void Both performed well in hydrocarbon reduction
content for each single layer (V), and weighted (abatement of 99–100%), while the pH value did
Average Void content (AV). Table 3 reports the val- not change significantly remaining around the neu-
ues measured in the five boxes of Figure 3, also tral value. Compared to system (a), the addition of
including the total Layer Volume (LV). a previously developed biofilm of bacteria did not
The vegetated boxes exhibited a higher water have a predominant role in pollutant abatement.
storage capacity with values ranging from 29 to As expected, in boxes (c) and (d) the level of
44% of voids. In particular, the vegetated biofilter suspended solids detected were higher in the out-
(c) proved to be the best in water retention due to flow than in the inflow due to the presence of the
the considerably high void content (58%) of the organic soil layer. CDW materials released Cr in
CDW aggregate layer. The organic soil layers of box (c) and Fe in box (d). Both systems, like that
vegetated boxes were able to store water from 25 to of Table 5, added Ni and Cu to the outflow. Other
31 dm3 in a layer with a total volume of 0.106 m3. heavy metals experienced a significant reduction in
Considering the two gravel layers of box (b), WSV their concentration levels.
reduces when particle size decreases. When considering results at the same depth
The paved boxes showed a lower average void from the surface, the layer consisting of OS and
content equal to 28%, due to low WSC of sub- CDW (system d) seems to perform better than the
grade and subbase layers. The pervious concrete layer with OS only. This suggests that CDW are
layers functioned as a reservoir, collecting in total more performant when used in combination with
35 dm3/m2. organic soil.

159

TIS2017_Book.indb 159 2/8/2017 8:02:23 PM

 Table 7 presents the results obtained from pervi- the geotextile by itself promotes the fixation
ous pavements. At a depth of 40 cm, the box (e) of naturally developing microbial communi-
had a low capability in hydrocarbon reduction ties, thus confirming previous investigations in
(91%). But at 90 cm, the hydrocarbons in output literature;
water is reduced to 1500 μg/l, which corresponds 8. the differences observed in the pollutant removal
to an abatement of around 99%. In the case of box capacity of systems formed in cylindrical col-
(f), the reduction was greater than 99% at both umns and boxes require further investigation;
depths. Water passing through the two systems col- 9. the outflow concentration values of hydrocar-
lected suspended solids: their concentration rising bons, suspended solids, and heavy metals for all
(at a depth of 90 cm) from 1 mg/l to 43.3 mg/l and the systems satisfy the acceptance limits of Ital-
14.0 mg/l in boxes (e) and (f) respectively. ian regulations for wastewater.
As expected, water filtrated through concrete
increased its pH from 6.4 to 11.3 and 11.8 in sys-
tems (e) and (f) respectively. However, the pH ACKNOWLEDGEMENTS
decreased when the water passed through the
subgrade, reaching values equal to 10.9 and 11.5 The research performed by the Department of
respectively. Environment, Land and Infrastructure Engineer-
The levels of Cd, Mn, Pb, and Zn were greatly ing of the Politecnico di Torino, the Biosearch
reduced by the two paved systems. Although the Ambiente s.r.l., and the Buzzi Unicem s.p.a. was
majority of heavy metals were removed, their partly founded by Regione Piemonte in the frame
increase in the subgrade layer revealed that this of WIN_STREET Project (“Water IN STReet
material released some of the pollutants into the design with Environmental Engineering Tech-
outflow water. nologies—for urbanized areas”, code F.E.S.R.
2007/2013).
4 CONCLUSIONS

Looking at the results, the following conclusions REFERENCES

can be drawn:
Abbott, C.L. & Comino-Mateos, L. 2003. In-situ
1. pervious pavements are more effective in pol- Hydraulic Performance of a Permeable Pavement
lutant removal than vegetated biofilters made Sustainable Urban Drainage System. Water and
up of organic soils in combination with other Environment Journal, 17(3), 187–190.
materials; Ahiablame, L.M., Engel, B.A., & Chaubey, I. 2012.
2. this finding could be due to the differences in Effectiveness of low impact development practices:
layer permeability, which determines the rate of literature review and suggestions for future research.
Water, Air, & Soil Pollution, 223(7), 4253–4273.
stormwater filtration, with water moving fast in Bayon J.R, Jato-Espino D., Blanco-Fernandez E., Castro-
vegetated biofilters, and very slow in pavement Fresno D. 2015. Behaviour of geotextiles designed for
systems; pervious pavements as a support for biofilm develop-
3. this difference in filtration rate partially explains ment. Geotextiles and Geomembranes, 43, 139–147.
the differences in pollutant removal perform- Collins, K.A., Hunt, W.F. & Hathaway, J.M. 2006.
ance, since bacteria may have more time to Evaluation of various types of permeable pavements
treat the same stormwater when the water flows with respect to water quality improvement and flood
slowly; control. In 8th International Conference on Concrete
4. furthermore, denser materials present a greater Block Paving. San Francisco, CA: American Society
of Agricultural and Biological Engineers.
surface area for the biofilm development, the Coupe, S.J., Smith, H.G., Newman, A.P. & Puehmeier,
hydrocarbons and metal sorption (in which T. 2003. Biodegradation and microbial diversity
liquid substances are attracted by solid par- within permeable pavements. European Journal of
ticles as a consequence of physical-chemical Protistology, 39(4), 495–498.
processes); Davis, A.P., Hunt, W.F., Traver, R.G., Clar, M., 2009.
5. vegetated biofilters may benefit if an interme- Bioretention technology: overview of current
diate layer with low permeability coefficients is practice and future needs. Journal of Environmental
used at the bottom; this condition will certainly Engineering, ASCE, 135, 109–117.
be investigated in the near future; Demuzere M., Coutts A.M., Gohler M., Broadbent
A.M., Wouters H., Lipzig van N.P.M., Gebert L.,
6. the pollutant reduction performance of CDW 2014. The implementation of biofiltration systems,
increases when it is combined with organic rainwater tanks and urban irrigation in a single-layer
soil; urban canopy model. Urban Climate, 10, 148–170.
7. pollutant reduction is not affected by microbial Dierkes, C., Holte, A. & Geiger, W.F. 1999. Heavy metal
film inoculated through the geotextile, while retention within a porous pavement structure. In Proc.

160

TIS2017_Book.indb 160 2/8/2017 8:02:23 PM

 the Eighth International Conference on Urban Storm urban surface runoff using non-vegetated bioswale.
Drainage. Ecological Engineering, 95, 706–713.
Dietz, M.E. 2007. Low impact development practices: Lim W., Lim H.S., Hu J.Y., Ziegler A., Ong S.L. 2015.
A review of current research and recommendations Comparison of filter media materials for heavy
for future directions. Water, air, and soil pollution, metal removal from urban stormwater runoff using
186(1–4), 351–363. biofiltration systems. Journal of Environmental
Dige, G. 2011. Green infrastructure and territorial Management, 147, 24–33.
cohesion. The concept of green infrastructure and its Mazer G., Booth D., Ewing K., 2001. Limitations to
integration into policies using monitoring systems. vegetation establishment and growth in biofiltration
Technical Report 18. European Environment Agency, swales. Ecological Engineering, 17, 429–443.
Copenhagen, Denmark. Newman, A.P., Puehmeier, T., Kwok, V., Lam, M., Coupe,
Drapper, D., Tomlinson, R. & Williams, P. 2000. Pollutant S.J., Shuttleworth, A. & Pratt, C.J. 2004. Protecting
concentrations in road runoff: Southeast Queensland groundwater with oil-retaining pervious pavements:
case study. Journal of Environmental Engineering, historical perspectives, limitations and recent develop-
ASCE, 126(4), 313–320. ments. Quarterly Journal of Engineering Geology and
Eisenberg, B., Collins Lindow, K. & Smith, D.R. 2015. Hydrogeology, 37(4), 283–291.
Permeable Pavements. Reston VA: American Society Pagotto, C., Legret, M. & Le Cloirec, P. 2000. Comparison
of Civil Engineers. of the hydraulic behaviour and the quality of highway
Fach, S. & Geiger, W.F. 2005. Effective pollutant reten- runoff water according to the type of pavement. Water
tion capacity of permeable pavements for infiltrated Research, 34(18), 4446–4454.
road runoffs determined by laboratory tests. Water Pratt, C.J., Newman, A.P. & Bond, P.C. 1999. Mineral
Science and Technology, 51(2), 37–45. oil bio-degradation within a permeable pavement:
Ferguson, B.K. 2005. Porous pavements. CRC Press. long term observations. Water science and technology,
Gilbert, J.K. & Clausen, J.C. 2006. Stormwater runoff 39(2), 103–109.
quality and quantity from asphalt, paver, and crushed Prince George’s County 1999. Low-impact development
stone driveways in Connecticut. Water Research, design strategies: An integrated design approach.
40(4), 826–832. Prince George’s County, MD Department of Envi-
Gilbert, J.K. & Clausen, J.C. 2006. Stormwater runoff ronmental Resources.
quality and quantity from asphalt, paver, and crushed Rahman Md. A, Imteaz M.A., Arulrajah A., Piratheepan
stone driveways in Connecticut. Water Research, J., Disfani M.M. 2015. Recycled construction and
40(4), 826–832. demolition materials in permeable pavements systems:
Holman-Dodds, J.K., Bradley, A.A. & Potter, K.W. 2003. geothecnical and hydraulic characteristics. Journal of
Evaluation of hydrologic benefits of infiltration based Cleaner Production, 90, 183–194.
urban storm water management. JAWRA Journal Reddy K.R., Xie T., Dastgheibi S. 2014. Removal of
of the American Water Resources Association, 39(1), heavy metals from urban stormwater runoff using
205–215. different filter materials. Journal of Environmental
Hurley S.E. & Forman R.T.T. 2011. Stormwater ponds Chemical Engineering, 2, 282–292.
and biofilters for large urban sites: Modeled arrange- Sansalone, J.J. & Buchberger, S.G. 1997. Characterization
ments that achieve the phosphorus reduction target of solid and metal element distributions in urban
for Boston’s Charles River, USA Ecological Engineer- highway stormwater. Water Science and Technology,
ing, 37, 850–863. 36(8–9), 155–160.
Kadlec & Wallace 2009. Treatment Wetlands, 2nd ed. Scholz, M. & Grabowiecki, P. 2007. Review of permeable
CRC Press, Boca Raton, FL. pavement systems. Building and Environment, 42(11),
Kayhanian, M., Anderson, D., Harvey, J.T., Jones, D. & 3830–3836.
Muhunthan, B. 2012. Permeability measurement and US EPA (US Environmental Protection Agency) 1999.
scan imaging to assess clogging of pervious concrete Stormwater technology fact sheet. Porous pavement.
pavements in parking lots. Journal of Environmental Washington, D.C: Office of Water, EPA 832-F-99-023.
Management, 95(1), 114–123. US EPA (US Environmental Protection Agency) 2000.
Kazemi F. & Kelly Hill K. 2015. Effect of permeable Low impact development (LID), a literature review.
pavement basecourse aggregates on stormwater qual- Washington, D.C: Office of Water, EPA-841-B-00-005.
ity for irrigation reuse. Ecological Engineering, 77, van Leeuwen, K. 2015. Too little water in too many cities.
189–195. Integrated Environmental Assessment and Manage-
Legret, M., Colandini, V. & Le Marc, C. 1996. Effects ment, 11(1), 171–173.
of a porous pavement with reservoir structure on the Zinger Y., Blecken G.-T., Fletcher T.D., Viklander M.,
quality of runoff water and soil. Science of the Total Deletic A. 2013, Optimising nitrogen removal in exist-
Environment, 189, 335–340. ing stormwater biofilters: Benefits and tradeoffs of a
Li J., Jianga C., Leib T., Li Y., 2016. Experimental retrofitted saturated zone. Ecological Engineering, 51,
study and simulation of water quality purification of 75–82.

161

TIS2017_Book.indb 161 2/8/2017 8:02:23 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Slope hazards and risk engineering in the Canadian railway network

through the Cordillera

J.L. Rodriguez, R. Macciotta & M. Hendry

University of Alberta, Edmonton, Canada

T. Edwards & T. Evans

Canadian National Railway, Canada

ABSTRACT: The Canadian railway industry is a significant contributor to the Canadian economy,
with safe and fluid operations being priorities for the industry. The presence of the Cordillera in Western
Canada exposes railway corridors to numerous slope hazards with the potential to endanger rail-
way personnel, operations, equipment and infrastructure. Railways in Canada have a long history of
operation delays and infrastructure damage caused by slope instabilities. In this regard, slope hazards are
a substantial aspect of the risk engineering approaches adopted by the railways crossing the Cordillera.
This paper summarizes some engineering approaches applied to slope hazards along some sections in the
Canadian Cordillera. This is illustrated with three landslide case studies with the potential to negatively
affect railway operations. The risk engineering approaches adopted at each location are also presented,
highlighting the importance of proactive investigation to increase the knowledge about the particular
hazard with the aims of continuously enhancing the risk management strategies.

1 INTRODUCTION

Canadian railways have over 50,000 km of tracks

connecting urban areas, production regions and
ports across the country (Minister of Transport
2016), and are a significant contributor to the
Canadian economy. Most of the operations consist
of freight transportation by the two major railway
companies in Canada: Canadian Pacific Railway
(CP) and Canadian National Railway Company
(CN).
The railway network in western Canada crosses
the Cordillera (in British Columbia and western
Alberta) and connects the cities of Vancouver and
Prince Rupert with other provinces and markets
in the US (Figure 1). The corridors through the Figure 1. Canadian western railway network.
cordillera are exposed to varied ground hazards,
including slope instabilities, flooding, embank-
ment failures, ground subsidence, frost heave, and hazard control very challenging. In this context,
snow avalanches. The most frequent ground haz- adequate risk engineering strategies are neces-
ards in these corridors are slope instabilities (rock sary to maintain operations within tolerable risks.
falls, rock slides, soil slides, debris flows) (Macci- These strategies require a thorough understand-
otta et al. 2013). Examples of some of these are ing of the slope instability mechanisms in order to
illustrated in Figure 2. assess their likelihood and the potential impact to
The large extent of these corridors and the envi- the railway infrastructure and operations.
ronmental characteristics of the region (extreme The complexity of slope instability mecha-
weather events and complex geology) make ground nisms is reflected on the uncertainty associated

163

TIS2017_Book.indb 163 2/8/2017 8:02:25 PM

 framework, slope instabilities can be identified,
analyzed and evaluated, in order to take objec-
tive actions to control the hazards and minimize
risk (Porter 2013), therefore minimizing traffic
disruptions.
CN and CP have proactively been adopting
risk engineering measures to control potential and
existing slope instabilities in the Cordillera. These
control measures include displacement monitoring
and early warnings (slide fences), track deflection
monitoring, slope stabilization, Protection meas-
ures (ditches, embankments, rock sheds, rock fall
meshes), and operation strategies (i.e. patrols in
front of trains, reduced speeds). Slope stabilization
can involve water control (drainage and surface
water management), earth movement, retaining
structures, reinforcement (piles, anchors), or combi-
nations of the above (USGS, 2000). Although these
can prove effective, they are often cost-prohibited
and technically challenging. In these cases, com-
binations of protection, warning and operation
strategies provide feasible options.
This paper presents some examples of risk engi-
neering applied to slope hazards along railway cor-
ridors crossing the Cordillera.

2 ROCK FALLS ALONG THE FRASER

RIVER VALLEY AND ALONG CN’S
SQUAMISH SUBDIVISION

Two corridors with long histories of rock fall

occurrences are discussed. CN and CP main
lines travel along one of the corridors, which
runs parallel to the Fraser River valley between
Vancouver and interior British Columbia.
A particular section of this corridor is charac-
terized by steep slope cuts to accommodate the
railway alignment (Macciotta et al. 2013). The
other corridor (CN’s Squamish subdivision)
runs along Anderson and Seton lakes, and links
the city of Vancouver with the city of Lillooet,
British Columbia. This corridor is characterized
by steep slopes of sheared and weathered rock,
where rock falls frequently originate (Figure 3a)
(Macciotta et al. 2015b).
Rock falls are a frequent ground hazards along
these corridors, where rock slides and debris flows
are also present but less frequent. Proactive control
strategies in place include: scaling loose rock blocks,
slope reinforcement (rock anchors and shotcrete
Figure 2. Examples of slope instabilities along rail- surfaces), rock fall catchment nets and drapery
way corridors through the Cordillera: a) soil landslides,
b) rock slides, c) rock falls.
systems, and catchment ditches. Slide fence sys-
tems that detect material potentially blocking the
tracks, and protection sheds have been particularly
with assessing slope instability likelihood and adopted along the section within the Fraser River
consequences. Risk engineering measures under corridor. Patrols in front of trains, and train speed
such uncertain conditions are best dealt with constrains are characteristic of the section along
within a risk management framework. In this the Anderson and Seton lakes. Examples of some

164

TIS2017_Book.indb 164 2/8/2017 8:02:25 PM

 Figure 3. a) Steep rock slopes along mountainous rail-
way corridors in western Canada, b) railway equipment
removing fallen debris, and c) patrol in front of train.

control approaches are illustrated in Figure 3, and

Figure 4.
These risk engineering approaches control the
slope hazards along these corridors. Furthermore,
CP and CN keep records of slope instabilities that
include the location, volume, date, and environ-
mental conditions. These allow quantification of
the most hazardous sections of track, slope insta-
bility frequencies and their influence in the fluidity
of operations.
The extensive database on slope instabilities has
promoted research with the objective of enhanc-
ing the current risk engineering approaches
implemented by CP and CN. Macciotta et al.
(2015b) analyzed the trigger mechanisms of rock
falls along Anderson and Seton lakes to predict
periods of time when rock falls are more likely
to occur. These trigger mechanisms included
weather conditions, seismic events, and biologi-
cal influences. The results from this study helped
propose a rock fall risk control chart based on
the freeze-thaw cycles and precipitation rates
along the study area (Figure 5). This chart is
meant to aid scheduling maintenance work under
nonhazardous periods, increase awareness on the
hazardous periods, and increase the fluidity of Figure 4. Examples of rock fall hazard control: a) sheds
operations. (for rock fall, debris flow and avalanche protection),
b) rock anchors, c) attenuator nets.
165

TIS2017_Book.indb 165 2/8/2017 8:02:27 PM

 Figure 5. Weather base criteria for rock fall hazard along
the Squamish Subdivision, (after Macciotta et al. 2015).

3 RIPLEY LANDSLIDE

The Ripley landslide was initially identified in 1951

(Hendry et al. 2015) and did not show significant
signs of movement until 2005 when building a sid-
ing that required the construction of a retaining
wall to increase the construction area.
The increase in landslide displacement rate
was identified by shearing of two piezometers
within the Ripley Landslide and the development
of tension cracks and a back scarp in 2007. The
landslide moves at a rate between 25 and 180 mm/
year (Bunce & Chadwick 2012; Hendry et al. 2015).
A risk assessment in terms of loss of life (train
crew), environment, and operations concluded that
the high costs associated with slope stabilization
works were not reflected in significant reduction of
the overall risk to operations (Bunce and Chadwick
2012). The risks associated with displacement of the
landslide have been successfully managed through
regular inspections and maintenance. The risk engi-
neering approach adopted by CN and CP include Figure 6. Instrumentation at the Ripley Landslide:
displacement monitoring with a GPS system and a) GPS station, b) reflector for satellite radar technology,
continuous, in-place, down-hole, inclination meas- c) location of piezometer, Shape Accel Array, reflector
urements (Slope Inclinometers and recently Shape and data acquisition system.
Accel Arrays) that provide early notice of excessive
movement. Other approaches adopted at this site and their effect on track geometry have allowed
are groundwater monitoring, scheduled surveys proposing an enhanced early warning system based
using LiDAR imaging, regular track geometry sur- on displacement rates, trends of deformation and
veys and visual inspections (Figure 6). track quality (Macciotta et al. 2015).
The Ripley Landslide has also served as labora- The correlations found in these studies provide
tory for landslide research and assessment of new the basis to assess other landslide movements in the
monitoring technologies. Technologies that have area and optimize the risk engineering approaches
been deployed include an acoustic monitoring sys- adopted.
tem, satellite radar technology, and a network of
optical fiber to monitor deformations at the retain-
ing wall (Hendry et al. 2015, Huntley et al. 2014). 4 10-MILE SLIDE
Ongoing research activities include investigating
the relationship between slope displacements and The 10-Mile Slide is a landslide located in the
groundwater pressures, and the response of the province of British Columbia, near the town of
railway track to cumulative landslide displace- Lillooet, North-East of Vancouver. The importance
ments. Research on the landslide displacements of the landslide is associated with the location of a

166

TIS2017_Book.indb 166 2/8/2017 8:02:31 PM

 highway (Highway 99) and a railway line (now part
of CN) within its boundaries (Figure 7). The land-
slide is approximately 200 m wide and 140 m high,
with an approximate volume of 750 000 m3. The
10-mile Slide is sliding on a through-going shear
surface at velocities up to 10 mm/day.
The landslide related risks to the railway have
been managed through scheduled track geometry
measurements, visual inspection of the track
before each train passes (patrols), track deforma-
tion monitoring, and regular track maintenance.
A retaining wall was installed immediately
downslope from the railway tracks to prevent
deformations caused by loosening of materials
associated with the slope deformations and delay
retrogression of the landslide. Construction of the
retaining wall strengthened a cost-effective risk
Figure 7. 10-mile Slide and location of the GPS moni-
toring system. management strategy along the railway section.
Active research activities are also part of CN’s
monitoring efforts. Displacement monitoring of
some of the piles that compose the retaining wall
has allowed for the measurement of the response of
the wall to retrogression of the landslide. Applica-
tion of new technologies for landslide monitoring
also include aerial LiDAR and photogrammetry,
drone surveys, installation of a newly developed
cost-efficient GPS system, and monthly ground-
based LiDAR image acquisitions (Figure 8).
Currently, CN has decided to stabilize the sec-
tion of the landslide to maintain the alignment of
their track by supporting the slope with an array
of deep piles.

5 CONCLUSION

This paper illustrates the adoption of different risk

engineering approaches to slope instabilities along
railway corridors in the Canadian Cordillera.
These corridors are exposed to varied slope insta-
bilities, including rock falls and soil landslides. The
most suited engineering approach depends on the
type of slope instability addressed, its volume, its
particular mechanism and the resources available
for risk control. Table 1 presents a summary of the
approaches discussed.
The cases presented show the importance of a
proactive approach towards slope management.
Such proactive approach requires a thorough
understanding of the slope instability mechanisms
in order to assess their likelihood and the potential
impact to the railway infrastructure and opera-
tions. Furthermore, promoting research activities
has proven to enhance the risk engineering strate-
gies through increased understanding of the slope
Figure 8. 10-mile Slide instrumentation and survey: hazards. In turn, this is reflected in increased fluid-
a) GPS Unit, b) downhole slope displacement readings, ity of operations and resource optimization.
c) geotechnical investigation and instrument installation, The results and risk management techniques
d) remote sensing (LiDAR imaging). presented in this paper are site specific and do not

167

TIS2017_Book.indb 167 2/8/2017 8:02:33 PM

 Table 1. Risk Engineering Strategies applied in each Research Council of Canada (NSERC), Canadian
case history. Pacific Railway (CR), Canadian National Railway
Company (CN); and the Canadian Rail Research
Case Risk Engineering Strategies Laboratory (CaRRL) (www.carrl.ca), which is
Rockfall along the Includes: funded by NSERC, CP, CN, the Association of
Fraser River • Scaling loose rock blocks American Railways—Transportation Technology
Valley • Slope reinforcement (rock Center Inc., Transport Canada, the National
anchors and shotcrete surfaces) Research Council of Canada, and Alberta
• Catchment nets and drapery Innovates—Technology Futures.
systems
• Catchment ditches
• Slide fences
REFERENCES
• Protection shed
• Records keeping
Cui, Y., Miller, D., Nixon, G., and Nelson, J., 2015. British
Rockfalls along Includes:
Columbia digital geology. British Columbia Geologi-
CN’s Squamish • Scaling loose rock blocks
cal Survey, Open File 2015-2. Accessed August 25th,
Subdivision • Slope reinforcement (rock
2016. http://www.empr.gov.bc.ca/Mining/Geoscience/
anchors and shotcrete surfaces)
BedrockMapping/Pages/BCGeoMap.aspx
• Catchment nets and drapery
Hendry, M.T., Martin, C.D., Choi, E., Edwards, T. &
systems
Chadwick, I. 2013, “Safe Train Operations over a
• Catchment ditches
Moving Landslide.”, 10th World Conference on
• Patrol trucks
Railway Research, November 25–28.
• Speed constrains
Hendry, M.T., Macciotta, R., Martin, C.D. & Reich, B.
• Records keeping
2015, “Effect of Thompson River elevation on
Ripley Slide Includes: velocity and instability of Ripley Slide”, Canadian
• Down-hole surveys Geotechnical Journal, vol. 52, no. 3, pp. 257–267.
• Slope Inclinometers Huntley, D., Bobrowsky, P., Zhang, Q., Sladen W.,
• Shape Accel Arrays Bunce, C., Edwards, T, Hendry, M., Martin, D., and
• Groundwater monitoring Choi, E. 2014. “Fiber optic strain monitoring and
• Ground Surveys evaluation of a slow-moving landslide near Ashcroft,
GPS British Columbia, Canada”. Proceedings of World
Satellite Radar Landslide Forum 3, 6 p. Beijing, China.
Visual inspections Macciotta, R., Cruden, D.M., Martin, C.D., Morgen-
LiDAR imaging stern, N.R., & Petrov, M. 2013. “Spatial and temporal
• Acoustic monitoring system aspects of slope hazards along a railroad corridor in
• Network of optical fiber to the Canadian Cordillera”. In P.M. Dight (Ed.) Slope
monitor Stability 2013: International Symposium on Slope
10-mile Slide Includes: Stability in Open Pit Mining and Civil Engineering,
• Patrol trucks Brisbane, Australia. pp. 1171–1186.
• Ground Surveys Macciotta, R., Hendry, M., & Martin, C.D. 2015. “Devel-
GPS oping an early warning system for a very slow landslide
Satellite Radar based on displacement monitoring”. Natural Hazards,
Visual inspections 1–21. http://doi.org/10.1007/s11069-015-2110-2.
LiDAR imaging Macciotta, R., Martin, C.D. & Cruden, D.M. 2015a,
Photogrammetry “Probabilistic estimation of rockfall height and kinetic
Drone energy based on a three-dimensional trajectory model
• Slope Inclinometers and Monte Carlo simulation”, Landslides, vol. 12,
• Track maintenance no. 4, pp. 757–772.
Macciotta, R., Martin, C.D., Edwards, T., Cruden,
D.M. & Keegan, T. 2015b, “Quantifying weather
provide a guide that can be directly implemented on conditions for rock fall hazard management”, Georisk:
other slope instabilities. However, the case histories Assessment and Management of Risk for Engineered
presented in the paper highlight the importance of Systems and Geohazards, vol. 9, no. 3, pp. 171–186.
implementation of adequate risk engineering, and Minister of Transport 2016, Transportation Canada
the role of investigation and research to achieve 2011, Comprehensive Review, Transport Canada,
safe and fluid operations. Minister of Public Works and Government Services.
Porter M. & N. Morgenstern 2013, Landslide Risk
Evaluation—Canadian Technical Guidelines and Best
Practices related to Landslides: a national initiative for
ACKNOWLEDGMENTS loss reduction, Geological Survey of Canada.
Spiker, E.C. & Gori, P.L. 2003, National Landslide
This research was made possible by the (Canadian) Hazards Mitigation Strategy: A Framework for Loss
Railway Ground Hazard Research Program, which Reduction, U.S. Geological Survey, USGS Publica-
is funded by the Natural Sciences and Engineering tions Warehouse.

168

TIS2017_Book.indb 168 2/8/2017 8:02:36 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Computational model to conform aerodrome geometry to ICAO

design standards

E.J. Silva & N.D.F. Gualda

Escola Politécnica, University of São Paulo, Brazil

ABSTRACT: The article presents a computational model to conform an aerodrome geometry to the
ICAO Annex 14 standards. These standards impose geometric characteristics to runways, taxiways and
aprons, as well as protection surfaces free of interferences according to the project aircraft. The model
generates a KML file that can be read in a virtual globe such as Google Earth or in a GIS desktop as
QGIS. It encompasses both CAD and GIS routines written in Python language. It allows automatic anal-
ysis of possible aerodrome geometry conflicts. Model functionality has been verified for both hypothetic
and real world instances. The paper shows results of its application to the Viracopos Airport in Campinas,
Brazil (SBKP). A DTM (Digital Terrain Model) of the SBKP site was achieved with subsequent genera-
tion of the runway and of the associated OFZ (Obstacle Free Zone) surfaces using SRTM (Shuttle Radar
Topography Mission) data.

1 INTRODUCTION 1.1 Paper structure

Section 1 defines paper objectives and structure.
ICAO Annex 14 to the Convention on Interna-
In section 2, aerodrome components, their geom-
tional Civil Aviation (ICAO, 2004) sets forth man-
etry and related design standards are discussed.
datory design standards for international airports
Section 3 presents related existing software and
of 191 countries and is a main technical basis for
their features. In section 4, the proposed model is
domestic airports as well. Given a project aircraft,
described. Its application to a real world instance is
those standards require geometric characteristics
presented on section 5. Section 6 presents the study
for runways, taxiways and aprons, as well as pro-
conclusions.
tection surfaces free of interferences related to
airport and surrounding objects. Airspace inter-
ference outside the aerodromes must be checked
as buildings, antennas and other objects must not 2 AERODROME GEOMETRY
protrude a set of imaginary surfaces that start
in the runway vicinity and slope upwards and Whereas airside denotes the portion of an aero-
outwards. drome intended for aircraft operations, landside
This article, developed as part of a doctoral stands for passengers and cargo facilities (e.g.:
re-search, recognizes the complexity involved in terminals and parking lots). Textbooks, such as
assessing the compliance of a given aerodrome Horonjeff, et al. (2010) and Kazda & Caves (2007),
design to applicable design standards, especially present a comprehensive approach to both the air-
regarding its geometric aspects. After analyzing side and landside design principles.
existing tools and the aerodrome geometric char- This research tackles only the airside, from
acteristics, a computational model was developed the perspective of Annex 14 to the Convention
to analyze conformity of an aerodrome geometry on Inter-national Civil Aviation (ICAO, 2004).
to the ICAO Annex 14 standards (ICAO, 2004). This document establishes Standards and Recom-
The model incorporates and integrates CAD mended Practices (SARPs) to be adopted by 191
(Computer Aided Design) and GIS (Geographic signatory States in their international airports.
Information System) routines. Besides the auto- While standards are mandatory, recommended
matic generation of the related geometric entities, practices are optional, but States may turn them
the proposed model performs geometric con- obligatory, too. When standards cannot be met, an
flicts analysis, allowing the assessment of existing aeronautical study may be necessary to assure that
objects as to Annex 14 design standards. Results the TLS (Target Level of Safety) is attained. Silva
are shown for both real world and fictitious test & Gualda (2013) show the application of design
instances. standards flexibilizations to São Paulo/Guarulhos

169

TIS2017_Book.indb 169 2/8/2017 8:02:37 PM

 International Airport airside for A380–800 and 2.1.2 Protection areas in the air
B747–8 aircraft operation. In order to protect aircraft in the air, several imagi-
For domestic airports in Brazil, ANAC (National nary surfaces apply, which exclude fixed objects.
Civil Aviation Agency) adopts a design standards Additionally, during operations, temporary sur-
framework quite close to that of ICAO. faces are created on the ground and in the air to
FAA—Federal Aviation Administration (FAA, exclude movable objects, such as other aircraft and
2014) in the USA, on the other hand, issues a service vehicles. Approach and take-off operations
design standards framework for domestic air- require their own protection surfaces.
ports that substantially differs from the ICAO Annex 14 (ICAO, 2004) obstacle limitation sur-
one. faces apply to aerodrome planning and design,
As a result, the modelling proposed in this re- whereas PANS-OPS (ICAO, 2006) concerns flight
search fits most world international airports and procedures design and obstacle assessment.
domestic airports in countries where ICAO stand- Figure 1 shows the basic geometry of obstacle
ards are adopted. For different standards, such as limitation surfaces according to ICAO.
those imposed by FAA in the USA, model adapta- This set of surfaces control land use in the aero-
tions are required. drome neighborhood, especially the construction
Aerodrome airside design encompasses both of buildings and other infrastructure. On preci-
con-figuration analysis and sizing of physical com- sion approach runways, additional OFZ (Obstacle
ponents such as runways, taxiways and aprons, as Free Zone) surfaces are required. Such surfaces
well as their safety areas on the ground and in the are closer to the runways and represent significant
air. criteria for positioning taxiways, bypasses, runway
holding position marks and holding areas. The
standards for aerodrome obstacle limitation sur-
2.1 Airside components: geometry and standards
faces depend on the ARC code number and on the
Several aerodrome geometry criteria are estab- approach procedures to be adopted – e.g. visual,
lished according to ARC (Aerodrome Reference non-precision, or precision one.
Code), a classification that accounts for aircraft
dimensions and performance. 2.1.3 Airside geometry representation
Besides physical components, such as runways The airside geometry representation should be as
and taxiways, the airside comprises protection simple as possible, for model efficiency.
areas on the ground and imaginary surfaces in the Runway and taxiway centerlines can be repre-
airspace. sented by lines or points with a set of x,y,z coor-
dinates. Runways and taxiways, as well as other
2.1.1 Protection areas on the ground physical airside components, can be represented as
Protection areas on the ground surround runways polygons. This is also true for protection surfaces
and taxiways. on the ground or in the air. At most, polygonised
Runway Strip is a rectangle that surrounds the sur-faces are necessary to duly account for the
runway with a width that depends on the ARC complex intersections that occur between airspace
code number. It starts at a given distance before protection surfaces. Fortunately, surfaces do not
the threshold and extends after the runway end overlap, as the lower surface rules.
or the stopway end. Such given distances depend
on the ARC code number and on whether it is an
instrument runway or a non-instrument one.
The Runway Strip follows the runway centerline
profile and constitutes the starting edge of some of
the airspace protection surfaces. For this reason, a
Runway Strip must be represented as a 3D polygon
or surface rather than a simpler 2D geometry.
Runway End Safety Area (RESA) is an area on
the ground longitudinally adjacent to a Runway
Strip and has similar geometric characteristics to
a Runway Strip.
Taxiway Strips exclude movable objects at given
distances from the taxiway centerline. It is required
for taxiways other than aircraft stand taxilanes. At
curved sections, the path followed by aircraft criti-
cal points such as wingtips and stabilizers must be Figure 1. Geometry of obstacle limitation surfaces.
analyzed to assure proper clearances. Source: (ICAO, 2004, pp. 4–2).

170

TIS2017_Book.indb 170 2/8/2017 8:02:38 PM

 In curves, wheels and other parts of an aircraft are also movable objects of complex geometric
describe a mathematical complex pattern, even if representation.
the nose wheel follows an arc of a circle. This has
implications for designing pavement fillets and for
analyzing safety clearances during taxi, as well as 3 AERODROME ANALYSIS AND
on apron maneuvers. Such complex patterns may DESIGN SOFTWARE
be well represented by a sequence of points, even
though a larger number of points are necessary for A web search has shown no open source software
higher precision. related to aerodrome design, but some related com-
As an analogy, an aerodrome can be represented mercial software. Table 1 summarizes those types
by a non-continuous adjacent set S of surfaces of software and their main features.
zS = fS(x,y). Different “z” surfaces can be repre- No mention was found that existing computa-
sented by layers (e.g.: runways, protection areas on tional tools could be called by programming pro-
the ground, airspace protection areas). cedures, allowing the extension of their functions,
Therefore, every layer of the aerodrome geom- multiple scenario evaluation or optimization.
etry can be represented by points, lines, polygons
and polygonised surfaces, avoiding the use of solid
representations. 4 MODEL DEVELOPMENT

2.1.4 Obstacle representation This section presents key aspects of the model
When evaluating airspace infringement by a build- development.
ing or antenna, a single point representation seems
reasonable. However, objects closer to runways will 4.1 Conceptual background
deserve a more complex procedure. Geospatial and CAD (Computer Aided Design)
For movable objects, a movement simulation concepts, of fundamental importance to the model,
or an investigation of the critical position is nec- are briefly introduced in the following sub items.
essary. TRB (2010) presents a case of airspace
infringement, in which a crane operating over a 4.1.1 Geospatial considerations
1,000 rail at Oakland Port became an obstacle Computer handling of geospatial data can be
in only some of its possible positions. Aircraft performed through user implementations or with

Table 1. Aerodrome design and analysis software.

Obstacle Surface ArcGis for

Feature 3DAAP AeroTurn3D* PathPlanner* Planner Aviation: Airports

License Freeware, but Proprietary Proprietary Proprietary Proprietary.

requires Requires
Autodesk ArcGis for
Map 3D Desktop
Aircraft docking analysis
Jet Blast analysis
Airspace protection surfaces
infringement
Fillets analysis and design
Control Tower and runway
line of sight analysis
ALP (Airport Layout Plan)
support
Ground clearances and
protection areas analysis
Aerodrome optimization
Source (FAA & Transoft Transoft Simtra (2016) ESRI (2016)
Planning Solutions Solutions
Technology, (2016) (2016)
Inc, 2005).

Source: Several sources mentioned on the table.

*: Succeeded by AviPLAN.

171

TIS2017_Book.indb 171 2/8/2017 8:02:38 PM

 the aid of the so-called GIS software (Geographic i. Geometry generation: points, lines, polygons
Information System). For developers, GDAL and mathematical functions are usually adopted
and OGR free open source libraries deserve con- as primitives for 2D geometry construction. By
sideration, as they “are the backbone of many utilizing mathematical functions, fewer control
Desktop GIS software” (Leidig & Teeuw, 2015). points might be required, design traces can be
For the interested reader, Chen et al. (2010) com- smoothed and direct calculations of areas might
pares an extensive list of free open source GIS be easier. For 3D geometries, the same afore-
software. mentioned primitives can be used, in addition
Although civil engineering designs can often to cones, pyramids, spheres, particle systems
disregard the shape of the Earth, the greater the and others. Choosing primitives directly influ-
distance involved, greater the error introduced ences the easiness and the flexibility to represent
will be. An aerodrome extends over considerable different shapes. As detailed by Hosaka (1992),
distances, especially when airspace protection sur- once a geometry is generated, operations may
faces are considered. Moreover, terrain topography be executed, such as translations, scaling, rota-
and obstacles are usually surveyed and stored over tions, mirroring, bending, etc. In computational
a geodesic coordinate system. In some cases, even geometry, such operations are usually per-
the final designed geometry will be more conven- formed via matrix calculations.
iently represented on a geodesic system. ii. Model construction paradigm: Monedero
When geodesic data is transformed into a pro- (2000) contrasts variants programming, which
jection in the Cartesian coordinate system, two is a static model, and the interactive one. In
sources of distortions are expected: scale distor- variants programming, the model is completely
tion and height distortion (Burkholder, 1993). generated by an internal programming proce-
Both distortions can be treated with proper adjust- dure, and in case altering part of it is necessary,
ing factors or with a simplified approach, in which an entire new model must be generated. This
projection is properly selected and a maximum approach requires complete anticipation of the
height imposed to keep errors within an acceptable involved geometries and their relationships. On
range. the other hand, an interactive model, sometimes
From the above considerations, the local topo- called “parametric design” or “history-based
graphic plan (ABNT, 1998), was considered as the design”, allows the user to change the whole
projection system to be adopted. It is restricted to or parts of the model, employing an internal
a squared area of 100 km X 100 km and can keep data structure, such as a graph tree that keeps
errors as small as 1:20,000. Adaptations were made track of the sequence used to build the model
to ensure complete numeric consistency on direct (Monedero, 2000). Additionally, a feature based
and inverse transformations using the matrix approach (Bidarra & Bronsvoort, 2000) can be
transformations described by Dal’Forno, et al. utilized in the model to represent entities based
(2011) and Andrade (2003). on their semantic structure (runways and their
Another key component of this proposed mod- protection surfaces, for example). For further
elling is a DTM (Digital Terrain Model) to bridge discussion on CAD construction paradigms
the gap between terrain topography data and auto- refer to Shah (1998).
matic retrieval of height at a given x,y coordinate.
For this, a TIN (Triangulated Irregular Network)
4.2 Proposed model
was adopted. Delaunay triangulation is performed
over terrain input data, considering only 2D (x,y) Figure 2 shows the proposed model basic
information. From this triangulation, a bilinear structure.
interpolation model is set, considering the eleva- Input data refer to ICAO design standards, crit-
tion of each of the terrain input points. Li, et al. ical aircraft (ARC), site available areas, site topo-
(2005) can be consulted for a deeper discussion on graphic data, and surrounding fixed and movable
alternatives for DTMs. objects. With these data it is possible for the airport
designer to propose an airside geometric configu-
4.1.2 CAD considerations ration (number, position, length and orientation
A CAD (Computer Aided Design) is a computa- of runways, taxiways, airport fixed and movable
tional system for digital design and for draft gen- objects). The model generates protection areas and
eration in 2D and/or in 3D. A geometric design surfaces and performs the analysis of conformity
comprises a set of geometries and the underlying of the airside geometry and of existing objects to
relationships. Consequently, a geometric design the Annex 14 standards.
has at least two concerns, namely: i) geometry Implementation considers that the model may
generation; and ii) model construction paradigm. be extended in the future, for simulation or opti-
These topics are here further analyzed: mization purposes, for example. Bearing this in

172

TIS2017_Book.indb 172 2/8/2017 8:02:38 PM

 primitive can be transformed back into a simpler
one, reliably and efficiently, this hybrid approach
deserves consideration.
Regarding the model construction paradigm,
the aerodrome geometry is strongly constrained
by design standards so that different aerodrome
alternatives differ only due to very objective and
enumerable design criteria, such as orientation and
number of runways, length of runways, runway/
taxiway separations, etc. To facilitate revision of
a given aerodrome geometry, it is convenient to
represent the aerodrome components individually.
Therefore, a feature-based approach (Bidarra &
Bronsvoort, 2000) was utilized to represent these
Figure 2. Proposed model basic structure. components.
Source: Prepared by the authors. For the above reasons, aerodrome geometry is
generated by an internal programming procedure
(static model), but its components are created
mind, the following criteria were set for choosing based on a feature-based approach. The latter
the most suitable programming language: availa- feature shall facilitate a future development of the
bility of libraries to deal with geospatial and CAD model to account for simulation or optimization
manipulations; support to 2D and 3D visualiza- applications.
tion, as well as animations; possibility of develop- Obstacles are assessed against the infringement
ing an integrated GUI (Graphic User Interface); of a protection surface by a routine that: i) first
easiness of native data structures; existence of a compares both convex hulls; and ii) in case they
wide collaborative community; commitment to overlay, height tests are performed. Protection
openness; and independence from IDEs (com- surfaces are transformed into TINs (Triangulated
mand line execution). Irregular Networks) and interpolated height is
The above criteria led the Python language retrieved just as done with DTMs.
to be selected. This is an interpreted, high-level,
multi-paradigm, multi-platform programming 4.2.2 File formats
language, promoted by the Python Software With the constant progress of GIS software and
Foundation (PSF, 2016). Although computational data acquisition technologies, it is interesting that
performance is not deemed as a strong Python designs can be translated into standard interop-
point, tools such as Cython allow speeding up erable data formats. The comparison of different
execution, by means of a special compiling tech- layers, such as DTMs, protection surfaces, noise
nique (Cython, 2016). Visualization can be gener- contours, satellite and aerial imagery, soil occu-
ated by Python with the Matplotlib library and a pation patterns, and roads, for example, might
GUI can be developed via wxPython toolkit, for provide valuable insights and analysis power
example. to decision makers. Airports are an example of
infrastructure that keeps several relationships with
the surrounding environment. It was found that
4.2.1 Geometry generation and model
the KML (Keyhole Markup Language) stand-
construction paradigm
ard could be used to write files that can be easily
Consideration of Python language indicated that
read in GIS software such as the free open source
points, lines and polygons would be the most
QGIS and in a virtual globe such as Google Earth.
suitable primitives, facilitating CAD operations,
For this task, Simplekml Python library, licensed
visualization with the Matplotlib library and
by GNU standard, has been chosen. All other file
exportation to KML (see item 4.2.2). Moreo-
exchange requirements have been kept as simple
ver, from the discussion in item 2.1, it can be
as supportable in CSV (Comma-Separated Val-
assumed that geometries to be generated do not
ues) format.
require complex representations, as design stand-
ards explicitly require the representation of given
points and lines. 4.2.3 Pseudocode
Note that using complex primitives can simplify A pseudocode of the proposed model is presented
and speed up geometry construction coding. Their in Figure 3. Algorithm 1 generates the aerodrome
pitfalls, however, lie on storage, manipulation, visu- geometric model, which is an input to Algorithm 2
alization, or other tasks. Nevertheless, if a complex that allows obstacle assessment.

173

TIS2017_Book.indb 173 2/8/2017 8:02:38 PM

 5.1 Digital terrain modelling
As discussed in section 2.1, DTM (Digital Terrain
Modelling) is essential for aerodrome geometry
analysis and design. To assess the methodology
described in item 4.1.1, SRTM (Shuttle Radar
Topography Mission) elevation data were used to
create a TIN model, for an arbitrary delimitation of
the Airport of Viracopos (SBKP). SRTM data were
obtained from USGS (2015) in GeoTIFF raster for-
mat (1 arc-second global, published on 2014). Data
were pre-processed on QGIS software to generate
a CSV input file with 23,893 height points, corre-
sponding to 1-m contour lines. This DTM is shown
in Figure 4, at WGS84 horizontal datum. Visualiza-
tion was generated by the Matplotlib library.

5.2 Aerodrome geometry

Figure 5 depicts the Runway, the inner portion
of the Runway Strip and the OFZs of the SBKP
Airport. As SRTM data were used for DTM con-
struction, differences to Google Earth data are
apparent in the picture. It is also possible to visu-
alize the irregularities of the Inner Transitional
OFZ higher contour. As prescribed, 45 m height is
attained at irregular orthogonal distances from the
runway centerline, given the runway height profile.
This visualization is possible by importing the gen-
erated KML file into Google Earth software.

5.3 Obstacles
Aircraft are movable objects situated inside the
aerodrome and very close to protection surfaces.
Hence, after a geometric configuration is gener-
ated, aircraft must be also generated and positioned
on taxiways/runways in the different situations they
are allowed to operate. This is essential for obstacle
assessment and for the subsequent conformity vali-
dation of a geometric configuration. Aircraft height
is retrieved automatically from the DTM and cor-

Figure 3. SBKP aerodrome site TIN DTM. Graphic

output generated by the Matplotlib Python library.
Source: Prepared by the authors from USGS (2015)
data.

5 MODEL APPLICATION
Figure 4. SBKP aerodrome OFZ surfaces as viewed on
This section presents details of the model applica- Google Earth after importing a model generated KML file.
tions for both real world and fictitious instances. Source: Prepared by the authors.

174

TIS2017_Book.indb 174 2/8/2017 8:02:38 PM

 Figure 5. Obstacle assessment graphic output generated from the computational model for a fictitious instance.
Graphic generated through Matplotlib Python library.
Source: Prepared by the authors.

responds to the maximum of two heights: DTM at paradigm for DTMs were also implemented and
nose wheel; and DTM at the main gear. Figure 6 validated.
depicts a fictitious test instance output automati- Along with GIS, CAD features regarding geom-
cally generated by the model. In this picture, the etry generation and internal representation were
aircraft considered, a B747–8, waits to enter the considered. User interaction can be replaced by an
Runway at 90 m from the Runway centerline. How- internal programming procedure.
ever, as the aircraft is positioned at different angles, The model was implemented in Python lan-
wingtips and tail tips can protrude a protection guage and applied to the case of Campinas Inter-
surface. In Figure 6, while at 90° only aircraft nose national Airport in Brazil (SBKP), demonstrating
is in the imminence of infringing the Inner Transi- its functionality. Graphic results were generated by
tional OFZ, at 30° the vertical stabilizer infringes the Matplotlib Python library. Additionally, model
this OFZ surface. The model can perform obstacle interoperability with the KML standard is provided,
assessment in a 3D environment and present a list allowing results visualization on Google Earth.
of detected violations for the different combina- Designers can benefit from the automation pro-
tions of objects and protection surfaces. vided by this type of model, which favors analyses
of design alternatives regarding their compliance
with ICAO standards.
6 CONCLUSIONS

A clean-sheet computational model to conform an ACKNOWLEDGEMENTS

aerodrome to ICAO Annex 14 is presented, bridg-
ing a gap in the literature. The authors acknowledge CAPES (Coordenação
Modelling requirements of the aerodrome de Aperfeiçoamento de Pessoal de Nível Supe-
geometry were pinpointed. Simple geometric rior/Coordination for the Improvement of Higher
primitives were employed to represent aerodrome Education Personnel) for a Doctoral Fellow-
components: runways and taxiways, as well as on ship (Bolsa de Doutoramento), CNPq (Conselho
ground protection areas and airspace protection Nacional de Desenvolvimento Científico e Tec-
surfaces were represented by lines and polygons nológico/ Brazilian National Council of Scientific
(2D and 3D). and Technological Development) for a Research
Geospatial concerns were treated, as the dis- Productivity Fellowship (Process 307622/2015-0)
tortions related to geographic projections, as well and LPT/EPUSP (Laboratório de Planejamento e
as the construction of a DTM (Digital Terrain Operação de Transportes da Escola Politécnica da
Model), including the DTM height retrieval. A Universidade de São Paulo—Transportation Plan-
local topographic coordinate system and a TIN ning and Operation Laboratory of the School of

175

TIS2017_Book.indb 175 2/8/2017 8:02:41 PM

 Engineering of the University of São Paulo) for ICAO, 2006. International Civil Aviation Organization.
technical support. Procedures for Air Navigation Services: Aircraft
Operations (PANS-OPS). Doc 8168.
Kazda, A. & Caves, R. E., 2007. Airport Design and
REFERENCES Operation. Bingley: Emerald Group Publishing, 2nd
Ed.
ABNT, 1998. NBR 14166 - Rede de Referência Cadastral Leidig, M. & Teeuw, R., 2015. Free software: A review,
Municipal—Procedimento, Rio de Janeiro: ABNT. in the context of disaster management. International
Andrade, J. B. d., 2003. Fotogrametria. 2ed. Curitiba: Journal of Applied Earth Observation and Geoinfor-
SBEE. mation, Volume 42, pp. 49–56.
Bidarra, R. & Bronsvoort, W., 2000. Semantic fea- Li, Z., Zhu, Q. & Gold, C., 2005. Digital terrain mod-
ture modelling. Computer-Aided Design, 32(3), pp. eling. s.l.: CRC Press.
201–225. Monedero, J., 2000. Parametric design: a review and
Burkholder, E. F., 1993. Design of a Local Coordinate some experiences. Automation in Construction,9(4),
System for Surveying, Engineering, and LIS/GIS. pp. 369–377.
Journal of Surveying and Land Information Systems, Planning Technology Inc., 2004. Three-Dimensional
53(1), pp. 29–40. Airspace Analysis Programs (3DDAAP) process
Chen, D., Shams, S., Carmona-Moreno, C. & Leone, A., implementation: airport layout plan, navigational
2010. Assessment of open source GIS software for aid screening, web coordination, & photogrammetry
water resources management in developing countries. standards. [Online]
Journal of Hydro-environment Research, 4(3), pp. Available at: http://www.3daap.com/ [Accessed 18 Aug
253–264. 2015].
Cython, 2016. C-extensions for Python. [Online] Avail- PSF, 2016. Python Software Foundation. [Online]
able at: http://cython.org. [Accessed 28 Jul 2016]. Available at: https://www.python.org/psf/. [Acessed
Dal’Forno, G. L., Aguirre, A. J., Hillebrand, F. L. & 28 Jul 2016].
Gregório, F. d. V., 2011. Transformação de coorde- Shah, J. J., 1998. Designing with Parametric CAD: Clas-
nadas geodésicas em coordenadas no plano topográ- sification and comparison of construction techniques.
fico local pelos métodos da norma NBR 14166/1998 In: Geometric Modelling: theoretical and compu-
e o de rotações / translações. A Mira: Agrimensura e tational basis toward advanced CAD applications.
Cartografia. Tokyo: Springer Science+ Business Media, LLC, pp.
ESRI, 2016. ArcGIS for Desktop Extensions. ArcGIS 53–68.
for Aviation: Airports. [Online] Silva, E. J. & Gualda, N. D. F., 2013. São Paulo/Guarul-
Available at: http://www.esri.com/software/arcgis/exten- hos International Airport Airside Compatibility Anal-
sions/aviation/airports/features. [Accessed 26 Jul 2016]. ysis for A380–800 and B747–8 Aircraft Operation.
FAA, 2014. Federal Aviation Administration. AC Journal of the Brazilian Air Transportation Research
150/5300–13 A. Airport Design. Includes Change 1. Society, 9(1), pp. 9–22.
Washington: U.S. Department of Transportation. Simtra, 2016. A Transoft Solutions Company. Obstacle
FAA, T. C. A. F. T. I. L. & Planning Technology, Inc, Surface Planner. [Online]. Available at: http://www.
2005. Three-Dimensional Airspace Analysis Pro- simtra.com [Accessed 26 Jul 2016].
grams. User Manual: Program Installation and Menu Transoft Solutions, 2016. Transoft Solutions. [Online]
Commands: AutoCAD Map 3D 2005, s.l.: s.n. Available at: http://www.simtra.com/. [Accessed 26 jul
Horonjeff, R., McKelvey, F., Sproule, W. & Young, S., 2016].
2010. Planning and Design of Airports. s.l.: McGraw- TRB, 2010. Transportation Research Board. Airport
Hill Education, 5 Edition. Cooperative Research Program. Report 38. Under-
Hosaka, M., 1992. Modeling of curves and surfaces in standing airspace, objects, and their effects on air-
CAD/CAM. s.l.:Springer-Verlag. ports, Washington: s.n.
ICAO, 2004. International Civil Aviation Organiza- USGS, 2015. United States Geological Survey. Shuttle
tion. Annex 14 to the Convention on International Radar Topography Mission (SRTM) 1 Arc-Second
Civil Aviation. Volume I. Aerodrome Design and Global. [Online] Available at: https://lta.cr.usgs.gov/
Operations. SRTM1 Arc [Accessed 29 Jul 2016].

176

TIS2017_Book.indb 176 2/8/2017 8:02:42 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A speed model for curves of two-lane rural highways based on

continuous speed data

Raul Almeida & Luís Vasconcelos

CITTA and Department of Civil Engineering, Polytechnic Institute of Viseu, Viseu, Portugal

Ana Bastos Silva

CITTA and Department of Civil Engineering, University of Coimbra, Coimbra, Portugal

ABSTRACT: The concept of design consistency refers to the conformance of road geometry with driver
expectancy. An inconsistency in road design can be described as a geometric feature or a combination of
features with unusual or extreme characteristics that may lead to unsafe driving behavior. A handful of
methods and approaches have been trying to model design consistency. One of the most used criteria to
evaluate design consistency is based on operating speed measurement. The present Portuguese design
manual proposes a method to evaluate the geometric design consistency that is usually considered too
simplistic, assuming that the speed on the curve is uniform and the values of the acceleration and decel-
eration are based on constant values. To investigate the limitations of this approach, field measurements
were recorded using a vehicle equipped with a precision data logger that combines video footage with
kinematic variables. The analysis of the resulting database has shown that some assumptions of tradi-
tional speed models do not hold and has motivated the development and calibration of a speed model on
curves for two-lane rural highways. This model can differentiate the driving dynamics of a set of drivers
on two-lane rural highways under a wide range of geometric conditions and can be integrated in a global
methodology to assess design consistency on two-lane rural highways.

Keywords: Design consistency, Speed profile, Operating speed, Two-lane rural highways, Global posi-
tioning system

1 INTRODUCTION the 85th percentile of the distribution of observed

speeds in real environment.
The objective of transportation is usually described It is important to define road designs that do
as the safe and efficient movement of goods and not require significant changes in the geometric
people. In the specific case of road transportation, characteristics of consecutive elements, and inher-
this objective presupposes the existence of a road ently abrupt changes to driver behavior. Therefore,
design that does not defraud the natural expecta- the need to predict the behavior of drivers in these
tions of drivers. Thus, speed is normally used as situations has resulted in various methodologies to
one of the most relevant factors to assess the per- evaluate the design consistency of two-lane rural
formance of a road infrastructure. highways.
In recent decades, speed has been a major focus
of interest for researchers in many fields, resulting
in numerous operating speed models that can be 2 LITERATURE REVIEW
applied to roads from different countries.
Design speed is set according to the desired The necessity to check the speed at individual ele-
functionality for a particular road. It establishes ments of the road may, in many cases, prove to be
the starting point for the selection of most of the overly restrictive and inadequate to ensure a com-
geometrical features of the road, which enables it fortable and safe road design. There are many meth-
to correspond to the expectations of the drivers ods to ensure a good balance regarding contiguous
in terms of travel time, economy and comfort. It elements of the road and design consistency for
should also be consistent with the expected speeds two-lane rural roads. Those methods can be catego-
by drivers on the same road. rized into four major groups: a) methods that use
Operating speed is also a strong indicator of traffic behavior parameters (including operating
road performance, and is normally estimated by speed profiles); b) methods that use indices related

177

TIS2017_Book.indb 177 2/8/2017 8:02:44 PM

 to road geometry; c) methods that use driver work- such as the curve radius, the Degree of Curvature
load; d) methods that use checklists to fulfill design (DC), the Curvature Change Rate (CCR), and the
consistency criteria (Fitzpatrick et al. 2000a). deflection angle.
A consistent road design can be obtained when The model developed by Morrall & Talarico
free-flow speed variations are made in a gradual (1994) in their study to determine side friction on
way along the road. With this goal in mind, it is two-lane rural highways, consider the degree of
vital to have adequate models to estimate operat- curvature as the only feature that has a significant
ing speed depending on the geometric characteris- effect on operating speed.
tics of the road. With these models it is possible to Passetti & Fambro (1999) investigated the effect
define design consistency criteria with particular of the transition curves relative to operating speed
regard in establishing limits to speed differential on curves, considering the curvature (1/R) as the
between two consecutive geometrical elements, as only relevant feature to develop the their speed
recommended by the present Portuguese design model.
manual. The radius is considered by Misaghi & Hassan
The method used in the present Portuguese (2005) as the most statistically significant parame-
design manual to evaluate design consistency on ter for estimating the operating speed, according to
two-lane rural highways consists first to determine the study conducted on several horizontal curves.
operating speed in the various elements of the road Other parameters, like road width have
and then calculate the operating speed profile. This been studied by several authors using different
method assumes that the acceleration and decel- approaches, as is the case of Lamm et al. (1988)
eration are constant, and that the operating speed that proposed several speed models for different
is also constant along the curved sections of the categories of road width. Krammes et al. (1995)
road. used the approaching speed on tangent in their
Between the various parameters that may be speed model, seeking to translate the influence of
associated with operating speed, some are more the driver expectations through variables that char-
significant than others, such as the maximum acterize the upstream and downstream sections of
legal speed, number of lanes, road type, percent- the road, considered in the literature as being par-
age of heavy vehicles, weather, visibility, period of ticularly related to the field of vision of drivers.
the day, and factors related with driver behavior Operating speed, according to Gibreel et al.
(Ye et al. 2001). (1999), is the most widely used criterion to evalu-
Although there is already a significant number ate design consistency, and is usually defined as the
of studies relating to operating speed, it continues speed corresponding to the 85th percentile of the
to be justifiable the development of studies that speed distributions of a number of vehicles operat-
allow to identify the parameters that affect operat- ing in free-flow conditions.
ing speed, particularly in the development of con- Several models have been developed to estimate
tinuous operating speed profiles on two-lane rural operating speed on curves. However, the model
highways (Cafiso & Cerni 2012). format, the independent variables, and regression
Typically, the design manuals of different coun- coefficients are substantially different from one
tries establish the operating speed on road sections model to another. This may prove that there is
through a focused analysis on specific geomet- differences in driver behavior from one region to
ric elements of the road. In the United States, another, and highlights the lack of a model that is
AASHTO (2011) proposes a number of recom- universally accepted by the scientific community.
mendations regarding design speed and operating The procedure to collect the data in most of this
speed according to the type of road. This proce- cases was performed using equipment that would
dure is also found in the present Portuguese design necessarily have to be handled by an operator (e.g.
manual regarding design speed and operating LIDAR speed gun). These devices are likely to
speed. present several problems during field data gath-
According to Lobo et al. (2013), the problem of ering, such as human error and influence on the
speed modeling is an issue that concerns not only driver behavior. Another method is by the use of
road infrastructure management entities but also pavement sensors. Though, since it is a bulkier
researchers from academia, resulting in models equipment, takes a longer time to install, and it
developed for different geometric elements, type of remains visible to the driver, which normally tends
vehicles and environmental conditions. These stud- to influence driver behavior.
ies have led to various methodologies to evaluate Most of these methods allow the data to be
design consistency on two-lane rural roads. collected at the medium point of the curve. Thus,
Road geometry has been identified as the most the models that result from such methods assume
conditioning factor regarding operating speed, that operating speed remains constant inside the
and can be characterized by various parameters curve, such as the model developed by Lamm

178

TIS2017_Book.indb 178 2/8/2017 8:02:44 PM

 et al. (1999). This model uses the CCR variable model and the results will become more accurate in
as the only explanatory variable for estimating case of a similar sample of data.
the operating speed on curves, and assumes that Continuous speed profiles allow to study driver
the acceleration and deceleration occurs only behavior namely the variations of operating speed
at the tangent, and is equal to ±0.85 m/s2. Similar throughout the various elements of the road, and
values are considered by the present Portuguese the points where those variations occur. These
design manual (±0.80 m/s2). phenomena could not be studied before the use of
Another model was developed by Ottesen & continuous speed data, mainly concerning speed
Krammes (2000), which is essentially based on the variations within the curved sections.
same assumptions that the model by Lamm et al. According Pérez-Zuriaga et al. (2010) this tech-
(1999), where the data is collected with speed guns nique has some limitations, particularly regard-
in various spots on tangent and curves along two- ing the equipment used as it can influence driver
lane rural roads. behavior, though in a minimal way, since they are
The present study presents a method to collect aware of being observed and tested.
continuous speed data based on Global Position- However, this method has a great advantage
ing System (GPS) that allows to collect a large because it uses a portable GPS device that col-
amount of data without significant influence on lects a large amount of continuous data for further
the drivers, which facilitates the development of analysis and that allows to build continuous speed
speed models. profiles for individual drivers.

4.1 Equipment and road characteristics

3 OBJECTIVES
The equipment used for data gathering was a
The objective of this study is to develop an oper- data logger from Race Technology Ltd (DL1
ating speed model on curves that uses geometric MK3 model). The data logger has a built in 6 g
characteristics of the road as explanatory vari- accelerometer with three axes and a 20 Hz GPS (20
ables. It can be used as a surrogate method to esti- recordings per second), enabling to reference the
mate the operating speed on curves to the present data for periods of time in relation to the position
Portuguese design manual, and also be incorpo- on the road segment.
rated in a global methodology to evaluate design This equipment has an accuracy of about three
consistency. meters in terms of positioning and an accuracy
The analysis is based on continuous speed data in relation to speed of around 0.1 km/h. In addi-
contrary to other models that use spot-speed data. tion to the GPS functionality, it is also possible to
This procedure allowed the development of the connect external sensors. This device was comple-
speed model and also the opportunity to analyze mented with another device (VIDEO4 model) that
the speed variations from tangent to curved sec- allows video recording using a high definition cam-
tions, and within the curved sections. era placed inside the vehicle perfectly synchronized
The research allowed to assess a new methodol- with the data logger.
ogy to obtain the geometrical characteristics of the The data was collected in two two-lane rural
road based on the data collected and using specific road segments in Portugal. The gathering and
software related to road design. Therefore, this processing of data allowed to differentiate the driv-
paper also evaluates a new methodology for col- ing dynamics of a set of drivers on two-lane rural
lecting and processing continuous speed data on highways.
two-lane rural highways. The segments of road are characterized by a
A comparative analysis is also performed with low traffic volume with an Annual Average Daily
the proposed model that is developed using con- Traffic (AADT) of approximately 6500 vehicles.
tinuous speed data, and other speed models that Since this study was focused on horizontal align-
used different methodologies to collect speed ments, the sample data was limited to relatively
data. flat stretches (grades less than ±4%). A total of 47
curves were analyzed for each direction of traffic.
The radius of the circular curves varied between
4 DATA COLLECTION METHODOLOGY 40 m and 500 m.
The pavement was in good condition for all the
The analysis presented in this paper is not based on extension of the road. The road segments presented
spot-speed data gathering using equipment placed a total length of about 13.5 km long with pavement
in various locations of the road as observed in shoulders throughout most of its extension. Road
other studies, but based on continuous speed data width had an average of 7 m. The data was col-
collected over time and length. Consequently, the lected between August and September during work

179

TIS2017_Book.indb 179 2/8/2017 8:02:44 PM

 days under dry weather. None of the road segments
presented important intersections.
All the drivers had more than five years of
driving experience and were allowed a testing
period before each field session. All of the laps
were recorded in real time with the support of the
installed equipment and checked afterwards with
the video footage to guarantee that the drivers
were always in free-flow conditions.

4.2 Data analysis and modeling

The data obtained from the equipment allowed the
modeling of the road segments with the accuracy
needed to obtain the various geometric charac-
teristics of the horizontal alignments. The coor- Figure 2. Example of a partial curvature diagram and
dinates taken from the data logger were exported speed profile of a single driver.
with the use of the data review program provided
with the equipment. Using the software AutoCAD
Civil 3D it was possible to determine the beginning ments. These diagrams were combined with actual
point and end point of each element of the road speed profiles of drivers to analyze speed varia-
segments (tangents, curves, and transition curves). tions and driver behavior on curves (Fig. 2).
Consequently, all the geometric characteristics of
each road segment were obtained (the radius was
obtained from the trajectory but it can be consid- 5 OPERATING SPEED MODEL
ered to be very close to the real axis radius).
The software provided by the equipment offers Many of the methods used to evaluate design con-
an extensive analysis of the journeys made by the sistency on two-lane rural highways are based in
drivers on the road segments (Fig. 1), i.e. path view the analysis of operating speed. The data gathering
overlaid with satellite maps, speed and acceleration methodology used in this paper was chosen because
diagrams, synchronized video footage with the it is the most representative in recent studies and
position on the road and the corresponding speed one that allows to observe driver behavior concern-
diagram, etc. ing speed and geometric design of the road when
Since the equipment takes into account GPS, compared to other methods of data collection.
it is assumed that some positioning errors might This method is similar to that adopted by Pérez-
occur, but since the recording are taken at intervals Zuriaga et al. (2013) and other researchers that
of 0.1 seconds those errors did not influence the explored continuous speed data gathering, like
modeling of the road segments. the study of Cafiso & Cerni (2012), but with some
With the road segments characteristics retrieved differences, namely in obtaining and modeling the
from the software analysis, there was the need to geometric characteristics of the road, and data
construct curvature diagrams of all the road seg- reduction and analysis.
One of the main advantage of using this data
gathering procedure was the possibility to col-
lect a substantial amount of data which allowed
the development of the operating speed model
on curves. Although many observations were col-
lected and validated under free-flow conditions,
the sample is not representative of all the driv-
ers that drive in a particular road under the same
circumstances.
The operating speed model developed in this
paper can only be used as a tool to evaluate design
consistency when associated to operating speed
profiles and consistency thresholds.
Taking into consideration all the parameters
of the horizontal alignment on both segments of
Figure 1. Display of the software analysis software pro- the road, the most important variables are the
vided with the data gathering equipment. curve radius, curve length, and deflection angle.

180

TIS2017_Book.indb 180 2/8/2017 8:02:44 PM

 The curve radius varied from 40 m to 500 m, the
curve length from 33 m to 232 m, and the deflec-
tion angle from 9.85° to 120.75°.
With the purpose to develop several regression
models, the relations between these parameters
were studied. The radius of the curve was the vari-
able that revealed the highest correlation to operat-
ing speed, and the length of the curve revealed the
least correlation. Therefore, the radius of the curve
was used as the explanatory variable to develop
different models.
To develop the operating speed model the mini-
mum value of the observed speed data was used in
each of the 47 curves for each driver. A total of 284
observations were obtained and validated.
The assumption that the operating speed on
curve is constant was not observed. From the Figure 4. Proposed operating speed model on curves.
analysis of the continuous speed profiles, most of
deceleration and acceleration length occur within
the curve (Fig. 3). Only in very specific situations
a constant speed on curves was observed, for for curves with large radius this variable is not as
instance, curves that are very close to each other significant in the choice of speed as for curves with
and have a small tangent between them, or curves small radius. This analysis is consistent with the
with similar radii. These situations do not allow model developed by Pérez-Zuriaga et al. (2010).
the driver to increase or decrease the travel speed,
taking the decision to maintain a constant speed
between these elements (see rightmost curves in 6 COMPARISON OF MODELS
Fig. 3).
From a single regression analysis resulted the The proposed model was compared with others
exponential model shown in Figure 4 with the in the literature. All models that were considered
radius as explanatory variable and R2 = 0.511. for the analysis were applied to curves with radii
It can be obtained by the following equation: between 40 m and 500 m to comply with the pro-
posed model in this study.
V85 81 29 e ( −3.86 / R)
(1) The model developed by Morrall & Talarico
(1994) on two-lane rural highways in Canada
was calibrated with observations from 9 curves
where V85 = operating speed on curve (km/h);
with pavement shoulders. The curve radius varied
R = radius (m).
between 290 m and 3490 m. The model presents
The model shows a high slope for small radii and
the degree of curvature as the explanatory variable,
decreases as the radius gets bigger. It indicates that
and has an R2 = 0.631. This model assumes that the
speed on curve is constant, and it is represented by
the following equation:

V85 e ( 4.561− 0.00586 DC ) (2)

where V85 = operating speed (km/h); DC = degree

of curvature (°/100 m of arc length). The vari-
able DC present in this model can be defined as
one degree per 100 m of arc length (SI units),
hence it is possible to relate this variable to the
variable curve radius, i.e. DC = 5729.578/R,
where R is the radius (m).
Another model was developed on two-lane rural
roads in the United States by Passetti & Fambro
(1999) with observations from 12 transition curves
and 39 circular curves. The model uses the inverse
Figure 3. Example of a partial speed profile of a single of the radius as the explanatory variable, and has
driver. an R2 = 0.68. This model also assumes that the

181

TIS2017_Book.indb 181 2/8/2017 8:02:45 PM

 speed on curve is constant, and it is represented by Table 1. Comparison of models.
the following equation:
Model Equation
V85 103 9 3020 5 / R (3) Kanellaidis et al. (1990) V85 129 88 623 1 / R

where V85 = operating speed (km/h); R = radius Morrall & Talarico (1994) V85 e ( 4.561− 0.00586 DC )
(m).
Passetti & Fambro (1999) V85 103 9 3020 5 /R
R
In another study conducted by Fitzpatrick et al.
(2000b), the authors also developed a model to esti- Fitzpatrick et al. (2000b) V85 MC 104 82 3574 51/R
mate the V85 for vehicles traveling on two-lane rural 0% ≤ G ≤ 4%
highways. The speed data were collected in 176 Proposed model V85 81 29 e ( −3.86 / R)

locations in the United States. The model uses the

inverse of the radius as the explanatory variable.
Two equations were developed for grades less than
±4%, but only the one that best fits in this analysis
will be used in the comparison. Other grades were
considered by these researchers to develop more
equations. The model presents the inverse of the
radius as the explanatory variable, and has an R2
coefficient of 0.76. It can be represented by the fol-
lowing equation:

V85 MC 104 82 3574 51 / R; 0% G 4% (4)

where V85MC = operating speed on curve midpoint

(km/h); R = radius (m); G = grade (%). It was noted
by the authors that operating speed on curves is
very similar to the speed on tangents when the
curve radius is equal or higher than 800 m. On
curves with radii below 250 m operating speed
decreases very rapidly.
Kanellaidis et al. (1990) developed a model in
Figure 5. Comparison of operating speed models.
Greece that studied the relation between operat-
ing speed on curves and various geometric charac-
teristics of the road. The data was collected in 58 7 CONCLUSIONS
locations, and the model can be represented by the
following equation: The necessity to have methods to evaluate design
consistency holds to prevent unexpected changes
V85 129 88 623 1/ R (5) to the geometrical characteristics of consecutive
road elements and combinations of elements that
where V85 = operating speed (km/h); R = radius (m). do not comply with driver expectations. Some of
The inverse of the squared root is the independ- the methods that exist today to estimate design
ent variable in Equation 1, which is similar to the consistency on two-lane rural highways are based
model developed by Kanellaidis et al. (1990). From on operating speed profiles that do not comply
other models shown on Table 1 that also use the with actual driver behavior. Thus, it is necessary
radius of the curve as the explanatory variable, and to have models that are able to predict operating
performing a comparative analysis of those models speed concerning road characteristics and driver
with the proposed model, it can be concluded that behavior.
this model tends to underestimate operating speed The methodology adopted for data gathering
on curves (Fig. 5). That can be explained because allowed to model genuine operating speed profiles
the operating speed used to develop this model was and thereby analyze in greater detail the behavior
always the minimum observed in each curve from of drivers on several elements of the road. How-
the continuous speed data, when the other models ever, it was not conclusive if this type of equip-
used spot-speed collection methods, usually at the ment has less influence than other data gathering
medium point inside the curve, which not always methods.
correspond to the geometrical point where the The speed models that result from this type of
minimum speed occurs. analysis will be more accurate that those that use

182

TIS2017_Book.indb 182 2/8/2017 8:02:46 PM

 data gathering on specific points of the road, such Pendleton, O.J., Messer, C.J. (1995). Horizontal align-
as the midpoint of the curve. Thus, this method- ment design consistency for rural two-lane highways.
ology allows to determine the beginning point Report No. FHWA-RD-94-034. Texas Transporta-
and end point of both acceleration and decelera- tion Institute, U.S. Department of Transportation,
FHWA—Federal Highway Administration.
tion phenomena, as well as the minimum speed Lamm, R., Choueiri, E.M., Hayward, J.C., Paluri, A.
observed for each driver inside the curve. (1988). Possible design procedure to promote design
It was observed that the speed is not constant on consistency in highway geometric design on two-lane
the curve, as is assumed in various methods to eval- rural roads. Transportation Research Record, Vol.
uate design consistency on two-lane rural roads. 1195. Washington, D.C., TRB—Transportation
The use of the software AutoCAD Civil 3D Research Board, pp. 111–122.
for modeling the road segments, combined with Lamm, R., Psarianos, B., Mailaender, T. (1999). High-
the instrumented vehicle for field data gathering, way design and traffic safety engineering handbook.
resulted in an innovative methodology that allows McGraw-Hill.
Lobo, A., Rodrigues, C., Couto, A. (2013). Free-flow
in an expeditious way to obtain the geometric char- speed model based on Portuguese roadway design fea-
acteristics of the horizontal alignment, revealing a tures for two-lane highways. Transportation Research
procedure that is essential to the development of Record—Journal of the Transportation Research
operating speed models on two-lane rural roads. Board, Vol. 2348. Washington, D.C., TRB—Trans-
portation Research Board, pp. 12–18.
Misaghi, P., Hassan, Y. (2005). Modeling operating speed
REFERENCES and speed differential on two-lane rural roads. Journal
of Transportation Engineering, Vol. 131(6). ASCE—
AASHTO. (2011). A policy on geometric design of high- American Society of Civil Engineers, pp. 408–418.
ways and streets. 6th Edition. Washington, D.C., Morrall, J.F., Talarico, R.J. (1994). Side friction demanded
American Association of State Highway and Trans- and margins of safety on horizontal curves. Transporta-
portation Officials. tion Research Record, Vol. 1435. Washington, D.C.,
Cafiso, S., Cerni, G. (2012). New approach to defin- TRB—Transportation Research Board, pp. 145–152.
ing continuous speed profile models for two-lane rural Passetti, K., Fambro, D. (1999). Operating speeds on
roads. Transportation Research Record - Journal of curves with and without spiral transitions. Transporta-
the Transportation Research Board, Vol. 2309. Wash- tion Research Record—Journal of the Transportation
ington, D.C., TRB—Transportation Research Board, Research Board, Vol. 1658. Washington, D.C., TRB
pp. 157–167. - Transportation Research Board, pp. 9–16.
Fitzpatrick, K., Wooldridge, M.D., Tsimhoni, O., Collins, Pérez-Zuriaga, A.M., Camacho-Torregrosa, F.J., García-
J.M., Green, P., Bauer, K.M., Parma, K.D., Koppa, García, A.G. (2013). Tangent-to-curve transition on
R., Harwood, D.W., Anderson, I., Krammes, R.A., two-lane rural roads based on continuous speed profiles.
Poggioli, B. (2000a). Alternative design consistency rat- Journal of Transportation Engineering, Vol. 139(11).
ing methods for two-lane rural highways. Report No. ASCE—American Society of Civil Engineers, pp.
FHWARD-99–172. Texas Transportation Institute, 1048–1057.
U.S. Department of Transportation, FHWA-Federal Pérez-Zuriaga, A.M., García-García, A., Camacho-Tor-
Highway Administration. regrosa, F.J., D’Attoma, P. (2010). Modeling operating
Fitzpatrick, K., Elefteriadou, L., Harwood, D.W., Col- speed and deceleration on two-lane rural roads with Glo-
lins, J.M., McFadden, J., Anderson, I.B., Krammes, bal Positioning System data. Transportation Research
R.A., Irizarry, N., Parma, K.D., Bauer, K.M., Pas- Record—Journal of the Transportation Research
setti, K. (2000b). Speed prediction for two-lane rural Board, Vol. 2171. Washington, D.C., TRB—Trans-
highways. Report No. FHWA-RD-99-174. Texas portation Research Board, pp. 11–20.
Transportation Institute, U.S. Department of Trans- Ottesen, J., Krammes, R. (2000). Speed-profile model for
portation, FHWA-Federal Highway Administration. a design-consistency evaluation procedure in the United
Gibreel, G.M., Easa, S.M., Hassan, Y., El-Dimeery, I.A. States. Transportation Research Record—Journal of
(1999). State of the art of highway geometric design the Transportation Research Board, Vol. 1701. Wash-
consistency. Journal of Transportation Engineering, ington, D.C., TRB—Transportation Research Board,
Vol. 125(4). ASCE-American Society of Civil Engi- pp. 76–85.
neers, pp. 305–313. Ye, Q., Tarko, A., Sinha, K. (2001). Model of free-flow
Kanellaidis, G., Golias, J., Efstathiadis, S. (1990). Driver’s speed for Indiana arterial roads. Transportation
speed behavior on rural road curves. Traffic Engineer- Research Record—Journal of the Transportation
ing and Control, Vol. 31(7/8). London, England, Prin- Research Board, Vol. 1776. Washington, D.C., TRB—
terhall Limited, pp. 414–415. Transportation Research Board, pp. 189–193.
Krammes, R.A., Brackett, R.Q., Shafer, M.A., Ottesen,
J.L., Anderson, I.B., Fink, K.L., Collins, K.M.,

183

TIS2017_Book.indb 183 2/8/2017 8:02:47 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Rheological characterization of cold bituminous mastics produced

with different mineral additions

C. Godenzoni, M. Bocci & A. Graziani

Dipartimento di Ingegneria Civile Edile e Architettura, Università Politecnica delle Marche, Ancona, Italy

ABSTRACT: Nowadays, cold bituminous mixtures produced with bitumen emulsion (CBEMs) repre-
sent one of the most attractive alternatives with respect to traditional hot mix asphalt. In particular, the
use of CBEMs is increasing because they combine economical and environmental efficiency. CBEMs can
be schematized as composite of coarse aggregate particles (virgin and/or reclaimed) covered and bound
by a bituminous mastic (fresh bitumen and filler-sized particles), also incorporating part of the fine aggre-
gate particles (sand). Therefore, the mechanical behavior of the whole CBEMs is strictly connected to the
behavior of its associated Cold Bituminous mastic (CBm). The main objective of this study is to evaluate
the effect of different mineral additions and volumetric concentrations on the LVE properties of CBms.
Complex shear modulus (G*) was measured on CBms prepared with Calcium Carbonate (CC) or cement
(CEM) at 0.15 and 0.3 as mineral addition ratios. In addition, the influence of curing time (1 and 3 days)
on LVE properties, was also investigated. The testing protocol included Brookfield viscosity measure-
ments during mixing phase and Complex shear modulus testing, performed at temperatures ranging from
5°C to 60°C. The results showed that CBm composition strongly affects the viscosity and the mechanical
response of CBms. In particular, at higher temperatures and higher concentration ratios, phase separation
was observed. The linear viscoelastic behavior showed an evolutive trend at longer curing times, especially
for CBms prepared with cement.

1 INTRODUCTION to achieve suitable volumetric properties [Grilli

et al., 2016]. Second, CBEMs can be defined as
The use of Cold Bitumen Emulsion Mixtures evolutive materials because their physical structure
(CBEMs) for pavement construction, mainte- evolve over time. This process, known as curing,
nance and rehabilitation constantly increased over leads to the improvement of the mechanical prop-
the last 20 years [Liebenberg et al., 2004; Serfass erties (e.g. stiffness and strength) until reaching a
et al., 2004; Santagata et al., 2009, Nassar et al., log-term cured state [Jenkins and Moloto, 2008;
2016a]. Economic and environmental benefits, Godenzoni et al., 2016a; Graziani et al., 2016].
with respect to both hot and warm mixtures, derive Because of the simultaneous presence of both
from the reduction of energy necessary for heat- bituminous and hydraulic binders, the curing proc-
ing both bitumen and aggregates. Moreover, when ess of CBEMs is actually due to the interaction
emulsion-based materials are used for pavement of different physical and chemical mechanisms
recycling, consumption of virgin aggregate can be [Brown and Needham, 2000]. Curing mechanisms
significantly reduced [Stroup-Gardiner, 2011]. The related to the bituminous phase basically involve
composition of CBEMs can be extremely variable emulsion breaking and water expulsion.
and affecting their mechanical response. In partic- Huang and Di Benedetto (2015) proposed a
ular, CBEM properties vary depending on the pro- useful schematization of HMA as a multi-phase
portions in which binders, bitumen and cement, material:
are mixed.
– bituminous mastic, composed by bituminous
When comparing the composition and the behav-
binder and filler-sized particles that establishes
ior of CBEMs to that of conventional Hot-Mix
the connection between bituminous binder and
Asphalt (HMA) mixtures, two aspects are of the
mixture;
outmost importance. First, water instead of heat-
– fine aggregate mixture or mortar, composed by
ing is used in order to allow CBEMs laydown and
bituminous mastic and fine aggregate particles,
compaction. Indeed, moisture content is the main
that fills voids between the coarsest aggregate
factor controlling the rheological properties of
particles;
CBEMs in the fresh state, in particular their ability
– coarse aggregates skeleton.

185

TIS2017_Book.indb 185 2/8/2017 8:02:49 PM

 The same representation can be adopted for with
CBEMs. In this case, cold bituminous mastic
(CBm) is composed by bitumen emulsion (bitu- τ *( ) τ 0 ⎡⎣cos (ω t ) + i i (ω t ) ⎤⎦ (2)
men droplets dispersed in water) and filler-sized
particles. Over time, water evaporates and emul- γ *( ) γ 0 ⎡⎣cos (ω δ ) + i sin
si (ω δ ) ⎤⎦ (3)
sion breaking occurs; bitumen droplets coalesce
forming a continuous thin film of bitumen that where τ0 and γ0 are the amplitude of stress and
englobes filler-sized particles. In addition to filler- strain respectively, ω is the angular frequency, t is
sized particles commonly used with HMA (e.g. the time and δ is the phase angle.
recovered dust), other mineral additions such as By substituting these expressions (periodic sinu-
cement or hydrate lime can be included in CBEMs soidal loads) in the previous Equation (1), the fol-
[Nassar et al., 2016b]. lowing form can be obtained:
CBEMs can also be analyzed at different time-
scales because water content evolves over time as τ 0 iδ
well as the physical state of the mixture, i.e. from G * (iω ) = e G1 + iG
G2 (4)
γ0
the fresh state to the cured state.
This feature is noticeable at mixture level and
at each smaller level of investigation (mortar or As reported in Equation 4, G* can be decom-
mastic-level). posed in two components
The performance of CBEMs mainly depends – G1 = G*⋅cos (δ), is the storage modulus, used as
on the rheological properties of its CBm. In this indicator of the elastic behavior of the material
context, it is important to investigate the linear vis- (elastic behavior);
coelastic (LVE) properties in order to understand – G2 = G*⋅sen (δ), is the loss modulus, used as
how the proportion of each constituent can affect indicator of the dissipated energy when a load
the whole LVE behavior of the associated mixture. is applied to the material and, hence, represents
Different proportions, generating various micro- the viscous component of the material behavior
structures, can produce a wide range of bitumi- (viscous behavior).
nous material behaviors [Elnasri et al. 2013].
In addition, considering LVE properties of The phase angle represents the relative amounts
CBm, mineral addition used as emulsion breaking of recoverable and non-recoverable deformation,
regulator, material stiffener and filler, plays a fun- therefore, it concurs to identify the material behav-
damental role. ior: when δ tends to 0° the bitumen behaves as an
The main objective of this study is to evalu- elastic material, whereas δ values close to 90° a vis-
ate the effect of different mineral additions and cous behavior is detected.
volumetric concentrations on the LVE properties
of CBms. The complex shear modulus (G*) was 3 EXPERIMENTAL PROGRAM
measured on CBms prepared with limestone filler,
i.e. Calcium Carbonate (CC) or cement (CEM). 3.1 Materials
In addition, the influence of curing time (1 and
3 days) was also investigated. A cationic overstabilized bitumen emulsion desig-
nated as C 60 B 10 (EN 13808) was adopted to pre-
pare the cold bituminous mastics. This emulsion
2 BACKGROUND ON RHEOLOGICAL was selected because it is commonly used during
PROPERTIES OF BITUMINOUS cold recycling operations. Its breaking behavior is
MATERIALS evaluated by measuring the mixing stability with
cement (UNI EN 12848).
The mechanical behavior of CBms can be analyzed Bitumen emulsion properties were reported in
using the same approach commonly adopted for Table 1.
bituminous binders. Two mineral additions were selected:
Bituminous materials exhibit a thermo-viscoe- – Portland limestone cement II A/LL, strength
lastic behavior, i.e. dependent to applied tempera- class 42.5 R (UNI EN 197-1);
ture and loading frequency [Pellinen et al., 2007]. – limestone filler, i.e. finely ground calcium car-
In this context, G* is often used to identify their bonate (CaCO3).
LVE response and it is defined as:
Portland cement and calcium carbonate were
τ * (t ) characterized determining the Rigden voids and
G* = (1) the fineness (specific surface) using the Blaine
γ * (t ) method (Table 2).

186

TIS2017_Book.indb 186 2/8/2017 8:02:49 PM

 Table 1. Characterization of mineral additions. Table 3. Summary of CBms produced.

Characteristics of the bitumen emulsion Residual Curing

Mineral bitumen Concentration time Sample
Property Value addition [g] ratio [day] code
Water content (UNI EN 1428) 40%
pH value 3 None 100 0 1 BE_1d
Settling tendency @ 7 days 8% 3 BE_3d
Breaking value (UNI EN 13075-1) 180% Cement 0.15 1 CEM0.15_1d
Mixing stability with cement (UNI EN 12848) ≤0.2 g 3 CEM0.15_3d
Application temperature 5–80°C 0.3 1 CEM0.3_1d
3 CEM0.3_3d
Calcium 0.15 1 CC0.15_1d
carbonate 3 CC0.15_3d
Table 2. Characterization of mineral additions.
0.3 1 CC0.3_1d
Parameter Cement Calcium carbonate 3 CC0.3_3d

Rigden voids [%] 40.18 31.14

UNI EN 1097-4
Specific surface [m2/g] 0.41 0.51
UNI EN 196-6

Two different volumetric concertation ratios

(volume of mineral addition over volume of bitu-
men) of mineral addition were considered for
CBms preparation: 0.15 and 0.3. These values
were calculated starting from the mass concentra-
tion ratio and considering the particles densities of
1.02, 3.15 and 2.69 g/cm3 for bitumen, cement and
calcium carbonate, respectively. The adopted volu-
metric concentration ratios were intended to simu-
late the common composition of CBms, where
mass ratio between bitumen and cement is usually
around 1 [Godenzoni et al., 2016b].
CBms were coded with an alphanumeric code Figure 1. Mastic preparation: a) mixing by mechani-
composed by letters referring to the types of min- cal device; b) and c) pouring of liquid mastic in silicone
eral addition adopted (CEM or CC, for cement or molds (8 mm and 25 mm diameter).
calcium carbonate, respectively) and a number rep-
resenting the volumetric concentration ratio (0.15 and the lower rotation speed rate equal to 489.3
or 0.3). rpm was selected in order to avoid the premature
For example, a CBm produced using cement at emulsion breaking (Fig. 1a).
0.15 of volumetric concentration ratio, was coded CBms were mixed continuously for one hour at
as CEM0.15. In addition, bitumen emulsion was room temperature and sampled after 15, 45 and
also tested and coded as BE (without mineral 60 minutes (end of mixing phase) of mixing in
additions). order to check the mastic homogeneity by measur-
ing CBm viscosity and reject mastic when sedimen-
tation phenomena occur [Santagata et al., 2016].
3.2 Mastics production
The homogeneity control was carried out by
A standardized procedure for the production of measuring material viscosity using a Brookfield
CBms is not available. In this experimental study rotational viscometer.
25 g of pre-wetting water was initially added to At the end of this phase, mastics were poured
100 g of calcium carbonate or cement and stirred in silicone molds with different dimensions (8 and
by hand for one minute, in order to facilitate mix- 25 mm) (Fig. 1b and c). Then, CBm specimens
ing and avert breaking of the emulsion (Table 3). were cured for one and three days at 40°C, in a
Afterwards, bitumen emulsion was added and temperature controlled chamber.
stirred by hand until obtaining a homogeneous In particular, the production of small size test
compound without any lumps. Then a high-shear specimens has avoided separation between the
mixer was adopted for the second mixing phase materials characterized by different densities,

187

TIS2017_Book.indb 187 2/8/2017 8:02:50 PM

 obtaining homogenous samples for rheological
laboratory testing.
At the end, CBms codes were updated adding
information about the curing time (1d and 3d, for
one and three days of curing, respectively). For
example, a CBm produced using cement at 0.15 of
volumetric concentration ratio and tested after one
day, was coded as CEM0.15–1d. The list of CBms
produced is given in Table 3.

Figure 2. Dynamic shear rheometer with different

3.3 Testing program and protocols measuring system: a) 8 mm and b) 25 mm parallel-plate-
3.3.1 Brookfield rotational viscometer geometry.
In order to quantitatively assess the occurrence of
segregation phenomena, CBm samples were char-
acterized in terms of dynamic viscosity (η). Meas-
urements were performed by means of a Brookfield
rotational viscometer operated at ambient tem-
perature (about 25°C). About 10 ml of CBm were
poured into the cylinder. The spindle (S21) was
lowered into the samples and subsequently, the test
was run at different shear rates γ (ranging from 50
up to 200 s–1).

3.3.2 Strain-sweep and frequency-sweep tests

Rheological measurements (G*) were performed
using a stress/strain controlled Dynamic Shear
Rheometer (DSR) device equipped with parallel
plate-plate geometry. Strain-Sweep (SS) tests were
preliminary performed with the aim to investi-
gate the LVE limit and define a suitable range of
strain level for both BE and CBms [Dondi et al.,
2014; Frigio et al., 2016]. SS tests were performed
at 0°C by using 8 mm parallel-plate-geometry
and 2 mm gap, applying a constant frequency of
10 rad/s (1.59 Hz). A unique strain level of 0.01%
was adopted as LVE limit for all materials, in order
to simplify the testing procedure. This value was
selected based on the LVE limit detected with
CEM mastic, even if BE and CC mastic were char-
acterized by higher LVE limits.
Frequency-Sweep (FS) test was conducted in a
range of frequencies between 0.01 and 10 Hz, at
the temperatures of 0, 10, 20, 30, 40, 50 and 60°C.
The 8 mm plate-plate measuring system with 2 mm
gap was adopted from 0 up to 40°C and 25 mm
plate-plate measuring system with 1 mm gap was
used from 30 to 60°C (Fig. 2). Figure 3. Dynamic viscosity measurements as a func-
tion of shear rates, at different sampling times: a) mastic
CC0.3 and b) mastic CEM0.3.
4 RESULTS AND ANALYSIS
and to evaluate segregation phenomena [Santagata
4.1 Dynamic viscosity (η)
et al., 2016; Wang et al., 2016].
Viscosity provides an evaluation of the flow char- In Fig. 3, an example of results obtained from
acteristics of materials. In this study, measurements CC0.30 and CEM0.30 mastics was reported.
in terms of dynamic viscosity were carried out in As commonly observed for suspension, a typi-
order to verify the homogeneity of CBm samples cal shear thinning behavior was detected for both

188

TIS2017_Book.indb 188 2/8/2017 8:02:50 PM

 CC0.3 and CEM0.3 mastics: viscosity decreases The rheological parameters |G*| and δ of BE,
with the increasing in shear rate showing an asymp- CC0.15, CC0.30, CEM0.15 and CEM 0.30, cured
totic trend. As expected, viscosity values increase 1 and 3 days are reported, in terms of Black dia-
with mixing time for CC0.3 mastic; differently, a gram, in Fig. 4. Results represent the averaged val-
clear behavior over time cannot be distinguished ues calculated on two replicates specimens.
for CEM mastic. Considering the CC mastic (Fig. 4a and Fig. 4c),
As it can be observed at higher shear rate it can be noted that at both curing times (i.e. 1 and
(γ = 200 s−1), viscosity reached a constant value 3 days) and volumetric concentration ratio (i.e. 0.15
of about 60 Pa⋅s and 100 Pa⋅s for CC0.30 and and 0.30) did not affect the mechanical response of
CEM0.30 mastics, respectively. this material. In fact, the reported Black diagrams
A similar behavior was also detected for CC0.15 is practically superposed. At high temperatures,
and CEM0.15 mastics but viscosity measurements |G*| and δ were 85 Pa and 85°, respectively while at
were characterized by lower values due to the low temperatures, |G*| and δ were 131⋅106 Pa and
lower volumetric concentration ratio of mineral 35°, respectively. In particular, results carried out
addition. on CC0.15 and CC0.30 were approximately the
According to previous results, it can be con- same of those obtained on BE.
cluded that CBm samples can be considered homo- The rheological behavior detected for CEM
geneously mixed, i.e. no segregation phenomena mastic was totally different: the influence of both
occur during the mixing phase and consequently, concentration ratio (i.e. 0.15 and 0.3) and curing
can be adopted for mechanical testing. time (i.e. 1 and 3 days) was marked (Fig. 4b and
Fig. 4d).
4.2 Complex shear modulus (G*) Considering CEM0.15-1d mastic: at high tem-
peratures, |G*| and δ were 5000 Pa and 51°, respec-
In order to analyze G*, graphs as Black diagram, tively while at low temperatures, |G*| and δ were
where the norm of complex shear modulus (|G*|) 93⋅106 Pa and 38°, respectively. Differently with
is reported as a function of phase angle (δ), were CEM0.15-3d mastic, at high temperatures, the |G*|
adopted. and δ were 64 Pa and 51°, respectively while at low

Figure 4. Black diagram, effect of concentration ratio: a) BE, CC0.15 and CC0.3 mastics at 1 day of curing; b)
BE, CEM0.15 and CEM0.3 mastics at 1 day of curing; c) BE, CC0.15 and CC0.3 mastics at 3 days of curing; d) BE,
CEM0.15 and CEM0.3 mastics at 3 days of curing.

189

TIS2017_Book.indb 189 2/8/2017 8:02:51 PM

 temperatures, |G*| and δ were 21⋅106 Pa and 38°,
respectively.
Considering CEM0.30-1d mastic: at high tem-
peratures, |G*| and δ are 1.9⋅106 Pa and 27.4°,
respectively while at low temperatures, |G*| and δ
were 208⋅106 Pa and 27.7°, respectively.
Differently with CEM0.30–3d mastic, at high
temperatures, |G*| and δ were 2⋅106 Pa and 18°,
respectively while at low temperatures, |G*| and δ
were 158⋅106 Pa and 26°, respectively.
According to previous results, it is possible to
assess that an increasing in concentration ratio led
to an increase of |G*| and, consequently, a reduc-
tion in δ. Fig. 4b highlighted that the mechanical
behavior of CEM mastic was strictly related to the
amount of cement added in the material.
In fact, increasing cement content, stiffness
properties move from those achieved with BE (vis-
coelastic fluid behavior) to those obtained with
CBEM (viscoelastic solid behavior).
This observation is probably due to an increase
in the development of cementitious products (at
higher cement content and curing time) that stiff-
ened the structure of CEM mastics.
Although, the effect of curing time on
CEM0.15 mastic was evident, it was less marked
with respect to that obtained analyzing CEM0.3.
As it can be noted from Fig. 4d, CEM0.15 mas- Figure 5. Black diagram, effect of type of mineral
tic cured 3 days was not characterized by a unique addition: a) CC0.15 and CEM0.15 at 1 day of curing; b)
trend in the Black diagram. This peculiarity can CC0.3 and CEM0.3 at 1 day of curing.
be related to phase separation phenomena that
probably occur in the mastic, at high temperatures Whereas, the adoption of cement at high tempera-
(from 40°C to 60°C). tures, promoted a strong reduction in phase angle,
In Fig. 5, the comparison between CC and i.e. rheological proprieties more related to a viscoe-
CEM mastics at both volumetric concentration lastic solid material (solid-like behavior, δ→0°).
ratios (0.15 or 0.3), was reported considering one
day of curing.
As it can be observed from Fig. 5a, at volumetric 5 CONCLUSIONS
concentration ratio equal to 0.15, the rheological The present research focuses on the rheological
response at lower temperatures of both CC0.15 and characterization of two cold bituminous mastics
CEM0.15 mastics were practically superposed. At produced in laboratory with bitumen emulsion
temperature of 20°C, the Black diagram was char- and two mineral additions: cement and calcium
acterized by two different trends for the considered carbonate. The influence of both volumetric con-
CBms: G0 values were almost the same for CC0.15 centration ratio (0, 0.15 and 0.30) and curing time
and CEM0.15 mastics but, the adoption of cement (1 and 3 days) on linear viscoelastic properties (i.e.
as mineral addition caused a strong reduction in δ complex shear modulus G*) was evaluated. Con-
when temperature increased (δ changes from 85° sidering the overall results, the following conclu-
with CC0.15 to 50° with CEM0.15, at temperature sions can be drawn:
of 60°C). This observation highlighted the cemen- – The adoption of Brookfield rotational viscom-
titious microstructure of CEM0.15 mastic pro- eter for evaluating the homogeneity of CBm
moted a higher viscoelastic response with respect samples during production, was satisfactory.
to that detected with CC0.15 mastic.
Moving towards higher concentration ratio – The influence of the selected curing times was
(Fig. 5b), CC0.30 and CEM0.30 mastics exhibited negligible on CC mastics and slightly affected the
a totally different mechanical behavior: at high LVE properties of CEM mastics at all volumetric
temperatures, the addition of calcium carbonate concentration ratios. Shorter curing times (e.g.
promoted rheological properties more close to a vis- few hours) should be considered to better analyze
coelstic liquid material (liquid-like behavior, δ→90°). the effects of curing time at mastic-scale analysis.

190

TIS2017_Book.indb 190 2/8/2017 8:02:52 PM

 – The volumetric concentration ratio strongly Grilli, A., Graziani, A., Bocci, E., & Bocci, M. Volumet-
affects the LVE properties of CEM mastics at ric properties and influence of water content on the
both curing times. Increasing of cement volu- compactability of cold recycled mixtures. Materials
and Structures, 1–14.
metric concentration led to an increase of |G*|
Huang, S. C., & Di Benedetto, H. (Eds.). (2015).
and a reduction in δ (viscoelastic behavior) Advances in Asphalt Materials: Road and Pavement
probably related to the stiffening contribution Construction, Chapter 13 “Paving with asphalt emul-
of the developed cementitious structure. sion”. Woodhead Publishing.
– The effect of the type of mineral addition Jenkins, K. J., & Moloto, P. K. (2008). Updating bitumi-
was comparable with that achieved increasing nous stabilized materials guidelines: mix design report.
cement concentration ratio. Moving from cal- Phase II–Curing protocol: improvement. In Technical
cium carbonate to cement as mineral addition, Memorandum Task 7.
the LVE behavior evolved from that of liquid Liebenberg, J. J. E., & Visser, A. T. (2004). Towards a
mechanistic structural design procedure for emulsion-
material (viscous behavior) to that of solid mate-
treated base lavers. Journal of the south african insti-
rial (viscoelastic behavior). tution of civil engineering, 46(3), 2–9.
Nassar, A. I., Mohammed, M. K., Thom, N., & Parry,
T. (2016b). Mechanical, durability and microstructure
properties of Cold Asphalt Emulsion Mixtures with
REFERENCES
different types of filler. Construction and Building
Materials, 114, 352–363.
Brown, S., & Needham, D. (2000). A study of cement Nassar, Ahmed I. and Thom, Nicholas and Parry, Tony
modified bitumen emulsion mixtures. Asphalt Paving (2016) Examining the effects of contributory factors
Technology, 69, 92–121. on curing process of cold bitumen emulsion mixtures.
Dondi, G., Mazzotta, F., Sangiorgi, C., Pettinari, M., In: Fourth International Chinese European Workshop
Simone, A., Vignali, V., & Tataranni, P. (2014). Influ- on Functional Pavement Design, 29 June 6–1st July
ence of cement and limestone filler on the rheological 2016, Delft, Netherlands. (In Press)
properties of mastic in cold bituminous recycled mix- Pellinen, T., Zofka, A., Marasteanu, M., & Funk, N.
tures. Sustainability, Eco-efficiency, and Conservation in (2007). Asphalt Mixture Stiffness Predictive Mod-
Transportation Infrastructure Asset Management, 61. els (With Discussion). Journal of the Association of
Elnasri, M., Airey., G. & Thom, N. 2013. Experimental Asphalt Paving Technologists, 76.
Investigation of Bitumen and Mastics under Shear Santagata, E., Baglieri, O., Tsantilis, L., & Chiappinelli,
Creep and Creep-Recovery Testing. The Airfield and G. (2016). Storage stability of bituminous binders
Highway Pavement Conference. Los Angeles. Califor- reinforced with nano-additives. In 8th RILEM Inter-
nia 9th-12th June. national Symposium on Testing and Characterization
Frigio, F., Ferrotti, G., & Cardone, F. (2016). Fatigue of Sustainable and Innovative Bituminous Materials
rheological characterization of polymer-modified (pp. 75–87). Springer Netherlands.
bitumens and mastics. In 8th RILEM International Santagata, F. A., Bocci, M., Grilli, A., & Cardone, F.
Symposium on Testing and Characterization of Sus- (2009). Rehabilitation of an Italian highway by cold
tainable and Innovative Bituminous Materials (pp. in-place recycling techniques Vol. 2 (pp. 1113–1122).
655–666). Springer Netherlands. Proceedings of the 7th international RILEM sym-
Godenzoni, C., Cardone, F., Graziani, A., & Bocci, M. posium on advanced testing and characterization of
(2016a). The effect of curing on the mechanical behav- bituminous materials, Rhodes, Greece.
ior of cement-bitumen treated materials. In 8th RILEM Serfass, J. P., Poirier, J. E., Henrat, J. P., & Carbonneau, X.
International Symposium on Testing and Charac- (2004). Influence of curing on cold mix mechanical per-
terization of Sustainable and Innovative Bituminous formance. Materials and structures, 37(5), 365–368.
Materials (pp. 879–890). Springer Netherlands. Stroup-Gardiner, M. (2011). Recycling and reclamation
Godenzoni, C., Graziani, A., & Corinaldesi, V. (2016b). of Asphalt pavements using in-place methods (No.
The Influence Mineral Additions on the Failure Prop- Project 20–05 (Topic 40–13)).
erties of Bitumen Emulsion Mortars. In 8th RILEM Wang, H., Yang, J., & Gong, M. (2016). Rheological
International Conference on Mechanisms of Cracking Characterization of Asphalt Binders and Mixtures
and Debonding in Pavements (pp. 327–333). Springer Modified with Carbon Nanotubes. In 8th RILEM
Netherlands. International Symposium on Testing and Charac-
Graziani, A., Godenzoni, C., Cardone, F., & Bocci, M. terization of Sustainable and Innovative Bituminous
(2016). Effect of curing on the physical and mechani- Materials (pp. 141–150). Springer Netherlands.
cal properties of cold-recycled bituminous mixtures.
Materials & Design, 95, 358–369.

191

TIS2017_Book.indb 191 2/8/2017 8:02:52 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A comprehensive methodology for the analysis of highway

sight distance

M. Castro & C. De Santos-Berbel

Department of Civil Engineering, Transport and Territory, Technical University of Madrid, Spain

L. Iglesias
Department of Geological and Mining Engineering, Technical University of Madrid, Spain

ABSTRACT: As one of the main elements of geometric design, sight distance must be considered
carefully for the safe and efficient operation of highways. An application developed on Geographic
Information Systems (GIS) was conceived for the three-dimensional estimation of sight distance on high-
ways, as opposed to conventional two-dimensional techniques, which may underestimate or overestimate
the actual visibility conditions. It is capable of computing the available sight distance of a highway section
given the driver’s eye height, the target height, the vehicle path and an elevation model. The outcome can
be studied in detail with the aid of the tools and capabilities developed, including sight-distance graphs.
The influence of the input features, such as the nature of the elevation model, its resolution and the spac-
ing between path stations on the results accuracy was analyzed. The interpretation of results is also essen-
tial to explain sight distance deficiencies and provide insight into the effect of roadside elements on those
results. In addition, the sight-distance graph permits the detection and characterization of sight-hidden
dips, an undesirable shortcoming in the spatial alignment of highways. The versatility of GIS enables,
moreover, an integrated research of highway safety. It allows the incorporation of diverse operational
factors such as accident data, traffic volume, operating speed and design consistency to detect and diag-
nose potentially hazardous spots or, eventually, identify the factors involved in a particular accident. This
paper describes the methodology utilized and reviews the main issues through case study examples.

1 INTRODUCTION precise algorithm to compute the available sight

distance. Besides algorithms based on line-of-sight
As one of the main elements of geometric design, loops, procedures based on viewsheds were devel-
sight distance must be considered carefully for oped to study sight distance on highways (Castro
the safe and efficient operation of highways. In et al. 2011, Jha et al. 2011).
response to this, highway geometric design stand- Computer-aided applications for road design
ards in different countries set minimum sight dis- estimate and compare available sight distances to
tance threshold values (Ministerio de Fomento stopping sight distance and passing sight distance.
2016, AASHTO 2011, FGSV 2012). In order They also include visualization tools that simu-
to facilitate the geometric design of roads, some late the driver’s perspective while travelling (Kühn
guidelines propose two-dimensional analytical et al. 2011, Castro 2012). Such visualization tools
procedures to estimate the available sight distance. are utilized to supervise proper 3-D alignment
Nevertheless, these procedures may not be practi- coordination, although it requires this checking
cal since they consider separately horizontal and procedure is performed by experienced engineers
vertical alignment, which may lead to overestimate (Larocca et al. 2011).
or underestimate the actual available sight distance Methods based on line-of-sight loops enable the
(Ismail & Sayed 2007). It is more common instead, depiction of sight-distance graphs. These charts
to develop algorithms based on line-of-sight represent on the horizontal axis the stations where
loops in three dimensions (3-D). Such procedures the driver is sequentially placed, and on the vertical
retrieve the cross-sectional profile of the terrain axis the sight distance variables ahead each driver
below the line of sight between the observer and position (Kühn & Jha 2011, Castro et al. 2014).
the target locations, detecting whether the vision Besides the comparison available and required
is obstructed. Ismail and Sayed (2007) devised a sight distances, such charts result advantageous to

193

TIS2017_Book.indb 193 2/8/2017 8:02:54 PM

 evaluate the 3-D alignment coordination (Roos & iteratively from every station on the vehicle path
Zimmermann, 2004, Jha et al. 2011, Castro et al. towards the stations ahead, determining whether
2015a). The German Road and Transportation each target is seen by the driver. Once the loop from
Research Association provided a framework a station has been completed, the virtual driver is
both for virtual perspective generation and sight- moved forward to the next station, where an iden-
distance graphs on the design of rural highways tical loop is launched. Available sight distance is
(FGSV 2008). Campoy-Ungría (2015) proposed a defined as the distance, measured along the vehicle
procedure to estimate available sight distance on path, between the driver’s position and the farthest
highways based on prismatic line-of-sight buffers target seen without interrupting the line of sight.
launched directly on a high-density LiDAR cloud According to this definition, the algorithm checks
of points, not requiring any terrain surface. lines of sight from the driver position as shown in
A Geographic Information System (GIS) is use- Figure 1. At station i, the available sight distance is
ful to calculate sight distances because it gathers determined by the station i+2 (line of sight in light
numerous advantages. Besides the 3-D treatment of grey), which is the furthest one seen before the
the sight distance issue, it enables safety integrated first line of sight is blocked (station i+3, with line
analyses accounting for other factors such as geo- of sight in dark grey). Software stores the binary
metrics, accident data, traffic volume, operating value of visibility of every line of sight.
speed and design consistency at ease (Altamira et al. Figure 2 shows the complete process to study
2010; Castro & De Santos-Berbel 2015). Nowadays, sight distance on highways, including the input
affordable data sources are available to characterize
the highway and the roadsides with higher accu-
racy, which GIS is capable to handle and exploit at
ease. Khattak and Shamayleh (2005) assessed high-
way safety through GIS data visualization.
This paper reviews the GIS-based methodol-
ogy developed to study sight distance on highways.
The main issues are contemplated by means of
case study examples. Following this introduction,
the second section provides detailed description
of the methodology. Particular attention is paid
to inputs (driver’s eye height, target height, vehi- Figure 1. Available sight distance estimation through
cle path and elevation model) and output (sight lines of sight.
distance graph). In the third part, case studies are
presented. The first case illustrates the influence of
roadside elements (vegetation) through the use of
a Digital Terrain Model (DTM) and a Digital Sur-
face Model (DSM). The second one describes how
to detect and analyze shortcomings in the spatial
alignment of highways through the use of sight-
distance graph. The third one proposes a solution
based on GIS tools (lines of sight) and multipatch
datasets in order to represent overhanging roadside
features (e.g. cantilever traffic signal) adequately.
Finally, conclusions are presented.

2 METHODOLOGY

2.1 Sight distance algorithm

An application developed on ArcGIS was con-
ceived for the 3-D estimation of available sight dis-
tance on highways and has already been validated
by the authors (Castro et al. 2014). It is capable of
computing the available sight distance of a high-
way section given the driver’s eye height, the target
height, the vehicle path and an elevation model.
The computational routine for available sight Figure 2. Flowchart of the sight distance study
distance estimation launches a line-of-sight beam procedure.

194

TIS2017_Book.indb 194 2/8/2017 8:02:54 PM

 and output data as well as the auxiliary tools The shape of roadside elements is another fact
developed. to bear in mind. DSMs from airborne LiDAR
can hardy model vertical roadside features, such
as traffic signs or guard-rails, although they could
2.2 Input data
represent considerably larger vertical devices (e.g.
Each particular highway requires two important gantries) depending on the resolution. In general,
datasets to study sight distance. Whereas a Digital roadside vertical equipment is better depicted by
Elevation Model (DEM) is necessary to recreate DMSs derived from terrestrial LiDAR.
the highway environment, a file containing the Regarding the trajectory, points that compose
points that define vehicle trajectory is needed. driver’s path could be obtained from several sources.
With respect to DEMs, the two types mentioned If the highway geometrics are known, a theoretical
in the previous section are available for sight dis- vehicle path can be extracted. Otherwise, it might
tance modelling: DTM’s and DSM’s. Either of be deduced from inventories or precise-enough
them is handled by the application as long as it cartographic data. Moreover, the track of a GNSS
is of the form of a Triangular Irregular Network receiver mounted on a car driving along the stud-
(TIN). To choose one over the other is not a trivial ied highway constitutes a reliable data source for
matter owing to their influence in results. A DTM this input when highway geometrics are unknown
is a 3-D representation of the terrain surface which or not reliable. The application has tools that sim-
depicts exclusively the elevation of the bare ground. plify its treatment. Points should have an attribute,
However, the reality contains many more elements namely station, which indicates their distance to
influencing sight distance than the bare ground. the origin measured along this trajectory.
Features such as vegetation, traffic signs, build- As generic inputs, driver’s eye height and target
ings and many other elements are not included in height must be considered. Those values are usually
a DTM. These models comprise such additional taken from highway design standards (AASHTO
roadside features, making available further infor- 2011, Ministerio de Fomento 2016).
mation about features by the roadsides which could The accuracy and resolution of these models,
limit the available sight distance. However, where along with the spacing between the path stations,
overhanging features are present by the roadsides, come also into play. The effect of these factors has
the intrinsic features of DSMs hamper sight dis- been studied by the authors (Castro et al. 2015b).
tance analysis. These surfaces do not support two This was tested by comparing the available sight
points on its surface having the same horizontal distance results of elevation models of different
projection while their elevation values are different. resolutions and varying the spacing between sta-
This fact hinders a lifelike representation of over- tions on each model. The elevation model resolu-
hanging features, which is particularly troublesome tion ranged from 1 to 5 m, all built up of a squared
when they are partially located above the road, as mesh. The paths tested had stations no closer than
occurs for tree crowns or cantilever signals. 1 m and no further than 20 m. It was found that
Current techniques provide cost-effective high the DEM resolution has a larger effect on outcome
resolution DEMs. The remote sensing LiDAR than the spacing between stations.
(Light Detection and Ranging) devices emit a pulse To avoid the issues created by overhanging fea-
beam. The pulse is received back if any surface tures in the DSM, the use of multipatch datasets
is hit, which allows its geospatial location and its can be contemplated. This supports the insertion
characterization (Topcon 2010). Usually, those data of roadside elements such as cantilever signals,
might have been collected by airborne surveying or gantries or overpasses to achieve an adequate sight
terrestrial surveying. The latter ones are known as distance modeling.
Mobile Mapping Systems (MMS). Whereas the air-
borne LiDAR is able to capture around one point
2.3 Output
per two m2, the MMS is able of deliver more than
200 points per m2 when closer to the sensor. How- The results, stored by the application after the com-
ever, the area covered from the airborne standpoint putational process, are plotted on the sight-distance
is much greater and the performance much higher, graph. Also, software may retrieve the longitudinal
whilst the MMS are limited by elements that may profile between the observer and observed points
create shadow areas beyond them. In both cases, at the request of the user. This feature permits the
raw data usually contain much noise. Items not detection of the area that obstructs vision.
forming part of the static landscape are captured, Due to the GIS geolocation capabilities, the out-
such as vehicles or even wildlife. Moreover, the come can be shown on the map. Mapped features
abovementioned overhanging elements, i.e. aerial facilitate the integrated analysis along with other
power lines, hamper the use of DSM. All those factors. Furthermore, results may be exported in
points are therefore entitled to be removed. full detailed reports. The 3D visual inspection of

195

TIS2017_Book.indb 195 2/8/2017 8:02:55 PM

 the scene modelled is also possible in ArcSCENE
to get a better understanding of modelling issues.

3 CASE STUDIES

For the purpose of illustrating the capabilities of

the methodology described hereby, several case
studies are presented. The first case illustrates
the influence of roadside elements (vegetation)
through the use of a DTM and a DSM. The
second case describes how hidden dips are char-
acterized throughout the use of sight-distance
graph. The third one proposes a solution based
on GIS tools (lines of sight) and multipatch data- Figure 4. Sight distance graph comparing results using
DTM and DSM.
sets in order to calculate and analyze the influ-
ence of a cantilever traffic signal in the visibility
of truck drivers. All cases correspond to two-lane Figure 4 shows the sight distance graph super-
rural highways located in the region of Madrid posing the results of both a DTM and a DSM.
(Spain). When a DTM is used as input, the minimum avail-
able sight distance is 55 m (slashed line) whereas
the corresponding to the DSM is 15 m (solid black
3.1 Roadside elements
line) around station 3150. The latter one is more
Roadside elements such as vegetation, traffic signs in line with reality. This value would not comply
or buildings may be taken into account in the sight with the stopping sight distance set at 50 m by the
distance studies when using a DSM instead of a Spanish standard (Ministerio de Fomento 2016).
DTM. Regarding the roadside vegetation, trees Moreover, the available sight distance is reduced
and plants may reduce available sight distances, around 40 m all the way in front of the curve.
especially when it comes to forests and densely This difference is highlighted in medium grey in
wooded areas. A sub-section of highway M-611 Figure 4, which represents the sections that are
was selected to illustrate this issue. The design seen when the input is the DTM whilst the study
speed is assumed to be 40 km/h. A horizontal with DSM determined they cannot be seen. This
curve of radius 23 m is flanked by respective spi- example shows how significant is the influence of
rals and tangents. Figure 3 shows the actual view vegetation by the roadsides by means of the choice
of a vehicle approaching a right curve, where a of DEM.
densely wooded area is found close to the inner In addition, a study carried out by the authors
roadside. In this case, both an airborne DTM and found that sight distance results using airborne
an airborne DSM arranged in a 1-m square mesh DTM, airborne DSM and MMS DSM were all
were used. The vehicle path was derived from car- statistically significantly different (Castro et al.
tographic data and was discretized into stations 2016). The differences were particularly relevant in
spaced 5 m apart. sub-sections where the available sight distance was
shorter.

3.2 Hidden dips

Some combinations of horizontal and vertical
alignments might produce shortcomings in the
driver’s perspective. Hidden dips are a common
shortcoming on highways where the profile adjusts
the terrain more strictly than the horizontal align-
ment. A hidden dip is produced where the driver
is able to see two separate sections of the roadway
while the stretch in between remains concealed.
This typically occurs where a sag follows a crest
curve on a rather straight horizontal alignment.
It is essential to avoid potentially hazardous spots
Figure 3. Real view of curve where sight distance is lim- within the hidden section, such as intersections or
ited by vegetation. unexpected changes in direction. Moreover, these

196

TIS2017_Book.indb 196 2/8/2017 8:02:55 PM

 alignments may mislead drivers at the beginning 165 m only. Thus passing should be prohibited all
of a passing maneuver, since oncoming traffic along the hidden dip range.
remains unnoticed in the hidden section. Furthermore, the German guidelines for the vis-
The sight-distance graphs generated by the appli- ualization of rural roads (FGSV 2008) describe the
cation developed are suitable to detect and analyze conditions under which a hidden dip is potentially
hidden dips. Figure 5 shows a straight section of hazardous for drivers, regardless of the design
M-611 highway on a rolling profile where there speed. Three conditions must be fulfilled simulta-
is a hidden dip. The corresponding sight-distance neously along a range of at least 60 m: The hidden
graph is illustrated in Figure 6. Sections seen by sections must not spread out beyond 600 m from
the driver are colored in light grey, in medium grey the driver, the hidden section has to cover more
stations not seen and in dark grey target stations in than 75 m and the depth of diving must exceed
the dip where an object of 0.75 m height would not 0.75 m (hence the area in dark grey in Figure 6).
be seen. The reason for this value is explained later. These values are largely exceeded in the present
The maximum dip depth is 3.9 m according to the case, reporting a measure of risk exposure during
longitudinal profile between the observer and the passing maneuvers. Hence this spot may be poten-
observed points retrieved by software. tially hazardous.
The perspective shortcoming is first noticed at Similarly, the Swiss standard (VSS 1991) deter-
station 710. When the driver reaches station 725, mines the maximum distance to consider the
the available sight distance is 380 m, a stretch of reappearing section at 500 m for that speed. That
240 m remains hidden, and a further segment is occurs only from station 920, therefore the hazard-
seen over again up to 905 m. The hidden dip ranges ous stretch would range 140 m.
up to station 1060, totaling 350 m.
As this type of shortcoming may produce pass-
3.3 Overhanging elements
ing issues, the study of passing sight distance results
interesting. In this section posted speed is 90 km/h. In this case study, a cantilever traffic signal in
For this value, the current Spanish standard (Min- a highway section was simulated. The aim is to
isterio de Fomento 2016) demands a passing sight analyze the effect of the location of this signal on
distance of 205 m along a section of 340 m or drivers sight distance, avoiding the problems asso-
larger. However this threshold is exceeded along ciated to the use of a DSM. As in previous cases,
a DEM is needed. In this case an airborne DTM
arranged in a 1-m square mesh was utilized. In
addition, the cantilever traffic signal was modeled
as a multipatch file taken from an open library
(Trimble 2015). This multipatch was placed on
5 different locations (Table 1) along a section of
highway M-104. The cantilever traffic signal loca-
tion covers exactly the width of the traffic lane
where signing applies.
The highway horizontal alignment is composed
by a right horizontal curve, followed by a long
tangent and a left curve. In the vertical alignment,
there is a sag curve approximately on the middle
Figure 5. Real view of hidden dip on straight alignment. of the tangent, between the two horizontal curves.
The cantilever traffic signal locations 1 and 2 are
supported by the roadside on different spots of the
right curve. Location 3 is at the beginning of the
sag curve and location 4 is at the lowest point of

Table 1. Locations where

cantilever signal was placed.

Location Station (m)

1 5260
2 5305
3 5400
4 5450
5 5630
Figure 6. Sight-distance graph of hidden dip.

197

TIS2017_Book.indb 197 2/8/2017 8:02:55 PM

 the sag curve. Location 5 is between the sag vertical
curve and the left horizontal curve. Figure 7 shows
a 3D view made using ArcSCENE, where the dif-
ferent locations of the cantilever traffic signal
considered are depicted. The right and the left
horizontal curves overlap approximately with two
crest vertical curves.
The clearance height was 5.5 m, according to the
current Spanish standard (Ministerio de Fomento,
2016) whereas the maximum height of the struc-
ture is 6.81 m. Moreover, such standard requires
that the impact of gantries and cantilever signals
on sight distance is checked. Figure 9. Sight-distance graph corresponding to canti-
In this study, geometric characteristics of a the- lever signal at station 5260 (location 1).
oretical truck path were considered. Therefore the
observer height was set at 2.5 m. According to the
Spanish design standard (Ministerio de Fomento
2016), the path considered was that resulting of the
parallel offset of 1.5 m from the highway center-
line. Also, according to the same standard, target
height was set at 0.5 m. Figure 8 shows the sight-
distance graph when the calculation was launched
without cantilever traffic signal. There is a first
zone of shorter sight distance due to the first hori-
zontal curve and the first crest vertical curve (mini-
mum available sight distance of 115 m at station

Figure 10. Sight-distance graph corresponding to can-

tilever signal at station 5450 (location 4).

4855). Then, the available sight distance increases

until drivers approach the second horizontal curve
and the second crest vertical curve (135 m of mini-
mum sight distance at stations 5680–5715).
Figure 9 shows sight-distance graphs corre-
sponding to location 1 of the cantilever traffic
signal. In location 1, there are some lines of sight
corresponding to driver location between stations
4940 and 5240 that intersect traffic signal post,
but its effect is negligible. The cantilever signal
Figure 7. View in ArcSCENE of the cantilever traffic itself has no effect on sight distance. Similarly,
signal possible locations. Figure 10 shows the sight-distance graph corre-
sponding to location 4 of the cantilever traffic sig-
nal (near station 5450). Comparing Figures 9–10,
it can be noticed that not only the cantilever traffic
signal effect moves (due to the change of location)
but also the non-visible area becomes larger. This
latter effect is due to signal location at the lowest
point of the sag curve. As a result, the cantilever
signal intercepts much more lines of sight, produc-
ing a more relevant hidden area.

4 CONCLUSIONS

The proposed procedure demonstrates the poten-

Figure 8. Sight-distance graph without traffic signal. tial and versatility of GIS in highway sight distance

198

TIS2017_Book.indb 198 2/8/2017 8:02:56 PM

 studies. The inputs needed for the study, namely Development Fund (FEDER). Research Project
the driver’s eye height, the target height, the vehi- TRA2015-63579-R (MINECO/FEDER).
cle path and the elevation model were described.
The importance of the resolution and nature of
the DEM was particularly emphasized. To achieve REFERENCES
precise results, it is desirable to use high resolution
models (1 node per m2). The outcome can be stud- Altamira, A.L., Marcet, J.E., Graffigna, A.B. & Gómez,
ied in detail with the aid of the tools and capabili- A.M. 2010. Assessing available sight distance: an indi-
rect tool to evaluate geometric design consistency. In
ties developed, including the sight-distance graph, Proceedings of the 4th International Symposium on
line-of-sight profiles and mapped features. Sight- Highway Geometric Design.
distance graphs permit a detailed analysis of road- American Association of State Highway and Transportation
side elements which limit sight distance. Officials (AASHTO) 2011. A Policy on Geometric Design
Throughout three case studies on in-service of Highways and Streets. Washington DC: AASHTO.
highways, the strengths and capabilities of the Campoy-Ungría, J.M. 2015. Nueva metodología para la
methodology were proved. To take account of the obtención de distancias de visibilidad disponibles en
effect of roadside elements or vegetation on sight carreteras existentes basada en datos LiDAR terrestre.
distance, DSMs must be used instead of a DTM. Doctoral dissertation. Valencia: Universidad Politéc-
nica de Valencia.
In particular cases, the reduction of available sight Castro, M., Iglesias, L., Sánchez, J.A. & Ambrosio, L.
distance is especially significant while considering 2011. Sight distance analysis of highways using GIS
these entities. The second case study showed how tools. Transportation Research Part C: Emerging Tech-
to detect and characterize sight-hidden dips. The nologies, 19(6): 997–1005.
parameters that may indicate the risk exposure of Castro, M. 2012. Highway design software as support of
these sections, namely range, length of hidden sec- a project based learning course. Computer Applica-
tion, dip depth minimum available sight distance tions in Engineering Education, 20(3): 468–473.
and distance to reemerged stretch can be identified Castro, M., Anta, J.A., Iglesias, L. & Sánchez, J.A. 2014.
at ease with the tools provided. GIS-based system for sight distance analysis of high-
ways, Journal of Computing in Civil Engineering, 28(3):
A DSM leads to biased sight distance model- 04014005.
ling where there are overhanging features because Castro, M., De Blas, A., Rodriguez-Solano, R. & Sanchez,
lines of sight are obstructed by the model surface J.A. 2015a. Finding and characterizing hidden dips in
even below the overhanging feature. The proposed roads. Baltic Journal of Road and Bridge Engineering,
GIS-based procedure may contemplate multipatch 10(4): 340–345.
structures to model them, overcoming the difficul- Castro, M & De Santos-Berbel, C. 2015. Spatial analysis
ties inherent to the presence of overhanging ele- of geometric design consistency and road sight dis-
ments. The third case study showed how to model tance. International Journal of Geographical Informa-
properly a section with a cantilever traffic signal tion Science, 29(12): 2061–2074.
Castro, M., García-Espona, A. & Iglesias, L. 2015b.
and its real impact on sight distance. The easiness Terrain model resolution effect on sight distance on
to place multipatch objects is an additional advan- roads. Periodica Polytechnica: Civil Engineering, 59(2):
tage provided by this procedure. This simplifies the 165–172.
simulation of object location to evaluate its possi- Castro, M., Lopez-Cuervo, S., Paréns-González, M. &
ble effects on sight distance. Also, due to the avail- De Santos-Berbel, C. 2016. LIDAR-based roadway
ability of multipatch datasets libraries, modelling and roadside modelling for sight distance studies. Sur-
effort is reduced. vey Review, 48(350): 309–315.
Therefore the GIS based methodology pre- Forschungsgesellschaft für Straßen- und Verkehrswesen
sented is useful not only to study sight distance but (FGSV) 2008. Hinweise zur Visualisierung von Entwür-
fen für außerörtliche Straßen. Bonn: FGSV Verlag.
also seek for potential safety issues though inte- Forschungsgesellschaft für Straßen- und Verkehrswesen
grated analysis. Diverse operational factors such as (FGSV) 2012. Richtlinien für die Anlage von Land-
accident data, traffic volume, operating speed and straßen. Bonn: FGSV Verlag.
design consistency can be incorporated to locate Ismail, K. & Sayed, T. 2007. New algorithm for calculating
and diagnose potentially hazardous spots or, even- 3D available sight distance. Journal of Transportation
tually, to identify the factors involved in a particu- Engineering, 133(10): 572–581.
lar accident. Jha, M.K., Karri, G.A.K. & Kühn, W. 2011. New three-
dimensional highway design methodology for sight
distance measurement. Transportation Research
Record: Journal of the Transportation Research Board,
ACKNOWLEDGEMENTS 2262: 74–82.
Khattak, A.J. & Shamayleh, H. 2005. Highway safety
The authors gratefully acknowledge the financial assessment through geographic information system-
support of the Spanish Ministerio de Economía based data visualization. Journal of Computing in Civil
y Competitividad and European Regional Engineering, 19(4): 407–411.

199

TIS2017_Book.indb 199 2/8/2017 8:02:57 PM

 Kühn, W., Volker, H. & Kubik, R. 2011. Workplace simu- Roos, R. & Zimmermann, M. 2004. Quantitative meth-
lator for geometric design of rural roads. Transporta- ods for the evaluation of spatial alignment of roads,
tion Research Record: Journal of the Transportation In Proceedings of the 2nd International Conference of
Research Board, 2241: 109–117. Società Italiana di Infrastrutture Viarie (SIIV).
Larocca, A.P., Da Cruz Figueira, A., Quintanilha, J.A. & Topcon 2010. IP-S2 specifications. Available from
Kabbach Jr, F.I. 2011. First steps towards the evalua- Internet: <http://www.topcon.co.jp/en/positioning/
tion of the efficiency of three-dimensional visualiza- products/pdf/ip_s2.pdf>.
tion tools for detecting shortcomings in alignment’s Trimble 2015. Sketchup PRO. Available from Internet:
coordination. In Proceedings of the 3rd International <http://buildings.trimble.com/products/sketchup-pro>.
Conference on Road Safety and Simulation. Vereinigung Schweizerischer Strassenfachleute (VSS)
Ministerio de Fomento 2016. Norma 3.1-IC: Trazado. 1991.Linienführung;OptischeAforderungen(SN-640140).
Madrid: Ministerio de Fomento. Zurich: VSS.

200

TIS2017_Book.indb 200 2/8/2017 8:02:57 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Performance improvements of asphalt mixtures by dry addition

of polymeric additives

M. Ranieri & C. Celauro

Department of Civil, Environmental, Aerospace, Materials Engineering, University of Palermo, Palermo, Italy

L. Venturini
Iterchimica s.r.l., Suisio, Bergamo, Italy

ABSTRACT: This paper shows the results of an experimental study concerning the development and
optimization of asphalt mixtures for binder and base courses, improved with specifically engineered
additives. The focus was on the mechanical improvements of the mixtures as achievable via dry
modification with polymeric additives by making use of aggregate and bitumen of average quality, as
locally available, in order to limit the consumption of virgin materials. The results allowed interesting
conclusions to be drawn about the use of polymeric additives for these mixtures. In particular, the
modified mixtures proved to have better performance in terms of both permanent deformation resistance
and stiffness modulus. Moreover, it is possible to use lower binder contents, thus proving the economic
feasibility of this modification, when considering the advantage of consuming mixtures with components
of average quality for production of highly performing mixtures.

1 INTRODUCTION ent sectors. The polymer market shows a strong

potential for development and diversification;
In recent years growing traffic has demanded the reasons of this success are clear: low weight,
higher quality pavements. Binder and base courses workability, versatility, hygiene, different selec-
must have structural characteristics such as not to tion options, recycling and recovery (Celauro et al.
allow degradations. A method that can improve 2004).
the quality of the pavements, avoiding degrada- There are two main processes for adding poly-
tion, with benefits (and savings) regarding the need mers to asphalt mixture: by modifying the bitumen
for maintenance, is the addition of polymers in the (PMB—wet process); or by adding the polymers
asphalt mixtures (Bense 1983, Serfass et al. 2000). during the mixing phase (PMA—dry process)
In addition awareness of environmental prob- (Pettinari et al. 2014).
lems has led to research being focused on devel- The wet process needs specific equipment (for
opment of new technologies of additives that can mixing and to facilitate the reaction with bitumen
improve the mechanical performances of asphalt at high temperatures), while this is not required for
mixtures. Among these specially engineered addi- the dry process. Therefore, the dry process is easier
tives are interesting solutions, both to meet the to implement.
related environmental issues, in Italy and in all The aim of this study was optimization of
other economically developed countries, like for mixtures that do not necessarily employ high-
example the productive reuse of secondary raw performance materials, making use of locally avail-
materials, and to solve specific technical require- able stone aggregate and bitumen, also aiming to
ments of pavements by improving the physical and improve the traditional mixtures as made possible
mechanical properties of the mixtures. by dry addition of suitable polymers.
The use of additives makes it possible to increase
the performance of asphalt mixtures, by decreas-
2 MATERIALS AND METHODS
ing production and laying costs, and to reduce the
environmental impact.
The experimental plan was developed by using the
It also favors sustainable development because
following materials (aggregates, binders, additives,
it makes it possible to reduce consumption of valu-
mixtures):
able natural resources (Venturini et al. 2016).
In this context plastics are the materials with • aggregates (Dmax = 16 mm);
the highest degree of diffusion in many differ- • binder (neat bitumen 50/70);

201

TIS2017_Book.indb 201 2/8/2017 8:02:59 PM

 • additive (plastomeric compound); Table 2. Physical and mechanical characteristics of the
• mixtures: available aggregates.
α. traditional mixture;
Fractions
β. mixture with plastic additive.
The mix design was achieved by carrying out: Characteristics a2 a3 a4 a5 filler

1. Marshall test with 4 percentages of bitumen Bulk specific 2.85

and CE = 75 blows per face (compaction effort), weight, g/cm3
according to the EN 12697-34:2012 standard; (EN 1097-7:2008)
2. compactability test with a gyratory compac- Apparent specific 2.82 2.83 2.84 2.85
tor (D = 150 mm), in accordance with the EN weight, g/cm3
12697-31:2007 standard. (EN 1097-6:2013)
Los Angeles 22.10 20.19 20.64 20.12
Finally, tests on optimized mixtures were per- abrasion, %
formed in order to assess the following properties: (EN 1097-2:2010)
Sand equivalent, 91.38 90.41
rutting resistance, according to the EN 12697- % (EN 933-8:1997)
22:2007 standard, method B; Voids ratio 0.80 0.79 0.71 0.73
complex stiffness modulus, according to the EN (CNR 65:1978)
12697-26:2012 standard, annex B; Absorption 0.64 0.51
fatigue resistance, according to the EN 12697- coefficient
24:2012 standard, annex D. (EN 1097-6:2013)

2.1 The stone aggregates

For composition of mixtures it appears appropri-
ate to refer to aggregates with a diameter between
0 and 16 mm, a suitable range for binder and base
courses. The aggregates used in the mixtures were
crushed limestone from a quarry whose composi-
tion and physical and mechanical properties are
given respectively in Table 1 and Table 2.
The mixture target (a typical dense graded high
modulus asphalt) was obtained for the composi-
tion of the fractions provided by implantation and
by appropriate sub-fractions (see Fig. 1).

Table 1. Composition of the aggregates’ available

fractions.

Passing
Figure 1. Mix grading curve.
%

Fractions The mixture of aggregates was then subjected to

the pycnometer method for determination of the
Sieve a2 a3 a4 a5 real specific weight and the value was 2.842 g/cm3,
mm (20/25) (10/15) (6/10) (0/6) Filler according to the EN 1097-3:1999 standard.
32 100 100 100 100 100
24 100 100 100 100 100 2.2 Bitumen
20 94.73 100 100 100 100
The bitumen used in this research was neat bitumen
12 18.33 99.96 100 100 100
50/70, more easily available in contexts such as in
8 0.77 85.67 99.91 99.70 100
Italy; its characteristics are reported in Table 3.
4 0.51 38.3 72.62 97.78 100
2 0.49 15.66 41.54 70.60 100
0.4 0.44 6.62 16.84 26.09 99.31 2.3 Additive
0.18 0.41 5.28 11.21 17.19 93.78
0.075 0.33 4.01 6.42 10.57 74.25 The additive used was a particular polymeric
compound of selected polymers, Low-Density

202

TIS2017_Book.indb 202 2/8/2017 8:02:59 PM

 Table 3. Characteristics of the bitumen used in the mix-
tures studied.

Characteristic Unit Value Standard

Specific weight g/cm3 1.033 EN 3838:2005

at 25°C
Penetration dmm 68 EN 1426:2015
at 25°C
Softening point °C 50.5 EN 1427:2015
Ring and Ball
Penetration Index −0.21 EN 12591:2009
Temperature Fraas °C −12 EN 12593:2015
Ductility at 25°C mm >100 ASTM D113:2007 Figure 2. DSC test results on plastomeric compound.
Viscosity at 60°C Pa⋅s 255.5 EN 13302:2010
Viscosity at 100°C Pa⋅s 3.917 EN 13302:2010
Viscosity at 135°C Pa⋅s 0.435 EN 13302:2010
106°C and it is usually for low-density polyethyl-
Viscosity at 150°C Pa⋅s 0.222 EN 13302:2010 ene, the second one at 120°C is for the high-density
Mixing temperature °C 155 EN 13302:2010 polyethylene, while the last peak at 160°C is for
(η = 0.17 Pa⋅s)
polypropylene.
Compaction °C 145 EN 13302:2010
temperature
Since the area under the curve is proportional
(η = 0.28 Pa⋅s) to the mass and the two areas at 120°C and 160°C
After RTFOT: are quite low, the quantity of polypropylene and
Change in mass % −0.19 EN 12607-1:2015 High-Density Polyethylene (HDPE) is minimal.
Penetration at 25°C % 44 EN 1426:2015 The other peaks are not important because they
Softening point °C 64.5 EN 1427:2015 refer to crystallization of polymers at low tem-
Ring and Ball peratures. This test confirms that the plastomeric
Viscosity at 60°C Pa⋅s 668 EN 13302:2010 compound is a low-density polyethylene with a
small quantity of high-density polyethylene and
polypropylene.
This additive has low affinity with the bitumen
and therefore, for practical needs, it is much more
Polyethylene (LDPE), Ethylene (EVA) as well
advantageous to add additives in the asphalt mixture
as others with low molecular weight and average
(known in the technical literature as DRY method):
melting point, that is semi-soft and flexible gran-
the additive is added to the hot aggregates, before
ules. The plastomeric compound was not designed
mixing with the bitumen (Celauro et al. 2004). In
for modification of the bitumen, but rather to
view of this, the optimal process to make a mix-
improve the mechanical performance and durabil-
ture is the succession of the following components:
ity of the asphalt mixtures (Iterchimica S.r.l 2015).
aggregates, additives, bitumen and filler.
The physical properties are:
• aspect: granules;
• colour: shades of grey; 3 FORMULATION TESTS
• dimensions: 2.00–4.00 mm;
• softening point: 160°C; 3.1 Marshall Stability
• melting point: 180°C;
The first step in the formulation of these mixtures
• melt index: 1÷5;
was applying the Marshall Method only to the tra-
• apparent density at 25°C: 0.40–0.60 g/cm3.
ditional mixture in order to assess the physical and
• specific weight: 0.934 g/cm3.
mechanical characteristics typically considered in
A DSC (Differential Scanning Calorimetry) test the Italian specifications.
was carried out on this material in accordance with Four percentages of bitumen were selected
the ISO 11357-3:2011 standard, which makes it (b’1 = 4.8%, b’2 = 5.2%, b’3 = 5.5%, b’4 = 5.9%, by
possible to characterize thermal behaviour because weight of the aggregates) and four specimens for
it provides temperature and enthalpy values cor- each percentage were produced, for a valid repeti-
responding to glass transition, melting point and tion. The volumetric properties (v, air voids, and
crystallization through heating from −60°C to VFB, voids filled with bitumen) were determined
180°C, cooling from 180°C to −60°C and again according to the requirements of the EN 12697-
heating; the results are shown in Figure 2, where 8:2003 standard. The calculation of the maximum
there are three peaks: the first is more or less at specific weight (γt) of the mixture was performed

203

TIS2017_Book.indb 203 2/8/2017 8:02:59 PM

 according to the “C” (mathematical) process Table 5. Marshall test results in accordance with MIT
specified by the EN 12697-5:2010 standard, while specification.
the calculation of the apparent specific weight
(γapp) was performed according to the EN 12697- Course
6:2012 standard. The results of the Marshall test Required results Unit Base Binder
are reported in Table 4 and Figures 3–4, where S
Marshall Stability, S KN 8 10
Table 4. Marshall test results (2 × 75). Marshall Ratio KN/mm >2.5 3 ÷ 4.5
Residual voids % 4÷7 4÷6
B v S F R VFB γapp
% % KN mm KN/mm % g/cm3
is the Marshall Stability, F is the Marshall Flow
4.8 6.03 12.24 4.27 2.87 69.22 2.46 that is the corresponding displacement and R the
5.2 2.88 13.39 4.07 3.29 82.90 2.52 Marshall Ratio S/F.
5.5 2.52 14.67 4.40 3.33 86.05 2.52 As can be seen from the graphs above, regarding
5.9 1.82 13.30 5.33 2.49 89.75 2.53 the Italian specifications, such as the example in
Table 5 (MIT 2001), the requirements are respected
for bitumen contents between 4.8 and 5.1%.

3.2 Gyratory compactor

After some information obtained from the Marshall
method the second step in the mix design of these
mixtures was to consider some compactibility char-
acteristics also considered by the Superpave method
introduced in 1992 by the Strategic Highway
Research Program SHRP (Cominsky et al. 1994).
The mixtures were subjected to gyratory com-
paction in such a way that a measure of the internal
stability of the mixture during the compaction was
obtained too, i.e. the gyratory shear ratio σ = S/P,
where S is the shear stress, and P is the ram pres-
sure. Acquisition of this feature was carried out on
specimens compacted up to N = 200 rpm, in order
to evaluate the behaviour of the mixtures in the dif-
ferent conditions of densification that affect it from
Figure 3. Marshall Stability and displacement, at dif-
ferent bitumen contents.
the time of laying and throughout the design life
years. In general, an increasing trend of the shear
ratio σ in the initial stage of compaction (approxi-
mately, in the first 50 rpm) and stabilization even
beyond the maximum value of N, together with
fulfillment of the volumetric requirements (VMA,
VFA), ensure a correct formulation and good sta-
bility during operation (Roberts et al. 1996).
This test gives a good idea of the job-site density
values, according to course thickness. Conducted
ahead of the other mechanical tests, this test is used
to make a preliminary selection or screening of mixes,
and for optimizing the asphalt mix composition.
The gyratory compactor test was initially car-
ried out on two specimens with different bitumen
contents (4.9, 5.1 and 5.4%) and three different
plastomeric compound contents (P.0 = 0% of poly-
mer, P.3 = 0.3% of polymer, P.6 = 0.6% of polymer,
by weight of mineral aggregates). Later, the speci-
mens with additives being subjected to bleeding
Figure 4. Marshall Ratio and Marshall voids, at differ- due to the excessive binder content, it was also
ent bitumen contents. decided to carry out the test, for the percentage 0.3

204

TIS2017_Book.indb 204 2/8/2017 8:02:59 PM

 of the plastomeric compound, on two specimens From the values reported in Table 6 it can be
with two lower bitumen contents (4.3, 4.6). The observed that for the same aggregate skeleton the
densification curves recorded during the gyratory workability does not depend on the bitumen or
compaction made it possible to obtain parameters additive content. Instead, when the bitumen or
of the regression lines, K and C1, that respectively additive content increases, the values of the initial
define the workability and the self-densification of densification C1 and, consequently, the compact-
these mixtures (see Table 6). ness at any number of revolutions also increases.
Shear ratio values, determined during the test as
mentioned above and automatically recorded by
Table 6. Values of workability and self-densification. the equipment, are reported in Figure 5.
The shear ratio lines in Figure 5 show that the
% Gmm = C1 + k * log (N)
b p specimens with a percentage of 0.3 of plasto-
Mixture % % C1 K R2 meric compound and 4.3 or 4.6 of bitumen, and
specimens without any percentage of plastomeric
AC 4.3/P.3 4.3 0.3 0.8009 0.0757 0.9980 compound but with the percentage 4.9 or 5.1 of
AC 4.6/P.3 4.6 0.3 0.8072 0.0739 0.9984 bitumen, mobilize shear ratio values that are main-
AC 4.9/P.0 4.9 0.0 0.7972 0.0753 0.9989 tained constant during the design life years. By
AC 4.9/P.3 4.9 0.3 0.8141 0.0788 0.9952 contrast, a slight excess of bitumen and a slight
AC 4.9/P.6 4.9 0.6 0.8278 0.0748 0.9900 excess of additive cause a fall of the shear ratio and
AC 5.1/P.0 5.1 0.0 0.8190 0.0710 0.9988 therefore the content is not optimal.
AC 5.1/P.3 5.1 0.3 0.8168 0.0786 0.9940 Regarding the voids, Table 7 shows the aver-
AC 5.1/P.6 5.1 0.6 0.8274 0.0755 0.9880 age values of specimens prepared with the gyra-
AC 5.4/P.0 5.4 0.0 0.8141 0.0788 0.9952 tory compactor that, considering for example the
AC 5.4/P.3 5.4 0.3 0.8345 0.0741 0.9837 requirements on the air voids (Anas S.p.A 2009) at
AC 5.4/P.6 5.4 0.6 0.8366 0.0742 0.9702 10, 100 and 190 rpm, reported in Table 8, confirm
the results seen in Figure 5.

Table 7. Average values of the air voids of the speci-

mens at 10, 100 e 190 gyrations.

No. of gyrations
b p
Mixture % % 10 100 190

AC 4.3/P.3 4.3 0.3 12.6 4.7 2.8

AC 4.6/P.3 4.6 0.3 12.1 4.4 2.6
AC 4.9/P.0 4.9 0.0 13.0 5.1 3.2
AC 4.9/P.3 4.9 0.3 11.1 2.7 1.0
AC 4.9/P.6 4.9 0.6 10.2 2.0 0.7
AC 5.1/P.0 5.1 0.0 11.2 3.8 2.0
AC 5.1/P.3 5.1 0.3 10.8 2.4 0.8
AC 5.1/P.6 5.1 0.6 10.2 1.9 0.6
AC 5.4/P.0 5.4 0.0 9.9 1.7 0.4
AC 5.4/P.3 5.4 0.3 9.6 1.4 0.3
AC 5.4/P.6 5.4 0.6 9.5 1.0 0.3

Table 8. Air void values with varia-

tion in the number of gyrations (Anas
S.p.A requirements).

N° gyrations % voids

10 11 ÷ 15
100 3÷6
Figure 5. Shear ratio lines of the traditional mixture 190 ≥2
and mixture with plastomeric compound.

205

TIS2017_Book.indb 205 2/8/2017 8:03:00 PM

 4 TESTS ON OPTIMIZED MIXTURES Table 9. Values of WTSair (Wheel-Tracking
Slope in air) of the mixtures studied.
4.1 Permanent deformation resistance
WTSair
The wheel tracking test was carried out consider-
ing two percentages of bitumen 50/70 as well as Mixture mm/103 cycles
two percentages of plastomeric compound (i.e. P.0
the control mixture and P.3 the one with 0.3% of AC 4.9/P.0 0.10
polymer), and making overall two slabs for each AC 5.1/P.0 0.13
AC 4.3/P.3 0.01
mixture, with dimensions 305 × 305 × 50 mm and
AC 4.9/P.3 0.01
air voids content v = 4.5%. The average values are
reported in Figure 6.
In Figure 6 it is possible to note that, for the
The fact is that this parameter increases when the
same percentage of plastomeric compound, there
percentage of bitumen increases, while it decreases
is no substantial difference regarding rut depth; by
when the percentage of additive increases.
contrast, at the same percentage of bitumen, the
rut depth values of the mixture with the plasto-
meric compound are half of those of the mixture 4.2 Stiffness and fatigue cracking resistance
without the plastomeric compound.
The loading configuration adopted in the fatigue
Moreover, the introduction of the plastomeric
tests (as well as in the complex modulus tests) was
compound leads to almost constant results with
sinusoidal bending on prismatic specimens con-
variation in the bitumen content in the mixture, for
strained at the two outer clamps and point loads at
the contents studied, unlike what happens for tra-
the two inner clamps (4 point bending beam test).
ditional mixtures, which always have an increase in
The deformation was kept constant during the test
rut depth with an increase in the bitumen content.
(controlled strain).
This leads to the conclusion that the introduction
The tests were conducted at 20°C and 10 Hz
of the polymer, making permanent deformation
on beams with dimensions 400 × 45 × 50 mm
resistance indifferent to variations in the bitumen
and obtained from a slab of dimensions
content, is advantageous in the case of mixtures
400 × 305 × 50 mm. The deformation was 350 με for
which require particularly high bitumen contents.
short duration tests and 150 με for long-term tests.
The additive used thus appears to be particu-
The fatigue criterion used was the classical one,
larly advantageous regarding permanent deforma-
referenced as Nf50. It corresponds to the number of
tion resistance, since it reduces rutting, increasing
cycles for which the modulus decreases to 50% of
the cohesive ability of the bituminous mixture, as
its initial value. The initial value was calculated at
already demonstrated by many tests (Iterchimica
the 100th load cycle. The value of the strain ampli-
S.r.l 2015). A confirmation can be obtained by
tude leading to failure at one million cycles is here-
calculating an additional parameter as the wheel-
after called “ε106”.
tracking slope in air (WTSair), that is the average
The complex modulus tests were carried out in
rate at which rut depth increases with the number
the same way as the fatigue tests, with the only dif-
of passages (see Table 9).
ference that the test did not finish when the beams
broke or when the value was half its initial value
(at 100th cycle), and was only carried out up to
150 cycles. The deformation was 25 με and the
frequencies were 1, 10, 30 and 1 Hz. The com-
plex modulus tests and the fatigue tests were car-
ried out on optimized mixtures, that is ones with
a content of 4.9% of bitumen for the traditional
mixture, 4.3% of bitumen and 0.3% of plastomeric
compound.
The complex modulus test parameters are
reported in Table 10 and show what it was reason-
able to expect: the complex modulus values are
highest at low temperatures and high frequencies,
and are lowest at high temperatures and low fre-
quencies; besides, the mixture with the plastomeric
compound has higher values than the traditional
mixture at the same temperature and frequency
Figure 6. Trend of rut depth in the wheel-tracking test. conditions.

206

TIS2017_Book.indb 206 2/8/2017 8:03:00 PM

 Table 10. Values of complex modulus and phase angle Table 11. Fatigue line parameters of the mixtures
of the mixtures studied. studied.

|E*|, Mpa φ, Deg. Mixture A B R2 ε10

6

Frequency, Hz Frequency, Hz AC 4.9/P.0 6328 −0.262 0.9828 169.5

AC 4.3/P.3 4532 −0.240 0.9797 164.5
Mixture 1 10 30 1 10 30

AC 4.9/P.0 3692 7282 8756 36 24 23 the stiffness of the bituminous mixture (dynamic
AC 4.3/P.3 4168 7528 9188 30 20 18 modulus) was direct compression on cylindrical
specimens. The tension was kept constant during
the test (controlled stress), which was carried out
according to the EN 12697-26 standard, annex D.
The tests were conducted at 10, 20, 30 and
40°C and six frequencies for temperature (20, 10,
5, 1, 0.5 and 0.1 Hz) on cylinders with dimensions
100 × 150 mm and obtained after compaction with
a gyratory compactor. The stress levels applied were
chosen in such a way that the strain response was kept
within 50–150 με. The complex modulus tests were
carried out on the optimized mixtures seen before,
with two percentages of voids (2.5% and 5.5%).
The isotherms obtained from the triaxial cell were
used for determination of the Master Curves. The val-
ues of the shift factor were calculated and optimized
according to the formula of Arrhenius (Celauro et al
2010). In this way, by horizontal translation of the
Figure 7. Fatigue lines. shift of isotherms relating to the test temperatures it
was possible to construct the master curve at a refer-
Fatigue test results made it possible to obtain the ence temperature of 20°C for each bituminous mix-
regression lines (Wöhler curves) shown in Figure 7, ture. Figures 8 and 9 show the master curves for the
i.e. Eq. (1). two mixtures with 2.5% and 5.5% voids.
From Figure 8 it is interesting to note that in
ε = a*N−b (1) the whole range of frequencies studied the mixture
with the plastomeric compound provides higher
For each fatigue lines the following parameters dynamic modulus values than those of the tradi-
were calculated in order to make a judgment on tional mixture.
performance: Compared to the previous case, in Figure 9 it is
interesting to note that at low frequencies the mix-
• “a” is a constant and it depends on the physical
ture with the plastomeric compound still provides
and mechanical characteristics of the material,
higher dynamic modulus values than the tradi-
test temperature and frequency;
tional mixture, while at high frequencies it provides
• slope of the fatigue lines (b);
similar values.
• coefficient of determination (R2);
For the mixtures studied it is observed that the
• admissible strain level at N = 106 loading appli-
dynamic modulus always increases with an increase
cations (ε106) in order to characterize fatigue
in the percentage of the voids in the mixture (in the
resistance.
range of investigation).
Table 11 summarizes these parameters for the However, while for the mixtures without addi-
mixtures studied. tives this increase in the mechanical performance
The fatigue lines have high values of the regres- (manifested by an upward shift of the Master
sion coefficients R2 (see Table 11) and this means Curve of the dynamic modulus), is observable, both
that the results are slightly dispersed and very at low and high frequencies in the test (and hence
reproducible. The fatigue lines being very close both at high and low temperatures), in the case of
to each other, one can conclude that the mixtures the mixture with the plastomeric compound this
studied offer comparable performances, in terms increase is much more limited, the values of the
of fatigue resistance. Finally, the complex modulus dynamic modulus being very similar for the two
was also evaluated by means of a triaxial cell. The different percentages of voids studied (the Master
loading configuration adopted for characterizing curve are close).

207

TIS2017_Book.indb 207 2/8/2017 8:03:01 PM

 Moreover, the presence of the plastomeric com-
pound allows optimization of mixtures with the
lowest binder contents and also makes it possible
to obtain higher complex modulus values than tra-
ditional mixture without additive (35% more) and
leads to the same performance regarding fatigue
resistance. From triaxial cell tests it can be said
that the dynamic modulus always increases with an
increase in the percentage of voids in the mixture.
However, while for the traditional mixture there is
a slight increase in the mechanical performance,
both at low and high frequencies of the test (and
hence both at high and low temperatures), in the
case of mixtures with the plastomeric compound,
there is substantial insensitivity of this parameter
characteristic to variation in the air void content.
Figure 8. Master curves with v = 2.5%. A possible future development is a thorough
statistical study considering several kinds of bitu-
men and sources of aggregate, as locally available
in order to highlight the performance sensitivity to
aggregate-bitumen combinations.

REFERENCES

Anas 2009. Capitolato Speciale d’Appalto per lavori

stradali. Norme tecniche.
Bense, P. 1983. Enrobés armés par déchets de matières
plastiques. Bull Liaison Lab Ponts Chauss, 99–106.
Celauro, B. et al. 2004. Effect of the construction process
on the performance of asphalt paving mixes with
Crumb Rubber from scrap tires. 3rd Eurasphalt &
Eurobitume Congress, ISBN 908 0288 446, Vienna,
757–769.
Celauro, C. et al. 2010. Production of innovative, recycled
Figure 9. Master curves with v = 5.5%. and high-performance asphalt for road pavements.
Resources, Conservation and Recycling, 54 (6):
337–347.
This leads to the consideration that the Cominsky, R.J. et al. 1994. The Superpave Mix Design
mechanical performances of the mixture are less Manual for New Construction and Overlays. Report
dependent on the variation in air void content, SHRP-A-407, Strategic Highway Research Program-
with substantial insensitivity to this parameter. National Research Council-Washington, DC.
Iterchimica S.r.l 2015. SuperPlast. Conglomerati
bituminosi ad elevate prestazioni. Libretto Tecnico.
5 CONCLUSIONS Iterchimica S.r.l, Bergamo.
MIT 2002. Studio a carattere pre-normativo delle Norme
Tecniche di Tipo Prestazionale per Capitolati speciali
Limited to the materials studied, the laboratory d’Appalto. MIT Ministero delle Infrastrutture e
tests reported in can be considered adequate to dei Trasporti-Ispettorato per la Circolazione e la
indicate the possibility of using the plastomeric Sicurezza Stradale.
compound in order to improve mechanical per- Pettinari, M. et al. 2014. The effect of Cryogenic Crumb
formances of bituminous mixtures produced with Rubber in cold recycled mixes for road pavements.
component of average quality. Construction and Building Materials 63: 249–256.
The optimized mixtures show good stability Roberts, F.L. et al. 1996. Hot Mix Asphalt Materials,
values and compaction and the positive influence Mixture Design, and Construction. NAPA Research
that the plastomeric compound has on these mix- and Education Foundation (2), Lanham-Maryland.
Serfass, J.P. et al. 2000. Enrobé bitumineux modifiés
tures regarding permanent deformation resistance au polyethylene. Revue Génerale des Routes et
(at the same bitumen content, the mixture with the Aérodromes 787: 47–57.
plastomeric compound is more resistant than the Venturini, L. et al. 2016. New technologies in road
traditional mixture, with rut depth values reduced pavement design-Increase of the service life. Second
to less than half). Serbian Road Congress. June 9–10.

208

TIS2017_Book.indb 208 2/8/2017 8:03:01 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

An application of the parallel gradation technique to railway ballast

M. Esposito, F.S. Capaldo, S. Andrisani & B. Festa

University of Naples “Federico II”, Naples, Italy

ABSTRACT: Minimizing the decay phenomenon of ballast performances is essential to preserve func-
tions and substructure over time.
The track design usually ignores the degradation of ballast and associated plastic deformations result-
ing from the passage of many wheels. This is due to a lack of understanding of failure mechanisms of
this material and the absence of stress-strain constitutive models, which include plastic deformation and
breakage of the particles under a large number of load cycles.
This limitation involves an oversimplified design of the ballast and technological deficiencies in the
construction, which will impact on frequent and expensive maintenance. A good understanding of the
ballast behavior and parameters that control its performance will help reduce ballast maintenance costs
while at the same time preserving «effectiveness» and «efficiency».
In order to keep high standards related to ballast fulfillments, several railway standards select material
based on physical and geometric requirements, but there is never a direct correlation with the deformation
behavior of the track: using a stiffness parameter of subgrade would allow one to consider the track as a
beam resting on an elastic foundation.
The aim of this study is to characterize the ballast by means of experimental tests, in order to assess the
resilient modulus provided by dynamic triaxial tests, simulating the passage of the wheels.
According to UNI-EN13286–7 and ASTM-D2850–87 the specimen to be tested must have a diameter
D greater than 5Dmax (maximum particle size of the material) and height H = 2D.
Since the ballast contains aggregates too large (Dmax = 6–7 cm) to be tested in the common triaxial
apparatus, this study applies the parallel gradation technique to railway ballast. The aim of this approach
is to obtain a scaled down grading curve compatible with the available equipment, keeping shape of the
particles, surface roughness and mineralogical nature.

1 THE BALLAST beating the ballast during its descent, can gener-
ate dynamic forces stressing the ballast and causing
1.1 Introduction scaling, size redistribution and rubbing; also, the
lift of material from the foundation can occur in
The railway track is placed on the ballast, which
cases of a lack of substructure or in the presence
plays the role of elastic soil. Its functions are to:
of a poorly constructed substructure or even con-
− distribute vertical loads on the road body; tamination of the crushed stone, generated by the
− ensure the track geometric design conditions loss of loads, vegetation residues or other atmos-
(levels and alignments due to building and main- pheric agents (Lichtberger, 2010).
tenance conditions); Minimizing the decay phenomenon of ballast
− absorb stresses induced in the track by trains; performances is essential to preserve its functions
− confer elasticity to the track in order to allow and the substructure over time.
the arming to react elastically to the stresses to
which it is subjected;
− allow drainage of rainwater; 1.2 Ballast fulfillments
− absorb longitudinal stresses due to temperature
There are no universal standards to this material’s
changes;
characteristics, such as size, shape, hardness, fric-
− provide a filter between the track and environ-
tion, texture, abrasion resistance and the particle
ment against vibrational phenomena.
size composition which can allows to obtain the
These properties can be affected by the pres- optimum performance under all load types and
ence of the operating loads that, due to the rolling with all types of subgrade; therefore, a large vari-
process, determine the lifting of the sleeper that, ety of materials are employed for the formation of

209

TIS2017_Book.indb 209 2/8/2017 8:03:03 PM

 the layer (i.e. basalt, limestone, granite, dolomite, Table 2. Granulometric curve AS 2758.7.
gneiss and quartzite).
In order to ensure the functions which the bal- Sieve n° Passed percentage
last must carry out, and which have been defined mm %
in section 1.1, it must have characteristics such as 63.00 100
size, shape, particle size, surface roughness, particle 53.00 85–100
density, resistance, durability, hardness, resistance 37.50 20–65
to attrition and water. 26.50 0–20
In Italy the activity of management and mainte- 19.00 0–5
nance of the railway network, as well as the design, 13.20 0–2
construction and start-up of the plants is provided 9.50 –
by RFI (Rete Ferroviaria Italiana). 4.75 0–1
According to the Technical Standards provided 1.18 –
by RFI, and in compliance with the UNI EN 0.075 0–1
13450, the stone used for a new railway ballast or
for renewal and maintenance, must comply with
special requirements. The UNI EN 13450 con- Table 3. Granulometric curve AREMA.
tains all the requirements that must be followed
for the identification of the material; in Table 1, Sieve n° Passed percentage
for instance, granulometric curve fulfillments are inch %
summarized.
In addition to the Italian requirements, different 2 1/2 100
standards and technical specifications have been 2 90–100
issued by different railway organizations around 1 1/2 60–90
the world to meet their design requirements. In 1 10–35
general, as already seen with Italian standards, 3/4 0–10
the ballast has to be edgy, strong, hard and tough 3/8 0–3
under the expected traffic and in difficult environ- No 4 0–0.5
mental conditions, and have a uniform particle size
distribution.
For example, the Australian law is the AS 2758.7: the ballast must fulfill in terms of granulometric
“Aggregates and rock for engineering purposes. curve according to American standards (AREMA,
Railway ballast” (AS2758.7, 1996) (Table 2). 2010), are reported in Table 3 by way of example.
The AREMA (The American Railway Engi-
neering and Maintenance-of-Way Association) is
1.3 Mechanical characterization of ballast
a North American rail industry group. Like RFI
in Italy, it publishes technical specifications for As discussed above, various railway standards in
the design, construction and maintenance of rail the world (RFI, Australian and AREMA) select
infrastructure, thus defining the requirements in the ballast for railway use, based on physical tests
the United States and Canada. to assess the correct form of aggregates, granu-
The material intended to be used as a railway lometric curve, abrasion resistance and climate
ballast, is defined in terms of physical and geo- changes, but the identified parameters never have
metrical characteristics, responding to grain size, direct correlation with the deformation behavior
elongation, specific weight, the friction resist- of the track and which would allow to schematize
ance requirements, etc. The requirements which its behavior as a beam resting on elastic founda-
tion, characterized by the parameter K. the triaxial
test is probably the most appropriate test to assess
Table 1. Granulometric curve UNI EN 13450. the material’s ability to withstand vertical forces.
The stress state of the ballast during the passage
Sieve n° Passed percentage of the train can be, with a good approximation,
mm % well described by the compression phase in a triax-
ial test, limited to the area which is located below
80.00 100
the sleepers, and by the traction phase with refer-
60.00 100
ence to the area between two successive sleepers
50.00 70–99
(Kaya M., 2004), as shown in Figure 1.
40.00 30–65
The aim of the study is, therefore, to character-
31.50 1–25
22.40 0–3
ize the ballast by means of experimental tests, by
referring to the Resilience Modulus MR, which is

210

TIS2017_Book.indb 210 2/8/2017 8:03:03 PM

 Figure 1. Stresses in ballast.

used to represent the load subgrade bearing capac- Figure 2. Contact stresses between two spheres.
ity when it is stressed by mobile loads, such as road
or rail loads. MR is defined as the ratio between
the deviator stress q and the corresponding axial mula for the maximum stress at the contact of two
recovered deformation and is evaluated, for exam- bodies.
ple, through a triaxial test. The maximum contact stress, σNmax, is located at
r = 0, where the radius of the contact area is a. This
calculation assumes the two objects are spherical
1.4 Parallel gradation scaling technique and of the same radius, R. P is the compressive
According to UNI-EN13286-7 and ASTM- force acting on the particles and G is the shear
D2850-87 the test specimen must have a diameter modulus of the materials. This equation assumes
D greater than 5*dmax (maximum particle size of that both materials exhibit the same elasticity and
the material) and height H = 2D, in order to reduce Poisson’s ratio, ν (Hertz, 1956).
the edge effect given by the plates during the test.
With reference to a railway ballast, the maxi- ⎧ 3P 4Ga
⎪ σN = =
mum size of the aggregates is less or more 6–7 cm, ⎪
max
2π 2
( υ )π R
involving D ≥ 35 cm and H = 70 cm. ⎨ 1 1 (1)
Since the ballast contains aggregates too large to be ⎪ ⎡ 3P ( - υ ) ⎤ 1 ⎡ 3 ( υ ) PR ⎤ 3
a =
⎪ ⎢ 4E ⎥ 2 ⎢ = ⎥⎦
tested in the equipment that are located in the com- ⎩ ⎣ ⎦ ⎣ G
mon test laboratories, the application of the parallel
gradation scaling technique is proposed, in order to This relationship for particles having a perfect
model this material. The objective of this technique is geometric similarity shows that the values of con-
to retain the shape of the particles, the surface rough- tact stresses and strains are independent of particle
ness and the mineralogical nature in order to obtain size. Laboratory data of tests run on quartz, which
a reduction of the material that is characterized by is a highly elastic material, have found that the
a maximum dimension of the aggregates compatible coefficient of friction is constant and independent
with that of the available apparatus. of both contact area and normal load. Therefore,
This technique was used for the first time by it appears that the deformational characteristics
Lowe with reference to materials used for the dams of elastic rock materials should not depend on the
and then recalled by other authors (M. Kaya, 2004) grain size of the material tested (Sevi, 2008).
(Sevi, 2008) in following studies aimed at testing However, other studies conducted by (Kaya et al.
the ballast. 2004; Kaya et al. 1997) on the same material, have
A model sample of perfect spheres (Figure 2), surprisingly led to completely contrasting results
regardless of grain size, would closely duplicate than Sevi and shown in Figure 3:
the contact stresses and void ratio characteristics Many predictive formulas have been developed
of a larger prototype gradation. Thus in mod- to estimate the modulus of the material, taking
els of coarse gradations, where the only differ- into account the dependence on the stress state,
ence between prototype and model sample is the but the hyperbolic model is the most widespread
difference in size of particles, the model sample (Janbu, 1968):
should closely duplicate the behavior of the larger
prototype. ⎛σ ⎞
n

The theoretical basis of the parallel gradation E Kpr ⎜ 3 ⎟

K (2)
scaling modeling scheme is based on the Hertz for- ⎝ pr ⎠

211

TIS2017_Book.indb 211 2/8/2017 8:03:03 PM

 where σ3 = confining stress, pr = reference pressure In this study two different scaled down granulo-
(atmospheric pressure, used to make conversion metric curves have been studied and the hyperbolic
from one system of units to another), K and n are model has been applied to both of them.
dimensionless numbers.
With reference to the Resilient modulus, hyper-
bolic model (Figure 4) gives the same result both 2 TRIAXIAL TEST
the loading e unloading phases and becomes:
2.1 Introduction
n
⎛σ ⎞ The triaxial test is aimed to reproducing a given
Eur K ur pa ⎜ 3 ⎟ (3)
⎝p ⎠ stress state on a sample and later, through the pore
a
pressure measurement, even the evolution of the
effective stresses up to breaking conditions.
Two steps are involved in evaluating the modu-
Observing Figure 5, it is possible to get an idea of
lus parameters K and n. The first is to determine
the trend of the total and residual vertical deforma-
the values of Ei for each test, and the second is to
tions deriving from a triaxial test at constant con-
plot these values against σ3 (on log-log scales) to
fining pressure and with the deviatoric stress that
determine the values of K and n. The value of K is
varies with a sinusoidal law as a function of time.
equal to the value of Ei/pa when σ3/pa = 1.
The total and residual deformation curves, which
The value of n is the slope of the line on this
increase with the number of load applications, show
plot, and may be determined graphically (Duncan
parallel trends after a certain number of loading
et al. 1980).
cycles; therefore, the resilient deformation first var-
ies significantly with the number of cycles and then
it assumes an almost constant value. MR is then
calculated as the ratio of q and the corresponding
resilient deformation εra, stabilized around a single
value (Giannattasio et al. 1989).

2.2 The equipment

Triaxial tests were performed using the equipment
INSTRON 8502, provided by the road laboratory
LaStra of DICEA (Department of Civil, Architec-
tural and Environmental Engineering; University
of Naples Federico II) and showed in Figure 6.

Figure 3. Resilient modulus versus mean particle size,

varying confining pressure.

Figure 4. Loading and unloading modulus. Figure 5. Vertical and horizontal deformations.

212

TIS2017_Book.indb 212 2/8/2017 8:03:04 PM

 Figure 7. Adopted granulometric curve.

Table 4. Converted sieves.

EN Sieve n° Converted sieve Passed percentage

mm mm %

63.00 20.00 100

50.00 15.87 81.53
40.00 12.70 43.06
31.50 10.00 5.28
Figure 6. INSTRON 8502 available at the LaStra. 22.50 7.14 0.53
0.063 0.02 0.48

The press is equipped with a servo-hydraulic Table 5. Granulometria 1.

frame that allows the loading of samples for cyclic
Sieve n° Passed perc. Retained percentage
or static tests and connected to a suitably equipped mm % %
computer with control software.
The diameter D of the sample is 10 cm and the 20.00 100.00 0.00
height H is 20 cm. 16.00 82.10 17.90
12.50 40.28 41.82
2.3 The material 10.00 5.28 35.00
8.00 1.96 3.32
This study analyzes basaltic material taken from 0.063 0.00 1.96
a quarry in Campania (Italy), selecting a granulo-
metric curve compatible with RFI 2012 Specifica-
tions, as showed in Figure 7. Table 6. Granulometria 2.
With reference to the INSTRON 8502, in
a sample with D = 10 cm and H = 20 cm the Sieve n° Passed perc. Retained percentage
mm % %
maximum particle size dmax of the material to be
tested is instantly assessable through the limit 10.00 100.00 0.00
dictated by the specification UNI EN 13286-7: 8.00 82.10 17.90
dmax = D\5 = 20 mm. 6.30 41.49 40.61
The conversion coefficient c is, therefore, 4.00 1.96 39.53
obtained through the ratio between the diameter 2.00 0.51 1.45
of the maximum standard sieve Dsetstand (max) in the 0.063 0.00 0.51
real granulometric curve and dmax:

c = Dsetstand (max)\ dmax = 3.15. Finally, since the previous operation led to the
determination of sieves that do not match to the
The coefficient thus obtained allows the determi- real ones, then a linear interpolation must be car-
nation of the sieve sizes corresponding to the reduced ried out in order to derive the passing percentage
grain size, obtained simply as the ratio between the trough a real sieve and determine the final reduced
size of the sieves of the real curve and c: size curve, called “Granulometria 1”.

213

TIS2017_Book.indb 213 2/8/2017 8:03:05 PM

 Figure 8. Granulometria 1.

Figure 10. Static triaxial tests.

Figure 9. Granulometria 2.

Reasoning similarly and setting the conversion

coefficient c = 6.30 a second granulometric curve,
called “Granulometria 2”, has been obtained.

2.4 Static tests

In order to highlight the rheological model rep-
resentative of the ballast behavior and to make a
clear distinction between linearity and plasticity
field, static triaxial tests were carried out.
Tests were conducted controlling the axial s and
assuming a final axial strain of 20%, in order to
leave unchanged mechanical characteristics and
the resistance of the material; the corresponding
advance speed of the test is 0.0002 cm/s. Figure 11. Elastic Modulus, versus confining pressure.
Tests were performed with different confining
pressure, both for Granulometria 1 and 2, but they
were not always carried to term, because of system- depending on whether the test was performed in
atic arrests of the apparatus. In any case, the post the elastic or plastic range (defined by the output
processing of the outputs showed that in almost all of static tests).
the tests the specimen reached the breakage, intended Since the train speed varies point by point, it is
as the depletion of the resistance (Figure 10). important to study the influence of the load on the
From the analysis of static triaxial tests it is frequency behavior of the ballast; hence, the tests
possible to derive the modulus of elasticity of the were conducted with frequencies 1, 5, 10 and 20 Hz.
ballast, obtained as the slope of the linear portion In this stage we were not interested in studying the
(Figure 11). fatigue behavior of the material, therefore the speci-
mens were subjected to 500 load cycles, cycles from
1 to 5, from 149 to 155 and from 495 to 500 were
2.5 Cyclic tests
acquired, with a resolution of 100 points per wave.
Cyclic tests were carried out controlling the load. Tests in the elastic range were carried out on the
Vertical stress was increased up to σ3 in order to same sample before and after tests in the plastic
achieve the spherical state; then the deviator stress range, with the aim of verifying the influence of
was varied cyclically, varying in a different range the stress state on the linear behavior.

214

TIS2017_Book.indb 214 2/8/2017 8:03:06 PM

 An example of output of the dynamic tests is
given in Figure 12.
MR was evaluated on 151 load wave.

2.6 Results
Outputs were classified as a function of particle
size and were divided in two classes, depending on
whether or not they were the result of tests carried
out in the linearity range.
By resorting to mathematical software Math-
cad, the above-mentioned moduli were finally rep-

Figure 14. Trend of K and n.

Table 8a. M for Granulometria 1, Prova 2.

A1 mean st.dev.

2,13E+05 9,45E+03
by Tests Analysis μ−2σ μ+2σ
f Mr 1,94E+05 2,32E+05
Figure 12. Cyclic test on “Granulometria 1”, elastic Hz kPa test A1
range (q = 0–100 kPa), σ3 = 8psi. 1 2,19E+05 ok
5 2,18E+05 ok
Table 7. K and n from linear regression. 10 2,15E+05 ok
20 1,99E+05 ok
K
Granulometric curve kPa n R2

Gran. 1 Linear range 4.6 × 104 0.378 0.611 resented in a bi-logarithmic diagram (σ3 on x-axis)
Plastic range 9.2 × 104 0.327 0.792 reported in Figure 13, obtaining the intercept val-
Gran. 2 Linear range 1.4 × 104 0.648 0.960 ues of the regression line on the ordinate axis (K)
Plastic range 5.9 × 104 0.421 0.756 and the corresponding slope (n).
Finally, the variability of K and n was correlated
to the mean diameter of the aggregates of both
granulometric curves.
Regressions (see Figure 14) show a growing
trend of K function with increasing average diam-
eter and a decreasing trend of the slope function.

3 CONCLUSIONS

This article proposed an approach to study the

bearing capacity of the ballast.
In fact, the analysis of the literature shows that
the ballast is not completely defined in terms of
stiffness.
The Resilient modulus is the best parameter
to schematize the bearing capacity of the ballast,
because it is a dynamic parameter used to represent
the load bearing capacity of the substrate when this
is stressed by mobile loads (i.e. road or rail loads).
The above parameter was, therefore, studied by
Figure 13. Linear regression (Granulometria 1). means of static and dynamic triaxial tests, varying

215

TIS2017_Book.indb 215 2/8/2017 8:03:07 PM

 Table 8b. M for Granulometria 1, Prova 3. (a and b) and 9 (a and b) and in Figure 15, with
reference to the elastic range.
A1 A2 mean st.dev. Furthermore, the regressions (elastic and plastic
range, cfr Figure 13) allow to correlate, with good
2,08E+05 2,81E+03
approximation, the resilient modulus to the stress
by Tests Analysis μ−2σ μ+2σ
state applied.
f Mr Mr 2,02E+05 2,13E+05
Additional tests should be conducted, intro-
Hz kPa kPa test A1 test A2
ducing at least one other reduced grain size, in
1 2,05E+05 2,10E+05 ok ok
order to increase the sample size and improve the
5 2,08E+05 2,10E+05 ok ok
evaluations.
10 2,07E+05 2,12E+05 ok ok
20 2,04E+05 2,07E+05 ok ok
ACKNOWLEDGMENTS
Table 9a. M for Granulometria 2, Prova 3. This work would not have been possible without
the dedication and constant support of the Las.
A1 A2 mean st.dev.
Tra laboratory technicians Alfredo Caruso, Anna
1,82E+05 2,55E+03 Formisano and Attilio Sannino, who are gratefully
by Tests Analysis μ−2σ μ+2σ acknowledged.
f Mr Mr 1,77E+05 1,87E+05
Hz kPa kPa test A1 test A2
1 1,81E+05 1,85E+05 ok ok REFERENCES
5 1,78E+05 1,84E+05 ok ok
American Railway Engineering and Maintenance-of-
10 1,79E+05 1,83E+05 ok ok
Way Association (AREMA), 2010. Part 2: Ballast.
20 1,82E+05 1,83E+05 ok ok ASTM D 2850–87. (s.d.). Standard Test Method for
Unconsolidated-Undrained Triaxial Compression Test
on Cohesive Soils.
Table 9b. M for Granulometria 2, Prova 4. Australian Standard, 1996. AS2758.7: Aggregate and rock
for engineering purposes—Part 7: Railway Ballast.
A1 A2 mean st.dev. Duncan, J., Byrne, P., Wong, K., & Mabry, P. 1980.
Strenght, stress-strain and bulk modulus parameters for
1,95E+05 3,79E+03 finite element analysis of stresses and movement in soil
by Tests Analysis μ−2σ μ+2σ masses. College of Engineering, Office of Research
f Mr Mr Services, University of California, Berkeley: Report
1,88E+05 2,03E+05
N. UCB/GT/80–01.
Hz kPa kPa test A1 test A2 Giannattasio, P., Caliendo, C., Esposito, L., Festa, B., &
1 1,91E+05 1,98E+05 ok ok Pellecchia, W. 1989. Portanza dei sottofondi. Fondazi-
5 1,91E+05 1,99E+05 ok ok one Politecnica per il mezzogiorno d’Italia; Napoli
10 1,91E+05 1,98E+05 ok ok Hertz, H. 1899. The principles of mechanics presented in a
20 1,93E+05 2,00E+05 ok ok new form. New York: Dover Publications.
Janbu, N. 1968. Soil compressibility as determined by
oedometer and triaxial tests. European Conference
on Soil Mechanics and Foundation Engineering,
(p. 19–25, Vol.I). Wiesbaden.
Kaya, M., Jernigan, R., Runesson, K., & Sture, S. 1997.
Reproducibility and conventional triaxial tests on ballast
material. USA: University of Colorado, Department
of Civil Environmental and Architectural Engineering.
Kaya, M. 2004. A study on the stress-strain behavior of
railroad ballast materials by use of parrallel gradation
technique. Middle East Technical University.
Lichtberger, B. 2010. Manuale del binario. Hamburg:
Eurail Press; ISBN: 978–3-7771–0408–9
RFI, 2011. Technical Standards RFI DTCINC SP IFS
Figure 15. Mr versus f (Hz). 010 A. Roma: Gruppo Ferrovie delle Stato S.p. A.
Sevi, A. 2008. Physical modeling of railroad ballast using
the parallel gradation scaling technique within the cycli-
confining pressure, deviator stress and load frequency, cal triaxial framework. Missouri: Ph.D. thesis
for a scaled down granulometric curve, in order to UNI EN 13450, 2003. Aggregati per massicciate ferroviarie.
adapt the test size to the available triaxial cell size. UNI EN 13286–7, 2006. Miscele non legate e legate con
Results seem to show that the modulus is inde- leganti idraulici—Metodi di prova. Part 7: Prova trias-
pendent from load frequency, as shown in Tables 8 siale ciclica per miscele non legate.

216

TIS2017_Book.indb 216 2/8/2017 8:03:07 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Rolling resistance impact on a road pavement life cycle carbon

footprint analysis

Laura Trupia, Tony Parry, Luis Neves & Davide Lo Presti

Faculty of Engineering, Nottingham Transportation Engineering Centre (NTEC),
University of Nottingham, UK

ABSTRACT: In terms of methodology, there is a continuing discussion related to the introduction of the
impact of road pavement surface properties on rolling resistance in pavement Life Cycle Assessment (LCA).
The aim of this paper is to analyse if the current level of knowledge of this component is sufficient to
be implemented in the pavement LCA framework. The study compares the CO2 emissions, calculated with
two different rolling resistance models in the literature and performs a sensitivity test on the pavement
deterioration rate, for two UK case studies.
The rolling resistance models and the pavement deterioration rate significantly affect the LCA results.
The results show that the methods of modelling and the methodological assumptions need to be trans-
parent in the analysis of the impact of the pavement surface properties on fuel consumption, in order to
be interpreted by decision makers and implemented in the LCA framework.

Keywords: Rolling Resistance, Pavement surface properties, Life Cycle Assessment, Carbon Footprint

1 INTRODUCTION and complex parts of the life cycle that requires

investigation in different areas of scientific litera-
The transport sector and particularly the road ture: rolling resistance, albedo, carbonation, light-
transport sector, is responsible for the emission ing and leachate (Santero and Horvath 2009).
of a large amount of Greenhouse Gases (GHG) Rolling resistance is one of the forces resisting
contributing to global climate change. In 2009, vehicle movement partly due to the energy loss
transport accounted for around a quarter of UK associated with the Pavement-Vehicle Interaction
GHG and road transport was the most signifi- (PVI), due to the physical interaction between pave-
cant source, accounting for 68% of total transport ment and tyre. Much of the rolling resistance can
GHG emissions (UK Department of Energy & be tracked to tyre properties, but it is also affected
Climate Change 2015). by other parameters related to the characteristics of
Life Cycle Assessment (LCA) is a standard- the pavement, such as the pavement surface prop-
ized approach to quantify the direct and indirect erties. The impact of the pavement surface on roll-
environmental impacts associated with a product, ing resistance has been an area of study for many
process or service. In the road transport sector years because of its effect on vehicle fuel consump-
this approach is promising because it can estimate tion and emissions and the opportunity to reduce
the impacts of infrastructure and operations on them with conventional maintenance strategies.
the environment and then assist highway authori- The pavement surface properties that affect rolling
ties, companies and government institutions in resistance include roughness and macrotexture,
evidence-based decision-making (Muench et al. usually represented by parameters International
2014). Over the last years, the LCA approach has Roughness Index (IRI) and Mean Profile Depth
begun to permeate into the planning, construction, (MPD) or Mean Texture Depth (MTD) (Sandberg
operation, and maintenance processes in highway et al. 2011)
asset management (Matute et al. 2014). However, Calculating the impact of pavement surface
the complete introduction of this approach in the properties on the rolling resistance and then on
asset management decision making process is not vehicle fuel consumption is complex, although over
possible yet, due to an incomplete understanding the last years, some studies have been performed to
and uncertainty regarding the impact of some rel- estimate the emissions related to these components
evant phases and components of a road pavement (Hammarström et al. 2012; Wang et al. 2014; Wang
LCA. The use phase is one of the most relevant et al. 2012).

217

TIS2017_Book.indb 217 2/8/2017 8:03:10 PM

 However, the different rolling resistance mod-
els and the incomplete knowledge related to the
influence of specific variables and assumptions on
the results, generates a high level of uncertainty in
their interpretation.
In the UK, not only are there no significant stud-
ies involving national case studies on the impact of
the rolling resistance on the LCA of a pavement, but
there are is also a lack of general pavement deterio-
ration models able to predict the change of uneven- Figure 1. Outline process.
ness and texture depth over time (deterioration rate
of IRI and MPD). The change in these parameters
Table 1. Case study details.
may be different for each lane, since it depends on the
traffic volume and type, on the surfacing type and on Case study A17 A1
the regional climate. While, some empirical models
to describe the deterioration rate of IRI and MPD AADT 15372 51502
have been developed (Lu et al. 2009; Tseng 2012), Road type A—road single Motorway dual
these models are calibrated for specific areas and carriageway carriageway
maintenance treatments and are not applicable to Construction 2009 2009
each case study (in these models, the value of MPD year
tends to increase over time, which is not typical in the Analysis 20 20
UK, where MPD may decrease over time). period
This generates a further level of uncertainty Surface Hot Rolled Thin surfacing
materials Asphalt (HRA)
concerning the introduction of PVI rolling resist-
ance into the LCA approach. In order to introduce
pavement LCA results into the decision making
process of highway authorities, governmental insti- GHG emissions due to PVI rolling resistance,
tutions and companies, it is necessary that meth- using two models from the literature and a sensitiv-
ods of modelling and the assumptions in LCA and ity test performed on the pavement surface proper-
carbon footprint studies are transparent and lead ties deterioration rates.
to consistent results. In this study, only the CO2 emissions are con-
This paper analyses the effect of the pavement sidered for the GHG estimation, since this is the
surface properties on vehicle fuel consumption and biggest component of the vehicle tailpipe CO2e
the impact of some methodological assumptions emissions (over 99.8%) (Wang et al. 2014).
related on the results. To do this, it compares the The study is applied to two case studies, with
CO2 emissions calculated with two different rolling different Annual Average Daily Traffic (AADT)
resistance models and performs a sensitivity test and pavement design (see Table 1).
on the pavement deterioration rate.
The main aim of this study is to evaluate if the 2.1 Calculation of the tailpipe CO2 emissions
existing approaches to incorporating the impact with VTI and UCPRC models
of PVI rolling resistance into the pavement LCA
framework can be done with sufficient confidence The CO2 emissions due to the effect of the pavement
for decision making in the UK. surface properties on vehicle fuel consumption for
In this paper, the authors will extend the anal- two different UK case studies were estimated using:
ysis performed for a previous case study (Trupia the model developed at the University of Califor-
et al. 2016 (in press)) to a second case study, with nia Pavement Research Center (UCPRC, Davis)
a higher volume of traffic, in order to consider the (Wang et al. 2014) and the model developed by the
research questions: Are the rolling resistance mod- Swedish National Road and Transport Research
els ready for implementation in pavement LCA? Institute (VTI), within the European Commission
Can they be applied in the UK? How do pavement project Miriam (Models for rolling resistance In
deterioration and the models used to describe it Road Infrastructure Asset Management systems)
affect the results? (Hammarström et al. 2012). Details related to the
models are provided in the references, but the key
elements are summarised here.
2 METHODOLOGY In the UCPRC model, the vehicle CO2 emis-
sion factors are a continuous function of MPD
Figure 1 shows the outline of the process used and IRI, but the coefficients in the function are
in this research. It involves the calculation of the different for each combination of the categorical

218

TIS2017_Book.indb 218 2/8/2017 8:03:10 PM

 variables (pavement, road and road-access type, obtained from a software VETO, based on a theo-
vehicle type). The CO2 emissions for a specific retical model developed at VTI to calculate fuel
vehicle type can be calculated directly, based on consumption and exhaust emissions from traffic
the analysed pavement segment’s MPD and IRI due to various characteristics of vehicles, roads
values by using equation (1) and multiplying by and driving behavior (Hammarström and Karls-
the vehicle mileage travelled. son 1987; Karlsson et al. 2012)
For both models, only the CO2 emissions
TCOO2 a1 × MPD a2 × IRI IIntercept (1) directly related to the pavement surface properties
(IRI, MPD) are calculated; the other terms of the
where TCO2 is the tailpipe CO2 emission factor, equations are considered equal to zero, since their
the terms a1, a2 and Intercept are the coefficients estimation is not the aim of the study.
derived from the linear regression, depending on Applying the equations previously defined,
surface type and access type, year and vehicle type, it is possible to estimate the total CO2 emissions
IRI is the road roughness (m/km) and MPD is the related to the pavement condition in terms of IRI
macrotexture (mm). The Intercept term identifies and MPD (see Figure 2), namely the total com-
the total CO2 emissions related to the total driv- ponent (total area, representing the total CO2
ing resistance, excluding the impact of the pave- emissions related to the IRI and MPD). This total
ment condition, which is estimated from the other component can be considered as the sum of the
two terms. To develop this equation, the authors basic component (dark grey area, representing the
used two different software models: Highway value of emissions if the IRI and MPD remain
Development and Management Model - version constant over time—no deterioration) and the
4 (HDM-4) (PIARC 2002), an empirical - mecha- deterioration component (light grey area, equal
nistic model software tool to perform cost analysis to the difference between the first two and repre-
for the maintenance and rehabilitation of roads, senting the emissions due to the deterioration of
calibrated for US conditions, and MOVES (Motor the pavement properties during the study analysis
Vehicle Emission Simulator), the US EPA highway period, in terms of IRI and MPD).
vehicle emission model based on national data. Pavement engineering studies tend to focus on
HDM-4 was used to estimate the rolling resistance, the deterioration component because of the oppor-
while MOVES was used to model the vehicle emis- tunity to reduce these PVI emissions, taking direct
sions as a function of the rolling resistance action on the road surface condition, through
The VTI model includes a general rolling resist- appropriate maintenance. Obtaining pavement
ance model (equation (2)) to estimate the contri- condition improvements is in general more rapid
bution of the rolling resistance to the total driving and easy than other approaches to reduce roll-
resistance and a fuel consumption model (equa- ing resistance emissions that involve technology
tion (3)) to calculate the vehicle fuel consump- improvements or traffic reduction.
tion (Hammarström et al. 2012). Once the fuel In this paper, all the components were esti-
consumption related to a specific type of vehicle mated, since they can provide a better understand-
was estimated using this model, it was converted ing of the behaviour of the two rolling resistance
to CO2 emissions, assuming the conversion proc- models.
ess proposed by International Carbon Bank &
Exchange (ICBE) (2010).

Fr m1 × g ×
( IRI V MPD )
(2)
FCSS = 0.286 ×
⎛ ⎛ 1.209 + 0.000481 × IRI V + 0.394 × MPD⎞ 1.163 ⎞
⎜ ⎜ + 0.000667 × V 2 + 0.0000807
0 × ADC × V 2 ⎟ ⎟
⎜⎝ ⎝ − 0.00611 × RF + 0.000297 × RF 2 ⎠ ⎟⎠
× V 0.056
(3)
The rolling resistance model developed by VTI
is mainly based on empirical data from coastdown Figure 2. Total CO2 emissions, divided into basic com-
measurements in Sweden; the fuel consump- ponent (dark grey area) and deterioration component
tion model has been calibrated based on results (light grey area) (from Trupia et al. 2016 (in press)).

219

TIS2017_Book.indb 219 2/8/2017 8:03:10 PM

 Table 2. Pavement deterioration rate, in terms of IRI Table 3. CO2 emissions due to pavement surface condi-
and MPD, during the analysis period. tion (average deterioration).

MPD IRI A17

Scenario mm m/km
Emission of CO2 (ton)
Average deterioration 1.8–0.8 1.0–2.3
A17 Worst deterioration 1.5 1.0–5.0 Model Basic Deterioration Total
No deterioration 1.8 1.0
Average deterioration 1.6–0.6 1.0–2.3 VTI 10272 −634 9638
A1 Worst deterioration 1.3 1.0–5.0 UCPRC 1170 217 1387
No deterioration 1.5 1.0
A1

Emission of CO2 (ton)

2.2 Sensitivity test on pavement deterioration rate
Model Basic Deterioration Total
The use of these models, correlating pavement sur-
face properties to vehicle fuel consumption and VTI 109344 −4205 105139
emissions, requires as an input parameter, the esti- UCPRC 18058 4586 22645
mation of pavement condition deterioration rate
with time (in terms of IRI and MPD).
As mentioned in the introduction, in the UK components, with the exception of the deteriora-
there are no models able to predict the deteriora- tion component, where the two models provide a
tion rate of these parameters over the years. For negative term.
this reason, the time progression of IRI and MPD The difference in results obtained for the basic
on the assessed road segments over the analysis component shows an interesting aspect related to
period (20 years) is generated according to litera- the comparison of the two models; regardless of
ture data for specific maintenance strategies (Aavik the deterioration in the IRI and MPD over the
et al. 2013; Jacobs 1982; Wang et al. 2014). analysis period, the two models return consider-
In order to take into account the uncertainty ably different results (10272 ton against 1170 ton
related to these parameters and the range of poten- in the A17 and 109344 ton against 18058 ton in
tial impact during the use phase, different scenarios the A1). This difference reflects the substantially
of deterioration of IRI and MPD are considered different estimated total components. The differ-
for the two case studies (see Table 2) and compared ent validation of the two models, together with
in a sensitivity analysis. the different approaches used, can be considered
The average deterioration values include an initial the main reason for this significant difference
and final condition value and a linear change with in the results: indeed, the models were calibrated
time is assumed. This is also the case for the IRI val- for different countries with different input data, in
ues in the worst deterioration scenario. The MPD in terms of weather, vehicles, and roads.
the worst deterioration scenario and the MPD and Particularly significant are the results obtained
IRI for the no deterioration scenario are held con- for the deterioration component, which take into
stant. Note that in the average deterioration scenario, account the evolution of the pavement surface prop-
the MPD falls with time from a high initial value; this erties. The deterioration term for the VTI model
is common in the UK were high MPD values are is negative for both case studies and this means
specified for new surfacing to assist in provision of that overall the deterioration in pavement surface
high-speed wet skidding resistance. properties produces a reduction of the vehicle tail-
pipe CO2 emissions over the years, rather than an
increase as was expected. The negative term related
3 RESULTS to the deterioration component is due to the differ-
ent impact given to the IRI and MPD terms by the
Table 3 summaries the results obtained for the two two models (see Figure 3 and Figure 4).
case studies, using the average deterioration rate The VTI model assigns to the MPD term a
scenario (see Table 2). As explained above, the dete- greater impact on the rolling resistance and on the
rioration component is calculated as the difference emission estimate than for IRI (even at high speed,
between the total component and the basic compo- which increases the impact of the IRI); the opposite
nent (represented by the No deterioration scenario). consideration is true for the UCPRC model, where
For both case studies, the two models pro- the IRI term has a larger impact. The significance
vide substantially different results; this difference of this different behaviour becomes particularly rel-
amounts to one order of magnitude for all the evant for pavement surfaces where the IRI tends to

220

TIS2017_Book.indb 220 2/8/2017 8:03:11 PM

 Table 4. Sensitivity test results for the A17 case study.

A17

Emission of CO2 (tonne)

VTI UCPRC

Scenario B* D** T*** B D T

A 10272 −634 9638 1170 217 1387

W 10272 500 10772 1170 1134 2304
No 10272 0 10272 1170 0 1170
Figure 3. Impact of IRI and MPD in the VTI (above)
% compared to the Average scenario
and UCPRC (below) models for the A17 case study.
A 100 100 100 100 100 100
W 100 −79 112 100 522 166
No 100 0 107 100 0 84

* Basic component
** Deterioration component
*** Total component
Table 5. Sensitivity test results for the A1 case study.

A1

Emission of CO2 (tonne)

Figure 4. Impact of IRI and MPD in the VTI (above) VTI UCPRC
and UCPRC (below) models for the A1 case study.
Scenario B D T B D T
increase and the MPD tends to decrease, as in these
A 109344 −4205 105139 18058 4586 22645
case studies. The VTI model gives a negative value
W 109344 4716 114059 18058 19634 37693
for the deterioration component, because the MPD
No 109344 0 109344 18058 0 18058
term decreases faster than the IRI term increases.
Therefore, the pavement surface properties deterio- % compared to the Average scenario
ration and the models used to describe them have
a significant impact on the emissions results. This A 100 100 100 100 100 100
consideration is confirmed by the sensitivity test W 100 −112 108 100 428 166
performed on this input variable that shows how No 100 0 104 100 0 80
the IRI and MPD deterioration rate can change
the rolling resistance results in a pavement LCA see nent increases over time, so the absence of dete-
(Table 4 and Table 5). The two tables provide the rioration minimizes the total emissions.
results of the sensitivity test obtained from the two − In the VTI model, the deterioration component,
models in terms of basic component, deterioration under the average condition of pavement deteri-
component and total component (see Figure 2), oration, tends to decrease, producing an overall
taking into account three different scenarios of reduction in the calculated emissions. This effect
deterioration in IRI and MPD during the analysis levels off under the “worst deterioration” pave-
period, Average (A), Worst (W) and No deteriora- ment condition, when the IRI effect is larger
tion (No) as defined in Table 2. than the MPD effect.
The results obtained performing the sensitivity − This means that in both models, the CO2 emis-
test for the two case studies show: sions are significantly higher in the case of the
worst pavement deterioration scenario.
− For both models, the range of potential impact
− The results shows the CO2 emissions due to the
due to the PVI is wide;
pavement roughness are very sensitive to the
− The lowest emissions in the two models occur
pavement surface deterioration over time.
under different pavement deterioration rate sce-
narios (no deterioration in the UCPRC model It is clear that the choice of model used to esti-
and average deterioration in the VTI model). mate the PVI rolling resistance CO2 emissions and
In the UCPRC model the deterioration compo- the surface deterioration model chosen both have a

221

TIS2017_Book.indb 221 2/8/2017 8:03:11 PM

 large influence on the results. As mentioned above, should be used carefully and the results interpreted
the differences in the models are due to the coun- with caution, especially considering the different
try where they were developed and hence validated weight given to the IRI and MPD terms by the two
and the modelling approach taken. models.
In the development of these models, different The sensitivity test performed confirms that the
variables come into play affecting strongly the results are sensitive to the pavement deterioration
results obtained. In particular, some methodologi- rate of IRI and MPD.
cal choices (such as rolling resistance measurement In the UK, currently there are no validated mod-
methods, road surface measures, direct or indirect els able to predict the relationship between rolling
estimation of emissions) and site specific elements resistance and pavement surface properties and the
(weather, vehicle types and technology, type of use of the models in the literature could lead to
roads, pavement design models and deterioration) unreliable results.
can play a relevant role, producing a rolling resist- In addition, there are no models to predict the
ance and fuel consumption model that is not suit- deterioration of roughness and texture depth over
able for every geographical location. time depending on maintenance treatments, traffic
The two models assessed in this research are very volume and type, surface properties and materials.
different, both in terms of methodological approach This gap has to be filled before introducing the use
and in terms of site specific elements. The UCPRC phase in the pavement LCA framework.
models was developed in California, by using two As a result of this investigation, the authors
software tools (HDM-4 – calibrated for US condi- believe that overall, further research is necessary
tions—and MOVES—a US EPA highway vehicle to develop standardized procedures for PVI rolling
emission model based on national data) and imple- resistance emission estimates, so to obtain compa-
menting IRI and MPD data related to US roads. rable and reliable pavement LCA results beneficial
The VTI rolling resistance model is based on in the decision making process. Furthermore, it
empirical data from coastdown measurements in is recommended to any national or regional road
Sweden and the fuel consumption model was cali- authority that wants to introduce the use phase
brated using calculated values from VETO, a theo- into the pavement LCA framework, to at least
retical model to estimate the fuel consumption and carry out an exhaustive calibration and validation
exhaust emissions of vehicles. Different climates, of the existing models presented in this work, if
type of roads, pavement deterioration processes not developing their own models to accurately pre-
and models, traffic composition and vehicle tech- dict evolution of pavement surface properties of
nology characterize California and Sweden affect- their road networks over time.
ing the models developed and the results produced.
The two models consider the impact of the pave-
ment surface properties, IRI and MPD, in different ACKNOWLEDGMENTS
ways. In the UCPRC model the IRI has a larger
impact on the rolling resistance than the MPD This study was funded from an Engineering and
and the opposite is true for the VTI model. This Physical Sciences Research Council (EPSRC) Doc-
difference is particularly significant in this case toral Training Grant (DTG) - Faculty of Engi-
study, where the MPD falls over time, producing neering. The authors would like to thank staff at
opposite results when the two models are used; in Lincolnshire County Council and Highways Eng-
the UCPRC model, the deterioration component land for providing data for the case studies.
is positive, since the impact of the increase in IRI
is larger than that due to the reduction in MPD;
while for the VTI model the deterioration compo- REFERENCES
nent is negative. Therefore, the pavement condition
deterioration over time has a strong impact on the Aavik A, Kaal T, Jentson M Use Of Pavement Surface
rolling resistance, significantly affecting the results. Texture Characteristics Measurement Results In Esto-
This is confirmed by the sensitivity test performed nia. In: XXVIII International Baltic Road Conference,
on the IRI and MPD deterioration rate that Vilnius, Lithuania 26–28 August, 2013.
showed that the CO2 emissions due to PVI rolling Hammarström U, Eriksson J, Karlsson R, Yahya M-R
resistance are very sensitive to this factor. (2012) Rolling resistance model, fuel consumption
model and the traffic energy saving potential from
changed road surface conditions, VTI Rapport
748 A.
4 CONCLUSIONS Hammarström U, Karlsson B (1987) VETO: Ett dator-
program för beräkning av transportkostnader som
The significantly different results obtained for both funktion av vägstandard, VTI meddelande, ISSN
case studies show that rolling resistance models 0347–6049.

222

TIS2017_Book.indb 222 2/8/2017 8:03:11 PM

 Jacobs F (1982) M40 High Wycombe By-pass: Results of Santero NJ, Horvath A (2009) Global warming poten-
a Bituminous Surface-texture Experiment. Transport tial of pavements. Environmental Research Letters
and Road Research Laboratory, UK. 4:034011.
Karlsson R, Carlson A, Dolk E (2012) VTI - Energy Use Trupia L, Parry T, Neves L, Lo Presti D (2016 (in press))
Generated by Traffic and Pavement Maintenance: Rolling Resistance Contribution To A Road Pavement
Decision Support for Optimization of Low Rolling Life Cycle Carbon Footprint Analysis The Interna-
Resistance Maintenance Treatments, MIRIAM SP3. tional Journal of Life Cycle Assessment doi:10.1007/
Lu Q, Kohler ER, Harvey JT, Ongel A (2009) Investi- s11367–016–1203–9.
gation of Noise and Durability Performance Trends Tseng E (2012) The construction of pavement per-
for Asphaltic Pavement Surface Types: Three-Year formance models for the California Department of
Results, UCPRC-RR-2009–01. Transportation new pavement management system.
Matute JM, Chester M, Eisenstein W, Pincetl S Life- Department of Civil and Environmental Engineering -
Cycle Assessment for Transportation Decision Mak- University of California Davis.
ing. In: Transportation Research Board 93rd Annual UK Department of Energy & Climate Change (2015)
Meeting, 2014. vol 14–1287. 2014 UK Greenhouse Gas Emissions, Provisional
Muench ST, Lin YY, Katara S, Armstrong A (2014) Road- Figures.
print: Practical Pavement Life Cycle Assessment (LCA) Wang T, Harvey J, Kendall A (2014) Reducing green-
Using Generally Available Data. Paper presented at the house gas emissions through strategic management of
International Symposium on Pavement LCA 2014, highway pavement roughness Environmental Research
Davis, California, USA, October 14–16 2014. Letters 9:034007.
PIARC (2002) Overview of HDM−4. The Highway Wang T, Lee I-S, Harvey J, Kendall A, Lee EB, Kim C
Development and Management Series Collection. (2012) UCPRC Life Cycle Assessment Methodology
Sandberg U, Bergiers A, Ejsmont JA, Goubert L, Karls- and Initial Case Studies for Energy Consumption and
son R, Zöller M (2011) Road surface influence on tyre/ GHG Emissions for Pavement Preservation Treat-
road rolling resistance. Report MIRIAM SP1_04. ments with Different Rolling Resistance.

223

TIS2017_Book.indb 223 2/8/2017 8:03:11 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Intermediate stage traffic technical solution of prince Branimir

Street in Zagreb

M. Šimun
University of Applied Sciences Zagreb, Croatia

S. Mihalinac
ZG-projekt Ltd., Zagreb, Croatia

K. Rehlicki
Elipsa-S.Z. Ltd., Zagreb, Croatia

ABSTRACT: Traffic network of each city is determined by Urban Master Plan (UMP) which reflects in
traffic corridors of specified width. The urban master plan is plan of the local level and in our case covers
220 km2 for the city of Zagreb. The plan gives indicators for the construction, renovation and protection
of space in the area of Plans scope and to ensure spatial plan preconditions for realization of traffic and
infrastructure projects. On the area of a large city is a complex interdependence of traffic infrastructure.
The quality of traffic infrastructure is manifested as accessibility and mobility of space and advancement
of new traffic corridors as a basis for a sustainable development. In this paper, the traffic technical solu-
tion of prince Branimir Street in Zagreb, road section from Zavrtnica to Vjekoslav Heinzel Street, has
been elaborated. Construction of this road section was initiated in 2012 through the I. phase of construc-
tion but only part of the width of the planned traffic corridor. As a result of that property-right issues
were partially solved. A fully built, in a full corridor width, was planned for II. phase of construction. In
the meantime the property-right issues were solved but still not in a complete width of the planned cor-
ridor. Because of that, it has been approached to preparation of technical solution which would keep the
elements constructed in a I. phase, and built new ones in intermediate stage which would also be incorpo-
rated in the II. phase of construction. In intermediate stage, elements which improve traffic conditions,
should be constructed, such as enhancement of traffic capacity and satisfying level of service. In the paper
were analyzed all the components and traffic technical solutions of intermediate stage.

1 INTRODUCTION with satisfactory level of service. In other parts

of the city is very low capacity of road network,
1.1 General reduced availability of the city center and there is
a problem of structure and organization of traffic
Traffic network of the city of Zagreb is determined
flows. Also, the development of transport network
by Urban Master Plan (UMP)—Urban master
doesn’t follow the development of the city. Planned
plan of the city of Zagreb (2007) which reflects in
traffic corridors in other parts of the city are char-
traffic corridors of specified width for road catego-
acterized by no kind of construction (not started
rized as urban highway, city avenue, the main street,
any construction in the planned corridor) or kind
city streets and corridors of public transportation
of contruction only in the part of the planned
areas. The plan gives indicators for the construc-
width or discontinuity of development corridors
tion, renovation, modernize and upgrade existing
with the level of development of planned width or
streets, solve stationary traffic and improve bicycle
in full width of planned corridor.
traffic. Transport situation in the city of Zagreb is a
The cause of this condition partly lies in the
direct consequence of Urban Master Plan (UMP)
fact that it is an urban area where there are built
planned traffic corridors. Figures 1 and 2 showing
residential business and commercial objects and
Urban Master Plan (UMP) of the city of Zagreb—
any construction of traffic corridors is actually a
purpose of space and road network. Parts of the
“puncture” through the built area with unavoidable
city in which they are built traffic corridors into full
demolition of existing buildings. Prince Branimir
width, from the point of road capacity, function

225

TIS2017_Book.indb 225 2/8/2017 8:03:13 PM

 Figure 1. Urban master plan of the city of Zagreb—purpose of space.

Figure 2. Urban master plan of the city of Zagreb—traffic network.

226

TIS2017_Book.indb 226 2/8/2017 8:03:13 PM

 Figure 3. Grip of the project.

Street, as one of the main traffic corridors from the Between 2005 and 2008 it has been constructed
center to the eastern part of the city is an example a part from Marin Drzic Avenue to Zavrtnica in a
of discontinuity corridor, where for many years the section with dual two-lane carriageway each with
corridor was interrupted on the part of the Avenue two traffic lanes and pedestrian-bicycle way.
Marin Drzic to the Vjekoslav Heinzel Street and Continuity of traffic corridor couldn’t be real-
all because of the residential business and com- ized because there hasn’t been built a part from
mercial objects within the plan of the early traffic Zavrtnica to Vjekoslav Heinzel Street in a length
corridor. of 350 meters so that this part has been analysed
Few years back on the part of Zavrtnica to the in this paper.
Vjekoslav Heinzel Street was built only one part In Urban Master Plan (UMP), in range of a grip,
of the planned width of the corridor as a kind of prince Branimir Street has been planned in a width
“junction” of the old and new Branimir Street. The of 50 meters. On the south, it borders on railway
construction of “junction” has been made the con- corridor and Urban Master Plan (UMP) Heinzel-
tinuity of traffic corridor but due to insufficient Radnicka-railway line. Figure 3 showing grip of the
road capacity as a result of partially development project. On the north side corridor borders on Urban
the junction is only a “bottleneck”. Master Plan (UMP) area bounded with streets Ban-
The quality of traffic infrastructure is manifested javciceva-Heinzel-prince Branimir-Zavrtnica.
as accessibility and mobility of space and advance-
ment of new traffic corridors as a basic for sus-
tainable development—Book of regulations (NN 2 DEVELOPMENT OF PHASES
110/2001) & Book of regulations (NN 34/2005). OF CONSTRUCTION

The building of the section initiated in 2012

1.2 Position and role in city traffic network
through the I. phase of construction within par-
Prince Branimir Street has been planned by the tial width of the planned street corridor as a result
Urban Master Plan (UMP)—Urban master plan of of unsolved property-right issues—Decision 2006
the city of Zagreb (2007) to be a traffic corridor of (12). The building of entire width corridor was
main city street from main railway station to town planned through the II. phase of construction.
district Sesvete and represents one of the main traffic The I. phase of construction was finished in 2012.
corridors from the center to the east side of the town. In years after achieving the I. phase of construc-
Figure 2. showing prince Branimir Street in city road tion, additional property-right issues were solved.
network. For years there were only two separated But since they were not solved entirely, it was not
parts: part from main railway station to Marin Drzic possible to build the planned II. phase of construc-
Avenue and part from Vjekoslav Heinzel Street to tion. That situation led to finding a new technical
Dubrava Street in city district Sesvete. solution with the aim to improve the conditions

227

TIS2017_Book.indb 227 2/8/2017 8:03:18 PM

 of ever-increasing traffic: increasing capacity of construction is on intersection with Heinzelova
and achieving satisfying level of service. All com- Street. Design of intermediate stage not allow
ponents of road design build in the I. phase and direct left turn on intersection with Heinzelova
planned in the II. phase of construction were ana- Street. Left turn is achievable indirectly on inter-
lyzed and resulted in a traffic and road design of section with Zavrtnica Street. Figure 5 showing
an intermediate stage of construction. All design technical solution situation of intermediate stage
components built in the I. phase of construction of construction.
are preserved and upgraded with additional com- The II. phase of construction planned to be acti-
ponents, entirely compatible with the planned II. vated as soon as property-right issues are solved.
phase of construction—Decision 2006 (16). Figure 6 showing technical solution of planned the
II. phase of construction.
Each component of road design built in the I.
3 TECHNICAL SOLUTION phase and planned in the II. phase of construc-
tion were analyzed and redesigned, resulting in
Basic design of the I. phase of construction was sin- new combined technical solution of intermediate
gle lane carriageway with no additional traffic lanes stage—Rehlicki (2016). Cross-sections of a corri-
at Zavrtnica and Heinzelova intersections (Elipsa- dor defined through the I. phase, the II. phase and
S.Z. Ltd., 2012). Footpath in minimum width was intermediate stage are shown in Figure 7.
designed only on south side of carriageway. At time
when designed, it was not sufficient for existing traf-
3.1 Horizontal alignment
fic volume, but was only possible due to unsolved
property-right issues. The I. phase resulted in con- In the I. phase of construction was built one sin-
nection of existing unlinked segments of prince gle lane carriageway, partially on south and north
Branimir Street but with congestions on Heinzelova side of planned corridor. Horizontal alignment is
Street intersection. Figure 4 showing technical solu- defined with two curves 80 meters in radius con-
tion situation of the I. phase of construction. necting three straights. Design speed of 50 km/h is
In the II. phase of construction was planned determined by curves radius.
dual-two lane carriageway with additional traf- In the II. phase of construction are planned two
fic lanes for left and right turns at Zavrtnica and duo lane carriageways, one on south and one on
Heinzelova intersections—Elipsa-S.Z. Ltd. (2014). north side of planned corridor. Horizontal align-
Footpath in minimum width was designed on south ment is defined with one axis. Design speed is
side of carriageway. On north side of carriageways 50 km/h and determined by traffic regulation.
footpath with cycle lanes was designed. Intermediate stage, according to drive direction,
Intermediate stage is presented with combined has two horizontal alignments. For drive direction
carriageways from the I. and the II. phase of con- west-east horizontal alignment completely match
struction. Intermediate stage in fully use carriage- the I. phase of construction. For drive direction
way built in the I. phase of construction and build east-west horizontal alignment match north car-
north side corridor carriageway as planned in the riageway horizontal alignment of the II. phase of
II. phase but with modification on east side neces- construction.
sary to form intersection. Main difference in design Traffic corridor is defined by an unique axis
of intermediate stage and design of the II. phase set in central median on which are based all the

Figure 4. The I. phase of construction.

228

TIS2017_Book.indb 228 2/8/2017 8:03:19 PM

 Figure 5. Intermediate stage of construction.

Figure 6. Planned the II. phase of construction.

Figure 7. Cross-sections of a corridor.

229

TIS2017_Book.indb 229 2/8/2017 8:03:20 PM

 other ele-ments of cross-section across the width 3.3 Pavement structure
of the planned corridor—Korlaet (1995). Because
The same pavement structure as defined in the I.
of phase building in the I. phase and traffic tech-
phase is retained in the planned II. phase of con-
nical solution, it was necessary to define separate
struction. As shown in Figure 10 total thickness
axis that will best define the elements of the cross-
of existing pavement structure is 61 cm. Exten-
section for the purpose of phase building. In a
sion of existing carriageway while maintaining the
transport-functional sense, considering directions
same layer thickness, cross fall and subgrade fall
of the vehicle, traffic technical solution consists
requires, on extended side, pavement structure in
of two axis, different for every vehicle di-rection.
total thickness of 51 cm. The frost-resistant pave-
Figure 9 showing defined axis.
ment structure requires total thickness (depth) of
57 cm. Due to keeping existing pavement layer
3.2 Vertical alignment thickness and insuring frost-resistant pavement
structure, it was necessary to reduce subbase layer
Vertical alignments of the I. phase, the II. phase
thickness from 30 cm down to 26 cm and reduce
and intermediate stage are shown in Figure 8. Ver-
subgrade fall from 4% down to 3%.
tical alignment deviation in intermediate stage in
Layer thickness of pavement and subbase, based
regard to the II. phase of construction is result
on layer thickness and subgrade fall, designed in the
of keeping existing carriageway on north side of
I. phase of construction are recalculated and rede-
corridor, build in the I. phase of construction, and
fined for intermediate stage of construction—Babic
super elevation, changing grade from existing car-
(1997). Layer thickness and subgrade fall of interme-
riageway to carriageway planned through the II.
diate stage pavement structure is shown in Figure 10.
phase of construction.

Figure 8. Vertical alignments.

Figure 9. Defined axis.

230

TIS2017_Book.indb 230 2/8/2017 8:03:22 PM

 Figure 10. Intermediate stage pavement structure.

Table 1. Comparative overview carriageway upgrade-

ability through phases of construction. time period, require a coordinate work of large
number of city and municipal services. It includes
First Intermediate second planning and harmonization of activities, solving
phase stage phase property-right issues and insurance of financial
resources, both in the design stage and building
Carriageway stage. A transportation system is characterized
built, planned [m2] by incoherence of urban areas, low road capacity,
2304 4534 6820 reduced availability of town center, adverse traf-
to remove in second phase [m2] fic effect on environment and increased degree of
1104 81 - motorization. Systematic construction of main
realized in year 2012 - 2016 traffic corridors should be a priority and the only
cost in all* 6.8 - 9.0
measure of sustainability of existing traffic net-
[million Kn**]
work. The level of construction should correspond
* Realized cost in all include: carriageway, footpath, traffic to development stage of urban area. This is achiev-
lights, traffic signs, road markings, sewer, lighting, telecom- able only through multi phase of construction. The
munication infrastructure, landscaping, urban equipment. development of the corridor is achieving through
** 1 Kn = 0.133€. construction phasing in a way that next phase is a
suffix to a previous phase. It is unacceptable to tear
down allready built part of infrastructure, except
3.4 Validity and sustainability of
in case when there was a long period between two
intermediate stage
phases. Traffic system is inimitable so it is neces-
Validity and sustainability of the intermedi- sary to continuosly invest in it’s development.
ate stage technical solution is shown in Table 1. Today’s low road capacity of traffic infrastruc-
Almost 50% of carriageway built in the I. phase of ture and lowered availability of town center is
construction should be removed to be upgraded to caused not only by inappropriate building but also
the II. phase technical solution. Unlike the I. phase by absence of systematic traffic control. Only com-
of construction, less than 2% of carriageway built bination of measures based on urban planning can
in intermediate stage should be removed to be able contribute to growth of motorization degree and
to upgrade to technical solution of the II. phase of increased road capacity.
construction. As shown in this case, originally planned con-
struction through two phases, due to the impos-
sibility of solving property-right and legal issues
4 CONCLUSION in the planned time frame, led to new phase of
construction—intermediate stage. Each phase cor-
Building of traffic corridors in a full width repre- responds to traffic needs but taking into account
sent a complex project, which extends through long the spatial possibilities at its time of construction.

231

TIS2017_Book.indb 231 2/8/2017 8:03:24 PM

 In May 2016 property-right issues were solved Decision 2006 (16) http://www1.zagreb.hr/slglasnik/
entirely. After four years in use, 48% of carriage- index.html#/home.
way built in the I. phase was removed and in July Elipsa-S.Z: Ltd. 2012, Prince Branimir Street project—
2016 was built full width corridor according to the I. phase of construction.
Elipsa-S.Z: Ltd. 2014, Prince Branimir Street project—
planned the II. phase of construction. So there was the II. phase of construction.
no need for application of intermediate stage. This Korlaet, Z. 1995. An introduction to the design and con-
paper was not intended to cover feasibility analysis structions of roads, Zagreb, Faculty of civil engineer-
of construction for each phase but to show that ing University of Zagreb.
for each stage of development there is appropriate Rehlicki, K. 2016, Technical solutions of prince Bran-
technical solution. imir Street in Zagreb, Diploma thesis, University of
Applied Sciences Zagreb.
Urban master plan (UMP) of the city of Zagreb, deci-
REFERENCES sion 2007 (16) http://www1.zagreb.hr/zagreb/slglas-
nik.nsf/7ffe63e8e69827b5c1257e1900276647/fc7bc
5b1e3f36f97c12573a90032dc3d/$FILE/Odluka%20
Babic. B. 1997. Design of pavement structures, Zagreb,
o%20dono%C5%A1enju%20Generalnoga%20
Croatian Society of Civil Engineers.
urbanisti%C4%8Dkog%20plana%20grada%20
Book of regulations, NN 110/2001 www.nn.hr.
Zagreba.pdf.
Book of regulations, NN 34/2005 www.nn.hr.
Decision 2006 (12) http://www1.zagreb.hr/slglasnik/.
html#/home.

232

TIS2017_Book.indb 232 2/8/2017 8:03:24 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Research and analysis of dolomite aggregate for resistance

to fragmentation

L. Šneideraitienė & D. Žilionienė

Department of Roads, Vilnius Gediminas Technical University, Vilnius, Lithuania

ABSTRACT: Mechanical and physical properties of aggregates, used for producing asphalt mixture,
influence the quality indices of pavement structure, i.e. functionality, reliability and durability. Investiga-
tion of various mechanical and physical properties of numerous aggregates by using test methods of EU
standards Lithuania started in 2004. Requirements for resistance to fragmentation of aggregate, the tech-
nical specifications of mineral materials in Lithuania and in Latvia have been determined. The resistance
to fragmentation of Lithuanian high quality dolomite aggregate meets Lithuanian and Latvian require-
ments according to TRA ASPHALT 08 Technical Specifications for Road Asphalt Mixtures and Ceļu
SPECIFICATIONS 2015 for course mineral materials, bituminous mixtures and surface treatments.

1 INTRODUCTION Bulevičius et al. (2011) indicate that since

2004 Lithuania began to investigate mineral physi-
According to Bulevičius et al. (2011) the principal cal and mechanical indicators by EN standard test
cause of defects in road pavement is the impact methods. Many tests of resistance to fragmentation
of heavy vehicles which influence functionality done with numerous minerals from different man-
of road pavement, driving comfort and traffic ufacturers, e.g. EN 1097-2:2010 Tests for Mechani-
safety. In order to improve the properties of road cal and Physical Properties of Aggregates—Part
pavement and the quality of traffic the scientists 2: Methods for the Determination of Resistance to
of Lithuania, Latvia and various countries carry Fragmentation impact resistance (SZ) and the coef-
out investigations of structural pavement layers, ficient of Los Angeles (LA) indicator values pro-
analyse the effect of their properties on pavement vided. Regulations for resistance to fragmentation
performance, mechanical and physical properties for Lithuanian rock determined too (Figure 1).
of asphalt mixtures used for structural pavement Rock processing takes place according to
layers, etc. Aggregates (crushed stone, dolomite, technological schemes for the manufacture of
gravel) differ in their size, shape and working con- bulk products used in the production of other
ditions of the material that depend on the size and construction products (Fook et al. 1988; Labus
nature of loads, working temperature, environmen- et al.2016).
tal aggressiveness, etc. Mechanical and physical
properties of aggregate, used for producing asphalt
mixture, influence the quality indices of pavement
structure, i.e. functionality, reliability and dura-
bility. Aggregate suitability to asphalt mixtures is
dependent on its mechanical and physical proper-
ties. Aggregate properties determine the mode of
production of asphalt mixture, the thickness of
structural pavement layers and other peculiarities
of road pavement structure. In each case, selection
of aggregate shall be economically justified. The
selected aggregate shall be cheap and easily avail-
able. When a stronger aggregate is used, the service
life of road pavement structure is longer, the struc-
ture is more reliable, the individual structural pave-
ment layers are thinner, the costs of aggregate and
other materials are lower, etc. Often the aggregate
is selected not because of its better properties but Figure 1. Category value of rock resistance to fragmen-
because it is cheaper. tation by TRA MIN 07 (Lithuania).

233

TIS2017_Book.indb 233 2/8/2017 8:03:27 PM

 Dolomite is one of the most available sedimen- In addition, requirements for mineral materials
tary rocks in the territory of Lithuania and Latvia indicated in TRA MIN 07 Technical Specifications
(Haritonovs et al. 2014, Skrinskas et al. 2010). for Road Mineral Materials when resistance to frag-
Dolomite quarries in Lithuania contain hundreds mentation of aggregates is determined in accordance
million tons of this material. Either, high quality with LST EN 1097-2 presented in TRA ASPHALT
dolomite produced applying special extraction 08 by asphalt quality and type.
technology, and the mechanical properties of this
material are similar to granite (Šernas at al. 2016).
Haritonovs et al. (2015) specify a reason to 2 TEST METHODS AND REQUIREMENTS
import dolomite aggregates from other countries
because Latvian dolomite lacks the mechanical One of the most accurate tests to determine coarse
strength for motorways according to the Latvian aggregate strength is the test of resistance to frag-
Road. Latvian dolomite of its low quality, mainly mentation in accordance with LST EN 1097-2:2010
LA value, cannot use for average and high volume Tests for Mechanical and Physical Properties of
roads (Haritonovs, Tihonovs 2014; Haritonovs Aggregates − Part 2: Methods for the Determination
et al. 2015). Therefore, mostly importable mag- of Resistance to Fragmentation. The main method
matic rocks (granite, diabase, gabbro, basalt) or of this test lies in the determination of resistance
dolomites are used and makes asphalt concrete to fragmentation by use in Latvia of the Los Ange-
expensive (Haritonovs et al. 2014). les Test method. However, the alternative Impact
According Šernas et al. (2016), it is a cost-effec- Test method is the most frequently used in Lithua-
tive to use local road construction materials (on nia. For this reason, the specimens tested by both
that ground and dolomite) because the transporta- methods to determine the strength of Lithuanian
tion increases price of ready to use materials about dolomite aggregates and to discover a relationship
30 percent and the mechanical and physical prop- between these two tests. In consequence, LA and
erties of the materials are practically comparable SZ values show the same property of tested mate-
to these properties of importable mineral materi- rial but the methods of tests are different.
als. Considering these facts and to avoid ruts on In the Los Angeles test method of coarse
road pavements, there is very important question aggregate, 10/14 mm size fraction (5000 ± 5)
of using local high quality dolomite aggregates in g specimen together with ten steel balls, each
producing asphalt mixes, which properties would Ø 45–49 mm in diameter and weighing in total
be similar or better comparing to asphalt concrete 4690–4860 g, rotated for 500 revolutions at a con-
mixes with granite aggregates. Šernas et al. (2016) stant speed of 31–33 min-1 in a closed drum. In
used empirical and performance based tests for the Impact Test times by a weight falling from
designed asphalt concrete mixtures with high qual- 370 mm height. After the tests, the percentage
ity dolomite aggregates. These tests showed that: mass loss of material, having passed the control
sieve, is calculated.
− asphalt concrete mix AC 11 VS with high qual-
Using the Los Angeles Test method, it is allowed
ity dolomite aggregates could be used for the
by the standard to test also other aggregate frac-
asphalt concrete pavement wearing course with
tions alternative to the standard 10/14 mm size,
ESALs ≤ 3.0 million during design period, and
whereas, the Impact Test method allows to test.
− use of stone mastic asphalt mix SMA 11 S with
Therefore, when using the Los Angeles Test
high quality dolomite aggregates for asphalt
method a 10/14 mm size fraction of coarse aggre-
concrete wearing course should be evaluated
gate selected.
additionally, taking into account the change
The test results assessed according to their con-
of mechanical and physical properties of these
formity to the categories. For example, LA20 when
aggregates during road maintenance.
the mass loss is ≤ 20%, SZ shows the percentage
Respectively Vaitkus et al. (2015) state that mass loss of material having passed five test sieves,
dolomite aggregates commonly used in asphalt e.i. SZ18 when the mass loss is ≤ 18% (Table 1).
concrete pavement for the base course of surfacing LA calculated by the formula:
and road foundation layers’ mixtures.
This paper focus on experimental tests of 5000 − m
Lithuanian high quality dolomite aggregates for LA = (1)
50
resistance to fragmentation (SZ value) and Impact
Test (LA value). Analysis of test results showed where m – retained mass on a 1.6 mm sieve, g.
that dolomite meets Lithuanian TRA ASPHALT SZ calculated as:
08 Technical Specifications for Road Asphalt Mix-
tures and Latvian Ceļu SPECIFICATIONS 2015
SZ = ⎛ ⎞ %,
M
standards for coarse mineral materials, bituminous (2)
⎝ 5⎠
mixtures and surface treatments.

234

TIS2017_Book.indb 234 2/8/2017 8:03:27 PM

 Table 1. Values and categories of Los Angeles coefficient (LA) and Impact Resistance (SZ).

Los Angeles Category Value of resistance Category

Requirements coefficient LA to fragmentation% SZ

TRA MIN 07 ≤ 20 LA20 ≤ 18 SZ18

(Lithuania) ≤ 25 LA25 ≤ 22 SZ22
≤ 30 LA30 ≤ 26 SZ26
Ceļu SPECIFICATIONS ≤ 20 LA20 _ _
2015 (Latvia) ≤ 25 LA25
≤ 30 LA30
≤ 40 LA40

Note: According to TRA MIN 07 requirements for surface dressing SZ18 (LA20) and SZ22
(LA25), and according to Ceļu SPECIFICATIONS 2015 - LA20.

Figure 3. Results of Los Angeles coefficient (LA) of

the dolomite aggregates.
Figure 2. Experiment plan of dolomite aggregates for
resistance to fragmentation.

where M – the sum of percentage mass passing five

test sieves.
Table indicates Los Angeles coefficient (LA)
and impact resistance (SZ) requirements of cate-
gories for course aggregates by Ceļu SPECIFICA-
TIONS 2015 and course aggregates for bituminous
mixtures and surface treatments according to TRA
MIN 07 requirements.

3 EXPERIMENTAL RESEARCH
Figure 4. Results of resistance to fragmentation (SZ)
of the dolomite break-stone.
Figure 2 presents the experimental plan of the high
quality dolomite aggregates of resistance to frag-
mentation by SZ and LA tests. Whatever test results of high quality dolomite
obtained according to EN 1097-2 after 17 tests by
LA method and 17 tests by SZ method. Figure 3
3.1 Object and research methods
and Figure 4 show results of these tests.
High quality dolomite aggregates produced at AB
“Dolomite” quarry “Petrašiūnai-2” in Lithua-
3.2 Results
nia under the special production technology. In
2014−2016 period, the experimental studies of The received values of the high quality dolomite
resistance to fragmentation for this dolomite done aggregates for resistance to fragmentation accord-
in AB "Dolomite" laboratory. ing to EN 1097-2:2010 by Los Angeles method

235

TIS2017_Book.indb 235 2/8/2017 8:03:27 PM

 showed that all 17 samples meet requirements of tures and Latvian Ceļu SPECIFICATIONS 2015
the Los Angeles coefficient category LA20 for the requirements.
aggregates used in bituminous mixtures and for Although the experimental results are positive
surface dressing according to TRA MIN 07, TRA and it is necessary to continue a long-term research
ASPHALT 08 and Ceļu SPECIFICATIONS 2015. and tests on the high quality dolomite aggregates
Maximum value obtained of sample No. 7 and what enable to use local high quality dolomite
sample No. 13 is 15.9. The lowest value was 14.4 aggregates instead expensive importable granite
of the sample No. 6. The average value of this test aggregates in producing asphalt concrete mineral
was 15.2. mixes.
The received values of the high quality dolomite
aggregates for resistance to fragmentation accord-
ing to EN 1097-2 by Impact Test method showed REFERENCES
that all 17 samples meet requirements of the SZ
Bulevičius, M.; Petkevičius, K.; Žilionienė D.; Čirba, S.
value for the aggregates used in bituminous mix- 2011. Testing of Mechanical-Physical Properties of
tures and for surface dressing according to TRA Aggregates, Used for Producing Asphalt Mixtures,
MIN 07 and TRA ASPHALT 08. For SZ18 the and Statistical Analysis of Test Results, The Baltic
maximum value was 17.9% (sample No. 12), and Journal of Road and Bridge Engineering 6(2): 115−123.
the lowest value was 15.8% (sample No. 8). Aver- http://dx.doi.org/10.3846/bjrbe.2011.16.
age value was 17.2%. Fook, P.G.; Gourley, C.S.; Ohkere, C. 1988. Rock Weath-
Positive results of the experiments shows neces- ering in Engineering Time, The Quarterly Journal
sity to continue a long-term research of the high of Engineering Geology 21: 33−57. http://dx.doi.
quality dolomite aggregates. org/10.1144/GSL.QJEG.1988.021.01.03.
Haritonovs, V.; Zaumanis, M.; Brencis, G.; Smirnovs, J.
2013. Performance of Asphalt Concrete with Dolo-
mite Sand Waste and BOF Steel Slag Aggregate. The
4 CONCLUSIONS Baltic Journal of Road and Bridge Engineering 8(2):
91−97. http://dx.doi.org/10.3846/bjrbe.2013.12.
Lithuanian high quality dolomite aggregates Haritonovs, V., Tihonovs, J., Smirnovs, J. 2015. High
according to EN 1097-2 Tests for Mechanical and Modulus Asphalt Concrete with Dolomite Aggre-
gates. In IOP Conference Series: Materials Science
Physical Properties of Aggregates—Part 2: Methods and Engineering 96(1): 1−10. IOP Publishing. http://
for the Determination of Resistance to Fragmenta- dx.doi.org/10.1088/1757–899X/96/1/012084
tion by LA method test results correspond the min- Haritonovs, V., Tihonovs, J. 2014. Use of Unconventional
eral materials, used for bituminous mixtures and Aggregates in Hot Mix Asphalt Concrete. The Baltic
surface dressings according to Lithuanian TRA Journal of Road and Bridge Engineering 9(4): 276−282.
MIN 07 Technical Specifications for Road Mineral http://dx.doi.org/10.3846/bjrbe.2014.34
Materials, TRA ASPHALT 08 Technical Specifica- Labus, M.; Such, P. 2016. Microstructural Character-
tions for Road Asphalt Mixtures and Latvian Ceļu istics of Wellbore Cement and Formation Rocks
SPECIFICATIONS 2015 requirements of the Los under Sequestration Conditions, Journal of Petroleum
Science and Engineering 138: 77−87. http://dx.doi.
Angeles coefficient category LA20. org/10.1016/j.petrol.2015.12.010.
Lithuanian high quality dolomite aggregates Skrinskas, S.; Gasiūnienė, V. E.; Laurinavičius, A.;
by SZ method test results in the total value corre- Podagėlis, I. 2010. Lithuanian Mineral Resources,
sponds Lithuanian TRA MIN 07 Technical Speci- Their Reserves and Possibilities for Their Usage in
fications for Road Mineral Materials and TRA Road Building, The Baltic Journal of Road and Bridge
ASPHALT 08 Technical Specifications for Road Engineering 5(4): 218−228. http://dx.doi.org/10.3846/
Asphalt Mixtures requirements for mineral materi- bjrbe.2010.30.
als used in bituminous mixtures and surface dress- Šernas, O.; Vorobjovas, V.; Šneideraitienė, L.; Vaitkus,
ings (impact resistance category SZ18). A. 2016. Evaluation of Asphalt Mix with Dolomite
Aggregates for Wearing Layer, Transport Research
Lithuanian high quality dolomite aggregates Procedia 14: 732−737. http://dx.doi.org/10.1007/
meets the Lithuanian and Latvian requirements s00603–011–0174.
for bituminous mixtures and surface treatments Vaitkus, A.; Vorobjovas, V. 2015. Dolomito skaldos pan-
according to Lithuanian TRA ASPHALT 08 audojimas asfalto viršutinio sluoksnio mišiniams,
Technical Specifications for Road Asphalt Mix- Lietuvos keliai 1(34): 57−61. ISSN 1392–8678.

236

TIS2017_Book.indb 236 2/8/2017 8:03:28 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Design vehicles and roundabout safety—review of Croatian design

guidelines

Š. Bezina, I. Stančerić & S. Ahac

Faculty of Civil Engineering, University of Zagreb, Zagreb, Croatia

ABSTRACT: In this paper, review of Croatian national guidelines is given, with the emphasis on the
geometric design and fastest path (speed profile) analysis. The review is presented through several theo-
retical examples of single-lane suburban roundabouts with various external radii, designed according to
the recommendations of these guidelines. Design vehicles used in this analysis are truck with trailer, semi-
trailer truck and tri-axle bus. These vehicles occupy different swept path widths when moving through the
roundabout with the same external radii, which leads to differences in resulting entrance and exit widths
and consequently the speed profiles on roundabouts. These differences, as well as the impact of analyzed
design vehicles on the roundabout geometric design and speed profile, are also presented.

1 INTRODUCTION 2 BACKGROUND

Compared to the conventional intersections, Main differences between roundabout design

roundabouts can provide numerous benefits, guidelines and standards are related to the design
such as improved intersection safety and capac- of roundabout elements (roadway edges, entrance
ity, reduced maintenance costs and air pollution and exit widths, islands, circular lane widths), fast-
(Ahac et al. 2014; Hydén et al. 2000; Mandavilli est path analysis and sight distance tests (Ahac
et al. 2003). Their design process is highly iterative et al. 2014, 2016a,b; Bastos Silva & Seco 2005; Gal-
and usually consists of several steps: leli et al. 2014; Montella et al. 2013; CROW 1998,
FGSV 2006, RS 2011, SETRA 1998, TRB 2010,
− initial geometric design of elements,
TSC 2011 & VSS 1999).
− performance and safety checks (swept path
analysis, fastest path analysis and sight distance
tests), and 2.1 Geometric design elements
− final design of elements.
As a rule, the dimensions of the roundabout ele-
These steps are prescribed by national guidelines ments are determined on the basis of the design
or standards. Each step is equally important; if any vehicle swept path width, with the addition of lat-
of them is ignored, previously mentioned benefits eral protective widths. Design vehicles in European
can be annulled. (FGSV 2001, RS 2011) and American guidelines
In this paper, review of Croatian guidelines (HC (AASHTO 2011) are semi-trailer trucks, trucks
2014) for the design of roundabouts on national with trailers and buses with the maximum allow-
roads will be presented via 12 theoretical examples able length. Those three types of design vehicles
of single-lane suburban roundabouts with vari- have different dimensions (length, wheel base) and
ous external radii. These theoretical examples will they sweep different path widths when driving on
be designed according to the analyzed Croatian the circular arc. Because of that, resulting dimen-
guidelines (HC 2014), for following design vehi- sions of the roundabout elements differ, especially
cles: truck with trailer, semitrailer truck and tri- entrance and exit widths, circulatory road width,
axle bus. as well as entry and exit radii (Tables 1, 2).
The emphasis of this review will be on the
geometric design and fastest path analysis. Main
2.2 Speed analysis
objective of the study presented in this paper is
to compare the Croatian guidelines (HC 2014) One of the most important roundabout design
with the existing European (CROW 1998, FGSV feature, which is affecting vehicle speed on round-
2006, RS 2011 & TSC 2011) and American (TRB about, and consequently roundabout safety, is
2010) guidelines, and to recommend possible deflection around a central island (Bastos Silva
improvements. & Seco 2005). The influence of deflection on

237

TIS2017_Book.indb 237 2/8/2017 8:03:30 PM

 Table 1. Recommended lane widths (CROW 1998,
FGSV 2006, TRB 2010, VSS 1999).

Circulatory
Entrance Exit roadway

Country m m m

Germany 3.5–4.0 3.75–4.5 6.5–9.0

Switzerland 3.0–3.5 3.5–4.5 5.0–7.0
The Netherlands 3.5–4.0 4.0–4.5 5.0–6.0
USA 4.2–5.5 / 4.5–6.0

Table 2. Recommended entry and exit radii (CROW

1998, FGSV 2006, TRB 2010, VSS 1999). Figure 1. Deflection analysis according to (FGSV
2006).
Entry radius Exit radius

Country m m

Germany 14–16 16–18

Switzerland 12 14
The Netherlands 8–12 12–15
USA 15–30 15–60

roundabout performance can be evaluated by

defining the radius (or radii) of the centerline of
a vehicle traveling along the so-called fastest path
through the roundabout and then calculating the
vehicle speed (Galleli et al. 2014). This theoretical
fastest path is representative of most used trajec-
tory by drivers under free flow conditions when Figure 2. Geometry features used in determination of
minimizing their driving discomfort (Bastos Silva the fastest path radius according to (CROW 1998, TSC
& Seco 2005). 2011).
At first glance, this roundabout performance
check seems straightforward: firstly, fastest path is − by constructing the fastest paths through the
defined, then the path radii is measured or calcu- roundabout, measuring the path radii and then
lated, and at the end, the vehicle speed is estimated calculating the vehicle speed (as described in
based on speed-radius relationship. However, American (TRB 2010) and Serbian guidelines
inconsistencies in design standards and practices (RS 2011), and shown in Figure 3).
concerning procedures for the definition of the
deflection, construction of the fastest path, vehi- According to (CROW 1998, TSC 2011), the
cle speed estimation and the speed profile require- radius of the driving curve is calculated with the
ments can be observed (Ahac, 2016a). aid of the following equation:
As mentioned above, speed analysis on
( ⋅ L) ( (U + ) )
2 2
roundabouts can be conducted using different
approaches: R= (1)
U +2
− by measuring the roundabout’s geometry fea-
tures and checking the achieved deflection (as where R [m] = radius of the driving curve;
described in German guidelines (FGSV 2006), L [m] = distance between the tangent of entry
and shown in Figure 1); radius and the tangent of the exit radius; and U
− by measuring the roundabout’s geometry fea- [m] = distance between the edge of the central
tures and calculating the path radii and vehicle island and the right side of the lane of the access
speed (as described in Dutch (CROW 1998) and road, measured at the tangent point (Fig. 2). The
Slovenian guidelines (TSC 2011), and shown in design of the roundabout is correct if the radius of
Figure 2); the driving path is between 22 and 23 m.

238

TIS2017_Book.indb 238 2/8/2017 8:03:30 PM

 Figure 3. Fastest path radii according to (TRB 2010, Figure 4. Fastest path lateral clearances.
RS 2011).

The relation between the speed on the path 3 REVIEW OF CROATIAN GUIDELINES
curve and the radius of this curve is defined with FOR ROUNDABOUT DESIGN
the following equation:
Latest Croatian guidelines for roundabout design
Vi 74 Ri (2) (HC 2014) were published in June 2014, and their
usage is mandatory for the intersections on the
where Vi [km/h] = predicted design speed; and state roads. Geometric design of suburban round-
R i [m] = radius of the driving curve. Acceptable abouts according to these guidelines is carried out
vehicle speed through the roundabout should in nine major steps:
be in the range from 30 to 35 km/h (CROW
1998). − Selection of the external radius.
According to (TRB 2010, RS 2011), speed anal- − Determination of the circulatory roadway width
ysis procedure is composed from following steps: by using movement trajectory of two-axle design
construction of the fastest path on the analyzed vehicle driving in a full circle.
roundabout, measurements of path radii (Ri), and − Determination of the central island truck apron
estimations of vehicle speed based on speed-radius width by using movement trajectory of heavy
relationship: goods vehicle (truck with trailer or semi-trailer
truck) driving in a full circle.
− Selection of the approach roadway lane width
Vi 127 Ri ⋅ ( f e) (3) and splitter island form and length.
− Designing the outer roadway edge on entry and
where Vi [km/h] = predicted design speed; exit: selection of the entrance and exit width.
Ri [m] = radius of curve; f [−] = side friction factor; − Control of the entry angle and roadway widen-
and e [−] = superelevation. ing severity.
According to the analyzed documents, superel- − Swept path analysis on the roundabout for
evation values are assumed to be +0.02 (TRB 2010) the design vehicle and for all movement
or +0.025 (RS 2011) for entry and exit curves and directions.
−0.02 (TRB 2010) or −0.025 (RS 2011) for curves − Determination of the fastest path and vehicle
around the central island. speed through the roundabout.
According to (TRB 2010, RS 2011) fastest path − Conducting the sight distance tests.
should be drawn with the prescribed distances to
the particular geometric feature: (a) painted edge As in all abovementioned design standards,
line of the splitter island, (b) a right curb on the roundabout design process described in Croatian
entrance and exit of the roundabout, and (c) a curb guidelines (HC 2014) is iterative: if the initial
of the central island, as shown in Figure 4. Ameri- roundabout design does not meet the requirements
can guidelines (TRB 2010) define these distances for unobstructed design vehicle movement, travel
as (a) = 1.0 m, (b) = (c) = 1.5 m, while Serbian speed and sight distance, it has to be re-defined.
guidelines (RS 2011) define only (b) = 1.5 m and The design vehicle selection, swept path analysis
(c) = 2.0 m. According to (TRB 2010), acceptable and fastest path analysis given in these guidelines is
entry vehicle speed is 40 km/h. briefly described below.

239

TIS2017_Book.indb 239 2/8/2017 8:03:30 PM

 3.1 Design vehicles and swept path analysis radius (Rul), exit radius (Riz), and entrance width
(e). Circulatory roadway width (u) is determined
According to Croatian guidelines (HC 2014) round-
based on two-axle vehicle movement trajectory,
about swept path analysis can be performed with
while driving in a full circle. In order to determine
two design vehicles: semitrailer truck (16.5 m long)
the radius of this circle, roundabout’s outer radius
and truck with trailer (18.75 m long). Dimensions
is reduced by the protective lateral width, which is
of those vehicles are in compliance with (European
1.0 meter. Recommended and borderline dimen-
Committee Directive).
sions of abovementioned design elements for small
Two-axle vehicles are also mentioned in the
and medium sized suburban single-lane round-
guidelines (HC 2014) because their swept paths
abouts are given in Table 3.
are used for the determination of the circulatory
After the selection of basic design elements,
roadway width. Namely, required widths of a cir-
entry angles (Φ) and roadway widening severity
culatory roadway are specified for different wheel
on entrance (S) are checked. Severity on entrance
base and front overhang lengths of two-axle vehi-
is the dimensionless parameter of roundabout
cles and external radii. Their choice should be har-
design, which represents relationship between
monized with the needs of users at the proposed
entrance width and lane width. The dimensions
roundabout location and with the consent of the
of both entry angle and severity on entrance, as
road administration.
well as entrance widths, depend on the design of
Swept path analysis is conducted by drawing
the splitter island and applied entry and exit radii.
the design vehicle (body) movement trajectories
Borderline dimensions for the entry angles are
in all possible directions on roundabout blueprint.
0–77°, and recommended values are 20–40°, while
It ensures that conditions for unobstructed vehi-
both borderline and recommended values for the
cle movements on roundabout entry and exit are
severity on entrance are between 0 and 2.9.
achieved. The minimum protective lateral width
along the trajectories is 0.5 m (exceptionally 0.3 m)
on all segments, except on the outside edge of the 3.3 Speed analysis procedure
circulatory roadway where the minimum lateral
Speed analysis procedure described in Croatian
width is 1.0 m.
guidelines (HC 2014) is based on the procedure
described in Dutch (CROW 1998) and Slovenian
3.2 Geometric design elements (TSC 2011) guidelines. According to these guide-
lines, speed analysis on roundabout is conducted
According to the Croatian guidelines (HC 2014),
by measuring the roundabout’s geometry features
on small and medium sized suburban round-
and calculating the path radii and vehicle speed.
abouts (with outer radius Rv from 11 to 25 meters)
However, there are some changes in Croatian
the splitter island is triangular (Fig. 5.). Recom-
guidelines (HC 2014) (Fig. 6).
mended length of these islands (m) is between 15
to 50 meters. − Geometry feature L that is used in calcula-
Required speed profiles on roundabouts are tion of the path radii (Eq. 1) and vehicle speed
achieved by proper design of the right curb on (Eq. 2) is measured via fastest path through round-
entrances and exits, i.e. the proper selection of entry about—it is defined as the distance between the
tangent of entry radius and the tangent of the
exit radius on the fastest path.
− The vehicle path is determined by the clearance of
1.0 m from the roadway edges or the roundabout

Table 3. Recommended and borderline dimensions

(HC 2014).

Recommended Borderline

Element m m

Entrance width (e) 4.0–7.0 3.6–10.0

Lane width (v) 3.0–3.5 2.5–7.0
Entry radius (Rul) 8.0–20.0 6.0–25.0
Exit radius (Riz) 10.0–25.0 8.0–50.0
Circulatory roadway 4.5–6.0 4.0–9.0
Figure 5. Design elements of suburban roundabouts. width (u)

240

TIS2017_Book.indb 240 2/8/2017 8:03:31 PM

 4.1 Roundabout design
After the selection of the external radii (Rv), lane
widths (v), and shape and length of the splitter
islands (m), circulatory roadway widths (u) were
defined as the sum of the lateral protective widths
and width (sv + Δv), and rounded to 0.1 m (Table 4,
Figs. 5, 7). Widths (sv + Δv) were determined on
the basis of the two-axle vehicle movement trajec-
tories, while driving in a full circle of radii Rv-1
(Table 4). All trajectories were drawn by software
(Autodesk Vehicle Tracking 2015).
In this study, the two-axle vehicle used to
determine (sv + Δv) widths was 12.0 m long bus
from German guidelines (FGSV 2001), because
Figure 6. Fastest path construction elements (HC dimensions of the two-axle vehicle needed for
2014). the construction of the movement trajectories are
not specified in Croatian guidelines (HC 2014).
Croatian guidelines (HC 2014) only give (sv + Δv)
design elements: (a) painted edge line of the widths for a specific set of Rv-1 values (8.0, 9.0,
splitter island, (b) a right curb on the entrance 11.0, 12.5, 15.0, 18.0, and 21.5 m). Initial entrance
and exit of the roundabout, and (c) a curb of the widths (eul), exit widths (eiz) and outer edge radii
central island. (Rul and Riz) were also selected.
− Acceptable vehicle speed through the round- After an initial roundabout design was finished,
about should be in the range from 25 to 40 km/h. swept path analyses were conducted for all design
vehicles. The trajectories were drawn by a software
The differences between Croatian guidelines
(HC 2014) and aforementioned foreign standards
and guidelines (CROW 1998, RS 2011, TRB 2010,
Table 4. Applied dimensions of roundabouts design
TSC 2011) inspired the research described shown elements.
below. The aim of this research was to establish if
and in what way these discrepancies influence the Outer radius [m] 17.5 20.0 22.5 25.0
results of speed analyses.
m [m] 30.0 30.0 30.0 30.0
v [m] 3.5 3.5 3.5 3.5
4 STUDY Rul [m] 13.0 13.0 13.0 13.0
Riz [m] 15.0 15.0 15.0 15.0
For the purpose of this study, theoretical examples Rv-1 [m] 16.5 19.0 21.5 24.0
of single-lane suburban roundabouts with vari- sv+Δv [m] 4.94 4.62 4.32 4.11
ous external radii were designed (Rv = 17.5, 20.0, u [m] 6.5 6.2 5.9 5.7
22.5, and 25.0 m). Initial design elements of these u’ [m] 1.0 1.0 1.0 1.0
roundabouts were defined according to the ana-
lyzed Croatian guidelines (HC 2014).
Swept path analysis was performed for three
design vehicles (truck with trailer 18.75 m long,
semitrailer truck 16.5 m long, and tri-axle bus
15.0 m long). Because tri-axle buses play an
important role in passenger transport on long-
distance lines in Croatia, this type of vehicle was
included in the research, even though Croatian
guidelines (HC 2014) do not define buses as
design vehicles.
For each design vehicle, four roundabouts with
different external radii were designed. On those
roundabouts fastest path analysis was performed,
and vehicle speed was estimated according to
Croatian (HC 2014), Dutch (CROW 1998) and
American guidelines (TRB 2010), as described
below. Figure 7. Suburban roundabout scheme.

241

TIS2017_Book.indb 241 2/8/2017 8:03:31 PM

 (Autodesk Vehicle Tracking 2015) for all move- Table 7. Applied dimensions of roundabouts design
ment directions: straight, right, left, U-turn and elements for three-axle Bus (B).
circular. Based on these trajectories circular swept
path widths (sv + Δv’) were measured and follow- Roundabout Id. 17.5B 20.0B 22.5B 25.0B
ing elements were defined: truck apron widths (u’), Outer radius [m] 17.5 20.0 22.5 25.0
entrance widths (eul), exit widths (eiz), outer edge sv+Δv’ [m] 5.95 5.43 5.05 4.77
radii (Rul) and outer edge radii (Riz) (Tables 4–7). eul [m] 5.50 6.00 6.50 6.50
Entry angles (Φ) and roadway widening severity eiz [m] 6.00 6.25 6.50 7.25
on entrance (S) were also controlled. Obtained val- Φ [°] 43.3 46.5 49.0 49.1
ues of entry angles were within limit range, while S [-] 0.33 0.33 0.35 0.38
severity on entrance values were within the recom-
mended range (Tables 5–7).
Swept path analysis showed that selected three-
axle bus occupies greatest turning width, compared Table 8. Speed analysis variant parameters.
to semi-trailer truck and truck with trailer, and
that it requires the greatest entry and exit round- Clearances
about width. In addition, out of the 12 defined
Variant m Radius Speed
roundabouts some of them have the same elements
dimensions (Tables 5–7). CRO-1 a = b = c = 1.0 Eq. 1 Eq. 2
− Roundabouts with the outer radius of 20.0 m CRO-2 a = 1.0, b = c = 1.5 Eq. 1 Eq. 2
dimensioned for the design vehicle truck with USA-1 a = b = c = 1.0 measured Eq. 3
trailer (20TT) and semi-trailer truck (20ST). USA-2 a = 1.0, b = c = 1.5 measured Eq. 3
− Roundabouts with the outer radius of 22.5 m NLD – Eq. 1 Eq. 2
dimensioned for the design vehicle truck with
trailer (22.5TT) and semi-trailer truck (22.5ST).
Therefore, the vehicle speed estimation was con- with the Croatian (HC 2014), The Netherlands
ducted only on 10 examples of roundabouts. (CROW 1998) and American guidelines (TRB
2010). In these analyses, different fastest path
clearances from the roadway edges were taken into
4.2 Fastest path speed analysis
consideration, to test the effect of their changes on
For each roundabout fastest path analysis and vehi- the resulting speed (Table 8).
cle speed estimation were conducted in accordance Following assumptions were also made.
− Superelevation values are +0.025 for entry and
Table 5. Applied dimensions of roundabouts design exit curves, and –0.025 for curves around the
elements for Truck with Trailer (TT). central island.
− Side friction factor varies with vehicle speed and
Roundabout Id. 17.5TT 20.0TT 22.5TT 25.0TT is determined in accordance with (AASHTO
2011).
Outer radius [m] 17.5 20.0 22.5 25.0
sv+Δv’ [m] 5.39 4.95 4.62 4.38 In order to investigate the effect of different
eul [m] 5.25 5.75 6.00 6.00 approaches to the construction of the fastest paths
eiz [m] 5.75 6.25 6.50 6.50 and different equations for the vehicle speed cal-
Φ [°] 42.0 41.7 47.0 47.2 culation, speed analyses conducted according to
S [-] 0.44 0.40 0.44 0.56 (TRB 2010, HC 2014) included the definition of
the fastest path radii on entrances (R1), around the
central islands (R2), and on exits (R3).
Table 6. Applied dimensions of roundabouts design
elements for Semi-trailer Truck (ST).
5 DISCUSSION
Roundabout Id. 17.5ST 20.0ST 22.5ST 25.0ST

Outer radius [m] 17.5 20.0 22.5 25.0

The vehicle speed calculated in accordance with
5.72 5.13 4.84 4.56
the Croatian guidelines (HC 2014) with clearances
sv+Δv’ [m]
eul [m] 5.25 5.75 6.00 6.00
of 1.0 m from the central island and the roadway
eiz [m] 5.75 6.25 6.50 6.75
edges exceed the recommended value of 40 km/h.
Φ [°] 42.0 41.7 47.0 47.2 They are also higher than calculated speed on the
S [-] 0.44 0.40 0.44 0.56 fastest path with clearance of 1.5 m. The devia-
tions are in range from 2 km/h to 4 km/h, and they

242

TIS2017_Book.indb 242 2/8/2017 8:03:32 PM

 decrease with an increase of an external radius of
the roundabout.
The vehicle speed calculated in accordance
with the Dutch guidelines (CROW 1998) is within
the recommended limits (30 to 35 km/h), and is
smaller than the speed obtained in accordance
with the Croatian guidelines (HC 2014). This dis-
crepancy was expected, due to differences in the
definition of length L, and consequently the cal-
culated radius of the driving curve. The vehicle
speed around the central island (Fig. 8) calculated
in accordance with the American guidelines (TRB
2010) is significantly smaller than the speed calcu-
lated in accordance with the Croatian guidelines
(HC 2014). Average difference is 14 km/h for the
1.0 m path clearance, and 12 km/h for the 1.5 m
Figure 8. Vehicle speed around the central island. path clearance.
Even though the entrance and exiting widths
of roundabouts designed for three-axle buses are
larger than those designed for semi-trailer truck
and truck with trailer (0.25 and 0.50 m), speed
deviation rate around the central island is neg-
ligible (1 km/h). These differences are larger for
speed on entry and exit paths (Figs. 9, 10), where
larger lane widths needed to accommodate three-
axle buses led to the deviation of speed from 1 to
3 km/h.

6 CONCLUSIONS

Results from swept path analysis on theoretical

examples of roundabouts showed that three-axle
bus with the length of 15.0 meters needs largest
dimension of the design elements. That is because
this vehicle takes up a greater width when turning
than the semi-trailer truck and truck with trailer. It
Figure 9. Entry vehicle speed.
is evident that the impact of these vehicles on the
performance and safety of roundabouts must not
be ignored, especially in tourist countries such as
Croatia, where three-axle buses play an important
role in the transport of passengers on intercity and
international routes.
According to the results from fastest path analy-
sis and vehicle speed estimations on the theoretical
examples of roundabouts, it can be concluded that
the vehicle path clearances have the greatest impact
on vehicle speed. After that comes the impact of
the design vehicle type, because they have an effect
on the roundabout’s entrance and exit width. It
should be noted that 1.0 m clearance does not
always ensure unhindered passage of a passen-
ger car through the roundabout. Because of that,
larger minimum clearances from the central island
and right curb on the entrances and exits are rec-
ommended, for instance 1.5 meters. All this indi-
cates to the need to revise the guidelines in terms of
Figure 10. Exit vehicle speed. vehicle path clearances.

243

TIS2017_Book.indb 243 2/8/2017 8:03:32 PM

 REFERENCES FGSV (Forschungsgesellschaft für Straßen und
Verkehswesen). (2006). Merkblatt fur die Anlage von
AASHTO (American Association of State Highway Kreisverkehren. FGSV K 1000, Köln, Germany.
Transport Officials). (2011). A Policy on Geometric Gallelli, V., Vaiana, R. & Iuele, T.: 2014. Comparison
Design of Highways and Streets. Washington DC. between simulated and experimental crossing speed
Ahac, S., Džambas, T., Stančerić, I. & Dragčević, V. 2014. profiles on roundabout with different geometric fea-
Performance checks as prerequisites for environmen- tures, Procedia – Social and Behavioral Sciences 11:
tal benefits of roundabouts. In Stjepan Lakušić (ed.), 117–126.
Road and Rail Infrastructure III, Proceedings of the HC (Smjernice za projektiranje kružnih raskrižja na
Conference CETRA 2014, Split, 28–30 April 2014. državnim cestama). (2014). Hrvatske ceste d.o.o.
Zagreb: Department of Transportation, Faculty of Hydén, C. & Várhelyi, A. 2000. The effects on safety, time
Civil Engineering, University of Zagreb. consumption and environment of large scale use of
Ahac, S., Džambas, T. & Dragčević, V. 2016a. Review of roundabouts in an urban area: a case study, Accident
fastest path procedures for single-lane roundabouts. Analysis and Prevention 32: 11–23.
In Stjepan Lakušić (ed.), Road and Rail Infrastruc- Mandavilli, S., Russell, E.R. & Rys, M.J. 2003. Impact
ture IV, Proceedings of the Conference CETRA 2016, of Modern Roundabouts on Vehicular Emissions, In
Šibenik, 23–25 May 2016. Zagreb: Department of Proceedings of the 2003 Mid-Continent Transporta-
Transportation, Faculty of Civil Engineering, Univer- tion Research Symposium, Ames, 21–22 August 2003.
sity of Zagreb. Ames: Iowa State University.
Ahac, S., Džambas, T. & Dragčević, V. 2016b. Sight Montella, A., Turner, S., Chiaradonna, S. & Aldridge,
distance evaluation on suburban single-lane round- D. 2013. International overview of roundabout design
abouts, Građevinar 68(1): 1–10. practices and insights for improvement of the Italian
Autodesk Vehicle Tracking. (Integrated swept path anal- standard, Canadian Journal 0f Civil Engineering 40:
ysis software). (2015). Autodesk, San Rafael, CA. 1215–1226.
Bastos Silva, A. & Seco, A.: Trajectory deflection influ- RS (Pravilnik o uslovima koje sa aspekta bezbednosti
ence on the performance of roundabouts, In Euro- saobraćaja moraju da ispunjavaju putni objekti i drugi
pean Transport Conference (ETC) Association for elementi javnog puta). (2011) Službeni glasnik RS, br.
European Transport, Strasbourg, 2005. 50/2011
CROW (Dutch Technology Platform for Transport, SETRA (Service d’Etudes sur les Transports, les Routes
Infrastructure and Public Space). (1998). Eenheid in et leurs Aménagements). (1998). Aménagement des
rotondes. CROW 126, Ede, Netherlands. carrefours interurbains sur les routes principales,
European Committee Directive. (2002). “Directive Carrefours plans—guide technique.
2002/7/EC of the European Parliament and of the TRB (Transport research board). (2010). Roundabouts:
Council of 18 February 2002 amending Council An Informational Guide, 2nd Edition. Washington,
Directive 96/53/EC laying down for certain road vehi- D.C.
cles circulating within the community the maximum TSC (Krožna križišča, TSC 03.341: 2011). (2011).
authorised weights in international traffic and the Ljubljana: Ministrstvo za infrastrukturo in prostor –
maximum authorised weights in international traffic.” Direkcija RS za ceste.
OJ L 67, Brussels, Belgium. VSS (Vereinigung Schweizerischer Strassenfachleute).
FGSV (Forschungsgesellschaft für Straßen und (1999). Schweizer Norm SN 640 263: Knoten mit
Verkehswesen). (2001). Bemessungsfahrzeuge und Kreisverkehr. Zürich.
Schleppkurven zur Überprüfung der Befahrbarkeit
von Verkehrsflächen, Forschungsgesellschaft für
Strassen- und Verkehrswesen. Köln: FGSV.

244

TIS2017_Book.indb 244 2/8/2017 8:03:32 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Laboratory investigation into the modification obtained of paving

bitumen with different plastomeric polymers

E. Saroufim & C. Celauro

Department of Civil, Environmental, Aerospace, Materials Engineering, University of Palermo, Palermo, Italy

N. Khalil
Department of Civil Engineering, University of Balamand, Lebanon

ABSTRACT: The use of polymer/bitumen modified blends in road paving applications has been
increasingly advancing from the technological point of view. Plastomeric polymers are known to be
amongst the most widely used in bitumen modification. Therefore, this paper presents the influence and
degree of modification of several plastomeric-modified bitumen in terms of rheological properties and
morphology. Conventional tests such as penetration grade, Ring-and-Ball softening point and storage
stability were performed. Rheological properties of the modified blends were characterized in terms of
dynamic mechanical analysis via frequency sweep tests using the Dynamic Shear Rheometer (DSR). The
morphology of the samples is investigated by means of fluorescent light optic microscopy. New processed
and mixed polymers are introduced into the analysis. The results confirm that the fundamental properties
and morphology of the modified bitumen are directly dependent on the type of polymer in use. Overall,
the polymer-modified bitumen improves the original bitumen properties: remarkably stable blends were
found with overall good performance especially for treated polymers. The microscope pictures provide
useful information on the different types of interaction between polymers and the bitumen.

1 INTRODUCTION The latter would be better classified as thermoplas-

tic polymers, since the term “plastomer” simply
Bitumen is known to be a viscoelastic material with refers to their processability, thus recalling some
properties suitable for pavement construction. But technological properties they offer.
bitumen alone has poor mechanical performance, As reported in the literature, elastomeric modi-
it is hard and brittle in cold weather and soft in fication works by making the binder more elastic
hot environments (Blanco et al., 1996). With the and reducing its stiffness (Robinson, 2005). Plasto-
increase in traffic load nowadays, pavement sur- meric modification of bitumen, in contrast, makes
faces are highly stressed and affected by elevated bitumen stiffer and reduces its temperature suscep-
traffic volume and temperature variations. Con- tibility, particularly at high in-service temperatures
sequently the need to enhance pavement surfaces (McNally, 2011). This is important in Mediterra-
and minimize maintenance problems has allowed nean regions where high temperatures affect long-
the field of bitumen/polymer modification to term road conditions (Viola and Celauro, 2015).
grow and become important in designing opti- Thus, plastomers would be an appropriate modifier
mally performing pavements. Only small amounts to sustain the needs for road construction in warm
of additives are sufficient to maintain bitumen’s climates. They are also known to reduce the risk of
own advantage and improve the desired proper- rutting (Robinson, 2005, McNally, 2011), which is
ties (Blanco et al., 1996). It has been recorded amongst the major distresses of flexible pavements
that pavements with polymer modification exhibit and is particularly severe in hot regions.
greater resistance to rutting and thermal cracking, Although significant research has been done in
and decreased fatigue damage, stripping and tem- this area, plastomeric PMBs are still poorly under-
perature susceptibility (Yildirim, 2007). stood scientifically, due to the complex nature of
There is a wide variety of polymers that have bitumen and its interaction with the polymer sys-
been used as additives to bitumen for producing tem (Brule, 1996). This paper highlights this cat-
so-called PMBs (polymer modified bitumens). Nor- egory of modification.
mally, two main types of polymers are popular for Plastomers are known to offer several advan-
bitumen modification: elastomers and plastomers. tages beside some drawbacks in terms of PMB

245

TIS2017_Book.indb 245 2/8/2017 8:03:34 PM

 performance, depending on the polymer type taken The polymers investigated in this research are
into consideration (Zhu et al., 2014). For instance, Low-Density Polyethylene (LDPE), High Density
polyethylene-based polymers, once added to bitu- Polyethylene (HDPE) and Ethyl-Vinyl-Acetate
men, allow limited improvement in elasticity and (EVA) copolymer supplied by Versalis. The latter
potential storage stability problems, whereas Eth- contains 28% vinyl-acetate.
ylene-Vinyl Acetate (EVA) in appropriate content HDPE, EVA and LDPE were used as received.
gives relatively good storage stability to the final LDPE was processed (P-LDPE) in a Brabender
blends (Zhu et al., 2014). Consequently, the plas- mixer at 210°C for 30 min at a rotational speed
tomeric modification process exposes challenges of 64 rpm. Finally, LDPE and EVA were blended
in this domain since pavement construction tech- in the same mixer but in lower processing condi-
niques require the binder to remain stable before tions, at 180°C for 10 min at a rotational speed of
and after application on the road (Navarro et al., 64 rpm. The EVA content in the blend was
2004). Therefore, not only enhancement in rheo- 20% wt/wt.
logical properties is required but also production The modified binders were prepared at 180°C
of blends that will resist separation under high using a Silverson High shear mixer. The polymer
temperature storage, yet remain at low cost. content for all the modified binders was kept con-
Modification of binders is a complex mecha- stant and equal to 5% by weight as it has been
nism that requires simplification. Therefore, differ- seen in other studies to be an effective content
ent types of polymers, including new compositions, for modification (Pérez-Lepe et al., 2003) and it
were used in order to create a link between the is, therefore, one of the most common contents
rheological properties and stability of the PMBs, for practical applications. An additional mix was
and their influence on morphology. In this con- done with reduced polymer content (2.5%), only
text, conventional tests are carried out on different for HDPE, for comparison purposes regarding the
plastomeric polymer/bitumen blends. Apart from final properties of the blend. Mixing was carried
the typical empirical/technological tests on road out for an extra 30 min after the entire polymer
bitumen, the rheological properties of the poly- sample was fed, at a constant rate, into the heated
mer/bitumen are investigated through dynamic bitumen to produce homogenous mixtures. The
mechanical analysis while the morphology is evalu- modified samples were then sealed in aluminum
ated by assessing the state of dispersion of poly- containers and stored for further testing.
mers into the bitumen matrix. Empirical tests such as penetration and sof-
tening point were run according to the follow-
ing standards: EN 1426:2007 and EN 1427:2007,
2 MATERIALS AND METHODS respectively. The results are shown in Table 2.
After mixing with the base bitumen, the modi-
2.1 Physical properties and classical tests fied binders underwent the storage stability test
according to EN 13399:2004 Standard in order to
The physical properties of pure bitumen from
monitor the phase separation between binder and
vacuum distillation having a penetration grade
polymers because it is considered a critical aspect
50/70 are shown in Table 1. Five different polymers
of modification results.
were used for modification of one binder in order
to highlight the effect of the polymer used on the
final blend produced.
Table 2. Conventional tests on polymer modified
binders.

Table 1. Conventional properties of the base bitumen. T T

Units of Requirements Intermediate High

Characteristics measurement Values
Characteristics Pen at 25°C SP
Grade 50/70 Test method EN 1426 EN 1427
Penetration dmm 68 Unit of measure dmm °C
Softening point °C 50.5 PURE BITUMEN 50/70 68 50
Fraas breaking point °C –12 LDPE 35.7 56
Ductility mm over 100 HDPE 25.7 58
Increase in softening °C 3.5 LDPE/EVA 35 61
point (after RTFOT) P-LDPE 31.7 57
Increase in penetration % 64.7 EVA 40.1 61.1
(after RTFOT) 2.5% HDPE 33 55.5

246

TIS2017_Book.indb 246 2/8/2017 8:03:34 PM

 2.2 Dynamic Shear Rheometer 14023:2010 Standard, the specific storage stabil-
ity requirement for PMBs is that the softening
The viscoelastic properties of the polymer/bitu-
point difference between top and bottom (ΔTR&B)
men blends are investigated by means of Dynamic
is ≤ 5°C after 3 days of storage.
Mechanical Analyses (DMA). Frequency sweep
The stability results are shown in Table 3. The
tests were carried out within the linear viscoelas-
large increase in temperatures for LDPE and
ticity range using a Dynamic Shear Rheometer
HDPE in both percentages indicates that the sam-
(Physica MCR 1). Depending on the tempera-
ples are unstable and macroscopic phase separa-
ture range the tests were carried out with parallel
tion readily occurs during their storage at high
plates of diameter 8 or 25 mm. The temperatures
temperatures, which leads to undesirable problems
ranged from −10°C to 80°C. The rheological prop-
during transportation. Surprisingly, the process-
erties of the binders were measured in terms of
ing of LDPE as well as the addition of EVA to
their complex modulus G* and phase angle δ in a
the LDPE show significant stability. These results
controlled-strain loading mode.
emphasize the possibility of transforming previ-
ously unstable polymer/bitumen blends into abso-
2.3 Fluorescence microscopy lutely stable ones. Two methods are suggested in
this paper, one through treatment of virgin poly-
Fluorescent microscopy is the most valuable method
mer with other components and one through a
for studying the phase morphology of polymer
mechanical process.
modified bitumen, as it allows direct observation
of different microstructures of modified bitumen.
This is because the polymer-rich domains appear 3.3 Stiffness master curve
yellow while the asphaltene-rich domains remain The effect of polymer addition on the complex
black (Lesueur, 2009). This allows qualitative evalu- modulus loss and storage modulus of binders can
ation of the homogeneity and the structure of the
modified bitumen (Sengoz and Isikyakar, 2008a). Table 3. Storage stability.
In this study, the preparation of samples implied
pouring of the blends, which were thoroughly Requirements Storage stability
stirred to ensure homogeneity, into small elements Characteristics ΔTR&B
subjected immediately to a cooling regime. Prior to Method EN 13399
fluorescent image analysis, the frozen samples were Test EN 1427
sharply cut and placed on a lamella to examine the Unit of measurement °C
fractured part. The images were taken with a mag- LDPE 28
nification of 16:1 using a trinocular OPTIKA® HDPE 20
microscope, N-400FL model. LDPE/EVA 5
P-LDPE 0.4
EVA 1.75
3 RESULTS AND DISCUSSIONS 2.5% HDPE 10

3.1 Conventional test results

It can be seen from Table 2 that the addition of the
selected polymers to bitumen shows a decrease in
the Penetration (Pen) and an increase in the Sof-
tening Point (SP) for all the modified blends. How-
ever, the modification does not significantly affect
the softening point in relation to penetration: the
increase in the softening point is relatively uniform
for all the blends in the range of the 5–10% for all
the polymers, while there is a considerably larger
decrease in penetration values registered, especially
for the HDPE at 5%. This roughly indicates that
the bitumen in use becomes—in any case—harder
and less susceptible to temperature variation.

3.2 Storage stability

Storage stability is one of the most critical aspects Figure 1. Complex modulus |G*| (Pa) master curves at
of asphalt modification. According to the EN Tref = 30°C.

247

TIS2017_Book.indb 247 2/8/2017 8:03:35 PM

 polymer and its content. In the low frequency
domain, which corresponds to high temperatures,
a significant increase in stiffness for all the PMBs
is spotted. This is clearly seen for 5% HDPE, with
a value approximately equal to 1.5 kPa, which
denotes solid-like behaviour and improvement of
elasticity. Furthermore, the plateau region seen
for this polymer is an indication of a predomi-
nant polymer network within the modified binder
(Airey, 2002, Pérez-Lepe et al., 2003), while all the
other PMBs, at very low frequencies, show similar
asymptotic behaviour to each other.
Thus the high temperature stiffness is enhanced
for all the PMBs, and particularly for 5% HDPE.
The presence of the plateau region at high tem-
peratures (low frequencies) triggered the need to
investigate the behaviour of HDPE in lower per-
Figure 2. Storage modulus master curve G' (Pa), at centages for a better understanding of the mate-
Tref = 30°C. rial, and also to investigate whether this behaviour
and the overall stability are replicated. At low fre-
quencies, the 2.5% curve of HDPE has an asymp-
totic shape similar to that of the base binder but
with higher stiffness values; while 5% HDPE tends
to form a plateau. With lower polymer content, the
behaviour of the modified bitumen remains close
to that of the straight-run bitumen. This could
be an additional indicator that the rheological
response at 5% HDPE is strongly contributed by
the polymer. Nevertheless, both blends are found
to be unstable, as seen previously. Thus at a cer-
tain concentration of HDPE the blend receives its
properties from the polymer itself.
At low temperatures (high frequencies), this is
no longer the case, since all the mixtures exhibit
similar trends, apart from those with P-LDPE and
pure EVA which attain lower storage modulus
values.
The results shown in Figures 2 and 3 for G′
Figure 3. Loss modulus G˝ (Pa) at Tref = 30°C.
and G′′ comply with the complex modulus master
curve. The addition of polymer ends up with an
be observed in Figures 1, 2 and 3 respectively. The increase in the modulus values at high temperature,
reference temperature, Tref, is set at 30°C for the which is beneficial for limiting permanent defor-
application of the Time-Temperature Superposi- mation in road structures.
tion Principle (TTSP). All the polymer/bitumen
blends displayed higher complex modulus when
3.4 Phase angle master curves
compared to the pure binder. A clear increment
of the results is evident in the low frequency (high Figure 4 represents the phase angle master curves
temperature) zone. To a certain extent, this indi- for all the binders studied. A clear change in the
cates an improvement in the mechanical behaviour shape of the phase angle curves is observed after
of the base binder. On the other hand, an increase polymer addition. This indicates an improvement
in |G*| in the low frequency range is very common in elastic behaviour after modification.
in polymer modification. It can indicate that the Once again, HDPE modification at 5% displays
modifier has made the base binder less suscepti- a small plateau region at intermediate frequencies
ble to temperature (Asgharzadeh and Tabatabaee, followed by a reverse slope in the lower frequency
2013). The considerable difference in the degree of range. This plateau means that the elastic and
modification depends on the modifier type: this viscous complex modulus components vary in
proves that the properties of the blend primarily the same proportion so that the phase angle does
depend on the rheological characteristics of the not change (Da Silva et al., 2004). Such a plateau

248

TIS2017_Book.indb 248 2/8/2017 8:03:35 PM

 Table 4. Letter coding description according to
EN13632.

Phase continuity P: Continuous polymer phase

B: Continuous bitumen phase
X: Continuity of both (crosslinking)
Phase description H: Homogeneous*
I: Inhomogeneous
Size description S: Small (< 10 μm)
M: Medium (from 10 μm to 100 μm)
L: Large (>100 μm)
Shape description r: Round, cylindrical
s: Elongated
o: Other

* If no structure is visible but fine, homogeneous, slightly

yellow particles are detected, the product should be called
homogeneous.
Figure 4. Phase angle (tan δ) master curves at
Tref = 30°C.
The morphologies of the PMBs vary in accord-
ance with the differences in the characteristics of
the polymers used (in terms of type, density, etc.).
region for the phase angle, together with a rela- However, a desired morphology is illustrated in the
tively high modulus (as much as possible insensi- form of a biphasic condition in which a polymer-
tive to temperature variation) could be favourable rich phase and an asphaltene-rich phase coexist in
to binder deformation resistance (Lu and Isacs- a micro-scale metastable equilibrium stage (Polacco
son, 1997). In relation to this comportment, it is et al., 2006). From the different shapes depicted in
worth recalling that plateau rheological curves, Figures 5 (a) to (e) and listed in Table 5 it can be
in general, are a characteristic of the viscoelas- deduced that the morphology is correlated with the
tic response of rubbers. Therefore, they imply a properties previously examined in this paper. An
polymer-rich elastic network (Airey, 2003, Da inception of a polymeric network can be roughly
Silva et al., 2004) emphasizing the polymer elastic seen for P-LDPE, where a uniform dispersion of
properties which govern the overall behaviour of small irregularly shaped polymer particles is spot-
the blend. ted with a smaller quantity of medium particles
having non-uniform dispersion and a tendency to
group. In any case, based on previous studies (Lu
3.5 Fluorescence microscopy
and Isacsson, 2000, Sengoz and Isikyakar, 2008b)
Bitumen is a complex material. Its components polymeric phases are known to be strongly evident
are typically simplified by arranging them into a when at least 6% of polymer is used in the blend, i.e.
colloidal structure (Lesueur, 2009). Addition of at a slightly higher percentage than the one used in
polymer makes it even more complex. Morphology this study. The P-LDPE blend, which has acceptable
facilitates the analysis of the internal structure, as mechanical performances and excellent storage sta-
it investigates the polymer phase dispersion in the bility, displays uniformly distributed small particles
bitumen matrix by means of a simple microscopic with some parts of polymers linked together in clear
visualisation process. shapes. This morphology indicates a structured mate-
In this study, the morphology of the produced rial, because of the interconnection in the dispersed
PMBs is represented in Table 5 as a general char- phase. Thus, amongst the polymers studied in this
acterization related to the nature of the continuous research P-LDPE proves to be a promising modi-
phase as well as a description of the phases and fier for this grade of bitumen. As a new approach
shapes in a letter coding format according to EN in polymer modification, P-LDPE polymer gives
13632: 2005 (Table 4). the blend satisfactory mechanical performances
It was not possible to take representative micro- and stresses the fact that available resources can be
photographs of the blend with EVA, since the treated for enhancing the properties of the desired
dispersion was visible only for couple of seconds blends. The same applies to the LDPE/EVA stable
before the frozen material quickly melted under blend, where elongated small particles are uniformly
the microscope lamp. This could be a result of the distributed in the bitumen matrix.
EVA concentration in the blend and its low melting On the other hand, 5% HDPE, which showed an
temperature (around 40°C). improvement in the stiffness modulus, has a similar

249

TIS2017_Book.indb 249 2/8/2017 8:03:36 PM

 Table 5. Morphological characterization through letter
abbreviation.

Phase Phase Size Shape

PMB continuity description description description

LDPE B I S r/o
HDPE B I M r/o
P-LDPE B H/I S/M r/o
LDPE/EVA B H S s/r
EVA – – – –
2.5% HDPE B I S r/o

morphological phase separation, which is attributed

to Brownian coalescence, followed by gravitational
flocculation and later on leads to creaming (Habib
et al., 2011). The difference here between HDPE
and P-LDPE blends is in the degree of homogeneity
of the distribution of these particles in the bitumen
mass. Unlike P-LDPE, HDPE does not show any
uniform distribution of smaller polymer particles.
This indicates that the polymer is not well dispersed
within the bitumen, implying that the blend, at high
temperature ranges, primarily acquires polymer
properties. The mechanism of the polymer-rich
phase separation has also been proved to be due
to creaming, preceded by coalescence, as seen in a
previous study on HDPE (Pérez-Lepe et al., 2006).
Finally, stabilization effects are seen in this research
for the LDPE, either by mechanically process-
ing the polymer or by adding another copolymer
(EVA) known for its relatively good storage stabil-
ity. In the same perspective, a similar approach is
desirable when working with HDPE too.
As a result, successful modification should
be reflected not only in the enhanced rheologi-
cal properties and satisfactory storage stability at
macroscopic level, but in the dispersed polymeric
phase on the microscopic scale as well.

4 CONCLUSIONS

The addition of polymers to bitumen enhances the

mechanical properties of the modified binder. Even
though plastomers are known to be an effective and
economical modifier for the bitumen, they reveal
some major drawbacks, as mentioned in the intro-
duction. The main disadvantages are typically seen
in mechanical property deficiencies and also regard-
Figure 5. Microphotographs of the polymer/bitumen
blends studied. ing the stability in storage and transportation of
these PMBs. Hence overcoming their drawbacks is
an advance in bitumen modification and a challenge
structure to that of the P-LDPE blend in terms of for researchers in the field of paving materials.
“large particle” morphology but with less intercon- In this paper, a variety of polymer/bitumen
nection between the bitumen and polymer. HDPE blends was studied, at typical polymer contents.
previously proved to be unstable. One of the main The selected polymers were of the plastomeric
causes of this instability is the tendency of PMB to type. The results in terms of storage modulus and

250

CH033_156.indd 250 2/8/2017 8:55:55 PM

 phase angle indicate an improvement in perform- Habib, N. Z., Kamaruddin, I., Napiah, M. & Tan, I. M.
ance for all PMBs. The highest enhancement of 2011. Rheological properties of polyethylene and
elastic behaviour is registered for the blend pro- polypropylene modified bitumen. International Jour-
duced with 5% HDPE. On the other hand, the nal Civil and Environmental Engineering, 3, 96–100.
Lesueur, D. 2009. The colloidal structure of bitumen:
same blend also showed the highest stiffness in Consequences on the rheology and on the mecha-
penetration values, evident storage instability nisms of bitumen modification. Advances in colloid
and large circular spheres with no homogenous and interface science, 145, 42–82.
distribution of smaller particles in micrographs. Lu, X. & Isacsson, U. 1997. Characterization of Styrene-
This indicates that the mechanism of separation Butadiene-Styrene Polymer Modified Bitumens-
of the polyethylene from the bitumen is preceded Comparison of Conventional Methods and Dynamic
by the coalescence and flocculation phenomenon Chemical Analyses. Journal of testing and evaluation,
of the dispersed phase. On the opposite side, the 25, 383–390.
more stable bitumen/polymers blends, represented Lu, X. & Isacsson, U. 2000. Modification of road bitu-
mens with thermoplastic polymers. Polymer Testing,
by LDPE undergoing mechanical processing and 20, 77–86.
LDPE mixed with EVA, have improved rheological Mcnally, T. 2011. Polymer modified bitumen: properties
properties. They show a homogenous dispersion of and characterisation, Elsevier.
small particles in the bitumen. Thus micrographs, Navarro, F. J., Partal, P., Martínez-Boza, F. & Gallegos,
regardless of the type of polymer, can be a useful C. 2004. Thermo-rheological behaviour and storage
indicator of the rheo-stability of the blend. stability of ground tire rubber-modified bitumens.
Processing of LDPE and addition of EVA Fuel, 83, 2041–2049.
to LDPE showed an effective way of stabilizing Pérez-Lepe, A., Martínez-Boza, F. J., Gallegos, C.,
the virgin polymer. Considering the good results González, O., Muñoz, M. E. & Santamaría, A. 2003.
Influence of the processing conditions on the rheo-
obtained for the HDPE modifier, in terms of rheo- logical behaviour of polymer-modified bitumen. Fuel,
logical performances, further stabilization methods 82, 1339–1348.
should be developed for the HDPE/bitumen blend Pérez-Lepe, A., Martínez-Boza, F. J., Attané, P. &
in order to stabilize the blend and avoid filler type Gallegos, C. 2006. Destabilization mechanism of
behaviour or, in other terms, the hardening (stiff- polyethylene-modified bitumen. Journal of Applied
ening) effect of the polymer. Polymer Science, 100, 260–267.
Polacco, G., Stastna, J., Biondi, D. & Zanzotto, L. 2006.
Relation between polymer architecture and nonlinear
REFERENCES viscoelastic behavior of modified asphalts. Current
opinion in colloid & interface science, 11, 230–245.
Robinson, H. 2005. Polymers in asphalt, iSmithers Rapra
Airey, G. D. 2002. Rheological evaluation of ethylene
Publishing.
vinyl acetate polymer modified bitumens. Construc-
Sengoz, B. & Isikyakar, G. 2008a. Analysis of styrene-
tion and Building Materials, 16, 473–487.
butadiene-styrene polymer modified bitumen
Airey, G. D. 2003. Rheological properties of styrene
using fluorescent microscopy and conventional test
butadiene styrene polymer modified road bitumens.
methods. Journal of Hazardous Materials, 150,
Fuel, 82, 1709–1719.
424–432.
Asgharzadeh, S. & Tabatabaee, N. 2013. Rheological
Sengoz, B. & Isikyakar, G. 2008b. Evaluation of the prop-
master curves for modified asphalt binders. Scientia
erties and microstructure of SBS and EVA polymer
Iranica. Transaction A, Civil Engineering, 20, 1654.
modified bitumen. Construction and Building Mate-
Blanco, R., Rodríguez, R., García-Garduño, M. &
rials, 22, 1897–1905.
Castaño, V. M. 1996. Rheological properties of styrene-
Viola, F. & Celauro, C. 2015. Effect of climate change
butadiene copolymer–reinforced asphalt. Journal of
on asphalt binder selection for road construction in
Applied Polymer Science, 61, 1493–1501.
Italy. Transportation Research Part D: Transport and
Brule, B. 1996. Polymer-modified asphalt cements used
Environment, 37, 40–47.
in the road construction industry: basic principles.
Yildirim, Y. 2007. Polymer modified asphalt binders.
Transportation Research Record: Journal of the
Construction and Building Materials, 21, 66–72.
Transportation Research Board, 48–53.
Zhu, J., Birgisson, B. & Kringos, N. 2014. Polymer modi-
Da Silva, L. S., De Camargo Forte, M. M., De Alencas-
fication of bitumen: Advances and challenges. Euro-
tro Vignol, L. & Cardozo, N. S. M. 2004. Study of
pean Polymer Journal, 54, 18–38.
rheological properties of pure and polymer-modified
Brazilian asphalt binders. Journal of materials science,
39, 539–546.

251

TIS2017_Book.indb 251 2/8/2017 8:03:37 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Long-term monitoring of half warm mix recycled asphalt containing up

to 100% RAP

José Manuel Lizárraga, Antonio Ramírez, Patricia Díaz & Miguel Martín
Sacyr Construcción, Madrid, Spain

Francisco Guisado
Repsol, Madrid, Spain

ABSTRACT: This paper presents the main results from a project-site investigation to compare the
mechanical performance differences between half warm mix recycled asphalt (HWMRA) and a con-
ventional mix (HMA) section in Spain. To this end, HWMRA was manufactured using total recycled
contents up to 100% RAP at temperatures below 1000C. This study was therefore performed in three
phases. In the first phase, a preliminary laboratory testing analysis was conducted to ascertain the optimal
mix design solution through the influence of different curing times in a forced-draft convection oven at
50°C. A second phase assessed the viability of manufacturing these mixes in a discontinuous asphalt mix-
ing plant adapted with a secondary flow-parallel drying drum, while the final phase involved the construc-
tion of a road test section formulated with two emulsion contents. At the current time, little difference is
seen regarding the performance of the HWMRA compared to the control HMA section.

1 INTRODUCTION significant advantages regarding environmental

conservation and lowers energy consumption by
Environmental concerns about the preservation 50%, thus saving significant costs compared to
of the non-renewable resources and increasing conventional HMA (Olard et al. 2008; Blankendaal
damage to the environment from greenhouse gas et al. 2013). Moreover, the reduction of half-warm
(GHG) emissions have created increasing aware- mix production temperatures implies significant
ness and rising costs in raw materials (Tutu & emission reductions of 58% carbon dioxide and
Tuffour 2016). In this way, the reuse of recycled 99.9% sulfur dioxide (Rubío et al. 2013). Despite
material is happening more than ever before in a the environmental and money saving advantages
sustainable technique practice, and cost-effective that the use of these types of mixtures implies,
method able to generate significant economic there are some concerns and uncertainties about
savings in rehabilitation activities of up to 34% the mechanical behavior and life span of these
(Kandhal & Mallick 1997). These materials still mixes. For this reason, the reuse of high RAP con-
retain high costs and good properties, which there- tent is still downgraded in other layers (base and
fore should be reused in manufacturing and laying binder courses) rather than replacing it within
of new mixtures in road pavements (Botella et al. asphalt surface courses, thus laboratory testing
2015) and also to favorably reduce environmental analysis along with field investigations is required
burdens by 23% (Chiu et al. 2008). to demonstrate the performance and effectiveness
As regards cleaner production technologies, of these mixes.
they can be classified according their manufac- Recent studies have been focused on showing
turing temperatures as Cold Mix Asphalt (CMA) that reducing temperatures to half-warm produc-
spread usually at room temperatures, Half-Warm tion range and incorporating RAP should not com-
Mix Asphalt (HWMA) spread below 100°C, and promise the performance of these asphalt mixtures
Warm-Mix Asphalt (WMA) spread between (Dinis-Almeida et al. 2012; Swaroopa et al. 2015). In
110 and 140°C. In this way, half-warm mixes in addition, most of the current asphalt mixing plants
combination with reclaimed asphalt pavement are not completely equipped neither for manufactur-
(HWMRA) can be considered as an innovative ing nor to deal with total recycling production con-
concrete material produced at temperatures around tents up to 100% RAP. In fact, the content of RAP
60–700C lower than those used in the production to be added in new asphalt mixtures is still limited
of conventional HMA and yet still have similar and downgraded to less than 20% RAP in asphalt
mechanical performance. The reduction in manu- batch plants (Zaumanis & Mallick 2014) because
facturing and compaction temperatures entails of the difficulties in mixing the RAP, together with
253

TIS2017_Book.indb 253 2/8/2017 8:03:38 PM

 the virgin aggregates. Nevertheless, there are some tested in the laboratory, specifically, two HWMRA
examples of reusing total contents up to 100% RAP with total recycled contents of up to 100% RAP and
in a modified asphalt mixing plant using an addi- two emulsion binder contents (2.5% and 3.0%) and
tional flow parallel dryer drum for RAP heating a conventional mixture with a binder 50/70 penetra-
(García & Lucas, 2014). To validate this half-warm tion grade. The characterization of the materials, as
mix design procedure, a set of preliminary labora- well as the mixture itself, was based on European
tory tests were performed to evaluate whether the Committee for Standardization (CEN) standards.
use of high rates of RAP are feasible in developing
recycled mixes with a high level of durability and
2.3 RAP characteristics
performance (García et al. 2011). To this end, both
formulations of HWMRA (2.5 and 3.0% by weight As was already previously mentioned in the 2.1 sec-
of the mix) were produced in the laboratory to tion, the RAP was obtained from damaged road
accurately ascertain their volumetric and mechani- pavement through a Wirtgen (W/1000) cold milling
cal performance, and subsequently these asphalt machine. The RAP was transported to the asphalt
mixes were reproduced in a modified discontinuous mixing plant to be classified into two fractions, that
asphalt mixing plant capable of reaching up to 100% is, the fine fraction (0/5 mm) and coarse fraction
RAP production at temperatures below 100°C. A (5/25 mm). This was done with the aim of ensur-
third phase involved the construction and deploy- ing gradation, optimal binder contents in the final
ment of HWMRA in a service road using two dif- mix design and also with the aim of avoiding exces-
ferent emulsion contents (2.5 and 3%), respectively. sive mix heterogeneity. For this purpose, both RAP
These mixes were intended for use in binder courses fractions were homogenized, quartered, treated and
for low traffic load categories. In short, perform- characterized. This was done in terms of residual
ance is compared in the laboratory and after plant binder content per EN 12697–1, RAP binder recov-
manufacture, laying and compaction in the field ery through the centrifuge extractor method per EN
with the aim of increasing the confidence in using 12697–3, penetration test EN 1426, softening point
these cleaner production technologies. per EN 1427, and gradation curves per EN 933–2,
that is, black and white grading curves, where black
curves are the RAP gradation before the extraction
of the aged binder, while the white curves are the
2 HALF WARM MIX RECYCLED
RAP gradation once the residual binder has been
ASPHALT
extracted. The average binder content value was
7.02% in the fine fraction and 3.66% in the coarse
2.1 Service road structure
fraction by total weight of aggregates as collated in
The test section was built on a service road paral- Table 1. Moreover, the recovered binder showed an
lel to the A1 motorway, between km 203 and 204, average penetration value of 17 dmm and a soften-
located in the town of Lerma, province of Burgos in ing point of 67.3°C. The physical binder charac-
Spain. The service road is currently being subjected terization and particle grain size distribution of the
to T2 traffic loads with an average daily intensity RAP are collated in Tables 2–3.
of heavy vehicles equal to or greater than 200 and The aggregate grading curves of both
lower than 800. The service road consists of two HWMRA fell within the limits stipulated for
lanes with one way traffic on a carriageway width of dense graded asphalt concrete mixtures (AC16
10 m. The rehabilitation works consisted of milling D) in accordance with the following proportion
a layer of 5 cm thickness, which was removed and of RAP fractions: 40% fine fraction (0/5 mm)
replaced using half warm mix recycled asphalt for and 60% coarse fraction (5/25 mm), as illustrated
binder course with the following dimensions: 400 m in Figure 1.
long and a thickness of 5 cm over its entire carriage-
way width. The remaining 300 m were rebuilt with Table 1. Determination of the binder content.
a conventional hot mix asphalt layer (AC16 D). In
this way, the slow traffic lane was designed with 3% Fraction Unit 0/5 mm 5/25 mm
emulsion, whilst the fast traffic lane was formulated
binder/aggregate (b/a) % 7.02 ± 0.1 3.66 ± 0.07
with 2.5% emulsion content. Finally, the asphalt con-
crete surface course (AC16 D) was paved in the total
length of 700 m and with a thickness of 40 mm. Table 2. Physical characterization of the binder.

Properties Unit Test method Value

2.2 Material and methods
Penetration 0,1dmm EN 1426 17
For the successful development of this research Softening point °C EN 1427 67.3
project, both HWMRA were characterized and

254

TIS2017_Book.indb 254 2/8/2017 8:03:38 PM

 Table 3. Particle grain size distribution: white and phase, coating with RAP aggregates and to guar-
black. antee a thick asphalt emulsion. In this regard, the
bituminous emulsion was formulated with high
RAP residue content greater than 61% by weight of the
0/5 mm 5/25 mm emulsion. For emulsion production, a bitumen
Sieve size 50/70 dmm penetration grade was employed. The
UNE (mm) white black white black asphalt emulsion meets the current pavement recy-
cling specifications according to EN 13808 stand-
31.5 100 100 100 100 ard. The characterization of the bitumen emulsion
22.4 100 100 99.7 99.2 consisted of analyzing the viscosity, water content,
16 100 100 95.6 84.5 sieving, residual binder content and penetration as
8 100 100 58.9 27.4 collated in Table 4.
4 84.9 76.7 26.1 1.2
2 60.5 44.3 19.4 0.9
0.5 34.4 15.6 12.0 0.7 3 METHODOLOGY
0.25 25.6 8.1 9.5 0.6
0.063 14.0 1.3 5.5 0.2 This project investigation was divided into three
main phases. The first phase consisted of a pre-
liminary field investigation together with a labo-
ratory testing program to accurately evaluate the
optimal mix design solutions through a compre-
hensive accelerated ageing study (0, 24, 48 and
72 h) in a forced-draft convection oven at 500C.
This study enables the reproduction of field
pavement conditions one year after construction
(Bowering 1970; Thenoux & Jamet 2002). The
second phase assessed the feasibility of manufac-
turing these mixes in a modified discontinuous
asphalt mixing plant equipped with a secondary
flow parallel drum. Cores were extracted and
tested after construction for determining density,
Indirect Tensile Strength (ITS) at 150C, stiffness
modulus (ITSM) at 200C and Indirect Tensile
Strength Ratio (ITSR) through the water sensitiv-
ity test, in June 2012. The last phase involved the
construction process (laying and compaction) of
a test road section to validate both the short and
Figure 1. Aggregate grading curves of HWMRA and long-term mechanical performance of these mixes
HMA.
under real conditions. Finally, a set of monitor-
Table 4. General technical specifications for cationic ing campaigns were also performed to control and
bitumen emulsion (CSS-1). assurance the quality of these mixes in the field.
In April 2014, new pavement cores were taken
Test Emulsion and tested to verify the evolution of volumet-
Property method Unit CSS-1 ric characteristics in terms of air void contents,
resistance to Indirect Tensile Fatigue Test (ITFT)
Viscosity at 250C EN12846-1 S 23
at 20°C, stiffness modulus (ITSM), water sensitiv-
pH EN 12850 – 3.0
Water content EN 1428 % 38.8
ity and Indirect Tensile Strength (ITS) two years
Sieving EN 1429 % 0.01 after construction. Specific details are thoroughly
Residual binder content EN 1431 % 61.2 shown in the following sections.
Penetration of residual EN 1426 dmm 66
binder at 25°C
4 PRELIMINARY LABORATORY STUDY
2.4 Bituminous emulsion
4.1 Mix design
Both HWMRA have been manufactured with a
cationic slow setting bituminous emulsion (CSS-1), HWMRA samples were initially produced in the
which was selected as the asphalt binder to pro- laboratory using RAP and bituminous emulsion,
vide adequate workability during the compaction thus, the initial conditions began by heating the

255

TIS2017_Book.indb 255 2/8/2017 8:03:39 PM

 RAP at temperatures between 900 and 1000C and Table 5. Mechanical-volumetric properties of HWMRA.
bitumen emulsion at 500C. The Superpave Gyra-
tory Compactor (SGC) was chosen and employed % Emulsion content
for compaction assessment of the half warm mixes
Properties 0% 2.0% 2.5% 3.0% 3.5%
in accordance with EN 12697–31. The compaction
of the samples was performed at a temperature of Maximum 2.481 2.428 2.407 2.389 2.377
700C, thus the number of gyros to be applied on density, (g/cm3)
asphalt samples was adapted to obtain final air Apparent density, 2.268 2.327 2.328 2.338 2.339
void contents between 2.5 and 4.0%, so it was nec- ssd, (g/cm3)
essary to apply, at least, 65 load cycles to obtain the Geometric 2.164 2.256 2.258 2.278 2.289
required mix performance. The compaction began density (g/cm3)
using the following conditions shown below: Air voids (%) 8.6 4.2 3.3 2.1 1.6
Modulus (MPa) 3754 3034 2891 2861 2364
i. Internal angle velocity: 0.82 0C ITSd, (MPa) 1.55 1.87 2.13 1.93 1.72
ii. Constant speed of rotation: 30 rpm ITSw, (MPa) 1.08 1.82 2.08 1.89 1.69
iii. Vertical contact pressure load: 600 kPa
iv. Number of load cycles: 65 gyrations ITSR (%) 69.5 97.3 97.6 98.1 98.3
v. Mold diameter: 100 mm
The determination of the optimal mechanical a certain curing period to develop their ultimate
performance of HWMRA was conducted accord- mechanical properties such as: stiffness modulus
ing to the apparent density curves under saturated and ITS (Godenzoni et al. 2016). For this reason,
surface dry (ssd) conditions, as per EN 12697-6, mechanical tests were carried out to ascertain
maximum density per EN 12697-5, air void con- whether it could have had a loss of water content
tents per EN 12697-8, water sensitivity test per EN by evaporation during the laboratory compaction
12697-12, and stiffness modulus per EN 12697-26, phase; however, most of the water content was
in comparison with different emulsion contents lost during the mixing process between 60 and 120
ranging from 2.0 to 3.5% (at 0.5% increments). The seconds. The development of ultimate mechanical
target properties of the mix design were set out in performance of these mixes can be largely attrib-
accordance with the resistance to water action, uted either to emulsion breaking or moisture loss
indirect tensile strength and air void contents. (Bocci et al. 2011; Serfass et al. 2004). In this way,
The retained strength ratio was fixed above 90%, a servo-pneumatic mechanical testing machine
air void contents should be between 2.5 and 4.0%, was used to determine both ITS and stiffness
and Indirect Tensile Strength (ITS dry) at 15°C modulus. Therefore, ITS test consisted of apply-
between 1.5 and 2.0 MPa according to the traffic ing a constant deformation rate of 50 ± 2 mm/min
load category (García & Lucas, 2014). In order after a conditioning period of 2 hours, according
to determine the retained strength ratio, asphalt to EN 12697-23. Thereafter, the laboratory speci-
samples were prepared using two-thirds (2/3) of mens were compared along with specimens that
the total compaction energy used for determining still hadn’t been subjected to an accelerated curing
their mechanical and volumetric properties in the process, thus, the stiffness modulus and ITSdry
laboratory. The average retained strength values values remained relatively constant within the
for all mixtures were higher than 90%, except for first 24 hours. Nevertheless, ITS at 15°C showed a
0% emulsion content, as indicated in Table 5. sharply increase of cohesiveness at the end of the
curing process between 2.13 and 2.34 MPa, while
the stiffness modulus at 20°C varied from 2891 to
4.2 Influence of curing on HWMRA
3642 MPa, as collated in Table 6.
After HWMRA had been characterized, a new
set of laboratory specimens were prepared with
4.3 Wheel tracking test
the optimal emulsion content of 2.5%, to ana-
lyze the influence of curing as optimization of To complete the preliminary laboratory study of
the mix design. In this way, HWMRA 2.5% were the mechanical performance of HWMRA, the
then assessed and tested after being subjected to resistance to permanent deformation was per-
an accelerated ageing process (0, 24, 48 and 72 h) formed using the wheel tracking test at tempera-
through a forced-draft convection oven at a tem- tures of 50 and 60°C, respectively. This test was
perature of 500C. This was done to evaluate both performed with duration of 10,000 load cycles, per
volumetric characteristics and mechanical behav- procedure b in air, with a contact load pressure of
ior such as: (i) Indirect Tensile Strength (ITS) at 700 N, with a frequency of 26.5 ±1 load cycles/min,
15°C and (ii) stiffness modulus at 20°C. Typically, according to EN 12697-22.
these types of mixes manufactured at low temper- Prior to the wheel tracking test, the prismatic
atures and with bitumen emulsion tend to require specimens were manufactured with the following

256

TIS2017_Book.indb 256 2/8/2017 8:03:39 PM

 dimensions: 300 mm length ⋅ 300 mm width ⋅ 50 mm Table 7. Wheel tracking test at 50°C and 60°C.
height, and thereafter compacted until reaching
up to 98% of the benchmark density through the HWMRA 100% RAP
vibratory steel roller device in accordance with EN 2.5% 2.5% 3.0%
12697-33. Testing 50°C 60°C 60°C
The wheel tracking test was also conducted for
HWMRA 2.5%, at 50°C, to evaluate the influence of Apparent density, ssd, (g/cm3) 2.302 2.328 2.330
the temperature on mix resistance to permanent defor- Deformation at 5000 load/cycles, 0.52 2.11 2.48
mation in asphalt binder courses, since these mixes RDAIR, (mm)
are exposed to significant lower thermal gradients Deformation at 10.000 load/ 0.86 2.66 3.19
compared to surface courses. In this sense, the aver- cycles, RDAIR, (mm)
age deformation slope at 50°C, was 0.068 (mm/1000 Wheel tracking slope, WTSAIR, 0.068 0.109 0.143
load cycles) between 5000 and 10,000 load cycles, (mm/103 load cycles)
and at 60°C was 0.109 (mm/load cycles). Moreover, Proportional rut depth, 1.42 3.47 5.3
PRDAIR, (%)
HWMRA 3.0% was also assessed and tested at 60°C,
resulting in a slope deformation of 0.143 (mm/1000
load cycles) and with an average proportional rut ess began by weighting the right proportion of
depth of 5.3%. In relation to what has been men- RAP, heating it in the drum dryer at 100°C, and
tioned above, HWMRA 2.5% meets the minimum heating the bituminous emulsion at 50–60°C. All
strengths required for the potential T2 traffic loads. the components were then mixed together in the
Table 7 shows the mean apparent density, deforma- mixer, transported by the bucket elevators to the
tion slope values and proportional rut depths. discharge hopper, and after which the resulting
asphalt mix was unloaded to the delivery trucks at
100°C. Another crucial aspect to reach up to 100%
5 ON-SITE ROAD TESTING half-warm mix production was the proper charac-
terization of the RAP. To this end, the RAP was
5.1 Reproducibility of the HWMRA in an classified into two fractions, the fine (0/5 mm) and
modified asphalt plant coarse fraction (5/25 mm). For this case study, the
After the preliminary laboratory testing phase, it modified batch plant was fixed at a maximum theo-
was decided to evaluate the feasibility of manu- retical output capacity of 120 t/h. Figure 2 shows a
facturing these types of asphalt mixes in a modi- schematic overview of the pretreatment and
fied discontinuous asphalt mixing plant used for processing of RAP, together with the installation
conventional HMA production. However, it was of specific equipment. The facilities of the modi-
necessary to make some adjustments to reach up fied discontinuous asphalt plant are illustrated,
to 100% RAP production at half warm tempera- as follows: 1. Grinding roller mill 2. Screening 3.
tures. To achieve this, the batch plant was adapted Belt conveyor of the coarse fraction (5/25 mm) and
with a secondary flow-parallel drying drum with 4. Fine fraction (0/5 mm) 5. Cold feeding bins 6.
a differentiated inlet ring for giving special treat- drum dryer 7. Vertical RAP elevator 8. Flow-paral-
ment to each RAP fraction and for driving off the lel drum and 9. Vertical virgin aggregates elevator.
moisture content in the fine fraction. This batch
plant features a delayed combustion chamber, 5.2 Spreading and compacting
wherein the RAP is heated through the hot gases For the final phase to validate this technology
before coming into contact the flame, thus, there under real conditions, a pilot road test section was
is no direct contact either with the burner flame previously built near the asphalt mixing plant to
or at high temperatures zones. An additional hot define the optimal compaction procedure. The
gases recirculation circuit was installed for poten- paving process of these mixes was conducted with
tial suction improvement, partial thermal energy the same machinery used for actual field pavement
recovery, and replacement of secondary fresh air construction of conventional mixtures, namely, an
for temperature control. The manufacturing proc- asphalt paver, a pneumatic road roller and vibra-
tory road roller compactor. Spreading and laying
Table 6. Curing of the HWMRA samples at 500C. of the HWMRA mixes began using a Volvo (ABG
8820) asphalt paver in which the vibrating screed
Curing time 0h 24 h 48 h 72 h was fixed at a paving speed of 2.5 m/min and
loading frequency of 1700 rpm to induce a high
Apparent density (g/cm3) 2.313 2.308 2.300 2.315
precompaction degree on the asphalt mixture.
Air voids, Vm, (%) 3.9 4.1 4.4 3.8
Modulus (MPa) 2891 2984 3077 3642
A Dynapac (CC422) double-drum vibratory road
ITSdry (MPa) 2.13 2.11 2.17 2.34
roller with a maximum theoretical operating weight
of 10.4 ton, together with a Dynapac (CP271) pneu-

257

TIS2017_Book.indb 257 2/8/2017 8:03:39 PM

 Figure 2. Facilities of the discontinuous asphalt mixing plant using a secondary flow parallel drum (UM 260 t/h).

Table 8. Volumetric and mechanical results of

HWMRA.

HWMRA 100% RAP

Properties 2.5% 3.0%

Binder content (%s/m) 6.17 6.25

Density, ssd, (g/cm3) 2.364 2.365
Figure 3. Monitoring process: a) manufacturing below Air voids, Vm, (%) 3.0 2.65
100°C and b) spreading at temperatures between 85 and Modulus at 20°C (MPa) 5248 5057
90°C. ITS dry (MPa) 2.80 2.92
ITSR (%) 92.2 90.3
matic tyre road roller of 27 ton (mass with maximum
ballast) were employed to achieve the benchmark
density above 98% of the asphalt samples that have 6.2 Cores performance
previously been prepared for determining their
In June 2012, more than 32 pavement cores were
mechanical and volumetric properties in the labora-
extracted after each mix was compacted for deter-
tory. To this end, the compaction protocol adopted
mining thicknesses, density, air void contents, stiff-
during the road construction consisted of applying
ness modulus at 20°C, ITS at 15°C. Moreover, in
at least four double passes with the vibratory roller
April 2014, almost two years after construction of
compactor and four passes with the pneumatic road
the service road, and with mixtures being subjected
roller. In addition, from manufacturing until final
to T2 traffic loads; more than twelve new pavement
paving were continually monitored through an infra-
cores were extracted for each type of asphalt mix-
red thermographic camera system (FLIR B360) to
ture to assess the long-term in-situ performance,
detect possible homogeneity defects generated dur-
specifically, the resistance to fatigue at 20°C. These
ing laying and compaction as illustrated in Figure 3.
scheduled coring campaigns were performed to both
analyze the short-term performance and their evolu-
tion in terms of density, air voids, stiffness modulus,
6 RESULTS OF THE TEST ROAD SECTION
ITS, as collated in Table 9.
6.1 Asphalt samples
6.3 Particle size distribution
Both half-warm formulations (2.5 and 3.0%) were
taken after in-plant manufacturing to evaluate the The control and quality assurance of these mixes
reproducibility of these mixes in a modified batch was done in terms of grading, particle grain size
plant. Hence, more than six laboratory specimens distributions and binder contents in the final mix
were compacted using the gyratory compactor by design. The final binder content for HWMRA
applying the same standard conditions, which have (2.5%) was 6.17% and 6.25% (of the total weight
previously been obtained in the preliminary mix of aggregates) for HWMRA 3%. The average
design. This was performed for determining air void binder content resulting from batches-in black
contents, stiffness modulus at 20°C, water sensitivity didn’t vary by a percentage greater than ± 0.3% by
and ITS dry at 15°C. The average stiffness modulus total weight of aggregates regarding the prelimi-
values were above 5000 MPa, resulting in accord- nary laboratory mix design. In this context, both
ance with those obtained for the conventional mix- half warm mixes fell within the limits stipulated
tures. From the ITSR test, both HWMRA showed for dense graded asphalt concrete mixtures (AC16
retained strengths above 90%, as collated in Table 8. D), see in Figure 4.
258

TIS2017_Book.indb 258 2/8/2017 8:03:39 PM

 Table 9. Performance of pavement cores extracted in Table 10. Fatigue laws for HWMRA and HMA cores.
2012, and two years after construction in 2014.
HWMRA 100% RAP
HWMRA – 100% RAP HMA
Regression constants 3.0% 2.5% AC16D
Cores – 2012 Cores – 2014 HMA
Cores a (μm/m) 0.0007 0.0009 0.0006
Properties 2.5% 3.0% 2.5% 3.0% 2014 b (−) 0.159 0.177 0.149
R2 0.7906 0.8727 0.9195
Height (mm) 53.6 55.2 54.2 57.0 50.9 Modulus (MPa) 3284 3792 4287
Apparent 2.309 2.250 2.277 2.307 2.315 ε6 (μm) 63 69 75
density, (g/cm3)
Air voids, 5.2 7.4 4.27 6.46 –
Vm, (%) frequency of 10 Hz and with a loading time of 0.1 sec-
Modulus at 4758 3431 3505 4990 5399 onds, according to EN 12697-24 (AENOR, 2003).
20°C (MPa) The failure criterion was defined using the classi-
ITS in dry at 2.64 2.29 2.54 2.43 2.85 cal approach as the number of cycles to 50%, (Nf),
15°C (MPa) reduction in the initial stiffness modulus measured at
ITSR (%) 90.2 94 79.2 88.4 94.5 100th load cycle. The deformation slope and fatigue
resistance of the HWMRA (3.0%) and HMA cores
were practically identical, resulting in a very simi-
lar fatigue resistance behavior. For the stress levels
tested, fatigue performance of HWMRA (2.5%)
was slightly better than that of the control mixture.
Nevertheless, the deformation slope of HWMRA
(2.5%) was slightly steeper than that of the other
two mixtures, which therefore indicates that for
greater number of load cycles the fatigue perform-
ance could be somewhat worse than that of the
other two mixtures, as depicted in Figure 5. The
regression constants, stiffness modulus and micros-
trains (a, b, R2, ε6) are collated in Table 10.

7 CONCLUSIONS

Figure 4. Aggregate grading curves of both HWMRA This research project was focused on demonstrating
(2.5–3%). the suitability of manufacturing half-warm mixes
with total recycled contents of RAP at temperatures
below 1000C and using a cationic bituminous emul-
sion for use in asphalt binder courses. In this regard,
this research project involved the mechanical char-
acterization of the optimal mix design solution in
the laboratory, in-plant manufacturing, laying, con-
struction and monitoring of its performance over
time. The optimal job mix formula was chosen and
employed after an accelerated ageing process (0,
24, 48 and 72h) in a forced-draft convection oven.
The main conclusions that can be drawn from this
research project are the followings:
− Short and long term in-situ performance of the
half-warm mixes with total recycled contents of
RAP showed similar results to those obtained for
conventional mixtures in terms of water sensitivity,
Figure 5. Fatigue laws of HWMRA 100% RAP and stiffness modulus and indirect tensile strength.
HMA. − The aggregate grading curves of both HWMRA
formulations fell within the limits stipulated for
6.4 Fatigue test
dense graded asphalt concrete mixtures (AC16
In April 2014, the resistance to fatigue cracking was D) and the binder contents obtained in the field
carried out under controlled stress mode at 200C, were quite similar than that of the job mix for-
using a harvesine load waveform with a loading mula. Thus, these types of mixes were successfully
259

TIS2017_Book.indb 259 2/8/2017 8:03:41 PM

 reproduced in a modified batch plant with a sec- Bocci M, Grilli A, Cardone F, & Graziani A. (2011). A
ondary flow-parallel drum. study on the mechanical behaviour of cement–bitumen
− The average deformation slope (WTSair) for treated materials. Construction and Building Materials,
HWMRA 2.5% reached at a temperature of 50°C, 25(2), 773–778. doi:10.1016/j.conbuildmat.2010.07.007
Botella R, Miró R, Díaz P, Ramírez A, Guisado F, & Moreno
was of 0.068 (mm/load cycles) and proportional E. (2015). Sustainable Urban Surface Asphalt Layers.
rut depth (PRDair) of 1.42%, resulting in good Springer, 11, 619–627. doi:10.1007/978–94–017–7342–3
results. However, the wheel tracking test conducted Bowering RH. (1970). Properties and behavior of foamed
at 60°C can be considered hardly representative, bitumen mixtures for road buildings. Canberra, Aus-
for both base and binder courses since these mix- tralia. In Proceedings of the 5th Australian Road
tures are commonly exposed to lower heating tem- Research Board Conference (pp. 38–57).
peratures compared to wearing courses. Chiu CT, Hsu TH, & Yang WF. (2008). Life cycle assess-
− Cores extracted two years after construction, ment on using recycled materials for rehabilitating
showed that both HWMRA (2.5% and 3%) had asphalt pavements. Resources, Conservation and Recy-
cling 545–556. doi:10.1016/j.resconrec.2007.07.001
a similar mechanical performance to the conven- Dinis-Almeida M., Castro-Gomes J., & Antunes, M.D.L.
tional hot asphalt mixture, however, HMA cores (2012). Mix design considerations for warm mix recy-
showed higher stiffness modulus and ITSdry val- cled asphalt with bitumen emulsion. Construction
ues than that of the other two half-warm mixes. and Building Materials, 28(1), 687–693. doi:10.1016/j.
Despite this, HWMRA (3.0%) cores displayed a conbuildmat.2011.10.053
significant increase of the stiffness modulus and García JL, Guisado F., Paez A., & Moreno, E. (2011).
ITS dry two years after construction of the serv- Reciclado total de mezclas bituminosas a bajas temper-
ice road, which therefore could be related to the aturas. Una propuesta para su diseño,caracterización
evolution of their mechanical properties for these y producción. ASEFMA, 1–12.
García JL, & Lucas, F. (2014). Mezclas templadas con
types of mixtures manufactured with emulsion. reutilización del RAP con tasa alta y tasa total. Apli-
− It appears the decrease of the stiffness modulus cación, Experiencias reales y resultados. Asfalto y
and ITS at 15°C was largely produced due to Pavimentación, IV(14), 51–65.
pavement cores extracted at center of the travel Godenzoni C, Cardone F, Graziani A, & Bocci M. (2016).
lane, so these cores showed lower density than The Effect of Curing on the Mechanical Behavior of
those values obtained in 2012. Cement-Bitumen Treated Materials. In 8th RILEM
International Symposium on Testing and Characteri-
This project is a firm commitment for design- zation of Sustainable and Innovative Materials (pp.
ing, manufacturing and applying new cleaner pro- 879–890). doi:10.1007/978–94–017–7342–3
duction technologies using total recycled contents Kandhal PS, & Mallick RB. (1997). Pavement Recycling
of RAP at low temperatures. It is therefore con- Guidelines for State and Local Governments Partici-
cluded that these findings encourage continuing pant’s Reference Book.
endeavors in developing these sustainable asphalt Olard F, Le Noan C, Beduneau E, & Romier A. (2008).
mixtures for road pavements. Current research is Low energy asphalts for sustainable road construction.
In Proceedings of the 4th Eurasphalt and Eurobitume
being carried out to further evaluate the long-term congress (p. 12). Copenhaguen, Denmark: European
in-situ performance of this technology. Asphalt Pavement Association (EAPA).
Rubío MC, Moreno F, Martínez-Echevarría J, Martínez
ACKNOWLEDGEMENTS G, & Vázquez J.M. (2013). Comparative analysis of
emissions from the manufacture and use of hot and
half-warm mix asphalt. Journal of Cleaner Produc-
I would like to express my fully acknowledgement to tion, 41(2013), 1–6. doi:10.1016/j.jclepro.2012.09.036
European Union’s Seventh Framework Programme Serfass J., Poirier J., Henrat J., & Carbonneau, X. (2004).
for Research, Development and Demonstration Influence of curing on cold mix mechanical performance.
under grant agreement number 607524, Marie Curie Materials and Structures, 37 (April 2003), 365–368.
Initial Training Network actions (ITN) FP7-PEO- Swaroopa S, Sravani A, & Jain PK. (2015). Comparison
PLE-2013-ITN for its financial support. Moreover, of mechanistic characteristics of cold, mild warm and
we would also like to express our acknowledgement half warm mixes for bituminous road construction.
to Eduardo Ortega of Sacyr’s R&D laboratory. Indian Journal of Engineering and Materials Sciences,
22, 85–92.
Thenoux G, & Jamet A. (2002). Foamed Asphalt Technol-
REFERENCES ogy. Revista Ingeniería de Construcción, 17(2), 84–92.
Tutu KA, & Tuffour YA. (2016). Warm-Mix Asphalt
AENOR. Asociación Española de Normalización y and pavement sustainability: a review. Journal of Civil
Certificación. Normas UNE EN 12697. Métodos de Engineering, 6, 84–93. doi:http://dx.doi.org/10.4236/
ensayo para mezclas bituminosas en caliente. Madrid. ojce.2016.62008
Blankendaal T, Schuur P, & Voordijk H. (2013). Reduc- Zaumanis M, & Mallick RB. (2014). Review of very high-
ing the environmental impact of concrete and asphalt: content reclaimed asphalt use in plant-produced pave-
a scenario approach. Journal of Cleaner Production, ments: state of the art. International Journal of Pavement
66 (2014), 27–36. doi:10.1016/j.jclepro.2013.10.012 Engineering, 1–17. doi:10.1080/10298436.2014.893331

260

TIS2017_Book.indb 260 2/8/2017 8:03:42 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Fiscal commitments to encourage PPP projects in

transport infrastructure

N. Nose
World Bank, Washington, DC, USA

C. Queiroz
Claret Consulting LLC, Washington, DC, USA

M. Nose
International Monetary Fund, Washington, DC, USA

ABSTRACT: This paper aims at analyzing how different types of fiscal commitments to Public Private
Partnerships (PPP) can be effectively utilized in the preparation and implementation of PPP projects in
transport infrastructure. The instruments include risk mitigation instruments offered by international
financial institutions and public financial support for infrastructure projects. In PPP projects in roads,
it is important to identify risks and allocate responsibility for the identified risks between the public and
private sectors. In particular, allocating revenue related risks is critical because it involves uncertainty
for future demands. However, not all public sector agencies are capable to take those risks, especially in
developing economies, due to their insufficient fiscal condition and limited track record of similar type
of projects. In order to attract private sector investment to PPP road projects in those countries, the fiscal
commitments discussed in the paper could be utilized to mitigate the risks.

1 INTRODUCTION However, not all public sector agencies are

able to accept such risks. Especially in develop-
Managing risks is crucial for the preparation and ing economies, the public partner may not be suf-
implementation of PPP program in roads and ficiently creditworthy or does not have a proven
other modes of transport. The main risks of PPP track record in the eyes of private financiers to be
projects are those affecting gross revenue of the able to attract private investments without external
project, in addition to those leading to significant support. In those cases, several types of possible
cost increases, such as changes in design during financial instruments can be used to facilitate the
construction. Revenue related risks usually reflect mobilization of private capital to finance transport
uncertainty in both the predictability of future PPP projects in a country where financing require-
demand (e.g., traffic volumes) and the willingness ments substantially exceed budgetary or internal
of users to pay tariffs. resources.
In order to manage risks associated with PPP Several International Financial Institutions
programs, parties concerned should identify risks (IFIs) and other international or bilateral organi-
for each stage of the project and allocate respon- zations can support those countries with financ-
sibility for the identified risks between the public ing instruments consisting of risk mitigation
and private sector. The best approach is not to try instruments and public sector financial support
to transfer all risks to the private sector, as this including, for example, the World Bank Group
would result in less interest by the private sector (International Development Association, Interna-
or a much higher cost to the public sector. As a tional Bank for Reconstruction and Development,
result, risk allocation is a very important compo- International Finance Corporation, Multilateral
nent in the assessment of any PPP project. The Investment Guarantee Agency) and the Asian
general rule is that risks need to be allocated to the Development Bank. Such organizations can pro-
party that is best capable to manage them. This vide resources in several areas, including private
means that the government would need to take infrastructure financing products, such as viability
some risks that it can manage better or because gap funds, financial intermediary loans, equity,
the costs of the private sector assuming such risks and partial risk guarantees (World Bank, 2014).
would be too high. Some of the IFIs can also finance Special Purpose

261

TIS2017_Book.indb 261 2/8/2017 8:03:44 PM

 Vehicles (SPV) on a risk sharing basis. In the par- allocated, their likelihood of occurrence, and impact.
ticular case of the World Bank Group, this can be If the private sector is transferred a risk that it can-
done through its International Finance Corpora- not control (for example inflation being higher than
tion (Web-1, 2016). Another World Bank affiliate, forecast) it will either take a very conservative sce-
the Multilateral Investment Guarantee Agency nario (such as assuming a very high inflation rate)
(Web-2, 2016) (Web-3, 2014) offers political risk or simply not accept it (and therefore will not make
insurance. Several types of possible financial sup- any proposal, thus reducing competition). The risk
port are discussed in the next paragraphs. allocation exercise requires a very good understand-
ing of the market and project finance principles in
order to allocate the risk in a way that balances the
2 FINANCIAL INSTRUMENTS, RISK public and private sector concerns and interests
SHARING AND GUARANTEES (Queiroz and Lopez-Martinez, 2012).
The preparation of a risk matrix would help the
In the particular case of roads, a study (Queiroz government to decide which risk should be allo-
and Kerali, 2010) suggests that toll road traffic cated to which party. The risk matrix should be
forecasts are characterized by large errors and prepared with a legal perspective in mind because
considerable optimism bias. As a result, financial it should provide the basis for drafting the PPP
specialists need to ensure that transaction structur- legal agreement/ concession agreement.
ing remains flexible and retains liquidity such that The risk allocation matrix should be updated
material departures from traffic expectations can and refined as project preparation evolves. It is
be accommodated. usually prepared with the support of transaction
Risks should be identified for each stage of a experts and in consultation with potential bidders.
project, and responsibility should be allocated for Ultimately the risk allocation will determine if a
the identified risks. PPP project is financeable (i.e., lenders will not
As PPP are legally long-term contractual agree- finance it if they believe the risk allocation is not
ments, responsibilities should be clearly defined as appropriate), so the public sector should remain
they will determine the costs that the public and flexible when designing such a matrix (Queiroz and
private partners will ultimately pay. For example Lopez-Martinez, 2012).
construction risk is usually transferred to the pri- The Global Infrastructure Hub (GIH) provides
vate sector, which means that it will be responsible good examples of risk matrices, showing the alloca-
(and will not be able to claim additional compen- tion of risks between the public and private sectors
sation) for delays and cost-overruns in complet- in typical PPP transactions, along with mitigating
ing the works. The best approach is not to try to measures, within the transport, energy and water
transfer all risks to the private sector, as this would and sanitation sectors (GIH, 2016).
result in less interest (or no interest, i.e., no bid- In road projects in developing economies, there
ders) by the private sector or a much higher cost may be a perception of a risky environment for
to the public sector. As a result, risk allocation is private investment, and the use of risk mitigation
a very important component in the assessment of instruments can help reduce the risk perception
any PPP project (Robert, 2009). and facilitate private sector investment.

2.1 Risk matrix 2.2 Risk mitigation

A good practice in preparing risk matrices is to Several instruments can be used to facilitate the
adopt the following structure for each potential mobilization of private capital to finance transport
risk of the project: infrastructure PPP projects in developing econo-
mies, where financing requirements substantially
a. Description of the risk
exceed budgetary or internal resources. Risk miti-
b. Proposed allocation of the risk (usually two col-
gation instruments are financial instruments that
umns – ‘Grantor’ and Concessionaire’- and one
transfer certain defined risks from project finan-
of them gets checked for a particular risk)
ciers (lenders and equity investors) to creditworthy
c. Comments
third parties (guarantors and insurers) that have a
The general rule is that risks need to be allocated better capacity to accept such risks. These instru-
to the party that is best capable to manage them. ments are especially useful when the public partner
This means that the government would need to take is not sufficiently creditworthy or do not have a
some risks that it can manage better or because proven track record in the eyes of private finan-
the costs of the private sector assuming such risks ciers to be able to attract private investments with-
would be too high. The private sector will price the out support. The advantages of such instruments
risk of the project based on how individual risks are are multifaceted (Matsukawa and Habeck, 2007):

262

TIS2017_Book.indb 262 2/8/2017 8:03:44 PM

 a. The public sector is able to mobilize domestic
and international private capital for infrastruc-
ture implementation, supplementing limited
public resources.
b. Private sector lenders and investors will finance
commercially viable projects when risk mitiga-
tion instruments cover those risks that they per-
ceive as excessive or beyond their control.
c. Governments can share the risk of infrastruc-
ture development using limited fiscal resources
more efficiently by attracting private investors Figure 1. Structure of a highway concession contract
and World Bank Project-based guarantee (Queiroz,
rather than having to finance the projects them-
2006).
selves, assuming the entire development, con-
struction, and operating risk.
illustration also applies to other modes of trans-
Commonly used risk mitigation instruments port infrastructure.
include guarantees and insurance products. Guar- Political risk insurance typically requires a speci-
antees typically refer to financial guarantees of fied period during which claims filed by the insured
debt that cover the timely payment of debt serv- are to be evaluated, before payment by the insurer
ice. Procedures to call on these guarantees in the is made. PRIs usually insure foreign direct invest-
event of a debt service default are usually relatively ments against losses related to single or a combi-
straightforward. In contrast, insurance typically nation of: Currency inconvertibility and transfer
requires a specified period during which claims restrictions; Expropriation; War, civil disturbance,
filed by the insured are to be evaluated, before terrorism, and sabotage; Breach of contract; and
payment is made by the insurer. Examples of risk Non-honoring of sovereign financial obligations
mitigation instruments available that could apply (Web-5, 2014). Unlike project-based guarantees,
to transport infrastructure PPP projects include PRIs do not require a counter-guarantee from a
(Matsukawa and Habeck, 2007): host government. Premium rates are decided on
a. Partial Risk Guarantees, currently known as a per-project basis and vary by country, sector,
project-based guarantees; and transaction and the type of risk insured. PRIs are
b. Political Risk Insurance (PRI). provided by private PRI providers, national export
credit agencies (ECAs) and multilateral agencies
Project-based guarantees are designed to pro- such as MIGA. Figure 2 provides an illustration
vide risk mitigation with respect to key risks that of how a PRI can apply to a highway concession
are essential for the viability of specific invest- contract; such illustration also applies to other
ment projects. Project-based guarantees generally transport infrastructure.
cover government-related risks, which are risks A recent case that utilized its guarantee instru-
within the control of the government and pub- ment for a road project is MIGA’s support to
lic entities. There are two main types of project- Vietnam’s BT20 National Highway 20, providing
based guarantees: loan guarantees and payment non-honoring of sovereign financial obligations
guarantees. Loan guarantees cover defaults of cover of $500 million to the project’s lenders to
debt service payments and could be granted for cover a 15-year loan and associated costs to finance
public sector or private sector projects. Loan the road’s construction works (Web-7, 2014). The
guarantees could protect commercial lenders project consists of the rehabilitation and upgrad-
financing a private sector project from debt serv- ing of the National Highway 20 that extends from
ices defaults caused by government actions or the Dong Nai to Lam Dong provinces of Vietnam,
inactions. This type of guarantee was previously under a build-transfer contract between the BT20
known as a partial risk guarantee (PRG) (World Consortium and the Ministry of Transportation.
Bank, 2016). They typically cover the full amount BT 20, Cuu Long Joint Stock Company, is a con-
of debt. Payment is made only if the debt default sortium comprised of Dong Mekong Construc-
is caused by risks specified under the guarantee. tion Production Trade Service Co., PetroVietnam
Such risks are political in nature and are defined Construction Joint Stock Company, Cuu Long
on a case-by-case basis. Project-based guarantees Corporation for Investment, Development and
are offered by multilateral development banks Project Management of Infrastructure, and Build-
(World Bank, 2016) and some bilateral agen- ing Materials Corporation No. 1 Co (Web-8, 2014).
cies. Figure 1 provides an illustration of how a Goldman Sachs (loan arranger and lending syn-
project-based guarantee can apply to a highway dicate member) has provided loans to BT 20 to
concession contract. Mutatis mutandis, such support the construction financing of the phase

263

TIS2017_Book.indb 263 2/8/2017 8:03:44 PM

 Figure 2. Structure of a highway concession contract
and MIGA political risk insurance (Web-6, 2014). Figure 3. Structure of MIGA political risk insurance to
BT20 National Highway 20 Project.

1 of the BT20 National Highway 20 Project with

the unconditional guarantee provided by Ministry quality of the service it is contracted to provide.
of Finance. Figure 3 provides an illustration of A third option is for governments to pay a fee per
MIGA’s support to the BT20 National highway 20. user, such as the number of vehicles using a toll
World Bank Group guarantee instruments road (World Bank, 2012). The latter is known as
(Web-9, 2016) have been relatively underused, shadow toll, a concept created for Design, Build,
partly due to policy constraints. Recent reforms, Finance and Operate—DBFO (Bain and Wilkins,
however, have made them more flexible and acces- 2003) roads in the United Kingdom, and has also
sible helping the World Bank Group (IBRD/IDA, been used in other countries such as Finland and
IFC and MIGA) to work together more effectively, Portugal (Web-10, 2014).
tackling clients’ needs and catalyzing private sec- Subsidies to PPPs help make sure projects that
tor participation (World Bank, 2013). Accordingly, will produce a net economic or social gain can be
this should contribute to mobilizing support to commercially financed. There are two broad rea-
potential PPP transport projects. sons why an economically justified project may not
be financially viable. First, infrastructure projects
can create public benefits that are not reflected in
2.3 Public sector financial support
the price consumers are willing to pay for the serv-
Road and other transport PPP projects in develop- ice, such as a toll road that creates third-party ben-
ing economies should seek to minimize the need for efits by increasing mobility and lowering vehicle
public financial support in order to maximize the emissions. Second, user fees can be deliberately set
benefits of a concession relative to its costs. How- at a low level to keep them at a socially acceptable
ever, public financial support may be appropri- level (Web-10, 2014).
ate if it helps ensure the mobilization of required When estimating the minimum required levels
amounts of private capital. of subsidies and/or PPP guarantees that will make
Overall, the type and level of government finan- a PPP project attractive to private investors, coun-
cial contribution to a PPP project should be lim- tries should note that the provisions of subsidies
ited to what is required to attract private financing and guarantees without proper assessment of the
and promote a successful project. fiscal cost may lead to budget shortfalls in the future
Following the recent global financial crisis, because of direct or contingent liabilities. In this
the higher cost and lower availability of loans regard, governments should establish a high qual-
and an increase in risk aversion, governments ity Public Financial Management (PFM) system to
have had in several cases to resort to increased select projects with high value-for-money and to
support to PPPs to enable them to go forward. ensure the long-term commitments will not over-
This has taken the form of subsidies (or grants burden the public budget in future (Engel, Fischer,
to the concessionaire) as well as increased risk- and Galetovic, 2014). Government officials in
bearing. developing countries need a relatively simple, user
A PPP subsidy is a direct government contribu- friendly tool that assesses potential fiscal costs and
tion or grant to pay for a portion of costs that is risks in PPP projects quickly so several options can
not repaid by the concessionaire. Governments can be tested in a short period at low cost. The financial
provide subsidies by making upfront cash contri- models included in the Toolkit for PPP in Roads
butions to pay for capital costs (i.e., construction and Highways and/or PPP Fiscal Risk Assessment
subsidies). Alternatively, once a project has been Model (P-FRAM) are useful tools that can be used
built, governments can make regular payments to for such purpose, with relatively small training
the private company based on the availability and required (Web-11, 2016; Web-13, 2016).

264

TIS2017_Book.indb 264 2/8/2017 8:03:44 PM

 2.4 Risk sharing and forms of PPP turning them into successful PPP projects, such as
the Perpignan-Figueras Rail Concession.
Different forms of transport infrastructure conces-
Developing countries tend to prefer PPPs to
sion contracts, such as availability fee, shadow tolls,
conventional public procurement in financing
Build-Operate-Transfer (BOT), and Build-Own-
infrastructure to relieve their budget pressure. This
Operate (BOO), provide increased risk transfers to
highlights the importance of fiscal prudence in
the private sector. Under shadow tolls, BOT and
using PPPs for infrastructure investments. Coun-
BOO the demand risks are borne by the private part-
tries assuming long-term liabilities (e.g., availabil-
ner, but under shadow tolls the concessionaire does
ity payments) and issuing PPP guarantees (e.g.,
not assume the risks associated with toll collection
project-based guarantees) should improve their
(e.g., sensitivity of demand to the level of toll rates).
PFM system to ensure the appropriate project
The cost of public sector risk bearing is an
appraisal and selection, for example by utilizing
important element to consider when requesting
available user-friendly financial models.
and evaluating PPP proposals. One of the key
premises that should be considered using PPPs is
the optimum allocation of project risks to the part-
ner that is best able to manage them cost effectively. REFERENCES
Consequently, to truly assess the impact of private
sector involvement, governments need to adopt an Aldrete R., Bujanda A. & Valdez-Ceniceros G.A. 2010.
approach to quantify the short-term impacts of Valuing Public Sector Risk Exposure in Transporta-
the project on the public budget and the long-term tion Public-Private Partnerships, at: http://utcm.tamu.
potential cost of the risks the government chooses edu/publications/final_reports/Aldrete_08–41–01.pdf
Bain R. 2009. Error and Optimism Bias in Toll Road
to retain (Aldrete et al, 2010). Traffic Forecasts. Transportation, 469–482.
Even countries more advanced in the implemen- Bain R. and Wilkins M. 2003. The Evolution of DBFO
tation of PPP projects have resorted to financial Payment Mechanisms: One More for the Road at:
instruments to attract private investors. This is the http://www.robbain.com/Evolution%20of%20DBFO.
case, for example, of France and Spain that jointly pdf
launched the Perpignan—Figueras Rail Conces- Engel, E., Fischer, R. & Galetovic, A. 2014. The Eco-
sion, which provides a link between French and nomics of Public-Private Partnerships: A Basic Guide,
Spanish rail systems, thus reducing travel times and Cambridge University Press, NY. USA.
transport bottleneck (Web-12, 2016). The project, European Commission. 2004. European Commission
Resource Book on PPP Case Studies Perpignan-
which combines high-speed trains (travelling at up Figueras Rail Concession, France & Spain, at: http://
to 350 km) as well as freight convoys (moving at ec.europa.eu/regional_policy/sources/docgener/
120 km/h), received a state subsidy covering 57% guides/pppresourcebook.pdf
of the construction cost, as well as bank guaran- Global Infrastructure Hub-GIH. 2016. Allocating Risks
tees (European Commission, 2004). in Public-Private Partnership Contracts. Sydney,
Australia. http://www.globalinfrastructurehub.org/
content/uploads/2016/07/160610-GIHub-Allocating-
Risks-in-PPP-Contracts-2016-Edition.pdf
3 SUMMARY AND CONCLUSIONS Matsukawa T. & Habeck O. 2007. Review of Risk Miti-
gation Instruments for Infrastructure Financing and
There are financial instruments that can help make Recent Trends and Developments. Trends and Policy
transport infrastructure PPP projects more attrac- Options No. 4, Public-Private Infrastructure Advi-
tive to the private sector, potentially resulting in a sory Facility (PPIAF) and the World Bank, at http://
more competitive selection of the private partner. www.ppiaf.org/ppiaf/publications/Trends%20and%20
Policy%20Options
Even if infrastructure projects in developing Queiroz, C. 2006. The Potential of Private Financing to
economies are economically and socially justified, Enhance Road Infrastructure in the Baltic States. In
often they are not, per se, able to attract private 26th International Baltic Road Conference, Kures-
investors. However, they may become feasible saare (Saaremaa), Estonia, 28–30 August 2006. http://
PPP projects if appropriate support is given to the www.bjrbe.vgtu.lt/news/news002.php
project, particularly through financial instruments Queiroz, C. & H. Kerali. 2010. A Review of Institu-
such as guarantees and subsidies. Potential PPP tional Arrangements for Road Asset Management:
projects in developing economies may want to take Lessons for the Developing World. World Bank
advantage of one or more of the financial instru- Transport Paper No. TP-32: at http://go.worldbank.
org/6HDCYBMRT0
ments discussed in the paper, such as capital subsi- Queiroz, C. & A. Lopez-Martinez. 2012. Legal Frame-
dies or through proper involvement of multilateral works for Successful Public Private Partnerships.
partners. in “The Routledge Companion to Public-Private
Even more advanced economies, such as France Partnership”, at: http://www.routledge.com/books/
and Spain, have granted subsidies to projects, details/9780415781992/

265

TIS2017_Book.indb 265 2/8/2017 8:03:44 PM

 World Bank. 2012. Best Practices in Public- Web-5: http://www.miga.org/investmentguarantees/
Private Partnerships Financing in Latin Amer- index.cfm, consulted 10 August 2014.
ica: the role of subsidy mechanisms, at: http:// Web-6: http://siteresources.worldbank.org/EXTGUAR-
api.ning.com/files/JgnTHgt50ouqoEiQqvTYC- ANTE/Resources/structure.pdf, consulted 15 August
edJiYj9tuAn-qH*A6j39vd-fQ72wjVqJmhEM2- 2014.
CjumJlv9ZKv9b7 s7rU0Obc2XfiYOsVHsnFZAA4/ Web-7: http://www.miga.org/news/index.cfm?aid = 3660,
BestPracticesroleofsubsidiesmechanisms.pdf consulted 15 August 2014.
World Bank. 2013. Enhancing the World Bank’s Web-8: http://www.miga.org/documents/BT20_
Operational Policy Framework on Guarantees. National_Highway_ESRS_April_24_2013.pdf, con-
http://consultations.worldbank.org/consultation/ sulted 15 August 2014.
modernizing-operational-policy-guarantees-public- Web-9: http://web.worldbank.org/external/default/main?
consultations menuPK = 64143540&pagePK = 64143532&piPK = 6
World Bank. 2014. World Bank Group Guarantee Prod- 4143559&theSitePK = 3985219, consulted 15 August
ucts, at: http://treasury.worldbank.org/bdm/pdf/Bro- 2016
chures/WBG_Guarantees_Matrix.pdf Web-10: http://pppnetwork.ning.com/page/financing-
World Bank. 2016. World Bank Group Guarantee Prod- public-private-partnerships-best-practices-in-latin-
ucts, at: https://ppp.worldbank.org/public-private- ame, consulted 15 March 2014
partnership/sites/ppp.worldbank.org/files/documents/ Web-11: http://www.ppiaf.org/sites/ppiaf.org/files/docu-
wbg-guarantees_final_0.pdf ments/toolkits/highwaystoolkit/index.html, consulted
Web sites: Web-1: http://www1.ifc.org/, consulted Sep- April 2016
tember 2016. Web-12 http://www.eiffage.com/en/le-groupe-eiffage/
Web-2: http://go.worldbank.org/3FAJ6 LS510, consulted a65_motorway.html, consulted 2016
September 2016. Web-13: http://www.imf.org/external/np/fad/publicinvest-
Web-3:http://www.miga.org/whoweare/index.cfm?stid= ment/pdf/PFRAMmanual.pdf, consulted April 2016.
1792, consulted 5 August 2014.

266

TIS2017_Book.indb 266 2/8/2017 8:03:44 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Mix design and volumetric analysis of hot recycled bituminous mixtures

using a bio-additive

E. Bocci
Università eCampus, Novedrate (CO), Italy

F. Cardone
Università Politecnica delle Marche, Ancona, Italy

A. Grilli
Università degli Studi di San Marino, San Marino, San Marino Republic

ABSTRACT: Nowadays, the use of Reclaimed Asphalt (RA) as a constituent material for hot Asphalt
Concretes (AC), is gaining increasing interest because of the important technical and environmental
advantages. In addition, the reuse of the old bitumen from the RA, allows the required amount of new
bitumen to be reduced with clear economic returns. However, only a small percentage of RA can be recy-
cled in new ACs (typically less than 30%), as the excess of oxidized aged bitumen may lead to a brittle
behavior of the pavement layer. In order to avoid this issue, when high amount of RA are used, specific
additives are recommended in order to restore the bitumen characteristics leading to a mixture with the
expected mechanical properties. The present paper deals with the use of a bio-based Additive (A) for the
production of AC for binder layer containing high quantities of RA. In particular, the study focuses on
the mix design phase, highlighting the effects of RA and A on the volumetric properties of the mixtures.
The experimental program included four ACs with no RA, 40% of RA, 40% of RA and additive A,
50% RA and additive A respectively. Results showed that the optimum bitumen content decreased when
decreasing RA content or when adding A. Moreover, the use of A allowed obtaining the desired volumet-
ric properties even with a significantly lower amount of virgin bitumen to be added to the mixture without
penalizing its mechanical response.

1 INTRODUCTION cation (Petersen & Glaser 2011, Read & Whiteoak

2003). Consequently the binder becomes harder
The ordinary road pavement maintenance proce- and more brittle (Stimilli et al. 2014), thus more
dure, consisting of milling of distressed asphalt lay- prone to cracking, with significant disadvantages
ers before overlaying, determines the production of on the mix behavior (Karlsson & Isacsson 2006,
large amounts of Reclaimed Asphalt (RA) (Bocci Molenaar et al. 2010).
et al. 2010). As this material includes precious In addition, the maximum amount of RA to
components as bitumen and mineral aggregate, be reused in AC production depends not only on
it should not be relegated to a waste product but the ability to correct the physicochemical charac-
should be advantageously exploited through hot teristics of the aged bitumen but also on the type
and cold recycling techniques (Karlsson & Isacs- of processing of the mix plant (Frigio et al. 2014).
son 2006, Al-Qadi et al. 2007). Particularly, hot For these reasons, the amount of RA which is
recycling of RA allows important economic and typically recycled in hot mixtures does not exceed
environmental benefits to be achieved, through the 30%.
reduction of aggregate and bitumen supplying and To enable an increase in RA content, the use of
thus the preservation of natural resources (Grilli specific additives is strongly recommended. The
et al. 2015, Olard et al. 2008). specific additives are intended to mobilize and
However, it is typically supposed worldwide restore the mechanical properties of the aged bitu-
that Asphalt Concretes (AC) with high RA con- men, in order to achieve adequate blending with
tent do not provide good performances because the fresh virgin binder and suitable workability
of the aged binder that RA contains. The ageing and mechanical performance of the mix (Chen
mechanism is complex and it is accompanied by et al. 2007, Grilli et al. 2013, Shen et al. 2007, Zau-
negative effects as oxidation and chemical modifi- manis et al. 2014).

267

TIS2017_Book.indb 267 2/8/2017 8:03:47 PM

 The present research project deals with the use ture weight), according to the Italian technical
of a specific bio-based Additive (A) to produce AC specifications;
using a high amount of RA, without penalizing – compaction of 3 specimens for each mixture
the mix performance and complying with the Ital- by means of a GC at 180 gyrations, in order
ian specifications. In particular, the study focuses to evaluate the overall compaction curve (EN
on the mix design phase, in order to highlight the 12697-10);
effects of RA and A on the volumetric properties – calculation of the bitumen content (optimum
of the mixtures. bitumen content) that allows 4% of air voids
content (Vm) to be obtained at 100 gyrations
and checking of Vm at 10 and 180 gyrations to
2 OBJECTIVE AND EXPERIMENTAL respect the Italian specification.
PROGRAM
Finally, volumetric and mechanical characteris-
tics were evaluated in detail on the design mixtures
This paper deals with hot recycling of RA using
following the present protocol:
a bio-based Additive (A). This study provides the
volumetric mix design of four ACs containing no – production of the mixture using the calculated
RA, 40% RA, 40% RA plus A, 50% of RA plus A, (optimum) bitumen content;
respectively. The main objective was to determine – compaction of 3 specimens for each mixture by
the effect of RA content and A on the volumetric means of a GC at 180 gyrations for the volu-
properties of the AC. In addition, the effectiveness metric analysis and the study of the compaction
of A to produce AC with high percentages of RA curve (EN 12697-10);
complying Italian specifications (Autostrade per – compaction of 3 specimens by means of a GC
l’Italia 2008, ANAS 2010) was evaluated. at 100 gyrations and determination of Indirect
To this aim, the experimental program included Tensile Strength ITS (EN 12697-23) and Indi-
the mix design and testing of four mixtures rect Tensile Coefficient ITC.
including:
ITC is an empirical mechanical parameter which
– no RA (00RA), as reference AC; can be related to the stiffness properties of the mix-
– 40% of RA (considering its aggregate weight) by ture, and be calculated according to Eq. 1:
granular mixture weight (40RA);
– 40% of RA and 6% of A by aged bitumen weight π ⋅ D ITS (1)
(40RA6A); ITC =
2 ⋅ δ v ,max
– 50% of RA and 6% of A by aged bitumen weight
(50RA6A).
where D is the specimen diameter and δv,max is the
Table 1 summarizes the compositions of the vertical deformation corresponding to the load
mixtures. peak during ITS test.
Mixtures were compacted by means of a Gyra-
tory Compactor (GC) according to EN 12697-31.
The mix design procedure consisted in the volu-
3 MATERIALS
metric analysis of the mixtures, carried out through
the following steps: 3.1 Aggregate
– selection of appropriate constituent materials Virgin limestone aggregates, limestone filler and
and aggregate gradation; two fractions of RA were sampled in a selected
– production of four mixtures using different mix plant and characterized in terms of grada-
bitumen contents (4.3, 4.8, 5.3 and 5.8% by mix- tion (Fig. 1). In particular, the RA after bitumen
extraction (gradation of the aggregates in the
RA) was sieved adopting the wet sieving method.
Table 1. AC composition. The bitumen content in the RA, measured by mix
weight, was 4.6% in the fine fraction and 3.2%
RA content A content in the coarse one. In order to optimize plant RA
Mixture [% by mix] [% by aged bit.] management it was used approximately 2/3 of fine
RA and 1/3 of coarse RA for the mix design.
00RA 0 0
The bitumen recovered from RA according to
40RA 40 0
the procedure defined by EN 12697-1 showed a
40RA6A 40 6
penetration of 18 dmm (EN 1426) and a softening
50RA6A 50 6
point of 75°C (EN 1427).

268

TIS2017_Book.indb 268 2/8/2017 8:03:47 PM

 Figure 2. Mix gradations.

Figure 1. Grading of aggregates and RA (after bitu- Table 3. Virgin bitumen properties.
men extraction).
Property Unit Norm Value
Table 2. Aggregate proportioning.
Penetration at 25°C dmm EN 1426 53
% by weight Softening point °C EN 1427 48
Loss of mass after RTFOT % EN 12607-1 0.08
Fraction 00RA 40RA 50RA
Retained pen. after RTFOT % EN 1426 55
8/16 Gc 90–20 45 34 34 Increase of softening point °C EN 1427 16
4/10 Gc 85–15 7 3 2
4/8 Gc 90–10 5 0 0
0/4 Gf 90 35 21 12 Table 4. Bio-additive properties.
16 RA 0/12 0 13 16
12 RA 0/8 0 27 34 Property Unit Value
Filler 8 2 2
Flash point °C >295
Dynamic viscosity at 25°C mPa⋅s 71
The different fractions of RA and virgin aggre- Kinematic viscosity at 0°C mm2/s 355
gate were proportioned in order to obtain a fixed Kinematic viscosity at 40°C mm2/s 43
gradation curve complying with the aggregate Kinematic viscosity at 100°C mm2/s 9
band for a binder course (ANAS 2010). Density g/cm3 0.93
Table 2 shows the percentage of each fraction in
the different ACs and Figure 2 plots the gradation
curves. Table 5. Physical properties of the binders.

Property VB RB RB5A
3.2 Virgin bitumen
The virgin binder used in the present study was a Penetration at 25°C [dmm] 53 18 32
50/70 bitumen (EN 12591). The characteristics of Softening point [°C] 48 75 65
the bitumen are shown in Table 3. Dynamic viscosity at 100°C [Pa⋅s] 4.9 19.4 13.0
Dynamic viscosity at 135°C [Pa⋅s] 0.54 1.35 0.95
Dynamic viscosity at 160°C [Pa⋅s] 0.18 0.30 0.27
3.3 Bio-Additive (A)
The recycling Additive (A) used in the present
study is a tall oil derived from the processing of bitumen recovered from RA (RB) and bitumen
pine wood in paper industry. It is a miscible, long- recovered from RA plus 5% of A (RB5A) by
lasting and sustainable bio-based material able weight have been determined (Table 5). Hence,
to mobilize oxidized RA binder and restore aged on the basis on the results obtained with 5%
bitumen physical properties. The characteristics of of A and previous experiences using empirical
A are shown in Table 4. tests, the optimum content of A was fixed as 6%
In order to identify the optimum dosage of A, by aged bitumen weight to achieve even better
the physical properties of Virgin Bitumen (VB), performance.

269

TIS2017_Book.indb 269 2/8/2017 8:03:47 PM

 3.4 Mixtures
The mix production process simulated the operative
procedure at the plant. In particular, for the 00RA
mixtures, both virgin aggregate and bitumen were
heated at 150°C. For the hot recycled ACs, 15% of
cold RA was added to the mix, whereas the virgin
aggregate and the remaining part of RA (25% for
mixtures 40RA and 40RA6A, 35% for the mixtures
50RA6A) were heated at 170°C. The bio-additive A
was directly sprayed on the RA particles (before the
heating in the case of hot-added RA).

4 MIX DESIGN RESULTS

Figure 5. Air voids content vs bitumen content, mix
4.1 Determination of the optimum bitumen
40RA6A.
content
Figures 3–6 show the air voids content as a func-
tion of bitumen dosage in order to calculate the
optimum bitumen content for the different mix-
tures. Bitumen introduced by RA was considered

Figure 6. Air voids content vs bitumen content, mix

50RA6A.

entirely reactivated. Note that 40% and 50% RA

carried 1.7% and 2.0% of aged bitumen in the mix-
ture, respectively.
Figure 3. Air voids content vs bitumen content, mix Establishing a target Vm of 4% at 100 gyrations,
00RA. an optimum bitumen content equal to 4.8% by mix
weight was identified for the mix 00RA (Fig. 3).
The mixture 40RA showed an optimum bitu-
men content equal to 5.0% by mix weight (Fig. 4),
corresponding to 3.3% of virgin bitumen and 1.7%
of aged bitumen from RA. This slight increase in
total binder content with respect to the mix 00RA
was probably due to the higher viscosity of the oxi-
dized binder included in the RA.
For the mix 40RA6A (Fig. 5), the optimum
bitumen content resulted equal to 4.8% by mix
weight (3.1% of virgin bitumen), denoting that
the bio-additive effectively mobilized the bitumen
from the RA, improving the compactability of the
mixture.
When increasing the RA content up to 50% (mix
50RA6A), the total bitumen content corresponding
Figure 4. Air voids content vs bitumen content, mix to 4% of Vm was found to be 5.0% (3.0% of virgin
40RA. bitumen) by mix weight (Fig. 6). This indicates

270

TIS2017_Book.indb 270 2/8/2017 8:03:48 PM

 that the presence of more 10% of RA determined Table 6. Volumetric properties of the design mixes.
a decrease of the mix workability, related to the
lower virgin bitumen/aged bitumen ratio. Vm,10 Vm,100 Vm,180 VMA,100 VFB,100
As it can be verified from the Figures, the design Mixture [%] [%] [%] [%] [%]
bitumen content values also allowed the fulfilment 00RA 12.1 4.5 3.1 15.8 71.5
of the volumetric limits provided by Italian Speci- 40RA 13.2 3.9 2.2 15.5 74.9
fications (Autostrade per l’Italia 2008) at 10 gyra- 40RA6A 13.4 4.4 2.8 15.5 72.4
tions (Vm between 11 and 15%) and 180 gyrations 50RA6A 13.9 4.5 2.7 15.7 71.5
(Vm ≥ 2%).

4.2 Analysis of the volumetric properties tion (N = 0), whereas m is the slope and represent
the mix workability.
The compaction curves, which describe the reduc- Figures 7 and 8 show the values of Vm,0 and m
tion in Vm when increasing the number of gyrations for the different ACs as a function of the bitumen
N, were determined according to EN 12697-10 for content respectively. It can be observed that, for all
all mixtures. In particular, the trend of these curves the mixtures, the intercept Vm,0 tended to decrease
is linear in a semi-logarithmic plane: when increasing bitumen content. This is due to
(2) the higher volume of binder that fills the voids
Vm Vm ,0 − m l g N
between the aggregate particles of the loose AC.
Differently, the slope m showed an increas-
where Vm,0 is the intercept of the y-axis and indi- ing trend with bitumen content up to a maxi-
cates the air voids at the beginning of the compac- mum value approximately in correspondence of
B = 5.3%. Over this value (B = 5.8%), the coeffi-
cient m exhibited a slight decrease. This indicates
that the compaction is promoted when increasing
binder content, as it helps to lubricate the aggre-
gate surface reducing the friction between the solid
particles. However, this effect is not evident for
high contents of binder, as the excess of bitumen
probably hinders the efficient stress distribution
through the aggregate skeleton.

5 CHARACTERIZATION OF THE
DESIGN MIXTURES
The mixtures including the optimum bitumen con-
tent were produced in order to verify the volumet-
ric and mechanical properties according to Italian
Figure 7. Coefficient Vm,0 vs bitumen content. Specifications.

5.1 Volumetric properties

Table 6 shows the Vm of the design mixtures at 10,
100 and 180 gyrations. Moreover, the Void in the
Mineral Aggregate (VMA) and the Void filled with
Bitumen (VFB) are presented.
It can be observed that, in general, the ACs
showed comparable volumetric properties, despite
the added virgin bitumen content was significantly
different (4.8%, 3.3%, 3.1% and 3.0 by mix weight
for the mixtures 00RA, 40RA, 40RA6A and
50RA6A, respectively). All the results fulfilled the
requirements at 10 (Vm between 11 and 15%), 100
(Vm between 3 and 6%) and 180 (Vm higher than
2%) gyrations. In addition, the values of VMA and
VFB proved to satisfy the requirements defined
by SHRP (Cominsky 1994), i.e. VMA higher than
Figure 8. Coefficient m vs bitumen content. 13% and VFB between 65 and 75%.

271

TIS2017_Book.indb 271 2/8/2017 8:03:48 PM

 The average value of the coefficients m and Vm,0 The ITS of the mixtures 00RA, 40RA6A and
of the design mixtures, determined through the lin- 50RA6A were comparable and met the accept-
ear interpolation of the compaction curves (in the ance range defined by Italian Specifications. On
semi-log plane), are presented in Figure 9. the contrary the mixtures 40RA exceeded the
From the graph it can be observed that the mix- maximum limit of ITS, denoting a not satisfying
tures 40RA and 50RA6A showed higher m and mechanical response. This indicates that, using
Vm,0 values with respect to the mixtures 00RA and the proper dosage of A (in this case 6% by aged
40RA6A. It can be assumed that this result is related bitumen weight), high percentages of RA, up to
to the higher total bitumen content (5.0% for 40RA 50%, can be used attaining satisfying mechanical
and 50RA6A vs 4.8% for 00RA and 40RA6A). performance and comparable to those of the tra-
Indeed, in the mix design it was observed that both ditional AC (i.e. 00RA mixture). In addition, the
m and Vm,0 increased when increasing bitumen con- 40RA and 50RA mixtures with A allow the high-
tent up to 5.3% (Figs. 8 and 9). est advantages to be reached in terms of amount
Comparing the results for the ACs with the of recycled materials.
same bitumen content (i.e. 00RA with 40RA6A
and 40RA with 50RA6A), it can be noted that the
presence of the bio-additive determined an increase
6 CONCLUSION
in m denoting the capability of A to improve the
mix workability.
The present paper deals with the use of a spe-
cific bio-based Additive (A) to produce AC using
5.2 Strength properties a high amount of RA, without penalizing the
mix performance and complying with the Italian
The ITS and ITC of the designed mixtures, com-
specifications. In particular, the study provides the
pared with the technical specification limits (ANAS
volumetric mix design of four ACs containing no
2010), are shown in Figure 10.
RA, 40% RA, 40% RA plus A, 50% of RA plus A,
respectively. The main objective was to determine
the effect of RA content and A on the volumetric
properties of the ACs and verify the design mix-
tures to comply with Italian specifications.
The analysis of the experimental results allowed
the following conclusions to be drawn:
– an optimum total bitumen content of 4.8% by
mix weight was identified for the mixes 00RA
and 40RA6A, whereas 5.0% by mix weight was
determined for the mixes 40RA and 50RA6A;
– as 40% and 50% RA carried 1.7% and 2.0% of
aged bitumen in the mixture (which was con-
sidered as entirely reactivated), the virgin bitu-
men content resulted 3.3%, 3.1% and 3.0%
respectively for the mixes 40RA, 40RA6A and
Figure 9. Coefficients m and Vm,0 for the design mixes. 50RA6A;
– the slope m of the compaction curves showed a
maximum value approximately in correspond-
ence of B = 5.3%, indicating that the compaction
is promoted when increasing binder content, as
it helps to reduce the friction between the aggre-
gate particles, but the excess of bitumen prob-
ably hinders the efficient stress distribution
through the solid skeleton;
– the designed mixtures showed comparable volu-
metric properties and fulfilled the requirements
by Italian Specifications and SHRP in terms of
Vm, VMA and VFB;
– the slope m for the designed mixtures increased
with bitumen content and in presence of A,
denoting the capability of the bio-additive to
Figure 10. ITS and ITC values for the design mixes. improve the mix workability;

272

TIS2017_Book.indb 272 2/8/2017 8:03:49 PM

 – using the proper dosage of A, good mechani- EN 1427. 2007. Determination of softening point.
cal performances can be obtained even for high European Committee for Standardization, Brussels,
contents of RA (up to 50%). Belgium.
Frigio, F., Pasquini, E., Partl, M.N. & Canestrari, F. 2014.
Future investigations will regard the full scale val- Use of reclaimed asphalt in porous asphalt mixtures:
idation of the hot recycled mixtures and the deter- laboratory and field evaluations. Journal of Materials
mination of the stiffness and fatigue properties. in Civil Engineering 27(7).
Grilli, A., Bocci, E. & Bocci, M. 2015. Hot Recycling
of Reclaimed Asphalt using a Bio-based Additive.
8th International RILEM SIB Symposium, Ancona,
REFERENCES Italy.
Grilli, A., Bocci, M., Cardone, F., Conti, C. & Giorgini,
ANAS 2010. Capitolato speciale di appalto, Parte E. 2013. Laboratory and in-plant validation of hot mix
2ª Norme tecniche per pavimentazioni stradali/ recycling using a rejuvenator. International Journal of
autostradali. Pavement Research and Technology 6(4): 364–371.
Al-Qadi, I., Elseifi, M. & Carpenter, S.H. 2007. Reclaimed Karlsson, R. & Isacsson, U. 2006. Material-related
asphalt pavement—A literature review. Report No. aspects of asphalt recycling: State-of-the-art. Journal
FHWA-ICT-07-001, Illinois center of transportation. of Materials in Civil Engineering 18(1): 81–92.
Autostrade per l’Italia 2008. Capitolato speciale di Molenaar, A., Hagos, E. & Van de Ven, M. 2010. Effects
appalto, parte seconda, opere civili. of Aging on the Mechanical Characteristics of Bitu-
Bocci, M., Canestrari, F., Grilli, A., Pasquini, E. & Lioi, minous Binders in PAC. Journal of Materials in Civil
D. 2010. Recycling techniques and environmental Engineering 22(8): 779–787.
issues relating to the widening of a high traffic volume Olard, F., Noan, C., Bonneau, D., Dupriet, S. & Alvarez,
Italian motorway. International Journal of Pavement C. 2008. Very high recycling rate (>50%) in hot mix
Research and Technology 3(4): 171–177. and warm mix asphalts for sustainable road construc-
Chen, J.S., Huang, C.C., Chu, P.Y. & Liu, K.Y. 2007. tion. Proceedings of the 4th Eurasphalt and Eurobi-
Engineering characterization of recycled asphalt con- tume Congress, Copenhagen, Denmark.
crete and aged bitumen mixed recycling agent. Journal Petersen, J.C. & Glaser, R. 2011. Asphalt Oxidation
of Materials Science 42: 9867–9876. Mechanisms and the Role of Oxidation Products on
Cominsky, R.J. 1994. The Superpave Mix Design Manual Age Hardening Revisited, Road Materials and Pave-
for New Construction and Overlays. Strategic Highway ment Design 12(4): 795–819.
Research Program, Washington DC, United States. Read, J. & Whiteoak, D. 2003. The Shell Bitumen Hand-
EN 12591. 2009. Specifications for paving grade bitu- book. Fifth edition. Shell UK Oil Products Limited.
mens. European Committee for Standardization, Shen, J., Amirkhanian, S. & Miller, J.A. 2007. Effects of
Brussels, Belgium. rejuvenating agents on superpave mixtures containing
EN 12697-10. 2007. Test methods for hot mix asphalt— reclaimed asphalt pavement. Journal of Materials in
Compactibility. European Committee for Standardi- Civil Engineering 19(5): 376–384.
zation, Brussels, Belgium. Stimilli, A., Ferrotti, G., Conti, C., Tosi, G. & Canestrari,
EN 12697-23. 2006. Test methods for hot mix asphalt— F. 2014. Chemical and rheological analysis of modified
Determination of the indirect tensile strength of bitumens blended with “artificial reclaimed bitumen”.
bituminous specimens. European Committee for Construction and Building Materials 63:1–10.
Standardization, Brussels, Belgium. Zaumanis, M., Mallick, R.B., Poulikakos, L. & Frank, L.
EN 12697-31. 2007. Test methods for hot mix asphalt— 2014. Influence of six rejuvenators on the performance
Specimen preparation by gyratory compactor. Euro- properties of Reclaimed Asphalt Pavement (RAP)
pean Committee for Standardization, Brussels, binder and 100% recycled asphalt mixtures. Construc-
Belgium. tion and Building Materials 71: 538–550.
EN 1426. 2007. Determination of needle penetration.
European Committee for Standardization, Brussels,
Belgium.

273

TIS2017_Book.indb 273 2/8/2017 8:03:49 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A water curtain-based wayside protection system for tunnel fire safety

S. Terribile & G. Malavasi

Department of Civil Building and Environmental Engineering, University of Rome “La Sapienza”, Rome, Italy

P. Firmi & A. Pranno

Rete Ferroviaria Italiana S.p.A., Technical Department, Infrastructure Standard, Rome, Italy

ABSTRACT: Tunnel fire safety plays a key role in the railway operations and many procedural and
technological improvements have been realized in order to prevent and mitigate the fire risk. The fire
prevention and protection procedures can be approached through two main different strategies: the deter-
ministic/prescriptive based approach and the performance-based approach. In the study presented in this
paper, the possibility of application of the water curtain-based wayside fire protection system, as one of
the auxiliary systems for improving and upgrading the fire safety in the railway tunnels for the smoke
compartmentation, has been studied and analysed. A finite element-based dynamic simulation model has
been developed in order to simulate different fire scenarios and to assess the water curtain-based fire pro-
tection system performances in terms of efficiency against the toxic smokes and gases propagation into
the railway tunnel. Some preliminary numerical results have been presented.

1 INTRODUCTION scientific techniques (e.g. engineering, chemistry,

physics, computer science, etc.), social sciences
In the context of the railway tunnel safety, derail- (e.g. sociology, psychology, etc.), medical sciences
ment, collision and fire are considered critical (e.g. biology, physiology, etc.) and economical sci-
scenarios (Kuesel et al. 1996). Indeed, in a closed ences (e.g. statistics, financial management, etc.)
environment (like a railway tunnel), the fire con- (D.M. 09/05/2009). The scientific feature of the
sequences can be more serious than ones that can performance-based approach is also related to the
happen in an open space: this is due to several fact that several Computational Fluid-Dynamics
factors such as the reached high temperature, the (CFD) codes and software tools can be applied
smoke and toxic gases production and propaga- in order to simulate the fire development and the
tion and the possible visibility lack (Malavasi & smoke and hot gases propagation.
Rainoldi, 2013). In the last years, the legislative framework for
The fire prevention and protection procedures the tunnel fire safety has been characterised by a
can be approached through two main different continuous evolution. In the European context, the
strategies: the deterministic/prescriptive based first legislation, recognised as reference guideline,
approach and the performance-based approach. was the UIC Codex 799-9 R “Safety in Railway
The prescriptive-based approach consists of Tunnels (SRT)” (UIC Fiche 799-9, 2002); then, the
normative and rules that impose to respect specific Technical Specification for Interoperability “Safety
obligations to each kind of activities in order to in Railway Tunnels” (TSI-SRT) was published in
be compliant with the fire safety constraints (mini- 2008 (first version) (TSI 07/03/2008) and in 2014
mum safety requirements). (second and current version) (TSI 12/12/2014): the
The performance-based approach consists of TSI-SRT is the cross-cutting mandatory legislation
the application of engineering principles, rules and because it defines minimum requirements for three
expert judgments: in tunnel fire safety, they are strictly related railway subsystems (Infrastructure,
based on the scientific assessment of the combus- Energy, Operations).
tion phenomenon, the fire effects and the human In the Trans-European railway Network (TEN)
behaviour, aiming at the human life, assets and context, the orographic complexity of Italy is the
environment protection, the fire risk and its effects main reason of a very high concentration of tun-
quantitative assessment (Beard & Carvel, 2005). nels (Fig. 1). In Italy, nowadays, the Interministe-
In the last years, many researches and studies rial Decree 28/10/2005 “Sicurezza nelle Gallerie
were conducted on the fire dynamics and people Ferroviarie” is identified as reference national nor-
behaviour evacuation, taking into account several mative: it defines a set of minimum requirements to

275

TIS2017_Book.indb 275 2/8/2017 8:03:51 PM

 reactor exchanging heat, mass and quantity of
motion with the dispersing gas cloud. Fluid cur-
tains are also used to mitigate the consequences of
flammable and toxic substances dispersion in the
environment.
In order to reduce and/or eliminate the released
gaseous and solid substances, the design criteria
must pay attention to three main performance
functionalities:
• Absorption (due to the chemical interactions
among gas particles and liquid droplets)
• Dilution (due to the “air entrainment” effect,
generated by the volumetric flow rate)
• Temperature reduction (due to energy exchange
in terms of heat)
Figure 1. The tunnel distribution in the European rail-
way network. A water curtain spray directed vertically, hori-
zontally or at different angles (e.g. 45°) through a
smoke and hot gases can have a number of mitiga-
be respected through the adoption of prescriptive tion effects:
measures (only for railway tunnels with more than
• Mechanical effects of acting as a barrier to the
1000 m of length); in certain conditions, it suggests
passage of smoke and gases;
a performance-based approach in order to identify
• Mechanical effects of dispersion and dilution by
the tunnel risk level (D.M. 28/10/2005).
air entrainment;
• Mechanical effects of imparting upward quan-
tity of motion to smoke and gases;
2 WATER CURTAIN-BASED PROTECTION
• Thermal effects by cooling in case of smoke and
SYSTEMS
hot or burning gases;
• Physic-chemical effects of absorption of smoke
In the last applications, the water curtain-based
and gases (with or without chemical reaction)
protection systems have been used as one of the
suitable solutions in order to mitigate the smoke From the operative point of view, the water
and hot gases, released from a fire source in closed curtain-based protection system functions are based
spaces. These systems have been recognized as a use- on the fire source identification and isolation through
ful technique to mitigate major industrial hazards a set of water-curtains, which compose a distributed
because it combines attractive features to different system at fixed distance along the railway tunnel or
types of risks (such as pool fire, fuel release, etc.). in the cross-section (Fig. 2). As regards the railway
Nowadays, in the industrial context, three main context, a questionnaire-based survey has shown
typologies of water-based protection systems that in Europe the application of such systems is not
exist:
• Water spray (composed by one or more shaped
sprays, placed side by side)
• Water curtain (composed by a continuous water
wall)
• Water-vapor curtain (composed by one or more
sprays, placed side by side, or by water vapor
screen)
Generally, fixed installations allow rapid
response to an emergency, especially when the
water spray can be activated automatically by gas
sensors; there are also clear advantages in achiev-
ing mitigation without bringing the lives of emer-
gency response crews and site personnel in danger.
The advantage of a mobile monitoring system is
the flexibility of when, where and how to apply.
In case of accidental flammable gas releases, Figure 2. Example of a water curtain-based protection
the spray curtain may be used as a direct-contact system installation inside a railway tunnel.

276

TIS2017_Book.indb 276 2/8/2017 8:03:51 PM

 so diffused; only in Austria, the railway IM (OBB) turbulent flows. A Direct Numerical Simulation
has performed some studies for the possible instal- (DNS) a simulation in computational fluid dynam-
lation of the water curtain-based systems along the ics in which the Navier-Stokes equations are numeri-
passenger railway station platforms. cally solved without any turbulence model. This
Up to now, in Italy, some applications of the means that the whole range of spatial and temporal
water curtain-based protection systems have been scales of the turbulence must be resolved. All the
realized into the underground railway passenger spatial scales of the turbulence must be resolved in
stations in order to create a physical water-based the computational mesh, from the smallest dissipa-
barrier between the fire source and the passenger tive scales (Kolmogorov scales), up to the integral
evacuation path. scale L, associated with the motions containing
These surveys explains better that the application most of the kinetic energy. RANS techniques are
of this kind of systems along the railway tunnels is based on the theory that the turbulent fluid motion
a possible innovative work in order to improve the is described by averaged and time-varying motion;
fire safety levels (UpTun Project, 2008). Unfortu- the parameters have to be averaged by time: this pro-
nately, inside the railway tunnels, the application duces a reduction on the computational time because
of the automatically activated protection systems the average fluid motion scale are bigger than tur-
is forbidden due to technical and procedural limi- bulent fluid motion ones. These techniques requires
tations (due to the presence of electric power sup- the application of other equations (k-ε model, k-ω
ply systems), which have to be switched off during model, etc.) in order to solve the problem.
a fire emergency. For this reason, it is important to
note that specific protection systems (activated by
3.2 Turbulence models
remote control centre) can be installed inside the
railway tunnel in order to generate a physical bar- Turbulent flows of a single Newtonian fluid, even
rier to the smoke and hot gases propagation and to those of quite simple external geometry (such as
facilitate the passenger evacuation. a fully-developed pipe flow) are very complex and
their solution at high Reynolds numbers requires
the use of empirical models to represent the
3 METHODOLOGY unsteady motions. It is self-evident that the addi-
tion of particles to such a flow will result in:
In order to study the fire characteristics, the smoke
• Complex unsteady motions of the particles that
and hot gases development and propagation and
may result in non-uniform spatial distribution
the related water curtain-based protection system
of the particles and, perhaps, particle segrega-
inside a closed environment (like a railway tunnel),
tion. It can also result in particle agglomeration
the risk theory, the thermo-fluid-dynamics equa-
or in particle fission, especially if the particles
tions, the fire combustion concepts and water jet
are bubbles or droplets.
theory have been taken into account.
• Modifications of the turbulence itself caused by
the presence and particles motions. The turbu-
3.1 Equations of motion lence could be damped by the presence of par-
ticles, or it could be enhanced by the wakes and
The physical law of a thermo-fluid dynamic system
other flow disturbances that the motion of the
can be represented by a set of six differential equa-
particles may introduce.
tion in six variables (also known as “Navier-Stokes”
equations): the three velocity components (u, v, w) The Kolmogorov scales are the smallest scales
in x, y, z spatial directions, the temperature (T), used in order to describe a turbulent flow. Basically,
the pressure (p) and the density (ρ). This set is very the Kolmogorov theory introduces the concept that
sensible to the spatial-temporal changes of the ini- the smallest scales have to be universal in order to
tial conditions in terms of solution convergence. describe the turbulence phenomena: in other words,
The Navier-Stokes equations consists of a time- the scales have to be similar for each turbulent flow
dependent continuity equation for conservation of and they are dependent to two parameters (ε is the
mass, three time-dependent conservation of quan- mean rate of dissipation per unit mass of fluid in
tity of motion equations and a time-dependent a time unit and ν is the fluid kinematic viscosity).
conservation of energy equation (Mc Grattan et al. The three main Kolmogorov scales are “Length
2015). In order to solve the Navier-Stokes equa- Scale”, the “Time Scale” and “Velocity Scale”.
tions, several mathematical techniques can be used:
the main three are Large Eddy Simulation (LES),
3.3 Fire and smoke dynamics
the Direct Navier-Stokes (DES) and the Raynolds
Average Navier Stokes (RANS). A Large Eddy Sim- The fire dynamics into a tunnel can be studied and
ulation (LES) is a popular technique for simulating analysed through the following main parameters:

277

TIS2017_Book.indb 277 2/8/2017 8:03:52 PM

 • Heat Release Rate (HRR) During the pre-processing phase, the standard
• Flame Length procedure require the following steps:
• Natural Ventilation • Definition of the problem geometry (physical
• “Backlayering” Effect regular or non-regular bounds);
During the fire spread, it produces energy • Division of the volume (occupied by the fluid)
(in terms of heat) that is time-dependent and it into discrete cells (the mesh): the mesh may be
changes in function of the calorific power and the uniform or non-uniform;
combustion speed. • Definition of the physics (equations of motion,
The fire growth speed depends to the ignition enthalpy, radiation, species conservation, etc.)
process, the flames propagation and the combus- and relative modelling;
tion rate. Each object, subjected to the combustion • Identification of the boundary conditions: this
process, is characterized by growth time and the involves specifying the fluid behaviour and prop-
combustion speed. During the initial phase, a fire erties at the boundaries of the problem.
can be modelled by the “T-squared curve” (Eq. 1): • Definition of the initial conditions (only for tran-
sient problems).
Q = αt2 (1) • After that, the simulation has to be started and
the equations are solved iteratively in a steady-
Where Q = heat release rate [kW]; α = fire growth state or transient way.
coefficient [/]; and t = time [s]. • Finally, a postprocessor has to be used for the anal-
Generally, fire can be classified into four main ysis and visualization of the resulting solution.
classes in function of the growth speed (slow,
medium, fast and ultra-fast). In the fire safety engi- The purpose of this research study was to inves-
neering, the flame length is an important factor to tigate the possibility of installation of a water
be considered in order to better understand the curtain-based protection system in railway tunnels
fire diffusion inside the tunnel. The flame length is and propose a new method to assess their perform-
defined as the distance between the fire centre and ances in terms of efficiency (considered as radia-
the flame impact section (Braubaskas, 1980). tion blockage, smoke propagation obstruction and
The natural ventilation is a very complex param- visibility improvement) by using a CFD simulation
eter to be studied in order to understand the fire model, developed with the free and open-source
behaviour. Inside railway tunnels, the natural ven- FDS and SMOKEVIEW software tools.
tilation is mainly longitudinal and it has influence The FDS model has been applied for two main
on the oxygen flow with consequently fire alimen- objectives:
tation and HRR increase. • To compare the experimental data, collected
During a tunnel fire, another important aspect during the tests, with the numerical results;
to be considered is the “Backlayering” phenom- • To compare the following two main scenarios: the
enum. Even if the longitudinal ventilation has first based on the absence of the water curtains in
been generated, the flames and the combustion order to study the fire growth and smoke and hot
products can move towards opposite direction: gases behaviour inside the railway tunnel; the sec-
this effect happens when the heat release rate and ond based on the activation of the water curtains
the air speed are equal to critical values. during time (distributed in variable number along
the tunnel length) in order to mitigate the effects
of smoke and hot gases propagation and increase
4 COMPUTATION FLUID-DYNAMICS the passenger safety and evacuation facilitation
(CFD) SOFTWARE APPLICATION levels and the rescue teams accessibility.
Computational Fluid Dynamics (CFD) is fluid The main developments steps of the model,
mechanics-related science that uses numerical referring to commands of the FDS code, have been
algorithms and methods to solve and analyse the reported below:
problems that involve fluid flows. Due to the high
• Geometry: “Obstruction”
complexity of the physical and chemical phenom-
• Computational Domain: “Mesh”
ena that characterized the fluids flow (such as inter-
• Boundary Conditions: “Vent” (inlet and outlet)
actions liquids-gases, two-phases flow, turbulence,
• Fire: “HRR-time curve” (t-squared)
etc.), computers have to be used in order to per-
• Water curtain system: “Nozzle flow rate”,
form the calculations required to simulate the inter-
“Droplet Initial Velocity”, “Spray Pattern
action of liquids and gases with surfaces defined
Shape”, “Liquid Water Droplet”
by boundary conditions. In all of these approaches,
the same basic procedure has to be followed to The results of a case study have been reported
design and develop the simulation model. in Section 5.

278

TIS2017_Book.indb 278 2/8/2017 8:03:52 PM

 5 CASE STUDY For the mesh dimension optimization, the
results have been divided in three main sets:
In the study presented in this paper, a finite ele-
• Coarse: suggested cell dimension (δx) is 60.22 cm
ment-based dynamic simulation model of a rail-
and the total number of cells is 81000;
way tunnel has been developed in order to assess
• Moderate: suggested cell dimension (δx) is
the performance of the water curtain-based fire
24.09 cm and the total number of cells is
protection system for the mitigation of smoke and
1244160;
toxic gases propagation.
• Fine: suggested cell dimension (δx) is 15.06 cm
The simulation model of the railway tunnel has
and the total number of cells is 4665600;
been defined in some different steps regarding the
geometry, the fire source, the water curtain-based At this first step of the research study, the cell
fire protection system and the boundary conditions dimension has been selected between the coarse and
inside the tunnel (Table 1). The simulated domain is a moderate classes (δx = 50 cm) in order to obtain the
single-tube, double track railway tunnel: it includes results compliant with the current available com-
the tunnel structural elements, the ballast, the pas- putational resources: this value has been set due to
senger coach (as fire source) and the water-based the fact that the railway tunnels are particular civil
protection system. This domain has specified over- work in which the longitudinal dimension (x) is
all dimensions equal to 200.0 m × 10.5 m × 7.85 m. predominant to the others. The purpose is to study
The mesh of domain consist in 137632 cubic cells the fire and smoke behavior along the entire length
of side equal to 0.5 m. of railway tunnel. A grid sensitivity analysis for
As regards the mesh dimension, the average size the grid resolution verification will be performed.
of the discretization cell (computational grid) is In the future simulations, a multi-mesh approach
related to the characteristic diameter of the fire: should be adopted in order to reduce the cell.
this parameter indicates the goodness of the reso- The protection system (located at 125 m from the
lution grid, according to the following expression fire source) has been composed by four single water
(Mc Grattan et al. 2015): shields: two of them displaced at the lateral sides
(n°1 and n°2), one at the top side of the railway
D* (Q / c T ρ ∞ g )2 /5 (2) tunnel (n°3) and the last one (n°4) along the ballast
in the middle of the railway trackside (Fig. 3).
The fire model has been represented by a single
where D* = characteristic fire diameter [m]; passenger coach with the flames, smoke and gases
Q: total heat release rate [kW]; ρ∞ = air density source located on the roof of the rolling stock. The
[kg/m3]; c∞ = air specific heat [kJ/kg * K]; T∞ = air fire source has been characterised by analytic func-
temperature [K]. tion of Heat Release Rate (HRR) with a maximum
As input, Q = 10000 kW, ρ∞ = 1.205 kg/m3, value equal to 10 MW.
c∞ = 1 kJ/kg * K, T∞ = 293 K, √g = 3.13 m/s2. As The simulations have been done for a time of
output, the fire characteristic diameter (D*) has about 26 minutes.
been calculated equal to 2.41 cm.

Table 1. Railway tunnel simulation model set-up.

Value
Input parameters u.o.m

Tunnel Length 200 m

Tunnel Width 10.5 m
Tunnel Height 7.25 m
Mesh Cell Size (x-direction) 0.5 m
Mesh Cell Size (y-direction) 0.5 m
Mesh Cell Size (z-direction) 0.5 m
Max Heat Release Rate (HRR) 10 MW
Fire Growth Coefficient (α) 0.01
Combustion Reaction Polyurethane Foam
Water Shields Pressure 5 bar
Water Shields Flow Rate 383 l/min, 283 l/min
Liquid Particles Mean Diameter 1 mm Figure 3. Water curtain-based protection system con-
Longitudinal Wind Speed 0.5 m/s figuration: nozzles installation on the railway tunnel
cross-section.

279

TIS2017_Book.indb 279 2/8/2017 8:03:52 PM

 Four sets of monitoring devices (such as ther- • Fractional Effective Dose—FED [/]
mocouple, opacimeter, etc.) has been located at dif- • Volume Fraction [mol/mol]:
ferent distances from the water-curtain protection Carbon Monoxide (CO)
systems (space step equal to 10 m) at two different Carbon Bioxide (CO2)
heights (+1.75 m and +7 m) in order to identify all Oxygen (O2)
the possible critical areas for the human life safety Soot
and sustainability and to monitor the parameters
along the evacuation paths in terms of length and
time. Moreover, two measurement devices has been 5.2 Life safety and evacuation key parameters
located near the water curtain (“sensor x” at −10 m
before and “sensor y” at +10 m beyond) in order to The assessment of the improvement provided by
study the behaviour of smoke and hot gases with the innovative water curtain-based fire protection
the presence of the protection system (Figs. 4–5). system in terms of life safety and evacuation has
been done into the simulation model taking into
account the “Human life safety critical condi-
5.1 Performance key parameters tions”. The human life safety critical conditions
The performance assessment of the innovative are the limit conditions under which a person
water curtain-based fire protection system has can be exposed in case of fire. The design verifi-
been done into the simulation model taking into cation consists to avoid the occurrence, through
account the following parameters: prevention and protection measures against fire,
of conditions more severe than each of the limits
• Temperature [°C] (Fava & al. 2013).
• Visibility [m]
• Heat Flow [kW/m2]
5.3 Preliminary simulation results
The simulation model of the railway tunnel, devel-
oped under the input conditions described above,
returns the following preliminary results. As
regards the “Temperature” parameter (expressed in
°C degrees), in the proximity of the water curtain
protection system at 1.75 m and 7 m, it is possible
to note a reduction due to the cooling effect of the
water particles on the smoke and hot gases gen-
erated by the fire. Both at the railway tunnel ceil-
ing (measurement devices set B) and on the lateral
evacuation sidewalks (measurement devices sets
Figure 4. Measurement devices longitudinal configura- C), the temperature decreases during time beyond
tion: sensors sets installation along the railway tunnel. the water curtain (Figs. 6–7): despite that for the
railway tunnel fire safety there are no commonly
recognised standard thresholds, the temperature
and visibility values could be compared to fixed

Figure 6. Comparison of temperature between the sen-

sor x (−10 m before the water curtain) and sensor y (+10
Figure 5. Measurement devices transversal configura- m beyond the water curtain), located at the ceiling of the
tion: sensors sets installation on the railway tunnel. railway tunnel (height = 7 m—Set B).

280

TIS2017_Book.indb 280 2/8/2017 8:03:53 PM

 Figure 7. Comparison of temperature between the
Figure 8. Comparison of visibility between the sensor
sensor x (−10 m before the water curtain) and sensor y
x (−10 m before the water curtain) and sensor y (+10 m
(+10 m beyond the water curtain), located on the railway
beyond the water curtain), located at the ceiling of the
tunnel lateral sidewalk (height = 1.75 m—Set C).
railway tunnel (height = 7 m—Set B).

thresholds as, for example, in the metro systems

(ISO/TS 13571-2002).
As regards the “Visibility” parameter (expressed
in m), in the proximity of the water curtain protec-
tion system at 1.75 m and 7 m, the first results have
shown that the system doesn’t produce effective
improvements in terms of visibility conditions. It
is important to note that the presence of the water
particles provides a time delay effect on the degra-
dation of visibility conditions, both at the railway
tunnel ceiling (measurement devices set B) and
on the lateral evacuation sidewalks (measurement
devices sets C) (Figs. 8–9). Figure 9. Comparison of visibility between the sensor
The numerical results have shown that the water x (−10 m before the water curtain) and sensor y (+10 m
curtains are capable to block the radiation gener- beyond the water curtain), located on the railway tunnel
ated by the fire and reduce the temperature. lateral sidewalk (height = 1.75 m—Set C).
The use of the Lagrangian approach in the FDS
model can give the following advantages: the chemical substances concentration, further
simulations will be done in order to assess the effi-
• The Lagrangian model is slightly faster to com-
ciency of the system, but the expected results are
pute than the Eulerian one.
not promising due to the same drag effects.
• The Lagrangian model enables to calculate the
initial velocity of the droplets.
• The specification of the initial trajectories of the
6 CONCLUSIONS AND FUTURE WORK
droplets is considerably simpler than with the
Eulerian approach.
In the research study proposed in this paper, an
• With the Lagrangian approach, an accurate rep-
innovative finite element-based dynamic simula-
resentation of the droplet size distribution can
tion model has been developed in order to assess
be modeled.
a water curtain-based fire protection system per-
A limitation to be highlighted (in the modeling formances in terms of efficiency against the heat
of the water droplets by using the Lagrangian and combustion products propagation into the
approach) is that the FDS code enables the user railway tunnel. This kind of systems have to be
to apply a two-phase model in which the smoke designed in order to perform some key functions:
and hot gases are studied as a continuous pattern
• Dilution (due to the “air entrainment” effect,
characterized by thermal energy, mass and velocity
generated by the volumetric flow rate)
and, on the contrary, the discharging curtain not
• Temperature reduction (due to energy exchange
as a continuous water layer but a transient discrete
in terms of heat).
pattern with air space inside the curtain; so, each
water curtain is not capable to use for the blockage An interesting property is that the FDS simula-
of smoke spreading, as the smoke particle would tion model can be applied in order to optimize the
transmit through the transient air space. As regards water curtain system configuration along the railway

281

TIS2017_Book.indb 281 2/8/2017 8:03:54 PM

 tunnel cross-section. The model can assess the sys- D.M. 28/10/2005. Sicurezza nelle Gallerie Ferroviarie.
tem performances by varying the nozzles number Decreto Ministeriale Italiano.
per each water curtain (n°4, 6, 8), the 3D position of Deliverable n°251. September 2008. Engineering Guid-
each single nozzle and the type/commercial model ance for Water Based Fire Fighting Systems for the
Protection of Tunnels and Sub Surface Facilities.
(in terms of orifice diameter). Some possible tests UPTUN Project. WP 2 Fire development and mitiga-
may be performed taking into account the variation tion measures.
of water flow [l/min] and pressure [bar]. Fava, P. & al. A.A. 2012–2013. Simulazione di un Incen-
As future works, a detailed research study will dio in Galleria tramite il Software FDS. Tesi di Laurea
be done in order to assess the water curtain-based Magistrale. Università degli Studi di Padova.
protection system performance in terms of com- ISO/TS 13571:2002. Life-Threatening Components of
bustion products and FED concentration in order Fire-Guidelines for the Estimation of Time Available
to guarantee the life safe conditions for passenger for Escape Using Fire Data. ISO Guidelines.
evacuation. Moreover, the calibration of simulation Kuesel, T.R. et al. 1996. Tunnel Engineering Handbook.
Malavasi, G. & Rainoldi, G. 2013. Il Rischio Incendio
model will be performed taking into account any nelle Gallerie Ferroviarie. L’elaborazione di Piani di
collected experimental data. Then, a model sensi- Emergenza Esterna. IV Convegno Nazionale-Sicurezza
tivity analysis will be done in relation to the assess- ed Esercizio Ferroviario. Giugno 2015. Roma, Italy.
ment of the water curtain-based protection system McGrattan, K. et al. November 2015. Fire Dynam-
performance characteristics, the comparison to life ics Simulator Technical Reference Guide-Volume1:
safe conditions thresholds (to be identified and Mathematical Model. National Institute of Standards
quantitatively defined) and the effects of variation and Technologies (NIST) Special Publication 1018-1-
of the railway tunnel geometrical features. As last 6th Edition.
future development, a simple evacuation model McGrattan, K. et al. November 2015. Fire Dynamics
Simulator-User’s Guide. National Institute of Stand-
will be integrated in the core one in order to study ards and Technologies (NIST) Special Publication
more in detail the smoke and hot gases propaga- 1019-6th Edition.
tion in relation to the people evacuation. TSI 07/03/2008. Technical Specification of Interoperabil-
ity relating to Safety in Railway Tunnels in the trans-
European Conventional and High-speed Rail System.
REFERENCES Technical Specification for Interoperability.
TSI 12/12/2014. Technical Specification for Interoper-
Babrauskas, V. 1980. Flame Lengths under Ceiling, Fire ability relating to “Safety in Railway Tunnels” of the
and Materials, 4, 3,119–126. Rail System of the European Union. Technical Speci-
Beard, A. & Carvel, R. 2005. The Handbook of Tunnel fication for Interoperability;
Fire Safety. UIC Fiche 799-9. September 2002. Safety in Railway
D.M. 09/05/2009. Direttive per l’attuazione dell’approccio Tunnels. UIC Codex.
ingegneristico alla sicurezza antincendio. Decreto
Ministeriale Italiano.

282

TIS2017_Book.indb 282 2/8/2017 8:03:55 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

3D control of obstacles in airport location studies

D. Gavran, S. Fric, V. Ilić, F. Trpčevski & S. Vranjevac

Department for Roads, Railroads and Airports, Faculty of Civil Engineering, University of Belgrade, Serbia

ABSTRACT: While looking for a potential site of a new airport, meteorological and environmental
analyses, as well as navigational analyses, are all of the ultimate importance. Potential airport location
failing to pass any of these three checks is considered absolutely inappropriate for further airport
development. Especially in mountainous regions, navigational analyses are concentrated on the control of
obstacles which is synonym for the analyses of Obstacle Limitation Surfaces. Obstacle Limitation Surfaces
are imaginary surfaces defining the volume of airspace in the vicinity of the airport that should ideally
be kept free from obstacles so as to provide for safe aircraft operations either during an entirely visual
approach or during the visual segment of an instrument approach. Also, establishment of such a protected
volume of airspace prevents the uncontrolled growth of manmade obstacles in the vicinity of the airport.
The paper presents contemporary 3D techniques of orienting the runway so as to minimize the extent of
protrusions through the Obstacle Limitation Surfaces. These techniques are based on the triangulated 3D
model of terrain surface and on the models of Obstacle Limitation Surfaces. The techniques in concern
are demonstrated on an airport located in the mountainous Balkan region.

1 INTRODUCTION of protrusions through the Obstacle Limitation

Surfaces. The techniques in concern are based on
Most of the airport projects are concentrated on the triangulated 3D model of terrain surface and
expansions and reconstructions of the existing facil- on the models of Obstacle Limitation Surfaces.
ities. Even when working on airport master plans, Officially, primary products of these analyses are
these master planes are usually confined to existing The Aerodrome Obstruction Chart—Type A and
airport locations. But, when working on a master The Aerodrome Obstruction Chart—Type B.
plan of a new airport, the whole planning proc- In addition, the deployment of triangulated 3D
ess starts with the, so called, location study (ICAO surfaces introduces isopachyte plan, which further
1987, FAA 2007, Horonjeff et al. 2010). Practi- improves the control of obstacles.
cally, meteorological and environmental analyses, as These techniques are demonstrated on a Trebinje
well as navigational analyses, are carried out for all airport located in the mountainous Balkan region
promising airport locations near the city in concern. (Fig. 1). Trebinje is small city in southern Herce-
Potential airport location failing to pass any of these govina with population of 25,000, the number
three checks is considered absolutely inappropriate expands to 70,000 in summer months. It is merely
for further airport development. Especially in moun- 25 km inland from the Croatian historical coastal
tainous regions, navigational analyses are concen-
trated on the control of obstacles which is synonym
for the analyses of Obstacle Limitation Surfaces.
Obstacle Limitation Surfaces are imaginary sur-
faces defining the volume of airspace in the vicinity
of the airport that should ideally be kept free from
obstacles so as to provide for safe aircraft operations
either during an entirely visual approach or dur-
ing the visual segment of an instrument approach.
Also, establishment of such a protected volume of
airspace prevents the uncontrolled growth of man-
made obstacles in the vicinity of the airport.
The paper presents contemporary 3D techniques
of orienting the runway so as to minimize the extent Figure 1. Future Trebinje international airport.

283

TIS2017_Book.indb 283 2/8/2017 8:03:57 PM

 city of Dubrovnik. The vicinity of Dubrovnik cre-
ates an opportunity for the new Trebinje airport
(Nikolić et al. 2009) to compete for passengers with
the existing Dubrovnik airport.
Despite the fact that navigational, climatologic
and environmental analyses together are crucial in
the search for an optimal airport location, it imme-
diately became apparent that the proper setting of
the approach procedures in relation to the existing
ground features would be decisive. The vicinity of
existing airports in Mostar, Podgorica, Tivat and
Dubrovnik, as well as the recently established bor-
ders between Croatia, Montenegro and Bosnia and
Hercegovina, imposed further limitations.
Figure 3. Grid terrain model.

2 MODELLING TERRAIN SURFACES

cell is a part of a curved (twisted) surface. Since
As a rule, planning or design process starts from the cutting of the longitudinal profiles and cross
the digital terrain model. The most widely adopted sections, volume calculations and other geometri-
terrain model for civil engineering purposes is cal analyses are much easier to execute (or to pro-
TIN—Triangulated Irregular Network model gram) on simple triangular facets, even when the
(Fig. 2). By definition, TIN model connects terrain grid terrain model is at the disposal, at the start of
points by using non-overlapping triangles tending the design process it is “exploded” into triangles.
to be as much equiangular in plan projection as In fact, each “twisted” grid cell (defined with four
possible (Green & Sibson 1978, Gavran 1996). points) could be easily exploded into two triangles
By simple editing (switching triangles’ edges) it (each one defined with three points).
is possible to incorporate any kind of manmade or
natural feature (ridge, escarpment, pavement edge)
into the TIN, making the model identical to the 3 MODELLING OBSTACLE LIMITATION
natural surfaces. For rough examination of large SURFACES
areas of the terrain, the grid model could be quite
appropriate (Fig. 3) (Gavran 1996, Petrie & Ken- Having the digital terrain model completed the
nie 1987). control of obstacles moves to the modelling of
The generation of a grid model is much easier to obstacle limitation surfaces. The shapes and sizes
program than that of a TIN model. But, for subse- of obstacle limitation surfaces are published by
quent geometrical analyses, the TIN model is much ICAO—International Civil Aviation Organization
simpler to work with. In fact, each triangular facet manuals (ICAO 2004). In general, these are imagi-
is a part of a simple plane (as the three triangle’s nary surfaces constructed around a particular
vertices define the perfect plane), while the grid runway. These are approach and take-off surfaces
(extending up to 15 km in front of each runway’s
threshold), the inner horizontal surface (circular
surface with the radius of 4 km, 45 m above the
lower threshold), the conical surface (climbing at
the grade of 20% around the perimeter of an inner
horizontal surface, and having the width of 2 km)
and the transitional surface climbing from the run-
way strip up to the inner horizontal surface at the
rate of 1:7 (7:1 in American format, or cca 14%).
The entire set of surfaces is moved and rotated
(together with the runway) in order to minimize
terrain protrusions. Obstacle limitation surfaces
are also checked against the natural (trees) and
manmade (buildings, towers, power lines) features.
To proceed with the obstacle control, a triangu-
lated model of obstacle limitation surfaces should
be constructed (Fig. 4). In essence, standard sur-
Figure 2. TIN terrain model. faces (with the straight approaches) could be easily

284

TIS2017_Book.indb 284 2/8/2017 8:03:58 PM

 Figure 5. Aerodrome obstruction chart—Type B.
Figure 4. Obstacle limitation surfaces’ model.

modeled by using general purpose CAD systems, each promising position of the runway, the hidden
while curved approach or take-off paths could be line removal should be called in plan projection,
modeled by using software solutions intended for thus approximately indicating areas where obsta-
road modeling. In fact, it is quite simple to pro- cle limitation surfaces sink beneath the terrain
gram the construction of a triangulated surface surface.
following any reasonable curved centerline in plan To sharply delineate the terrain penetration
or profile projection (Gavran 1996). line, it is necessary to deploy specific (and rather
sophisticated) tool. This is the tool dealing with the
penetrating triangles: the terrain triangles and the
4 3D CONTROL OF OBSTACLES triangles forming the model of the obstacle limita-
tion surfaces.
3D control of obstacles is performed upon the In this particular case, our design team deployed
triangulated terrain model, on one side, and the the tool for decomposing penetrating triangles into
model of the obstacle limitation surfaces, on the the subtriangles that do not intersect any more,
other. but touch each other along the lines of intersec-
tion. The software had been used for years while
modeling intersecting cut and fill slopes (Fig. 6)
4.1 Producing standard graphical documents
(Gavran 1996, 2012).
The graphical documents representing the relation The software is supposed to work on triangu-
between the obstacles and the obstacle limitation lated cut/fill slope models. After decomposing the
surfaces are The Aerodrome Obstruction Chart— fill slopes’ triangles, the lower subtriangles (below
Type A and The Aerodrome Obstruction Chart— the intersection lines) are to be removed, while
Type B. Aerodrome Obstruction Chart—Type B is modeling cut slopes, the upper triangles are the
more illustrative (Fig. 5) (ICAO 1983, 1989). This surplus triangles.
is the map representing obstacle limitation surfaces The algorithm that handles intersection of mul-
in plan projection, as well as all natural and man- tiple triangles is a sophisticated one, because all of
made obstacles in the area. Apart from being a cru- the newly created subtriangles deriving from one
cial element of the airport location study, the Type “explosion” (between two particular triangles) and
B map accompanies the flight crew on the route touching each other perfectly, must be checked
to a particular airport. The map informs the crew again for potential “explosions” with the rest of
on the most prominent obstacles surrounding an the starting triangles. To speed up the process, fam-
airport. Based on these obstacles the crew decides ily relations are introduced between the triangles.
upon the procedures (turns) to be performed in the The pretriangles are the triangles belonging to
case of the abandoned approach etc. the starting set of triangles, while the subtriangles
One of the most important features of The Aero- created in the explosion of one particular triangle
drome Obstruction Chart—Type B are the thick are brothers (or sisters). Besides the brothers and
blue lines indicating terrain penetration through the sisters, each subtriangle has its mother and
the obstacle limitation surfaces. While positioning father: the triangle of origination and the triangle
the runway centerline, the model of these surfaces in relation to which the originating triangle was
is moved and rotated along with the runway. For exploded. As the algorithm starts to dissipate the

285

TIS2017_Book.indb 285 2/8/2017 8:03:59 PM

 Figure 6. Modelling intersecting cut and fill slopes.

triangles, the number of candidates for the “explo- of the pretriangles) is only a fraction of its mother.
sion” grows rapidly. Keeping track of family rela- Consequently, the counter for potential fathers
tions, unnecessary (impossible) “explosions” are (no1) starts from 0, if the candidate mother is one
skipped, making the software run faster. But, to of the pretriangles, or from her father’s index No
cut the long story short, to delineate the intersec- increased by 1, if the mother results from a pre-
tion between the terrain triangles and the trian- vious explosion. The program then cycles trough
gulated model of the obstacle limitation surfaces, list listri, searching for a potential father. The
only a small fraction of this algorithm should be fathers’ cycle goes to the end of that portion of
deployed. Only the intersection lines between the the list containing pretriangles only (no1 < = len),
pretriangles are to be generated and no pretriangles as pretriangles only could act as potential fathers.
decomposition is needed (left process on Fig. 6). Taking newly created triangles (resulting from the
The simplified algorithm of resolving triangles’ explosions) for candidate fathers has no sense—a
explosion is given in Figure 7. General idea of how triangle not colliding with a particular pretriangle
explosion goes is given in the upper left corner, could not collide with its children either.
while the main body of the program starts just Unlike father, potential mother could be any of
below. The list of triangles to be exploded (listri) the pretriangles, or newly created triangles. Spatial
contains particular triangles, each one defined by triangle could collide with two or more triangles.
its vertices (t1,t2,t3), name (en, which is important In this case, particular triangle must be exploded
only within exact programming environment), in the relation to the first one (father). Its children
color (col), layer (lay) and father’s number within (resulting from the explosion) will be touching the
the list (No, pretriangles have no father). The father’s plane by their edges, but would still collide
length of the starting set of triangles is len + 1 (as with some other triangles. Therefore, these new tri-
the list of the triangles is zero-based, the index of angles must be given a chance to act as mothers
the first pretriangle is 0, while the index of the last and collide with the second (or even third) trian-
one is len). As the program keeps running, the list gle. When a potential mother finds an appropriate
listri grows in size, while new triangles are added father, it explodes, variable codx takes value nil and
to the list. program takes the next triangle from list listri as a
Program takes triangles one by one and tries potential candidate mother.
to explode them. Taken triangle, as candidate Resolving a potential collision between a poten-
for explosion, is potential mother triangle and is tial mother (black triangle) and a potential father
drawn in black. At the start, variable codx takes (white one) starts in subroutine incalc. Actually,
“T” value. When (and if) the mother explodes, it incalc subroutine determines in how many points
changes to nil value. father’s edges penetrate mother’s plane. Mother
The question is: in collision with which triangle triangle resides in plane P, while father triangle
the mother could explode? If the mother triangle is resides in plane G. If these two planes are parallel,
one of the pretriangles (if its index no is less than or coincide, there is no collision. Also, if some of
or equal len), it could be any of the pretriangles. the mother’s edges reside within a father’s plane,
But, if the candidate mother is one of the newly there is no collision either (the potential mother
created triangles (resulting from previous explo- touches the potential father by its edge). In these
sions, no > len), then it could be exploded only cases, the program leaves subroutine incalc. Oth-
when colliding with the pretriangles positioned in erwise, the program proceeds with determin-
the list listri after her father. If the mother of the ing penetration points of father’s edges through
potential mother have not found a potential father mother’s plane. These points are stored in the list
before finding the actual one (father), neither the named intpnt. The rule is: no point could appear
potential mother could collide with the triangle in the list twice. If one of the father’s vertices
preceding the actual father—the potential mother resides exactly within a mother’s plane, it is logical
(if resulting from the explosion and not being one that this vertex would be added to the list intpnt

286

TIS2017_Book.indb 286 2/8/2017 8:04:00 PM

 Figure 7. Simplified algorithm of resolving triangles’ explosion.

twice—once for each of the two edges meeting at ces resides exactly within the mother’s plane, while
that vertex. But, this is prohibited and the maxi- the remaining two vertices are located on the same
mum length of intpnt list is 2. If, after analyzing all side of the mother’s plane. In that case, there is no
of the three father’s edges, the length of list intpnt collision. Also, there is no collision if the length of
stays at 1, it means that one of the father’s verti- intpnt list is 0.

287

TIS2017_Book.indb 287 2/8/2017 8:04:00 PM

 The collision between the potential mother and file resides in the upper part of the drawing. The
the potential father occurs (mothers’ explosion profile spans the length of the approach (or take—
takes place) only if the length of intpnt list equals off) path. All the obstacles marked in the plan pro-
2. In that case, the program proceeds to subroutine jection are placed at their distinctive elevations in
newtry which executes the explosion. The subrou- the profile. Runway and approach surface profiles
tine first determines how many of the father’s pene- are also superimposed.
tration points (points from intpnt list) reside within The software for marking the obstacles and the
the mother triangle (the number of these points correlation of obstacles in plan and profile projec-
is cou). If no point from intpnt list resides within tions is a rudimentary one (Gavran 1996, 2012). But
mother’s triangle, but any of the mother’s edges the most interesting part of the obstacle profile is
penetrates a potential father, then the collision the terrain itself. The terrain profile is not a simple
exists and the explosion is executed by subroutine longitudinal profile cut along the extended runway
(int0). If no point from intpnt list resides within centerline and following the approach path, nor
mother’s triangle, but no mother’s edge penetrates a kind of combination of the profiles generated
a potential father, then the collision does not exist along the diverging edges of the approach path.
and flow leaves subroutine newtri. If only one of The terrain profile is supposed to be a kind of a
the points from intpnt list resides within mother’s shadow profile (ICAO 1989). At each incremental
triangle, the explosion is executed according to step along the approach path, the maximum ter-
the scheme from int1 subroutine. For two points rain elevation is taken from the terrain cross sec-
residing within a mother’s triangle, subroutine int2 tion, providing the cross section spans the exact
is executed. Each one of the three subroutines exe- width of the approach (or take–off) path at this
cuting an explosion (int0, int1 and int2) explodes particular location. The profile outlines the exact
the mother’s triangle into the triangles that touch terrain shadow for the observer standing aside the
the father’s triangle either by their edges, or by approach path (Fig. 9).
their vertices. Graphically, the mother entity (tri- To produce such a profile, we turned again to
angle) is erased, but it remains within listri list. If the existing software tool intended primarily for
the mother’s triangle belongs to the starting set road design. This is the tool transferring 3D points
of pretriangles, it could even serve as a potential labeled with station/elevation attributes into the
father for some other triangles. Each newly created longitudinal profile.
triangle (emanating from the explosion) is drawn When setting the vertical alignment of the street,
by using mothers color and layer and is added to it is highly recommended to observe some specific
the end of the list listri. In fact, the entire group of points scattered in the vicinity of the centerline,
newly created triangles (ranging in size from 3 to 5) such as entrances to nearby houses, shop windows
is added to the end of listri. Within the list, newly etc. These points are marked with station / eleva-
created triangle is represented by its vertices, by its tion attributes and transferred from 3D space into
color, layer and father’s index (no1). the longitudinal profile.
The Aerodrome Obstruction Chart—Type A For the creation of a shadow terrain profile along
is a combination of plan and profile projection the approach path, only a small automation is added
(Fig. 8). In the lower part of the document there to the existing tool. Points are now automatically
is a relatively narrow plan depicting the approach attached to every vertex of each terrain triangle
surface, with all the obstacles marked with the enclosed within the approach path and then labeled
symbols proposed by ICAO. The longitudinal pro- with station / offset (and elevation) assemblies in
relation to the centerline of the approach path. The
“cloud” of points generated in this manner is then
transferred into the longitudinal profile generated

Figure 8. Aerodrome obstruction chart—Type A. Figure 9. Shadow terrain profile.

288

TIS2017_Book.indb 288 2/8/2017 8:04:01 PM

 along the centerline of the approach path. When For construction purposes, 1.00 m isopachytes
taken from the TIN model produced by explod- are suitable for ground remodeling projects, while
ing the grid model into triangles, this cloud nicely the interval of 1.00 cm is suitable for road resurfac-
reflects the terrain morphology (Fig. 10). Finally, ing projects. On road resurfacing projects isopach-
the outline of the shadow profile is redrawn manu- ytes may be used to represent the variable thickness
ally, through the highest points within the profile. of the leveling course (the course laid after the
scraping of the existing pavement and beneath the
newly applied wearing course).
4.2 Producing non-standard graphical documents
In the case of Trebinje airport, apart from the ter-
Besides the plan projection presented on The Aero- rain penetration line through the obstacle limitation
drome Obstruction Chart—Type B, some cross sec- surfaces, the idea was to somehow depict the sheer
tions (perpendicular to the runway centerline) are extent of this penetration. Therefore, the TIN model
always helpful. These cross sections usually contain representing the “thickness” of the penetration was
terrain and the obstacle limitation surfaces. But, created and contours were generated from such a
in the case of Trebinje airport we came to a con- model. By definition, these contours were isopach-
clusion that isopachytes’ projection would give a ytes. Bearing in mind the area to be covered, the scale
much clearer picture than any set of cross sections. of Type B map and the sole purpose of these isopach-
Till then, we had been using isopachytes only on ytes, the interval of 50 m was adopted (Fig. 12).
resufac-ing and ground remodeling projects. Iso- As the isopachytes resembled the general morphol-
pachytes are the contour lines delineating equal dif- ogy of the terrain, the picture of the terrain penetra-
ferences in elevation between the two triangulated tion extent became quite clear and our design team
surfaces (the proposed and the existing surface). was very satisfied with this graphical document.
At the location of each node (from both triangu- In essence, whenever one of the two surfaces to be
lated surfaces) the vertical difference between the compared is rather flat, then the general flow of the
two surfaces is determined and the new point, hav- isopachytes resembles the contours of the opposing
ing the elevation equal to that difference, is set at surface. In this particular case, obstacle limitation
this position. The TIN model generated from thus surfaces played the role of a “flat” surface, while iso-
positioned points represents the thickness between pachytes “imitated” contours of the terrain.
the two surfaces. On grading projects, the model Unlike Trebinje airport project, in some cases we
is negative in cut areas and positive in areas to be were not thoroughly satisfied with the isopachytes’
filled. Contours generated from such a TIN model application. When working on a dredging plan for
are isopachytes (Fig. 11) (Gavran 2012).

Figure 10. Transferring cloud of points into the profile. Figure 12. Isopachytes’ representing terrain protrusions.

Figure 11. Isopachytes’ application on grading projects.

289

TIS2017_Book.indb 289 2/8/2017 8:04:01 PM

 Kuwait harbor (Lukić et al. 2003), the isopachytes itation surfaces completed, it is exceptionally easy
generated between the existing and the proposed to generate contours from such a model (Fig. 14).
bottom of the harbor were hard to follow even
for the eye of a professional (Fig. 13). This hap-
pens whenever the vertical differences between the 5 CONCLUSION
two surfaces are relatively small and when the sur-
faces frequently change sides in the vertical sense The paper presents contemporary 3D techniques
(between cut and fill). of obstacle control, while working on airport loca-
Contours generated from the obstacle limitation tion studies. In general, 3D control of obstacles is
surfaces are always welcome. In plan projection they based on comparison between the two triangulated
give a general three-dimensional picture of the entire models: the surface terrain model and the model of
assembly of obstacle limitation surfaces. In munici- obstacle limitation surfaces.
pality plans they impose vertical limits on the struc- Primary products of these 3D analyses are The
tures planned in the area surrounding the airport. Aerodrome Obstruction Chart—Type A and The
With the triangulated 3D model of the obstacle lim- Aerodrome Obstruction Chart—Type B, as required
by ICAO. Generation of these graphical documents
is supported by software tools primarily developed
for terrain remodeling, grading and road design.
In addition, the deployment of triangulated 3D
surfaces introduces new documents, such as iso-
pachyte plan, which is not obligatory, but further
refines the control of obstacles.

REFERENCES

FAA, 2007. Airport Master Plans. Washington D.C.:

U.S. Department of Transportation, Federal Aviation
Administration, USA.
Gavran, D. 1996. Razvoj metodologije i tehnoloških
postupaka za prostorno projektovanje aerodroma (in
Serbian): doctoral thesis. Belgrade: Faculty of Civil
Engineering University of Belgrade. 252 p.
Gavran, D. 2012. GCM++ (Gavran - Civil Modeller).
User Guide. Belgrade. Autodesk. 86 p.
Green, P.J. & Sibson, R. 1978. Computing Dirichlet Tes-
sellations in the Plane. The Computer Journal 21(2).
Horonjeff, R., McKelvey, F.X., Sproule W.J., Young S.B.
2010. Planning and Design of Airports. Fifth Edition.
New York: McGraw Hill.
ICAO, 1983. Airport Services Manual—Part 6—Control
Figure 13. Kuwait harbor dredging plan. of Obstacles. Montreal: International Civil Aviation
Organization.
ICAO, 1987. Airport Planning Manual—Part 1. Montreal:
International Civil Aviation Organization.
ICAO, 1989. Aeronautical Charts—Annex 4 to the
Convention on International Civil Aviation. Montreal:
International Civil Aviation Organization.
ICAO, 2004. Annex 14 to the Convention on International
Civil Aviation. Montreal: International Civil Aviation
Organization.
Lukić, M. et al. 2003. Kuwait Port Dredging Plan.
Belgrade/Kuwait: Al - Tawbad.
Nikolić, D. et al. 2009. Ispitivanje lokacijskih uslova,
potvrda lokacije i dopuna Master plana aerodroma
Trebinje (in Serbian). Belgrade: Faculty of Civil
Engineering University in Belgrade and Public
Enterprise ‘’Airport Trebinje’’.
Petrie, G. & Kennie, J.M. 1987. Terrain Modelling in
Figure 14. Generating contours from obstacle limita- Surveying and Civil Engineering. Computer Aided
tion surfaces. Design 19(4).

290

TIS2017_Book.indb 290 2/8/2017 8:04:02 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Environmental and economic assessment of pavement construction and

management practices for enhancing pavement sustainability

J. Santos & V. Cerezo

IFSTTAR, AME-EASE, Route de Bouaye, Bouguenais, France

G. Flintsch
The Charles Via, Jr. Department of Civil and Environmental Engineering, Center for Sustainable
Transportation Infrastructure, Virginia Tech Transportation Institute, Virginia Polytechnic
Institute and State University, Virginia, USA

A. Ferreira
Road Pavements Laboratory, Research Center for Territory, Transports and Environment,
Department of Civil Engineering, University of Coimbra, Coimbra, Portugal

ABSTRACT: Stakeholders in the pavement sector have been seeking new engineering solutions to move
towards more sustainable pavement management practices. The general approaches for improving pave-
ment sustainability include, among others, reducing virgin binder and virgin aggregate content in hot
mix and warm mix asphalt, reducing energy consumed and emissions generated in the mixtures’ produc-
tion, applying in-place recycling techniques, and implementing preventive treatments. In this study, a
comprehensive and integrated pavement Life Cycle Costing-Life Cycle Assessment (LCC-LCA) model
was developed to investigate, from a full life cycle perspective, the extent to which several pavement engi-
neering solutions (hot in-plant recycling mixtures, warm mix asphalt, cold central plant recycling and
preventive treatments) are most efficient at improving the environmental and economic aspects of pave-
ment infrastructure sustainability, when applied either separately or in combination, in the construction
and management of a road pavement section located in Virginia, US. Furthermore, in order to deter-
mine the preference order of alternative scenarios, a multi-criteria decision analysis method was applied.
The results showed that the implementation of a recycling-based maintenance and rehabilitation strategy
where the asphalt mixtures are of type hot mix asphalt containing 30% RAP best suits the varied and
conflicting interests of stakeholders. This outcome was found to be robust even when different design and
performance mixtures and treatment type scenarios were considered.

1 INTRODUCTION mental regulations, have strengthened the com-

mitment of DOTs to deliver infrastructures in
With the recent launch of the Build America a more environmentally friendly way, while also
Investment Initiative (White House 2014)—a US using funds in the most economically responsi-
government-wide initiative designed to address the ble way possible. These factors have motivated
country’s pressing infrastructure investment needs DOTs, and the pavement community in general, to
and promote economic growth—many Depart- investigate sustainable engineering solutions that
ments of Transportation (DOTs) are likely to reduce both environmental impacts and the costs
renew their efforts both in the construction of new of road pavement construction and maintenance.
highway infrastructures and in the maintenance of Some examples of solutions with the potential to
those already built. improve pavement sustainability include (but are
The activities central to the construction, opera- not limited to): (1) asphalt mixes requiring lower
tion and maintenance of highway infrastructures manufacturing temperatures, such as Warm Mix
are notorious for the large quantities of natural Asphalt (WMA), half-warm mix asphalt and cold
materials and energy they consume, as well as asphalt mix technologies, (2) in-place pavement
for the considerable environmental impacts they recycling, (3) pavement preservation strategies and
generate. In addition, the environmental impacts preventive treatments, (4) long-lasting pavements,
of these activities, along with stringent environ- (5) Reclaimed Asphalt Pavement (RAP) materials,

291

TIS2017_Book.indb 291 2/8/2017 8:04:05 PM

 (6) reclaimed asphalt shingles materials, (7) indus- of how materials considerations, treatment typol-
trial wastes and byproducts, biobinders, etc. ogy, design, construction, and application tim-
Several studies have, to some extent, corrobo- ing can enhance pavement sustainability, and the
rated the environmental benefits with which the related tradeoffs in these stakeholders’ various
aforementioned solutions are a-priori associated. requirements.
However, many of these studies have applied meth- For this purpose, a comprehensive and inte-
odologies that disregard the environmental burdens grated pavement LCC-LCA model, encompass-
of some processes and pavement life cycle phases. ing six pavement life cycle phases into the system
In addition, as the primary goal of transportation boundaries, has been developed. They are as fol-
agencies is to provide maximum pavement per- lows: (1) materials extraction and production,
formance within budgetary constraints, an envi- (2) construction and M&R, (3) transportation
ronmentally advantageous solution might not be of materials, (4) work-zone traffic management,
preferable to a technically equivalent solution if (5) usage, and (6) End-of-Life (EOL).
it is not economically competitive. Furthermore, Finally, to account for the often conflicting
there are still some questions about (1) the cost of interests of the multiple pavement management
such solutions throughout their life cycle, (2) which stakeholders, the pavement construction and main-
factors are the key drivers of their economic per- tenance scenarios considered in this paper were
formance, and (3) which stakeholders benefit most further analyzed using a Multi-Criteria Decision
from the application of those solutions. Making (MCDM) method.
Addressing this multifaceted issue and pro-
viding answers to the aforementioned questions
requires multidimensional life-cycle modelling 3 METHODOLOGY
approaches, such as Life-Cycle Assessment (LCA)
and Life Cycle Costing (LCC). These approaches 3.1 Principles of the integrated pavement life
enable long-term economic and environmental fac- cycle costing and life cycle assessment model
tors to be included in the decision-making process
by providing a comprehensive and cumulative view The research work presented in this paper builds on
of both the environmental and economic dimen- the process-based LCA (P-LCA) and LCC models
sions of a given technical solution. However, it is introduced by Santos et al. (2015a,c) and Santos
important to underline that life-cycle modelling et al. (2015b), respectively, to develop a comprehen-
approaches alone will not necessarily determine sive and integrated pavement LCC-LCA model.
which solution is most suitable for a given purpose. The proposed pavement LCC-LCA model follows
Rather, the information that they provide should a cradle-to-grave approach and relies on a hybrid
be used as one component of a more comprehen- inventory approach that allows the sub-models
sive decision making process, which, among other to connect with one another by data flows. Spe-
aspects, will allow the assessment of tradeoffs cifically, it connects the monetary flows associated
among the interests of the multiple stakeholders. with exchanges of the pavement life cycle system
that are directly covered by the LCC model but for
which specific process data are either completely
2 OBJECTIVES or partially unavailable (e.g., interest on loan, taxes
and insurance of construction equipment, etc.).
The main objectives of this paper are as follows: In other cases happen that data are available, but
(1) to investigate, from a life cycle perspective, the their collection and subsequent analysis are highly
extent to which several pavement engineering solu- demanding, either time- or resource-wise and were
tions (i.e., hot in-plant recycling mixtures, Sasobit® therefore disregarded in the previous P-LCA mod-
WMA, cold central plant recycling [CCPR], and els (e.g., construction equipment manufacturing
preventive treatments) best improve the environ- and maintenance, on- and off-road vehicles tires
mental and economic dimensions of pavement manufacturing, lubricant oil production, etc.)
infrastructure sustainability, when applied either (Santos et al. 2015a,c). These data are combined
separately or in combination, in the construction with an input-output methodology for deriving
and management of a road pavement structure, (2) the underpinning environmental burdens. Thus, by
to raise awareness of the importance of consider- interactively integrating the strengths of Process-
ing materials and processes in extending the system based Life Cycle Inventory (P-LCI) and Input-Out
boundaries of environmental and economic life (I-O) LCI, the resources which are readily avail-
cycle assessments, which may eventually reverse able can be used in a more efficient, consistent and
the sustainability of a solution and (3) provide rational way and with less effort, helping to reduce
designers, contractors, local and state agencies, the “cutoff ” errors and improving the consistency
and road users with an improved understanding between the system boundaries of the pavement

292

TIS2017_Book.indb 292 2/8/2017 8:04:05 PM

 life cycle when analyzed concomitantly from the As far as the pavement maintenance and rehabil-
economic and environmental viewpoint. For this itation (M&R) is concerned, three main groups of
purpose, the pavement LCC-LCA model uses alternative M&R strategies (scenarios) were con-
Carnegie Mellon University’s Economic Input- sidered to be available for application over the PAP
Output Life Cycle Assessment tool (EIO-LCA) of the pavement structure. The first two groups
(CMUGDI 2010). This tool utilizes the Leontief’s of alternative M&R strategies, hereafter named
methodology to relate the inter-sector monetary VDOT and Recycling-based VDOT strategy,
transactions sectors in the US economy, compiled respectively, were based on the M&R plan outlined
in a set of matrices by the Bureau of Economic by VDOT (VDOT 2014), in which functional and
Analysis (BEA) of the US Department of Com- structural treatments and a major rehabilitation
merce, with a set of environmental indicators (e.g. are applied in pre-established years (M&R activi-
consumption of fossil energy, airborne emissions, ties ID 1, 2 and 3 in Table 2, respectively). Briefly,
etc.) per monetary output of each industry sector the functional and structural treatments consist of
of the economy. The environmental burdens at sec- patching and asphalt concrete overlays with dif-
tor level associated with a particular commodity ferent areas of application and thickness, whereas
under analysis is therefore calculated by multiply- the major rehabilitation comprises the milling
ing its monetary value, previously adjusted to US and replacement of the existing bounded layers.
dollars of the EIO-LCA model’s year according
to sector specific economic indices from the US
Department of Labor, by the respective sectorial Table 1. Identification of the alternative M&R
environmental multipliers obtained from the EIO- scenarios.
LCA model. The US 2002 EIO-LCA benchmark
consumer price model for the US economy was pre- Type of scenario Scenario ID Scenario name
ferred to the Producer Price Index model because
VDOT 1 HMA – 0% RAP
the monetary values of the commodities whose 2 HMA – 15% RAP
environmental burdens the study aimed to quan- 3 HMA – 30% RAP
tify are better represented by retail price (e.g., con- 4 Sasobit® WMA –
struction equipment acquisition, tire acquisition, 0% RAP
lubricating oil acquisition, etc.), which allows for 5 Sasobit® WMA –
further accounting of the environmental impacts 15% RAP
associated with their distribution to wholesalers. 6 Sasobit® WMA –
30% RAP
Recycling-based 7 HMA – 0% RAP
3.2 Scope of the study
VDOT 8 HMA – 15% RAP
3.2.1 Functional unit 9 HMA – 30% RAP
The functional unit considered in this case study 10 Sasobit® WMA –
was defined as a 1 km one-way road pavement sec- 0% RAP
tion of an Interstate highway in Virginia, with two 11 Sasobit® WMA –
lanes, each of which was 3.66 m wide. The Project 15% RAP
Analysis Period (PAP) was 50 years (VDOT 2014), 12 Sasobit® WMA –
30% RAP
beginning in 2011 with the construction of the
Preventive 13 Microsurfacing –
pavement structure. The annual average daily traf- maintenance 0% RAP
fic for the first year was 20,000 vehicles, of which 14 THMACO – 0% RAPa
25% were trucks (5% were single-unit trucks and
95% were combination trucks). The traffic growth
rate was set equal to 3% per year.
The initial pavement structure was designed Table 2. M&R activities considered in each scenario, and
using the pavement structural design method respective application years.
AASHTO’93 (AASHTO 1993) for flexible pave-
M&R activity ID
ments, as defined by the Chapter V-Pavement M&R
Evaluation and Design of the Virginia Department Scenario ID 1 2 3 4 5 6
of Transportation’s (VDOT’s) Manual of Instruc-
tions for the Materials Division (VDOT 2014). The 1 to 6 12, 44 2 32 – – –
assumptions considered during the design process 7 to 12 12, 44 22 – 32 – –
and the Hot Mix Asphalt (HMA) mixtures prop- 13 9, 17, 25, – 32 – 7, 15, 23, –
erties are presented in Santos et al. (2017). Based 41, 49 39, 47
14 10, 18, 27, – 32 – – 7, 16, 24,
on those assumptions, a pavement structure was
41 50 39, 47
designed with a structural number of 7.38.

293

TIS2017_Book.indb 293 2/8/2017 8:04:05 PM

 Furthermore, the two aforementioned M&R sce- Table 3. Coefficients of VDOT’s load-related PPPM
narios were considered to be different from each expressed by the Equation (1) for asphalt pavements of
other in that only conventional asphalt materials interstate highways.
and treatments were implemented in the first group,
M&R activity category CCI0 a b c
while in the second group the major rehabilitation
(M&R activity ID 6 in Table 2) was carried out CM 100 9.176 9.18 1.27295
through the combination of the CCPR in-place RM 100 9.176 9.18 1.25062
recycling technique, and conventional asphalt lay- RC 100 9.176 9.18 1.22777
ers. The recycling-based M&R activity was designed
in such a way that it provided equivalent structural
capacity to the non-recycling-based counterpart and
took into account VDOT’s surface layers require- critical condition index immediately after treat-
ments for layers placed over recycling-based layers ment; and a, b, and c are the load-related PPPM
(VDOT 2013). The third group of M&R strategies coefficients (Table 3).
consisted of preventive maintenance treatments. VDOT did not develop individual PPPM for
The VDOT and Recycling-based VDOT strat- PM treatments. Thus, in this case study, the con-
egies were also further divided into HMA and sidered PM treatments (i.e., microsurfacing and
Sasobit® WMA scenarios with three distinct RAP THMACO) were respectively modelled as an
contents (0%, 15% and 30%). As for the preven- 8-point and 15-point improvement in the CCI of
tive alternative maintenance strategies, two addi- a road segment that takes place whenever the CCI
tional scenarios were considered depending on falls below the trigger value of 85 (Chowdhury
the type of preventive treatments adopted: micro- 2011). Once the treatment is applied, it is assumed
surfacing and thin hot mix asphalt concrete over- that the pavement deteriorates according to the
lay (THMACO) [M&R activities ID 4 and 5 in PPPM of a CM, without reduction of the effective
Table 2, respectively). A summary of all consid- age. In the case of the application of CM, RM and
ered scenario names is provided in Table 1. For RC treatments, the CCI is brought to the condition
further details of the M&R activities and actions of a brand new pavement (CCI equal to 100) and
considered in the different scenarios the reader is the age is restored to 0 regardless of the CCI value
referred to Santos et al. (2017). Table 2 shows the prior to the M&R activity application.
M&R activities considered in each scenario and To estimate the environmental impacts and
the respective application years. costs incurred by road users during the pavement
In order to determine pavement performance usage phase resulting from vehicles traveling over
over time, VDOT’s pavement Performance Predic- a rough pavement surface, a linear roughness
tion Models (PPPM) were used. VDOT developed prediction model, expressed in terms of Inter-
a set of PPPM in units of Critical Condition Index national Roughness Index (IRI), was considered
(CCI) as a function of time and category of the last (Equation 2).
M&R activity applied. CCI is an aggregated indi-
cator ranging from 0 (complete failure) to 100 (per-
fect pavement) that represents the worst of either IRI (t ) = IRII 0 IIRII grw × t (2)
load-related or non-load-related distresses. VDOT
classifies M&R activities into five categories: (0) Do where IRI(t) is the IRI value (m/km) in year t,
Nothing (DN), (1) Preventative Maintenance (PM), IRII 0 is the IRI immediately after the application
(2) Corrective Maintenance (CM), (3) Restorative of a given M&R activity and IRII grw , which was set
Maintenance (RM), and (4) Reconstruction/Reha- at 0.08 m/km (Bryce et al. 2014), is the IRI growth
bilitation (RC). Using the base form corresponding rate throughout time. It was assumed that the appli-
to Equation (1), VDOT defines PPPM for the last cation of an M&R activity other than PM restored
three categories (Stantec Consulting Services and the IRI to the value of a brand new pavement (IRI
Lochner 2007). The coefficients of VDOT’s load-re- equal to 0.87 m/km). The IRI reduction due to the
lated PPPM expressed through the Equation (1) for application of a PM treatment was determined
asphalt pavements of Interstate highways are pre- based on the expected treatment life and assumed
sented in Table 3 (Stantec Consulting Services and that there was no change in the IRII grw value after
Lochner 2007). the PM application (the same assumption was also
made in the case of the remaining M&R activities).
⎛ 1⎞
ln⎜ ⎟ By assuming treatment life cycle periods of three
b × c ⎝ t⎠
C (t ) = CCI
CCI CI 0 ea (1) and five years (Chowdhury 2011), respectively, for
microsurfacing and THMACO preventive treat-
where CCI(t) is the CCI in year t since the last ments, the resulting reductions in the IRI value
M&R activity (i.e., CM, RM or RC); CCI0 is the were found to be 0.24 and 0.40 m/km.

294

TIS2017_Book.indb 294 2/8/2017 8:04:05 PM

 3.2.2 System boundaries, system processes, life To account for the fact that specific heat capaci-
cycle inventory data and main assumptions ties of minerals and fluids increase substantially
The pavement LCC-LCA model includes six with temperature, the equations presented by
pavement life cycle phases but the environmen- Waples and Waples (2004a,b) were adopted, taking
tal burdens associated with the EOL phase were the temperature of 20ºC as the reference tempera-
disregarded based on the consideration of the ture. The heating requirements for the aggregates
“cut-off ” allocation method. Also excluded from applied in bound layers other than surface lay-
the system boundaries were the environmental ers were modeled by considering the specific heat
burdens due to the transportation of the work- value of limestone (880 J/[kg/ºC]). In the case of
force to the workplace. Furthermore, only real the surface layers, the value for quartzite (1013 J/
monetary flows were accounted for in order to [kg/ºC]) and diabase (860 J/[kg/ºC]) were taken to
avoid double counting the environmental impacts represent the aggregate used in the surface mixes
(Swarr et al. 2011). and THMACO, respectively. With regard to binder
For the sake of brevity only the methodology and water, the third equation proposed by Gambill
used to model the materials production phase is (1957) and the equation developed by Somerton
presented in detail. The various models used for (1992), both cited and displayed in Waples and
each component of the remaining pavement life Waples (2004b), were adopted, respectively. The
cycle phases, as well as the main data required to initial moisture contents of fine and coarse aggre-
run those models, can be found in Santos et al. gates were assumed to be 3% and 1%, whereas
(2017). for RAP a value of 4% was assumed. As for the
In this case study it was assumed that all asphalt HeatingEffF, a value of 80% was adopted for the
mixes were produced through a natural gas-fired production of all mix types after calibrating the
conventional drum-mix plant. In a conventional model with the data corresponding to the HMA
drum mix plant, to prevent additional aging of the production in the case study of Munster, Indiana,
RAP binder, RAP is not heated directly. Instead, reported by West et al. (2014). The HMA mixing
the virgin aggregates are superheated before-hand temperature was set at 160ºC and the initial tem-
so that when the RAP is introduced into the drum perature of all raw materials other than bitumen
it is dried and heated by conduction. Such a super- was assumed to be equal to the ambient tempera-
heated temperature is likely to cause additional ture of 15ºC. Bitumen was assumed to be stored at
energy consumption, which may eventually offset 160ºC in heated tanks located in the asphalt plant
the RAP’s associated economic and environmental facility. The volume of natural gas required to heat
benefits. the insulated storage tanks was calculated based on
In order to assess these tradeoffs, along with the the total quantity of binder heated, the total time
sensitivity of the air emissions due to the variations the bitumen spent in the tanks throughout the pav-
in composition and manufacturing temperature ing season and the heat capacity of the tanks. As
of the mixes and the moisture content of the raw for the WMA, whose mix design was considered
materials, the heat energy required to produce the the same as that of the homologous HMA, it was
asphalt mixes was determined through an energy assumed that the addition of 1.5% of Sasobit® per
balance represented by Equation (3). mass of bitumen reduced the mixing temperature
by 25ºC in relation to the reference temperature of
T fi
160ºC. This assumption was based on the range
M

∑ mi ∫ Ci (TT dT
i =0
(
Lv × mwvfv mwvo )
values of reduction of temperature of 20–30ºC
commonly referred to in the literature (D’Angelo
T0 i
Q= (3) et al. 2008). Moreover, it was also assumed that the
Heating Eff
E F RAP used in WMA could be blended with new
asphalt binder at this lower temperature.
where Q is the heat energy required to produce In order to determine the air emissions resulting
the asphalt mixture (J); mi is the mass of material i from the mixing process of all mixes considered in
(kg); M is the total number of materials, including this case study, a methodology was developed based
water; Tfi is the final temperature of the material on the Emission Factors (EFs) published by the
i (ºC); Toi is the initial temperature of the material AP-42 study of HMA plants (US EPA 2004) cor-
i (ºC); ci(T) is the specific heat capacity coefficient, as responding to a natural gas-fired filter-controlled
a function of temperature, of material i [J/(kg/ºC)]; drum-mix plan, and the thermal energy required
Lv is the latent heat required to evaporate water to produce the asphalt mixes. Firstly, the aver-
(2256 J/kg); mwvf is the final mass of water vapor age EFs referring to the production of an HMA
(kg); mwvo is the initial mass of water vapor (kg); with 0% RAP were taken as reference. Secondly,
and HeatingEffF is a factor that represents the cas- as CO2 emissions primarily result from fuel com-
ing losses. bustion, the average emission of this Greenhouse

295

TIS2017_Book.indb 295 2/8/2017 8:04:06 PM

 Gas (GHG) was combined with the fuel emis- was adopted in this study to conduct the impact
sion coefficient (53.1 kg/MMBtu) reported by US assessment step of the LCA. The TRACI impact
Energy Information Agency (EIA) to determine categories used in the analysis include acidification
the quantity of natural gas whose combustion air, eutrophication air, human health criteria pollut-
would release the same amount of CO2 (US EIA ants, and photochemical smog formation. The time-
2013). Thirdly, for each mix, an EF multiplier was adjusted characterization model for the Climate
determined through the ratio between the thermal Change (CC) impact category that was proposed by
energy computed with Equation (3) and the ther- Kendall (2012) was used, as opposed to the tradi-
mal energy calculated according to the procedures tional time-steady International Panel on Climate
previously described. Finally, GHGs and air pol- Change model. Furthermore, three energy-based
lutant EFs from the mix’s production were derived indicators were also included in the assessment:
by multiplying the EFs taken as reference by the (1) primary energy obtained from fossil resources, (2)
EF multipliers. The values of the EF multipliers primary energy obtained from non-fossil resources
and the natural gas consumption requirements for and (3) feedstock energy. The feedstock energy was
producing all mixes considered in this case study fully allocated to the virgin binder, with none attrib-
can be found in Santos et al. (2017). Emissions and uted to RAP. This assumption aimed to avoid dou-
energy consumption due to the operation of the ble counting, as it should have been accounted for in
wheel loader at the asphalt plant facility were esti- the previous pavement system.
mated based on the rate at which the wheel loader In computing the life cycle costs, the concept of
could move aggregates (Santos et al. 2015c) and net present value was applied and a real discount
the methodology adopted by the US EPA’s NON- rate of 2.3% was adopted (OMB 2013).
ROAD 2008 model (US EPA 2010).
In addition to the process-based components
described throughout this section, the input-output 4 RESULTS AND DISCUSSION
life cycle inventory approach was adopted to esti-
mate the environmental burdens associated with 4.1 Overall performance
the manufacturing, repair, maintenance, interest
on loan and insurance of the asphalt plant setup In order to determine the preference order of alter-
and auxiliary equipment. The environmental bur- native scenarios, the MCDM Technique for Order
dens were amortized by applying the portion of of Preference by Similarity to Ideal Solution (TOP-
the asphalt plant’s setup and auxiliary equipment SIS) method was applied. Three main criteria were
depreciation that was allocated to the quantity of considered: Highway Agency Costs (HAC), Road
asphalt mixes consumed in a given construction User Costs (RUC) and environmental impacts. The
activity and considering the average annual pro- last criterion was further broken down into eight
duction of asphalt mixes. A similar approach was sub-criteria, each representing one environmental
adopted in the construction, M&R, and transpor- impact category. To inform decision-makers about
tation of materials phases for determining the envi- the consequences of the weighting in the ranking
ronmental burdens associated with the construction of the alternative scenarios, a combinatorial weight
equipment and hauling trucks, but taking as alloca- assignment method was used for the main criteria,
tion factors the number of usage hours and hauling while the weights assigned to the environmental
km traveled for a given construction activity. sub-criteria remained unchangeable and equal to
As for the economic dimension, the proposed those adopted by the US-based Building for Eco-
pavement LCC-LCA model accounts for the costs nomic and Environmental Sustainability (BEES)
incurred by the highway agencies during the con- software (Lippiatt 2007). Since the energy demand
struction and M&R of the pavement, by the road indicators considered in the proposed LCC-LCA
users when facing a disruption of the normal model are not available in the BEES software, they
traffic flow as a consequence of the constraints were given a weight of five points each, which is
imposed by a work zone traffic management plan, the weight assigned to the Fossil Fuel Depletion
and throughout the PAP as a consequence of the impact category considered in BEES. All the
pavement’s deterioration. weights assigned to the environmental sub-criteria
were posteriorly rescaled, so that the sum of their
values totaled 100 points. Thus, in the MCDM, the
3.3 Life cycle impact assessment and life cycle final weight of each environmental sub-criterion
costs computation is the value resulting from multiplying the weight
The US-based impact assessment methodology, the of the main environmental criterion by the weight
Tool for the Reduction and Assessment of Chemi- determined, as explained above. The best scenario
cal and other environmental Impacts 2.0 – TRACI for all possible weighting combinations between the
2.0 (Bare et al. 2011) developed by the US EPA, three main criteria is displayed in Figure 1 through

296

TIS2017_Book.indb 296 2/8/2017 8:04:06 PM

 method, output variations are induced by varying
one input factor at a time, while all others are held
at their default values. Table 5 presents the trian-
gular diagrams that display the best scenarios for

Table 4. Values of the main criteria for each M&R sce-

nario, as well as the respective raking, when the main cri-
teria are considered to be equally important.

M&R
Scenario CC HAC RUC
ID (tonnes CO2-eq) (K$) (K$) Ranking

1 4 262 974 3 044 11

2 4 217 880 3 044 7
3 4 178 768 3 044 3
Figure 1. Best scenarios for all possible weighting com- 4 4 263 995 3 044 12
binations of the main criteria. (Note: WHAC = weight 5 4 220 901 3 044 9
assigned to highway agency costs; WRUC = weight assigned 6 4 178 788 3 044 4
to road user costs; WEnv = weight assigned to environmen- 7 4 121 904 2 978 8
tal impacts). 8 4 082 822 2 978 5
9 4 049 722 2 978 1
a triangular diagram. Each point in the triangle 10 4 121 923 2 978 10
area corresponds to a specific weighting set and 11 4 083 840 2 978 6
the relative weights always add up to a total weight 12 4 047 740 2 978 2
of 1 (or 100%). From Figure 1, we can conclude 13 3 897 1 001 3 702 14
that of the competing scenarios, scenarios 9 and 14 3 404 1 059 2 956 13
14 rank best. Of those, if the decision is exclusively
based on either HAC or RUC, scenario 9 ranks Table 5. Results of the sensitivity and scenario analysis
best. Alternately, if the environmental perform- expressed in terms of the best M&R scenarios for all pos-
ance is the only criterion taken into account, then sible weighting combinations of the main criteria.
scenario 14 outperforms the remaining scenarios.
In order to give the decision-maker a clearer
picture of the drawbacks and advantages of each
possible technological solution, expressed in terms
of emissions, costs and consumption of natural
resources, Table 4 presents for each alternative
solution the values of the main criteria as well as
the respective raking when they are considered to
be equally important. For the sake of brevity, the
CC impact category is taken as representative of
the environmental criteria.

4.2 Sensitivity and scenario analysis

Due to their recent application, there is a lack of
results obtained from comprehensive field studies
about the long-term performance of road pave-
ments that incorporate new pavement engineering
solutions. Thus, it is pertinent to consider that new
paving materials/solutions may not be as durable
as conventional materials.
Having this aspect in mind, a sensitivity and
scenario analysis were conducted to examine how
variations across a set of parameters and assump-
tions affect the outcomes, and thereby, the relative
merits of the alternatives being compared. For this
study, the one-(factor)-at-a-time sensitivity analy-
sis method was used (Pianosi et al. 2016). In this

297

TIS2017_Book.indb 297 2/8/2017 8:04:06 PM

 all possible weighting combinations between the REFERENCES
three main criteria when each model parameter is
individually changed +60% with respect to the base American Association of State Highway and Transportation
scenario. Table 5 clearly shows that the position of Officials (AASTO), 1993. Guide for design of pavement
scenarios 9 and 14 in the ranking of the best com- structures. 4th ed. Washington, DC: American Associa-
promise solutions is robust. In general, only a slight tion of State Highway and Transportation Officials.
Bare, J., 2011. TRACI 2.0: The tool for the reduction
change is seen when considering scenarios different and assessment of chemical and other environmental
from the baseline. The only exception to this out- impacts 2.0. Clean Technologies and Environmental
come is observed when different IRI grow rates are Policy, 13 (5): 687–696.
considered. In these circumstances, for a given set of Bryce, J., Flintsch, G. & Hall, R., 2014. A multi criteria
weights, scenario 1 was found to be the best compro- decision analysis technique for including environmen-
mise scenario. This happens when the weight values tal impacts in sustainable infrastructure management
assigned to the RUC criterion are high while those business practices. Transportation Research Part D:
assigned to the remaining criteria are low, and can Transport and Environment, 32: 435–445.
be explained by the increase in the roughness-related Carnegie Mellon University Green Design Institute
(CMUGDI), 2010. Economic input-output life cycle
user costs associated with higher IRI grow rates. assessment (EIO-LCA), U.S. 2002 Industry Bench-
mark model. Available from: http://www.eiolca.net
Chowdhury, T., 2011. Supporting document for the devel-
5 SUMMARY AND CONCLUSIONS opment and enhancement of the pavement maintenance
decision matrices used in the needs-based analysis. Vir-
This paper presents a comprehensive and inte- ginia Department of Transportation, Maintenance
grated pavement LCC-LCA model to investigate Division, Richmond, VA.
the potential environmental and economic benefits D’Angelo, J., Harm, E., Bartoszek, J., Baumgardner,
resulting from applying in-plant recycling mix- G., Corrigan, M., Cowsert, J., et al., 2008. Warm-
mix asphalt: European practice, Report No. FHWA-
tures, WMA, cold in-place recycling techniques PL-08-007, Federal Highway Administration, U.S.
and preventive treatments throughout the life cycle Department of Transportation.
of a pavement structure. Gambill, W., 1957. You can predict heat capacities.
For the conditions considered in this case study, Chemical Engineering, 64: 243–248.
the recycling-based VDOT M&R strategy, with Kendall, A., 2012. Time-adjusted global warming poten-
asphalt mixtures of type HMA containing 30% tials for LCA and carbon footprints. The International
RAP, has been shown to be more compatible with Journal of Life Cycle Assessment, 17: 1042–1049.
highway agency and road users’ demands for afford- Lippiat, B., 2007. BEESRG 4.0: building for environmen-
able road maintenance and usage over its life cycle tal and economic sustainability technical manual and
user guide (The National Institute of Standards and
than the remaining technical solutions investigated. Technology Report No. 7423).
Furthermore, a sensitivity and scenario analy- Office of Management and Budget, 2013. Discount rates
sis were undertaken to assess the robustness of the for cost-effectiveness, lease purchase, and related analy-
outcomes in response to variations in some of the ses. Table of past years discount rates from Appendix
most relevant input values. The analysis showed C of OMB Circular No. A−94.
that variances to the key assumptions considered Pianosi, F., Beven, K., Freer, J., Hall, J., Rougier, J.,
when assessing the life-cycle environmental and Stephenson, D. & Wagener, T. 2016. Sensitivity analy-
economic performances of multiple pavement sis of environmental models: a systematic review with
construction and maintenance practices do not practical workflow, Environmental Modelling & Soft-
ware, 79: 214–232.
considerably alter the overall advantage of imple- Santos, J., Flintsch, G. & Ferreira, A. 2017. Environ-
menting a recycling-based VDOT M&R strategy, mental and economic assessment of pavement con-
where the asphalt mixtures are of type HMA con- struction and management practices for enhancing
taining 30% RAP. pavement sustainability, Resources, Conservation and
Recycling, 116:15–31.
Santos, J., Ferreira, A. & Flintsch, G., 2015a. A life cycle
ACKNOWLEDGEMENTS assessment model for pavement management: meth-
odology and computational framework, International
This work has been supported by the project Journal of Pavement Engineering, 16(3): 268–286.
EMSURE- Energy and Mobility for Sustainable Santos, J., Bryce, J., Flintsch, G. & Ferreira, A., 2015b.
Regions (CENTRO-07-0224-FEDER-002004) and A comprehensive life cycle costs analysis of in-place
recycling and conventional pavement construction and
by the Transportation Pooled Fund TPF-5(268) maintenance practices. International Journal of Pave-
National Sustainable Pavement Consortium. João ment Engineering, (available online). http://dx.doi.org/
Santos wishes to thank the Portuguese Foundation 10.1080/10298436.2015.1122190
of Science and Technology (FCT) for a personal Santos, J., Bryce, J., Flintsch, G., Ferreira, A. & Diefend-
research grant (SFRH/BD/79982/2011). erfer, B., 2015c. A life cycle assessment of in-place

298

TIS2017_Book.indb 298 2/8/2017 8:04:07 PM

 recycling and conventional pavement construction Virginia Department of Transportation (VDOT), 2014.
and maintenance practices. Structure and Infrastruc- Manual of instructions for the materials division.
ture Engineering: Maintenance, Management, Life- Virginia Department of Transportation Materials
Cycle Design and Performance, 11(9): 119–1217. Division.
Somerton, W., 1992. Thermal properties and temperature- Virginia Department of Transportation (VDOT), 2013.
related behavior of rock/fluid systems. Developments Project selection guidelines for cold pavement recycling.
in Petroleum Science, 37, Amsterdam: Elsevier. Virginia Department of Transportation Materials
Stantec Consulting Services & Lochner, H., 2007. Devel- Division.
opment of performance prediction models for Virginia Waples, D. & Waples, J., 2004a. A review and evalua-
department of transportation pavement management tion of specific heat capacities of rocks, minerals, and
system. Virginia Department of Transportation, Rich- surface fluids, part 1: mineral and non-porous rocks.
mond, VA. Natural Resources Researches, 13(2): 97–122.
Swarr, T., Hunkeler, D., Klöpffer, W., Pesonen, H.-L., Waples, D. & Waples, J., 2004b. A review and evaluation
Ciroth, A., Brent, A. & Pagan, R., 2011. Environ- of specific heat capacities of rocks, minerals, and sur-
mental life cycle costing: a code of practice. Pensacola face fluids, part 2: fluids and porous rocks. Natural
(FL): Society of Environmental Chemistry and Toxi- Resources Researches, 13(2): 123–130.
cology (SETAC). West, R., Rodezno, C., Julian, G., Prowell, B., Frank,
United Sates Energy Information Administration (US B., Osborn, L. & Kriech, T., 2014. Field perform-
EIA), 2013. Carbon dioxide emissions coefficients. ance of warm mix asphalt technologies (National
United States Environmental Protection Agency (US Cooperative Highway Research Program Report
EPA), 2010. Exhaust and crankcase emission factors No. 779). Transportation Research Board, Wash-
for nonroad engine modelling - compression-ignition ington, D.C.
(Report No. NR-009d). The White House, 2014. FACT SHEET: Building a 21st
United States Environmental Protection Agency (US century infrastructure: increasing public and private
EPA), 2004. AP-42: Compilation of air pollutant emis- collaboration with the Build America investment initia-
sion factors (Vol. 1: Stationary point and area sources, tive, Office of the Press Secretary.
Chap. 11: Mineral products industry 11.1).

299

TIS2017_Book.indb 299 2/8/2017 8:04:07 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Road route planning for transporting wind turbines in Europe

F. Autelitano, E. Garilli & F. Giuliani

Dipartimento di Ingegneria Civile, dell’Ambiente, del Territorio e Architettura—DICATeA, University of Parma,
Parma, Italy

ABSTRACT: The upward trend in wind energy production had created a challenge for the roadway
system. The road infrastructure had to accommodate a wide range of complex vehicle configurations for
the transport of wind turbine components (nacelles, tower sections and rotor blades), which are often
considered oversize/overweight loads by the transport authorities. The paper provided an overview of
established difficulties encountered during the road transport of wind turbine components, examining
how the growth in the size of them, had affected the transport industry and the route assessment proc-
ess. The authors proposed an operational planning strategy, based on maximum swept path width, to
facilitate the identification of optimal routes for rotor blades transport. The methodology, which does
not substitute a rigid swept path analysis, could be used by the road managing authorities for checking
planimetric restrictions and constraints on their road infrastructure and for a ready evaluation of possible
abnormal road transport corridors.

1 INTRODUCTION Road transport of wind turbine components

represents both an economic problem and a tech-
The use and development of renewable energy nical one (Cotrell et al., 2014; Neff & Bai, 2012).
technologies plays an important role at Euro- These components, which are considered abnor-
pean level to respond to climate change and move mal loads (overweight/oversize) by transport
towards a low carbon future. Among renewable authorities, do not fit easily with road infrastruc-
energy sources, wind power represents one of the ture and conventional transport methods and
most important in the world, with an extraordinary vehicles. On the one hand, transport is an impor-
development in the last few years (EAWA, 2016; tant capital cost component. Abnormal vehicles
IEA-ETSP & IRENA, 2016; WWEA, 2016a). must obtain permits for every state through which
Transport of wind turbine components (nacelle, they will travel and are frequently required to be
tower sections and rotor blades) from the factory floor accompanied by escort vehicles, with one escort
to the project site represents a critical part of the logis- in front and one behind. International literature
tic and cost structure of a wind project, which requires has listed that transport cost is approximately the
specialized vehicles and suitable infrastructures (Hau, 6–7% of capital cost of the entire wind project
2013). It involves handling sensitive and valuable (Cotrell et al., 2014; Ray, 2007; TPI Composite,
components that can weigh several tonnes and be well 2003). On the other hand, the ever-increasing
over fifty meters in length (Cotrell et al., 2014). There demand for energy and the growing popularity
are three potential stages of turbine component deliv- of green power has led designers and engineers
ery: construction, operation and decommissioning. to build larger turbine units, which provide bet-
Initially, all turbine components are delivered to the ter land utilization and maximization of electric-
site during the construction period for installation. ity production (Abu-Rub et al., 2014). As turbines
Throughout the operational phase of the wind farm, grow, it becomes more difficult to transport their
it is possible that individual components may require components, which are already pushing the limits
replacement. The expected life of the wind farm is of what can be carried by trucks. In the European
20–25 years, after which the turbines will be decom- context these large and oddly shaped trucks have
missioned (Gash & Twele, 2012). Some efficient trans- usually to pass through at least one major urban
portation solutions for delivering turbine components center and sometimes to travel through residential
were improved. But, even when other modes of trans- neighborhoods. Moreover, they often can not fit
port are employed (rail, water and air) for some or through small underpasses or around sharp cor-
most of the journey, trucks are usually required for ners and heavy shipments may not be able to pass
the last leg in order to carry turbine components in safely over certain bridges and overpasses (IWEA,
austere or landlocked territories. 2011; Shihundu & Morrall, 2011).

301

TIS2017_Book.indb 301 2/8/2017 8:04:09 PM

 From that standpoint, the paper examined sev- Table 1. World wind turbine market share in 2015.
eral established difficulties encountered during
the road transport of wind turbine components, Market share
providing a critical analysis of the strategies and
Manufacturer Nation (%)
technologies used to overcome them. The authors
wanted to propose theoretical and practical trans- Goldwind China 12.5
port strategies, mainly based on swept path analysis, Vestas Denmark 11.8
for a ready route assessment and the identification General Electric USA 9.5
of abnormal road transport corridors already at Siemens Germany 8.0
planning level. Gamesa Spain 5.4
Enercon Germany 5.0
United Power China 4.9
2 WIND POWER IN EUROPE Ming Yang China 4.1
Envision China 4.0
The wind power has raised worldwide dramatically CSIC Haizhuang China 3.4
more than other renewable energy resource, reach- Others – 31.4
ing a wind power generation capacity of more than
456 GW at the end of June 2016 (WWEA, 2016b).
Europe was the undisputed global leader in wind The functioning principle of wind turbine is to
energy technology until about the year 2010 and now convert the kinetic energy contained in the wind
represents about the one third of the global market into mechanical power which can be used for spe-
with 150 GW of installed wind power capacity: 138 cific tasks or transformed into electricity via elec-
GW onshore and 12 GW offshore. Germany is the tromagnetic induction. The basic elements of the
EU country with the largest installed capacity (47 wind power system are the blades, the rotor hub,
GW), followed by Spain (23 GW), UK (14 GW), the nacelle and the tower (Fig. 1). Specifically, the
France (11 GW) and Italy (9 GW) (EAWA, 2016; blades capture and convert the wind’s energy to
Wind Europe, 2016; WWEA, 2016b). The Euro- rotational energy which is transferred by a rotor
pean onshore wind market is still quite dynamic: hub to the rotor shaft. The rotor shaft is also
some countries have a considerable market growth connected to the gearbox, which changes the low
potential (Norway, Poland and Turkey) and others rotating speed from the blades to a high rotating
are still experiencing a modest growth (Germany, speed for input to the generator. The rotor shaft,
France and The Netherlands). But, there are some rotor brake, gearbox and generator components
saturated markets (Denmark and UK), in which lit- are housed within a nacelle. The rotor blades, rotor
tle or no new installed capacity will be added in the hub and nacelle are supported by and elevated on
future (Henzelmann et al., 2016). a tower. Most large wind turbines are up-wind
Regarding the markets and manufacturers, the horizontal-axis turbines with three blades (Gash &
FTI Consulting annual report (2016) on wind tur- Twele, 2012; IEA-ETSP & IRENA, 2016).
bine original equipment manufacturers said that For utility-scale (MW-sized) sources of wind
2015 had been the first year in which a Chinese energy, a large number of wind turbines are usu-
firm (Goldwind) took the top spot as the world’s ally built close together to form a wind plant, also
largest suppliers of turbine (Table 1). called wind farm. The onshore or land-based tech-
These data were strongly influenced by the surg- nology has evolved over the last 10 years to reduce
ing Chinese market that installed a record of 28.7 the Levelized Cost of Energy (LCOE) (Abu-
GW last year. Nearly all the capacity of Chinese Rub et al., 2014; IEA-ETSAP & IRENA, 2016).
suppliers was reserved for the home market, where Machines have become bigger with taller hub
foreign manufacturers accounted less than 5% of heights, larger rotor diameters and in some cases
market share. The acquisition of French trans- bigger generator, depending on the specific wind
port and electricity conglomerate Alstom, gave to and site conditions. The growing trends of emerg-
General Electric (GE) a firm foothold in Europe. ing turbine size and power from 1980 are shown in
In the light of these considerations, the European Figure 2 (EAWA, 2009; IEA-ETSAP & IRENA,
wind service market is currently dominated by 2016). Currently, manufacturers offer utility-scale
Vestas, GE, Siemens, Gamesa and Enercon, even turbines with rotor diameters ranging from 70 m to
though other suppliers, such as Nordex-Acciona 140 m, generators from 1.7 MW to 5 MW and hub
(Germay-Spain) and Senvion (Germany) plays heights from 60 m to 150 m. In addition, some sup-
an active role. Some, like Vestas and Enercon, are pliers have started to offer 6–8 MW onshore plat-
wind pure-players, whereas others, such as GE and forms with a rotor diameter over 150 m. Table 2
Siemens, count wind turbine as only part of their provides an overview of the commercially avail-
energy generation interest. able onshore solutions at 2016 offered by the most

302

TIS2017_Book.indb 302 2/8/2017 8:04:09 PM

 important European manufacturers. Each turbine
is characterized by an alphanumeric acronym: the
letter is representative of the manufacturer, the
number describes the rotor diameter in meters and
the number in brackets gives the rated power in
megawatts.

3 WIND TURBINE COMPONENTS AS

ABORMAL LOADS

These days heavier nacelles, bigger tower section

and longer blades, mean that wind turbine com-
ponents are often considered abnormal or super
loads by road transport authorities. The European
Union had a clear legislation since 1996 (Directive
96/53), recently revised (Directive 2015/719), on
allowed weights and dimensions (length, width,
Figure 1. Basic components of an horizontal-axis tur- height and overhang) in road transport, even
bine with three blades. though there are big differences between the rules
and procedures currently applied in the Member
States. According to these standards, the abnor-
mal loads are those that exceed maximal accepted
weight (overweight) and/or dimensions (oversize).
A load is considered overweight when it exceeds
40 t for articulated vehicles and 44 t for articulated
vehicles loaded with 12.00 m containers. A load is
defined oversize when the total length of the vehi-
cle exceeds 12.00 m or 16.50 m in the case of an
articulated vehicle (18.75 m for road train), or the
overall width exceeds 2.55 m, or the overall height
exceeds 4.00 m. Vehicle weight is a hard constrain,
since excessive loads can damage pavements or
exceed the bearing capacity of bridges and culverts.
The dimensional constrain underlie different road
transport problems. The principal factors limiting
the permissible height are the clearances under
any overhead bridges or power lines. Abnormally
Figure 2. Growth of wind turbines.
long and wide vehicles could have difficulty in

Table 2. Overview of the European commercially available onshore wind turbines at 2016.

NORDEX
VESTAS GE SIEMENS GAMESA ENERCON ACCIONA SENVION

V90(1.8/2.0) GE100(1.7) SWT82(2.3) G80(2.0) E70(2.35) AW70(1.5) MM82(2.0)

V100(1.8/2.0) GE103(1.7) SWT93(2.3) G87(2.0) E82(2.35) AW77(1.5) MM92(2.0)
V110(2.0) GE 82.5(1.85) SWT101(2.3) G90(2.0) E92(2.35) AW82(1.5) MM100(2.0)
V90(3.0) GE87(1.85) SWT108(2.3) G97(2.0) E103(2.35) N90(2.5) M122(3.0–3.2–3.4)
V105(3.45) GE116(2.0/2.3) SWT120(2.3) G114(2.0) E101(3.05–3.5) N100(2.5) M114(3.2–3.4)
V112(3.45) GE107(2.2/2.4) SWT101(3.0–3.2–3.4) G106(2.5) E115(3.2) N117(2.4) M104(3.4)
V117(3.45) GE88(2.5) SWT108(3.0–3.2–3.4) G114(2.5) E126(4.2) AW100(3.0) M144 (3.4)
V126(3.45) GE100(2.5–2.75) SWT113(3.2) G126(2.5) E141(4.2) AW116(3.0) M126(6.2)
V136(3.45) GE120(2.5–2.75) SWT130(3.3) G132(3.3) E126(7.58) AW125(3.0) M152(6.2)
GE103(2.75–2.85) SWT154(6.0) G128(5.0) AW132(3.0)
GE103(3.2) SWT154(7.0) G132(5.0) N100(3.3)
GE130(3.2) N117(3.0)
GE137(3.4) N131(3.0)

303

TIS2017_Book.indb 303 2/8/2017 8:04:09 PM

 travelling around the sharp curves, particularly and dimension depending on where they are in
in urban areas, in mountain passes, on highway the tower. The towers have a conical shape, with
ramps, over certain roads with short vertical curves a diameter that diminishes from the base up to the
and some bridges. Moreover, these loads can cause tower head, so that lower sections (base) are the
obstruction and danger to other road users. shorter and heavier whereas higher sections (top)
The wind turbine components are usually trans- are taller and lighter. The most critical section with
ported separately as a single abnormal indivisible respect to its transport is therefore the bottom
load with special vehicle and telescopic multi-axle one. Bottom tower sections are generally limited
flat-bed trailers and assembled on site. A single to 4.3–4.6 m in diameter, but already in the case
turbine can require up to 10 hauls, not counting of a 100 m tower, the bottom section diameter is
the ones for the crane: 1 nacelle, 1 hub, 3 blades already over 5 m so that this segment can scarcely
and 3 to 5 tower sections. The wind turbine nacelle be transported by road in one piece. The only way
represents an overweight load (Fig. 3). It can weigh out is manufacturing it in half-shells which must
between 70–120 t (3 MW) to 130–240 t (5 MW) be welded together at the installation site (Gash &
depending on its drive train concept. Dimensions Twele, 2012; Hau, 2013).
instead do not typically exceed standard construc- The blade transport challenge is caused by the
tion trailer dimensions. difficulty of transporting long blades around turns,
Single-piece towers are occasionally used, through narrow passages, and beneath overhead
whereas in nearly all cases the steel or concrete obstructions. Some manufacturers have attempted
towers are produced in 3 to 5 sections of 25–35 m to design turbine blades that can be broken into
(25–55 t). These sections are transported in a multiple pieces during transport and resemble on
horizontal position by low-bed trailers (Fig. 4) site. So far, these efforts have met with no success
or semi-trailer equipped with clamp lift adapters. because segmented turbines are too fragile after
The tower transport challenge is therefore twofold: reassembly and do not retain their structural integ-
each section can be considered as an overweight rity under strong winds (Hau, 2013). The rotor
and overheight load. Tower sections vary in length blades in the range of 35 and 40 m are mostly
transported on edge in a pack of three in a spe-
cially designed long trailer (Fig. 5). Longer rotor
blades are instead carried individually in vertical
position (Fig. 6). When the maximum blade chord
length exceeds 4 m the passage below most bridges
(overhead height more of 5 m) is no longer pos-
sible. In this case, a special blade pitching mecha-
nism is used to turn the blade into the horizontal
position before passing below bridges.
Some extendible trailers are equipped with an
extra moveable support beam on top of the load
floor. This makes it possible to retract the vehi-
cle during transport without moving the load, by
temporarily creating a shorter wheelbase even the
tightest bends can be negotiated. Recently, a new
Figure 3. Transport of a nacelle. rotor blade adapter (hydraulic lifter) was designed

Figure 4. Transport of a tower segment. Figure 5. Transport of a pack of three rotor blades.

304

TIS2017_Book.indb 304 2/8/2017 8:04:09 PM

 and barriers), accommodation of street furniture
(e.g. lighting, traffic signals, power lines) and pro-
tection and upgrade of road related structures (e.g.
bridges and culverts). Analysis of each constraint
or group of constraints includes any concerns
regarding vertical alignment and swept path analy-
sis to assess horizontal alignment.
Vertical alignment issues are difficult to assess
during a desk-based exercise. Theoretical limits
on longitudinal gradients and cross falls are given
in few turbine transport guidelines, but the limits
vary significantly. For a 3MW wind turbine with
a rotor diameter bigger than 90 m, Vestas and
Nordex transport guidelines state that road lon-
Figure 6. Transport of a longer rotor blade. gitudinal slopes and longitudinal radii (convex or
concave) must be less than 6–8° and 200 m, respec-
tively. Moreover, all roads must be clear of over-
head obstructions to a minimum height of 5 m to
allow the passage of high loads.
The horizontal alignment of the existing road
layout is checked by undertaking vehicle swept
path analysis, i.e. the evaluation of the envelope
swept out by the sides of the vehicle body, or any
other part of the structure of the vehicle (Pecchini
& Giuliani, 2013).
These calculations generally require special soft-
ware. Some authors (Flores et al., 2015; Godavar-
thy et al., 2016) have conducted field tests with a
wind blade transporter vehicle and compared its
swept path to that generated by the AutoTURN
Figure 7. Transport of a blade lifted with the blade software, concluding that this software vehicle sim-
adapter. ulation can reasonably model the vehicle’s turning
characteristics. Figure 8 represents a swept path
for lifting and tilting upwards the wind blade up to analysis of a blade transport (Nordex N117-2.4)
approximately 60º (Fig. 7). Thus, the position of elaborated with AutoTURN 10. This process that
the rotor blade can be adjusted to the road, guar- demonstrates if the vehicles can manoeuvre safely
anteeing a flexible transport. and efficiently within the site layout is very useful
when it is contextualized in a specific area consid-
ering a preordained truck/trailer configuration.
4 ROUTE PLANNING AND SWEPT PATH For a ready planimetric verification at planning
ANALYSIS level should be more interesting to approach the
problem from a theoretical point of view, consid-
The transport of wind turbine components pro- ering the maximum vehicle swept path width, i.e.
vides a preliminary phase, defined route planning, the amount of roadway width that a truck covers
which includes the research of fully possible deliv- in negotiating a turn (AASHTO, 2011). Assuming a
ery routes and the individuation of the most effi- no skidding or tyre slippage and an infinitely small
cient path which best satisfies technical, financial tyre contact, it may be demonstrated that the rear
and functional criteria (Bazaras et al., 2013; Neff steering wheels of a vehicle follow circular paths
& Bai, 2012). The route assessment process con- concentric to those traced out by the front wheels
sists in finding, with GIS and on-site inspection, while the steering angle is constant. Effect of driver
a route with adequate width and height clearances characteristics are minimized by assuming that the
(especially in tunnel and bridge) and suitable turn- speed of the vehicle for the turning radius is less
ing radii (Petraška & Palšaitis, 2012; Ray, 2007). than 15 km/h. The geometrical path of a vehicle
Once identified the most efficient solution, the with front and rear steering is showed in Figure 9.
recognition of critical points in the route is neces- The swept path width (Ro-Ri), is established by the
sary. The dimensional requirements of abnormal outer trace of the tractor front overhang (Ro) and the
vehicle/loads may require alterations to the exist- path of the inside trailer rear wheel (Ri) (AASHTO,
ing road infrastructure (e.g. widening on corners 2011). The Ro value is described by the relation:

305

TIS2017_Book.indb 305 2/8/2017 8:04:11 PM

 R0 = front overhang is the trailer wheelbase, i.e. the dis-
2 tance from the king pin to the rearmost trailer axle
⎛ ⎛ W⎞
2
W ⎞ group. For wind turbine components transport,
⎜ ⎜ Ri + t ⎟ + (WB
Bt ) + T ⎟ + (WB OT )
2 2
BT + FO
⎜⎝ ⎝ 2⎠ 2 ⎟⎠ this parameter depends on the longest elements,
that are the rotor blades. Table 3 summarises the
(1) blade dimensions of some of the latest commer-
cially avaliable onshore solutions of Vestas, which
where WBT and WBt are the tractor and trailer is the most important European wind turbine
wheelbases, WT and Wt are the tractor and trailer manufacturer.
widths and FOT is the tractor front overhang. Considering the maximum swept path width
Figure 9 shows that the most influencing param- as principal parameter, a methodology and a
eter on the radius of the outer trace of the tractor corresponding operational planning tool could be
developed to facilitate efficient preliminary route
plans for wind turbine components. The Vestas
turbines were chosen as reference. In parallel,
different vehicle configurations were selected, con-
sidering the same tractor (Scania LA6 × 4HSZ) and
several load floor extensions of a telescopic trailer
(Nooteboom TELE-PX Super Wing Carrier)
(Fig. 10). These configurations were chosen so that
the trailer floor base length (Lt) almost coincided
with the rotor blade length, as a common practice
of transporting
According to equation 1, it was constructed a
diagram (Fig. 11), which relates the inside trailer
tyre radius, which can be considered in first
approximation equal to the inside road curve
radius, the maximum vehicle swept path width
and the trailer length (blade length). Once fixed
the inside road curve radius, the user can eas-
ily define the maximum swept path width for a
predetermined blade length. Vestas transport
guidelines gives a minimum inside turning radius
ranging from 28 to 43 m for a blade length of
45 m and 55–60 m, respectively. The passage in a
more binding path element with an assigned max-
imum swept path width can be satisfied through
a reduction of the trailer length which requires
Figure 8. Example of a swept path analysis with Auto- a reduced inner radius. The extent of the vehicle
TURN.

Table 3. Blade dimensions of onshore Vestas wind tur-

bines at 2016.

Blade dimensions

Rated power Length Max. chord

kW Wind turbine m m

1.8–2.0 V90 44.0 3.9

V100 49.0 3.9
2.0 V110 54.0 3.9
3.0 V90 44.0 3.5
V105 51.2 4.0
V112 54.7 4.0
3.45 V117 57.2 4.0
V126 61.7 4.0
Figure 9. Turning characteristics of a typical tractor/ V136 66.7 4.1
trailer configuration truck.

306

TIS2017_Book.indb 306 2/8/2017 8:04:13 PM

 Figure 10. Tractor/trailer configurations for the rotor blade transport.

Figure 11. Maximum swept path width vs inside road curve radius for different trailer floor base lengths (Lt).

length reduction generates an equal value of the swept path width of about 33 m. At this point it
rear blade overhang increase. This overhang can is possible to easily evaluate a swept path width
be usefully compensated, where there are the con- reduction assuming gradually greater rear blade
ditions, by the lifting of the blade with the pos- overhangs, compatibly with the technical wind
sible use of a rotor blade adapter (Fig. 7). By way turbine transport guidelines. The user has to
of example, a truck/trailer transporting a rotor shift vertically downward considering the under-
blade of 65 m and negotiating a turning maneu- lying curves, which represent shorter trailers or
ver in a local urban context (inside road curve temporarily shortened configurations. Thus, the
radius of 50 m) can be considered. By plotting same blade should be transported, assuming for
in the graph a vertical line from the x-coordinate example a rear blade overhang equal to 20 m, by a
value of 50 m, it intercepts the curve Lt = 65 m trailer with a floor base length of 40 m which cov-
(no blade rear overhang) identifying a maximum ers a maximum swept path width of 16 m.

307

TIS2017_Book.indb 307 2/8/2017 8:04:14 PM

 5 CONCLUSIONS Flores, J., Chan, S., Homola, D. 2015. A field test and
computer simulation study on the wind blade; Proceed-
The upward trend in wind energy production had ings of the fifth International symposium on highway
created a challenge for the roadway system. The road geometric design, Vancouver, 22–24 June 2015.
Gasch, R. & Twele, J. 2012. Wind power plants: Funda-
infrastructure had to accommodate a wide range of mentals, design, construction and operation. Berlin
complex vehicle configurations for the transport Heidelberg: Springer-Verlag.
of wind turbine components (nacelles, tower sec- Godavarthy, R.P., Russell, E. & Landman, D. 2016.
tions and rotor blades), which are often considered Using vehicle simulations to understand strategies
oversize/overweight loads by the transport authori- for accommodating oversize, overweight vehicles
ties. The paper provided an overview of nacelles, at roundabouts. Transportation Research Part A 87:
tower sections and rotor blades transport chal- 41–50.
lenges, examining how the growth of wind energy Hau, E. 2013. Wind turbines. Fundamentals, technologies,
in Europe, and especially the growth in the size of application, economics. Third edition. Berlin Heidel-
berg: Springer-Verlag.
wind turbine components, had affected the trans- Henzelmann, T., Büchele, R., Hoff, P. & Wollgam, G.
port industry and the route assessment process. The 2016. Onshore wind power. Playing the game by new
authors proposed an operational planning strategy, rules in a mature market. Think act. Beyond main-
based on maximum swept path width, to facilitate stream, January, 1–19.
the identification of optimal routes for rotor blades IEA-ETSAP & IRENA. 2016. Wind Power. Technology
transport. The methodology, which does not substi- Brief. International Energy Agency, Paris and Inter-
tute a rigid swept path analysis, could be used by the national Renewable Energy Agency, Abu Dhabi.
road managing authorities for checking planimetric IWEA. 2011. Transport of abnormal load to wind farm.
restrictions and constraints on their road infrastruc- Irish Wind Energy Association, Naas.
Neff, M.R. & Bai, Y. 2012. Developing a multi-modal
ture and for a ready evaluation of possible abnor- freight movement plan for the sustainable growth of
mal road transport corridors. To effectively connect wind energy related industries. Report 25-1121-0001-
the European industrial centres, the identifications 467. Mid-America Transportation Center. Lincoln,
of these road corridors should provide an interna- NE.
tional cooperation and involvement, especially for Nelson, V. 2014. Wind energy. Renewable energy and the
strategic trans-European routes. environment. Second edition. Boca Raton, FL: CRC
Press.
Pecchini, D. & Giuliani, F. 2013. Experimental test of an
REFERENCES articulated lorry swept path. Journal of Transportation
Engineering 139(12): 1174–1183.
AASHTO. 2011. AASHTO Greenbook. A policy on Petraška, A & Palšaitis, R. 2012. Evaluation criteria and
geometric design of highways and streets. American a route selection system for transportating oversize
Association of State Highway and Transportation and heavyweight cargoes. Transport 27(3), 327–334.
Officials, Washington, DC. Shihundu, D. & Morrall, J. 2011. Oversize/overweight
Abu-Rub, H., Malinowski, M. & Al-Haddad, K. 2014. vehicles: An investigation into the safety and space
Power electronics for renewable energy systems, trans- requirements for alternative energy projects. Proceed-
portation, and industrial applications. Chichester: John ings of the 2011 Conference and Exhibition of the
Wiley & Sons Ltd. Transportation Association of Canada, Edmonton,
Bazaras, D., Batarliene, N., Palšaitis, R. & Petraška, A. 11–14 September 2011.
2013. Optimal road route selection criteria system for Ray, J.J. 2007. A web-based spatial decision support
oversize goods transportation. Baltic Journal of Road system optimizes routes for oversize/overweight vehi-
and Bridge Engineering 8(1), 19–24. cles in Delaware. Decision Support Systems 43(4),
Cotrell, J., Stehly, T., Johnson, J., Roberts, J.O., Parker, 1171–1185.
Z., Scott, G. & Heimiller, D. 2014. Analysis of TPI Composites. 2003. Cost study for large wind turbine
transportation and logistics challenges affecting the blades: WindPACT blade system design studies. Report
deployment of larger wind turbines: Summary of SAND2003-1428. Sandia National Laboratories.
results. Technical Report NREL/TP-5000-61063. Albuquerque, NM and Livermore, CA.
National Renewable Energy Laboratory. Denver West Wind Europe. 2016. The European offshore wind
Parkway Golden, CO. industry. Key trends and statistics 1st half 2016.
EWEA. 2016. Wind in power: 2015 European statistics. Bruxelles.
European Wind Energy Association, Brussels. WWEA. 2016a. WWEA assessment report. World Wind
EAWA. 2009. Wind energy—the facts: a guide to the Energy Resource Assessment, Bonn.
technology, economics and future of wind power. WWEA. 2016b. WWEA half-year report 2016. World
London: Earthscan. Wind Energy Resource Assessment, Bonn.

308

TIS2017_Book.indb 308 2/8/2017 8:04:14 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The method of the friction diagram: New developments

and possible applications

P. Colonna, N. Berloco, P. Intini & V. Ranieri

DICATECh—Department of Civil, Environmental, Building Engineering and Chemistry,
Polytechnic University of Bari, Bari, Italy

ABSTRACT: This paper presents an updating and an application of a tool developed by the authors:
the Friction Diagram Method. The model on which the Friction Diagram is based was explained in
previous papers and here it will be briefly summarized. A sensitivity analysis has been performed in order
to establish which input parameters mostly influence the percentage of Friction Capital that a vehicle is
using traveling on a given road layout with different boundary conditions. Furthermore, some possible
applications of the Friction Diagram to real cases will be shown. In particular, the examples carried out
show how the Friction Diagram, together with the speed and visibility diagrams, could be considered an
important tool for selecting the speed limits of a road, both in dry and in wet conditions. The implications
of the method provide some guidance for both the design of new road infrastructure and the improvement
of existing roads.

1 INTRODUCTION combination of both the horizontal and vertical

alignments. On the other hand, different environ-
Road friction is a crucial matter in road design, mental characteristics, speed motion conditions
maintenance and operation since it is related to and road geometry features are considered as well.
safety issues (Wallman & Astrom 2001, Li et al The impact of each variable on the output of the
2013). Indeed, if the vehicle requires more friction model, represented by the percentage of friction
than the ability of the road to provide it, then a used (FUSED [%]) is shown through the use of a
loss of friction between tire and road occurs. The sensitivity-uncertainty matrix.
phenomenon of road friction depends on several Furthermore, some considerations about the
features belonging to different spheres: the road, maximum safe speed with regard to the risk of
the vehicle and the environment. skidding in each combination of road geometry
All these influences were considered in a new and vehicles are given. Currently, the recognized
method developed by the authors: the Friction method by the Italian Standard (D.M. 6792–2001)
Diagram Method (FDM). This method is based on for controlling the speed of a road layout is the
the estimate of the safety level of a road by calcu- speed diagram. In the case of an existing road
lating and comparing for each road section the fric- infrastructure, the speed diagram could not be ver-
tion force that the road can potentially provide to a ified if the horizontal alignment does not allow the
driving wheel “FP” with the resultant force acting imposed acceleration and deceleration distances
on it “FD”. The quantity “FP” is a measure of the (considering an acceleration rate: a = 0.8 m/s2) or
“Friction Capital” available to the vehicle, while the if the speed of consecutive geometric elements
quantity “FD” represents the share of “Friction exceeds the fixed thresholds. In these cases, the
Capital” used by the vehicle. The model on which designer reduces the allowed speed limit in order to
the Friction Diagram is based was explained in pre- comply with the design speed range. Similar situa-
vious works (Colonna et al. 2014, 2016a, b, c) and tions occur for the sight distance diagram. If it is
it will be briefly summarized in next section. not possible to change the road geometric layout
In this paper, the possible applications of the or to eliminate the obstacles that limit the visibil-
Friction Diagram to real cases will be shown. ity, the designer reduces the speed limit in order
Different vehicles are considered, with particular to guarantee the stopping distances, or eventually
regard to the Design Critical Vehicles (DCVs), the the overtaking distances, with the new maximum
vehicles presenting the worst performance for each allowed speed.

309

TIS2017_Book.indb 309 2/8/2017 8:04:16 PM

 Therefore, the road speed limits do not consider • they are only function of speed, cross slope and
the phenomenon of friction and, consequently, the radius of horizontal curvature;
numerous factors which can greatly influence its • the motion condition is characterized by a speed
value according to the geometrical characteristics constant, equal to the operating speed;
of the road, the type of the vehicle, its motion con- • the vehicle is modeled as a material point, with-
ditions and the boundary conditions. out considering possible different geometric
It can occur that the speed limit (which is derived features which are able to influence the redistri-
from the study of the visibility and the speed dia- bution of the vehicle weight on the axles;
grams) definitely ensures the stopping distances • the topographic factor “n” can represent by
and the correct deceleration and acceleration itself all the possible longitudinal actions.
between the geometric elements of the road; but it
could not guarantee the friction of all vehicles for The applicability of the framework given by Lamm
all plano-altimetric combinations and for all pos- about road friction was limited only to the horizontal
sible environmental conditions. curves. It does not provide any indication about other
The Friction Diagram Method provides a road geometric types and other important features:
useful tool to prevent these dangerous situa- • the rolling resistance of tires (see European Reg-
tions, for both the design of new road sections ulation No. 1222/2009);
and the improvement of safety of existing road • the influence of road geometry on the distribu-
infrastructures. tion of the vehicle weight on the road surface;
The contemporary and iterative use of the speed • the environmental and surface conditions;
diagram, the sight distance diagram and friction • the forces acting on the vehicle along the road.
diagram allows to obtain, for each section of a
road, the speed which satisfies the three checks and If the maximum friction force provided by the
ensures greater road safety, defined in this paper as road to the vehicle (ALIM) is greater or equal to the
the safe speed. resultant (R) of the cross and longitudinal forces
acting on the vehicle (ALIM ≥ R), then a vehicle can
be considered as safe from the point of view of
2 THE FRICTION DIAGRAM METHOD friction. Hence, the quantity ALIM can be defined
as the Friction Potential, the limit performance of
2.1 Update of the 3rd criterion of Lamm the road in terms of friction.
Among all the authors who studied the phenom- Therefore, in order to correctly estimate the fric-
enon of road friction, Lamm is one of the most tion along a road layout, the comparison between
important. The outcome of his studies is the the two above defined quantities in all the possible
Lamm’s Third Criterion (Lamm et al. 1998, 1999, road geometric conditions can be made, once the
2002) which determines the level of safety design parameters reported in Table 2 are known.
of a curve road section (good, fair or poor) by The Friction Diagram Method FDM (Colonna
comparing the “side friction assumed” (fRA) with et al. 2016c) is a tool able to analyze the friction of
the “side friction demand” (fRD). a road layout considering at the same time physical,
The first coefficient “fRA” is the maximum cross dynamics and environmental factors. It provides an
friction coefficient that the road can provide to the evaluation of the skidding risk associated to a road
vehicle in the design conditions. The second coef- layout by introducing a synthetic parameter, the
ficient “fRD” represents the friction coefficient that Friction Used (FUSED). The Friction Used is plotted
a vehicle, considered as a material point, requires on the y-axis against the distances on the x-axis of
from the road when it navigates a curve at a con- a diagram. This diagram, the “Friction Diagram”,
stant speed equal to the operating speed “S85”. allows to individuate high-risk road sections with
Therefore, the Lamm’s Third Criterion allows to respect to friction by a simple visual inspection.
measure the level of safety by considering the dif- The Friction Diagram can be plotted below the
ference between fRD and fRA. These two terms abide road elevation profile, by connecting the values of
by the following hypotheses: the FUSED obtained for each cross section.

Table 2. Factors influencing ALIM and R.

Table 1. III Lamm’s criterion.
Factors influencing ALIM Factors influencing R
Level of safety design Recommended interval
Adherent weight Road geometry
Good fRA − fRD ≥ 0.01 Friction coefficient Vehicle speed
Fair −0.04 ≤ fRA − fRD < 0.01 Boundary conditions
Poor fRA − fRD < −0.04 Motion conditions

310

TIS2017_Book.indb 310 2/8/2017 8:04:16 PM

 The Friction Diagram Method FDM is based ding risk, FUSED will be close to 100%. Moreover, if
on three key concepts: the vehicle stability is not guaranteed according to
the FDM, then FUSED will be greater than its limit.
− Friction Potential “FP”;
− Friction Demand “FD”;
2.5 Methodological approach to implement
− Friction Used “FUSED”.
the Friction Diagram Method
In order to compute the FUSED value in each cross sec-
2.2 The Friction Potential tion of a road layout and to plot the related Friction
The Friction Potential is the maximum friction Diagram, the following steps have to be followed:
force that a road surface can provide to the vehi- • analysis of road geometry and motion conditions;
cle (Padmanaban & Pawar 2015). It varies with the • estimation of the Friction Potential;
adherent weight (Schaar 1993), the friction coef- • computation of the resultant of the longitudinal
ficient (Canale et al. 1998) and the road geometry. forces acting on the vehicle;
It is computed as follows: • computation of the resultant of the cross forces
acting on the vehicle;
|Fp| = Pa (g, s) × fa (c, s) [N] (1) • computation of the resultant R of all the forces
acting on the vehicle, equal to the Friction
where: Demand, except for the sense of the vector;
− Pa (g, s) = adherent weight on the driving • comparison between the Friction Demand and
wheels, depending on the speed “s” and the road the Friction Potential, by computing the Fric-
geometry “g”; tion Used.
− fa (c, s) = friction coefficient in any direction The FDM has been implemented in a spread-
(depending on the speed “s” and the surface sheet which allows the easy application of the
conditions “c”). method, once the “Road Layout Data” and the
“Vehicle Data”, reported in Table 3 are known.
2.3 The Friction Demand In particular, the variables considered in the
Friction Diagram Method for the calculation of
The Friction Demand is the friction force required the FUSED are the following: road friction coefficient
by the vehicle to the road. It is equal to the resultant (fa), cross slope ϕ and longitudinal slope αm, vehicle
force of all the longitudinal and cross forces acting characteristics (wheelbase [m] (PL) and front track
on the vehicle. It is computed as follows: [m] (PC), height of the center of gravity [m] (hG),
longitudinal section [m2] (SL) and cross section [m2]
FD = (L2 + C2)0.5 [N] (2) (SC), longitudinal drag coefficient [−] (C(N)L) and
cross drag factor [−] (C(N)C), mass (m)), resistance
where: class of the tires, vehicle speed, acceleration, decel-
− “L”, sum of all the longitudinal forces: rolling eration, wind speed and wind direction.
resistance, aerodynamic drag, slope resistance,
curve resistance, longitudinal wind force; 2.6 The design critical vehicle
− “C”, sum of all the cross forces: centrifugal The forces acting on the vehicle taken into account
force, cross component of the vehicle weight, by the FDM are the following:
cross wind force.
− weight force, depending on the vehicle mass;
− inertial actions, depending on the vehicle mass
2.4 The Friction Used and the speed variation;
The parameter introduced, able to immediately
define the skidding risk, is the Friction Used. It is Table 3. Road layout data and vehicle data needed for
defined as: the implementation of the method.

FUSED = FD/FP × 100 [%] (3) Road layout data Vehicle data

Friction coefficient Mass

If the Fused is less than 100% (if the Friction
Cross section Height of center of gravity
Demand is less than the Friction Potential), then
Travel direction Longitudinal and cross pitch
the vehicle is in safety conditions. Therefore, safety
Design speed Vehicle cross section
conditions are defined by the distance of the FUSED Allowed acceleration Vehicle longitudinal section
from its limit value. In very safe situations, FUSED Allowed deceleration Drag coefficients
will be low; while in situations related to high skid-

311

TIS2017_Book.indb 311 2/8/2017 8:04:16 PM

 − centrifugal forces, depending on the vehicle Since cross slope is considered in the static equi-
mass, the vehicle speed, the horizontal and verti- librium of a vehicle navigating a curve section, the
cal radii of curvature; spiral transition curve is divided into two sections:
− aerodynamic actions, depending on the wind “ante level section” (a. l. s.) and “post level section”
speed and the longitudinal and cross sections of (p. l. s.). They represent the sections before and
the vehicle. after the point where the cross slope of the external
lane is equal to zero (level section).
All the actions above reported depend on vehi- Three motion conditions were considered:
cle characteristics (Jadav Chetan & Patel Priyal constant speed, acceleration and deceleration. In
2012, Genta 1983, Zagati et al. 1983). Therefore, order to individuate the characteristics of the criti-
for a given road layout, different vehicles corre- cal vehicle related to each of the 20 combinations
spond to different FD and FP values. Hence, since of the geometric elements shown in Figure 1, the
FUSED is the ratio between FD and FP, different vehi- influence of the FUSED value on the 6 vehicle param-
cles are related to different FUSED, and to different eters shown in Table 4 and previously defined, were
Friction diagrams. The analysis of a road layout evaluated.
is not complete if only data belonging to a given Among the 120 resultant combinations (20 × 6),
type of vehicle are considered. In fact, in this case, 6 recurrent combinations representing theoretical
the Friction Diagram is valid only for that type of critical vehicles were identified. Each theoretical
vehicle. Preliminary analyses (Colonna et al. 2014; critical vehicle was considered both having a front
Colonna et al. 2016a) revealed that: wheel drive and a rear wheel drive.
− it is not possible to individuate a unique type of These critical vehicles should be considered
vehicle associated to FUSED values always greater merely theoretical, since each of them represents
than the other vehicles; an unfavorable combination of parameters among
− the difference between the FUSED values related all the possible combinations.
to different type of vehicles is included between Further developments of the research led to
0 [%] and 35 [%]. select a set of real critical vehicles, in order to repre-
sent the real different types of vehicles commercially
Therefore, the type of vehicle to be used for available. The mean characteristics of the chosen
computing the FUSED have to be identified, in vari- real critical vehicles are shown in Table 5. How-
ous boundary conditions, without underestimating ever, further research developments will eventually
the level of skidding risk. This means that the type include other types of vehicle according to their
of vehicle maximizing the FUSED value in different performances with respect to the skidding risk.
conditions has to be found.
For this aim, in the FDM, the “Design Critical
Vehicle—DCV” is introduced: it is the vehicle to
be considered for plotting the Friction Diagram
in different road geometric conditions. Using the
DCV allows road designers to control for skidding
risk by using a unique Friction Diagram. Other
studies by the authors revealed that (Colonna et al.
2016b, c):
− the vehicle associated to the maximum value of
FUSED varies with the specific type of plano-alti-
metric road section; Figure 1. Possible combinations of both the horizontal
− the value of FUSED varies with the type of trac- and vertical alignments.
tion (front or rear wheel drive) in uphill and
downhill road sections;
Table 4. Theoretical critical vehicle features.
− the motion condition (constant speed, accelera-
tion, deceleration) largely influences FUSED. Critical C(N)L C(N)C hG PL PC M
Hence, a set of critical vehicles, rather than a vehicle [−] [−] [m] [m] [m] [kg]
unique vehicle, was considered. Each vehicle of
I 1.15 1.15 1.50 6 2.50 7,031
the set is “critical” in a given combination of road
II 1.15 1.15 0.75 6 1.50 2,109
geometry and motion conditions.
III 1.15 1.15 1.50 1.7 1.50 1,195
Twenty road geometric conditions were stud- IV 1.15 1.15 0.75 1.7 1.50 598
ied, obtained by combining horizontal align- V 1.15 1.15 0.75 6 2.50 3,515
ment (4 conditions) and the vertical alignment VI 1.15 1.15 1.50 6 1.50 4,218
(5 conditions).

312

TIS2017_Book.indb 312 2/8/2017 8:04:16 PM

 Table 5. Real critical vehicle features.

Critical* C(N)L C(N)C hG PL PC M

vehicle [−] [−] [m] [m] [m] [kg]

I 0.32 1 0.76 2.33 1.46 972

II 0.33 1 0.84 2.78 1.61 1,829
III 0.38 1 0.87 2.56 1.54 1,647
IV 0.60 1 1.65 6.30 2.23 15,750
V 0.31 1 1.28 3.52 1.76 3,432
VI 0.70 1 1.10 3.57 1.73 3,579

*I—City Car; II—Sport Car; III—Cross-country;

IV—Bus; V—Van; VI—Light Truck.

Figure 2. The sensitivity-uncertainty matrix.

The further applications shown in this study are
based on the real critical vehicles. Therefore, for each
cross section characterized by a given geometry, the
ability of data, for each parameter three degrees of
maximum FUSED value among the six FUSED values
uncertainty have been assigned qualitatively: low,
computed for each real critical vehicle is chosen.
medium and high.
The Friction Diagram is obtained by connecting
Then, a sensitivity analysis was carried out
the FUSED values obtained for each cross section.
considering:
− 21 parameters, characterizing each of them with
3 SENSITIVITY ANALYSIS three values (standard, minimum and maximum);
− 27 plano-altimetric combinations;
3.1 Input parameters − 3 different motion conditions of the vehicle
(constant speed, uniform acceleration or
A sensitivity analysis has been performed in order
deceleration).
to establish which input parameters mostly influ-
ence FUSED. The several factors taken into account Therefore, it was possible to verify the varia-
(Colonna 2016) belonging to different spheres of tion of FUSED for each of the 5,103 cases analyzed,
influence, are listed as follows. coming to the following matrix of uncertainty vs
sensitivity.
− Vehicle features: Longitudinal drag coefficient
The first five parameters that mostly influence
C(N)L [−]; Cross drag coefficient C(N)C [−];
the model, ranked by decreasing importance, are:
Wheelbase PL [m]; Front track PC [m]; Height of
acceleration, speed and deceleration of the vehicle,
the center of gravity hG [m]; Mass M [Kg]; Roll-
the horizontal curve radius and, finally, the friction
ing resistance coefficient RRC [Kg/t];
coefficient.
− Road design features: Longitudinal slope m
The high degree of uncertainty related to accel-
[rad]; Cross slope ϕ [rad]; Radius of the circular
eration and deceleration is caused by the value
curve RC [m]; Parameter of the spiral transition
assumed by Italian standards, equal to ± 0.8 m/
curve A [m]; Length of the horizontal curve
s2. Realistically, these values can be appreciably
s [m]; Radius of the convex vertical curve RD [m];
greater. The medium degree of uncertainty related
Radius of the concave vertical curve RS [m];
to speed is caused by the use of theoretical and not
− Vehicle dynamics: Speed VV [km/h]; Friction
real or detected speeds. Similar considerations can
coefficient fa [−]; Acceleration rate A [m/s2];
be made for fa since the state and age of the surface
Deceleration rate D [m/s2];
course and the tires are often not considered.
− Environmental conditions: Air speed VA [km/h];
In order to get reliable results by applying the
Air direction (with respect to the vehicle) [°];
FDM, it is necessary to have as accurate as pos-
Air density [Kg/m3].
sible information about these parameters that are
able to influence the model more than the others.
3.2 The sensitivity-uncertainty matrix
For each analyzed parameter, the ranges of vari- 4 USING THE FRICTION DIAGRAM
ability were identified, in consideration of the METHOD TO DEFINE THE SAFE SPEED
boundary conditions, the types of vehicles and the
motion condition. In relation to the variability of In case of new road infrastructures design, the friction
each parameter and to the easy or difficult avail- diagram can be used as a “design tool”, conditioning

313

TIS2017_Book.indb 313 2/8/2017 8:04:16 PM

 the horizontal and vertical layout of the road in order road infrastructure are the speed diagram and the
to always ensure the friction used percentage (FUSED) sight distance diagram, both required by the exist-
less than 100%, for each road section. ing Italian Standard.
The friction diagram should be used in con- In the first iterative loop, the existing road infra-
junction with the speed and the sight distance dia- structure is approximated to a type of road defined
grams, being sure that, for each road section, a safe by the Italian Standard (even if the existing road
speed is assigned, within the design speed range, was built certainly long before the entry into force
in order to verify the three checks, conducted in of the Standard). Subsequently, the speed range of
both directions of travel and for different bound- that type of road is identified and it enables the
ary conditions. construction of the speed diagram (in both direc-
In the case of safety improving of an exist- tions of travel). Once the speed diagram is known,
ing road infrastructure, the friction diagram it is possible to plot the visibility diagram (in both
can be used as a diagnostic tool, useful to iden- directions of travel).
tify the road sections, characterized by a specific In the road sections where the verification of
plano-altimetric combination, in which the friction one of the two diagrams is not satisfied, the two
used could exceed the friction potential. Also in checks will have to be repeated, lowering progres-
this case, it should be used with the speed and the sively the speed in order to obtain a value able to
sight distance diagrams, for each travel direction satisfy both of them.
and for different boundary conditions. If this value is acceptable for the purpose of the
For improvement of existing roads, the safe level of service to ensure on the road infrastructure,
speed, intended as the lowest speed value among for its strategic function and for the type of traffic,
the three speed values which satisfy respectively the recommended improvements will avoid varia-
the visibility, the friction and the speed diagram, tions on the road geometry and will be focused on
assumes an even more important value. other types of countermeasures. Otherwise, if at
The two possible uses of the friction diagram, the same time, there is the availability of a high
although apparently similar, are conceptually dif- budget and there are no territorial limiting restric-
ferent from each other. tions, functional improvements on the geometry
In the case of a new road, the designer can define will be implemented.
the geometrical elements of the road in such a way The above described procedure, however, does
that the three checks are satisfied. Depending on not eliminate the risk of skidding of a generic road
the function of the new infrastructure, on its inter- vehicle because:
actions with the land use and the type of expected
− the speed diagram does not take into account
traffic, it will be possible to fix for each road sec-
the vertical alignment of the road;
tion the minimum safe speed. The safe speed will
− the two tests do not take into account the differ-
influence the road geometry, as a result.
ent types of road vehicles;
The higher is the strategic value of the infra-
− the two tests do not take into account the condi-
structure, the higher is the level of service to be
tions of the road surface (new, dry, wet, slippery,
guaranteed and, therefore, the higher is the value
smooth);
of the minimum safe speed to ensure in the differ-
− the two tests do not take into account the differ-
ent sections of the road.
ent motion conditions of vehicles.
In the case of safety improvement of existing
road infrastructures, it is firstly necessary to dis- The friction diagram is therefore the tool that,
tinguish three types of intervention that depend contextually used with the speed and visibility
on the available budget, the function of the road diagrams, ensures higher road safety. The loop
under study and the boundary conditions: described above should be conducted simultane-
ously with the three procedures. For each road sec-
− functional improvements on the geometry (high
tion, there will be a threshold value of the speed
budget, strategic function of the road and not
which satisfies the speed diagram, a threshold
limiting boundary conditions);
value of the speed which satisfies the visibility dia-
− functional improvements that exclude geomet-
gram and a threshold value of the speed which sat-
ric variations (medium budget, not important
isfies the friction diagram. The safe speed, for that
function of the road and limiting boundary
section, will be the lowest speed among these three
conditions);
threshold values and it will be used for the subse-
− minimal functional improvements (low budget,
quent plotting of the other two diagrams.
negligible function of the road, and limiting
Once completed the diagnosis of the road, using
boundary conditions).
the conditions previously explained, the improve-
The diagnostic tools that it is possible to use ment interventions can be better identified. If
simultaneously for the verification of an existing improvements on the road geometry are not

314

TIS2017_Book.indb 314 2/8/2017 8:04:17 PM

 feasible, it will be possible to implement actions to
improve the visibility, i.e. with the possible elimina-
tion of obstacles in the vicinity of the road, and/or
improve the friction, i.e. by remaking the surface
course of the road pavement. Surely, even in the
case of low budget available, it will be possible to
operate on the speed limits, redesigning the road
signs.
The new project of road signs should be based
on the identification of the safe speed and joined by
solutions that ensure compliance with the imposed
speed limits (speed control) in order to guarantee
the success of the countermeasure.
The example reported in this section refers to a
rural road located in the District of Bari, in Italy
(S.P. 239). Only a part of the road is shown for
graphic reasons, selected in order to support the
procedure previously described.
The road is not a strategic infrastructure in the
national road network context and it is located in
an environmentally protected area with a lot of
ancient olive trees considered natural archaeologi-
cal monuments. The section shown in the example
has a constant longitudinal slope equal to 4.5%. In
order to improve the road safety, interventions on
the geometry are therefore excluded as well as the
removal of olive trees for increasing the visibility.
Considering the cross section of the road, about
eight meters wide, the road has been approxi-
mated to the road type—F2-defined by the Italian
Standard, with a design speed ranging from 60 to
100 km/h.
Hence, considering a maximum design speed
equal to 100 km/h, it is possible to note (Fig. 3)
that none of the three checks is satisfied (for the
entire length or for some parts of the section con-
sidered). Once the same checks were carried out
with a maximum speed of 70 km/h, it is possible to
note that the visibility and the speed checks are sat-
isfied, while the friction check is not met in a small
part of the road section and for only one direction
of travel (“return” direction in the Figure 3).
Further calculations have been conducted itera-
tively in order to obtain the speed satisfying the
friction check in every part of the road section. Figure 3. Searching the safe speed in a road section.
Its value is equal to 66.67 km/h. This speed is the
safe speed for the road under study and it elimi- sections where it is possible to allow it. In the exam-
nates the skidding risk for any vehicle and for wet ple carried out, the control of the new speed limits
conditions. has been guaranteed with an automatic speed con-
In the re-design process of the road signs, trol system with the possibility to sanction users
the limit of 70 km/h must therefore be fixed on the who do not comply with them.
entire road but not in approaching the curve for the
return direction, where it will be equal to 60 km/h.
The study of the entire infrastructure leads to 5 CONCLUSIONS AND FURTHER
various safe speeds for each road section. The RESEARCH
selection of speed limits will be carried out trying
to uniform as much as possible these speed values This paper presents an update of a tool developed
and at the same time to ensure a higher speed in by the authors: the Friction Diagram Method.

315

TIS2017_Book.indb 315 2/8/2017 8:04:17 PM

 A sensitivity analysis has been carried out in order in real time. The proposed method should be diver-
to identify the input parameters on which it is nec- sified for roads where traditional and self-driving
essary to pay more attention for the proper imple- vehicles can circulate together and roads exclu-
mentation of the FDM. sively devoted to next-generation vehicles.
At the same time, in the example shown before,
the attention is focused on the road sections in
REFERENCES
which, while the speed and the visibility diagrams
are verified, the risk of skidding is still present. Canale, S., Leonardi, S. & Nicosia, F. 1998. Nuovi criteri
The application of the method leads to identify the progettuali per una politica di sicurezza stradale, Atti
“safe speed”, namely the speed able to verify all the del XXIII Convegno Nazionale Stradale dell’A.I.P.C.R.,
three checks about speed, visibility and friction. Verona, Italy.
Therefore, it provides a rational tool for improving Colonna, P., Berloco, N., Intini, P., Perruccio, A. & Ranieri,
existing road infrastructures and selecting safety V. 2016a. Proposal of a new method for analyzing the
countermeasures, including the re-design of the road safety conditions related to friction. International
road signs and the speed limits, under dry and wet Journal of Systems Applications, Engineering & Develop-
conditions. This can be useful since sometimes the ment, 10, 87–96. University Press, UK.
Colonna, P., Berloco, N., Intini, P., Perruccio, A. & Ranieri,
speed limits are not selected in a rational way by the V. 2016b. Sicurezza stradale—un approccio scientifico a
road Agencies or designers, setting speed limit val- un problema tecnico e comportamentale. WIP Edizioni,
ues that could not eliminate the risk of skidding. Bari, Italy.
Regarding the critical vehicles considered in this Colonna, P., Berloco, N., Intini, P., Perruccio, A. & Ranieri,
paper, compared to six ideal vehicles considered in V. 2016c. Evaluating the skidding risk of a road layout for
the previous studies, six real vehicles were identi- all types of vehicle. Transportation Research Record: Jour-
fied. While the theoretical critical vehicles represent nal of the Transportation Research Board, 2591, 94–102.
the maximum unfavorable conditions among all the Colonna, P., Perruccio, A. & Ranieri, V. 2014. Influence
possible combinations, the real ones represent the of vehicle characteristics on skidding risk, SCORE@
POLIBA2014 Conference, Bari, Italy.
prevailing traffic composition on the road under European Parliament Council. 2009. Regulation (EC) No.
study and, at the same time, represent the real dif- 1222 on the labelling of tyres with respect to fuel effi-
ferent types of vehicles commercially available. ciency and other essential parameters.
Future insights will deepen the analysis by divid- Genta, G. 1983. Meccanica dell’autoveicolo. Levrotto &
ing the critical vehicles into two categories: the light Bella.
vehicles (city car, sport car, cross-country vehicles Italian Ministry of Infrastructures and Transports. 2001.
etc.) and the heavy vehicles (buses, vans, light trucks Functional and geometric standards for roads construc-
etc.). In this way, the safe speeds, and consequently tion. D.M. n. 6792. Rome. Italy.
the speed limits, could be differentiated for the two Jadav Chetan, S. & Patel Priyal, R. 2012. Parametric Analy-
sis of Four Wheel Vehicle Using Adams/Car, Interna-
macro categories. With this purpose, the use of var- tional Journal of Computational Engineering Research.
iable message signs would be appropriate, showing Lamm, R., Psarianos, B., & Cafiso, S. 2002. Safety evalu-
the speed limits in case of dry or wet conditions, for ation process for two-lane rural roads: A 10-year
passenger cars and heavy vehicles. review. Transportation Research Record: Journal of the
Considering the current and future develop- Transportation Research Board, 1796, 51–59.
ments related to self-driving cars, the speed that Lamm, R., Psarianos, B., & Mailaender, T. 1999. High-
averts the skidding risk could play an even more way Design and Traffic Safety Engineering Handbook.
important role in the future. The technological McGraw-Hill.
equipment of future vehicles could autonomously: Lamm, R., Psarianos, B., Choueiri, E. M. & Solilmezoglou
G. 1998. A practical safety approach to highway geomet-
− properly modulate the acceleration and decelera- ric design International studies: Germany, Grece, Leba-
tion between consecutive geometric elements of non and USA. Transportation Research Circular. USA.
the road (with databases containing the design Li, Y., Liu, C., & Ding, L. 2013. Impact of pavement con-
features of the traveled road); ditions on crash severity. Accident Analysis & Preven-
tion, 59, 399–406.
− measure the sight distances (using a variety of Padmanaban, S. & Pawar, P. 2015. Estimation of Tire Fric-
techniques such as radar, lidar, GPS, odometry) tion Potential Characteristics by Slip Based On-Road
and therefore modulate the speed to ensure the Test Using WFT, SAE Technical Paper 2015-26-0225.
stopping distance; Schaar, A. 1993. Driving simulation with Innovative Tools,
− to measure in real time the friction coefficient Automobile Technical Journal (ATZ), 95, 256–262.
and other environmental factors in order to Wallman, C.G. & Astrom, H. 2001. Friction Measurement
obtain FUSED and therefore modulate the speed Methods and the Correlation Between Road Friction
to avert the skidding risk. and Traffic Safety, Swedish National Road and Trans-
port Research Institute, VTI, Linkoping, Sweden.
In this case, it would be necessary to implement Zagati, E., Zennaro, R. & Pasqualetto, P. 1983. L’assetto
an algorithm able to apply the FDM on board and dell’autoveicolo. Ed. Levrotto & Bella.

316

TIS2017_Book.indb 316 2/8/2017 8:04:18 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The relationships between familiarity and road accidents:

Some case studies

P. Intini, P. Colonna, N. Berloco & V. Ranieri

DICATECh—Department of Civil, Environmental, Building Engineering and Chemistry,
Technical University of Bari, Bari, Italy

E. Ryeng
Department of Civil and Transport Engineering, Norwegian University of Science and Technology,
Trondheim, Norway

ABSTRACT: Familiarity with the route can lead to distraction, inattention and more dangerous
behaviors. Conversely, unfamiliar drivers could be unaware of possible dangers hidden in the road
environment. In order to inquire in detail the relationships between accidents and familiarity, a database
composed of 633 fatal and injury accidents (over the period: 2005–2014) related to 84 sections of two
important two-way two-lane rural Norwegian highways (E6, E39) was investigated. Familiarity of drivers
with the place of the accident was defined by considering a distance measure from the residence. Two sites
characterized by high percentages of namely familiar and unfamiliar accidents (selected basing on distance
of involved drivers from residence) were analyzed to a micro-scale level in order to find possible recurring
patterns and related factors. Familiar drivers were found to be over-involved in hitting vehicles in rear-end
accidents, while only some indications without clear patterns were found for the unfamiliar accidents.

1 INTRODUCTION different cross sections), which are part of the

familiar route. The concept of route familiarity is
1.1 The concept of familiarity with a road a sort of “selective experience” independent from
the driving experience in the strictest sense of the
A very common situation for a driver is to be
word (that is e.g. measurable in terms of years of
familiar with the road on which he/she is traveling.
driving license and annual mileage). The driving
This happens because many travels from a given
experience is related to the act of driving itself and
origin to a given destination are often repeated
so, it can influence the capability related to a given
over time (during the same week, for the whole
task (see e.g. Fuller 2005) to be accomplished by
year and more). A typical example is the home to
the driver. On the other hand, familiarity is related
work travel, that is the commuting trip. It is rea-
to the act of driving on a specific route and so, it
sonable to assume that, having to reach the same
can influence the behavior of the driver in a similar
destination from the same origin, the path cho-
way like experience does. In fact, the expectations
sen is usually the same. In fact, the main aim of a
of the drivers, given other boundary conditions
driver is the maximization of the utility related to
being constant, are most likely to meet reality
a travel (see e.g. Noland 2013). Once all possible
because the driver already knows what to expect
routes by car from a given origin to a given des-
by the road and so he is “experienced” about it (but
tination have been explored, it is most likely that
at the same time he can be more or less experienced
the driver chooses, from that moment on, the most
with the act of driving itself).
useful solution for the same trip. Repeating the
Studies in literature have suggested that famili-
same travel on the same route several times makes
arity can influence driving behavior inducing
the drivers likely familiar with that route.
inattention and changes in perception (Yanko &
Being familiar with a given route means, by def-
Spalek 2013; Martens & Fox 2007). This can hap-
inition of the adjective “familiar” (Oxford Univ.
pen because drivers focus on other thoughts differ-
Press Dictionary 2016), having a good knowledge
ent from driving, since the driving task could have
of its characteristics for long or close association.
become less demanding. However, parallel to the
This implies that the driver has a good knowledge
increased proneness to distraction, a familiar driver
of the features of the roads (in general, including
could be more skilled at more dangerous behaviors
segments and intersections, rural and urban, with

317

TIS2017_Book.indb 317 2/8/2017 8:04:19 PM

 on the familiar route, like speeding and rule break- fact that often this information is required when
ing (see e.g. Rosenbloom et al. 2007), especially the some specific analysis is performed, since it is not
more aggressive drivers (Colonna et al. 2016). normally included into accident databases as a
On the other hand, it is clear that also unfamiliar variable. However, the inclusion of a “familiarity”
drivers have some behavioral weak points. In fact, if variable related to a given road accident would be
the road system is not able to self explains (Theeu- complex too. In fact, police officers should ask to
wes & Godthelp 1995), expectations of unfamiliar drivers involved in the accident if they were famil-
drivers could be different by the real road situation iar, barely familiar or not familiar with the place
inducing errors or unsuitable maneuvers. of the accident. The problem is that this is only a
However, familiarity and unfamiliarity can be perception by the driver which is not an objective
seen as two features of the same phenomenon if measure. Some drivers could state to feel confident
drivers’ perception of risk is taken into considera- with a given road only after some travels on it and
tion. Indeed, the subjective perception of risk can vice versa. On the other hand, it should be con-
be able to influence driving tasks, but this can hap- sidered that possible driving behavioral changes
pen at two different levels: one internal to the driver due to familiarity belong to the subjective sphere
and the other external to the driver (Colonna & of the driver. Therefore, the more reliable relation-
Berloco 2011). In the internal sphere of risk (the ship between accidents and familiarity should take
driver’s unconscious), the perceived risk is com- into account the subjective feeling of being famil-
pared with a personal constant budget of safety iar with a given road on which the accident hap-
that a driver is ready to take, and this difference pened. Since this information cannot be found in
(supposed to be constant as deduced from Wilde databases, it could be achieved through surveys.
1982) determines the behavior in the medium/long A survey asking about previous road accidents
term period. Instead, in the external sphere of as a driver, over a short period of time (in order to
risk (the driver’s consciousness, that is practically acquire realistic information) could be widespread.
the interface driver-road-environment), the driver Questions about the familiarity of drivers with the
could have to suddenly face a real risk given by the place of the accident (i.e. unfamiliar, barely famil-
road environment different from the perceived one: iar or familiar) should be included in this survey.
the more is the difference, the more demanding will Anyway, it is evident that also this strategy shows
be the task of avoiding the danger. All drivers (in some weak points. First of all, a huge number of
this case both the familiar and the unfamiliar driv- surveys should be widespread for this aim, consid-
ers) have a conscious and an unconscious sphere, ering the response rate and the detail of the infor-
and so they are exposed to both the internal and mation required (to be a driver, to have experienced
the external risks. However, for familiar drivers, it some recent road accidents as a driver, to be able to
is possible that the acquired knowledge of the road report detailed characteristics about the accident).
environments can alter the equilibrium between Furthermore, self-reported information could not
the perceived risk and the target level of risk in the correspond to reality in some cases.
medium/long period, resulting in possible pitfalls. Therefore, in order to find relationships between
So, in this case, a higher danger could come from road accidents and drivers’ familiarity, some indi-
the internal sphere (unconscious). Instead, for the rect measures could be used. They have the advan-
unfamiliar drivers, a higher hazard could come tage of being objective, but the disadvantage of
from the external sphere: they do not know the being not easy to achieve or not being directly
road and so, they could be completely unaware of related to subjective familiarity. For example, one
possible sudden dangers hidden in the road. could measure drivers’ familiarity with a road by
using the frequency of traveling on it on a given
time period (daily, weekly, monthly, yearly...). This
1.2 Accidents and familiarity
is a common measure of familiarity (see e.g. Liu &
Given that driving experience is normally consid- Ye 2011; Brown et al. 2005) since a driver who trav-
ered a variable to be accounted while performing els daily or close to daily on a given road is most
statistical analyses related to road accidents; one likely familiar with it. Anyway, again, this is not a
should expect that also drivers’ familiarity with the measure that can be found in a common accident
road on which the accident occurred is taken into database.
account. Actually, in some sources, the familiarity Instead, an accident database could include
of drivers is contemplated as a factor taken in con- information about the origin of the drivers (see e.g.
sideration (see e.g. Liu & Ye 2011; Baldock et al. Blatt & Furman 1998; Yannis et al. 2007). The zip
2005). code associated to the place of residence could be
Establishing if a driver was familiar with the an indirect measure to be used for the aim of asso-
road on which the accident happened is a com- ciating road accidents to familiarity. This measure
plex matter to address. It is even perplexed by the is used in this study since it can be representative of

318

TIS2017_Book.indb 318 2/8/2017 8:04:19 PM

 the familiarity of drivers with the road. In fact, it is 2.2 Measures
most likely that a driver whose residence is placed
As explained in the introductory section, the meas-
very far from the place of the accident, is not famil-
ure used to define familiarity of the drivers with
iar with it. On the other hand, it is most likely the
the place where the accident occurred is their dis-
opposite concept: a driver who is traveling on a road
tance from the place of residence. In particular,
close to the place of residence is most likely famil-
this distance was computed by considering the
iar with that road. This common sense rule could
road itinerary associated to the shortest travel time
be violated by some exceptions: e.g. if the driver’s
between the center of the city/town of residence
permanent residence is different from the real resi-
and the exact place where the accident occurred.
dence placed very far from the permanent one or, if
Once the measure of distance from the residence
the driver is familiar with some roads very far from
has been defined, there is still the problem of defin-
his residence for different reasons (e.g.: for traveling
ing a threshold distance after which a driver can be
always to the same vacation place or for working
considered as unfamiliar with the road. Finding
reasons, as it happens to the professional drivers).
an exact distance boundary for the aim explained
However, since the aim is to find recurrent patterns
above is a not realistic approach. Instead, defin-
at particular sites characterized by very high or
ing a short distance from residence below which
very low accidents to familiar drivers and not the
the driver can be considered most likely familiar
analysis of a single accident only, then errors due to
and, defining a very long distance above which the
the wrong classification of drivers based on the dis-
driver is supposed to be unfamiliar is a more real-
tance criterion could be tolerated, since the analysis
istic strategy. In this study, a distance of the place
considers several accidents and not single crashes.
of the accident from 0 to 20 km from the place of
residence was used to consider drivers as familiar
1.3 Aim of the study with the road where the accident occurred. This
distance was chosen according to a report in which
Hence, since relationships between road accidents
average car commuting travels were computed for
and drivers’ familiarity are not easy to describe, the
Norway: 15.8 km for car drivers and 21.7 km for
aim of this study is to make a contribution in this
car passengers (Hjorthol et al. 2014). A distance of
sense. In particular, the main objective is the iden-
the place of the accident greater than 200 km from
tification of recurrent accident patterns related
the place of residence was used to consider driv-
both to unfamiliar and familiar drivers. This will
ers involved in the accidents as unfamiliar with the
be addressed by inquiring accidents at particular
road where the accident occurred. This distance
sites selected through the methods described in
was set by weighting the definition of long trip
next section.
(100 km) from the same Norwegian report used
for the definition of commuting travels (Hjorthol
2 METHODS et al. 2014) and the distance above which the plane
is the preferred mean of transport (300 km accord-
2.1 Accident database ing to Hjorthol et al. 2014; 400 km according to
Thrane 2015). Based on these threshold distances
A part of an accident database provided by the
two areas of familiarity can be defined, as sum-
Norwegian Public Road Agency (NPRA) was used.
marized in Table 1.
It regards 84 two-way two-lane rural road sections
of two important Norwegian highways: the E6
(37 sections) and the E39 (47 sections). The Nor- 2.3 Sites inquired in detail
wegian itinerary of the E6 connects the southern
For the aims of this study, some road sites have to
boundary with Sweden to Kirkenes in the North
be selected for a detailed analysis of the possible
of Norway, for a length of 2628 km. Instead, the
Norwegian itinerary of the E39 connects Trond-
heim to Kristiansand, for a length of 1140 km. Table 1. Ranges of distance used to define familiarity
The database used for this study includes 10 of drivers involved in the accidents.
years of accident data, from 2005 to 2014. In par-
ticular, the accident database is composed by 633 Distance* (km)
fatal and injury traffic accidents in which at least Drivers
Range From To involved
one vehicle was involved along the considered
sections. It provides general information about 1 0 20 Familiar
accidents, including the exact localization of the 2 200 – Unfamiliar
accident along the road section, and specific infor-
mation about the vehicles and the involved units *Distance from the drivers’ place of residence to the
(drivers and passengers). place where the accident occurred.

319

TIS2017_Book.indb 319 2/8/2017 8:04:19 PM

 accident patterns for which drivers’ familiarity Therefore, the sites seem comparable in terms of
can play a not negligible role. Therefore, sites in traffic, speed limits and widths. Instead, a notice-
which large percentages of unfamiliar drivers able difference in terms of driveway density and
were involved in accidents and sites in which large number of curves per km can be observed. Those
percentages of familiar drivers where involved in differences were considered in the analysis of the
accidents were inquired. In order to individuate results of the accident reconstructions.
patterns, road sites where less than 10 accidents
occurred were not considered. The definition of
drivers’ familiarity was based on the distance val- 3 RESULTS AND DISCUSSION
ues shown in Table 1.
Road sites having experienced accidents in 3.1 “Familiar” site
which at least one unfamiliar driver was involved
The “familiar” site inquired experienced eleven
for a percentage greater or equal than 70% of the
accidents in ten years, having involved 23 vehi-
total number of accidents were analyzed. They are
cles. Among these eleven accidents, eight of them
here defined as the “unfamiliar” road sites. In the
were classified as “familiar”. This means that, in
same way, road sites having experienced accidents
more than 70% of those accidents (eight), all driv-
in which all drivers were familiar for a percentage
ers being resident from 0 km to 20 km from the
greater or equal than 70% of the total number of
place of the accident, were involved. Among the 23
accidents were analyzed too. They are defined as the
vehicles involved, missing zip codes associated to 6
“familiar” road sites. In the latter procedure of sites’
vehicles were found. A summary of the accidents
selection, the accidents in which at least one driver
together with the indications about the vehicles
was defined as unfamiliar were considered, rather
and the drivers involved is given in Table 3.
than requiring that all drivers were unfamiliar. This
choice is based on the consideration that drivers
distant more than 200 km from the residence can Table 3. Details about the “familiar” site.
be considered as an “anomaly” in the traffic flow
among the other drivers. Moreover, the percentage Vehicle/Drivers’ familiarity**
of 70% was chosen since it was the higher percent-
# Type* A B C D
age allowing to individuate at least a road site in
each of the two categories (unfamiliar and familiar 1 Rear-End/ Car Car
road sites). The minimum of 10 accidents for each Angle (ND) (NC)
site was set in order to obtain meaningful percent- 2 Rear-End/ Car Car
ages of familiar and unfamiliar accidents. Angle (ND) (F)
Characteristics of the two road sites selected by 3 Rear-End/ Motorcycle Unknown
using these criteria are summarized in Table 2. Angle (F) (ND)
Moreover, both sites have posted speed limits 4 Head-on Car Bicycle
included between 50 and 80 km/h. Road widths (ND) (F)
measured in correspondence of the accidents at the 5 Run-off Car
two sites according to the database, are included (F)
between 6 and 8.6 meters (maximum 8.3 meters for 6 Rear-End/ Car Car Car
Angle (F) (F) (F)
the site 2). The driveway density (number of drive-
7 Run-off Truck
ways or minor intersections per kilometer) is 2.1 (F)
for the site 1 (corresponding to 5 total driveways) 8 Rear-End/ Tractor Tractor Car
and 1.1 for the site 2 (corresponding to 6 total Angle (F) (ND) (NC)
driveways). The site 1 is characterized by 8 curve 9 Rear-End/ Car Car
sections along the segment (3.3 curves per km), Angle (F) (F)
while the site 2 is characterized by 24 curve sections 10 Rear-End/ Car Camper Car Camper
(4.4 per km). Three out of the total curves belong- Angle (F) (ND) (F) (F)
ing to the site 2 are near to 90-degree curves. 11 Run-off Car
(NC)

Table 2. Characteristics of road sites inquired. *Rear-End and Angle accidents were grouped together
since most of the crashes classified in the database as
AADT Length Vehi- “while turning from the same direction” were described
Site (veh/day) Road (m) Crashes cles as rear-end crashes.
**Classification of familiarity: F = familiar (≤20 km),
1 (Familiar) 6697 E39 2400 11 23 U = unfamiliar (≥200 km), NC = not classified
2 (Unfamiliar) 5165 E6 5500 10 19 (20–200 km), ND = missing data. The definition “camper”
includes campers, vans, cars with trailer, light trucks.

320

TIS2017_Book.indb 320 2/8/2017 8:04:19 PM

 of the vehicle A, considered familiar, fell asleep
while driving as found out from the report.
The accidents 6 and 7 happened on a straight
road section. The accident 6 occurred on wet road
and in good visibility conditions. The cars A and
B were stopped on the roadway, and the car B was
rear-ended by the car C coming from behind. All
the drivers were considered familiar. The accident
7 happened on icy road and in good visibility con-
ditions. The driver of the truck A tried to overtake
another vehicle but he lost control of its vehicle,
and he hit the road barrier. The driver of the vehi-
cle A was considered familiar.
The accidents 9 and 10 happened on a curve
section (showing a very high radius of curvature)
in correspondence of a driveway. The accident 9
occurred in good weather, road and visibility con-
ditions. The car B is stopped in order to turn to the
driveway, and it was rear-ended by the car A. Both
drivers were considered as familiar. The accident
10 happened in good weather, road and visibility
conditions. The vehicles A and B were stopped in
order to turn to the driveway. The car C was suc-
cessful in braking behind the vehicles A and B.
Instead, the vehicle D coming from behind, failed
in stopping and so, the preceding vehicles were
rear-ended. All the drivers were familiar, except for
the driver of the vehicle B, showing missing data.
Figure 1. Schemes of accidents occurred at the two The other three accidents (1, 8, 11) were two
road sites. rear-end/angle crashes and a run-off road crash.
Some general considerations can be made on
considering the results of the accident analyses.
The accidents analyzed, for the aim of finding Most of the accidents were rear-end crashes (7 out
familiar accident patterns, were the number 2, 3, 4, of 11, and five of them were considered “famil-
5, 6, 7, 9, 10 of the list shown in Table 3. Schemes iar” accidents). The other recurrent type is the
representing collisions are reported in Figure 1. run-off road (3 out of 11, and two of them were
The accident 2 happened on a straight road “familiar”). Only one head-on crash was recorded
before a curve section in good weather, road and (familiar).
visibility conditions. The car A was rear-ended by At this road site, the percentage of missing
the car B. The driver of the vehicle B was consid- zip codes (resulting in missing classification of
ered familiar while any measure was available for the familiarity of drivers) is close to 25%, higher
the vehicle A. than the average (less than 15%) of all the sample
The accidents 3, 4 and 5 occurred in the same of road sites inquired. This occurrence makes the
curve road section in correspondence of a driveway. interpretation of results more complex. Anyway,
The accident 3 happened in good weather, road among the 8 familiar accidents analyzed, familiar-
and visibility conditions. The motorcycle A was ity was not defined for only 4 drivers. Except for
rear-ended by the unknown vehicle B. The driver the accident 3, these drivers were rear-ended or
of the vehicle A was considered familiar while any otherwise hit. Therefore, even with missing data, it
information was available for the vehicle B. The is possible to affirm that the driver who hit the cars
accident 4 happened on wet road and in bad vis- in the rear-end accidents was considered familiar
ibility conditions (but with presence of road light- in 4 out of 5 of the familiar crashes (4 out of 7
ing). The cyclist B, coming from the driveway, hit total rear-end crashes). This is a recurring pattern
the car A traveling on the main road. The cyclist B at this site.
was considered familiar, while any information was However, the argument of a recurring pattern in
available for the vehicle A. The accident 5 occurred rear-end accidents related to familiar drivers have
on wet road and in good visibility conditions. The to be analyzed in more detail. In fact, the road site
car A, ran off the road when entering in the curve is characterized by a great number of driveways
from the previous straight road section. The driver connecting nearby peripheral residential areas.

321

TIS2017_Book.indb 321 2/8/2017 8:04:19 PM

 In some cases, they do not seem to be adequately Table 4. Details about the “unfamiliar” site.
signaled along the road or they are not provided
by turning lanes to simplify the maneuver. It Vehicle/Drivers’ familiarity*
seems logical that, at this road site, a high share
# Type* A B C D
of rear-end accidents is present, since they can be
related to the presence of driveways (and indeed, 1 Head-on Camper Camper
5 rear-end crashes out of 7 were near a driveway). (U) (U)
On the other hand, the presence of a nearby resi- 2 Other Car
dential area can be likely related to a high share of (U)
familiar drivers in the traffic flow at this road site. 3 Run-off Camper
This occurrence can be responsible for the over- (U)
representation of familiar drivers in the accidents 4 Head-on Motorcycle Truck
happened there. (NC) (NC)
Nevertheless, in four out of five familiar rear- 5 Head-on Camper Bicycle
end accidents (among the seven total rear-end (ND) (U)
accidents), the crash started with a familiar driver 6 Run-off Car
(NC)
hitting from behind other cars. Finding who were
7 Head-on Car Camper
at fault was obviously not coherent with the aim
(NC) (U)
of this paper and anyway, it is an information not 8 Head-on Car Camper Car
derivable from the information present in the data- (NC) (NC) (U)
base. However, this information is useful since it 9 Rear-End/ Car Truck Camper
can be likely related to drivers’ speed. A rear-end Angle (NC) (NC) (U)
crash is often caused by a car traveling at high 10 Rear-End/ Car Car
speed that fails in braking promptly. The braking Angle (NC) (NC)
process can be perplexed by wet road or bad
weather conditions (e.g. this was observed at this * See footnotes to Table 3.
site too). Familiarity of drivers with the road site
could have been responsible for higher speeds (as
introduced in the first section), potentially related The accident 1 happened at a curve section on
to the rear-end crashes. Moreover, as stated in icy road, during rain and in good visibility condi-
the introduction, familiarity of drivers can be tions. The vehicle A invaded the opposite lane in
related to distraction (potentially linked to rear- curve resulting in a head-on crash with the vehicle
end crashes too) and more dangerous behaviors. B coming from the other direction. Both drivers
The accident 5 (familiar driver fell asleep) and the were considered unfamiliar with the road.
accident 7 (familiar driver who tried a likely dan- The accident 2 occurred in a curve section in
gerous overtaking maneuver on icy road) can be good weather and road conditions but without
related to this tendency. However, given the small road lighting at night. The car A hit a moose
number of items, these should be considered only crossing the roadway. The driver was classified as
as indicators. unfamiliar.
The accident 3 happened at a curve section in
good weather, road and visibility conditions. The
3.2 “Unfamiliar” site
vehicle A ran off the road on the right and it tipped
The “unfamiliar” site inquired experienced ten over. The driver was classified as unfamiliar.
accidents in ten years, having involved 19 vehicles. The accident 5 happened at a curve section in
Among these ten accidents, seven of them were good weather, road and visibility conditions. The
classified as “unfamiliar”. This means that, in the bicycle B, invaded the opposite lane in curve result-
70% of those accidents, at least one driver being ing in a head-on crash with the vehicle A coming
resident from more than 200 km from the place of from the other direction. The cyclist was classified
the accident, was involved. Among the 19 vehicles as unfamiliar, while no zip code was available for
involved, only one vehicle was not associated to the other driver.
a zip code. A summary of the accidents together The accident 7 occurred at a curve section on icy
with the indications about the vehicles and the road in good weather and visibility conditions. The
drivers involved is given in Table 4. car A lost control and invaded the opposite lane
The accidents analyzed, for the aim of individu- resulting in a head-on crash with the vehicle B com-
ating recurring unfamiliar accident patterns, were ing from the other direction. The driver of the vehi-
the number 1, 2, 3, 5, 7, 8, 9 of the list presented in cle B was considered unfamiliar, while the driver of
Table 4. Schemes representing those collisions are the vehicle A came from a distance between 20 and
reported in Figure 1. 200 km (neither familiar nor unfamiliar).

322

TIS2017_Book.indb 322 2/8/2017 8:04:20 PM

 The accident 8 happened on a straight road road site is a segment of a rural rolling/mountain-
section in good weather, road and visibility condi- ous road with no important city placed at a close
tions. The car A invaded the opposite lane result- distance. Given these conditions, it is expected that
ing in a head-on crash with the vehicle B coming a significant part of the traffic flow can be com-
from the other direction. Then, the car C coming posed of unfamiliar drivers. This can explain the
from the same direction, was damaged by junks of over-representation of unfamiliar drivers in the
vehicle B. The driver of the vehicle C was classified studied accidents. Some of them were involved in
as unfamiliar, while the others were neither famil- the possible causation of accidents likely due to
iar nor unfamiliar. driving errors. This is coherent with the concepts
The accident 9 occurred in a curve section (show- stated in the introduction: if the road environ-
ing a high radius of curvature) in correspondence ment is particularly demanding, presenting sud-
of a driveway, in good weather, slippery road and den surprises, the unfamiliar drivers could be not
without road lighting. The truck B was moving on able to correctly address the road demand, since
the left in order to turn to the driveway, while it their expectations do not meet reality. However, in
was rear-ended by the vehicle C. Then, both the this case, no recurrent patterns were found for the
vehicles B and C were rear-ended by the vehicle A. unfamiliar accidents.
The driver of the vehicle C was classified as unfa-
miliar, while the other drivers were neither familiar
nor unfamiliar. 4 CONCLUSIONS
The other three accidents (4, 6, 10) were a
head-on, a run-off and a rear-end/angle crash. A micro-analysis of the accidents occurred at spe-
Some general considerations can be made on cific sites selected for their noticeable high share
considering the results of the accident analyses. of familiar and unfamiliar drivers (based on the
Most of the accidents were head-on crashes (5 out distance between the accident place and the resi-
of 10, and in four of them there was at least one dence) was conducted.
unfamiliar driver involved). Two run-off crashes In detail, two road sites were further analyzed:
(one of them with the presence of an unfamiliar the site 1, showing a high concentration of crashes
driver) and two rear-end crashes (one of them for familiar drivers and the site 2, showing a high
with the presence of an unfamiliar driver) were concentration of crashes for unfamiliar drivers.
recorded. Only one crash with an animal was found Some interesting results were highlighted
(unfamiliar). through the reconstruction of accidents. In fact,
It is important to note that even if they are some recurrent accident patterns were individu-
considered as two divided categories, the head-on ated, especially for site 1. There, most of accidents
and the run-off crashes have some common traits were rear-end crashes, likely due to the high den-
to be evaluated for the aims of this study. In fact, sity of driveways along the road segment. Familiar
the head-on accidents are often caused because a drivers were found to be in most cases the driv-
vehicle invades the opposite lane. The reasons for ers hitting vehicles from behind, resulting in the
these invasions could be various, but except for the rear-end crash. This is coherent with the possible
case of the driver who fall asleep or is distracted, tendency to speeding and distraction for familiar
they can be related to some maneuvering errors. drivers. Instead, at the site 2, most of the accidents
The same errors can be related to the run-off road were head-on collisions, likely associated to the
crashes together with the speed selection. Consid- high demanding road geometry of the road seg-
ering together the head-on and the run-off crashes, ment (as well as run-off-road crashes). Unfamiliar
in five out of seven total accidents at least one drivers were found to be in some cases the drivers
unfamiliar driver was involved. In three out of five who lost control or who anyway invaded the oppo-
of those accidents, the unfamiliar driver invaded site lane, resulting in crashes. This is coherent with
the opposite lane or ran off the road at curves. In the possible difference for them, between expecta-
only one of the inquired cases, the invasion of the tions and the surprising reality (Colonna & Berloco
opposite lane happened in a road section essentially 2011). However, in this case, no clear patterns were
straight (even if included between two curves). individuated for the unfamiliar accidents.
This case could be better related to distraction or Anyway, these results come from the analysis
fatigue, but in that case (accident 8) the driver was of a very limited number of case studies. Further-
not classified as unfamiliar. Therefore, from this more, these sites were selected by setting empirical
analysis, only some indications about unfamiliar threshold percentages of drivers’ familiarity. There-
drivers can be taken, without the identification of fore, these interpretations should be mainly taken as
clear accident patterns. a cause for further reflections about the topic, that
Anyway, also in this case, it is important to should be analyzed more in depth. For example, at
analyze the context of the road site inquired. The the road sites inquired, the most recurrent types of

323

TIS2017_Book.indb 323 2/8/2017 8:04:20 PM

 accident are multi-vehicle crashes. The influence of Blatt, J. & Furman, S.M. 1998. Residence location of
familiarity on the accident type, considering more drivers involved in fatal crashes. Accident Analysis &
in detail the interactions between familiar and Prevention, 30(6), 705–711.
unfamiliar drivers in multi-vehicle crashes should Brown, J., Fitzharris, M., Baldock, M., Albanese,
B., Meredith, L., Whyte, T. & Oomens, M. 2015.
be analyzed by focusing on other specific sites. Motorcycle In-depth Crash Study (No. AP-R489-15).
Moreover, it is evident that several other factors Austroads.
(related to road, driver, environment) are related Colonna, P. & Berloco, N. 2011. External and internal
to the accidents besides the familiarity of drivers. risk of the user in road safety and the necessity for
In this paper, they were considered only locally, a control process. Xxiv PIARC World Congress.
as related to each accident. It is evident as well Mexico City.
that, the highlighted accident patterns could be Colonna, P., Intini, P., Berloco, N. & Ranieri, V. 2016.
explained by the differences themselves in the road The influence of memory on driving behavior: how
sites features, complementary to, or independent route familiarity is related to speed choice. Safety
science, 82, 456–468.
from drivers’ familiarity. Therefore, a more com- Fuller, R. 2005. Towards a general theory of driver behav-
prehensive statistical analysis would be needed for iour. Accident Analysis & Prevention, 37(3), 461–472.
further studies, in order to assess the importance Hjorthol, R., Engebretsen, Ø. & Uteng, T.P. 2014. Den nas-
of the familiarity as predictor variable of accident jonale reisevaneundersøkelsen 2013/14: nøkkelrapport.
occurring and type, but considering other vari- Transportøkonomisk institutt. Norway.
ables, such as for example drivers’ age and gender, Liu, C. & Ye, T.J. 2011. Run-off-road crashes: an on-scene
curvature rates, other geometric features, weather perspective (No. HS-811 500). NHTSA’s National
etc. (Russo et al. 2013). Center for Statistics and Analysis. Washington, DC.
However, some of the highlighted facts sug- Martens, M.H. & Fox, M.R. 2007. Do familiarity and
expectations change perception? Drivers’ glances and
gest that drivers’ familiarity/unfamiliarity could response to changes. Transportation Research Part F:
be better considered for road design and safety- Traffic Psychology and Behaviour, 10(6), 476–492.
based maintenance. Demanding and surpris- Noland, R.B. 2013. From theory to practice in road safety
ing road features should be avoided on segments policy: Understanding risk versus mobility. Research
where unfamiliar drivers are likely to travel (i.e. in transportation economics, 43(1), 71–84.
touristic and rural mountainous highways). On Rosenbloom, T., Perlman, A. & Shahar, A. 2007. Women
the other hand, possible shift of familiar drivers to drivers’ behavior in well-known versus less familiar
more risky behaviors could be taken into account locations. Journal of safety research, 38(3), 283–288.
at roads near residential areas (e.g. by modifying Russo, F., Biancardo, S.A., Busiello, M., De Luca M. &
Dell’Acqua, G. 2013. A statistical look at gender and
minor intersections into roundabouts for reducing age differences as related to the injury crash type on
speeds at intersections). However, further studies low-volume roads. WIT Transactions on the Built
are needed to better address the possible practical Environment, 134, 213–224.
applications. Theeuwes, J. & Godthelp, H. 1995. Self-explaining
roads. Safety science, 19(2), 217–225.
Thrane, C. 2015. Examining tourists’ long-distance
ACKNOWLEDGMENTS transportation mode choices using a Multinomial
Logit regression model. Tourism Management
The authors acknowledge Damiano De Gennaro Perspectives, 15, 115–121.
Wilde, G.J. 1982. The theory of risk homeostasis:
who, during his thesis, helped with the reconstruc- implications for safety and health. Risk analysis, 2(4),
tion of the accidents used throughout the paper. 209–225.
Yanko, M.R. & Spalek, T.M. 2013. Route familiarity
breeds inattention: A driving simulator study. Accident
REFERENCES Analysis & Prevention, 57, 80–86.
Yannis, G., Golias, J. & Papadimitriou, E. 2007. Accident
Baldock, M.R.J., Long, A.D., Lindsay, V.L.A. & risk of foreign drivers in various road environments.
McLean, J. 2005. Rear end crashes. Adelaide Univer- Journal of safety research, 38(4), 471–480.
sity. Australia.

324

TIS2017_Book.indb 324 2/8/2017 8:04:20 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Geotechnical asset management for Italian transport agencies:

Implementation principles and concepts

P. Mazzanti
NHAZCA S.r.l. and Department of Earth Sciences, University of Rome “Sapienza”, Rome, Italy

P.D. Thompson
Consultant, Bellevue, WA, USA

D.L. Beckstrand
Landslide Technology, Inc., Portland, OR, USA

D.A. Stanley
D.A. Stanley Consulting, Bellingham, WA, USA

ABSTRACT: Internationally, Transportation Asset Management (TAM) has been accepted practice for
many years. National, regional or state statutes and policies encourage transportation agencies to adopt
asset management principles, including performance- and risk-based management. Transport agencies
also develop their own short- and long-term plans, objectives and goals to guide agency programs.
Italy recently reorganized the Ministry of Infrastructure and Transport (MIT) and developed ten key
concepts to guide the Ministry. While all these concepts have application to TAM, several stand out,
including utility, simplification, involvement, maintenance and safety. Development of TAM for trans-
port systems can be a guiding force in improving development and management of transport systems.
Consideration of geotechnical assets is important in TAM development for Italy to meet national, regional
and agency goals. Italy has a highly-developed multi-modal system of surface transportation and thousands
of km of tunnels also due to the complex topography. The system ranges from roadways to airports to rail
transport and ferry and ship transport infrastructure, all of which depend on support from geotechnical
assets, whether embankments below pavement, earth retaining walls, rock slopes, etc. These type of assets
are often overlooked, but the value of geotechnical assets is likely to be a substantial percentage of the total
value of the system. The high geological and geomorphological activity of the Italian peninsula is a relevant
factor of stress for the national transportation network and, especially, for geotechnical assets. Furthermore,
many of these assets are in declining condition, because, like every transport agency in developed countries,
there are not sufficient funds to do everything that is needed. Maintaining and improving the condition of
geotechnical assets can be a key step for the achievement of the MIT goals.
Geotechnical Asset Management (GAM) can contribute to improvement to the condition, utility, safety
and life cycle cost of a transportation network. GAM will assist transport agencies to understand the cur-
rent and future condition of geotechnical assets, and through integrated life cycle cost and risk analysis, the
alternative actions available to rehabilitate, preserve and eventually replace assets. Geotechnical asset man-
agement offers guidance in supporting decision-making to preserve and improve the transport system.

1 INTRODUCTION
manner, through the management of assets for
present and future customers.” (Gordon, et al,
1.1 Role of agency goals and objectives
2011). TAM ties essential agency processes to a
and transportation asset management
performance management system that includes
This paper explores potential processes for creating target levels of service and performance metrics
a Transportation Asset Management (TAM) pro- that when monitored over time, give a picture
gram for the management of geotechnical assets of how the agency is progressing toward its tar-
associated with the Italian road system. gets. TAM benefits include optimized costs for
The purpose of a TAM program is: “to meet a the system, improved communication within the
required level of service, in the most cost-effective agency and with outside stakeholders, reduced

325

TIS2017_Book.indb 325 2/8/2017 8:04:22 PM

 risk, and improved asset condition and system are supported by or protected by geotechnical
performance. assets, whether soil embankments below high-
The Italian transport agency goals and objectives way surfaces, rock slopes and rockfall mitigation
have already been created, as discussed below, and devices above roadways and railroads, rock below
provide the basis for development of an asset man- bridge foundations or retaining walls that support
agement process. What remains is the development roadways and tunnels, that could not otherwise be
of a step-by-step roadmap to guide transport agen- constructed.
cies through the process of self-assessment, setting The relevance of geotechnical assets is even
agency targets and desired outcomes, incorporating greater in a geologically young country like Italy
TAM into transport system organizations, establish- that is characterized by intense geological and geo-
ing performance standards, developing TAM Plans, morphological processes. Landslides and rockfalls
developing TAM tools and processes (inventory pro- are the most common events that interact with the
cedures, monitoring and data collection, information transportation assets. However, erosion, floods,
systems, communicating with public and stakehold- earthquakes and volcanic activities interact quite
ers, etc.) through established agency protocols. often with the transportation network.
Ideally, Geotechnical Asset Management Moreover, settlement and subsidence are fre-
(GAM) programs should develop alongside or quent in low-lying areas, due to the compressibility
trailing TAM programs. However, TAM systems of recently deposited soils. Hence, instability prob-
may include all transportation assets and therefore lems can frequently affect embankments, roads,
may be much slower to move forward. But, even if runways/airstrips and tunnels. Due to the complex
the TAM system is not entirely developed, a GAM topography, the Italian transportation network
program can proceed apart from, but in alignment includes thousands of kilometers of tunnels, a sub-
and compatible with the TAM system. This can be stantial percentage of all the tunnels in Europe.
done by first making connections between GAM However, despite the complexity and extent of
and the broader agency goals and objectives and geotechnical assets, they are often constructed with
then following a roadmap that mirrors the TAM a “build and forget” mentality. But these assets
roadmap. The steps in a TAM/GAM roadmap can must be maintained throughout their lifecycle,
vary considerably, but Figure 1 shows an example. just as concrete or steel bridges and asphalt and
concrete pavements must be maintained. GAM
programs offer the ability and tools to make cost-
1.2 Why manage geotechnical assets?
effective decisions about geotechnical assets that
Geotechnical assets have a rarely noticed impor- contribute to the good performance, reduced risk
tance in transport networks: all transport assets and improved cost-effectiveness of the transport
system as a whole.

1.3 What are geotechnical assets?

One of the first steps in management is to decide
what geotechnical assets Italian transport agen-
cies are responsible for and what can be managed.
While taxonomic classifications have recently been
developed (Anderson, 2016), geotechnical assets
can generally be subdivided into tangible and
intangible assets. Tangible assets are those that
have been built to physically support the transpor-
tation system. Intangible assets include results of
geological or geotechnical data gathering. Table 1
contains examples of both types. Poor manage-
ment of either type results in decreased resilience,
inefficient stewardship, and higher life cycle costs.
In the USA, only a few programs are operat-
ing or under development for geotechnical assets.
Specific assets include: rock and soil slopes, retain-
ing walls and sand and gravel sources/quarries. In
addition, one GAM-based program focuses on
managing geotechnical hazards, such as rockfall,
Figure 1. Steps for Development of Geotechnical Asset landslides, debris flows, volcanic hazards, and
Management for Italian Transport Agencies. floods. As stated above, Italy has an abundance of

326

TIS2017_Book.indb 326 2/8/2017 8:04:22 PM

 Table 1. Example geotechnical assets. • adoption of safety measures, and
• conduct of research and experimentation in the
Physical assets Data assets field of roads and traffic and implementation of
Embankments Geologic Maps
the results.
Soil and Rock Slopes Geotechnical Reports (ANAS website http://www.stradeanas.it/,
Tunnels Road Closing Events/Costs accessed September 2, 2016).
Retaining Walls Boring Logs In addition, the MIT has engaged in strategic
Coastal Revetments Performance monitoring long-range planning with development of its July
data 2016 “Connect Italy—Strategies for transport
Foundations Laboratory Data infrastructure and logistics.” Together the mission
Geotechnical Instruments Geotechnical Event Data statement, Ten Key Concepts and the “Connect
Rockfall Control Measures Maintenance History Italy” strategic planning effort provide ample high
Geotechnical Instruments As-built Plans level guidance to create and execute a roadmap
Material Sources Photographs
for a performance-based asset management sys-
Levees Flood Records
tem for geotechnical assets. The “Connect Italy”
infrastructure strategy lays out the architecture
of the transport system, but importantly defines
each of these asset types and hazards. As part of the “objectives of transport policy in Italy.” This
a gap analysis, the agencies can determine which critical section lists: sustainable and safe mobil-
assets to manage. ity, quality of life and competitiveness in urban
and metropolitan areas, and support for indus-
1.4 High level transport agency goals trial policies of the supply chain (MIT website:
and objectives http://www.mit.gov.it/comunicazione/news/piste-
ciclabili/connetterelitalia-pubblicata-sul-sito-la-
The Italian transport system is the responsibil- nuova-strategia-del: accessed September 2, 2016).
ity of the Ministero delle Infrastrutture e dei Specific strategic objectives outlined in the docu-
Trasporti (Ministry of Transport and Infrastruc- ment include:
ture or MIT). The MIT recently reorganized and
adopted a mission statement: “Italy does not need • Useful infrastructure, streamlined and shared;
large or small projects, but useful ones.” The term • Modal and intermodal integration;
“Useful” applies to growth of the country, to mod- • Enhancement of existing infrastructure assets;
ern and efficient logistics, for rapid mobility, and a • Sustainable urban development.
safe, clean, convenient and widespread system of MIT and ANAS work together to operate the
transport. Useful also relates to recovery and liv- Italian network of roads and highways within the
ability of urban centers and for the defense and framework of mission statements, adopted key con-
maintenance of the country. (MIT website: http:// cepts, and strategic planning objectives. Incorpora-
www.mit.gov.it/ communicazione/news/nuovo-mit, tion of transportation asset management principles
accessed September 2, 2016) The MIT mission also into both agency’s business and technical processes
includes modernizing the existing road and motor- in accord with these objectives, concepts and strat-
way network. egies will help align the organizations, strengthen
As part of their efforts to modernize the agency, each agency and provide long term improvements
the MIT has developed “Ten Key Concepts” to to the road and motorway networks.
guide its operations and investments: Utility, Sim-
plicity, Participation, Sustainability, Care, Integra-
tion, Safety, Iron, South, and Europe.
2 SELF-ASSESSMENT AND GAP ANALYSIS
Since 1946, MIT has operated a government-
owned company, Azienda Nazionale Autonoma
The incremental process of advancement in asset
delle Strade (the National Autonomous Roads Cor-
management necessarily occurs in phases spread
poration or ANAS) whose mission is the “manage-
over many years. During that time, much can
ment of ordinary and extraordinary maintenance
change in an agency’s institutional and economic
of streets and highways,” including:
environment, in the needs of stakeholders, in the
• progressive improvement of the network of agency’s delivery capability, and in technology.
roads, A useful general approach to implement asset
• construction of new roads and highways, management is self-assessment. The American
• service of information to users, Association of State Highway and Transportation
• protection of the heritage of roads and Officials has published an example (Cambridge
highways, et al, 2002) which addresses business processes,

327

TIS2017_Book.indb 327 2/8/2017 8:04:22 PM

 data, tools, documentation, and training in the fol- the agency’s valuable infrastructure over a typical
lowing key areas: time frame of ten years. The GAM Plan serves sev-
eral important functions:
• Policy guidance
• Planning and programming • Documents the ways in which infrastructure
• Program delivery performance contributes to agency success.
• Information and analysis • Specifies an agreed means of measuring the con-
tribution of asset performance toward agency
The strategic self-assessment is couched in very
success.
general terms in order to be applicable to all trans-
• Provides a commonly-understood set of criteria
portation agencies and all types of assets. Volume
for evaluating and selecting investments.
2 of the AASHTO Asset Management Guide pro-
• Establishes a basis for incentives for superior per-
vides more detail (Gordon et al, 2011).
formance, internal and external to the agency.
In the specific domain of geotechnical assets,
• Engages stakeholders in a discussion of the
it is useful to think about how information about
tradeoff between funding and performance.
these assets enters into each of the business proc-
• Provides an objective method of allocating
esses addressed in the strategic assessment. For
resources among preservation, risk mitigation,
example:
and service improvement goals.
• Policies and criteria to guide inventory and con- • Specifies research requirements and analyti-
dition surveys, and the identification and priori- cal tools for relating current decisions to future
tization of needs. outcomes.
• Processes to ensure that geotechnical needs are • Provides the functional and quality assurance
considered in road and bridge projects, and in requirements for asset inventories and data col-
the maintenance budget. lection processes.
• Reliable tracking of actual costs of preservation, • Provides the justification for capital and operat-
reconstruction, risk mitigation, and incident ing expenditures on preservation, maintenance,
recovery on geotechnical assets? and risk mitigation.
• Efficient and effective delivery of planned inspec- • Documents the agency’s commitment and
tions, maintenance activities, and projects. accomplishments in the continuous improve-
• Maintenance of a complete inventory with effec- ment of its management processes.
tive presentation tools, such as maps and charts.
All of the basic components of asset manage-
• Capability to forecast future conditions and
ment and TAM Plans have been codified in vari-
needs.
ous standards documents in recent years. In the
• Incorporation of geotechnical needs within
United Kingdom, the authoritative source is Pub-
agency tools for planning of pavement and
licly Available Specification 55, volumes 1 and 2
bridge work.
(BSI 2008). In the United States, a basic framework
• Consideration of geotechnical risk and long-
is described in a financial management context in
term cost, including future maintenance cost,
Government Accounting Standards Board State-
within the design of new facilities.
ment 34 (GASB 1999), and in a strategic planning
Clearly many of these concerns apply to all context in Volume 1 of the AASHTO Guide for
types of assets, so in many agencies it may be more Asset Management (Cambridge et al 2002) as well
cost-effective to consider all infrastructure assets in as in ISO 55000. A more detailed adaptation of
the same self-assessment. This would be a prelude the same principles is New Zealand’s International
to developing a coordinated multi-year program Infrastructure Management Manual (IIMM,
that can fill all the gaps. NAMS 2006). AASHTO has built on this concept
in great practical detail with the AASHTO Trans-
portation Asset Management Guide, Volume 2: A
3 GEOTECHNICAL ASSET Focus on Implementation (Gordon et al 2011).
MANAGEMENT PLAN A key aspect of successful asset management
implementation, brought out in the IIMM and
A Geotechnical Asset Management Plan (GAM the AASHTO Guide, is the notion of continuous
Plan) is a written document, or a set of written improvement. A variety of human and automated
documents and databases, which describes the ingredients need to be improved in tandem. The
processes and outputs of agency GAM activities. amount of progress that can be made in asset man-
(Thompson 2016) The GAM Plan complements agement tools is limited by the human and organi-
other agency planning documents such as strate- zational readiness to use the technology, and vice
gic plans, service plans, and investment plans, but versa. In a more tangible sense, the technology to
focuses on the preservation and performance of produce quality asset management information

328

TIS2017_Book.indb 328 2/8/2017 8:04:22 PM

 depends on management willingness to accept over time. Decisions regarding life cycle manage-
asset management information in decision-making ment, capital programming, and maintenance pol-
(and to see the value and pay the cost of producing icy are evaluated according to predicted changes
this information); and management acceptance, in in measurable objectives. Research-based analysis
turn, depends on the quality of information that methods are used to forecast future performance
can be produced. A small improvement in the and to estimate the outcomes of agency actions.
decision making process must be matched by an Transportation agencies spend money construct-
incremental improvement in technology, which ing or acquiring geotechnical assets, and spend
then spurs the next small improvement in decision additional money over the lifespans of these assets
making. to keep them functioning as intended. In general,
These same principles are widely used in the pri- the intended function is to maintain (or refrain
vate sector, often taking the form of performance from disrupting) desired levels of safety, mobil-
management frameworks such as the Balanced ity, environmental sustainability, and economic
Scorecard and Six Sigma (Proctor et al 2010, Gor- efficiency of transportation service. Through its
don et al 2011). maintenance forces and contractors the agency
In 2015, the US Federal Highway Administra- implements treatments that maintain or enhance
tion published a draft of its proposed TAM Plan the characteristics of its geotechnical assets to
requirements (FHWA 2015). The rule specifies that minimize the frequency of disruptions. These
the plan shall cover at least a 10-year period, shall desirable characteristics make up a property called
be made easily accessible to the public, and shall resilience. In general, resilience is defined as:
establish a set of investment strategies that improve
or preserve condition and performance in support “… the capability of a system to maintain its func-
of national highway performance goals. The plan tions and structure in the face of internal and
is to be tied in to existing programming processes external change and to degrade gracefully when it
and must contain the following minimum content: must.” (Allenby & Fink 2005).
1. TAM objectives, aligned with agency mission; “Internal and external change” can be interpreted
2. Performance measures and targets; in the context of geotechnical assets as changes
3. Summary of asset inventory and condition; caused within the asset itself (i.e. normal deteriora-
4. Performance gap identification; tion) and change caused by external forces (natural
5. Life cycle cost analysis; extreme events). “Maintain its functions and struc-
6. Risk management analysis; ture” can be interpreted as the avoidance of trans-
7. Financial plan; portation service disruptions. “Service disruptions,”
8. Investment strategies. in turn, can be interpreted as unintended changes in
The risk-based TAM Plan is required to identify the safety, mobility, sustainability, or economic per-
the hazards affecting the movement of people and formance of the roadway. Based on this reasoning,
goods, assess the likelihood and consequences of a slope may be considered to have high resilience (or
adverse events, and evaluate and prioritize mitiga- low vulnerability) to the extent that it is sufficiently
tion actions. The life cycle cost analysis is a quanti- able to refrain from causing service disruptions due
tative network-level analysis that considers current to normal deterioration or adverse events. Examples
and desired condition levels, asset deterioration, of measurable factors affecting resilience are:
effects of adverse events, and treatment options
Rock slopes:
over the whole life of assets.
The proposed US rules require that TAM Plans • Ditch (or catchment) effectiveness: assesses how
address only pavements and bridges, but encour- often falling rocks reach the roadway, combining
age agencies to expand the scope to include any the effects of all design, mitigation, and geom-
or all infrastructure assets within the right-of-way etry concerns.
corridor. Alaska is the first state to develop a plan • Rockfall activity: assesses how active the slope is
for geotechnical assets which includes all of the in producing falling rocks, combining the effects
components required in the federal rules (Thomp- of all condition characteristics, geological char-
son 2016). acter, climate, and hydrology.
Embankments and soil slopes:
4 PERFORMANCE MEASURES • Roadway displacement or slide deposit: assesses
AND TARGETS the direct effect on the roadway surface of earth
movement, combining the effects of all rel-
Transportation Asset Management is a discipline evant condition characteristics and mitigation
that relies on performance measures that are tracked features.

329

TIS2017_Book.indb 329 2/8/2017 8:04:23 PM

 • Length of affected roadway and roadway imped- developed an example set of analysis tools as Excel
ance: assesses the geometry of the site. spreadsheets (Thompson 2016).
• Movement history: assesses the combined effect
of geological character, climate, hydrology, and
permafrost characteristics. 5 INVENTORY, ASSESSMENT,
AND MONITORING
Retaining walls
Transportation asset management relies on data
• Vertical and horizontal wall alignment: assesses
and analysis to improve decision making. Thus
one aspect of condition (deformation), com-
the first step, and most costly investment, in GAM
bining the effects of drainage, geometry, and
implementation is the creation of an accurate
foundation.
inventory of these assets, and establishment of a
• Impacts to the roadway: assesses the effect of
recurring process of field inspections to keep the
physical condition of the roadway as it relates to
inventory up-to-date and to monitor conditions
wall condition.
over time.
• Critical component health: assess all aspects of
wall condition other than deformation, such as
corrosion. 5.1 Inventory and assessment approaches
These characteristics are typically summarized Site-based geotechnical performance assessment
into an index (often on a scale where 100 is best and focuses on a limited set of visual observations of
0 is worst) or a set of states (such as good-fair-poor) the key physical properties that have most effect on
which can be recorded and tracked over time. deterioration rates or life cycle cost.
Acceptable asset performance is characterized
by determining whether a given asset satisfies a set 5.1.1 Focus on screening-level efforts
of level of service criteria, usually the same crite- Often, agencies may not have a good grasp on the
ria used in separating Poor performance from Fair. number, location, or type of geotechnical assets for
These criteria may vary by class of road, with high- which they are responsible. Inventories can start
volume roads or essential lifeline routes subject to with plan sets and online street viewer systems car-
more rigorous criteria. Additional criteria may be ried out by entry-level geotechnical personnel. Fol-
used to identify opportunities for preventive main- low-up, site specific condition assessments require
tenance or risk mitigation. the knowledge and judgement of more experienced
At the network level, performance is typically geotechnical personnel. Condition assessments
assessed in the form of average of asset perform- evaluate the asset according to a set of parameters
ance index values, or as the percent of the inventory where numerical scores are assigned to eight to
in Good or Poor condition. The agency can track twelve observable conditions then combined into
changes in this performance, and will typically report a single metric enumerating its condition. Subsur-
changes publicly in order to establish mutual account- face explorations should not be part of the effort.
ability for making progress toward shared goals. Field time for each condition assessment should
Future performance goals are expressed as per- take no longer than 20 to 30 minutes per asset. If
formance targets over a timeframe of 5–10 years. longer times are required, such as for long linear
Aspirational targets can be established to indicate assets such as levees, they can be broken up into
the desired direction of future performance trends. smaller segments or evaluated to a more detailed
For example, many agencies have an aspirational reporting level.
safety objective of “zero fatalities.” Aspirational
LOS and performance measures have been con- 5.1.2 Limit to observable criteria
sidered for unstable slopes in Alaska (Stanley & There are many unknowns in the geotechnical
Pierson, 2011). field. At the screening level, the asset’s condition is
However, there are more mature asset manage- effectively communicated by its performance. For
ment processes with fiscally-constrained perform- example, the performance of a rock slope is tied
ance targets, developed from a research-based directly to its rockfall activity and the effectiveness
analysis of deterioration rates, intervention costs, of the roadside catchment ditch. Retaining wall
action effectiveness, and the likelihood and con- performance is observable through its deforma-
sequences of service disruption, to indicate the tion, corrosion, and performance of the adjacent
expected performance level under a specific fiscal pavement. Initial observations of surface features
scenario. Such tools are common in the manage- of retaining walls is sufficient for screening level
ment of pavements and bridges, but are still in their assessments. More in-depth investigation of bur-
infancy for geotechnical assets. Alaska’s Depart- ied reinforcement elements is possible through
ment of Transportation and Public Facilities has exhumation, use of retrievable reinforcing sample

330

TIS2017_Book.indb 330 2/8/2017 8:04:23 PM

 coupons, and remote electrochemical testing, but Table 2. Example Rock Slope Evaluation Approach.
these methods are better suited for long-term
monitoring activities for high-risk structures. Score
Range Descriptor Rock Slope Condition
5.1.3 Utilize existing systems and data 80 to 100 Good Little to no rockfall activity
Many systems have been developed to assess the with good ditch effectiveness
condition of geotechnical assets. Systems such as 40 to 80 Fair Moderate activity with
the Rockfall Hazard Rating System (RHRS) (Pier- moderate to limited ditch
son & VanVickle, 1993) have been implemented effectiveness
along select Italian motorways (Budetta, 2004; 0 to 40 Poor Frequent or constant
Budetta & Nappi, 2013). Other systems have been activity with limited to no
generated for the evaluation of landslide slopes ditch effectiveness
and retaining walls by US transportation depart-
ments (Federal Highway Administration, states of
Washington, Ohio, Oregon, and Alaska). These
systems contain concepts utilized by the geotech- 1969) can be used to reduce the impacts of unpre-
nical community for decades and can be adapted dicted events and to systematically update the
for condition assessment and correlated to risk assessment.
and mitigation costs while still containing concepts In the last few decades, a strong technological
familiar to most geotechnical professionals (Beck- evolution has provided to the community several
strand and Mines 2016). tools to improve observational abilities and per-
form effective geotechnical monitoring that can
5.1.4 Data collection and storage support GAM programs. Ranging from remote
Data collection and storage platforms have satellite systems to contact apparatus today it
advanced in recent years. The advent of user- is possible to perform a multi-scale monitoring
friendly mobile devices combined with GIS data approach in space and time, thus supporting asset
collection applications, such as ESRI’s Collector management and decision making.
mobile application, has expanded opportunities Geotechnical monitoring tools can be classified
for efficient data collection and storage. These in two main groups: contact systems and remote
systems facilitate field data collection and subse- systems. Contact systems need the physical con-
quent storage directly into an online GIS platform tact between the instruments and the ground
that integrates readily into planning activities and while remote methods do not require such con-
analysis. tact as they are mainly based on sensors receiving
and, often, also emitting electromagnetic waves.
5.1.5 Simplify results A suitable combination of these tools may answer
Clear and concise communication of assessment most of the needs of management of geotechni-
results is critical. Results from existing systems, cal assets.
such as the RHRS, are clear to geotechnical per- Specifically, remote sensing systems like satel-
sonnel well versed in its use, but may be confusing lite InSAR, aerial LiDAR and Photogrammetry
to non-geotechnical personnel. GAM system out- may provide considerable information at regional/
put should be understandable to non-geotechnical national scale (thus controlling the entire trans-
personnel and use scores and descriptions that portation networks) while contact systems may
clearly communicate results. Descriptors such as allow for the continuous control of high-risk sites
‘Good’, ‘Fair’, and ‘Poor’ describe the meaning of where an early detection of potential threats due to
the scores as shown below. Scores that drop over adverse events can reduce the risk.
time from 100 toward 0 through Good and Fair to In some cases, continuous monitoring can be
Poor communicate deterioration. considered also the most effective and ultimate
solution for the management of high risk areas.
Over the last years ANAS and RFI (Rete Ferro-
5.2 Monitoring
viaria Italiana) – the two main players of the Ital-
Inventory and condition assessment provide a ian National Transportation networks that control
baseline for life cycle management. Uncertainty is more than 50,000 kms of roads—performed sev-
a constant companion in geotechnical engineering. eral successful applications that demonstrated the
Each asset can be characterized by unpredicted efficacy of the monitoring approach.
events, and unforeseen findings. Risk management Hence, Italy could be a leading country in the
principles may be applied in GAM programs to development of a “Smart GAM”, i.e. a manage-
reduce the effects of uncertainty. As noted above, ment plan that makes extensive use of geotechnical
observational methods (Terzaghi, 1937; Peck, monitoring tools.

331

TIS2017_Book.indb 331 2/8/2017 8:04:23 PM

 6 COMMUNICATE PERFORMANCE 7 CONCLUSION

An important concept in all transportation asset As Italy moves forward to modernize and improve
management systems is the ability to conduct effec- its transport network of roads, motorways and
tive communication about the program internally rails, it has an opportunity to follow the interna-
within the agency, between related transport agencies tional pattern of implementing transportation
(MIT, ANAS and railway agencies, e.g.), and with asset management as a business process and as a
the public and stakeholders including government set of tools to aid in management of geotechnical
administration and legislative bodies. In Italy, there assets. Italy has substantial numbers of geotechni-
must be robust communication systems that will cal assets at risk, such as rock and soil slopes and
assist developers of geotechnical asset management retaining walls throughout the country located in
programs in communicating about the process and to complex and active geological terrains as recently
collect and disseminate data and information about demonstrated by the August 24, 2016 earthquake
geotechnical assets and their condition, performance, in Central Italy that triggered more than 100
lifecycle costs, risk/safety issues, among others. landslides intersecting with transportation roads
An Italian example of communication is the (CERI website: http://www.ceri.uniroma1.it/index.
ANAS Road Management Tool, which provides php/2016/08/earthquake-central-apennines-italy/,
surveillance through remote cameras, traffic accessed September 10, 2016).
management, data collection and real time event Italy’s current effort to effect change through
management capabilities through a central control reorganizing the Ministry of Infrastructure and
room. Managers in the control room may dispatch Transport, utilizing “Ten Key Concepts” and the
resources as needed to roadway events. operations of the National Autonomous Roads
Alaska DOT has developed an ArcGIS On- Corporation, as well as comprehensive planning
Line-based mapping product that shows a snap- such as the “Connect Italy” all combine to pro-
shot of the state’s program for managing unstable vide the high-level goals and objectives necessary
slopes derived from databases of road network and to formulate and implement an asset management
geological data (Figure 2). plan for geotechnical assets. Conducting TAM and
A more traditional TAM presentation provides GAM programs to manage assets will further the
the public and other stakeholders a quick view agency goals and objectives.
of status of high-level agency-wide performance Aging infrastructure elements such as unstable
attributes through the use of DOT website per- slopes and retaining walls are often overlooked
formance dashboards or performance report pub- until they fail, sometimes catastrophically, but
lications, such as attributes of congestion, safety, they can be managed through TAM/GAM proc-
road surface condition, financial performance, esses, starting with gap analyses to assess the
management performance, customer service, and weaknesses in the transport agency’s management
timely completion of projects. of geotechnical assets. Agencies can then proceed
through inventory and assessment, beginning with
screening level basic efforts utilizing simplified
condition ratings.
Furthermore, the international leadership of
Italy in the field of monitoring technologies and
remote sensing can be a driving force for the devel-
opment of a “Smart Geotechnical Asset Manage-
ment” (SGAM) plan that includes the continuous
collection and storage of monitoring data for the
systematic update of the GAM in support of deci-
sion making. ANAS is one of the most advanced
and innovative agencies at global level in the appli-
cation of cutting edge monitoring and sensing
technologies for the control of geotechnical assets.
Technologies like satellite and terrestrial InSAR
and LiDAR, for example, have been commonly
used for the last several years for management
plans of single assets under critical conditions
(Mazzanti et al, 2015; Brunetti & Mazzanti, 2015).
With this technology, transport agencies can then
Figure 2. GIS heatmap for Alaska’s rock slope define standard procedures for the continuous geo-
performance. technical monitoring of high risk areas and assets.

332

TIS2017_Book.indb 332 2/8/2017 8:04:23 PM

 As all of these pieces of the puzzle come together, FHWA. 2015. Notice of Proposed Rule-Making on Trans-
transport agencies will develop the capability to portation Asset Management Plans. Federal Register
provide analysis-based decision-making support to 80:34, Page 9232 and 9236. Washington: US Federal
agency authority in planning and selecting projects Highway Administration.
GASB. 1999. Basic Financial Statements — and Manage-
and project elements best suited for improving per- ment’s Discussion and Analysis — for State and Local
formance and reducing life-cycle cost and risk for Governments. Government Accounting Standards
the transport network. Board Statement 34.
Gordon, M., Jason Smith, G., Thompson, P.D., Park, H.,
Harrison, F., & Elston, B. 2011. AASHTO Transpor-
REFERENCES tation Asset Management Guide, Volume 2: A Focus on
Implementation. Washington: American Association
Allenby, B. & Fink, J.. 2005. Toward inherently secure and of State Highway and Transportation Officials.
resilient societies. Science 309:5737, pp. 1034–1036. Mazzanti P., Bozzano F., Brunetti A., Esposito C., Mar-
Anderson, S.A., Schaefer, V.R., & Nichols, S.C. 2016. Tax- tino S., Prestininzi A., Rocca A., Scarascia Mugnozza
onomy for geotechnical assets, elements, and features. G., 2015. Terrestrial SAR Interferometry Monitoring
Compendium of Papers, TRB Annual Meeting, Wash- of Natural Slopes and Man-Made Structures. G. Loll-
ington, D.C., USA: Transportation Research Board. ino et al. (eds.), Engineering Geology for Society and
Beckstrand, D. & Mines, A. 2016. Jump starting a Territory, Volume 5, Springer International Publish-
geotechnical asset management program with exist- ing, Switzerland, 189–194.
ing data. Submitted for publication to Transporta- NAMS Steering Group. 2006. International Infrastruc-
tion Research Record: Journal of the Transportation ture Management Manual (IIMM). Auckland:
Research Board, in preparation. National Asset Management Steering Committee,
Brunetti A.& Mazzanti P., 2015. Monitoring an unsta- New Zealand.
ble road embankment for public safety purposes by Peck, R.B. 1969. Advantages and limitations of the
terrestrial SAR interferometry. Proceedings of the 9th observational method in applied soil mechanics. Géo-
International Symposyum on Field Measurements in technique, 19(2), 171–187.
Geomechanics (Sydney, 9–11 September 2015) pp. Pierson, L & VanVickle, R. 1993. Rockfall Hazard Rat-
769–780. ing System Participants Manual. Washington: US
BSI. 2008. Asset Management Part 1: Specification for Federal Highway Administration Report FHWA
the optimized management of physical assets. London: SA-93–057.
British Standards Institute, Publicly Available Specifi- Proctor, G., Park, H., Varma, S., & Harrison, F. 2010.
cation 55–1 (PAS 55–1). Beyond the Short-Term: Transportation Asset Man-
BSI. 2008. Asset Management Part 2: Guidelines for the agement for Long-Term Sustainability, Accountability,
application of PAS 55-1. London: British Standards and Performance. Washington: US Federal Highway
Institute, Publicly Available Specification 55–2 (PAS Administration Report FHWA-IF-10–009.
55–2). Stanley, D.A. & Pierson, L.A. 2011. Geotechnical asset
Budetta, P. & Nappi, M. 2013 Comparison between qual- management performance measures for an unstable
itative rockfall risk rating systems for a road affected slope management plan. In Proceedings: 62nd High-
by high traffic intensity Natural Hazards and Earth way Geology Symposium: pp. 133–152. Lexington,
System Sciences 13: 1643–1653. KY, USA.
Budetta, P. 2004 Assessment of rockfall risk along roads Terzaghi K., 1937. Settlement of structures in Europe
Natural Hazards and Earth System Sciences 4: 71–81. and methods of observations. American Soci-
Cambridge Systematics, Inc., Parsons Brinckerhoff ety of Civil Engineers. Proceedings, Vol. 63, pp.
Quade and Douglas Inc., Roy Jorgensen Associates 1358–1374.
Inc., & Paul D. Thompson. 2002. Transportation Asset Thompson, P.D. 2016. Geotechnical Asset Management
Management Guide, Washington: American Associa- Plan: Technical Report. Juneau: Alaska Department
tion of State Highway and Transportation Officials. of Transportation and Public Facilities.

333

TIS2017_Book.indb 333 2/8/2017 8:04:23 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Evaluation of workability of warm mix asphalt through CDI parameter

and air voids

D.M. Mocelin, L.A.T. Brito, M.G. Johnston, V.S. Alves, G.B. Colpo & J.A.P. Ceratti
Federal University of Rio Grande do Sul, Brazil

ABSTRACT: Warm Mix Asphalt (WMA) has grown in production over the past years due to its vari-
ous advantages, such as reduced greenhouse gas emissions, lower energy consumption and reduced com-
paction temperature among others. In Brazil, it was not until few years ago that warm mixes set off. With
typical temperature reductions of around 30°C below the Hot Mix Asphalt (HMA) production, the use of
surfactant additives has proved effective also in enhancing bitumen coating and lubrication of the asphalt
binder in the mixture—arguably, yielding and improvement in WMA workability at lower temperatures.
This research evaluated the workability of warm mixes accounting the Construction Densification Index
(CDI), obtained in the compaction curve of a Superpave Gyratory Compactor (SGC) and also by means
of air voids control achieved in Marshall specimens. Torque measurements during the mixing process
improved the discussion along with mineral particle distribution obtained from digital image process-
ing. The experiment tested a standard dense HMA and equivalent WMA varying both mixes on their
production and compaction temperatures at −15°C, −30°C and −45°C of the reference HMA (143°C),
totalling seven analysed mixtures. The results showed no workability variation on the mixes production
but an improvement on the compactability; the latter being highly sensitive to surfactant presence and to
temperature decrease. In general, the mixtures at reduced temperatures only reach the proper compaction
using WMA surfactant additive. The tests carried out allow the conclusions that the addition of WMA
surfactant additive enables a decrease in the compaction temperature of about, at least, 30°C without
significant loss in workability and improving particle homogeneity in the mixes.

1 INTRODUCTION be supplied at longer distances; lower compaction

temperature thresholds allow longer transports
Environment awareness over the past decades moti- periods. Prowell et al. (2012) also highlight better
vated all engineering fields for more sustainable compaction rates using WMA and the possibility
products and rational use of natural sources. Pave- of carrying out jobs in cold seasons or places of
ment engineering has followed such concern and cold weather.
the industry has made new products available to Different techniques can be used to allow tem-
reduce impact in road maintenance. Amongst such perature reduction in asphalt mixes; the majority
developments, temperature reduction in asphalt uses some type of additive blended in binder. Sur-
mix production has gained space worldwide. factant additives proved to be effective and easy to
Several benefits can be associated to tempera- use in warm mix production (Hurley & Prowell,
ture reduction of Hot Mix Asphalt (HMA), such 2006; Kvasnak, 2010; Bennert et al., 2010).
as the reduction in greenhouse gas emissions in- Surfactant additives have their origin from sur-
plant and during paving jobs, energy savings due face-active agents. These agents act in the interface
to lower fuel consumption, healthier working between aggregates and the asphalt binder provok-
environment. ing better lubricity between both and, hence, ena-
In addition, benefits to asphalt mix perform- bling an easier binder coating over aggregates at
ance due to reduced heat exposure during produc- lower than conventional temperatures.
tion may be considered. Standard temperatures for Temperature of HMA preparation and compac-
mixing HMA average 150°C; at this temperature, tion is a function of the binder viscosity. Because
light volatile particles present in the asphalt binder surfactant additives do not change significantly
can evaporate and oxidize the material, resulting in this property (Johnston et al., 2015), temperature
material’s drop in performance. recommendations for WMA typically follows addi-
Another immediate benefit of the so-called tive supplier, whose expertise recommends produc-
Warm Mix is the mix plant coverage; material can tion and compaction temperature ranges.

335

TIS2017_Book.indb 335 2/8/2017 8:04:25 PM

 This paper attempts to measure, yet in a qualita- 2.2 Asphalt Binder
tive fashion, how such additives improve asphalt
The asphalt binder used was a type 50/70 (pen gra-
mixes workability and what is the temperature
dation). Such binder is widely used in most of Bra-
reduction limit threshold that can be achieved with
zilian highways. The benefit of using a standard
this WMA technique. A better understanding of
asphalt binder—with no other modification—was
the densification process makes a parallel to field
also important for bias, allowing the effect of the
compaction, while monitoring the torque during
surfactant additive to become evident.
the mix process enables to determine the required
Results in Table 1 summarize both used binders:
effort in both HMA and WMA production.
a standard 50/70 used to produce the tested HMA
Finally, an investigation of the homogeneity
and the same binder with surfactant additive for
after compaction assesses the effect on tempera-
the tested WMA. The addition of surfactant did
ture reduction of the produced mix.
not change significantly the characteristics.

2.3 Surfactant additive

2 MATERIALS
The surfactant used in this research is available in
The study was carried out with asphalt mixtures with liquid form, allowing the additive to be mixed with
characteristics and materials commonly used in Bra- the binder on terminals and, therefore, distributed
zilian highways; ergo, national standards were applied. using normal binder supplying process. Due to their
Materials description are summarized below. improved convenience, provided no changes in the
production line of asphalt mix is required, this type
of additive are of growing use (NCHRP, 2011).
2.1 Aggregates
The surfactant additive was added to a ratio of
The aggregates used are of basaltic origin, from a 0.4% by weight of asphalt binder, as per supplier
quarry located in southern Brazil (Santo Antônio recommendation. Typically, such use allows a 30°C
da Patrulha/RS). The particle size distribution reduction in mix and compaction temperatures.
used is Figure 1. The used surfactant was Ingevity Evotherm® M1.

2.4 Mixtures parameters

Two different mixes were designed using the Mar-
shall method—HMA and WMA. Both used the
same aggregate gradation and binder type, only var-
ying incorporation of WMA surfactant additive.
The first step in the study was to determine
design characteristics; to do it, HMA was mixed at
153°C and compacted at 143°C, while WMA used
123°C & 113°C, (−30°C to HMA), respectively, fol-
lowing supplier’s recommendation.
The final characteristics for both mixes are
Figure 1. Particle size distribution. shown in Table 2 and evidence no significant
change in the parameters.
Table 1. Results of binder characterization. Important to notice that for WMA produc-
tion, binder is pre-heated at the same temperature
Binder + as HMA, regardless of the additive presence. The
Brazilian Binder Surfactant temperature reduction shall only be carried out for
Properties Standard 50/70 (WMA Additive) mixing and compaction.
Softening Point 46 min 49 48
Penetration 50–70 67 57 Table 2. Final characteristics of the asphalt mixtures.
Specific Gravity — 0.957 1.017
Brookfield 274 min 408 480 Standard WMA w/
Viscosity DNIT – HMA no surfactant
135°C (cP) Properties 031/2006 additive additive
Brookfield 112 min 237 242
Viscosity Binder Content (%) — 5.7 5.7
150°C (cP) Air Voids (%) 3–5 4.3 4.2
Brookfield 57–285 90 91 Voids w/asphalt (%) 72–82 76.9 78.2
Viscosity Gmm (kN/m3) — 25.06 25.04
177°C (cP) Gmb (kN/m3) — 23.98 24.04

336

TIS2017_Book.indb 336 2/8/2017 8:04:25 PM

 3 METHODOLOGY air voids from Marshall specimens, using standard
compaction effort of 75 blows per specimen side.
To investigate mixtures compactability, eight tests Masad et al. (1999), Tashman et al. (2001)
were performed with the superpave gyratory com- and Vasconcelos et al. (2005) used imaging tech-
pactor (SGC): four tests using standard unmodified niques to investigate homogeneity in asphalt mixes.
binder and the other four prepared with surfactant Aggregate scatter in both horizontal and vertical
additive. Each group of tests varied mixing tem- directions have direct effect in mechanical proper-
perature only, at: 143°C, 128°C, 113°C and 100°C. ties. Hence, homogenous distribution of materi-
This correlates to −0°C, −15°C, −30°C & −43°C of als during the compaction process is vital for an
the standard HMA compaction temperature. Tem- adequate behaviour of pavement layers.
perature reductions were always performed during In order to determine if the lower temperatures
mixing and compaction procedures. were affecting the homogeneity of the resulting
Going below the maximum temperature reduc- mix, 112 mm high by 150 mm in diameter speci-
tion recommended by the additive supplier and mens, produced in the SGC to designed air voids
extending the tests for the mixes with unmodified were cut into 10 mm slices, producing from each
binder enable to compare the densification process side of the slice one image (Figure 3).
via the Construction Densification Index (CDI) Images were then scanned and using FIGI/
calculated during compaction with the SGC. IMAGEJ software boundaries of aggregates were
The mixtures were prepared using a laboratory determined automatically by threshold limits on an
mixer with torque measurement sensor. Torque 8-bit grey scale. The minimum size aggregate cap-
was monitored on the blades moving mechanism tured were those with and area of 22.15 mm2, equiv-
throughout the mixing time. These measurements alent to material passing in a sieve #4 (4,75 mm).
allowed the evaluation of the workability of all Figure 4 illustrates, in simplicity, the sequence in
mixes during the production process, for the same the digital processing steps for particle detection.
temperature scheme. After aggregates were sized and orientation tagged,
CDI is defined as the area from the 8th gyration to their occurrence was grouped into three concentric
92% of Gmm in the densification curve (Figure 2).
Theoretically, CDI represents the work applied by
the roller to compact the mixture to the required
density during construction. The number of eight
gyrations is selected to simulate the effort applied by
a typical paver during the process of laying down
the mixture, while the 92% of Gmm is the density at
the completion of construction and the pavement is
open to traffic (Mahmoud and Bahia, 2004).
Because to construct CDI curves required high
number of gyrations, the resulting samples have
low air voids. Hence, controlling their final air
voids is of little interest. Nonetheless, as the air
voids represent one of the key parameters for a
Figure 3. Specimen slicing for image processing of par-
mix design, the same temperature scheme was also ticle sizing and orientation.
applied for a typical Marshall compaction, ena-
bling a cross analysis between CDI from SGC and

Figure 2. Hatched area corresponding to CDI Figure 4. Image processing steps to obtain particle size
determination. and distribution (HMA in detail).

337

TIS2017_Book.indb 337 2/8/2017 8:04:25 PM

 radial zones of 25 mm enabling to verify materials
homogeneity.

4 RESULTS

4.1 Construction Densification Index—CDI

The CDI is considered a volumetric parameter,
related to compactability of asphalt mixtures.
Higher CDI means a longer distance from the 8th
gyration to 92% of Gmm in the densification curve,
i.e., a difficulty to achieve a proper compaction
(see Figure 2). Mixes with lower CDI values are Figure 5. CDI to temperature variation.
deemed to have better constructability and, thus,
of preference, while excessively low values of CDI
could be an indication of a tender mixture and
Bahia, 2004; Nascimento, 2008; Soares, 2014),
should be avoided (Mahmoud & Bahia, 2004).
indicating good compactability of the mixes.
Table 3 show the results of CDI obtained for all
The CDI value is also dependent on granular
four tests carried out. The analysis of the results dem-
matrix skeleton. As particle size distribution of
onstrated that with greater temperatures, a reduction
both mixes used in this study have a high percent-
on CDI values suggests swifter compactability.
age of finer particles (55% smaller than 4.5 mm),
Despite the obvious expected results, for the
this may have contributed to low CDI values.
viscosity-temperature characteristics of bitumen,
Mahmoud & Bahia (2004) investigate a set of
it is evident that the addition of surfactant to the
mixes with different materials and binder contents,
binder provoked an increase in compactability,
finding a significant variation in densification
here inferred from CDI reduction values at each
behaviour of the mixtures tested. Results of that
tested temperature. Figure 5 presents the relation
study confirms that CDI results are highly depend-
of CDI to temperature summarized.
ent on the mineral skeleton and the type of binder.
The WMA mixtures at tested temperature of
Nonetheless, as all tests carried out in this study
113°C returned intermediate (27.9) CDI values to
have the same particle size distribution and binder
HMA mixtures at test temperatures of 143°C (25.4)
type, the comparative effect in unharmed.
and 128°C (28.8); this evidences clear improvement
in the densification process with the incorporation
of the additive. A variation of –30°C resulted in 4.2 Air Voids versus Compaction Temperature
near CDI values of the standard mixing HMA
Using standard Marshall compaction procedure to
temperature.
test resulting Air Voids (AV) at the same tempera-
These values of CDI obtained are similar to low
ture test sets used for CDI tests, higher AV values
compared to values found in literature (Mahmoud &
at lower temperatures would be expected.
Mix design properties (Table 2) of both HMA
Table 3. CDI results obtained for the mixtures. and WMA shows a target value of 4.3% AV for
HMA (compacted at 143°C) and 4.2% AV for
Construction Densification Index (CDI)
WMA (compacted at 113°C). Brazilian standard
Temperature No With DNIT 031/2006 for the asphalt wearing courses
(°C) additive Average additive Average states desired air voids percentage between 3% to
5%, averaging a designed target of 4% AV. The
100 43.40 43.40 38.41 38.41 experiment test results are recorded in Table 4.
* * Figure 6 presents the results obtained for all of the
113 51.72 39.34 24.22 27.92 analysed asphalt mixtures.
26.95 31.61 Sensibility of the mixtures to temperature and
128 28.95 28.81 15.68 18.69 the effect of the additive are evident in the results.
28.66 21.69 For the same temperatures, the mixtures with addi-
143 25.40 25.40 – – tive featured lower values of air voids than HMA.
* – A higher level of compactability can be credited
Average is a result of the two tested samples on each tem- to an increase in lubrication provided by the
perature. (*) means error has occurred during measure- surfactant.
ments and data was discarded. (–) no test at 143°C was The differences in the air voids were 1.5%, 1.6%
carried out for WMA (w/additive). and 1.1% for the temperatures of 100ºC, 113°C

338

TIS2017_Book.indb 338 2/8/2017 8:04:26 PM

 Table 4. Air voids results obtained. monitored mixer were performed in order to estab-
lish certain guidance. Rotation speed of the blades
Air Voids (%) were set to 20 rpm; the mixer’s drum was set to
rotate in reverse to the sense of the blades and the
Temperature No With
(°C) additive Average additive Average amount of material to be tested was limited to 20 kg
with a mixing time of 500 s. The mixer drum relies
100 6.34 6.2 4.85 4.7 on a controlled temperature environmental cham-
6.31 4.65 ber, thereby allowing mixing temperature control.
6.07 4.71 Torque measurements for the asphalt mixes
113 6.02 5.8 4.25 4.2 presented some level of variation throughout the
5.58 4.19 tests. Such variations are possibly associated to
5.92 4.27 particle size discontinuity, inherent to the material.
128 5.12 5.0 3.91 3.9 Final torque values, for each mix to all same six
5.08 3.98 temperature/mix combination afore mentioned,
4.87 3.82 were averaged discarding initial ten seconds and
143 4.11 4.3 – – when torque peaks exceeded mean value plus one
4.45 – standard deviation. This procedure avoided great
4.32 – variation observed when mixer blades met a coarse
lump against the drum spiking torque results.
(–) no test at 143°C was carried out for WMA (w/ Figure 7 illustrates the torque readings of a
additive).
mixture, depicting parameter variation and out-
lier values disregarded, as an example. The average
torque results obtained for the mixes are presented
in Figure 8.
Results show no significant change between the
different temperatures and mixes with and without
additive. Values ranged between a maximum of
22.5 N.m and a minimum of 20.5 N.m.
To test the torque apparatus responsiveness and
also guide results assessment, two very homogene-

Figure 6. Air voids versus temperature results.

and 128°C, respectively. A steeper trend in AV

increase with temperature variation—comparing
angular coefficient of regressions (0.0262 for
WMA & 0.0446 for HMA)—also corroborates to
the CDI values, indicating a considerably smoother
compactability.
For the HMA, a 15°C reduction shifted AV to
Figure 7. Results of torque measurements.
outside the upper limit threshold. Conversely, for all
three tested temperatures, WMA remained inside AV
limits. Despite adequate AV values at 100°C for the
WMA, caution is required to use such reduction in
WMA production; CDI values in section 4.2 shows
a closing pattern of WMA and HMA towards
100°C; below this point it is uncertain the surfactant
additives improvement in compactability.

4.3 Torque measurements—mixtures workability

Measurement of workability in asphalt mixes is
non-standardized; neither procedures nor stand-
ards for guiding the torque measurement method Figure 8. Torque measurement comparison for work-
was found. Hence, preliminary tests using a torque- ability assessment.

339

TIS2017_Book.indb 339 2/8/2017 8:04:26 PM

 ous and of lower kneading effort required—water
and sand—were used to test the equipment using
the same mixing parameters. Water, alone in the
drum, resulted in a 1.67 N.m torque reading and
sand, 9.78 N.m.
The tests revealed a good sensibility of the
equipment and allowed a benchmark of the
results, provided literature was not found about
such measurements and, yet if found, could greatly
vary according to equipment, amount of material,
rotation speed, among others.
Little torque variation indicates that both HMA
and WMA had similar workability under tested con-
ditions. Torque for WMA at 113° (21.5 N.m) nears the
Figure 9. HMA Specimen 2 – Faces 7 to 12.
torque required to mix HMA at 143°C (21.2 N.m);
this could be expected as the surfactant additive
ought to compensate the temperature loss and main-
tain workability of the mix. Notwithstanding, the
little torque alteration within the HMA temperature
variation suggests another secondary reasoning.
For all tests, the asphalt binder was heated up
to 153°C; it was done so because for WMA pro-
duction there is no alteration in temperature to
the added binder. Because the mixing time is fairly
short (500 s) it is arguable that the temperature
decay from 153°C to one of the mixing tempera-
tures (100°C to 143°C) – at which both environ-
mental mixing chamber and added aggregates
shall be—may happen at a steady low pace. The
mixture workability can then be attributed to the Figure 10. WMA Specimen 3 – Faces 7 to 12.
asphalt binder at its added temperature, resulting
in slight variation in the measured torque.

4.4 Homogeneity from imaging processing

The temperature reduction during mixing and
compaction processes demonstrated not to have
great impact in workability and granted WMA
with increased compactability. With the purpose
of analysing the effect of such variation in the
specimens homogeneity, image processing of sliced
samples enabled the evaluation.
This analysis was performed at standard tem- Figure 11. Image processing steps to obtain particle
peratures: mixing at 153°C and compaction at size and distribution in quadrants.
143°C for the HMA, while for WMA 123°C &
113°C respectively. After the image processing as
evidenced in the histogram and in the faces images
described it section 3, results indicated a consid-
in Figure 9.
erable more homogenous specimen for the warm
mixes.
Figure 9 illustrates six slices of a HMA bottom 5 CONCLUSIONS
half specimen, while Figure 10 illustrates an equiv-
alent imagery for a WMA specimen. Figure 11 This study aims to help understanding workability
brings a comparison of the resulting frequency and compactability improvements of surfactant
of aggregates greater than 4.75 mm in each of the agents used for WMA. The later was observed
three radial zones. with a better densification obtained via CDI meas-
Results clearly demonstrates a more homog- urements during compaction, whereas workability
enous distribution of aggregates across radial sec- was found unchanged for all tested conditions.
tion for the WMA. A concentration of coarser Analysis of the SGC tests demonstrated that
aggregates towards the edge of the specimen is greater temperatures lead to a CDI reduction sug-

340

TIS2017_Book.indb 340 2/8/2017 8:04:27 PM

 gesting swifter compactability. Although obvious, as laboratory partners for theirs technical contribu-
with rising temperatures asphalt binders have lower tion and financial support to the research group.
viscosity, the addition of surfactant to the binder
led to an increase in compactability, inferred from
CDI reduction values at each tested temperature. REFERENCES
WMA compacted at 113°C returned intermediate
Agência nacional do petróleo, gás natural e biocom-
(27.9) CDI values compared to HMA at 143°C (25.4) bustíveis. Cimentos Asfálticos de Petróleo. Resolução
and 128°C (28.8); this is a clear improvement in the n° 19 de 11 de julho de 2005 da ANP – Regulamento
densification process with the incorporation of the Técnico ANP n° 19/2005.
additive. A variation of −30°C resulted in near CDI bennert, T.; Reinke, G.; Mogawer, W.; Mooney, K., Assess-
values of the standard mixing HMA temperature. ment of Workability and compatibility of Warm-Mix
Sensibility of the mixtures to temperature and Asphalt. Transportation Research Record: Journal of
the effect of the additive were evident in Mar- the Transportation Research Board, No 2180, Wash-
shall compaction air voids control. For the same ington, DC., USA, pp. 36–47, 2010.
temperatures, mixtures with surfactant (WMA) Departamento nacional de infraestrutura de transportes.
DNIT 031: Pavimentos Flexíveis – Concreto Asfáltico –
featured lower values of air voids than HMA, also Especificação de Serviço. Rio de Janeiro, 2006.
indicating a higher compactability of the mix. This Hurley, G. e Prowell B., Evaluation of EVOTHERM®
can be credited to an increase in lubrication pro- for use in Warm Mix. National Center for Asphalt
vided by the additive. Technology (NCAT) - NCAT Report 06–02. Auburn,
Despite adequate air voids values were reached Alabama, USA, 2006.
at 100°C for the WMA, caution is recommended Johnston, M. G.; Bock, A. L.; Brito, L. A. T.; Ceratti, J.
to lower below this limit. CDI shows a matching A. P.; Ribeiro, R. Influência do período de condicion-
pattern of WMA and HMA towards 100°C. Below amento em estufa sobre os parâmetros de dosagem de
this temperature air voids may be met, but densi- misturas asfálticas mornas. XVIII Congresso Ibero
Latinoamericano del Asfalto, Bariloche, Patagônia,
fication starts to slow down what may result in Argentina, 2015.
unwanted collateral effects. Kvasnak, A. et al., Alabama Warm Mix Asphalt Field
Regarding the workability tests from torque Study: Final Report. National Center for Asphalt
measurements, the little variation in results indi- Technology (NCAT) – NCAT Report 10-XX. Auburn
cates that both HMA and WMA presented similar University, Auburn, AL, USA, 2010.
workability under tested conditions. Because the Mahmoud, A. F. F.; Bahia, H. Using the gyratory com-
mixing time was short, it is arguable that the tem- pactor to measure mechanical stability of asphalt
perature binder decay from 153°C to one of the mixtures. Wisconsin Highway Research Program
mixing temperatures (100°C to 143°C) may have 0092–01–02, Madison, Wisconsin, EUA, 2004.
Masad, E.; Muhunthan, B.; Shashidhar, N.; Harman, T.
happened at low pace. The mixture workability (1999) Internal Structure Characterization of Asphalt
can then be attributed to the asphalt binder at its Concrete Using Image Analysis. Journal of Comput-
added temperature, resulting in slight variation in ing in Civil Engineering, v. 13, N°2, April.
the measured torque. Nascimento, L. A. H. (2008) Nova Abordagem da dos-
Results in this study agrees with Prowell et al. agem de Misturas Asfálticas Densas com Uso do
(2012) findings, confirming that, despite the low Compactador Giratório e Foco na Deformação Per-
temperatures, WMA more easily achieves the manente. Dissertação de Mestrado – COPPE/UFRJ,
desired density due to an increase on lubrication Rio de Janeiro.
provided by the surfactant. In addition, the digital Prowell, B. D.; Hurley, G. C.; Frank, B. Warm-mix
asphalt: best practices. 3ª ed. Lanham: National
image processing of specimen sliced into chops Asphalt Pavement Association. Quality Improvement
have clearly demonstrated a better homogeneity of Series 125, 2012.
particle distribution in WMA. Soares, J. S. (2014) Investigação da Relação entre Parâmet-
An additional conclusion is that CDI ought to be ros da Compactação Giratória e de Deformação
further investigated to act as reference to determine Permanente em Misturas Asfálticas Densas. Disser-
the amount of temperature reduction for WMA tação de Mestrado – Programa de Pós-Graduação
additives. The CDI value obtained in a standard em Engenharia de Transportes, Universidade de São
HMA can be expected to be achieved at a lower tem- Paulo, São Paulo.
perature by means of using a warm mix additive. Tashman, L.; Masad, E.; Peterson, B.; Saleh, H. (2001)
Internal Structure Analysis of Asphalt Mixes to
Improve the Simulation of Superpave Gyratory
to Field Conditions. Journal of the Association of
ACKNOWLEDGEMENTS Asphalt Paving Technologists, v. 70, p. 605–655.
Vasconcelos, K. L.; Evangelista JR., F.; Soares, J. B.
The authors wish to express their gratitude for the (2005) Análise da Estrutura Interna de Misturas
products suppliers—Ingevity specialty chemical, Asfáticas. In: XVII Congresso Brasileiro de Pesquisa
Greca asfaltos & Triunfo Concepa—and all other e Ensino em Transportes, Recife, PE.

341

TIS2017_Book.indb 341 2/8/2017 8:04:28 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A laser profilometer prototype for applications in road pavement

management system

G. Cerni & A. Corradini

Università degli Studi di Perugia, Perugia, Italy

ABSTRACT: Correct planning of road maintenance should require intervention before road reliability
conditions become unsustainable. To this aim, it is necessary to evaluate the rate of road degradation,
which is strictly linked to road roughness, using high performance instruments like the laser profilometer.
The International Roughness Index, obtained from profilometric survey, provides basic information
regarding the general state of a road pavement, but does not take into account the characteristics of
distresses. In this context, the University of Perugia (on behalf of the Province of Perugia) developed a new
prototype of laser profilometer and an innovative post-processing software. This equipment enables global
and local analysis. The global analysis provides the intervention priorities of a whole road infrastructure,
while, with local analysis, the single distresses and the cross slopes of a limited road portion can be evaluated.
Information about the type of distress are very useful in order to plan maintenance work.

Keywords: pavement, profilometer, maintenance

1 INTRODUCTION (λ) of its components: microtexture (λ < 0.5 mm),

macrotexture (0.5 < λ < 50 mm), and megatexture
Road maintenance is the complex of operations and (50 < λ < 500 mm). Wavelengths longer than the
activities aimed to conserve the functional and struc- upper limit of megatexture are defined as roughness
tural characteristics of pavements over the useful (Henry 2000). Different sizes of texture will affect
lifetime of road infrastructure (CNR 1988, Durango pavement surface characteristics in different ways
& Madanat 2002, Robinson et al. 1998). In particu- (Sandburg & Ejsmont, 2002). Small texture affects
lar, the purpose of road pavement maintenance is friction, while large texture affects ride quality
to guarantee correct function maintaining adequate (Flintsch et al. 2002). Noise and rolling resistance
safety levels while efficiently employing economic are principally controlled by macrotexture and
resources (Dekker et al. 1998, Canale et al. 1998). megatexture. In general, it can be stated that micro-
Preventive diagnosis is guaranteed by the study and texture and macrotexture have good effects on ride
interpretation of deterioration curves, which need quality, while greater texture are poorly accepted.
to be elaborated employing the principal indicators Roe et al. report that increased macrotexture reduces
of pavement state, namely skid resistance, evenness, total accidents, under both wet and dry conditions
bearing capacity and noise levels. The monitoring (Roe et al. 1998). Furthermore, this study shows
over time of these parameters, combined with the use that increased macrotexture reduces accidents at
of appropriate threshold levels, permits the control of lower speeds than previously believed.
pavement damage. This approach allows the identifi- With the aim to convert a pavement evenness
cation of the maintenance interventions at an early measurement into a single number, the highway
stage, so as to optimize costs and limit inconvenience industry introduced the International Rough-
to road users. The monitoring of the functional state ness Index (IRI) as a quality indicator (Sayers
of a road can be carried out using high performance et al.1986). Lin et al. (2003) stated that IRI may
devices, such as the laser profilometer. completely reflect pavement distress conditions.
A laser profilometer is a measuring instrument Thus, it is feasible to use IRI as a pavement per-
used to investigate road pavement surface, paying formance index. Tighe et al. (2000) observed that
special attention to its longitudinal profile in terms IRI is significantly related to single-vehicle acci-
of surface texture (Sayers & Karamihas 1998). As is dent rate.
well known, texture is defined by the irregularities According to ASTM E 1926-08 (ASTM Stand-
on a pavement surface that deviate from an ideal, ards 2000), IRI is obtained from measured lon-
perfectly flat surface. Pavement texture has been cat- gitudinal road profiles and is calculated using a
egorized into three ranges based on the wavelength quarter-car model (Fig. 1). A quarter-car system

343

TIS2017_Book.indb 343 2/8/2017 8:04:30 PM

 Figure 2. IRI ranges represented by different classes of
road and velocity (replotted from Sayers & Karamihas
1998).

Figure 1. Quarter-car vehicle model.

is composed of two solid masses ms (sprung

mass) and mu (unsprung mass), which respec-
tively represent ¼ of the body of the vehicle and
one wheel of the vehicle (Jazar 2014). A suspen-
sion spring of stiffness ks and a shock absorber
with viscous damping coefficient c, connect the
two masses, while the sprung mass is in contact
with the real pavement surface by a tire spring of
stiffness ks. Figure 3. Profilometer prototype.
During the simulation, the quarter car system
runs over the longitudinal profile at a constant 2 OBJECTIVES
speed (V) of 80 kilometres per hour (km/h). The
passage over this profile induces dynamic excita- The state of degradation of a road surface repre-
tion to the quarter car system, generating a cer- sents a key parameter in the management pavement
tain level of movement in both the sprung and system. With the end to obtain data pertinent to the
unsprung masses (Zs e Zu) in relation to surface state of pavement surfaces, in order to determine
irregularities. As a result, IRI value for a given objective criteria for the prioritization of mainte-
section length (L) is computed according to nance works, the Provincia of Perugia decided to
Eqn 1. equip itself with a profilometer (Fig. 3).
As an alternative to the direct acquisition of
1 L /V
IRI = ⋅ ∫0 Z s − Z u dt (1) the instrument, the Provincia chose to construct
L a prototype in collaboration with the Engineer-
ing Faculty, involving the Department of Civil
The IRI is a dimensionless parameter (generally and Environmental Engineering and the Depart-
expressed as m/km or mm/m) and it is null for a per- ment of Electronic and Computer Engineering.
fectly smooth road surface. In addition, its typical The principal motivation for this choice related
range values change with pavement type and corre- to the possibility to take full advantage of the
sponding speeds (Fig. 2). Although this index pro- data recorded by the apparatus’s sensors, while
vides basic information regarding the general state containing costs. To this end, particular attention
of a road pavement, it does not take into account was given to the creation of a post-elaboration
the information related to the characteristics of the software capable of providing, as well as IRI
distresses that are associated with a specific level indexes, information relative to the typology, the
of unevenness. For this reason, two road segments, gravity and the localization of irregularities. In
with different typologies of surface irregularities, addition, cross slope values can be obtained by
could give comparable values of IRI. elaborating data provides by the instrumentation.

344

TIS2017_Book.indb 344 2/8/2017 8:04:30 PM

 Nowadays, the profilometer prototype is at an 4 PROFILOMETER POST-PROCESSING
early stage of use and it is employed to monitor SOFTWARE ANALYSIS
the major road networks within the Provincia of
Perugia. The development of the post-processing software
(Fig. 5) has represented the most difficult step of
the project. It is used to analyse and elaborate the
3 PROFILOMETER HARDWARE whole database in terms of heights (i.e. distances
COMPONENTS between road surface and sensors), accelerations,
velocities and position.
The profilometer was built using the following
components (Fig. 4):
4.1 Software set up
− a bar with a length equal to 2.20 m;
The first goal of this experimentation was to obtain
− 10 lasers installed on the bar every 20 cm;
longitudinal profiles (in terms of elevation) and
− an inertial unit installed at the center of the bar;
cross slopes as close as possible to the real ones. In
− a control unit for analogue/digital conversion;
fact, during profilometric surveys laser measure-
− a GPS device;
ments are generally affected by vehicle oscillations,
− an odometer;
but such contributions should not be taken into
− a computer inside the vehicle for data collect-
account in the final profile analysis. In this context,
ing.
an already known profile, especially realized for
Each laser measures the distance between the this application, was studied in a preliminary inves-
sensor and the pavement surface at rates up to tigation in order to correctly set up the instrumen-
20 kHz, while the inertial unit, evaluating acceler- tation. In particular, a PVC profile was laid down
ation and angular velocities along three orthogo- over the pavement surface and speed bumps were
nal axes, detects the vehicle movements and makes positioned along the wheel paths, causing longitu-
the profile independent from these contributes. dinal and transversal oscillations to the vehicle. The
The profilometer records the position using GPS intention was to implement an algorithm capable
technology, georeferencing both laser and iner- of providing, in these operative conditions, a pro-
tial measurements. In particular, the Geographic file as faithful as possible to the original one.
Information System used is the World Geodetic With the same aim, high-resolution measure-
System, established in 1984 (WGS84). In the ments were carried out on eight road portions (with
last version of the prototype, an odometer was an overall length of 389 m) using a laser scanner.
added with the aim to achieve greater precision This data was than compared with those provided
and address, together with the inertial unit, the by the post-processing software of the profilom-
possible loss of data. The computer inside the eter. The comparison in terms of IRI showed an
vehicle records measurements in real time, which average error equal to 0.55 mm/m. The same eval-
will then be transferred to a post processing soft- uation with regard to cross slopes resulted in an
ware. Such technology provide both longitudinal absolute error of 0.85% and 0.65%, respectively
profiles and cross slopes along the path of the before and after the introduction of the odometer
vehicle. in the prototype.
The introduction of the odometer not only
improved the calculation of distances and
velocities, but also enabled us to know the real

Figure 4. Profilometer hardware equipment: bar (a);

lasers installed on the bar (b); inertial unit (c); on-board
computer (d). Figure 5. Main interface of the post-processing software.

345

TIS2017_Book.indb 345 2/8/2017 8:04:32 PM

 path of the vehicle during GPS data loss. This was be distributed. The prototype program performs
made possible by setting a procedure in which both numerous attempts by shifting the error along the
velocities obtained from the odometer and angular section subject to examination. All the resulting
velocities provided by the inertial unit, step by step paths comply with the final direction. The criteria
were integrated. This technique allows the identi- for choosing between all attempts consists in set-
fication of the plan position at a time step j from ting the minimum value of the distance between the
the knowledge of the position at a time step j−1. “end integration point” and the “real end point”.
In order to verify procedure reliability, simulated Figure 6c shows the result obtained applying the
data losses with different durations were intro- correction procedure to a period of simulation of
duced into the GPS database (an example is shown no GPS signal equal to 96 seconds.
in Figure 6a).
The integration procedure required the knowl-
4.2 Location of critical point on the pavement
edge of the initial conditions, which consist of the
surface
coordinates for the starting point of the integra-
tion and the initial angle to the north point (initial The decision to realize on our own the equipment
direction). enabled us to maximize all the information provided
Integrating step by step, the estimated path by the apparatus’s sensors. In particular, through
shown in Figure 6b can be obtained, but it is clearly the data analysis carried out by the post-process-
affected by problems of drift. The problem of the ing software, different types of pavement distresses
drift is due to small errors that are added together could be highlighted and classified. Irregularities
during the integration phase, resulting in an error were divided into two main categories: structural
that is more consistent the longer the integration distresses (deep) and surface distresses.
time. This problem is generally difficult to resolve The structural distresses were associated with
because of its non-linear nature and although the the depressions of the road surface caused by
use of more sophisticated instrumentation can structural collapses. They could present significant
reduced it, it cannot be eliminated. extensions and induce longitudinal and transversal
However, a new post processing procedure was oscillations in the vehicle.
designed. This procedure takes into account the The surface distresses include localized sin-
final conditions, both real and obtained by integra- gularities such as potholes, marked cracks and
tion (in terms of positions and angles), and the total detachments of asphalt concrete, which caused
length covered, in a context of iterative solution. sudden movements of the vehicle.
In fact, it was observed that the angular error, i.e. This division stems from the need to distin-
the angle resulting from the difference between the guish between the types of intervention required
expected and the obtained by the integration, can- to restore the evenness of a damaged road. For
not be linearly distributed throughout the affected example, if accentuated depressions are present, a
section because the coincidence between the “end simple road resurfacing to restore the evenness will
integration point” and the “real end point” would not be sufficient, since the problem could mainly
not be obtained. Therefore, the strategy to proceed reside in the deeper layers of the pavement.
by attempts is to adjust two variables which are To this end, the post-processing software is
the length and the position of the stretch, within based on two methods for the detection of road
the affected section, where the angular error can irregularities:

Figure 6. Example of GPS data loss simulation (a); estimated path (b); final path corrected after iterative procedure (c).

346

TIS2017_Book.indb 346 2/8/2017 8:04:33 PM

 1. SSD Method (Separation of Superficial Dis- localized discontinuities, which may then be
tress), to identify surface discontinuities. analysed with the second SWVW method as illus-
2. SWVW Method (Shifting Windows of Variable trated in the following paragraph.
Width) to identify wave irregularities.
4.2.2 The SWVW method
For both the discontinuities, the software
In the SWVW method, Shifting Windows of Vari-
elaborates a summarizing index (structural and
able Width are applied in steps, relative to the size
superficial) which provides additional information
of the window itself, moving along the length of
beyond those obtained by analyzing IRI values.
the profile. In relation to their width, different
wave irregularities were identified using 6 classes
4.2.1 The SSD method
obtained by dividing the range of roughness in
The method consists in a repeated application of a
four portions and the range of megatexture into
Separation of Superficial Distress routine to ana-
two portions. Each portion has a constant ampli-
lyse and elaborate the peaks of the first derivative
tude in the logarithmic scale. In this way, six bands
of the elevation profile. In synthesis the SSD
are obtained as a function of wavelength λ, as
routine calculates the first derivate of the profile,
reported in Table 2.
where the value is compared point by point with a
Fourier spectral analysis is carried out within
predetermined threshold. In this way, the disconti-
each shifting window. The spectrum used is the
nuities, which present an accentuated slope in the
Power Spectral Density (PSD) of slope and not
profile, are identified.
that of elevation in order to obtain a more uniform
Based on a prototype procedure implemented in
spectral density graph so as to allow a more ade-
the software, it is possible to separate the irregu-
quate comparison between the various wave irregu-
larities with reference to their length λ and depth.
larities (Sayers and Karamihas 1998). This graph is
With reference to the length, surface disconti-
therefore the most indicated to assess which possi-
nuities are classified as followed (Table 1).
ble wave bands represent a more relevant part of the
In addition, based on the maximum distress
irregularities of the global profile. The area under
depth, three different classes can be established,
the PSD of slope quantifies the irregularities of the
i.e. low (from 10 to 20 mm), medium (from 20 to
longitudinal profile. Modifying the window width
30 mm) and high (greater than 30 mm).
chosen to analyse the profile, it can be noted that,
At the end of the routine, by removing the
for the same irregularities, the spectrum change
superficial discontinuities and applying a joining
considerably and other bands are also influenced.
polynomial, the SSD profile is obtained (Fig. 7).
It is therefore appropriate to use shifting windows,
This produces a continuous profile without
rather than fixed windows, which move along the
scanned profile with a width slightly greater than
Table 1. Surface discontinuities. the wavelength analysed. Essentially, the methods
Name Wavelength [mm] consist of 6 passages, one for each band analysed.
The preliminary phase conducted through SSD
Small potholes 50 < λ < 150 method resulted to be necessary to ensure the
Medium potholes 150 < λ < 500 elimination of noticeable irregularities, in relation
Big potholes 500 < λ < 1500 to slope, which influence the spectrum in reference
to all bands. To quantify each band irregularity, the
areas under the PSD curves between fixed upper
and lower band limits were calculated (Fig. 8). For
each computed area, three different levels (low,
medium and high) were defined in order to estab-
lish the quantity of irregularity within each band.

Table 2. Wave irregularities.

Band number Band name Wavelength [mm]

Band 1 Short microwave 50 < λ < 158

Band 2 Medium microwave 158 < λ < 500
Band 3 Large microwave 500 < λ < 1581
Band 4 Short macrowave 1581 < λ < 5000
Band 5 Medium macrowave 5000 < λ < 15810
Band 6 Large macrowave 15810 < λ < 50000
Figure 7. Profile obtained after SSD method application.

347

TIS2017_Book.indb 347 2/8/2017 8:04:35 PM

 Figure 9. IRI index analysis.

Figure 8. PSD of slope.

The detailed information obtained from these

two prototype methods supply plentiful data
useful for local analysis. However, the quantity
of information is so vast it is difficult to use to
organize global maintenance. For this reason, the
data is summarized into two main indexes. In par-
ticular, the SWVW procedure results provide the
structural index as relevant to the macrowaves
contribution, while the results obtained within the
Figure 10. Different road conditions in relation to
microwave ranges, together with those of the SSD damage.
method, determine the identification of the super-
ficial index.
As well as the IRI index, the structural and
the superficial indexes are represented in graphs
4.3 Software output
(an example is shown in Fig. 11). These indicate
In the post processing software, the IRI values for the type of distresses present, globally quantified
hundred meter stretches are calculated in order by the IRI index, and therefore provide detailed
to provide, in a simple and practical way, indica- information in order to address the choice of inter-
tions for maintenance interventions (Fig. 9). These vention. When these index values are higher than
interventions are defined based on IRI thresholds, prefixed thresholds, their labels are displayed in the
which have been set with reference to existing state video footage (Figure 10).
of a 95 km provincial roads sample. The study In addition, a graph of cross slope for every
revealed the necessity to modify the IRI limits, 10 meters is constructed to verify the presence of
with respect to those usually utilized for highways, curves with unacceptable values (an example of
according to the following categories: counter slope is presented in Fig. 12).
The recorded data are georeferenced (using
− IRI <3.5 where the pavement is in satisfactory
WGS84) on an orthophoto map to facilitate
conditions (24% of the analysed roads)
their geographical location (Figs. 13–15). In
− 3.5 < IRI < 5 where the pavement is slightly dam-
particular, the IRI index is recorded on the
aged therefore requiring frequent monitoring
orthophoto map according to the new classifica-
(35% of the analysed roads);
tion introduced.
− 5 < IRI < 7 where the pavement is damaged
With reference to distress localization, the valid-
requiring intervention (29% of the analysed
ity of the orthophoto map identification was also
roads);
verified by analyzing laser scanner data collected
− IRI >7 where the pavement is significantly dam-
on a later investigation. In addition, after this sec-
aged requiring urgent intervention (12% of the
ond laser scanner survey, which was carried only in
analysed roads).
specific road stretches, it was decided to conduct
In the actual version of the software, the elabo- site visits in the critical sections identified by the
rated results from the various sensors are synchro- program. Both the site visits and the laser scanner
nized with video footage. In the resulting video, the survey confirmed the optimal performance of the
chainage is recorded in the lower left of the screen post-processing software to identify actual distress
and the IRI value in the upper right (Fig. 10). (Cerni & Durantini 2013).

348

TIS2017_Book.indb 348 2/8/2017 8:04:36 PM

 Figure 11. Structural and superficial index.

Figure 14. Georeferenced indication for superficial

and/or structural distresses for every 20 meters.

Figure 12. Cross slope analysis.

Figure 15. Georeferenced indication of critical cross

slope for a curve.

index values, the necessary interventions can be

objectively prioritized in table form.

Figure 13. Georeferenced IRI index for every 100 meters.

5 CONCLUSIONS
4.4 Intervention priority
Correct planning of maintenance activities is
Distress analysis provides a structural and a super- essential to preserve both the structural and func-
ficial index for every 100 meters, which together tional characteristics of a road infrastructure, in
with the IRI index permit the construction of an consideration of the best use of available finan-
intervention priority program. From an opera- cial resources. To this aim, an accurate analysis of
tional point of view, the identification of this the existing situation by investigating parameters
intervention program is one of the most useful such as evenness is necessary. The elaboration of
instruments present in the post-processing soft- the pavement profile can supply indications of
ware in order to define correct road maintenance. not only the superficial conditions, but also reveal
Based on a decreasing classification in terms of deeper structural problems.

349

TIS2017_Book.indb 349 2/8/2017 8:04:37 PM

 Road profiles can be obtained by carrying out REFERENCES
field investigations using high performance instru-
ments, such as the laser profilometer. In this light, ASTM Designation E 1926. 2000. Standard Practice for
the Province of Perugia decided to adopt a pro- Computing International Roughness Index of Roads
filometer prototype, which represents, together from Longitudinal Profile Measurements. American
with the post-processing software, an important Society of Testing and Materials. 995–1012.
Canale, S. & Nicosia, F. & Leonardi, S. 1998. Program-
apparatus for the maintenance of its road net- mazione degli interventi manutentivi in base ai limiti
work. In fact, the instrumentation provides high di budget. Atti del XIII Convegno Nazionale Stradale
quality results, together with a high speed of data dell’A.I.P.C.R., Verona.
acquisition and an ease of use. Data reliability of Cerni, G. & Duranti, A. 2013. La realizzazione di un
both the longitudinal profile and the cross slope prototipo laser come strumento diagnostico per la
were verified through an initial trial set up survey, manutenzione programmata delle pavimentazioni
which certified the accuracy of the equipment stradali. Strade & Autostrade, 6.
settings. CNR n.125. 1988. Istruzioni sulla pianificazione della
The post-processing software enables both glo- manutenzione stradale. Consiglio Nazionale delle
Ricerche, Roma.
bal and local analysis. The global analysis provides Dekker, R. & Plasmeijer R.P., & Swart, J.H., 1998.
the intervention prioritization of a whole road Evaluation of a new maintenance concept for the
infrastructure, while the local analysis allows a preservation of highways. IMA Journal of Management
detailed evaluation of the single distresses and the Mathematics, 9(2): 109–156.
cross slopes of a limited road stretch. Durango, P.L. & Madanat, S.M. 2002. Optimal
The post-processing software carries out road maintenance and repair policies in infrastructure
irregularities detection employing two different management under uncertain facility deterioration
prototype methods: the SSD method classifies rates: an adaptive control approach. Transportation
surface discontinuities, while the SWVW method Research Part A, 36(9): 763–778.
Flintsch, G.W. & Al-Qadi, I.L. & Davis R. & McGhee
identifies structural distress (deep). The surface K.K. 2002. Effect of HMA Properties on Pavement
distresses include localized singularities such as Surface Characteristics. Proceedings of the Pavement
potholes and detachments of asphalt concrete, Evaluation, Conference, Roanoke, Virginia.
while structural distress are associated with Henry, J.J., 2000. Evaluation of Pavement Friction
road surface depressions caused by structural Characteristics—A Synthesis of Highway Practice.
collapse. NCHRP Synthesis 291, Washington, D.C.
The possibility of distinguishing different kinds Jazar R.N. 2014. Vehicle Dynamics: Quarter Car Model.
of distress is very useful in order to identify the 985–1026, New York: Springer.
typology of intervention, avoiding a simple road Lin, J.D. & Yau, J.T. & Hsiao, L.H. 2003. Correlation
Analysis Between International Roughness Index
resurfacing without evaluating the real cause of (IRI) and Pavement Distress by Neural Network,
deterioration. Transportation Research Board, Washington, D.C.
Furthermore, the post-processing software could Robinson, R. & Danielson, U. & Snaith, M. 1998. Road
be extended to take into account additional data Maintenance Management-Concepts and Systems.
available in the Provincia of Perugia database with Basingstoke, UK: McMillan Press Limited.
reference to traffic, accident and friction as well Roe, P.G. & Parry, A.R & Viner, H.E. 1998. High and
as evenness, in order to more effectively prioritize Low Speed Skidding Resistance: The Influence of
interventions. The development of a pavement per- Texture Depth, TRL Report 307, Crowthorne, U.K.
formance prediction model could also be a future Sandburg, U. & Ejsmont J.A. 2002. Tyre/Road Noise
Reference Book. Kisa, Sweden: Informex.
objective. Sayers, M.W. & Gillespie, T.D. & Paterson, W.D.1986.
Guidelines for the Conduct and Calibration of Road
Roughness Measurements. The World Bank, technical
ACKNOWLEDGEMENTS paper n. 46, Washington, D.C.
Sayers, M.W. & Karamihas, S.M. 1998. The little
The authors thank the Province of Perugia (Italy), book of profiling. The Regent of the University of
and in particular the staff of the Traffic Manage- Michigan. Courtesy of the University of Michigan
ment Department, for their support during the Transportation Research Institute.
GPS field surveys. The help of Dr. Andrea Duran- Tighe, S. & Li, N. & Falls, L. & C. & Haas R,
2000. Incorporating Road Safety into Pavement
tini in the software development has been greatly Management. Transportation Research Record Journal
appreciated. of the Transportation Research Board, 1699(1): 1–10.

350

TIS2017_Book.indb 350 2/8/2017 8:04:41 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Speed management in single carriageway roads: Speed limit setting

through expert-based system

N. Gregório, A. Bastos Silva & A. Seco

CITTA, Department of Civil Engineering, University of Coimbra, Coimbra, Portugal

ABSTRACT: A number of approaches has been developed and adopted to set speed limits in interur-
ban roads. Among others, the Australian family of applications (XLimits), later also adopted in the USA,
is a relevant example of this type of approach. This model enables the selection of adequate speed limits
by taking into consideration a wide set of explanatory variables, aiming at describing the infrastructure,
land use, local safety and operational characteristics. Hence, the major objective of this work is to pro-
vide a generalized decision-support methodology for speed limit setting in interurban single carriageway
roads, crossing different types of road environments, considering a range of objective variables. This work
thoroughly describes the methodological approach to be used in this phase, which involves the evaluation
of experts and the collection of a set of variables related with the road environment. Some early results,
regarding these evaluation, are presented and discussed.

1 INTRODUCTION Hence, the major objective of the current work

is to provide an integrated decision-support meth-
Speed is among the most important factors odology for speed limit setting in interurban sin-
involved in road operation, influencing its users gle-carriageway roads, crossing different types of
behaviour. Usually, the most noticeable effect of road environment, and taking into account a range
speed is its impact over road accidents, whether of significant and objective variables.
over occurrence risk or over severity, whose rela- A number of approaches have over the years
tionship has been demonstrated by several studies been developed and adopted to set speed limits
(Baruya 1998; Farmer et al. 1999; Kloeden et al. in interurban roads. Most of these methodologies
2001; Taylor et al. 2002; Nilsson 2004; Aarts & usually give prevalence to geometric features of the
van Schagen 2006; Elvik 2013). Speed also affects road layout, both in their vertical and horizontal
environmental conditions, such as emissions, fuel alignments, and especially in critical sections, such
consumption, noise and the overall quality of life as curves, intersections or in stretches with higher
(ERSO 2006; Kockelman 2006; Austroads 2010). slopes. Examples of this approach can be found in
Despite its widespread use and acceptance numerous studies and are usually the basis of offi-
throughout the world, no consensus has so far been cial guidelines and statutory documents. However,
achieved among practitioners on the most reliable given the road environment complexity, a wider set
and effective method to set the speed limit in each of factors related to the surrounding areas, safety,
road section. This is a major concern, since it leaves traffic and users need to be included. A robust
the technical community without definitive guid- methodology should consider the wider number of
ance in this field (Forbes et al. 2012). Furthermore, factors possible, weighting their significance in this
the ever growing urban developments around most process.
roads has, in many countries, led to the existence The current work aimed to deliver an accessi-
of not only purely rural environments, but also ble and easy to use methodology, based on data
disperse and non-consolidated built-up areas in which can be collected remotely and is easily meas-
their surroundings, with the boundaries between urable. Therefore, a different approach is consid-
these zones very often difficult to identify. This ered, which emphasizes factors related with the
has led to ever more complex road environments, prevailing road environment, especially focusing
where the traditional road design and management on road integration into its surrounding areas.
principles are no longer sufficient, with problems Previous work has already been done to develop
of coherence and homogeneity arising to speed a robust methodological approach, based on an
limit setting strategies (Aarts & van Schagen 2006; expert-based system (Seco et al. 2008; Correia &
Hauer 2009; Stuster et al. 1998). Bastos Silva 2010; Correia & Bastos Silva 2011;

351

TIS2017_Book.indb 351 2/8/2017 8:04:43 PM

 Bastos Silva et al. 2012; Bastos Silva et al. 2016; the surrounding environment is; and it cannot
Gregório et al. 2016). However, additional devel- be considered as objectively rational, since it is
opment was still needed (I) to compose a new data- based on an erroneous driver perception of speed
base, involving more itineraries crossing different impacts (Elvik 2010; Park & Saccomanno 2006;
zones, and thus broadening the model’s scope, as TRB 1998).
well as a wider set of explanatory variables, rep- Another engineering method, the Road Risk
resenting more accurately the road environment; Method, has been used in New Zealand and
(II) to develop and to apply a systematic method Canada. This method is similar to the Operating
of expert assessment in laboratory and data collec- Speed Method, since both determine the recom-
tion, enhancing its applicability and usability. This mended speed limit based on a previously defined
work intended to thoroughly describe the method- value, which is then adjusted having into account
ological approach that was composed to the over- several other factors. The difference between these
all process, and to present and discuss some early two methods arises from the fact that, while in
results of the experts’ assessment to be included in the former the base value corresponds to the 85th
the final model. percentile of the free flow speed, this method uses
the functional classification of the road instead
(Forbes et al. 2012). Therefore, speed limit can be
2 SPEED LIMIT SETTING METHODS reconciled with the functional nature of the road,
in a more adequate way to its prevalent use.
The scientific and technical community has been
using several speed limit setting methods, which
2.2 Harm minimization methods
can be divided into four major types: engineering
methods, economic optimization methods, harm The Harm Minimization strategies address the
minimization strategies and expert-based systems. speed limit setting problem on a road safety per-
spective, considering that it is against ethics to
allow situations in which accident occurrence is
2.1 Engineering methods
possible. Thus, they focus on the tolerance and
The engineering methods usually set speed lim- integrity of the human body in accident situations
its based on an analysis of traffic and the road to set the speed limit. It has been implemented in
environment on the section under study and its some countries, with Sweden (Vision Zero) and the
surrounding roads. This approach includes the Netherlands (Sustainable Safety) as the most rep-
Operating Speed Method, which has been widely resentative cases.
used throughout the world, but especially in the Vision Zero takes into account three fundamen-
United States, with examples in several states (such tal ethical imperatives: no individual can die or
as Illinois, for instance) (Forbes et al. 2012). This suffer chronic injuries as a consequence of acci-
method is based on the 85th percentile of speed dent; road safety cannot be considered as a mobil-
(V85) and its usual procedure includes setting ity function, but rather as a road safety function;
the speed limit in a value equal to or higher than a monetary value or cost can never be attributed
V85, eventually adjusting it according to specific or associated to human life (Vaa 1999). Further-
infrastructure and traffic conditions (Forbes et al. more, this approach also usually forgives the driver
2012). It takes into account that speed values equal in case of accident, since even the systematic non-
to or near a standard deviation over the mean observance of the established rules by the users
value (which is near that percentile) tend to cor- supposes their inadequacy (Whitelegg & Haq
respond to a minimum accident risk for the driver 2006). However, the activity of particularly vul-
(DfT 2013). It also takes into consideration that nerable users is not appropriately addressed (Vaa
this kind of speed limit is in accordance to the 1999), and its realism and rationality are also ques-
perception of a vast majority of users about which tioned (Elvik 1999; Rosencrantz et al. 2007).
is the adequate speed under specific traffic condi- The Sustainable Safety approach also consid-
tions, thus contributing to a more uniform speed ers that it is not ethically acceptable for a system
regime (FHWA 2009; Forbes et al. 2012). to allow frequent accident occurrence with fatal
Nevertheless, this method presents several dis- or serious injuries (Vaa 1999), aiming to create a
advantages: it unrealistically assumes that drivers road system in which accident occurrence is strictly
select their travel speed taking an adequate and limited by an intrinsically safe road environment.
objective consideration of road safety issues; it is Thus, speed limit setting must allow to influence
considered to be the cause of a gradual increase both traffic homogeneity, and road layout and
of the average operating speed; it produces an user behaviour predictability, and must be safe and
inadequate measure of speed consistency; it credible (Wegman & Aarts 2006). Lack of cred-
tends to be less effective the more residential ibility from the driver’s perspective may lead them

352

TIS2017_Book.indb 352 2/8/2017 8:04:43 PM

 to question and ignore it, and, thus, need to be function, the system is prepared to estimate the
avoided (Goldenbeld & van Schagen 2007). speed limit value, based on data related to each one
of the considered factors.
In general, the advantages of using this type of
2.3 Economic optimization methods
approaches are evidenced by practice, which shows
The Economic Optimization methods intend to its comprehensiveness, consistence and reliability,
attribute a monetary value to all costs related to as well as the fact of being easily reproducible in
mobility, including those which are due to acci- different contexts (Forbes et al. 2012).
dent-caused damage. Among the various available This type of systems has been widely applied in
methods, the Optimum Speed Limits approach several places in the last few decades, with Australia
has been the most disseminated. This methodol- (XLIMITS family programs) and USA (USLIM-
ogy was initially proposed by Oppenlander (1962), ITS) as the most representative cases. Although
and intends to regulate traffic speed from a gen- there is not any known case of its application in
eral society’s point of view, recognizing that indi- a European country so far, it is considered to be a
vidual users do not always select speed taking into relevant improvement of the engineering methods,
account the risk imposed over the other road user due to its more realistic results, involving more eas-
individuals. This is due to the non-consideration by ily available data and embracing a wider number
the traditional methods of the external costs asso- of factors.
ciated with mobility—like those arising from fuel
consumption, emissions, noise and accidents—,
which produces a market imperfection (Elvik 3 METHODOLOGICAL APPROACH
2010). Therefore, curves of the cost function asso-
ciated to each one of these factors must be devel- The model is an expert-based system, following
oped to the various road sections, under different the previously presented assumptions, as well as
traffic conditions. The optimal speed value must Seco et al. (2008), Correia & Bastos Silva (2010),
be set as the minimum point of the total function, Correia & Bastos Silva (2011), Bastos Silva et al.
corresponding to the minimum transport cost from (2012), Bastos Silva et al. (2016) and Gregório
the society’s point of view. et al. (2015). As it was previously stated, the funda-
This method is not currently prescribed by any mental component of such a system is the knowl-
authority or agency, though a number of stud- edge base, i.e. the database which incorporates
ies were carried out in countries like Sweden and the information provided by the experts’ technical
Norway. experience and knowledge, and whose content will
allow the system to take decisions.
2.4 Expert systems
3.1 Data collection
Expert-based systems are related with the engi-
neering methods, as they intend to address their A number of 4 experts are involved in this proc-
lack of consistency and uniformity, aiming at more ess. They were selected based on their experience
realistic results (Austroads 2005). These systems and expertise in the field of traffic engineering,
are computational based programs which are used and above all in speed management. These experts
to solve complex problems recurring to decision intervene in two main phases: they were intended
algorithms and a database, allowing it to simulate to analyse the final set of variables to include in the
the behaviour, reasoning, evaluation and decision model; and they independently attributed a speed
process of experts (MnDoT 2012). Among all limit value to each analysed road section, based on
the components that compound such a system, their experience and expertise. Their evaluation
the database—the knowledge base—is particu- must only be influenced by road functionalities and
larly relevant, since the system bases its decisions interaction with the surrounding environment, and
on it. This database includes information aris- the existing posted speed limits were disregarded.
ing from the experts’ knowledge and experience, Therefore, a number of road itineraries was
structured in tasks to execute and decisions to take selected, corresponding to road segments of inter-
(TRB 1998). urban highways, in which data collection is car-
The development of the model that establishes ried out. These itineraries are representative of
the relationship between the knowledge base and several characteristic situations of this type of
the factors shall be carried out by collecting infor- roads: they cross several types of environment,
mation in a set of representative cases. The cali- namely urban areas, transition areas and rural
bration of this model uses the speed limit values environments (intra-section variability); some of
previously devolved by the experts, establishing the them were selected in regions with dispersed urban
knowledge base. After the definition of the model occupancy (cases in most of northern and central

353

TIS2017_Book.indb 353 2/8/2017 8:04:43 PM

 Portugal), and other ones in regions with a more system’s function, which establishes the relation-
concentrated and consolidated urban pattern, like ship between the knowledge base and the included
in southern Portugal (inter-section variability); factors.
they do not present too irregular or heterogeneous Discrete choice models are based on the theory
layouts. Eventually, more than 130 km of roads of stochastic utility, in which the choice is car-
will be included in the database. ried out by the user aiming to maximize the util-
The process of speed limit attribution and data ity function. The utility function is composed as a
collection is carried out in successive road sections combination of known explanatory variables, the
of 200 meters, in which each itinerary is divided. systematic part of the utility, and its random part,
This value is considered to be a sufficiently short which is unknown (Ben-Akiva & Lerman 1985).
length to guaranty a homogeneous level of both Thus, this function has the following form:
road physical features and surrounding environ-
ment characteristics. On the other hand, this U in Vin + ε iin (1)
distance is also considered to be long enough to
enable experts to make stable assessments. The where Uin represents the utility function, given by
experts assisted, in laboratory, to video recorded decision-maker n to alternative i, Vin its systematic
in which one of itineraries, in both directions. This part and εin the error between the systematic part
video was recorded in dry weather conditions. and the true utility. The systematic part of the util-
Data related with each one of the variables ity function given by user n to alternative i is, in its
included in this model is collected through the turn, represented by the following expression:
examination of satellite/aerial imagery and by using
the videos recorded in the considered itineraries. U in = β 0i β ' X iin (2)

3.2 Model estimation where β0i is the specific constant of each alterna-
The model is estimated by using this data and the tive i, β’ represents the vector of weights and Xin is
speed limit values proposed by each expert in each a vector of attribute values for each alternative i,
road section, which is the dependent variable. This given by a decision-maker n.
model is a Multinomial Logit, presenting four The error between the systematic part of the
alternatives, which are contemplated in the expert utility and its true value can be regarded as the part
decision-making process. These alternatives are the of the utility which is unknown to the analyst, i.e.
following speed limit values: as its random part (this is the most widely accepted
theory).
− 30 km/h (18.6 mph), which has been associated The hypotheses assumed about the statistical
with zones within urban areas with a strong resi- distribution of the error term of the utility func-
dential, commercial or services function, with tion lead to the adoption of different types of dis-
high volumes of pedestrians and cyclists (it even crete choice models. The Logit models are one of
originated the concept of Zone 30, now widely them. The Multinomial Logit model, which is part
used all over Europe); of this family, was developed as a generalization of
− 50 km/h (31.1 mph), the statutory speed limit for a binary choice model in a context involving more
urban roads in Portugal, i.e. roads crossing areas than two alternatives.
where there is a significant urban occupancy This model is based on the assumption that the
and presence of multiple functions, other than error terms of the alternatives are all independent
those related to motorized through traffic; and identically distributed (IID), according to a
− 70 km/h (43.5 mph), which is considered as a Gumbel distribution, also known as type I extreme
representative value for a transition zone (subur- value distribution (Ben-Akiva & Bierlaire 1999).
ban or peri-urban context, with a disperse urban This distribution directly implies that any differ-
occupancy); ence between the error terms is logistically distrib-
− 90 km/h (55.9 mph), the statutory speed limit for uted. According to this hypothesis, the probability
interurban highways in Portugal, crossing rural of choice of alternative i in this model is the fol-
areas where there is none or negligible urban lowing (Ben-Akiva & Lerman 1985):
activity.
The reference alternative must be 90 km/h, since eVin
it is the statutory speed limit for interurban high- Pn (i ) P(U in U jjn ) = (3)
V
ways in Portugal. ∑ e jn
j Cn
Taking into account that speed limit setting is
essentially a discrete choice problem, the Multi-
nomial Logit model was selected to estimate the where Cn is the choice set.

354

TIS2017_Book.indb 354 2/8/2017 8:04:43 PM

 Having into account that distributions are iden- − INTCR: Number of intersections with 4 or more
tically distributed, it is considered that the mean legs along the section;
assumes the same value for all the errors. However, − INTEN: Number of 3-leg intersections along
the mean is irrelevant, since only differences in the section;
the utility matter for changing the probability of − NATER: Number of lateral local accesses, both
choice. The best method to estimate this model is public and private;
through maximum likelihood, using the following − AGRO: Number of rural roads and paths, namely
function: small local access roads without pavement.
N Comparing with previous work, it was decided
L* = ∏ ∏ Pn (i ) y in (4) to create new variables regarding each type of
n =1 j C n
intersection. A new variable was also created to
small rural roads, given their specific location in
where yin represents the binary variable which rural zones.
assumes value 1 if the decision-maker n chooses
alternative I, and 0 otherwise. By using logarithms, b. Pedestrian accesses:
the maximization of the likelihood function can be − NAPED: Number of public pedestrian paths
simplified in the following expression: along the section;
N
− NAPEH: Number of private pedestrian paths
L yin log( Pn (i )n ) (5) through buildings, such as doors and gates,
n =1 i C n which are directly connected to the road;
− NCRO: Number of formal pedestrian crossings
Finally, the obtained model must be validated, along the section;
by carrying out other trials in one or more itin- − SIDEW: Proportion of the section with pedes-
eraries, which have not been involved in previous trian sidewalk.
stages, and following a similar procedure.
A new configuration was given to variable
SIDEW, which used to be a binary variable and
3.3 Explanatory variables now corresponds to a proportion of the section.

The set of explanatory variables was selected c. Parking:

according to previous work already developed − GARAG: Number of accesses to off-road park-
and to some shortcomings previously identified. ing, namely garages and other private parking
Examples of new variables now included are those situations;
related with lateral barriers, intersections, road − NAPAR: Number of on-road parking places
lighting and parking. These variables were also along the section;
selected having into account the experts sugges- − IPAR: Existence of illegal parking along the sec-
tions, in a previous stage before their assessment. tion (binary variable);
Variables with a binary character represent the − LOG: Number of accesses to open spaces sepa-
existence of the feature. Likewise, discrete variables rated from the road, both public and private,
represent the number of elements of that type which including off-road parking areas, petrol stations
can be found along the section in analysis. Other vari- and private patios or terraces, among others.
ables, with a continuous character, represent a pro-
portion of the road section where the feature exists. A new variable was created to address the prob-
On the other hand, edification density is lem of illegal parking, which is widespread in vari-
expressed in number of edifications per 100 meters ous locations. Another variable was composed to
of road, while distance-related variables are meas- include accesses to public and private open spaces
ured in meters. To determine these variables, only near the road, and connected with it. As a result,
edifications located within a range of less than previously considered variables, such as the number
30 meters of the roadway are taken into account, of petrol stations, were removed.
according to Bastos Silva et al. (2012).
d. Infrastructure:
Every variable, except NCRO and ISLAND, are
determined for both sides of the road–R for the − ISLAND: Proportion of the section with the
nearside of the road; L for the offside. presence of a central island;
Portugal is a country with right-hand traffic. − RLIGH: Proportion of the section with the pres-
ence of road lighting;
a. Intersections and motorized accesses:
− SIGN: Number of speed control traffic lights
− ROUND: Number of roundabouts along the along the section;
section; − NBUS: Number of bus stops along the section.

355

TIS2017_Book.indb 355 2/8/2017 8:04:43 PM

 A new variable was added to represent the exist- fact that each expert always analyse each road in the
ence of road lighting along the section, which was exactly same conditions of other experts (avoiding
previously mentioned and identified as a relevant the influence of undesirable factors to this process,
indicator of urban development. A new configura- such as weather conditions). Laboratorial conditions
tion were given to ISLAND, which now represents also allow the expert to avoid the sense of continu-
a proportion of the section, instead of the mere ity between two successive sections. The suitability
existence of this feature. of these methodology was already tested, showing a
high degree of consistence (Gregório et al. 2015).
e. Density and visibility:
− ED: Edification density per 100 meters of road,
within 30 meters off the carriageway; 4 PRELIMINARY ANALYSIS OF RESULTS
− MD: Minimum distance between edifications
and the carriageway, for edifications within An analysis with some early results obtained for
30 meters off the carriageway (in meters); the experts’ evaluation was carried out, allowing to
− MED: Median distance between edifications take some preliminary conclusions. This analysis
and the carriageway, for edifications within included the assessment of two experts over 73 km
30 meters off the carriageway (in meters); of roads in both directions (a total of 146 km),
− OPAC: Proportion of the section with the corresponding to 730 road sections. These sections
presence of a physical barrier on the roadside, included urban, transition and rural zones, and
including walls, fences, long barriers or moats. were collected in different regions, as it was previ-
ously mentioned.
The first three variables were previously defined
Although each expert was allowed to use each one
as objective variables, to better represent the effect
of the four available alternatives, the 30 km/h alter-
of buildings and other structures near the road
native was not used in any section, which is mostly
over speed limit setting (Bastos Silva et al. 2012).
due to the characteristics of the road sections con-
These variables intended to substitute other sub-
sidered in this analysis. Thus, in case their assess-
jective variables, and their adoption allowed to
ments were not identical, experts could diverge by a
obtain more accurate results.
value of 20 km/h (if their assessments were 50 and
On the other hand, OPAC is a new variable added
70 km/h, or 70 and 90 km/h) or by 40 km/h (in case
to represent the optical effect that physical barriers
their assessments were 50 and 90 km/h).
on the immediate roadside have over the driver’s per-
To analyse the consistency of their assessments,
ception, namely regarding two main problems: these
a statistical test was performed. Since this analysis
structures provide both a sense of opacity (which
may be considered as a paired observations-like
harms visibility) and of porosity (when there are
situation—two results with the same nature for
access points) on the roadside. However, having into
each one of the cases, collected under the same
account the diversity of situations that may fit under
conditions and assumptions, by analysing the same
this variable, a wide range of different cases were
road sections –, a paired samples t-test, a paramet-
included, for the sake of simplicity and ease of use.
ric test, was performed. This test allows to compare
two population means in case the corresponding
two samples are correlated, and it assumes the fol-
3.4 Expert assessment
lowing hypotheses (Washington et al. 2011):
A certain degree of disagreement is expected
between the speed limit values proposed by each H0: μd = 0 vs. H1: μd ≠ 0 (6)
expert. Despite Multinomial Logit model being
prepared to deal with these dissimilarities between where μd is the difference between both sample
expert evaluations, it tends to produce models rep- means.
resenting average assessments, regardless of the fact The obtained results are shown in Table 1. The
that models would present probabilities for each SPSS software was used in this analysis.
alternative. Since in safety-related problems it is A p-value of 0.030 was obtained in this test, which
recommended to use conservative, rather than aver- implies that the null hypothesis cannot be rejected for
age, options when traffic operational conditions are a significance level of 1%, but can already be rejected
to be selected, a virtual conservative expert is used,
selecting for each road section the second most
restrictive speed limit value from the choices made Table 1. Results obtained for the paired samples t-test.
by each expert, as in Bastos Silva et al. (2012).
The trials are carried out in laboratory, in order to N df t p-value
assure more homogeneous and replicable conditions,
Paired t-test 730 729 −2.171 0.030
due to its more controlled environment and to the

356

TIS2017_Book.indb 356 2/8/2017 8:04:44 PM

 for levels of 5% or over. This is a slightly inconclusive as a whole, will express the same proportion of
result, although it indicates that the null hypothesis divergence, since it is more likely to occur in these
can be rejected for some of the usual significance lev- sections in transition zones, disregarding their pro-
els and, therefore, the difference in means between pensity to a more or less conservative approach.
the two samples is not significantly closer to 0. This
result implies that the two assessment sets are not
significantly similar in statistical terms. 5 CONCLUSION
In order to better analyse these outcomes, those
sections where the expert assessment diverged were This work aimed to present an accessible and easy
identified. Thus, within the 730-cases set, divergence to use methodology to set speed limits in interur-
occurred in 192 sections, corresponding to 26.3% ban single carriageway roads. Thus, a methodo-
of the set. However, situations involving 50 and logical approach involving an expert system was
90 km/h in the same section were not reported. Such developed, including a knowledge base composition
a divergence would be excessive, and could indicate through expert assessment on a number of road sec-
some inconsistency of the experts’ assessment and tions, and data collection regarding a set of factors
of the overall methodology. Indeed, having all the related to the road environment and its surrounding
divergent cases occurring in 50–70 and 70–90 km/h areas. The usual factors related with road geometry
situations is an expected and reasonable outcome. and layout were disregarded in this phase, given
The obtained results are presented in Table 2. the short length (200 m) of the road sections. The
All sections where divergent assessment occurred experts’ assessment was delivered in laboratory, by
belong to transition zones. This was again an watching video recorded in each one of the analised
expected outcome, since in these zones character- roads. In this phase, only some results of this evalu-
istics of both urban and rural areas prevail very ation process were presented and discussed, allow-
often in the same road section, under a peri-urban ing to conclude that the most problematic sections
or suburban environment, sometimes coexistent are those in transition zones. However, the overall
with other non-residential or mixed uses. This is a results were consistent and showed the adequacy of
particularly relevant issue for some variables, espe- the methodology. Further work shall now be carried
cially edification density, intersections and lateral out in order to develop the knowledge base and then
accesses (all those variables usually express average to estimate the final model. Moreover, further work
values in these sections). Besides, in these sections can still be developed about the expert assessment
usually problems related with speed limit inconsist- process, involving additional trials and, in a later
encies are more recurrent and severe. stage, a thorough analysis in each one of the sections
These results were expected and are entirely where divergent results were obtained, in order to
acceptable. In fact, similar experts’ evaluations identify the responsible factors.
were reported in a wide number of sections (nearly Finally, in later phases, longer stretches of road
75%), in different road environments and regions, must be considered, involving a number of adjacent
proving the adequacy of the method and of the sections, and particularly focusing on transitions
conditions under which this process was carried between them, in terms of road consistency and
out. On the other hand, experts diverged in their layout homogeneity, and intending to correspond
judgment in a number of cases that is big enough to drivers expectancy. Later on, this analysis will be
to justify the scope of this work, proving that the performed on a route perspective, involving a longer
problem being addressed is relevant and wide- road extension and focusing on issues related with
spread all over the road network. safety, economic and environmental effects.
The overall results proved that one of the experts
(Expert 1) is slightly more conservative, which also
justifies the use of the previously described vir- ACKNOWLEDGEMENTS
tual conservative expert method. In any case, it is
expected that the assessment of the four experts, The authors would like to thank the Foundation of
Science and Technology (Fundação para a Ciência
Table 2. Divergence cases in the experts’ assessment, e a Tecnologia—FCT), under the MIT-Portugal
where each expert was more conservative. Program.
50, in 50–70 km/h 70, in 70–90 km/h

No. % No. % REFERENCES

Expert 1 37 5.1% 74 10.1% Aarts, L., van Schagen, I. 2006. Driving Speed and the
Expert 2 34 4.7% 47 6.4% Risk of Road Crashes: A Review. Accident Analysis
and Prevention 38(2): 215–224.

357

TIS2017_Book.indb 357 2/8/2017 8:04:44 PM

 Austroads. 2005. Balance between Harm Reduction and of Road and Person (ality) Characteristics. Accident
Mobility in Setting Speed Limits: A Feasibility Study. Analysis and Prevention 39(6): 1121–1130.
Research Report AP-R272/05. Sydney: Austroads. Gregório, N., Bastos Silva, A., Seco, A. 2016. Speed
Austroads. 2010. Impact of Lower Speed Limits for Management in Rural Two-Way Roads—Speed Limit
Road Safety on Network Operations. Publication No. Definition Through Expert-Based System. European
AP-T143/10. Sydney: Austroads. Transport Conference 2015; Frankfurt am Main, 28–30
Baruya, A. 1998. Speed-Accident Relationships on Euro- September, 2015. Transportation Research Procedia 13:
pean Roads. MASTER Project. 9th International Con- 166–175.
ference Road Safety in Europe; Bergisch Gladbach, Hauer, E. 2009. Speed and Safety. Transportation
21–23 September, 1998. Research Record (2103): 10–17.
Bastos Silva, A., Seco, A., Gregório, N. 2016. Set- Kloeden, C.N., Ponte, G., McLean, A.J. 2001. Travel-
ting Speed Limits in Interurban Single Carriageway ling Speed and the Risk of Crash Involvement on Rural
Highways Using Experts Judgment. Transport 31(2): Roads. Report No. CR 204. Canberra: Australian
282–294. Transport Safety Bureau.
Bastos Silva, A., Seco, A., Santos, S. 2012. Setting Safe Kockelman, K. 2006. Safety Impacts and Other Implica-
Speed Limits in Two-Lane Rural Highways Using tions of Raised Speed Limits on High-Speed Roads.
Expert’s Judgments. 91st Transportation Research Final Report. NCHRP Document 90 (Project 17–23).
Board Annual Meeting; Washington, D.C., 22–26 Washington, D.C.: Transportation Research Board.
January, 2012. MnDoT. 2012. Methods for Setting Posted Speed Limits.
Ben-Akiva, M., Bierlaire, M. 1999. Discrete Choice Transportation Research Synthesis. TRS 1204. Saint
Methods and Their Applications to Short-term Travel Paul: Minnesota Department of Transportation.
Decisions. In Handbook of Transportation Science, Nilsson, G. 2004. Traffic Safety Dimension and the Power
Vol. 23. International Series in Operations Research Model to Describe the Effect of Speed on Safety. Doc-
and Management Science. Kluwer. toral Thesis. Lund: Lund Institute of Technology.
Ben-Akiva, M., Lerman, S. 1985. Discrete Choice Analy- Oppenlander, J. 1962. A Theory on Vehicular Speed
sis: Theory and Application to Travel Demand. Cam- Regulation. 41st Annual Meeting of the Highway
bridge: MIT Press. Research Board; Washington, D.C., 8–12 January,
Correia, G., Bastos Silva, A. 2010. Setting Speed Limits in 1962. Highway Research Board Bulletin (341): 77–91.
Rural and Interurban Two-Lane Highways Using Expert Park, Y., Saccomanno, F. 2006. Evaluating Speed Con-
Opinion Crossed with Measurable Road-Side Char- sistency between Successive Elements of a Two-Lane
acteristics. 89th Transportation Research Board Annual Rural Highway. Transportation Research Part A: Pol-
Meeting; Washington, D.C., 10–14 January, 2010. icy and Practice 40(5): 375–385.
Correia, G., Bastos Silva, A. 2011. Setting Speed Limits Rosencrantz, H., Edvardsson, K., Hansson, S.O. 2007.
in Rural Two-Lane Highways by Modeling the Rela- Vision Zero—Is It Irrational? Transportation Research
tionship between Expert Judgment and Measurable Part A: Policy and Practice 41(6): 559–567.
Roadside Characteristics. Journal of Transportation Seco, A., Bastos Silva, A., Galvão, C. 2008. Speed Man-
Engineering 137(3): 184–192. agement Model Development Applicable to Regional
DfT. 2013. Setting Local Speed Limits. DfT Circular and National Single Carriageway through Roads.
01/2013. London: Department for Transport. 2nd European Road Transport Research Arena—TRA
Elvik, R. 1999. Can Injury Prevention Efforts Go Too 2008; Ljubljana, 21–25 April, 2008.
Far? – Reflections on Some Possible Implications of Stuster, J., Coffman, Z., Warren, D. 1998. Synthesis of
Vision Zero for Road Accident Fatalities. Accident Safety Research Related to Speed and Speed Manage-
Analysis and Prevention 31(3): 265–286. ment. Report No. FHWA-RD-98–154. Washington,
Elvik, R. 2010. A Restatement of the Case for Speed D.C.: Federal Highway Administration.
Limits. Transport Policy 17(3): 196–204. Taylor, M., Baruya, A., Kennedy, J. 2002. The Relation-
Elvik, R. 2013. A Re-Parameterisation of the Power- ship between Speed and Accidents on Rural Single-
Model of the Relationship between the Speed of Traf- Carriageway Roads. Report TRL511. Crowthorne:
fic and the Number of Accidents and Accident Victims. Transport Research Laboratory.
Accident Analysis and Prevention 50: 854–860. TRB. 1998. Managing Speed—Review of Current Practice
ERSO. 2006. Speeding. European Road Safety Observatory. for Setting and Enforcing Speed Limits. TRB SR254.
Retrieved on 31 August, 2016, from http://www.erso.eu. Washington, D.C.: Transportation Research Board.
Farmer, C., Retting, R., Lund, A. 1999. Changes in Vaa, T. 1999. Vision Zero and Sustainable Safety: A Com-
Motor Vehicle Occupant Fatalities after Repeal of the parative Discussion of Premises and Consequences.
National Maximum Speed Limit. Accident Analysis Borlänge: Swedish National Road Administration.
and Prevention 31(5): 537–543. Washington, S.P., Karlaftis, M.G., Mannering, F.L. 2011.
FHWA. 2009. Manual on Uniform Traffic Control Devices Statistical and Econometric Methods for Transporta-
for Streets and Highways. 2009 Edition. Washington, tion Data Analysis. Second Edition. Boca Raton:
D.C.: Federal Highway Administration. Chapman & Hall/CRC.
Forbes, G., Gardner, T., McGee, H., Srinivasan, R. Wegman, F., Aarts, L. 2006. Advancing Sustainable Safety:
2012. Methods and Practices for Setting Speed National Road Safety Outlook for 2005–2020. Leidschen-
Limits: An Informational Report. Report No. FHWA- dam: SWOV—Institute for Road Safety Research.
SA-12–004. Washington, D.C.: Federal Highway Whitelegg, J., Haq, G. 2006. Vision Zero: Adopting a
Administration—Office of Safety. Target of Zero for Road Traffic Fatalities and Serious
Goldenbeld, C., van Schagen, I. 2007. The Credibility of Injuries, for the UK Department for Transport. Stock-
Speed Limits on 80 km/h Rural Roads: The Effects holm Environment Institute. Norwich: HMSO.

358

TIS2017_Book.indb 358 2/8/2017 8:04:44 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Effects of mineral fillers on bitumen mastic chemistry and rheology

R.M. Alfaqawi, G. Airey, D. Lo Presti & J. Grenfell

Nottingham Transportation Engineering Centre, University of Nottingham, Nottingham, UK

ABSTRACT: Age hardening of bitumen is one of the key factors determining the lifetime of an asphalt
pavement. This study attempts to investigate the effect of different fillers including hydrated lime on mas-
tic ageing. The Thin Film Oven Test (TFOT) was used for short-term ageing, and the Pressure Ageing
Vessel (PAV) was used to simulate long-term ageing. The changes due to the ageing are measured by the
changes in complex modulus |G*| of the mastics. Chemical changes in the binder were evaluated by the
means of Fourier Transform Infrared Spectroscopy (FTIR). Results indicated that different filler min-
eralogy and properties can significantly affect the rate of oxidative ageing, and this depends mainly on
the filler type and concentration. It also showed that hydrated lime reduces the hardening rate more than
limestone and granite filler and this correlated with the chemical changes measured by FTIR in terms of
rate of change of the carbonyl oxidative products.

1 INTRODUCTION This is also the same in the current specifications.

It is important to recognise that the most intimate
Age hardening of bitumen has long been consid- contact between bitumen and aggregate in a pav-
ered one of the main factors that can significantly ing mixture involves the fine aggregate and filler.
affect the durability of pavements. When bitumen is Anderson et al. (1992) stated that the properties
age hardened, this process is usually accompanied of the fine fraction dominate in terms of physico-
by stiffening and embrittlement of the binder, thus chemical interactions between the bitumen and the
asphalt pavements become brittle and their ability mineral surface, as the fines are embedded in the
to support traffic-induced stresses and strains may bitumen and, thus, the majority of the surface area
significantly reduce. In addition, excessive hardening is generated by the fine minerals. Therefore, the
can also weaken the adhesion between the bitumen mechanisms of asphalt mastic ageing are influenced
and aggregate, which contributes to a reduction of by the characteristics of the asphalt binder and the
the durability of pavements and eventually increases mineral filler properties, in addition to the molecu-
the maintenance cost (Wu, 2009). Even though it is lar interaction between them. The chemical and
only one component of asphalt mixtures, the overall rheological changes associated with ageing are well
performance of the asphalt pavement is largely influ- understood for neat bitumen as the result of much
enced by the viscoelastic properties of the bitumen. research on this subject (Petersen 2009, Read &
Generally, bitumen ageing takes place in two stages; Whiteoak 2003). However, for the ageing of asphalt
short-term ageing at high temperatures during asphalt mastics, many topics in this research area still need
mixing and paving and long-term ageing at ambient to be investigated Moraes & Bahia (2015).
temperatures during in-service. In bitumen ageing, Little and Petersen, (2005) investigated the unique
two mechanisms can be involved; the main one is irre- effect of hydrated lime on the performance-related
versible, which is characterised by chemical changes properties of asphalt pavements. They compared the
of the binder where two processes take place, oxida- effect of hydrated lime to a similar size filler com-
tion and loss of volatile components. The other mech- prised of limestone. They performed extensive labo-
anism is a reversible process called physical hardening ratory testing, on mastics and mixtures. The results
that can be attributed to molecular restructuring. confirmed that when HL is added to bitumen, its
The oxidation of the hydrocarbon components effect will range between that equal to other mineral
in bitumen is largely accepted as the major factor of fillers to a considerably greater effect, depending on
the hardening and stiffening of asphalt pavements the physicochemical interaction developed between
Petersen et al., (1974a). According to Moraes & the bitumen and the HL. This interaction has proven
Bahia (2015), even though binder ageing in pave- to be different when different binders are used.
ment always occurs when the binder is in contact In order to develop a better understanding of
with aggregates and mineral filler, in most ageing the way by which the mineral fillers affect bitu-
studies, asphalt binders are individually aged with- men ageing, an ageing study on the bitumen-filler
out accounting for aggregate-induced interactions. mastics including DSR and FTIR testing after dif-

359

TIS2017_Book.indb 359 2/8/2017 8:04:45 PM

 ferent ageing stages was carried out. Mastic ageing is an increase of the SSA with the increase of the
has its advantages compared to asphalt ageing. As fine particle content. Also for larger values of
mentioned above, the surface area of fillers is much SSA more bitumen is expected to be adsorbed by
larger than the coarse aggregate, which will make the filler. Consequently, the filler stiffening effect
the adsorbing and catalysing phenomenon more should increase with SSA. The BET method is
obvious. In addition, as the fillers in mastics can be used for the calculation of SSA.
viewed as being embedded in the binder, the effects
of ‘void content’ and ‘binder film thickness’, which 2.2.4 Fractional voids
were concerns during the mixture ageing studies, The voids in dry compacted filler or Rigden Voids
could be avoided in the mastic ageing. (RV) were determined with the Rigden equip-
ment in accordance with BS EN 1097-4:2008. A
filler sample is first compacted in a mould with the
2 MATERIALS AND EXPERIMENTS defined compaction system, and then the void con-
tent is calculated considering the compacted vol-
2.1 Materials ume and the filler particle density. Several studies
have concluded this test value to be closely related
Three types of fillers were used in this study, Gran- to the filler stiffening power (Moraes & Bahia
ite filler (G), Limestone filler (LS), and hydrated 2015, Bahia et al 2010).
lime (HL). All fillers were substantially finer than
63μm in particle size.
A 40/60 Pen bitumen with a penetration of 2.3 Mastic tests
45 dmm and a softening point of 50°C was used to
2.3.1 Mastic ageing procedures
prepare mastics.
The Thin Film Oven (TFOT) Test is a standard
Granite or limestone fillers were mixed with the
binder test used for measuring the combined
40/60 bitumen to produce mastics with 50% filler
effects of heat and air on a film of bituminous
by mass. Then the filler (Granite or Limestone)
binder. The purpose of the test is to simulate the
was replaced by hydrated lime at percentages of
effect of ageing on binders due to short-term age-
10% and 20% by mass.
ing during production in an asphalt mixing plant.
The test was carried out according to BS-EN
2.2 Filler tests 12607-2:2000. In this test, mastic samples with the
same volumes are placed in a TFOT pan to form
2.2.1 Geometrical properties
a thin layer of the same thickness, which are then
The morphology and geometrical properties (size
held in the TFOT oven at 163°C for 5 hours.
and shape) of the selected fillers for this study were
The Pressure Ageing Vessel (PAV) ageing pro-
assessed by a grading test and a scanning electron
cedure is used to simulate long-term field oxida-
microscopy imaging technique. A Beckman Particle
tive ageing of asphalt binders in accordance with
Size Analyser (LS200) was used for measuring the
BS-EN 14769:2012. In this procedure, samples are
particle size distribution of fillers. In this analyser, a
aged in a vessel pressurised to 2.1 MPa and heated
flux curve is generated which is a composite of the
to 90°C in order to simulate in-service ageing over
flux curves for all the particles that pass through a
an estimated 7 to 10 year period (Bahia 2014). The
laser beam during a sample run. The LS 200 system
mastics in this study are aged for 20 h at 90°C in the
converts the composite flux curve into a particle size
PAV after the short-term ageing in TFOT oven.
distribution. This device uses water as a liquid disper-
sant, so it was not suitable for testing the Hydrated
lime fillers as there is a reaction between water and 2.3.2 Dynamic shear rheometer
the hydrated lime-Ca(OH)2. So, the particle size dis- The rheological properties of the mastics were char-
tribution for hydrated lime was given by the supplier acterised in this study with a Bohlin Gemini model
as it was tested using a Malvern Mastersizer. DSR. A frequency sweep is performed with a 0.3%
strain amplitude over a range of 0.1 to 10 Hz and
2.2.2 Particle density temperatures from 0 to 50°C. The resulting complex
The particle density (SG) was measured by the shear modulus |G*| data is used to produce master
pycnometer method in accordance to BS EN 1097- curves using the time-temperature superposition
7:2008. This test is required for the Rigden voids principle, with shift factors following the Williams-
calculation. Landel-Ferry (WLF) equation. The selected refer-
ence temperature for the master curves is 30°C.
2.2.3 Specific surface area
The Specific Surface Area (SSA) was determined 2.3.3 Ageing index
to evaluate the stiffening power relation with the In order to show the effects of ageing on the
fineness of the filler. It is well known that there mechanical and rheological properties of mastics,

360

TIS2017_Book.indb 360 2/8/2017 8:04:45 PM

 an ageing index was calculated after each stage ers. In Figure 2 it can be seen from the microscopic
of ageing: short-term and long-term ageing. The morphology of the different filler particles by SEM
Ageing index was calculated using |G*| at 20°C and imaging that, hydrated lime has mostly granular
0.4 Hz according to Equation 1. type particles with very rough texture. However
Granite and Limestone fillers have mostly angular
∗
Gage type particles with smooth to slightly rough tex-
I x=
d mastic
Aging Inde ∗
(1) ture. The SEM imaging results can be correlated
Gun aged mastic
with the other filler tests. As the larger value of the
specific surface area SSA and RV of hydrated lime
are consistent with the granular shape of particles.
2.3.4 Chemical analysis of bitumen and mastics The higher SSA of hydrated lime could be related
Fourier transform infrared FTIR spectroscopy is to the internal pores within the hydrated lime par-
the most powerful means of detecting and identi- ticles and its rough surface compared to limesone
fying chemical bonds (functional groups) in either and granite fillers.
organic or inorganic materials. The asphalt binder’s
main functional groups can be quantified by using
the FTIR technique. Therefore FTIR is widely used 3.2 Mastics
to characterise the oxygen-containing functionali- 3.2.1 Dynamic Shear Rheometry (DSR)
ties found in bitumen, particularly those produced The master curves of the complex modulus G* of
during oxidation Wu, (2009). Typical wave numbers bitumen and mastics are shown in Figure 3 a to c,
for bitumen ageing products are sulfoxides S = O at it can be noticed that the oxidative ageing distorted
1030 cm−1 and C = O in carbonyls at 1700 cm−1. The the shape of the original 40/60 bitumen master
FTIR test was performed on the recovered bitumen curve by decreasing the slope, and thus decreasing
from mastics before and after ageing. Carbonyl the time-dependency of the material response in this
C = O ageing index was used to evaluate the effect frequency range. This effect for bitumen is in line
of ageing on different mastics and it was calculated with previous research by Moraes & Bahia (2015).
as follows: Carbonyl index (C = O): A1700/ΣA. On the other hand, the addition of mineral filler
The sum of the area ΣA represents: A1700 + A1600 to bitumen shifted the curve toward higher fre-
+ A1460 + A1376 + A1030 + A864 + A814 + A743 + A724 + quencies without changing the slope, and thus a
A(2953, 2923, 2862). more elastic response, which is expected to be due
to the addition of elastic filler particles.
An important observation is that ageing of
the mastics, regardless of filler mineralogy, also
3 RESULTS AND DISCUSSION
resulted in a rotation of the master curve in the
reverse direction of that caused by ageing of the
3.1 Filler
binder alone. The result of which is aged mastic
shows the gradation curves of the used fillers; it
can be seen that hydrated lime and limestone filler
almost have the same gradation, but granite filler
is significantly coarser. However more than 95%
of the filler particles are smaller than 63 μm. The
properties of the fillers such as the Specific Grav-
ity, Rigden Voids SSA are shown in Table 1.
These results are expected to show a strong
correlation with the filler particle shape and
morphology. The results show that, although the
hydrated lime and limestone have almost the same
grading and particle size, HL has significantly
higher Rigden Voids and specific surface area
compared to both the limestone and granite fill-
Figure 1. Gradation curves of fillers.

Table 1. Filler properties.

Filler type Specific gravity Rigden voids (%) Surface area (m2/g)

Granite (G) 2.66 47.15 1.26

Limestone (LS) 2.65 39.82 1.58
Hydrated Lime (HL) 2.22 61.62 2.24

361

TIS2017_Book.indb 361 2/8/2017 8:04:45 PM

 Figure 2. SEM imaging of fillers a) Granite, b) Limestone, c) Hydrated lime.

Figure 3. Oxidative ageing effect on master curves of bitumen and mastics: a) complex modulus of 40/60 bitumen,
b) complex modulus of granite mastics c) complex modulus of Limestone mastics, d) phase angle of granite mastics,
e) phase angle of limestone mastics.

master curves that are only shifted in comparison reduces the change in phase angle or increases the
to the neat, without the distortion and rotation phase angle in the range of frequencies tested. This
observed for ageing of binder alone. is a relative measure of the ability of mastics with
The master curves of the phase angle of original hydrated lime to dissipate shear stress. The increase
and PAV aged bitumen and mastics are shown in of phase angle by hydrated lime should be benefi-
Figure 3 (d and e), it is clear that the addition of cial to reduce cracks in aged pavements.
the mineral fillers to bitumen decreases the phase The effect of hydrated lime on the phase angle
angle in all cases. However after PAV ageing the was more significant with the granite mastics than
effect of hydrated lime replacement is clear. It the mastics with limestone.

362

TIS2017_Book.indb 362 2/8/2017 8:04:45 PM

 3.2.2 Ageing Index calculation It also can be seen that the combination of
As can be seen from Figure 3, the comparison 30%LS+20%HL is a stiffer mastic before ageing,
between the filler effect on stiffening and ageing however after short-term and long-term ageing, mas-
of mastics is not clear from the master curves. In tics with LS+HL are softer than mastics with G+HL
order to make the comparison more quantitatively, fillers, thus meaning this combination reduces the rate
the complex modulus (G*) values at 20°C, 0.4 Hz of ageing and oxidation within the mastics. The initial
for the 8 mm testing geometry were selected for the stiffening effect of HL could be related to its higher
ageing index calculations It can be seen in RV and the larger surface areas of hydrated lime and
Figure 4(a) that all mastics complex modulus limestone filler compared to the granite filler.
values increase with ageing.
In all cases after ageing, mastics containing 3.2.3 FTIR results
granite and hydrated lime fillers are stiffer than The infrared absorption spectrum between 500 cm−1
those containing limestone and hydrated lime filler and 4000 cm−1 has been recorded for bitumen
with the same percentages. It could also be noticed, re-covered from granite mastics, and the absorption
that by replacing with 20% of hydrated lime does bands belonging to asphalt functional groups that are
not make significant additional effect compared to affected by ageing, such as carboxylic acids (C = O)
the 10% of hydrated lime with granite fillers. were used to evaluate the increase in oxidative ageing
products, which can be determined through FTIR,
utilising a very strong band around 1700 cm−1.
Figure 5 shows the FTIR spectrum for the
recovered binder from mastics before and after
different ageing stages. It can be noticed that the
carbonyl area increases with the ageing of the mas-
tics. In addition, according to this scan result, the
carbonyl index was calculated and presented in
Figure 6. It can be seen that the binder recovered
from the mastic of hydrated lime with granite con-
tains less oxygenated products after both short—
and long-term ageing. These results can explain
the lower ageing index (G* ageing index) for mas-
tics including hydrated lime which means a slower
rate of increase of the heavier oxygenated products
in mastics containing hydrated lime with ageing.
In order to show the effect of hydrated lime on
the rate of stiffening of the mastics after ageing,
the mastics containing 50%G, 40%G+10HL, and
30%G+20HL were aged in the TFOT oven for dif-
ferent ageing times up to 20 hours, and the complex
moduli of the aged mastics were measured at 20ºC
and 0.4 Hz and presented in Figure 7(a). From these
results the ageing index was calculated as mentioned
before and the results are presented in Figure 7(b).
Figure 4. Complex modulus (G*) ageing index of filler It can be seen that, mastics containing hydrated lime
mastics. (a) Mastics stiffness (G*), (b) Ageing index. are initially stiffer than the granite only mastic.

Figure 5. FTIR spectrum for recovered bitumen of granite mastics.

363

TIS2017_Book.indb 363 2/8/2017 8:04:46 PM

 long-term ageing periods in the lab and the change
in rheological and chemical properties of the mas-
tics and the recovered binders of the mastics were
evaluated by means of the DSR and FTIR respec-
tively. The following conclusions could be drawn:
– The ageing of bitumen mastics is different from
the ageing of pure bitumen as can be seen from
the complex modulus master curves.
– The mineralogy and the properties of the fillers
can significantly affect the stiffening ratio of the
Figure 6. Change of Carbonyl of recovered binder bitumen mastics and that depends mainly on the
from mastics with ageing. filler type and concentration in the asphalt mastic.
– The effect of the hydrated lime replacement
on the phase angle change after PAV ageing
was measured. It was clear that hydrated lime
increase the phase angle after the PAV ageing.
This effect could be beneficial to reduce the
cracks of lime treated pavement after ageing.
– Replacing 10% or 20% of granite or limestone fill-
ers in the bitumen mastics by hydrated lime filler
is more effective in terms of reducing the rate of
ageing on the mastics. It can reduce ageing by up
to 50% after long-term ageing and this could be
used to produce more durable asphalt pavements.

REFERENCES

Anderson, D.A., Bahia, H.U. and Dongre, R. (1992) Rheo-

logical Properties of Mineral Filler-asphalt Mastics and
Their Relationship to Pave-ment Performance, Ameri-
can Society for Testing and Materials, Philadelphia.
Bahia HU, Faheem A, Hintz C, Al-Qadi I, Reinke G,
Dukatz E. Test methods and specification cri-teria
for mineral filler used in HMA. In: Program NCHR,
editor. NCHRP Project 9–45. USA: Trans-portation
Research Board; 2010.
Figure 7. (a) Change of G* at 20˚C with ageing time, BS EN 1097-4:2008, Tests for mechanical and physical
(b) G* ageing index. properties of aggregates. Determination of the voids
of dry compacted filler
However with ageing the rate of increase in BS EN 1097-7:2008, Tests for mechanical and physical
stiffness of the granite only mastics is higher than properties of aggregates. Determination of the parti-
cle density of filler. Pyknometer method
the mastics with hydrated lime. And it is clear that Moraes, Raquel, and Hussein Bahia. “Effect of Mineral
replacing 10% or 20% of Granite filler by hydrated Fillers on the Oxidative Aging of Asphalt Binders:
lime reduces the ageing index by almost 50% com- Laboratory Study with Mastics.” Trans-portation
pared to the mastic without hydrated lime. Research Record: Journal of the Transportation
These results prove that hydrated lime is a more Research Board 2506 (2015): 19–31.
effective filler than limestone and granite in terms Petersen, J.C., Barbour, F.A. and Dorrence, S.M.(1974a)
of reducing the rate of ageing and the production of Catalysis of Asphalt Oxidation by Mineral Aggregate
oxidative ageing products such as carbonyls. Thus, Surface and Asphalt Compo-nents, Proc. Association
this proves that there is a physicochemical interac- of Asphalt Paving Tech-nologists, Vol. 43, pp 162–177.
Petersen, J. C. A Review of the Fundamentals of Asphalt
tion between hydrated lime and the bitumen which Oxidation - Chemical, 9 Physicochemi-cal, Physical
is not present with limestone or granite fillers. Property, and Durability Relation-ships. Transporta-
tion 10 Research Circular E-C140, Transportation
Research Board, 2009.
4 CONCLUSIONS Read, J., and D. Whiteoak. The Shell Bitumen Hand-
book, Fifth Edition, 2003.
In this study the effect of different mineral fillers Wu, Jiantao. “The influence of mineral aggregates and
of the ageing of bitumen mastics were investigated. binder volumetrics on bitumen ageing.” Nottingham:
The bitumen mastics were aged for short—and University of Nottingham (2009).

364

TIS2017_Book.indb 364 2/8/2017 8:04:47 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Functional and durability properties evaluation of open graded

asphalt mixes

V.M. Kolodziej, G. Trichês, J.S. Ledezma, G.C. Carlesso, L.M. Jardín & R.M. Knabben
Universidade Federal de Santa Catarina, Santa Catarina, Florianópolis, Brasil

ABSTRACT: Open graded asphalt mixes are one solution to the current demand for roads that offer
quality, safety and comfort to users. However, these mixtures have problems of durability and mechani-
cal strength. In this context, the aim of this research is to obtain a mixture that presents a balance
between the permeability, acoustic and durability properties. This study evaluated mixtures with two
different aggregate gradations with a maximum aggregate size of 9.5 mm and a polymer modified
binder, through a series of laboratory testing. One of the two aggregate gradations was studied in Brazil
(G1), and another gradation was based on a work done in the United States (G2). The research con-
cluded that G2 mixture shows satisfactory performance related to acoustic, permeability and durability
properties.

1 INTRODUCTION Presence of water on road surfaces involves

safety problems for users by the loss of tire/pave-
From a roadway users perspective, the more inter- ment adhesion, the risk of aquaplaning and the
ested functions of a pavement are more related to decreased visibility, which increases the number of
its surface performance, including smoothness, traffic accidents in rainy days (Meurer Filho, 2001).
safety, low noise, and good visibility of markings A surface with high air-voids contents has the abil-
at night and during raining. ity to eliminate the water depth of pavement sur-
Traffic noise pollution has become a growing faces, which is formed during a rain. According to
problem because of the continuous increase of Oliveira (2003), pores of mixtures can be classified
traffic volume. Traffic noise is a primary source as effective pores (which contribute to passage and
of noise pollution that affects human health and to water storage), semi-effective pores (that not
life quality of bordering residents (WHO, 2011). contribute to water passing, but allow water stor-
Sources of noise produced by vehicle traffic are age) and ineffective pores (that not contribute to
divided into motor noise, aerodynamic noise storage or passage of water).
and tire/pavement noise (Rasmussen et al, 2007). Both the noise caused by tire/pavement contact
According to Bernhard and Wayson (2004), in light as the presence of water can be reduced or mini-
vehicle traffic situations and speeds above 50 km/h, mized by optimizing characteristics of pavement
the amount of tire/pavement noise exceeds motor surface. Open graded asphalt mixtures have the
and aerodynamic noise. capacity to absorb the tire/pavement noise and
Sound absorption of a road surface is considered allow a passage of water through its internal struc-
one of the parameters that influence the road noise ture (PIARC, 2013).
levels, which are generated by the contact tire/pave-
ment (Li et al., 2015). Existing studies show that the
2 EXPERIMENTAL DESIGN
sound absorption is effective to reduce tire/pave-
ment noise in the frequency range between 800 and
2.1 Materials and mix designs
1000 Hz (Sandberg & Ejsmont, 2002).
Sound absorption coefficient is the ratio of This experiment includes two aggregates gradations
the incident acoustic energy and acoustic energy and one SBS polymer modified asphalt binder type
absorbed by the surface. This value is always posi- 60/85. One granitic nature aggregate was used for
tive and can vary between 0 and 1, where 1 rep- all mixtures. The chosen aggregate gradation G1
resents a purely absorbing material (Gerges & was previously studied by Guimaraes (2012) in
Arenas, 2010). Factors such as air-void content, Brazil, and this mixture showed good resistance to
texture patterns, layer thickness and the structure abrasion loss and permeability. Aggregate grada-
of voids influence the sound absorption capability tion G2 was previously studied by Lu et al. (2009)
of surfaces (Liu et al., 2016). in the United States. In this study, however, there

365

TIS2017_Book.indb 365 2/8/2017 8:04:50 PM

 was a modification of the gradation G2, inserting 2.2.2 Porosity
a larger percentage of coarse aggregate (material In order to obtain the porosity of mixtures was
retained on 4.75 mm sieve), in order to improve its measuring the amount of water that penetrates
permeability. The aggregate gradations G1 and G2 inside the specimen for the upper face, when the
used in this study are showed in Table 1. side and bottom faces are waterproofed by adhe-
In order to obtain the optimum binder content sive tape and paraffin remaining on its upper face a
of the mixtures were evaluated four binder contents constant film of water for ten minutes. This proce-
(4.5%, 5.0%, 5.5% and 6.0%), being shaped three dure was performed as recommended by AFNOR-
specimens in Superpave Gyratory Compactor for NF-P-98-254-2, 1993.
each binder content. For each specimen was deter-
mined the air-void content, porosity and abrasion 2.2.3 Resistance to raveling
loss. The optimum binder content was the one corre- Resistance to raveling was evaluated using the
sponding to air-void content higher than 18%; poros- Cantabro test on specimens, following the proce-
ity higher than 10% and abrasion loss less than 20%. dure described in DNER-ME 383/99. Specifically,
compacted specimens were put inside a Los Ange-
2.2 Test methods les Abrasion machine drum without steel balls,
and the drum was turned for 300 revolutions in
The mix properties that are critical to pavement 10 minutes. The percentage of mass loss during
surface performance were evaluated in this study this process was calculated by Equation 2.
including air-void content, porosity, resistance to
raveling, moisture sensitivity, permeability and
acoustic absorption. ⎛ Pi Pf ⎞
D=⎜ × 100
Pi ⎟⎠
The Superpave Gyratory Compactor was used (2)
⎝
to fabricate specimens of 100 mm diameter for the
Cantabro test, moisture susceptibility and acous-
tic absorption tests. The specimens were fabricated where Pi = initial sample weight of the sample;
using 50 gyrations of the SGC, applied pressure of Pf = final weight of the sample; and D = Cantabro
600 kPa, and rotation angle of 1.25°. The LCPC abrasion loss.
(Laboratoire Central des Ponts et Chaussées) com-
pactor was selected to compact slabs specimens of 2.2.4 Moisture Susceptibility
dimensions 50 cm length, 18 cm width and 5 cm Moisture susceptibility of the mixtures was deter-
thick for permeability test. mined using the AASHTO T 283 test method
with some modifications as specified in ASTM D
2.2.1 Air-void content 7064. For moisture conditioned specimens, they
The air void content of specimens was calculated were first saturated at a vacuum of 87.8 kPa for
from the theoretical maximum specific gravity 10 minutes, and then submerged in water during
(Gmm) measured in accordance with ASTM D2041 the 16 hrs freeze cycle. After, in the 24 hrs thaw
and the bulk specific gravity (Gmb) measured using cycle, specimens were submerged in the 60°C water
the DNER-ME 117/94 (Equation 1). bath. Instead of five freeze/thaw cycles as specified
in ASTM D 7064, only one cycle was applied.
⎛ G Gmb ⎞
Vv = ⎜ mm ⎟⎠ × 100 (1)
⎝ Gmm 2.2.5 Permeability
The permeability of compacted slabs was determined
with LCS permeameter in accordance with Spanish
Table 1. Aggregate gradations used in the study. standard NLT-327/00. In this test is measured the
Percentage Passing (%) time required for the percolation of a column of
water in the asphalt mixture layer. The mixture per-
Sieve size mm G1 G2 meability characteristics were determined as a func-
tion of flow time in accordance to Equation 3.
9.5 100 100
6.3 43.07 – K 7.624 1.348 ln
l T (3)
4.75 12.51 55
2.36 12.51 14
where K = permeability coefficient; and T = water
1.18 12.51 12
flow time in LCS permeameter.
0.6 12.51 10
0.3 12.51 7
2.2.6 Acoustic Absorption
0.15 7.14 6
0.075 4.08 5
The acoustic absorption coefficient of a material
is the ratio of acoustic energy not reflected by the

366

TIS2017_Book.indb 366 2/8/2017 8:04:50 PM

 surface material for a normal incidence plane wave. mixture. In Figure 3 and Figure 4 are represented
The sound absorption coefficient as a function of the results of porosity of G1 and G2, respectively.
frequency was measured in accordance with ISO
10534–2:1998, using an impedance tube of 100 mm
3.3 Resistance to raveling
diameter.
One of the main failure modes of an open mix-
ture is the raveling, which is caused by vehicle traf-
3 RESULTS AND DISCUSSION fic and high temperatures. This failure mode was
evaluated for each specimen by the Cantabro test.
3.1 Air-void content The results are shown in Figure 5 and Figure 6 for
G1 and G2 mixtures, respectively. Analyzing this
Results of determination of air-void content for
results it can be seen that the resistance to raveling
the three specimens for each asphalt binder con-
increases as the binder content is increased. The
tent and each aggregate gradations G1 and G2 are
mixture G1 has lower resistance to raveling than
represented in Figure 1 and Figure 2, respectively.
the mixture G2, and the reason is explained by the
It is noted that air-void content in the mixtures
higher air-void contents and porosity identified in
decrease as it is increased the binder content. All
the first one that decreases its resistance. It can also
samples had air-void contents greater than 18%,
be seen in Figure 5, only for 6.0% binder content
which is the minimum required for open graded
the mix G1 meets the specification of less than 20%
friction course mixtures.
Cantabro abrasion loss. In the case of mixture G2,
the specification of less than 20% abrasion loss
3.2 Porosity holds for 5.5% and 6.0% binder contents.
In regard to test results determination of poros-
ity, all specimens had higher than 10% (minimum 3.4 Optimum binder content
required to open mixtures), and their values decrease
Based on the results obtained for the air-void
as it is increased the binder content. In general,
contents, porosity and abrasion loss for G1 and
higher porosity is related to higher permeability of

Figure 1. Air-void content of G1. Figure 3. Porosity of G1.

Figure 2. Air-void content of G2. Figura 4. Porosity of G2.

367

TIS2017_Book.indb 367 2/8/2017 8:04:51 PM

 Figure 5. Cantabro abrasion loss of mixture G1.

Figure 7. Indirect tensile strength test results.

Figure 6. Cantabro abrasion loss of mixture G2.

G2 mixtures, were selected the optimum binder

contents that met both the following conditions:
(1) air-void contents higher than 18%; (2) porosity Figure 8. TSR test results.
greater than 10%; and (3) abrasion loss less than
20%. As can be seen in Figures 1, 2, 3 and 4; all
binder contents met the air-void contents and Figure 8 shows the results of tensile strength
porosity specifications. However the results of the retained (TSR) for two mixtures (G1 and G2). It
abrasion loss of the mixture G1 (Fig. 5) show that is conclude that all mixtures showed values above
only for a binder content of 6.0% was achieved a 80% that is the minimum required by ASTM D
abrasion loss lower than 20%. The mixture G2 met 7064. The G2 mixture showed better perform-
the specification with 5.5% and 6.0% percentages ance with TSR value of 93.1% and the mixture
of binder. The optimum binder content selected for G1 showed greater damage caused by moisture
G1 was 6.0% and for G2 was 5.5% based on the with a value of 89.2%.
criteria of abrasion loss.
3.6 Permeability
3.5 Moisture susceptibility The permeability measurements were made at
Figure 7 shows the indirect tensile strength three points of each slab (right edge, left edge and
for each mix (G1 and G2) in both dry and wet center). The test results of permeability using the
(moisture-conditioned) conditions. The minimum LCS permeameter are shown in Figure 9. From the
value for the indirect tensile strength is 0.55 MPa analysis of results it was found that the mixture G1
in accordance with DNER-ES 386/99 specifica- has a permeability coefficient greater than G2. The
tion to open mixtures in Brazil. Only G2 mixture presence of a higher air-void contents and poros-
reaches values above 0.55 MPa for both wet and ity in the mixture G1 allows passage of a greater
dry conditions, while G1 mixture showed values amount of water through interior mixture. The
below the minimum specified for the two condi- K values for two mixtures exceeded the minimum
tions (wet and dry). value of 0.12 cm/s suggested by NCAT (National

368

TIS2017_Book.indb 368 2/8/2017 8:04:52 PM

 absorption performance in the frequency range of
800 to 1000 Hz, where the noise tire/road is more
intense.

4 CONCLUSIONS

This paper presents a laboratory study about the

performance (permeability, noise reduction and
durability) of two small aggregate size open-graded
asphalt mixes (G1 and G2). From the results of
a series of tests, the following conclusions were
obtained:

Figure 9. Permeability of mixes measured by LCS • The two mixes G1 and G2 with optimum binder
permeameter. percentages of 6.0% and 5.5%, respectively, met
the specifications of an air-void content higher
than 18%, porosity greater than 10% and abra-
sion loss less than 20%.
• In relation to results of indirect tensile strength,
G1 didn`t reach the minimum resistance value
required of 0.55 MPa, while G2 exhibited sat-
isfactory indirect tensile strength values. The
results of TSR of the two mixes were higher
than 80%, meeting specification.
• The permeability of the mixture G1 was higher
than the permeability of the mixture G2, due to
the presence of a greater percentage of air-voids
and porosity. Both mixes showed a satisfactory
performance for the permeability test, present-
ing results greater than 0.12 cm/s.
Figure 10. Spectra of sound absorption coefficients of
• G1 and G2 mixes showed high values of sound
mixes G1 and G2.
absorption coefficient within the range of fre-
quencies 800 and 1000 Hz, where the intensity
Center of Asphalt Technology) for the new genera- of noise generated by the tire/pavement contact
tions of open mixtures (Alvarez et al. 2006). is critical.
The mixture G1 showed better performance
3.7 Acoustic absorption than the mixture G2 from the point of view of
functional properties (permeability and sound
Figure 10 shows the spectrum of sound absorp- absorption), but didn´t reach the minimum value
tion coefficient versus frequency for each speci- of indirect tensile strength and had high abrasion
men compacted in SGC with 5 cm height and loss.
10 cm diameter for each mixture (G1 and G2). It Finally, the mixture G2 (American experience)
was observed that the maximum absorption coef- met all limits and specifications for all tests show-
ficients are similar for the two mixtures, but the ing a satisfactory performance in relation to acous-
peak sound absorption of the mixture G1 (0.84) tic, permeability and durability properties.
is higher than that of the G2 mixture (0.78). This
can be explained by the greater air-void contents
and porosity in the mixture G1, which allow a
greater passage of sound wave into the same. The REFERENCES
frequencies where the peaks have been reported
(resonance frequencies) were in the range between Alvarez, A.E.; Martin, A.E.; Estakhri, C; Button, J.;
750 Hz and 1000 Hz. The results showed that the Glover, C.J.; Jung, S.H. 2006. Synthesis Of Current
Practice On The Design, Construction, And Mainte-
greater air-voids contents and porosity is related nance Of Porous Friction Courses. FHWA/TX-06/0–
with greater peak sound absorption coefficients. 5262–1. Texas Transportation Institute. Austin, Texas.
The mixture G1, which contains approximately ASTM D2041. 2011. Standart Test Method for Theoreti-
23% of air-void and 18% of porosity, is the best cal Maximum Specific Gravity and Density of Bitu-
sound attenuation material because showed better minous Paving Mixtures.

369

TIS2017_Book.indb 369 2/8/2017 8:04:53 PM

 ASTM D7064/D7064M – 08. 2008. Standard Practice for Lu, Q.; Fu, P.C.; Harvey, J.T. 2009. Laboratory Evalua-
Open-Graded Friction Course (OGFC) Mix Design. tion of the Noise and Durability Properties of Asphalt
Bernhard, R.J. & Wayson R.L. 2004. An introduction Surface Mixes. Research Report: UCPRC-RR-
to tire/pavement noise of asphalt pavement. Asphalt 2009–07 University of California Pavement Research
Pavement Alliance. USA. 27 p. Center UC Davis, UC Berkeley. EEUU.
DNER-ES 386/99: Pavimentação—Pré-misturado a Meurer Filho, E. 2001. Estudo de Granulometria
quente com asfalto polímero—camadaporosa de Para Concretos Asfálticos Drenantes. Dissertação
atrito. 1999. Departamento Nacional de Estradas de (Mestrado) – Programa de Pós-graduação PPGEC-
Rodagem. Rio de Janeiro, RJ. UFSC, Universidade Federal de Santa Catarina.
Gerges, S.N.Y. & Arenas, J.P. 2010. Fundamentos y Con- Florianópolis.
trol del Ruído y Vibraciones.2. ed. Florianopolis: Nr NLT-327/00. 2000. Permeabilidad in situ de pavimentos
Editora, 787 p. drenantes con el permeámetro LCS.
Guimaraes, J.M.F. 2012. Concreto Asfáltico Drenante Oliveira, C.G.M. 2003. Estudo de propriedades mecâni-
em Asfaltos modificados por polímero SBS e borracha cas e hidráulicas do concreto asfáltico drenante. Dis-
moída de pneus. Dissertação (Mestrado) – Programa sertação de mestrado em Geotecnia. Universidade de
de Pós-graduação PPGEC-UFSC, Universidade Fed- Brasília.
eral de Santa Catarina. Florianópolis. PIARC (World Road Association). 2013. Quiet Pave-
ISO 10534–2 E: 1998 Acoustics—Determination of ment Technologies. Technical Committee D.2 Road
sound absorption coefficient and impedance tubes— Pavements.
Part 2: Transfer-function method. Rasmussen, R.O.; Bernhard, R.J.; Sandberg, U.; MUN,
Li M.; Van Keulen, W.; Tijs, E.; Van De Ven, M. 2015. E.P. 2007. The Little Book of Quieter Pavements.
Molenaar, A. Sound absorption measurement of road FHWA, USA. FHWA—IF-08–004. 37 p.
surface with in situ technology. Applied Acoustics 88: Sandberg, U.; Ejsmont J. 2002. Tyre/road Noise Refer-
12–21. ence Book. Informex, Kisa, Sweden.
Liu, M.; Huang, X.; Xue, G. 2016. Effects of double layer WHO. 2011. Burden of disease from environmental
porous asphalt pavement of urban streets on noise noise—Quantification of healthy life years lost in
reduction. International Journal of Sustainable Built Europe. World Health Organization. – Regional office
Environment 5: 183–196. for Europe and EC-JRC.

370

TIS2017_Book.indb 370 2/8/2017 8:04:54 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Re-use in asphalt pavements of fillers from natural stone

sawmilling sludge

E. Santagata, O. Baglieri, L. Tsantilis, G. Chiappinelli & P.P. Riviera

Politecnico di Torino, Turin, Italy

ABSTRACT: The experimental investigation reported in the paper focused on the use in asphalt pave-
ments of fillers originated from natural Stone Sawmilling Sludge (SSS). Performance characteristics of
bituminous mastics and mixtures prepared with two types of SSS fillers, differing in mineralogical origin
of stones and in stone cutting techniques, were compared with those of reference mastics and mixtures
containing standard mineral filler. In the case of mastics, the testing program included the evaluation of
their rheological properties at different temperatures and loading conditions. In the case of corresponding
mixtures, laboratory characterization was based on the evaluation of their workability, volumetric charac-
teristics, stiffness, and rutting potential. Experimental results indicated that SSS fillers can be conveniently
used in asphalt pavements yielding satisfactory levels of performance.

1 INTRODUCTION The study presented in this paper focused on the

performance evaluation of two types of SSS, sup-
The construction of flexible pavements requires plied from two Italian quarries and derived from the
large volumes of aggregates to be employed cutting of natural stones, added as a substitution
in bitumen-bound layers (Zoorob & Suparma of the filler fraction in asphalt concrete mixtures.
2000), leading to a massive exploitation of natural Such an application has already been documented
resources and to significant environmental con- in literature, although limited information was
cerns (Drew et al. 2002). In such a context, re-use gathered with respect to potential field perform-
of alternative materials such as recycled aggregates ance (Chandra et al. 2002, Akbulut & Gurer 2007,
allows a serious waste management problem to be Karasahin & Terzi 2007, Choudary & Chandra
solved and simultaneously contributes to reduce 2008). Thus, this specific aspect was addressed in
the depletion to which available natural materials the investigation described in the following, which
are exposed. included rheological tests carried out on mastics in
The manufacturing process of natural ornamen- several aging conditions and performance-related
tal stones (including marble and granite) generates tests performed on bituminous mixtures.
huge quantities of sawmill sludge (Akbulut &
Gurer 2007, Ribeiro et al. 2008, Yilmaz et al. 2010).
This by-product derives essentially from the cutting 2 MATERIALS AND METHODS
phase to which stone is subjected and, once dried,
it is considerably fine (Sarkar et al. 2006). 2.1 Bituminous mastics
Due to its variable chemical composition and Bituminous mastics considered in the experimen-
particle size, Stone Sawmilling Sludge (SSS) can be tal investigation were prepared by using two SSS
used as aggregate replacement in several construc- fillers that were supplied from two different Ital-
tion materials, such as ceramic and clay composites ian quarries, and a standard calcareous filler (FR),
(Acchar et al. 2006, Saboya et al. 2007) and cement used for comparative purposes. The two SSS fill-
concrete (Hwang et al. 2008, Pereira et al. 2008, ers differed not only in terms of stone mineralogy,
Topcu et al. 2009, Yilmaz et al. 2010, Corinaldesi but also with respect to employed stone cutting
et al. 2010, Aliabdo et al. 2014). Furthermore, SSS techniques. In particular, one was obtained from
has also been considered as a possible stabilizer diamond sawing (FD), while the other one derived
of plastic soil since its use leads to a reduction from bucksaw blade operations (FT). Particle size
of plasticity index and optimal moisture content, distribution and bulk density (ρf) of the three fill-
with a corresponding increase of unit weight and ers are presented in Tables 1 and 2, respectively.
CBR index (Okagbue & Onyeobi 1999, Agrawal & A single 70/100 penetration grade Bitumen
Gupta 2011, Sivrikaya et al. 2014). (BB) was used for the preparation of mastics and

371

TIS2017_Book.indb 371 2/8/2017 8:04:56 PM

 Table 1. Particle size distribution of fillers. Table 3. Penetration and softening point of bitumen.

% Passing pen25 [dmm] TR&B [°C]

Sieve size
[mm] FT FD FR 83 45.6

1.000 100 100 100

0.500 100 96.0 100
0.250 100 90.0 99.0 Table 4. Calculated filler-to-bitumen ratios.
0.125 97.0 85.0 97.0
0.063 80.0 80.2 89.8 Mastic Filler (P0.063)F (P0.063)M BM
code code (%) (%) (%) RF/B

Table 2. Bulk density of fillers. MT FT 80.0 8.0 5.6 1.14

MD FD 80.2 8.0 5.6 1.14
FT FD FR MR FR 89.8 8.0 5.6 1.28

ρf [g/cm3] 2.72 2.88 2.72

• viscosity tests (UNI EN 13302-2010), with the

mixtures. It was subjected to basic characterization
Brookfield viscometer at temperatures com-
tests for the determination of penetration at 25°C
prised between 125 and 190°C, in order to gather
(pen25) and softening point temperature (TR&B, with
information on filler-bitumen affinity;
the ring and ball apparatus) according to standards
• multiple stress creep recovery tests (AASHTO
UNI EN 1426 (2007) and UNI EN 1427 (2007),
TP 70-2010), with the Dynamic Shear Rheom-
respectively. Results are reported in Table 3.
eter (DSR) on short-term aged mastics at high
Bituminous mastics were prepared in the
in-service temperatures, for the evaluation of
laboratory by adopting a predefined filler-to-
rutting resistance;
bitumen ratio (by weight) (RF/B). Target values
• flexural creep tests (AASHTO T313-2010),
of such a parameter were calculated by means of
with the Bending Beam Rheometer (BBR)
Equation 1:
on long-term aged mastics at low tempera-
tures, for the assessment of thermal cracking
(P ) ((P )
R = 0.063 M 0.063 F (1) resistance.
F/B B ⋅100
M

2.2 Bituminous mixtures

where (P0.063)M is the filler fraction dosage adopted
in the job mix formula of the bituminous mixtures Bituminous mixtures were obtained by combining
(set at 8%), (P0.063)F is the percentage of filler frac- four mineral aggregate fractions with the same fill-
tion passing through 0.063 mm sieve, and BM is the ers (FT, FD, FR) and Bitumen (BB) employed for
design bitumen content of the bituminous mix- the preparation of mastics. Their job mix formula,
tures (by weight of mineral aggregates). which was the same for all filler types, is shown in
Input data and filler-to-bitumen ratios obtained Table 5. Particle size distribution curves of the four
from Equation 1 are listed in Table 4, where aggregate fractions and of the target mixture are
codes used to identify mastics are also indicated. displayed in Figure 1.
On account of the negligible differences existing Preparation of the bituminous mixtures was per-
between the calculated RF/B values of the three formed at 165°C by means of a BBMAX80 mixer
materials, in the experimental investigation a single (certified standard NF P 98-250-1).
value was adopted for all mastics, fixed at 1.2. The investigation performed on the bituminous
Mastics were considered in their original state mixtures focused on the following performance-
and after Rolling Thin Film Oven (RTFO) and related characteristics:
Pressure Ageing Vessel (PAV) treatments, carried • workability and volumetric parameters deter-
out to simulate short-term and long-term aging mined from the analysis of gyratory compactor
conditions, respectively (AASHTO T 240-2009, densification curves and resulting specimens
AASHTO R 28-2009). (UNI EN 12697-31-2004UNI EN 12697-6-
Mastics were subjected to the following tests: 2008-Proc. B; UNI EN 12697-5-2010; UNI EN
• softening point tests (UNI EN 1427–2007), with 12697-8-2003);
the ring and ball apparatus, for the determina- • master curves of the complex modulus deter-
tion of filler “stiffening power”; mined from the results of triaxial tests performed

372

TIS2017_Book.indb 372 2/8/2017 8:04:56 PM

 Table 5. Job mix formula of bituminous mixtures. 5%, thus complying with the minimum requirement
that is indicated in Italian specifications in the case
Aggregate size fraction Mass percentage of filler-to-bitumen ratios of 1.5 (ANAS S.p.A,
2010). Moreover, it can be noticed that the two
Coarse gravel 10.0
SSS fillers exhibited a stiffening power which was
Medium gravel 29.0
Fine gravel 39.0
significantly higher than that recorded for the ref-
Sand 14.0 erence calcareous filler, with the greatest stiffening
Filler 8.0 potential associated to the use of the filler obtained
Bitumen 5.6 from bucksaw blade operations (mastic MT).

3.1.2 Viscosity tests

Viscosity tests were performed at several tempera-
tures and shear rates in order to yield information
on the affinity between filler and bitumen.
Results are synthetically presented in Figure 2,
which shows viscosity values obtained for the base
bitumen and the three mastics at all temperatures
adopted for testing (125, 135, 150, 165, 175, 190°C)
at a single reference value of shear rate, equal to
6.8s−1.
As expected, mastics showed viscosities which
were higher than those recorded for neat bitumen
in the entire range of investigated temperatures.
Coherently with the outputs of softening point
Figure 1. Particle size distribution curves of mineral tests, mastics containing the two SSS fillers yielded
aggregate fractions and target mixture. viscosity values which were significantly higher
than those obtained for the mastic prepared with
the reference calcareous filler. However, in contrast
with the asphalt mixture performance tester on
with the outcomes of softening point tests, the SSS
cores obtained from gyratory compactor speci-
filler which affected binder viscosity to the great-
mens (UNI EN 12697-31-2004; AASHTO TP
est extent was that obtained from diamond sawing
79-2010; AASHTO PP61-2010);
• semi-circular bending tests performed on por-
tions of gyratory compactor specimens at an
intermediate in-service temperature (20°C) for Table 6. Softening points of mastics (TR&B) and their
the evaluation of the potential for crack prop- increments with respect to bitumen (ΔTR&B).
agation (UNI EN 12697-31-2004; UNI EN Mastic code TR&B ΔTR&B
12697-44-2010); [°C] [°C]
• wheel tracking tests carried out on roller com-
pactor slabs at a high in-service temperature MT 63.4 17.8
(60°C) for the assessment of anti-rutting poten- MD 57.9 12.3
tial (UNI EN 12697-33-2007; UNI EN 12697- MR 54.0 8.4
22-2004).

3 ANALYSIS AND RESULTS

3.1 Bituminous mastics

3.1.1 Softening point tests
Stiffening power represents an important charac-
teristic that can be taken into account in order to
gain an insight into the role played by filler in the
mastic phase.
As shown in Table 6, this parameter was evalu-
ated by comparing the softening points of mastics
with that of base bitumen.
It can be observed that for all mastics relative Figure 2. Viscosity values of Base Bitumen (BB) and
increments of the softening point were greater than mastics (MR, MT, MD).

373

TIS2017_Book.indb 373 2/8/2017 8:04:56 PM

 (mastic MD), thus suggesting its greater affinity prepared with SSS fillers, suggested by softening
with bitumen. In any case, it should be empha- point tests. However, no significant differences
sized that such a difference is not surprising since were found when comparing the two mastics con-
the two test procedures differ significantly in terms taining different types of SSS filler.
of strain conditions imposed to the test specimens The same observations can be made when consid-
and explore different temperature ranges. ering elasticity values obtained at 0.1 and 3.2 kPa.
As shown in Figures 5 and 6, from the analysis of
3.1.3 Multiple stress creep recovery tests both R0.1 and R3.2 parameters, it can be inferred
Multiple stress creep recovery tests were carried that when compared to the reference mastic, mas-
out on RTFO-aged bitumen and mastics in order tics MT and MD exhibited a higher capability to
to obtain information on various aspects of their recover damage at all investigated temperatures.
response under repeated loading, including non- Outcomes of the above described MSCR tests
recoverable creep compliance (parameter Jnr) and are in line with the results of rheological tests
elasticity (R). reported by Choudhary and Chadra (2008) on
Values of non-recoverable creep compliance are mastics prepared with granite dust.
reported in Figures 3 and 4, where results obtained
at two different levels of stress applied in the creep 3.1.4 Flexural creep tests
phase, 0.1 kPa and 3.2 kPa, are presented at the The rheological response of mastics at low tem-
four investigated temperatures (58, 64, 70, 76°C). peratures was assessed by carrying out flexural
The stiffening effect of filler is clearly high- creep tests on PAV-aged materials at -12 and -18°C.
lighted by the substantial difference existing In particular, flexural creep stiffness (S) and creep
between the results obtained for neat bitumen and rate (m) were evaluated at a standard loading time
for the three mastics. Moreover, mastics contain- of 60 seconds, thus obtaining parameters S60 and
ing SSS filler (MT and MD) showed lower values m60 (Figures 7 and 8).
of non-recoverable creep compliance with respect As expected, the presence of filler led to an
to reference mastic MR, regardless of temperature increase of S60 and a decrease of m60, thus indi-
and stress level. This outcome seems to confirm cating a higher flexural stiffness and a lower
the superior anti-rutting performance of mastics relaxation capability of mastics with respect to

Figure 3. Non-recoverable creep compliance at 0.1 kPa

(Jnr0.1). Figure 5. Elasticity values at 0.1 kPa (R0.1).

Figure 4. Non-recoverable creep compliance at 3.2 kPa

(Jnr3.2). Figure 6. Elasticity values at 3.2 kPa (R3.2).

374

TIS2017_Book.indb 374 2/8/2017 8:04:56 PM

 Table 7. Compaction parameters.

Compaction parameter Reference value

Mold diameter 150 mm

Angle of inclination 1.25°
Speed of rotation 30 rpm
Vertical pressure 600 kPa
Mass of mixture into mold 4440.8 g
Number of gyrations 100
Temperature 155°C

Figure 7. S60 values at −12 and −18°C. Table 8. Workability parameters.

Parameter CT CD CR

C1 (%) 76.4 76.6 76.4

K 9.2 9.5 9.0

Obtained results clearly indicate that the three

materials exhibit a quite similar behavior in terms
of both self-compaction and workability. With
respect to workability, only a slight increase of k
can be observed for mixtures CT and CD, in con-
trast with results obtained on mastics in terms of
stiffening power and viscosity.
Figure 8. m60 values at −12 and −18°C.
Volumetric tests were performed to determine
maximum density ρmc on loose materials, bulk
bitumen. These effects were slightly more pro-
density ρ (saturated surface dry) of compacted
nounced in the case of SSS fillers, thus suggesting
specimens, and void characteristics at 100 and
that their use can lead to a reduction of thermal
10 gyrations, V100 and V10, respectively. Average
cracking resistance.
results of three determinations are listed in Table 9.
Data listed in Table 9 indicate that the influ-
3.2 Bituminous mixtures ence of filler type on the final volumetric proper-
3.2.1 Workability and volumetric characteristics ties of mixtures is quite limited. However, it can be
In order to gather information on the workability noticed that, in agreement with the information on
of bituminous mixtures, cylindrical specimens with workability, the mixture containing the reference
a diameter of 150 mm were prepared by means of filler exhibited the highest V100 value and the low-
a gyratory compactor. A summary of the main est V10 value.
compaction parameters is presented in Table 7.
The effect of filler type on the compaction prop- 3.2.2 Complex modulus master curves
erties of the mixtures was evaluated by referring to Mechanical characteristics of bituminous mixtures
densification curves that were modelled according were assessed via dynamic tests performed in the
to the following Equation (2): triaxial configuration by means of the Asphalt
Mixture Performance Tester (AMPT). For the
C C +k
1
( )
g Nx (2) determination of the complex modulus (E*), tests
were carried out by applying a sinusoidal load at
frequencies comprised between 0.01 and 25 Hz at
where C is the compaction level reached after Nx three test temperatures (4, 20 and 40°C).
gyrations, C1 is the initial value of compaction Test specimens were prepared by means of the
(self-compaction) and k is the slope of the curve, gyratory compactor (diameter 150 mm, height
that is related to the workability of the mixture. 170 mm) and then cored and cut by sawing to
Results collected for the three mixtures pre- obtain diameters and heights of 100 and 150 mm,
pared with fillers FT, FD and FR, marked in the respectively. The mass of the mixture used for
study as CT, CD and CR, respectively, are synthe- compaction was specifically selected to obtain geo-
sized in Table 8, where each value is the average metrical percentages of voids equal to those of the
of three runs. specimens compacted at 100 gyrations (Table 9).

375

TIS2017_Book.indb 375 2/8/2017 8:04:57 PM

 Master curves of the complex modulus (E*) ver- fracture toughness (KIC) determined from experi-
sus reduced frequency (fr) were built at the refer- mental data are summarized in Table 10.
ence temperature of 20°C by modelling raw data Fracture toughness values indicate that the
according to the Hirsch model (Christensen et al., mixture containing the calcareous reference filler
2003). Results obtained for the three mixtures are exhibited the highest resistance to crack propaga-
presented in Figure 9. tion. By contrast, values of the strain at maximum
From the analysis of master curves obtained for force seem to indicate that materials CT and CD
mixtures CT, CD and CR it can be observed that showed a more ductile behavior than reference mix-
in the high frequency (i.e. low temperature) range ture CR. This last result is consistent with those of
the three curves follow the same trend, thus reveal- fatigue tests performed by Chandra et al. (2002) on
ing a negligible influence of filler type. However, in bituminous mixtures containing marble dust.
the low frequency (i.e. high temperature) range the
three mixtures exhibited a different behavior under 3.2.4 Wheel tracking tests
repeated loading. In particular, mixture CD exhib- The resistance of bituminous mixtures to perma-
ited the highest complex modulus values, followed nent deformation under loading was evaluated by
by CT and CR. This finding, that can be related means of wheel tracking tests performed at 60°C.
to the volumetric peculiarities of each mixture and Rectangular slabs were produced with a roller
to the intrinsic characteristics of their mastics, is compactor by imposing a target geometric den-
in good agreement with the outcomes of multiple sity equal to the geometric density obtained at
stress creep recovery tests. Hence, obtained master 100 gyrations on cylindrical specimens.
curves suggest that the two bituminous mixtures Results of wheel tracking tests are graphi-
containing SSS fillers can provide better anti- cally presented in Figure 10, where rut depth
rutting performance with respect to the reference (εr), expressed in percentage, is plotted versus the
material. number of wheel passes (N). Table 11 contains the
average values (from 2 repetitions) of the maxi-
3.2.3 Semi-circular bending tests mum permanent deformation (εpmax) and of the
Semi-Circular Bending (SCB) tests were carried slope of rutting curves (ΔPi/ΔlogN).
out at 20°C to assess the potential for crack propa- Slope values of rutting curves shown in
gation of bituminous mixtures. Table 11 suggest that mixtures containing the ref-
A half cylinder test piece (diameter 150 mm, erence calcareous filler (CR) and the filler derived
height 50 mm) with a center notch of 10 mm from bucksaw blade operations (CT) exhibit a
was loaded in the three-point bending configura- better anti-rutting behavior. Moreover, it can be
tion by imposing a constant deformation rate of observed that the two mixtures containing SSS
5 mm/min until failure. Average values (from six filler lead to the highest values of maximum per-
replicates) of strain at maximum force (εmax) and manent deformation. These findings are in partial

Table 9. Volumetric properties of mixtures.

Table 10. Results of SCB tests.
CT CD CR
CT CD CR
ρmc [kg/m3] 2550 2538 2577
ρ [kg/m3] 2414 2425 2527 εmax (%) 2.10 1.93 1.83
V100 (%) 5.3 4.5 5.8 KIC [N/mm3/2] 10.3 10.2 13.1
V10 (%) 14.4 14.0 13.5

Figure 9. Master curves of the complex modulus E*. Figure 10. Results of wheel tracking tests.

376

TIS2017_Book.indb 376 2/8/2017 8:04:57 PM

 Table 11. Permanent deformation parameters derived AASHTO R 28-2009. Accelerated aging of asphalt
from wheel tracking tests. binder using a pressurized aging vessel.
AASHTO T 240-2009. Effect of heat and air on a moving
CT CD CR film of asphalt binder (Rolling Thin Film Oven Test).
AASHTO T 313-2010. Determining the flexural creep
ΔPi/ΔlogN [-] 0.11 0.18 0.11 stiffness of asphalt binder using the Bending Beam
εpmax (%) 7.34 5.76 4.93 Rheometer (BBR).
AASHTO TP 70–2010. Multiple Stress Creep Recovery
(MSCR) test of asphalt binder using a dynamic shear
rheometer.
disagreement with the outcomes of multiple stress AASHTO TP 79–2010. Standard method of test for
creep recovery tests performed on mastics and of determining the dynamic modulus and flow number
dynamic tests carried out in the triaxial configura- for asphalt mixtures using the Asphalt Mixture Per-
tion on mixtures. However, it should be considered formance Tester (AMPT).
that notwithstanding their simulative character, Acchar W., Vieira F.A., Hotza D. 2006. Effect of marble
and granite sludge in clay materials. Materials Science
wheel tracking tests are typically associated to a and Engineering A 419: 306–309.
greater dispersion of results. These may be sig- Agrawal V. & Gupta M. 2011. Expansive soils stabiliza-
nificantly affected, especially in terms of maximum tion using marble dust. International Journal of Earth
deformation, by local biasing effects related to the Sciences and Engineering 4: 59–66.
initial contact between test slab and loading wheel. Akbulut H. & Gurer C. 2007. Use of aggregates pro-
duced from marble quarry waste in asphalt pavements.
Building and Environment 42: 1921–1930.
4 CONCLUSIONS Aliabdo A.A. Elmoaty. A.E.M.A., Auda E.M. 2014. Re-use
of waste marble dust in the production of cement and
concrete. Construction and Building Material 50: 28–41.
Based on the experimental results presented in ANAS S.p.A. 2010. Capitolato special d’appalto—
this paper, it can be concluded that fillers retrieved Norme tecniche.
from Stone Sawmilling Sludge (SSS), obtained Chandra S., Kumar P., Feyissa B.A. 2002. Use of marble
from both diamond and bucksaw blade opera- dust in road construction. Road Materials and Pave-
tions, can be used for the preparation of bitumi- ment Design 3(3): 317–330.
nous mixtures for road paving applications. Choudhary R. & Chadra S. 2008. Granite and marble
With respect to their anti-rutting potential, it dusts as filler in asphalt concrete. Proceedings of the
was found that the use of SSS fillers can provide First International Conference on Transport Infra-
improvements in the performance properties of structures, Beijing (China), 24–26 April 2008.
Christensen, D.W., Pellinen T.K. & Bonaquist R.F. 2003.
mixtures at high temperatures, as indicated by sev- Hirsch model for estimating the modulus of asphalt
eral performance-related parameters analyzed at concrete. Journal of the Association of Asphalt Pav-
both the mastic and mixture scale. On the other ing Technologists, vol 72, pp 97–121.
hand, the use of these alternative fillers may not Corinaldesi V., Moriconi G., Naik T.R. 2010. Characteriza-
provide beneficial effects with respect to the resist- tion of marble powder for its use in mortar and concrete.
ance to crack-related distresses. Thus, use of SSS Construction and Building Material 24: 113–117.
fillers should be carefully considered depending Drew L.J., Langer W.H., Sach J.S. 2002. Environmental-
upon environmental and traffic scenarios. ism and natural aggregate mining. Natural Resources
Further research should be performed in order Research 11(1): 19–29.
Hwang E.H., Ko Y.S., Jeon J.K. 2008. Effect of polymer
to validate the conclusions of this preliminary cement modifiers on mechanical and physical proper-
investigation by taking into account a wider array ties of polymer-modified mortar using recycled artifi-
of base materials. cial marble waste fine aggregate. Journal of Industrial
and Engineering Chemistry 14: 265–271.
Karasahin M. & Terzi M. 2007. Evaluation of marble
ACKNOWLEDGEMENTS waste dust in the mixture of asphaltic concrete. Con-
struction and Building Material 21: 616–620.
The study reported in this paper was funded by the Okagbue C.O. & Onyeobi T.U.S. 1999. Potential of mar-
CCIAA (Chamber of Commerce, Industry, Crafts ble dust to stabilise red tropical soils for road con-
struction. Engineering Geology 53: 371–380.
and Agriculture) of Verbano Cusio Ossola. Pereira F.R., Ball R.J., Rocha J., Labrincha J.A., Allen G.C.
2008. New waste based clinkers: belite and lime formu-
lations. Cement and Concrete Research 38: 511–521.
REFERENCES Ribeiro R.C.C., Correia J.C.G., Seidl P.R. 2008. Use of
aggregates produced from marble and granite quarry
AASHTO PP 61–2010. Standard practice for developing waste in asphalt pavements: a form of clean technol-
dynamic modulus master curves for Hot Mix Asphalt ogy. Proceedings of the 2008 global symposium on
(HMA) using the Asphalt Mixture Performance recycling, waste treatment and clean technology, Can-
Tester (AMPT). cun (Mexico), 12–15 October 2008.

377

TIS2017_Book.indb 377 2/8/2017 8:04:57 PM

 Saboya F., Xavier G.C., Alexandre J. 2007. The use of the UNI EN 12697-44-2010. Bituminous mixtures—Test
powder marble by-product to enhance the properties methods for hot mix asphalt—Part 44: Crack propa-
of brick ceramic. Construction and Building Material gation by semi-circular bending test.
21: 1950–1960. UNI EN 12697-5-2010. Bituminous mixtures—Test
Sarkar R., Das S.K., Mandal P.K., Maiti H.S. 2006. methods for hot mix asphalt—Part 5: Determination
Phase and microstructure evolution during hydrother- of the maximum density.
mal solidification of clay-quartz mixture with marble UNI EN 12697–6-2008 Bituminous mixtures—Test
dust source of reactive lime. Journal of the European methods for hot mix asphalt—Part 6: Determination
Ceramic Society 26: 297–304. of bulk density of bituminous specimens.
Sernas O., Strumskys M., Skrodenis D. 2014. Possibility of UNI EN 12697-8-2003. Bituminous mixtures—Test
use of granite fines in asphalt pavements. Proceedings methods for hot mix asphalt—Part 8: Determination
of the 9th International Conference “Environmental of void characteristics of bituminous specimens.
Engineering”, Vilnius (Lithuania), 22–23 May 2014. UNI EN 13302-2010. Bitumen and bituminous
Sivrikaya O., Kiyildi K.R., Karaca Z. 2014. Recycling binders—Determination of dynamic viscosity of bitu-
waste from natural stone processing plants to sta- minous binder using a rotating spindle apparatus.Yil-
biliseclayey soil. Environmental Earth Sciences 71: maz H., Guru M., Dayi M., Tekin I. 2010. Utilization
4397–4407. of waste marble dust as an additive in cement produc-
Topcu I.B., Bilir T., Uygunoglu T. 2009. Effect of waste tion. Material and Design 31: 4039–4042.
marble dust content as filler on properties of self- UNI EN 1426-2007. Bitumen and bituminous binders—
compacting concrete. Construction and Building Determination of needle penetration.
Material 23: 1947–1953. UNI EN 1427-2007. Bitumen and bituminous binders—
UNI EN 12697-22-2004. Bituminous mixtures— Determination of the softening point—Ring and Ball
Test methods for hot mix asphalt—Part 22: Wheel method.
tracking. Zoorob S.E. & Suparma L.B. 2000. Laboratory design
UNI EN 12697-31-2004. Bituminous mixtures—Test and investigation of the properties of continuously
methods for hot mix asphalt—Part 31: Specimen graded asphaltic concrete containing recycled plas-
preparation by gyratory compactor. tics aggregate replacement (Plastiphalt). Cement
UNI EN 12697-33-2007. Bituminous mixtures—Test and Concrete Composites 22(4): 233–242.Clayey
methods for hot mix asphalt—Part 33: Specimen pre- Soil”, Environmental Earth Sciences 71, 4397–
pared by roller compactor. 4407, 2014.

378

TIS2017_Book.indb 378 2/8/2017 8:04:57 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Risk compensation in a changing road environment

A. Oikonomou, P. Tafidis, P. Kyriakidis, S. Basbas & I. Politis

Aristotle University of Thessaloniki, Thessaloniki, Greece

ABSTRACT: The objective of this paper is to identify risk compensation effect i.e. the tendency
of drivers to adjust their behavior in changing conditions of the road environment. For this reason a
questionnaire survey was conducted in order to find out whether the effect takes place as a result of
changes in the road safety levels and to address any possible connection between aggressive driving and
driver behavioral adaptation. The survey was conducted after the beginning of road works in Egnatias’
Street in Thessaloniki. The analysis revealed that males comprise the majority of drivers not to adapt their
driving behavior while being the more aggressive at the same time.

1 INTRODUCTION exposed to higher levels of risk. Traynor (1993)

concluded that drivers react to changes in safety
According to World Health Organization (2009) levels by a variable which measures their aggres-
road accidents is one of the biggest problems glo- siveness in driving and another one dependent on
bally and the main reason for the death of children dangerous driving situations.
aged between 10 and 19 (WHO 2008). The urge to Aggressive driving behavior is a key factor in
deal with the problem in a more sufficient man- reducing road safety. Aggressive driving is usu-
ner has led to the study of drivers’ psychological ally characterized by the influence of disorders
aspects, that is, research and interpretation of their and leads the driver to impose its own desired level
driving behavior. Risk compensation, the tendency of risk on others. The three main categories of
of drivers to adjust their driving behavior when emotional disorders which affect driving behavior
they perceive changes in the road environment, the according to James and Nahl (2000) are impatience
vehicle or themselves, is among the most popular and inattentiveness, power struggle, recklessness
explanations about drivers’ behavior. and road rage.
The notion of risk compensation was intro- Typical traffic violations by aggressive drivers
duced in early 70’s and was highlighted as one of are: red and yellow light running, ignoring stop
the most prominent parameters to affect driver signs, excessive lane changing, speeding, tailgating,
behavior (Summala 1996). Lave and Weber (1970) blocking intersections, driving while drunk and
and Peltzman (1975) proposed risk compensa- bad behavior.
tion as an answer to the technological approach According to Vaa (2007) humans like every
in road safety research. Peltzman (1975) intro- other living organism display the tendency to adapt
duced the term as the drivers’ tendency to take to the constant environmental changes. Road envi-
higher risks whenever safety conditions improve, ronment is a special case of an environment that
the result being having more accidents despite the demands the driver’s constant adaptation so acci-
improvements in road safety. Naatanen and Sum- dents are avoided. Risk compensation effect also
mala (1974) pointed out that when it comes to the represents a special case of driver’s behavioral
driver safety it is not skill but the driver’s preferred adaptation and concerns involuntary decisions
driving pace that matters. Risk compensation as a made by the drivers driving an ABS equipped car
term is getting more common to indicate changes in situations where the distances between vehicles
in driver’s behavior towards road safety improve- are getting shorter (Sagberg et. al 1997). This proc-
ments (Levym & Miller 2000). ess is risk compensation because despite the intro-
A confirmation of risk compensation theory duction of a system that reduces road risk, drivers
comes from a study from McCarthy (1986), where compensate by choosing to drive faster or closer to
it is demonstrated that the drivers’ decision to use the car ahead.
a safety belt resulted in them undertaking higher Studying the effectiveness of the enforcement
risks while driving. Singh and Thayer (1992) found in the use of seat belts in New Jersey, Asch et al.
out that the risk compensation effect is more likely (1991) noticed that while road fatalities were
to occur to drivers that are not against getting reduced by 18.9%, the overall number of road

379

TIS2017_Book.indb 379 2/8/2017 8:05:00 PM

 accidents increased by 16.5%. Assum et al. (199), third section concerns the detection of drivers who
in their study about risk compensation in regard adopt an aggressive stance when they deal with traf-
with road lighting, found out that when road light- fic conditions that occur as a result of the ongoing
ing is sufficient drivers tend to drive faster, some- works. In particular whether they are speeding at
times even faster than during daylight, while when the affected area, yellow light running at intersec-
lighting is insufficient or non-existent they tend to tions with limited visibility, their stance towards
driver more carefully. unruly pedestrian behavior, maneuvering in nar-
The direct observation and accurate estimate rowing road sections, entering congested intersec-
of risk compensation is difficult due to the limited tions while vehicles ahead are already trapped and
capacity to instantly monitor drivers’ actions when their stance towards buses that are getting to move
they perceive changes on road safety levels. In this again after a stop. The questionnaire survey con-
study risk compensation is detected indirectly, sisted of one hundred (100) interviews. The inter-
through the actions drivers stated they perform in viewees were drivers that had just parked their cars
a changing road environment. The approach is to or were about to leave or waiting inside their cars
compare drivers’ stated behavior before and after at the parking sites. The questionnaire items of the
the changes and to examine possible factors affect- survey were coded into specific variables (Table 1
ing the results. Finally, it is important to high- and 2).
light that risk is not the only parameter affecting
driving behavior, although its better observation
2.2 Descriptive and inferential statistics
and understanding could lead to a more efficient
implementation of road safety measures. Variables of the survey were processed through
IBM SPSS 21. Table 3 and 4 presents descriptive
statistics for the variables along with normality
2 METHODOLOGY tests for skewness and kurtosis. In a reliability test
for internal consistency the group of variables con-
2.1 Questionnaire survey cerning drivers’ behavioral adaptation has a Cron-
bach’s Alpha of .825, while the group of variables
A questionnaire survey was conducted in order
concerning aggressive driving .633. In table 2 it is
to study whether risk compensation effect mani-
shown that the sample consists mostly by males.
fests as a result of changes in road safety levels
Most prevalent age group is 25–54. On average
and in order to investigate the relation between
they drive 5 to 6 days per week, 2.5 hours a day
aggressive driving and drivers’ behavioral adapta-
with cars of 1468 cc. Most usual trip purpose is
tion. The area of the study was Egnatias Street in
commuting. We can notice the variable AggrDr1
Thessaloniki, Greece, where the subway is being
for speeding through the works-affected area has a
constructed. The street is one of the major streets
of the city’s road network and an area of mixed
land uses like residence, commerce, city services, Table 1. Variables.
urban green etc. The reason of this choice was the
Variables Description
observed worsening road safety levels that increase
accident risk as a side effect of the ongoing works. Gender Interviewee gender
Drivers were interviewed in two open-air park- Age Group Interviewee age
ing sites nearby (Ancient Agora Square, Fillipou DrivYrs Interviewee driving years
Street), securing that the drivers’ majority were fre- DaysWeek Days per week driving
quent users of Egnatias Street. HrsDay Hours per day driving
The questionnaire comprises of three sections. carCC Engine displacement in cc3
In the first section there are general question items TripPurpose Interviewee trip purpose
regarding the driver profile and trip habits. These Adapt1 Adaptation due to lane curves
are gender, age group, overall driving years, driv- Adapt2 Adaptation due to limited visibility
ing days per week, driving hours per day and most Adapt3 Adaptation due to narrower traffic lanes
usual trip purpose. The second section concerns Adapt4 Adaptation due to pedestrian crossings
the adaptation of drivers’ behavior. The question- Adapt5 Adaptation due to fragmented bus lanes
naire items are about how much affected were the Adapt6 Adaptation due to lane poles
drivers by the special traffic conditions the ongo- AggrDr1 Speeding through the affected area
ing works brought about. In particular lane curves, AggrDr2 Behavior at intersections
limited visibility at intersections, locally narrowed AggrDr3 Behavior towards pedestrian crossings
street lanes, increased incidents of unruly pedes- AggrDr4 Maneuvering in narrowing traffic lanes
trian crossing behavior, fragmented bus lanes and AggrDr5 Behavior at congested intersections
sporadic traffic pole roadways separation. The AggrDr6 Behavior towards buses

380

TIS2017_Book.indb 380 2/8/2017 8:05:00 PM

 Table 2. Scale of measure. Table 3. Descriptive statistics.

Variables Scale of measure Variables Σ min max Mean StdDev

Gender 0: Male 1: Female Gender 100 0 1 0,30 0,461

Age Group 0: 18–24 1: 25–54, 2: 55–64 3: 65+ Age Group 100 0 3 0,90 0,628
DrivYrs 0: 0–5 1: 6+ DrivYrs 100 0 1 0,63 0,485
DaysWeek Open type DaysWeek 100 1 7 5,69 1,710
HrsDay Open type HrsDay 100 0,5 8 2,30 1,573
carCC Open type CarCC 100 1000 2500 1468 306,456
TripPurpose 0 = Commuting 1 = Professional TripPurpose 100 0 6 1,33 1,589
2 = Education, 3 = Personal, Adapt1 100 0 1 0,50 0,503
4 = Recreational 5 = Shopping Adapt2 100 0 1 0,27 0,446
6 = Other Adapt3 100 0 1 0,52 0,502
Adapt1 0: Permanent adaptation Adapt4 100 0 1 0,29 0,456
1: Non-permanent adaptation Adapt5 100 0 1 0,44 0,499
Adapt2 0: Permanent adaptation Adapt6 100 0 1 0,54 0,501
1: Non-permanent adaptation AggrDr1 100 0 1 0,04 0,199
Adapt3 0: Permanent adaptation AggrDr2 100 0 1 0,20 0,401
1: Non-permanent adaptation AggrDr3 100 0 1 0,70 0,261
Adapt4 0: Permanent adaptation AggrDr4 100 0 1 0,27 0,446
1: Non-permanent adaptation AggrDr5 100 0 1 0,26 0,440
Adapt5 0: Permanent adaptation AggrDr6 100 0 1 0,39 0,491
1: Non-permanent adaptation
Adapt6 0: Permanent adaptation
1: Non-permanent adaptation Table 4. Normality tests.
AggrDr1 0: Non-aggressive driver
Skewness Kurtosis
1: Aggressive driver
AggrDr2 0: Non-aggressive driver Std. Std.
1: Aggressive driver Variables Stat. Error Stat. Error
AggrDr3 0: Non-aggressive driver
1: Aggressive driver Gender 0,886 0,241 −1,240 0,478
AggrDr4 0: Non-aggressive driver Age Group 0,825 0,241 2,548 0,478
1: Aggressive driver DrivYrs −0,547 0,241 −1,736 0,478
AggrDr5 0: Non-aggressive driver DaysWeek −1,223 0,241 0,534 0,478
1: Aggressive driver HrsDay 1996 0,241 4,285 0,478
AggrDr6 0: Non-aggressive driver carCC 0,634 0,241 0,053 0,478
1: Aggressive driver TripPurpose 0,718 0,241 −1,070 0,478
Adapt1 0,000 0,241 −2,041 0,478
Adapt2 1,052 0,241 −0,912 0,478
big kurtosis with a value of 20.641. Also AggrDr3 Adapt3 −0,081 0,241 −2,034 0,478
for aggressive driving towards unruly pedestrian Adapt4 0,940 0,241 −1,140 0,478
crossings has a kurtosis value of 9.334. Adapt5 0,245 0,241 −1,980 0,478
Adapt6 −0,163 0,241 −2,014 0,478
2.3 Analysis AggrDr1 4,714 0,244 20,641 0,483
AggrDr2 1,541 0,246 0,381 0,488
Figure 1 presents the relation between gender AggrDr3 3,338 0,246 9,334 0,488
and driver behavioral adaptation for the six indi- AggrDr4 1,060 0,250 −0,896 0,495
vidual changes in the road environment that were AggrDr5 1,125 0,245 −0,750 0,485
included in the study. As it is seen there is a sta- AggrDr6 0,451 0,245 −1,835 0,485
tistically significant difference between males and
females, with males appearing not to adjust their
driving behavior as a result of the changes in the the driver inexperience (up to 5 years of driving) or
road environment. experience (more than 6 years of driving). Engine
Another parameter that seems to have sta- displacement seems to have a significant impact in
tistically significant relation with the preserva- driver behavioral adaptation. In Figure 3 we notice
tion of driving behavior is the driver experience. that driver behavioral adaptation due to frag-
Figure 2 shows the differentiation in driver behav- mented bus lanes tends to remain unchanged the
ioral adaptation (cases 1, 5 and 6) depending on more the engine displacement increases. The same

381

TIS2017_Book.indb 381 2/8/2017 8:05:00 PM

 Figure 1. Driver behavioral adaptation by gender.

Figure 4. Driver behavioral adaptation due to traffic

poles.

Figure 2. Driver behavioral adaptation by driving years.

Figure 5. Aggressive driving by gender.

tendency is present in the case of traffic poles road-

ways separation (Figure 4).
Figure 3. Driver behavioral adaptation due to frag- As it was expected males appear to drive more
mented bus lanes. aggressively than females. Figure 5 shows the two

382

TIS2017_Book.indb 382 2/8/2017 8:05:00 PM

 for the drivers in Thessaloniki that need to drive
through Egnatias street where the ongoing subway
works alter the road conditions frequently. From
the analysis it is derived that the effect is weaker
in males and in those who drive more aggressively
in general. However it is necessary to emphasize
that to study risk compensation effectively would
require observing the actual driver behavior over a
long time, a thing not possible in the present study
where the effect was identified

REFERENCES

Asch, P., Levym, D.T., Shea, D. & Bodenhorn, H. 1991.

Risk compensation and the effectiveness of safety belt
use laws: A case study of New Jersey. Policy Sciences
24(2): 181–197.
Assum, T., Bjernskau, T., Fosser, S. & Sagberg, F. Risk
compensation - The case of road lighting. Accident
Analysis and Prevention 31(5): 545–533.
James, L. & Nahl, D. 2000. Aggressive driving is
emotionally impaired driving. University of Hawaii.
Lave, L. B. & Weber, W. E. 1970. A Benefit-Cost Analy-
Figure 6. Adaptation and aggressive driving due to bus sis of Auto Safety Features. Applied Economics 2(4):
lanes. 265–275.
Levym, D.T. & Miller, T. 2000. Risk Compensation
Literature - The Theory and Evidence. Journal of
out of the overall six instances in the survey with Crash Prevention and Injury Control 6(2): 82–89.
a differentiation by gender that is statistically sig- McCarthy, P. 1986. Seat belt usage rates: A test of
nificant. These are yellow light running at intersec- Peltzman’s hypothesis. Accident Analysis and
Prevention18(5): 425–438.
tions of limited visibility (AggrDr2) and entering Naatanen, R. & Summala, H. 1974. A model for the role
congested intersections while vehicles ahead are of motivational factors in drivers’ decision-making.
jammed (AggrDr5). Accident Analysis and Prevention 6(3–4): 243–261.
Figure 6 shows that the majority of those Peltzman, S. 1975. The effects of automobile safety regu-
who adapt their driving behavior when they deal lation. Journal of Political Economy 83: 677–725.
with fragmented bus lanes drive more cautiously Sagberg, F., Fosser, S., Sætermo, I.V. 1997. An investiga-
towards a bus that is about to rejoin traffic after tion of behavioural adaptation to airbags and antilock
a stop. Similarly, the majority of drivers who do brakes among taxi drivers. Accident Analysis and
not adjust their driving when they deal with frag- Prevention 29(3): 293–302.
Singh, H. & Thayer, M. 1992. Impact of seat belt use on
mented bus lanes act more aggressively towards driving behavior. Economic Inquiry 30(4): 649–658.
buses that rejoin traffic after a stop. Summala, H. 1996. Accident risk and driver behavior.
Safety Science 22(1–3): 103–117.
Traynor, T. 1993. The Peltzman hypothesis revisited:
3 CONCLUSIONS An isolated evaluation of offsetting driver behavior.
Journal of Risk and Uncertainty 7(2): 237–247.
Risk compensation effect was studied extensively in Vaa, T. Modelling Driver Behavior on Basis of Emotions
the past mostly when safety technologies, that were and Feelings: Intelligent Transport Systems and Behav-
new at the time, were introduced. Subsequently, risk ioural Adaptations. In C. Cacciabue (ed), Modelling
Driver Behaviour in Automotive Enviroments: 208–232.
compensation does not get much attention despite London: Springer.
being present in drivers either when new technology World Health Organization 2009. Global Status Report
innovations are introduced or like in this study the on Road Safety: Time for Action.
road environment changes. Analysis in the present World Health Organization 2008. The Global Burden of
study reveals that risk compensation effect is active Disease: 2004 Update. Geneva.

383

TIS2017_Book.indb 383 2/8/2017 8:05:02 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A PSO algorithm for designing 3D highway alignments adopting

polynomial solutions

G. Bosurgi, O. Pellegrino & G. Sollazzo

Department of Engineering, University of Messina, Messina, Italy

ABSTRACT: Intelligent optimization algorithms for highway alignments have produced good results
so far. However, considering the numerous constraints and factors directly implied in the infrastructure
design, the researchers’ efforts usually focus only on simplifying the alignment choice, supporting
engineers in the design phase. Implementing strategic considerations regarding comfort and safety
would be also very important. In this paper, the authors propose a method for designing improved 3D
highway alignments using a specific optimization algorithm, based on a Swarm Intelligence technique,
adopting an innovative polynomial transition curve as the unique horizontal curvature element, called
PPC (Polynomial Parametric Curve). This geometric solution assures higher levels of comforts for users
than the traditional ones (clothoid—circular curve—clothoid), because the PPC shows more gradual
trends of the main dynamic variables involved while driving, and defines each whole curve through a
unique element, simplifying the design procedure. The authors provide technical and operational details
for improving a Swarm optimization model through the adoption of the PPC and prove the efficacy of
the proposed procedure through a specific significant example.

1 INTRODUCTION transition curves, eliminates abrupt variations of

driving behaviour. This also guarantees to handle
In recent years, various researchers have continu- in a proper way the dynamic variables that influ-
ally developed intelligent techniques for automati- ence safety and comfort. For this reason, differ-
cally defining road alignments and optimizing ent kinds of polynomial curves and spirals have
their geometric design. Generally, the research- been studied and the related results seem very
ers’ aim is to help designers in the choice of the significant.
optimal alignment in a specific study area and in Therefore, the authors propose a different and
compliance with some of the several constraints. improved optimization algorithm for defining 3D
Each specific design requires a very deep examina- road alignments. This procedure not only takes
tion of the area characteristics and of the related into consideration several kinds of constraints
constraints. Then, many innovative tools and algo- (economical, geometric, environmental, etc.), but
rithms have been produced during the last years, draws a road with higher levels of users’ comfort,
based on various and effective optimization tech- adopting the PPC (Polynomial Parametric Curve)
niques. However, there have been no attempts to as the unique curve element. The specific math-
define and propose innovative optimization algo- ematical formulation of this curve guarantees
rithms producing advanced alignments, which can a smoother and more gradual variation of the
improve comfort levels for drivers. All the existing main variables generally considered for checking
optimization methods consider traditional solu- comfort along the roads (Bosurgi and D’Andrea,
tions in the design of the highway alignments. 2012; Bosurgi et al., 2015). This paper suggests
Traditionally, the clothoid is assumed as the technical and operational solutions for introduc-
geometric element for the transition curves. ing the PPC in an existing Artificial Intelligence
Considering the increasing levels of traffic and (AI) algorithm, based on the Particle Swarm Opti-
vehicle speed, recent studies based on the evalua- mization (PSO) method, for optimizing highway
tion of numerical tests and practical experiments 3D alignments and minimizing total costs. Then,
have proposed the adoption of different curves, the algorithm designs alignments made up of a
assuring a smoother gradation of the steering composition of tangent sections and PPC on the
manoeuvre in accordance with actual drivers’ horizontal plane, while a succession of tangent
behaviour. A proper composition of the different sections and parabolic curves describes the verti-
geometric elements forming roads, considering cal profile of the road.

385

TIS2017_Book.indb 385 2/8/2017 8:05:04 PM

 2 ALIGNMENT OPTIMIZATION calculate the position x and the speed v after each
METHODS iteration are very simple:

The idea of optimizing the design of road align- xij (t ) = xij (t ) + vij (t
(t ) (1)
ments showed a slow but constant development
after 1970. The most famous approaches exploit vij (t ) = wijvij (t ) c1r1 ⎡⎣ gb j (t(t ) xij (t )⎤⎦
modern AI techniques and, in particular, the + c2 r2 ⎡⎣ pb,iij (t ) − xij ( )⎤⎦ (2)
Genetic Algorithms (GA), derived from Darwin’s
theory on genetic evolution (Jong and Schonfeld, ⎧vij (t ) if vij (t ) ≤ ± Vmax j
1999; 2003; Jong et al., 2000; Jha and Schonfeld, vij (t ) = ⎨ (3)
2000; 2004; Kang et al., 2012; Kim et al., 2007; ⎩±Vmax j iiff viji ((t ) V j
Maji and Jha, 2009; Samanta and Jha, 2012). Their
outcomes have been very encouraging and repre-
where i is i-th particle (variable), j the j-th dimen-
sented great advances in this topic. For instance,
sion, gb and pb are the best positions achieved by
some papers introduced considerations concerning
the swarm and by the i-th particle respectively, w
road interconnections, bridges, or sight distances
the inertial coefficient (used to consider the influ-
into the analysis, producing effective algorithms.
ence of previous speed), c1 and c2 are acceleration
The Swarm Intelligence (SI) and, in particular, the
constants, both r1 and r2 are random numbers in
PSO method resulted as the best alternative to the
[0,1] range, t is the time, and Vmax represents the
GA, especially in terms of execution time; various
limit value of the speed, for better performance.
papers have presented the excellent results of the
Finally, since GA and PSO show different advan-
related algorithms (Angulo et al., 2012; Shafahi
tages, various papers proposed hybrid methods
and Bagherian, 2013; Bosurgi et al., 2013).
for obtaining better results and overcoming some
Both GA and PSO start from some random
operational problems (Eberharth and Shi, 1998;
initial solutions and, after each cycle, improve and
Thangaraj et al., 2001; Juang, 2004; Shafahi and
correct them in order to minimize costs (defined
Bagherian 2013). This combination may assure a
through a specific fitness or cost function), and
more extensive exploration of the searching space,
observe the several constraints. The main differ-
typical of the GAs, and, due to the PSO, guaran-
ences concern the analytical methods adopted for
tees a higher calculation speed and improves the
modifying and optimizing the various solutions.
model convergence.
Since GAs consider an evolutionary theory in
which only the best ones survive and reproduce
after each cycle, the solutions of each cycle are 3 BRIEF NOTES ABOUT PPCS
rearrangements and combinations of the previ-
ous ones. The PSO, on the contrary, moves in the The transition curve is an essential element for
search space the different solutions, as it derives causing a gradual influence of the radial accelera-
from the observation of flocks of birds moving in tion in infrastructure alignments, guaranteeing a
the space and searching for food. In detail, the SI is better visual perception of the curve, and assur-
an AI technique introduced by Kennedy and Eber- ing a more realistic gradation of the steering speed
hart (1995), based on the social interaction typical according to the driver’s behaviour. Traditionally,
of particular simple groups of animals, that, with- for improving the comfort for users while driv-
out a leader and through simple and coordinated ing along curves, the clothoid represents the most
behaviours, can perform very complex global tasks. used transition curve in the design of road align-
Even if the individuals do not have a real collective ments. However, different authors examined the
awareness, the group is able to achieve exceptional effects of more powerful geometric solutions that,
goals. according to their results, may improve the levels
The hypothesis of the PSO is that each bird of comfort for users along curves (Changping,
(particle) does not know the real food position, but 2007; Can and Kuscu, 2008; Easa and Mehmood,
only its distance from it; then, it modifies its speed, 2008). Among the various solutions analysed, the
step by step, and thus its position, in order to mini- application of splines, polynomial curves, and
mize this distance. Similarly, each particle symbol- multispirals provided the best outcomes (Sànchez-
izes a possible solution, moving in the space of Reyes and Chacòn, 2003; Habib and Sakai, 2009;
the solutions, searching for global optimum. Each Jha et. al, 2010; Kobryń, 2011; Ziatdinova et al.,
particle changes its position according to its own 2012). Fifth degree polynomial curves showed
experience and by imitating other particle move- high geometric and computational versatility
ments and experience. The position is evaluated (Baykal et al., 1997; Tari, 2004; Baykal and Tari,
through a specific fitness (or cost) function to 2005), also in resolving some design cases with very
be maximized (or minimized). The equations to complex geometry. Bosurgi and D’Andrea (2012)

386

TIS2017_Book.indb 386 2/8/2017 8:05:04 PM

 Figure 1. PPC for transition between tangents.

introduced a parametric version of a fifth degree

polynomial curve and named it “Polynomial Par-
ametric Curve” (Fig. 1). It represents one of the
most promising solutions for improving comfort
along roads (Bosurgi et al, 2014, 2015).
The expression of the curvature κ of the PPC, Figure 2. Rate of change of radial acceleration (Bosurgi
represented by a fifth-degree polynomial as evi- and D’Andrea, 2012).
denced in equation (4), may properly adapt to
every different geometric situation by changing
the specific boundary conditions (for the curvature driver, numerous numerical examples confirmed
itself and for its derivative) that modify the curve the advantages of the PPC, assuming as references
characteristics. the main dynamic variables (rate of change of
radial acceleration, steering speed, and roll speed).
β The particular geometric characteristics of the
⎡1 ⎤
κ( ) = ( 5 4 3 2
f )⎥ [ / m ] (4) PPC guarantee the continuity of these variables
⎣R ⎦ involved while driving that directly influence the
driver's behaviour and the comfort. In general, a
where l = s/L is the normalized abscissa (L = total more gradual variation of the curvature produces
length of the curve and s = curvilinear abscissa), R smoother trends for the dynamic variables, and
represents the minimum value of the radius of the this helps users to drive better and in a more com-
osculating circle [m], the coefficients a, b, c, d, e, f fortable way. Fig. 2 provides a comparison of the
can be obtained by imposing the boundary condi- trend of the rate of change of radial acceleration
tions for the curvature and its derivative, and β is a between the PPC and the traditional solution (clot-
parameter useful for modifying the shape and the hoid-circular curve-clothoid). Although in this
length of the curve. particular example the PPC shows a higher peak
Practically, the PPC is defined using two dif- due to the specific adopted parameters—however
ferent parameters, called α and β, that influence lower than the physiological limit -, the related
its geometric shape and, thus, the trends of the smooth curve provides higher level of comforts
dynamic variables involved while driving. The first to the users rather than the traditional solution.
parameter, α, represents the normalized abscissa Moreover, the PPC does not imply discontinuities,
of the minimum curvature point and influences the producing very significant increases in the driving
symmetry of the PPC. The other one, β, is a scale comfort. The other dynamic variables present simi-
parameter: it amplifies the length of the curve and, lar smooth trend, confirming the high quality of
for a fixed value of the minimum radius, it provides this geometric solution.
an exponential geometric flexibility of the PPC in
the space.
The PPC assures better performances espe- 4 OPTIMIZATION PROBLEM
cially in terms of users’ comfort and seems to be MODELLING
more appropriate than the clothoid in the design
of roads. In order to prove this assessment, some The optimization problem concerns the choice of
considerations regarding the dynamic aspects an optimum backtracking 3D highway alignment
of the curve are very explanatory. Regarding the linking two known points in the 3D space in com-
checks on the motion and on the effects over the pliance with the various considered constraints.

387

TIS2017_Book.indb 387 2/8/2017 8:05:05 PM

 In this context, the optimum solution is the one randomly selects the x and y coordinates of the ini-
minimizing the total costs, defined through a tial alignment vertices. However, this could provide
proper cost function. In this specific case, the hori- improper solutions with critical loops and intersec-
zontal alignment is defined using the PPC. The tions between following straights. For overcoming
position of the vertices defining the alignment may this limitation, other algorithms can be adopted
represent the independent variable, according to for performing this task, such as specific methods
the research goals. Since three coordinates describe for selecting non back-tracking alignments (Jong,
this position in a 3D space, each solution is repre- 1998; Bosurgi et al., 2013).
sented by 3 np independent variables to optimize Regarding the vertical alignment, the algorithm
(Eq. 5), where np is the number of the internal ver- chooses the z coordinate of each vertex. Since an
tices of the alignment preliminarily fixed. appropriate query of the DTM provides the ground
elevation for each vertex, the algorithm slightly
si ⎡ xi1 yi zi1, xik yik , zik , xin , yin zin ⎤ (5) modifies these values, in order not only to main-
⎣ p p p ⎦
tain as close as possible the road and the ground
elevations, but also to define a smooth and correct
A Digital Terrain Model (DTM) reproduces vertical profile, in compliance with the adopted
the region topography with a very high detail road standards. For simplifying the analysis, the
level (Fig. 3). In the study region, it is possible vertices of the horizontal path are vertices of the
to evidence some areas characterized by various vertical profile also. Concerning the radius design,
constraints (economic, historical, environmental, the authors adopted particular methods, determin-
etc.). This characterization is handled through two ing random radius values within a given range for
specific rectangular grids easily adaptable to all both alignments. The aim of assuring the geomet-
conditions. The first grid (“Gridval”) provides the ric continuity of the alignment significantly condi-
unit cost for each square meter of road built in a tions the choice of the radii. In detail, in order not
selected cell, due to the economic (and natural or to induce intersections between the curves, main-
historical) value of the area and to the soil condi- taining also if needed a certain minimum straight
tions. The second one (“Gridrisk”), that may have segment, the algorithm must check the radii of
different sizes and characteristics, relates to the following curves. Special penalty functions will
environmental hazard of the study area (as well as charge additional costs to improper solutions, in
seismic, hydraulic, and geomorphological hazard). which the difference between following radii is too
The optimization algorithm can be divided into high or the radius values exceed the Road Standard
various steps: 1) data input and numerical check; boundaries.
2) initial representation of the region DTM and In general, the cost function to be minimized
of the study area; 3) solution initialization; 4) geo- is the key element of the optimization procedure
metric analysis of the straight segments; 5) PPC (Eq. 6). In this case, the total cost of each solu-
and parabolic curves fitting; 6) solution evaluation tion, CT, is the sum of the real cost, C, and various
and evolution using the swarm core; 7) operator penalties, P, useful to discard incorrect alignments
application; 8) stopping criteria checking and final by increasing the related cost in proportion to the
representation. related violation.
The initial solutions are very essential to solve
the problem properly and, in general, there are CT Clen + Cloc CE + PR PR + Pret
various alternatives for choosing them. The easier + PG PRv + Pddz PE (6)
method is a random definition of the several ini-
tial solutions for favouring a whole exploration where Clen is the length-dependent costs, Cloc the
of the study area; then, a particular algorithm location-dependent costs, CE the earthwork costs,
PRmin the minimum radius of circular curves pen-
alty, PR the ratio of following circular curves radii
penalty, Pret the minimum and maximum tangent
segment length penalty, PG the maximum grade
penalty, PRv the minimum radius of parabolic
curves penalty, Pdz the maximum height (depth) of
fill (cut) sections penalty, and PE the environmental
penalty.
To guarantee a rapid convergence and a larger
exploration of the search space, some specific
operators, derived from the GA theory, have been
introduced in the model. There are two differ-
Figure 3. DTM representation of the region. ent kinds of operators: stochastic and corrective

388

TIS2017_Book.indb 388 2/8/2017 8:05:05 PM

 ones. The stochastic operators facilitate the mov-
ing from local optima and the solution optimiza-
tion, through a more exhaustive exploration of the
search space. The corrective operators, on the con-
trary, modify specific infeasible solutions, charac-
terized by inappropriate values of deviation angles
or intersections between tangents.

5 PPC IMPLEMENTATION

Adopting the PPC in the horizontal alignment

of the road, each curve is described by only one
element and, as a result, design, calculation, and
representation are simpler than adopting the tradi-
tional composition. Even if there is not an explicit
expression for defining both the clothoid and the
PPC, considering the latter, the total number of Figure 4. Same PPC drawn for different number of ref-
erence points.
geometric elements decreases to the minimum
and, thus, from a computational point of view, the
advantages are very clear and important. In the Whether the procedure reduces some radii beyond
following, some operational and technical solu- the limit value, the related improper solutions are
tions to actually adopt the PPC in the PSO algo- later penalized, using an appropriate penalty func-
rithm are provided. tion. In general, for all the penalty functions related
to the curves, the authors have assumed the same
expression adopted for the previous models with
5.1 PPC fitting and design
circular curves, since it is easy to assess a certain
This phase involves the choice of the minimum similarity and correlation between their radius in
radius of the osculating circle for each PPC, in a traditional approach and the minimum radius of
compliance with comfort and geometric con- the osculating circle in the PPC.
straints, avoiding intersections between follow-
ing curve paths and maintaining, if needed, a
5.2 Location costs and environmental penalties
minimum residual straight. A specific procedure
assigns to each radius an initial value, chosen for Evaluating location costs and environmental pen-
favouring high service levels for users as the one alties represents a critical task for the algorithm,
corresponding to the maximum transversal grade because of the specific analytical formulation of the
and the maximum design speed, in compliance PPC. As said, since there is not an explicit expres-
with the Italian Road Standards (2001). sion describing the relations between the x and y
After the radius values are assigned, it is possi- coordinates of the curve points on the horizontal
ble to process the PPCs (considering typical values plane, only a point-by-point numerical calculation
of the parameters - β = 1.0 α = 0.5 - for simplifying permits to evaluate and represent the PPC. Each
the analysis and reducing the computational time). cell of the reference grids has a specific unit value
Each curve is numerically calculated and defined, for both economic and hazard classifications. The
according to the number of reference points prop- objective of this sub-step is to estimate the length
erly chosen. Obviously, a reduction in the number of road passing through each cell.
of points describing the PPC decreases the quality Although for straights this could be easily eval-
of the representation, but increases the perform- uated through a linear system of equations, the
ance of the numerical evaluation (Fig. 4). PPC formulation does not permit to follow this
Then, it is time to check the geometric consist- approach. Then, the authors proposed a differ-
ency of the alignment, looking for critical point of ent but equally efficient procedure. Since the grid
intersections between following curves that could characteristics and the x and y coordinates of the
happen if the tangents of the curves are too long PPC points are known, it is simple to locate the
compared to the available distance between the ver- indices of the cell in which each point is. The algo-
tices. If all the residual straights are sufficient, the rithm counts how many points are located in each
horizontal alignment is complete. If some straights cell and the segments that are one less. Moreover,
are shorter than the minimum value, the critical as the curve is practically defined point by point,
segments are further processed and corrected, by the distance between two following points along
reducing in sequence all the critical radius values. the PPC is constant (Eq. 7). For more clarity,

389

TIS2017_Book.indb 389 2/8/2017 8:05:06 PM

 Table 1. Points and segments for cells according to Fig. 5. [ ] = rl
rl p t ⋅ LsPPC rll f (10)
PPC i , j

Cell Points Segments

where rlp is the residual length in the i-th j-th cell of
i, j − 1 3 2 the previous boundary segment, t is the number of
i, j 3 2 segments in the i-th j-th cell (equal to the number
i, j + 1 2 1 of points in the cell minus one), LsPPC is the length
i + 1, j + 1 1 0 of each PPC segment, and rlf is the residual length
i+1, j+2 2 1 in the i-th j-th cell of the following boundary
segment.

5.3 Earth-costs
Regarding the earth-costs evaluation, the main
complications derive again from the specific ana-
lytical formulation of the PPC. In this context,
the earthwork volumes are evaluated using the
“average-end-area” method. In order to properly
evaluate these volumes, it is important to define a
method to locate the reference station useful for
reproducing the ground profile. Although both the
tangent points of the PPCs always constitute sta-
tion points, there are two possible approaches:
Figure 5. Location costs using PPC.
− all the PPC points are assumed as station points;
− if the station points are equally located along the
Table 1 provides the number of points and segment whole alignment, only few selected points of the
for each cell, according to Fig. 5. PPCs are considered as station points.
In the first case, it is easy to evaluate the ground
LPPC
LsPPC = (7) elevation for each station point (Fig. 6a), but the
nPPC computational costs in the calculation procedure
may increase widely. Considering the second sce-
where LsPPC is the length of each segment, LPPC is nario, the number of points is limited, but it is
the total length of the PPC and nPPC is the number necessary to define a specific rule for selecting the
of the segments. points along the PPC. In this study, the authors have
Regarding the boundary elements (i.e. seg- applied the second approach. In detail, since after
ments between points located in different cells), the stations are equally located along the alignment
the authors have assumed to consider the curve (St*) the related progressive distances are known as
segment as linear. This approximation produces a well as those of the tangent points, the algorithm can
very little error that may become irrelevant (10−3) easily select the closest PPC points (St) to the sta-
considering the whole length of the PPC, especially tion target point (Fig. 6b). Then, these become the
if the number of segments used for defining it is selected station points and the DTM model of the
large enough (>100). In detail, if Q is in the inter-
section point with the grid, A is the point in cell
{i − 1, j}, and B is the point in cell {i, j}, the length
of the two segments (AQ for cell {i − 1, j} and QB
for cell {i, j}, Fig. 5) can be evaluated through
expressions (8) and (9).

(x ) + (y )
2 2
li j AQ Q − xA Q − yA (8)

(x ) + (y )
2 2
li j QB B − xQ B − yQ (9)

Then, in this way it is possible to estimate how

long are the segments passing through each cell
and, for a generic cell i, j, the related length of the
PPC can be calculated using equation (10). Figure 6. PPC stations for vertical profile calculations.

390

TIS2017_Book.indb 390 2/8/2017 8:05:07 PM

 study area provides the related ground elevations,
for calculating the earthwork volumes and costs.

6 NUMERICAL EXAMPLE
AND DISCUSSION

For evaluating and practically using the proposed

method, the authors coded an original script. The
results of a specific example are useful for show-
ing the effectiveness of the entire procedure. The
Figure 7. Alignment evolution.
reference maps considered in this example have
rectangular shape and measure 4000 m on X side
and 3000 m on Y direction. The study area topog-
raphy is reproduced through a particular DTM. In
order to get a proper representation of the region
characteristics, the authors defined the DTM using
more than 15.000 points with various elevation val-
ues (Fig. 3). The reference grids for economic and
environmental hazard constraints have the follow-
ing resolution: the cell size is 175 m in direction X
and 120 m in direction Y.
The authors assigned four different unit costs
(useful to estimate location-dependent costs, from
C1 low to C4 very high) to each cell of the Gridval
matrix, in order to define an ideal representation of
the study area. Four different hazard levels (from
H1 low to H4 very high) describe the environmen-
tal hazard map. They are representative of seismic,
geomorphologic and flood hazard levels for each
cell, and relate to four different unit penalties. The
greater the hazard level, the higher the probabil-
ity that a critical event happens (or its magnitude
is greater), the greater the unit cost related to this Figure 8. Vertical profile evolution.
hazard level must be. Since these examples repre-
sent only a test of the algorithm efficient, for char- comparison of the previous figures proves the cor-
acterizing the test area, the authors have defined rect and effective evolution of the road in the study
ideal distribution for location-dependent unit costs area. In the first phase of the analysis, the solution
and hazard levels. is widely incoherent and improper from a techni-
The elaboration started after all calculation cal point of view: Fig. 7a represents an alignment
variants (population size: 20, max cycles number: with abrupt and not correct changes of curvature
500, Vmax = 500 m, Vmaxz = 10 m, w = 0.5, c1 = 1.5, passing also through some of the high cost areas.
c2 = 2.5), unit costs (pavement, maintenance, envi- Furthermore, the selected road does not fit the ter-
ronmental and earthwork costs), geometric and ritorial characteristics, as confirmed by the vertical
environmental unit penalties had been specifically alignment provided in Fig. 8a, in which the ground
defined through appropriate input boxes. The final and the road profiles are clearly inconsistent, with
solution is made up of the coordinates of 4 internal very high height differences.
vertices (np = 4). Moreover, the algorithm analysed Fig. 7b, 7c, 8b and 8c permit to follow the effi-
50 sections for describing the vertical profile of the cacious optimization of the alignment. Step by
road. The maximum number of cycles is 500, and step, the road is effectively improved and the curve
all the PPCs count 50 building points. In order to characteristics are more homogeneous. Further-
underline the model efficiency, the best alignments more, the critical areas are avoided everywhere,
found by the algorithm at cycles 5, 25, and 500 are and the road elevation is properly defined, in
presented below: Fig. 7 provides an evolution of compliance with terrain height and features. The
the horizontal alignment, while Fig. 8 confronts final solution is acceptable and shows an effec-
the vertical profiles for the different cycles. tive optimization of the alignment. In particular,
Although the new algorithm includes the PPC both the horizontal path and the earthworks are
and the related modifications, an observation and widely optimized. The algorithm drew a correct

391

TIS2017_Book.indb 391 2/8/2017 8:05:08 PM

 area. The resulting algorithm proposed in this
paper guarantees to not only minimize the related
costs and design an alignment in compliance with
the main constraints, but also improve the users’
comfort along the road, as a direct consequence of
the PPC dynamic advantages. Moreover, the PPC
permits to solve the curve and define the transition
between two straights using only a single element,
producing benefits also in terms of design, evalua-
tion, and drawing of the curve.
This procedure, as also proved by the numeri-
cal example provided by the authors, is very effi-
Figure 9. Cost evolution of the best solution on advanc- cient and can help engineers in the planning phase.
ing cycles. The cost function may be further modified and
improved, in order to keep into considerations
other important factors conditioning the alignment
alignment, reducing both geometric and economi- design. Future researches may introduce comfort
cal penalties and reducing the peaks difference considerations in the cost function, in order to
between the ground and the road profiles. try to optimize the road characteristics consider-
Moreover, plotting the total cost evolution of the ing also the peak values of the rate of change of
best solution as function of cycles on a logarithmic radial acceleration. The modification of the ver-
scale (Fig. 9) evidences the solution improvement tical curves might be equally interesting, defining
and optimization after each single computational 3D alignments increasing the comfort of user both
cycle. Cycle after cycle, the algorithm actually on the horizontal and vertical planes.
modifies the different solutions reducing the cost
and improving the quality of the alignment. It
should be noticed that, probably, more iterations REFERENCES
should have produced a more accurate solution
with excellent geometric features. However, since Angulo, E., Castillo, E., Garcia-Ròdenas, R. and
the aim of the paper is to evidence the introduction Sànchez-Vizcaìno, J. (2012). Determining Highway
of the PPC in this kind of optimization method Corridors. Journal of Transportation Engineering,
and other proofs of the effectiveness of the PSO 138, 5, 557–570.
approach are provided in literature, the authors Baykal, O. and Tari, E. (2005). A New Transition Curve
stopped only after 500 cycle the optimization pro- with Enhanced Properties. Canadian Journal of Civil
Engineering, 32(5), 913–23.
cedure. This number of cycles is enough to assure Baykal, O., Tari, E., Coskun, Z. and Sahin, M. (1997).
that the technical solutions adopted for handling New Transition Curve Joining Two Straight Lines.
the PPC in this PSO optimization method are effi- Journal of Transportation Engineering, 123(5),
cient and the alignment optimization is properly 337–45.
performed. The natural geometric characteristics Bosurgi, G. and D’Andrea, A. (2012). A Polynomial Par-
of the PPC assure better performance in terms of ametric Curve (PPC-Curve) for the Design of Hori-
comfort for users, assuring a smoother variation of zontal Geometry of Highways. Computer-Aided Civil
the dynamic variables involved while driving. and Infrastructure Engineering, 27, 303–312.
Bosurgi, G., Pellegrino O. and Sollazzo G. (2013). A PSO
Highway Alignment Optimization Algorithm Consid-
ering Also Environmental Constraints. Advances in
7 CONCLUSIONS Transportation Studies an International Journal, Sec-
tion B, 31, 63–80.
In this paper, the authors have proposed an inno- Bosurgi, G., Pellegrino, O. and Sollazzo G. (2014).
vative optimization algorithm, based on the PSO “Using Genetic Algorithms for Optimizing the PPC in
method, for designing highway alignments consid- the Highway Horizontal Alignment Design.” Journal
ering the PPC as unique curvature element. The of Computing in Civil Engineering, 10.1061/(ASCE)
adoption of the PPC guarantees higher levels of CP.1943–5487.0000452, 04014114. 30 (1), Jan. 2016.
comfort for users than the traditional transition Bosurgi, G., Pellegrino, O. and Sollazzo G. (2015). An
Algorithm Based on the PPC (Polynomial Para-
curves. In particular, the paper presented various metric Curve) for Designing Horizontal Highway
operational solutions for introducing the PPC in Alignments. Advances in Transportation Studies an
the optimization method. In general, the optimi- International Journal, Section A, 36, 5–20.
zation algorithm leads to define an efficient pro- Can, E. and Kuscu, S. (2008). Investigate of Transition
cedure useful for preliminary locating the most Curves with Lateral Change of Acceleration for High-
appropriate path for a road in a specific study ways Horizontal Geometry. Integrating Generations

392

TIS2017_Book.indb 392 2/8/2017 8:05:09 PM

 FIG Working Week 2008, Stockholm, Sweden 14–19 Geographic Information Systems. Computer-Aided
June 2008. Civil and Infrastructure Engineering, 15(4): 261–271.
Changping, W. (2007). Polynomial Curve Fitting Method Juang, C.F. (2004). A hybrid of Genetic Algoritghm
for the Design of Highway Horizontal Alignment. and Particle Swarm Optimization for Recurrent Net-
CNKI Journal, 03, 30–2. work Design. IEEE Transactions on systems, man
Easa, S.M. and Mehmood, A. (2008), Optimizing and cybernetics—Part B: Cybernetics, Vol. 34, No. 2,
Design of Highway Horizontal Alignments: New 997–1006.
Substantive Safety Approach. Computer-Aided Civil Kang, M.-W., Jha, M.K. and Schonfeld, P. (2012). Appli-
and Infrastructure Engineering, 23(7), 560–73. cability of highway alignment optimization models.
Eberharth, R.C. and Shi, Y. (1998). Comparison between Transportation Research, Part C, 21: 257–286.
Genetic Algorithms and Particle Swarm Optimiza- Kennedy, J. and Eberhart, R.C. (1995). Particle Swarm
tion. Lecture Notes in Computer Science Volume Optimization. Proceeding of IEEE International
1477, 611–616. Conference on Neural Networks, 1942–1948.
Habib, Z. and Sakai, M. (2009). G2 Cubic Transition Kim, E., Jha, M.K., Schonfeld, P., and Kim, H.S. (2007).
between Two Circles with Shape Control. Journal Highway Alignment Optimization Incorporating
of Computational and Applied Mathematics, 223, Bridges and Tunnels. Journal of Transportation Engi-
133–144. neering, 133, 2, 71–81.
Italian Road Standards (2001). “Ministerial Decree Kobryń, A. (2011). Polynomial Solutions of Transition
n. 6792 (11/05/2001) of the Ministry of Infrastructures Curves. Journal of Surveying Engineering, 137(3),
and Transports. Functional and geometric standards 71–80.
for building roads” Rome (in Italian). Maji, A. and Jha, M.K. (2009). Multi-objective Highway
Jha, M.K. and Schonfeld, P. (2000). Integrating Genetic Alignment Optimization using a Genetic Algorithm.
Algorithms and GIS to Optimize Highway Align- Journal of Advanced Transportation, 43(4), 481–504.
ments. Transportation Research Record, 1719: Samanta, S. and Jha, M.K. (2012). Applicability of
233–240. Genetic and Ant Algorithms in Highway Alignment
Jha, M.K. and Schonfeld, P. (2004). A highway alignment and Rail Transit Station Location Optimization,
optimization model using geographic information International Journal of Operations Research and
systems. Transportation Research, Part A, 38(6): Information Systems, 3(1), 13–36.
455–481. Sànchez-Reyes, J. and Chacòn, J.M. (2003). Polynomial
Jha, M.K., Kumar Karri, G.A. and Kuhn, W. (2010). Approximation to Clothoids via S-Power Series.
Selection of 3 D Elements for Different Speeds in the Computation Aided Design, 35, 1305–1313.
3D Modelling of Highways. 4th International Sym- Shafahi, Y. and Bagherian, M. (2013). A customized
posium on Highway Geometric Design, June 2–5, Particle Swarm Method to solve highway alignment
Valencia, Spain. optimization problem. Computer-Aided Civil and
Jong, J.-C. (1998). Optimizing highway alignments with Infrastructure Engineering, 28: 52–67.
genetic algorithms. Ph.D. Dissertation, University of Tari, E. (2004). The New Generation Transition Curves.
Mariland, College Park. ARI Bulletin of the Istanbul Technical University,
Jong, J.-C. and Schonfeld, P. (1999). Cost Functions for 54(1), 34–41.
Optimizing Highway Alignments. Transportation Thangaraj, R., Pant, M., Abraham, A. and Bouvry, P. (2001).
Research Record, 1659: 58–67. Particle swarm optimization: Hybridization pespectives
Jong, J.-C. and Schonfeld, P. (2003). An evolutionary and experimental illustrations. Applied Mathematics
model for simultaneously optimizing 3-dimensional and computations. Vol. 217, 12, 5208–5226.
highway alignments. Transportation Research, Part B: Ziatdinova, R., Norimasa, Y. and Kima, T.-W. (2012).
Methodol., 372: 107–128. Fitting G2 Multispiral Transition Curve Joining
Jong, J.-C., Jha, M.K. and Schonfeld, P. (2000). Prelimi- Two Straight Lines. Computer-Aided Design, 44,
nary Highway Design with Genetic Algorithms and 591–596.

393

TIS2017_Book.indb 393 2/8/2017 8:05:09 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

QEMS. ANAS quality evaluation management system

Eleonora Cesolini & Stefano Oddone

ANAS S.p.A., Rome, Italy

Sonia Gregori
Anas International Enterprise S.p.A., Rome, Italy

ABSTRACT: The ANAS Quality Evaluation Management System (QEMS) is an innovative tool that
enables the collection, the analysis and the evaluation of all the information related to road pavement for
new construction projects and for maintenance of existing roads. The system provides on-time quality
reports which include information about the quality of the activities involved in road construction and
maintenance, and data about the performances of the Key Suppliers (KS). Based on this, with the QEMS
you can check the quality performance of each activity performed by KS. In addition, the QEMS may
be used by each KS, who is allowed accessing only to its own results. This interactive tool is proposed as
a support to enhance the quality of road construction by providing a synthetic and on-time analysis of
performances, all the results are showed with different colors on geographical maps in order to facilitate
the user to read them.
In the QEMS, each project is divided into more audit areas each characterized by a unique pavement
design. Then, in each Audit Area, the activities of each KS (e.g., laying, compaction), are audited acquir-
ing selected information (laboratory test data, on site audit data, management quality audit data), defined
as parameters; each parameter is evaluated according to defined evaluation criteria based on weights,
optimal values and acceptable values. Within the same audit area all the activities are organized into a
hierarchical structure, called Activity Tree.
The QEMS is completely configurable and you can simulate more scenarios with various weight factors
in order to get reports with different evaluation criteria.

1 INTRODUCTION • allowing timely put in place of corrective actions,

helping in the improvement of the Quality Lev-
The Quality Evaluation Management System els (knowledge allows improvements);
(QEMS) is a software tool wholly conceived, • providing tools for communication between
designed and developed by ANAS. The tool ana- different parties (Public Works Authority,
lyze and evaluate the pavement from the designing General Contractors, Supervision Consultants,
phase to the construction and maintenance phase. etc.).
The software uses a Geographical interface to
show the results of analysis; it intends to be used by In summary, the main characteristics of QEMS
all stakeholders involved in pavement design, con- are:
struction and maintenance (commissioner, audi-
• flexibility: it allows to easily define and modify,
tors, general contractors, supervision consultants,
over the time, the performance evaluation model
material suppliers and production facilities,…).
(weighting factors, parameters, activities, layers,
The final aim of this interactive tool is to
mixtures, materials, user profiles, etc.), making
enhance the quality of road construction by:
possible to easily adapt QEMS to new require-
• providing synthetic and on-time analysis of ments or new technologies;
performances, highlighting any problem and • accessibility by all the stakeholders involved in
allowing putting in place fast and optimized the road pavement construction projects: each
solutions; user accesses only his data;
• supporting activities of performance evalua- • friendly interface, with an Intuitive Graphi-
tion, allowing each involved KS, to access and cal User Interface, including reports displaying
to monitor his own quality performances; results on geographical maps.

395

TIS2017_Book.indb 395 2/8/2017 8:05:11 PM

 2 KEY CONCEPTS & DEFINITIONS Projects. The Key Supply Chain Partners are clas-
sified in categories according their role (Pavement
The analyses of performances are based on the Design Consultant, Contractor, Supervision Con-
QEMS configuration and on all information sultant, Materials Supplier and Production Facili-
acquired during audit activities on pavement con- ties, Laboratories, etc.).
struction projects (the results of the audits).
To illustrate the QEMS, it is necessary to intro-
2.4 Sources and plants catalogue
duce some preliminary concepts and definitions,
described in the following sections. Production affects significantly the quality and the
performances of all types of mixtures. The engi-
2.1 Project neering properties and behaviors of materials used
in the construction of pavement layers influence
Set of the overall information related to a specific the overall performance of the pavement structure
construction or maintenance process of one or over its design life. In QEMS also plants have to
more roads. The definition of a Project includes: be defined and they are audited and evaluated to
• The geographical location of the Project, ensure that plants are capable of producing materi-
• List of the KSs involved in the delivery of the als according to the physical and engineering prop-
final road. erties as specified in the required standards.

An example of a geographical location of a

Project is displayed in Fig. 1. 2.5 Activities
Activities are the operations necessary to accom-
2.2 Audit Area plish the successfully deliver of the road con-
In QEMS Audit Area is a part of the Project con- struction projects whose phases and processes are
taining road sections with the same pavement design. carefully audited.
A project may have one or more Audit Areas. Road pavement constructions are complex sets
Each Audit Area is composed by one or more of multiple processes (or Activities) interconnected
Road Sections, with a start and end chainage. referred to as “Supply Chain”. The pavement
Each road section is represented on the map with construction (complex multiple processes), is so
a polyline (Fig. 2). divided in several single activities that are audited.
The Main Activity (which includes all the activi-
ties involved in the project of road Construction) is
2.3 Key Supply Chain Partners (KSs) named: “Overall Road Construction”. This Activ-
Key Supply Chain Partners (KSs) are the subjects ity includes sub-Activities (carried out by different
responsible for the successfully delivery of the KSs):

Figure 1. Example of a Project.

396

TIS2017_Book.indb 396 2/8/2017 8:05:11 PM

 Figure 2. Example of Audit Area.

• Pavement Design • Production: Mixing Temperature, Compaction

• Quality Management System Temperature, etc.;
• Pavement Construction • Materials and Mixture Properties.
These sub-Activities articulate into further Therefore, the evaluation of a single layer occurs
sub-sub-Activities, because they are performed by auditing a set of sub-activities.
different KSs or at different times. Weighting factors are assigned to each activity
For example: and each parameter to establish the performance
evaluation criteria.
• Pavement Design includes sub-activities: Geo-
technical Investigations, Subgrade Layer Design,
Traffic Design Estimation, etc.; 3 CONFIGURATION DATA
• Pavement Construction includes a set of sub-ac-
tivities for each pavement layer: Construction of 3.1 Materials and Pavement Layers
Surface Layer, Construction of Base Layer, etc.; In order to define the pavement structure of
For this, the Activities have been structured in a each Audit Area, in the QEMS it is necessary to
hierarchical tree structure. configure: Raw Materials, Material Types, Mix-
tures and Pavement Layers. For each pavement
layer must be specified the materials and mixture
2.6 Activity tree template used. Therefore, the evaluation of pavements qual-
ity needs to go through the quality assessment of
In QEMS the complex relations between activities the used materials. Follows an example list of lay-
are organized into a hierarchical structure defined ers with their mixtures and materials.
as Activity Tree.
Development of the three template stop when 3.2 Parameters
an activity can be measured through parameters.
Parameters are the basic items directly audited and
The evaluation of Parameters could be through
used as evaluation criteria to assess the quality of
laboratory tests or specific audits.
the Activities.
The evaluation of Activities at higher level occurs
For each Leaf Activity of the Activity Tree
through the evaluation of activities at lower level.
Template it is necessary to define the Parameters
For instance, the performance evaluation of a
to evaluate it.
HMA Layer depends on:
Parameters are for example, the temperature of
• Construction: Laying, Compaction, etc.; the asphalt, the thickness of a layers, etc.

397

TIS2017_Book.indb 397 2/8/2017 8:05:13 PM

 Figure 3. Activity Tree Template.

Table 1. Example of Pavement Layers, Mixtures and Activity containing the Parameter. The Parameters
Materials. so configured, added only once in the Activity Tree
Template, will be automatically available on the spe-
Pavement Layers Mixtures Materials cific Activity Tree of each Audit Area, depending
Asphalt Surface Hot Mix Neat Asphalt Binder; on the pavement structure of the Audit Area.
Course (SC) Asphalt PMB, Coarse
Aggregates, Crushed
Aggregates, Fine 4 AUDITS
Aggregates, Filler
Asphalt Base Hot Mix Neat Asphalt Binder; In each Project, Activities are audited to assess their
Course (BC) Asphalt PMB, Coarse quality upon specified criteria, therefore, in the
Aggregates, Crushed QEMS the audit results coincides with Parameters
Aggregates, Fine results. Different types of Parameters require dif-
Aggregates, Filler ferent types of audit with different type of results.
Sub base / road Sub base Crushed Aggregates, In the QEMS, three types of Audit are possible:
base (SB) Mixture Fine Aggregates,
Cement, Filler • Laboratory Testing Results Audit—Laboratory
test, to assess materials properties;
• On Site Audit—Audits performed by auditors
on construction as well as production sites, to
In the QEMS a Parameter can be single (eg.
assess Construction Practice and Production
the temperature of the HMA during compaction
Facilities.
activities) or composed by a group of Sub-Parame-
• Documental Audit—Audits of documents, man-
ters (eg. the Parameter is the grading curve and the
uals and procedures adopted by the KSs.
sub-Parameters are the percentages of passing for
the different diameters).
Some Parameters are related to specific Pavement
5 EVALUATION
Layers, so QEMS enables the definition of condi-
tion of existence of each Parameter depending on
5.1 Normalized Scores and Quality Indicators
the characteristics of the layer (i.e. Layer, Mixture,
Raw Material, and Material Type). These condi- In QEMS the evaluation process begins with the
tions must be compatible with those set for the Leaf evaluation of Parameters, all the values of them

398

TIS2017_Book.indb 398 2/8/2017 8:05:14 PM

 are converted in Normalized Scores. Normalized on testing procedures and accuracy of the test
Scores are real numbers from 0 to 100 indicating method; where tolerance was not clearly speci-
the quality level of the result of the Parameter fied, a Coefficient of 5% was generally used to
(100 = max quality). set tolerance limits.
Combining the results obtained for a Param-
Within the Optimal Range, the Normalized
eter in a defined time interval (i.e. combining more
Score is 100 while within the Acceptability Range
Normalized Scores using specific algorithms), we
the Normalized Score varies linearly from 100 to
obtain a result that expresses the quality of the
0. Outside of the Optimal and the Acceptability
Parameter in the defined time interval; such result
Range, the Normalized Score is zero.
is named: Parameter Quality Indicator (PQI).
Combining the PQIs of the Parameters of a
Leaf Activity, we obtain a result that expresses the
5.3 Evaluation of Activities in the Activity Tree
quality of the Leaf Activity in the defined time
template
interval; such result is named: Activity Quality
Indicator (AQI). The main activity “Overall Road Construction”
Both PQIs and AQIs (as well as the Normalized is organized in Sub Activities providing braches
Scores) are expressed with real numbers from 0 to Activities at different intermediate levels and Leaf
100 indicating the quality level (100 = max qual- Activities at the lowest level of the hierarchical
ity); this criterion provides values combinable and structure. Beside Activities layout, in the compu-
comparable one each other. tational model, the definition of Activities requires
also to define 2 characteristics:
5.2 Evaluation of the Normalized Scores • Weighting Factor—For each Activity, a weight
The criterion for the evaluation of the Normalized factor has to be defined depending on its own
Scores from Parameter Results depends on the importance; it is used to evaluate the AQI of the
typology of Parameter input data that may be: Activity at a higher level.
• Percentage of Worst Results—for each Activity
• Selection from a data list (e.g. on-site and docu- has to be defined the percentage of worst results
mental audits), to be considered in the computation; such value
• Input of a Numerical value (e.g. lab test is used to calculate the AQIs of the Activity
results). using a subset of QIs values (the worst) of his
If the Parameter input data is based on the selec- Sub-Activities or Parameters.
tion from a data list, the criterion is based on the The quality of the Leaves Activities is measured
direct assignment of a Normalized Scores to each upon Parameters (evaluation criteria). In light of
item of the data list, an example of such criterion this, once all the Activities were configured, then
is shown in Table 2. Example of criterion to assign in each Leaf Activity the Parameters selected as its
Normalized Scores to Data List type Parameters. own evaluation criteria were defined.
If the Parameter input data is a numerical value, The evaluation of the AQI of the Leaf Activities
the criterion for the evaluation of the Normalized depends on:
Scores is based on the definition for each Param-
eter (using the QEMS configuration functions), of • the values of the PQI of the Parameters of the
threshold limits. In this case, two ranges of Nor- Leaf Activity,
malized Scores are considered: • the Weighting Factors of the Parameters of the
Leaf Activity,
• Optimal Range—Parameter within standard • the defined Percentage of Worst Results defined
threshold limits (Score 100); for the Leaf Activity.
• Acceptable Range—Parameter within tolerance
limits; in this case, the tolerance limits depend The evaluation of the AQI of the Activities with
child activities depends on:
Table 2. Example of criterion to assign Normalized
Scores to Data List type Parameters.
• the values of the AQI of its child Activities,
• the Weighting Factors of its child Activities,
Parameter’s Value Normalized Score • the defined Percentage of Worst Results.
The configuration of the Activity Tree Template
Yes 100
requires the definition, for each Activity, of his
Observation—Acceptable 75
Observation—Partial Satisfactory 50
repeatability depending on the pavement struc-
Observation—Poor 25
ture (eg. the layer construction activity has to be
No 0 repeated for each layer). The QEMS system auto-
matically generates a specific Activity Tree for each

399

TIS2017_Book.indb 399 2/8/2017 8:05:15 PM

 Figure 4. Example of Report.

Figure 5. Example of Activity Quality Indicator.

Audit Area depending on the defined pavement indicator is said “Measured” because his value
structure. derives from the results obtained directly from the
For this reason, the audit results have to be col- measurements of the Parameters of the Activity;
lected in each audit area; despite this, the validity of • for each Activity with child Activities, is cal-
such results may be extended to the entire Project or culated the Activity Quality Indicator Derived
to multiple Projects. For this reason was introduced (AQID) on the base of the Activity QIs of his
in QEMS the concept of “scope” to be defined for child Activities; this indicator is said “Derived”
each Parameter. This aspect permits to introduce because his value derives indirectly from the
as the geographical validity of the audit results in results of other Activities.
the computation of the Quality Indicators.
In synthesis:
6 REPORTS
• for each Activity with Parameters (Leaf Activities),
the Activity Quality Indicator Measured (AQIM) is QEMS provides some powerful reporting func-
evaluated on the base of the Parameter QIs; this tionalities, including the following:

400

TIS2017_Book.indb 400 2/8/2017 8:05:15 PM

 Figure 6. Example of Parameter Quality Indicator.

• Performance Reports of Activities It is possible to browse the reports obtaining

• Performance Reports of KSC Partners detailed information on QIs and audit results.
The results of the evaluations are displayed on
geographical maps: The polygons of a Project or REFERENCES
the polylines of an Audit Area are shown with dif-
ferent colors depending on the Performances of Cesolini, Oddone & Cuciniello (2015) 3rd MESAT2015
Middle East Society of Asphalt Technologists—The
the Activities (AQI).
ANAS Quality Evaluation Management System
Within the map it is possible to navigate obtain- (QEMS) Applied to Road Pavement Construction.
ing reports; for example the Partner Quality Indi- Cesolini, Oddone & Gregori (2016) Transport Research
cator Report or the Activity Quality Indicator Arena TRA2016—ANAS Quality Evaluation Mana-
Report. gement System.

401

CH052_250.indd 401 2/8/2017 8:49:29 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Innovative approaches to implement road infrastructure concession

through Public-Private Partnership (PPP) initiatives: A case study

M. Arata & M. Petrangeli

Anas International Enterprise S.p.A., Rome, Italy

Francesco Longo
ANAS S.p.A., Rome, Italy

ABSTRACT: During recent decades, major public road authorities showed an increasing trend in seek-
ing for alternative ways to establish and implement road infrastructures concession systems, with the main
aim to make it doable in a reliable, socio-economically sustainable and cost effective manner.
This phenomenon is mainly due to increasingly higher risks related to the establishment of road conces-
sion system and risk related to the introduction of road toll payment requirements.
During the last three years, ANAS International Enterprise S.p.A., a subsidiary company of the Ital-
ian Highway Agency ANAS S.p.A., has been appointed for assessing the feasibility and comprehensively
structuring road concessions, through the implementation of a Public-Private Partnership (PPP) initia-
tive, for a large part of road network of the Republic of Colombia.
The main activities consisted in (i) the development of a financial model, consistent with the rules and
economic parameters related to the Colombian market; (ii) the assessment of the economic and financial
feasibility; (iii) carrying out studies and simulations for several concessions scenarios implementation,
including analysis of risks and technical, financial and legal assessment; (iv) preparation of tender docu-
mentation and delivery of technical support to the National Infrastructure Agency during the public-
private partnership awarding process.
Along with these activities, the implementation of an innovative road concession system allowed the
transfer and application of a number of best practices in various sectors. Indeed, improvement on the
following aspects have been achieved:
• Innovations in construction, including alternative solutions for road embankment slope stabilization;
barriers against rock falls; asphalt concrete mixtures modified with polymers;
• Technological innovations, including intelligent transport systems, telecommunications and informa-
tion infrastructure, subsystem tolls, transport operating centers, subsystem video surveillance cameras
hotspots, emergency call subsystem, weighing subsystem, lighting systems in urban areas and tolls;
• Innovations in financial structuring, including tailor-made regulating financial planning and monitor-
ing tools, introduction of innovative solutions for the use of appropriate discount rates aimed to ensure
investor profitability, assessment tool to manage public -and private-sides risks, assessing tools to ensure
value for money, implementation of alternative solutions for public participation in Project Financing ini-
tiatives, development and implementation of alternative solutions for private funding in the financing proc-
ess (including structuring infrastructure bonds), adoption of PPP structure type contracts provisions;
In this paper, the Authors describe in detail the concession implementation process, by particularly
focusing on criteria adopted in selecting the contract type and related aspects, and highlighting main
criticalities identified throughout the concession setting up process.
Aiming to give food for thought for further development of road infrastructure concession practice
implementation, this paper aims to provide the Reader with an analysis of an actual successful case study,
according to the trust demonstrated by the financial closing approved by the banks, and accurate and
well-balanced comments on advantages, disadvantages and criticalities faced.

1 ALTERNATIVES FOR ROAD to give to this Partnership, and the equilibrium

INFRASTRUCTURE CONCESSIONS of the sharing of the burdens and responsibilities
In the PPP acronym the most important P is the between the Public and the Private sectors, the key
third one: Partnership. It is the meaning we want of success of such initiatives.

403

TIS2017_Book.indb 403 2/8/2017 8:05:19 PM

 This has also been the lesson learned in Colom- spaces (to rent) and supplying the legal assessment
bia, where ANAS has been awarded two contracts for arrangements that capture part of the value
for the comprehensively structuring of around added from the infrastructure to the productive
3.000 km of roads, divided in 9 lots, of the fourth activities of the area.
generation of concessions, the so called 4G high- Once finalized the feasibility study, the Admin-
way program, after the first, second and third gen- istration can decide the concession’s duration and
erations failed precisely for the lack of balance. In whether keeping it fixed or variable, according to
fact in the first generation all risks were borne by the necessity of government grants (in the referred
the public, ending to be much more costly for the project some corridors needed from 30% up to 70%
government. of public funding), researching and selecting the
In the second generation the situation has been best international practices regarding the financing
inverted, with the result that the private partner of infrastructure.
have either failed or fled the contracts. In the third To this end ANAS had interviews with the man-
one a better equilibrium has been found, but still agement of banks and other lenders and analyzed
lack of clear rules ended up creating delays, low major bond issues in the field of infrastructure,
levels of services and complaints. in order to understand the current potentiality
Therefore it is important that the public admin- of project financing and related conditions so
istration proposing the concession properly plan as the current situation of the capital market in
the project and arrange a fair contract, financially Colombia.
attractive for the private sector, but clear and
detailed in defining not only its obligations, but the
level of service required (in the referred projects 2 RISK ANALYSIS
ANAS has expanded from 12 to 22 the service level
return index to be satisfied by the concessionaires Risk analysis is a fundamental step in the study of
during their operation and maintenance period). a PPP project, which allows to evaluate the feasibil-
We must start from the analysis of the economic ity of the project at 360 degrees. Within the foresee-
viability of the project. To that end we must have able risk, the estimate is to assess the probability of
an accurate analysis of traffic flows with a care- occurrence of the risks that have been exemplified
ful projection, integrating future development and, considering their level of impact, to evaluate
of the territory crossed. Being this difficult and their effects. In the specific Colombian experience,
uncertain, given the long duration of road conces- ANAS has developed the analysis in four steps:
sions, it is advisable not to oversize the infrastruc-
• Risk identification: at this stage the possible
ture at the beginning (e.g. designing it for the first
causes and effects of risk factors were identified
10-year period), that would result in high CAPEX
and allocated to the party (public and/or private)
and related interest costs, but to foresee, both in
with more capacity to manage it;
the design and in the contract, an optional second
• Risk evaluation: at this stage the analysis seeks
phase of works for road widening.
to assess the materiality of the risks. The proc-
But if cost optimization concern mainly the
ess includes an evaluation of the probability of
engineering of the construction, in terms of the
occurrence and the related impact on the costs.
proceeds must be financially considered that a
The result from these first two phases is the risk
road is not only a road. A road is a service corridor
matrix, where any risk is identified, described,
where run all networks, from communication and
assigned and evaluated qualitatively with respect
power lines to oil and gas, enabling their mainte-
to probability and impact;
nance with easy access. A road is an avenue where
• Risk valuation: at this stage, taking into account
users are clients with needs to be satisfied (petrol,
the qualitative assessment of the previous stage,
food stations, shops and garages) and beneficiar-
it developed a risk map where, for the risks map-
ies of advertisings. A road opens new commercial
ping in the Significant and Serious area, it is per-
benefits to developing areas.
formed a quantitative evaluation, incorporating
Therefore, another aspect that is worth con-
the criteria assessed in the qualitative stage a
sidering are this “secondary incomes” that might
numerical approximation;
result even in being self-sustaining of the project
• Risk mitigation: solutions and mitigation tools
feasibility without the need for tolling, for example
have been proposed to optimize the manage-
advertising, service areas, etc. A useful reflection
ment of risks.
where there are social problems of tolling accept-
ance. Also in this case some measures can be taken The result applying this methodology, led to the
as widening the expropriation width to accommo- definition of a risk financial plan with annual pro-
date other services (paying use), arranging service visions that the public entity will use both to hedge
areas with suitable structures and advertisement against project’s risks and to get a better idea of

404

TIS2017_Book.indb 404 2/8/2017 8:05:19 PM

 the real cost of the project and its feasibility. The with or without ground anchors) are characterized
methods used for the quantitative evaluation are by the use of materials as concrete and steel. A plus
the main methods of statistics, such as Chi Square point of these solutions is the fact that they can
and Bootstrap. In some specific cases, when his- solve situations of high criticality, while among the
torical information about the event under consid- weaknesses there are:
eration was not available, experts meetings on the
• Need of careful selection of materials and high
specific issue were organized, analyzing with statis-
quality controls;
tical methods their answers.
• High costs and complex organizations, espe-
In order to obtain a correct balance of risks
cially during construction;
between the public and private entities, it was con-
• Some types of intervention require specialized
sidered a form of shared risks that was beneficial to
technologies and skilled labor;
the two parties involved. For example for the risks
• The works are poorly adapted to future addi-
associated with extra cost arising from land acquisi-
tions, for example in the case of road widening,
tion or extra costs for environmental offsets: the pri-
and eventual demolition is expensive, as it is the
vate entity will assume the extra cost of up to 120%
landfill disposal of materials
of the initial value; the extra cost between 120% and
• Works require maintenance (typically drain-
200% of the initial value will be shared between the
age interventions), but material’s degradation is
two parties, to the extent of 70% for the public entity
inevitable.
and 30% for the private sector; the extra cost over
200% will be entirely supported by the public entity. The proposed innovative solutions include
The correct balance of risks is crucial to the applications of “bioengineering” and they can be
success of a PPP operation and for this analysis alternative or integrative to traditional solutions.
have taken into account the point of views of all Among the plus points are noted:
involved parties and evaluated the market accept-
• The required materials are available locally. Other
ance of the distribution designed. The discussion
materials are metal nets, cables, rods and screws,
and pre-analysis phase therefore is essential.
geomats, geogrid and geosynthetics. Such mate-
It is important to point out that one of the success
rials are very low volume. This reduces transpor-
factors of the risk analysis is the level of advance-
tation costs;
ment of the detailed design. The more you deepen the
• Environmental integration is optimal;
technical studies, the more it reduces the uncertainty
• The execution of the works is simple, small in
about the estimates. In Colombia, the development
size, with less impact on traffic deviations;
requested of the technical studies for the fourth gen-
• Specialized supervision is required, but unskilled
eration of highway concessions is very high, which
labors are sufficient for their erection;
allowed a more accurate risk as well as projects
• Interventions are flexible and easy to extend or
assessment, in general. With this fourth generation
integrated in the future;
of PPP projects has been possible to improve the
• With proper maintenance durability is significant.
distribution of the risks of the concessions, solving
the critical points of previous generations who had Material used in bioengineering reduces the
a distribution of risks which in some cases proved costs of works: based on our experience, savings
to be critical. In the 4G concessions in Colombia, over traditional interventions range between 25%
for example, the risk for lower toll revenues is now to 30%.
borne by the public body. However the traffic study Other bioengineering solutions used on the
has been developed to an advanced level, thus reduc- Colombian projects have been rock slope protection,
ing the uncertainty of the risk. The same considera-
tion can be done on all risks for which the analysis
carried out has decreased the uncertainty.

3 INNOVATIONS IN CONSTRUCTION

Main constructive innovations concerned road

embankment slope stabilization and asphalt
mixtures.

3.1 Slope stabilization

The traditional works type (reinforced concrete
retaining walls on direct foundation or on piles, Figure 1. Example of actual concrete retaining wall.

405

TIS2017_Book.indb 405 2/8/2017 8:05:19 PM

 According to the polymer applied and the
amount thereof, can be obtained increases in pave-
ment service life up to 100%, against an increase
cost of the mixture up to 25%.
Analyzing graphs, it is observed that for a poly-
mer modified asphalt, the life cycle is much greater
than that of conventional binders. Defined the Load
Equivalence (LE) as the structural performance of
an asphalt layer compared to another, for modi-
fied asphalt mixtures with Reactive Elastomeric
Terpolymer (RET), this is about 1.5 compared to
a conventional asphalt. Economic evaluations that
Figure 2. Example of (a) Rock fall nets (b) Rock slope have been conducted show that the use of polymers
protection. with a LE greater than 1.2, represents a saving on
the final cost of the project.
Consequently:
ground slope reinforcing (soil nailing) and Rock
fall nets. • Polymers are the economical choice when traf-
fic volumes are high or climatic conditions are
tough and the project needs to include the main-
3.2 Asphalt mixtures
tenance to allow the return of capital invested;
It is believed that one of the oldest methods of • Requested service levels of the road can be pre-
paving was the so called “Roman road”. Created to served for longer, requiring less maintenance
facilitate communications within the empire, this interventions more spaced in time, which means
road was developed in various stages and some of better service index and a greater availability of
its stretches are still in good condition. Nowadays the road by users;
asphalt mixtures and concrete are the most usual • The polymer modified asphalt mixtures also
pavement construction materials because they pro- allow us to optimize the use of natural resources.
vide a good support and allow passage of vehicles In our case, we reduced the environmental
without major damages. impact reducing the thickness of the asphalt,
An asphalt mix is usually a combination of using quality stone materials. In the first cor-
asphalt and aggregate minerals. The relative pro- ridors delivered, the use of the Dense Hot Mix
portions of these minerals determine the physi- modified polymers, represented a 6% savings in
cal and performance properties of the mixture. the construction cost. Ate
Conventional asphalt mixtures have properties
that meet the normal requirements of pavements;
however, when severe climate conditions and high
4 TECHNOLOGICAL INNOVATIONS
traffic volumes occur, it is necessary to seek appro-
priate and economically viable technical solutions
Traffic, especially in metropolitan areas, has
that enable the concessionaire to meet service levels
become an everyday problem difficult to solve. An
required, both at the beginning and at the end of
issue that causes undesirable effects as increased
operating time, making it vital for the maintenance
travel time or failure times in public transport,
cost of the road.
intolerable air pollution and noise levels that seri-
Since a large percentage of car accidents that
ously affect health and has its counterpart in sig-
occur every day are related to the deterioration of
nificant economic losses. A problem accentuated
the roads, it is very important to prolong the life
in recent decades by the increasing use of the auto-
of pavements. This is achieved by studying poten-
mobile, a diffuse decentralization process and an
tial changes in the designs. The benefits of these
increasing number of trips and amplitude thereof.
developments would impact both road safety and
As regards the environment, the situation is also
project economy.
worrying. Transportation is, in this area, respon-
Polymer modified asphalts are conventional
sible for over 60% of carbon monoxide emissions,
mixtures which are added with a polymer that
50% of nitrogen oxides and 33% of hydrocarbons.
enhances the properties, producing:
Moreover, the consequences of pollution on the
• Increased adhesion properties even in presence health of European citizens represent a cost of
of high humidity; 0.5% of EU GDP.
• Increased resistance to deformation and Colombia is no stranger to this problem, due also
fatigue; to the effects of population growth with economic,
• Reduced susceptibility to thermal variations. environmental and quality of life consequences.

406

TIS2017_Book.indb 406 2/8/2017 8:05:20 PM

 The purpose of implementing an Intelligent collection and the transport of data, is always
Transportation System (ITS) is to allow the collec- advisable to make preliminary pilots projects, led
tion, storage, processing, analysis and distribution by a severe test plan, where it is verified that the
of information related to road operation, in order product complies with all features offered. The
to improve traffic management and achieve a more resulting information can be used to explore other
efficient and safe operation of the infrastructure, business alternatives.
generating more information to users, reducing The concession contract imposes the contractor
travel time, reducing the cost of fuel consumption to use the mentioned technologies.
and environmental impact, among others.
Any technological platform achieves its true
operating profit, and therefore the return on 5 INNOVATION IN FINANCIAL
investment, when it come across both the techni- STRUCTURING
cal components (equipment) as the operation and
maintenance processes, needed for an operating According to ANAS experience, it is proposed, as
strategy and continuous improvement. Technology a new alternative, to use the Italian practice in the
alone does not achieve the goals without processes financial model, called Financial Regulatory Plan
and people involved. (FRP), where toll and additional revenues remu-
The major subsystems required in the design of nerate only admitted project’s costs: return on
road concessions are: capital invested, depreciation costs, operation and
maintenance costs. In this way the FRP seeks to
• Telecommunications infrastructure: consisting
determine if the project’s costs are balanced, at the
of hybrid solutions with fiber optic as backbone,
end of the concession, by the income received from
different wireless systems and mobile technol-
tolls and eventual government grants, discounted
ogy solutions, to connect all elements along the
at WACC (Weighted Average Cost of Capital).
corridor;
The pursuit of this balance allows to review
• Electronic Toll Collection System (ETC);
what was executed annually and to determine pric-
• Operational Control Center (OCC): room where
ing adjustment policies. In addition, FRP allows
centralized management of all subsystems is
to calculate the value that the contracting entity
made;
should pay to the operator in case of early termi-
• Variable (or Dynamic) Messaging Systems VMS
nation of the concession’s contract in a given year.
(DMS), integrated with information screens to
This FRP is an official document that is part of
provide useful information for travelers;
the concession contract and every five years it is
• Video surveillance cameras (high-speed) for
revised to allow small adjustments ensuring regula-
vehicle identification, counting and speed con-
tory balance, without waiting to have unexpected
trol systems;
concerns at the expiration of the concession.
• Weighing system;
To calculate the financial balance should be used a
• Intelligent lighting systems in critical areas.
discount rate in accordance with market conditions
Perceived benefits in projects integrating ITS and projects with similar levels of risk and techni-
are: better information in real time, faster response cal features. This rate corresponds to the WACC,
to emergencies, greater fluency in circulation, with which the projected cash flows are updated to
increased road safety, automatic vehicle and freight calculate the IRR (Internal Rate of Return) of the
location, environment amelioration, and, at least project. Thereby it is considered that companies
but not at last, development of related industries. are financed by two mechanisms: through financial
Studies indicate a clear contribution of ITS to resources, which are denominated debt or through
develop a more sustainable mobility in the sense direct investment, denominated equity.
of increasing services without increasing impacts. In infrastructure projects worldwide equity ratio
It has been found that ITS not only contribute to to investment value ranges from 20% – 30% while
increased productivity of transport systems, with the rest is financed by third parties. The financial
the positive impact this has on the economy, but leverage can be of 3 types, multilateral agencies,
that is in itself a market with high economic poten- international banks and local banks.
tial and interesting future projection. The trend is Multilateral agencies promote infrastructure
to make the ITS design by subsystems, allowing development in different countries and make long
proper integration of the different elements. An term development loans, compared to that of com-
important role to meet the functional requirements mercial banks, that can have the same duration of
is to establish a governance model and organiza- the concession, which makes it much more profita-
tion with appropriate policies. ble for the concessionaire. But this type of loan has
Due to the diversity of manufacturers and sys- a much longer and difficult process to be obtained
tems today existing on the market, both for the than a commercial bank credit.

407

TIS2017_Book.indb 407 2/8/2017 8:05:21 PM

 International banks have a capacity of debt 6 TENDERING PROCESS
much higher than local banks. Moreover, because
its desire for risk diversification, a project like this ANAS, classified among the leading Italian con-
is framed within their target investment. Unfortu- tracting authority, has experienced since many
nately these banks do not usually pay in local cur- years the concessions structure using the project
rency, and given the length of payback, create a financing tool for large infrastructure projects.
currency exchange risk very high to the project. It has designed and optimized tender procedures
Local banks are a source of leverage in local aimed at involvement of private partners, in com-
currency, avoiding exchange risk. This source has pliance with the Italian legislation on public works
lower time limits than multilateral banks and gen- and developed with reference to the international
erally do not gather the entire loan individually. scenario in the field of PPP contracts.
There is the possibility of creating a structure This type of procedure has fostered not only
of debt leverage through a hybrid of the above, the financial support of private sector in con-
making an international bank to grant a loan in junction with the public one for the execution of
foreign currency at a good rate at a local bank, large infrastructure projects, but also allowed the
and this use these currencies for their interna- insertion of project proposals made by private
tional operations and fit a loan in local currency subjects within the National Development Plan
to the concessionaire with the guarantees on the framework. In this way the role of “Private Pro-
project. moter” has been fully integrated in the govern-
In the following are described various financial mental planning process, contributing concretely
instruments used by private subjects to finance to its definition and to the implementation of its
infrastructure projects. action plans.
In large PPP projects, especially in the UK, it is ANAS experience in the field of PPP has been
not unusual that the Administration actively contrib- successful also at international level. It is the case
ute to finding the best financing conditions thereof, of Colombia, where ANAS has supported the
requiring the opening of proceedings for the election local contracting authority to the development of
of debt providers, known as “debt funding competi- a PPP concession system of new generation, which
tion”. The competition may not be adequate under the government is now using for the implementa-
conditions of limited financial liquidity. tion of its important infrastructure investment
The Project bonds are specific bonds issues plan.
related to a project, with return of capital depend- Referring in particular to the characteristics of
ing on the cash flows generated by the project itself. the contract experienced in Colombia, it should
In order to make the Project Bond an instrument be stressed that ANAS, on the basis of the Italian
with market appeal, encouraging the financing of model, has contributed to the adoption of a meth-
public infrastructure by the private sector, have to odology which provides the use of a single model
be introduced amendments to legislation, as: of Cotract in all concessions tendering, consisting
of a general part, such as applicable to each pro-
• Companies incorporated in order to build and
cedure, and a special one provided with specific
manage an infrastructure project can issue
attachments for each project.
bonds, with the approval of the local supervi-
The procedure involves a first phase of pre-
sory bodies, including during the construction
qualification, targeted to the selection of the
phase of infrastructure. These bonds are nomi-
potential concessionaires that satisfy the techni-
native and cannot be transferred to people who
cal and financial requirements provided in the
are not qualified as institutional investors;
tender instructions. Such kind of restricted pro-
• Until the beginning of the operation of the
cedure, well consolidated in Italy, guarantees a
infrastructure by the concessionaire, obligations
qualified participation for the benefit of the con-
can be guaranteed by the financial system;
tracting authority and the subsequent production
• To introduce a tax incentive to the signing of
process.
Project Bonds, in Italy a tax system like the taxa-
To note also the great importance, in this con-
tion of government bonds is foreseen;
text, of an extensive promotional plan, aimed to
• Project bond issue can also be used to allow
attract the world of investors both in business
refinancing credit operations or senior debt
and banking. In the Colombian case, the suc-
acquired previously.
cess of such promotional initiatives, implemented
Another possible tool to limit the risks associ- through specific “road show”, has contributed
ated with PPPs could be a public guarantee fund significantly to accelerate the procurement and
in order to provide part of the bank debt of the awarding process of the first concessions of
Licensor. “fourth generation”.

408

TIS2017_Book.indb 408 2/8/2017 8:05:21 PM

 7 CONCLUSIONS can be of 3 types, multilateral agencies, interna-
tional banks and local banks. There is the possibil-
In the PPP acronym, the most important P is the ity of creating a structure of debt leverage through
third one: Partnership. It is important that the pub- a hybrid of the above. Project Bond can also be a
lic administration proposing the concession prop- sound instrument with market appeal, but amend-
erly plan the project and arrange a fair contract, ments to legislation might be needed.
financially attractive for the private sector. Also the It is important, for the success of a PPP tender-
risk analysis is a fundamental step in the study of ing procedure, to foresee a prequalification phase
a PPP project. The definition of a Risk Financial and an adequate promotional plan to attract
Plan with annual provisions that the public entity investors.
will use to hedge against project’s risks is suitable.
To analyze the economic viability of the project,
we must have an accurate analysis of traffic flows REFERENCES
with a careful projection, integrating future devel-
opment of the territory crossed. However, it is 2020 European Fund for Energy, Climate Change and
advisable not to oversize the infrastructure at the Infrastructure (Marguerite).
Ahmed, Priscilla, Project Finance in developing coun-
beginning, but to foresee, both in the design and tries: IFCb lessons of experience/International
contract, an optional second phase of works. Finance Corporation, Washington, D.C., 1999;
Another aspect that is worth considering when Balzer, L.A. (2001). “Investment risk: A unified approach
structuring a PPP initiative are the “secondary to upside and downside returns”.
incomes” that might result even self-sustaining of Beato, p. (2007). Road Concessions: Read Lessons from
the project feasibility without the need for tolling. the Experience of Four Countries rned. Substainable
Some examples of secondary incomes are incomes Development Department uses.
by rents, by advertising, by sub concession fees, etc. Bent Flyvbjerg, Mette Skamris Hola and Søren Buhl
Cost optimization concern mainly the engineer- (2002) “Underestimating costs in public works
projects: Error or lie?”.
ing of the construction. Main innovations con- CAF: “public infrastructure and private equity—
cerned slope stabilization and asphalt mixtures. concepts and experiences and in”America and Spain”,
For the first aspect, applications of “bioengineer- 2010;
ing” can be alternative or integrative to traditional CEPREDENAC—UNDP: the local risk management.
solutions. Material used in bioengineering reduces Notions and clarifications on the concept and prac-
the costs of works, saving between 25% to 30% over tice. 101 pg. 2003 (digital Version).
traditional interventions. Concerning the asphalt Chu, j. (1999). “The BOOT approach to energy infra-
mixtures, polymer modified asphalts increase pave- structure management: a means to optimize the return
ment service life up to 100%, against an increase from facilities”. Facilities. Vol. 17, no. 12–13.
Dams (2000) World Commission “dams and develop-
cost of the mixture up to 25%. Polymers are the ment: A new framework for decision-making”.
economical choice when traffic volumes are high Dragutin Nenezić, Branko Radulović Analysis of
or climatic conditions are tough, requiring less Finance Options and Models, and Financial Support
maintenance interventions. Measures for Public-Private Partnerships in Serbia;
Regarding the ITS technologies applied to road Essinger, J. & Rosen, J. (1991). Using technology for risk
network, studies indicate a clear contribution of management. London: Woodhead—Faulkner.
these in developing a more sustainable mobility. Fitch Ratings, “Report update issuance of straight
It has been found that ITS not only contribute to bonds—concessionaire of the West”, July 2012;
increased productivity of transport systems, with Flanagan, R., & Norman, G. (1993). Risk management
and construction.
positive impact on the economy, but that is in itself Grimsey, D., & Lewis, M. (2002). “Evaluating the risks
a market with high economic potential and inter- of public-private parnertships for infrastructure
esting future projection for the development of projects”. International Journal of Project Manage-
related industries. ment. Vol. 20.
Finally it is proposed at the contracting author- Guida al PPP—Manuale di buone prassi, European PPP
ity, but it isn’t compulsory, to use the Financial Expertise Centre and technical Unità di Finzanza di
Regulatory Plan (FRP), where toll and additional Progetto, 2011;
revenues remunerate only the return on capital http://ec.europa.eu;
invested costs, depreciation costs, operation and http://financialtribune.com;
http://www.aipcr.it;
maintenance costs. FRP allows to calculate the http://www.ani.gov.co;
value of the concession’s contract in a given year, http://www.ambteheran.esteri.it;
ensuring regulatory balance. In infrastructure http://www.caf.com;
projects worldwide equity ratio to investment value http://www.eib.org;
ranges from 20% to 30%. The financial leverage http://www.fitchratings.com;

409

TIS2017_Book.indb 409 2/8/2017 8:05:21 PM

 http://www.hm-treasury.gov.uk; Lara, (February 2003). Risk Management in Toll Road
http://www.ice.gov.it; Concessions. Massachusetts Institute of Technology.
http://www.imf.org; Leonardo Freitas, Project Finance and Public-Private
http://www.infomercatiesteri.it; Partnerships: A Legal and Economics View from
http://www.PIARC.org; Latin America Experience, Business Law Interna-
http://www.sace.it; tional, September, 2010;
http://www.simest.it; Mills, A. (2001). “A systematic approach to risk manage-
http://www.smbcgroup.com; ment for construction”. Structural Suvey. Vol. 19, N ° 5.
http://www.stradeanas.it; Private toll roads: lessons from Latin America, Samuel
http://www.utfp.it; Carpenter and Raphael Barcham, Polytechnic Uni-
http://www.worldbank.org; versity of Madrid (Spain), and Norbridge Inc., USA;
Infrastructure, Project Finance: “Financing of infra- Thomson Reuters Eikon (various analysis);
structure and project bonds”, 2010; Tool Roads Latin America Special Report, FitchRatings
Jefferies, M. (2006). “Critical success factors of public- 2009;
private sector partnerships”. Engineering, Construc- Toolkit for Public-Private Partnerships in Roads and
tion and Architectural Management. Vol.13, N ° 5. Highways, PPIAF—World Bank (2009).
Juan Benavides, reforms to attract private investment in UNDRO: Natural Disasters and Vulnerability Analysis
road infrastructure; in Report of Expert Group Meeting (9–12 July, 1979).
Kahneman & Tversky, a. (1979). “Prospect theory: An Geneva.
analysis of decisions under risk”, econometric. Vol. 47.
Lanfranco Senn, Il finanziamento delle Infrastrutture di
Trasporto;

410

TIS2017_Book.indb 410 2/8/2017 8:05:21 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Highway functional classification in CIS countries

V.V. Silyanov
State Technical University—MADI, Russia

J.I. Sodikov
Tashkent Automobile Road Institute, Uzbekistan

ABSTRACT: Road infrastructure are not only costly to build but also expensive to maintain in order
to adequately meet public expectation. Managing costly asset requires systematic approach which assures
adequate decision in each step of project life-cycle namely planning, designing, building and managing.
Presently, highway functional classification is widely used in developed countries however in CIS (Com-
monwealth of Independent Countries) countries planning, designing and managing of highways merely
based on technical (or administrative) classification. Authors are suggesting highway functional classifi-
cation which takes into account not only technical (or administrative) classification but also functional
classification. Recommended functional classification requires justification in parameters such as speed,
safety, riding quality and environment. The parameters were estimated by utilizing Analytical Hierarchy
Process. Optimum spending was defined as a function of road network size and importance weights of
highway functional classification.

1 INTRODUCTION regulation in order to develop functional classifi-

cation. In a proposed new classification one of the
1.1 Brief statistics important principles is to provide road use proper-
ties, which would be defined by the rate of speed
There are 9 member states in CIS which are Arme-
provision, comfort levels and traffic safety by road
nia, Azerbaijan, Belarus, Kazakhstan, Kyrgyzstan,
classes, Devyatoe & Vilkovay (2006). In the Repub-
Moldova, Russia, Tajikistan, and Uzbekistan.
lic of Belarus road classification has functional
The extent of public road network is about 1 232
assignments. There are four classes, namely: high-
850 km, land area 22 045,1 thousand km sq, number
ways, high-speed roads, ordinary roads and low
of bridges and overpasses is 95 593. Total number
class roads. Functional assignments are for high-
of vehicle fleet is 359 720,4 thousand units. There
ways—has purpose to serve long distance com-
are about 1 million people work in road sector.
mute without serving near by territories, for high
speed roads—has purpose to serve local commutes
with high speed, ordinary roads—has purpose to
1.2 Perception functional classification
serve for general purposes, and finally low class
The review of literature revealed that developed roads serve for small town and villages, Design
countries implemented functional highway clas- norms (2016). In Uzbekistan road classification
sification for planning, designing, and managing has road type, road function, road class (technical
whereas CIS (Commonwealth of Independent category), and administrative class. Road network
Countries) countries utilize based on technical consist of following types: highways, city access
classification. Functional classification serves for highways, main, local and internal roads, SHNK
which purpose the roadway is served such as for 2.05.02–07 «Automobile Roads» (2008). Accord-
mobility or accessibility. Technical classification ing to intergovernmental standard of CIS coun-
serves for technical definition of roadway such as tries, road technical classification according to
geometric parameters. There have been a number road use properties and access function is divided
of attempts to implement functional classification into the following classes: highways, high speed
in CIS countries. It was recommended to review roads, partial speed roads, and ordinary roads,
existing Russian classification of roads, especially GOST 33382–2015 (2015). One of the fist works
in urban areas by taking into account land use, on road classification in USSR was carried out by
traffic accidents spots, vehicles and pedestrians Babkov F. He suggested that there were need in
flows, traffic safety issues, including traffic speed national and functional classification of highways.

411

TIS2017_Book.indb 411 2/8/2017 8:05:23 PM

 However, he proposed to utilize national classifi- direct traffic flow and satisfy other requirements
cation, but he mentioned that highway functional of highways. Partially high speed roads have multi-
classification should be based on the traffic flow, lane without median strip; access allowed at grade
which may not necessarily be coincident with the and grade separation. Ordinary roads serve for low
national one, Babkov & Zamakhayev (1967). Sily- volume traffic therefore they have one or two traf-
anov (2004) reviewed classification of Russian fic lanes; access allowed at grade and grade separa-
highways and there are two separate standards for tion (see Table 1).
rural (SNIP 2.05.02–85) and urban roads (SNiP As a matter of fact, CIS countries inherit the
II-60–75). Both standards don’t have highway same history therefore standards and norms
functional classification. are based on research during that period with
Paraphantakul (2014) analyzed worldwide road some modifications to fit todays realities such as
classification and found that number of road increase of axle load and change in vehicle compo-
classes vary from 4 classes in Greece to 12 classes sition. Table 2 shows road network length by road
in South Africa. Author combined the relation- category. Majority of roads are IV and V category
ship structure and the classification categories which make up about 63% and 74% for Uzbekistan
from various literature and produced eight general and Russia respectively.
road classification themes: access control, road The problem with existing road classification is
surface, usage, administration, link role, place sta- that it doesn’t fully satisfy to manage adequately
tus, transport mode, and function. In European,
North American and Australian classification
themes have similarities such as function, access Table 1. Highway Classification according to GOST
and link role are the most recognized themes. In 33382–2015.
the study, there is overlapping between function,
access and link role themes, which may lead to a Design traffic Number of
Class Category volume* traffic lanes
biased conclusion.
Highways IA More than 4
>7
Speed roads Ib–
2 PRESENT ROAD CLASSIFICATION Partial speed IB >14 3 or 4
roads IIA 3−7 3
Intergovernmental standard GOST 33382–2015 IIb–
on road classification among CIS countries divide 6 − 14
highways into the following classes: highways, high III 1− 3
speed roads, partial high speed roads, and ordinary
roads. Highways are dedicated to safe and uninter- 2−6
rupted traffic flow with high speed and volume, Ordinary IV
roads 0.1 − 1 2
which serve for long distance commute. Access is
not allowed to a certain types of transport means, 0.2 − 2
pedestrians and bicyclists. Measures are taken to V < 0.1 1
avoid wild and domestic animals on the roads. There
are requirements to which highways must satisfy: < 0.2

– Division opposite direction by installing median *– Thousand vehicle per day (in numerator—actual number
strip or separate design of carriageway for each of vehicles, in denominator—passenger car equivalent).
direction;
– At least two traffic lanes for each direction;
– Avoid intersection at grade with roads, bicy- Table 2. Road network length comparison by road cat-
cle and pedestrian lanes, and also wild animal egory between Uzbekistan and Russia.
migration routes, domestic animal crossings;
Uzbekistan Russia
– Access provision by grade separation;
– Prohibit bicycles, motorcycles, tractors, and also Road Length, Length,
other types of transportation which may disrupt category km Percentage km Percentage
traffic flow, except machinery for road repair or
emergency situations; I 1246 3% 4800 1%
– Prohibit access heavy weight, dangerous and II 2363 6% 27800 5%
oversized vehicles; III 9800 23% 109800 20%
– Limit access points in grade separation. IV 21333 50% 320200 59%
V 7788 18% 81600 15%
High speed roads are allowed to design at grade Total 42530 100% 544200 100%
level but under condition that it would not cross

412

TIS2017_Book.indb 412 2/8/2017 8:05:23 PM

 road assets. Road classification basically served
for planning and designing purposes. While road
maintenance takes into account road category for
budget allocation and repair. But key question is
whether the road section corresponds to the category
which was assigned during design or not, what kind
of purpose does certain road section serve for? In
order to properly answer these questions, the authors
are proposing functional highway classification.

3 PROPOSED HIGHWAY FUNCTIONAL

CLASSIFICATION

3.1 Highway functional classification

Two primary transportation functions of roadways,
namely mobility and access, and describes where
different categories of roadways fall within a con-
tinuum of mobility-access. In addition to mobility
and access, other factors that can help determine
the proper category to which a particular roadway Figure 1. Highway Functional Classification taking
belongs—such as trip length, speed limit, volume, into account road category.
and vehicle mix. The concept of functional classi-
fication defines the role that a particular roadway – Collector:
segment plays in serving this flow of traffic through • Major Collector;
the network. Roadways are assigned to one of sev- • Minor Collector;
eral possible functional classifications within a – Local
hierarchy according to the character of travel serv-
Figure 1 shows highway functional classification
ice each roadway provides. Planners and engineers
in relationship with road category. As a road cat-
use this hierarchy of roadways to properly chan-
egory is primary principle for designing and main-
nel transportation movements through a highway
taining roads it would be appropriate to introduce
network efficiently and cost effectively, Highway
functional classification in line with it.
Functional Classification Concepts (2013). In Ser-
Highway functional classification serve not
bia, functional classification was proposed as pri-
only for identifying what function is playing cer-
mary and secondary functions due to the fact, that
tain road section but also importance of that sec-
unique basic function of a certain rural road sec-
tion. As one of the performance indicators AADT
tion is virtually impossible to achieve and, in any
(Annual Average Daily Traffic) or VKMT (Vehi-
case, would be spatially, economically and environ-
cle Kilometers Traveled) plays significant role in
mentally unacceptable. Therefore, certain mixing
budget spendings. In order to assign importance
of functional tasks is necessary and requires the
factors to each functional classification analytical
definition of primary and secondary function of
hierarchy process was utilized.
a certain road section, Maletin & Tubić (2015). In
other words, it’s logical that certain road section
may serve different purposes such in Serbian case, 3.2 Analytic Hierarchy Process
local road (direct access road) has primary func-
The Analytic Hierarchy Process (AHP) proposed
tion as direct access to property, and secondary
by Saaty (1980) is a very popular approach to
function as collection/distribution of traffic flow.
multi-criteria decision-making that involves
To avoid misunderstanding in practical usage and
qualitative data. It has been applied during the
to simplify data management the authors came up
last several decades in various decision-making
with simplified highway functional classification.
problems. The method uses a reciprocal decision
Based on literature review, the relationship between
matrix obtained by pairwise comparisons so that
number of classification and level of country devel-
the information is given in a linguistic form. The
opment the following highway functional classifi-
pairwise comparison method was introduced in
cation was proposed for CIS countries:
1860, Fechner (1965). Based on pairwise compari-
– Arterial: son, Saaty proposes the analytic hierarchy process
• Major Arterial; as a method for multi-criteria decision-making.
• Minor Arterial; It provides a way of breaking down the general

413

TIS2017_Book.indb 413 2/8/2017 8:05:23 PM

 method into a hierarchy of sub-problems, which methods, large number of comparisons when many
are easier to evaluate. It utilizes multiple criteria, alternatives exists, Ramanathan (2001), Taslicali &
and provides a simple process for weighting por- Ercan (2006).
tions of the hierarchy that is difficult to enumerate
directly. This method describes a general multi-
factor decision problem in way that decision tree 4 RESULTS AND DISCUSSIONS
in each of the hierarchy levels include some types
of criteria. The idea behind making decisions is to One of the key purposes of introduction highway
compare the relative importance of each criterion functional classification is to achieve optimal road
in adequate manner so that it can fit into general asset management. There are two main factors
concepts. They are some of the concerns regarding which affect road asset management. One is finan-
application of AHP in making pair-wise compari- cial constraint, the other one is time. To manage
sons between alternatives, the more alternatives the efficiently road assets at network level importance
more problems occur regarding the consistency of weights were assigned to each highway functional
the comparisons, and the assignments of the scores. classification by analytical hierarchy process. First,
An expert makes decision regarding the relative importance was assigned for each functional clas-
importance of each criterion and then specifies a sification such as arterial, collector, and local.
preference, which is rated on a scale from 1 to 9, for Intensity of importance varies from 1 to 9; 1 –
each decision alternative. If there are n alternatives, equal importance, 3 – somewhat more important,
then n*(n–1)/2 pair-wise comparisons are needed. 5 – much more important, 7 – very much more
Obviously, for practical application of AHP, the important, 9 – absolutely more important, 2,4,6,8
number of alternatives must be reasonable limited. – intermediate values used when compromise is
The result of AHP is a prioritized ranking that needed. Tables from 4 to 6 describe alternatives
indicates the overall preference of each alterna- and parameters values and pairwise values.
tive. The expert responsible for adequateness of In appendix 1 is given detailed description of
assigned that they reflect the importance of the pairwise matrix and R code for calculating analyti-
issues. The consistency of the judgments of the cal hierarchy process for highway functional clas-
expert can be measured with a Consistency Ratio sification. The result is shown in Table 3.
(CR). The CR is calculated as follows:
Table 3. Assigning importance weights.

CI =
( − n)
(1) Weight Arterial Collector Local CR
n −1
CI Weight 1.00 0.57 0.27 0.15 0.09
CR = (2) Safety 0.47 0.27 0.13 0.07 0.00
RI
Speed 0.33 0.22 0.08 0.03 0.02
where, CI – consistency index, λ max = the eigen- Riding Quality 0.14 0.08 0.04 0.02 0.01
value corresponding to the principal eigenvector Environment 0.06 0.01 0.02 0.04 0.00
n = the number of alternatives or criteria being
compared RI = the random index, a dimensionless Table 4. Alternatives and parameters values.
value that is a step function of n.
CR of 0.1 or less is considered acceptable. If Riding
a decision-maker’s responses fail the consistency Speed Safety Environment Quality
test, then the analyst must repeat iteration until (60–120) (0–10) (0–10) (0–100)
consistent responses are obtained. Applications
Arterial 120 10 6 90
of AHP highlighted in various transportation
Collector 80 6 8 70
projects evaluation such as Kang & Lee (2006),
Local 60 4 10 50
Azis (1990), Hagquist (1994), Kim & Bernardin
(2002), Masami (1995), Tabucanon & Lee (1995),
Kengpol (2002). A popularity of AHP method in Table 5. Parameters pairwise values.
solving multicriteria problems in infrastructure
projects, clearly related to easy to use and the way Riding
it prioritize alternatives. The logic behind priori- Speed Safety Environment Quality
tization is based on expert’s judgment. Besides in
Speed 1 1/2 5 4
decision-making the method can use quantitative
Safety 2 1 5 4
and qualitative data. Despite of advantages there Environment 1/5 2 1 1/4
are some disadvantages such as scaling alterna- Riding Quality 1/4 1/4 4 1
tives from 1 to 9, weakness of priorities estimation

414

TIS2017_Book.indb 414 2/8/2017 8:05:23 PM

 Table 6. Alternatives pairwise values.

Speed

Arterial Collector Local

Arterial 1 3 6
Collector 1/3 1 3
Local 1/6 1/3 1
Safety
Arterial Collector Local
Arterial 1 2 4
Collector 1/2 1 2
Figure 2. Optimum spending depending on functional
Local 1/4 1/2 1 classification and network size.
Environment
Arterial Collector Local
importance weight, analytical hierarchy process
Arterial 1 1/3 1/6 was utilized. Based on importance weights and net-
Collector 3 1 1/2
work size, optimum spending on road maintenance
Local 6 2 1
was recommended. Previously, the authors pro-
Riding Quality posed simplified road asset management system
Arterial Collector Local for developing countries in which one of key fac-
tors was traffic, Sodikov & Silyanov (2015). Traf-
Arterial 1 2 5 fic volume changes over time and doesn’t precisely
Collector 1/2 1 2 reflect the importance of road section. On the
Local 1/5 1/2 1 other hand, highway functional classification has
direct relationship with traffic flow, besides it has
Based on Table 2 and Table 3 the relationship functional purpose with importance weight. There
between importance weight and network size were are number of recommendations proposed:
found. It’s obvious that network size increases • Highway functional classification not only divide
from arterial to local. Whereas importance factor roads into classes like in technical category divi-
decreases from arterial to local. At the network level sion, but also each class has functional purpose.
and multiyear programming, it can be assumed Based on which road agencies would be able to
that optimal spending under limited budget should adequately and efficiently manage their assets.
be primarily focused on major and minor arterial, • Continuously surveying road network for iden-
major collectors and secondarily on minor collec- tification and homogeneous sectioning under
tor and local roads (see Fig. 2). highway functional classification, would lead to
Highway functional classification plays indis- more appropriate decision-making.
pensable role in not only in planning, designing, • Based on optimal spending curve one can build
and maintenance but also in budget allocating. In decision tree model to predict prioritization road
above-mentioned research works shows that there maintenance program at network level.
is overlapping in road function between arterial
and collector roads, and between collector and Further research will focus on defining detailed
local roads. Along with road inventory and condi- requirements for each functional classification,
tion data collection there is a need in functional actual division road network according the new
classification survey. classification and development GIS map.
The present classification in CIS countries,
mainly based on technical category, which doesn’t
satisfy up-to-date requirements of road asset man- REFERENCES
agement. The research proposes that functional
classification in line with technical classification Azis, I.J., 1990, “Analytic Hierarchy Process in benefit-
cost framework: A post-evaluation of the Trans-
would efficiently manage road assets. Proposed
Sumatra highway project”, European Journal of Oper-
highway functional classification takes into account ational Research. Volume 48, Issue 1.
two functions such as mobility and access. The five Ali Kamil T. & Sami E., 2006. The analytic hierarchy &
type of functional classification, namely major the analytic network processes in multicriteria deci-
and minor arterial, major and minor collector, and sion making: a comparative study. Journal of aeronau-
local roads were recommended. In order to assign tics and space technologies, Volume 2, Number 4.

415

TIS2017_Book.indb 415 2/8/2017 8:05:24 PM

 Babkov V & Zamakhayev M.1967, Highway Engineer- International Conference on Computing and Decision
ing, Mir Publishers, Moscow. Making in Civil and Building Engineering.
Chutipong P. 2014, Review of Worldwide Road Classi- SHNK 2.05.02–07 «Automobile Roads», 2008. Gos-
fication Systems, Conference: National Transportation komarhitektstroy the Republic of Uzbekistan,
Conference, At Bangkok, Volume: 9. https://www. Tashkent.
researchgate.net/publication/284371945_Review_ Silyanov V.V., Review of the Roads Appraisal Proc-
of_Worldwide_Road_Classification_Systems. Last ess in Russia, technical paper, 2004. http://www.
accessed: 1/11/17 transport-links.org/transport_links/f ilearea/
Devyatoe M.M. & Vilkovay I.M 2006, Methodology of documentstore/331_Valentin%20Silyanov%20
functional classification of roads for the purpose of Paper%20-%202.12.04.doc. Last accessed: 1/11/17
their modernization. Proceedings of roads and bridges, Sodikov J.I. & Silyanov V.V. 2015, Road Asset Manage-
Issue 16/2, Moscow 2006. http://www.complexdoc.ru/ ment Systems In Developing Countries: Case Study
ntdpdf/541958/dorogi_i_mosty_sbornik_vypusk_162. Uzbekistan, Science Journal of transportation, Mos-
pdf. Last accessed:1/11/17 cow-China-Vietnam, 2015, Special Issue 6.
Design norms 2016, TKP 45–3.03–19–2006 (02250), Tabucanon, M.T. & Lee, H., 1995, “Multiple Criteria
Ministry of architecture and Building of the Repub- Evaluation of Transportation System improvement
lic of Belarus. http://www.npmod.ru/3/Perechen_1/ projects: the case of Korea”, Journal of Advanced
tkp30319.doc. Last accessed: 1/11/17 Transportation, 29(1), pp 127–134.
Fechner G.T., 1965, Elements of Psychophysics. Vol-
ume 1, (Holt, Rinehart & Winston, New York.Infor-
mational-analytical publication, 2012. Status quo APPENDIX 1
and development of roads in CIS countries. http://
www.e-cis.info/foto/pages/19824.doc. Last accessed: Analytical Hierarchy Process in R
1/11/17
GOST 33382–2015, Public domain automobile roads.
Four parameters such as Speed, Safety, Envi-
Technical Classification. ronment and Riding Quality were used to assign
Hagquist R.F., 1994, “High-precision Prioritization importance weight to each highway functional
Using Analytic Hierarchy Process Determining State classification namely, arterial, collector and
HPMS Component Weighting Factors”, Transporta- local.
tion Research Record 1429, pp 7–14 ## First it’s needed to load libraries:
Highway Functional Classification Concepts, 2013. library(ahp)
Criteria, and Procedures. Federal Highway library(data.tree)
Administration. ## Then, loading input file hfc.ahp and look to
Kengpol, A., 2002, “The Design of a Support System
(DSS) to Evaluation the Investment in New Distri-
the file:
bution Centre Using The Analytic Hierarchy Process ahpFile <- system.file("extdata", "hfc.ahp",
(AHP)”, Capital Investment Model and Transportation package = "ahp")
Model. cat(readChar(ahpFile, file.info(ahpFile)$size))
Kim, K. & Bernardin, V., 2002, “Application of an Ana- ## Version: 2.0
lytical Hierarchy Process at the Indiana Department ## Alternatives: &alternatives
of Transportation for Prioritizing Major Highway ## # Here, speed is assigned (max-120 kmph,
Capital Investments”, Journal of the Transportation in-60 kmph), safety(max - 10, min-0),
Research Board 1816, pp 266–278 ## # Environment(max-10, min-0), Riding
Masami S., 1995, “Analytic Hierarchy Process (AHP)-
Based Multi-Attribute Benefit Structure Analysis of
Quality (max-100, min-0).
Road Network Systems in Mountainous Rural Areas ## Arterial:
of Japan”, International Journal of Forest Engineering ## Speed: 120
Vol. 7, No 1. ## Safety: 10
Mihailo M. & Vladan T., & Marijo V., 2015. Functional ## Environment: 6
Classification Of Rural Roads In Serbia, International ## Riding Quality: 90
Journal for Traffic and Transport Engineering, 5(2): ## Collector:
184–196 ## Speed: 80
Ramanathan R. 2001, A note on the use of the analytic ## Safety: 6
hierarchy process for environmental impact assess-
ment. Journal of Environmental Management, 63,
## Environment: 8
27–35 ## Riding Quality: 70
R source code for assigning importance weights, 2016, ## Local:
at github.com https://github.com/JamshidSod/HFC. ## Speed: 60
Last accessed: 1/11/17 ## Safety: 4
Saaty T.L. 1980. The Analytic Hierarchy Process, ## Environment: 10
McGraw Hill International. ## Riding Quality: 50
Seunglim K., Seongkwan M.L., 2006. AHP-based dece- ## #
sion-making process for construction of public trans- ## # End of Alternatives Section
portation city model: case study of Jeju, Korea. Joint

416

TIS2017_Book.indb 416 2/8/2017 8:05:24 PM

 ## ########################### ## pairwise:
########## ## - [Arterial, Collector, 2]
## # Goal Section ## - [Arterial, Local, 4]
## # ## - [Collector, Local, 2]
## Goal: ## children: *alternatives
## # A Goal HAS preferences (within-level ## Environment:
comparison) and HAS Children (items in level) ## preferences:
## name: Assigning Importance Weight ## pairwise:
## preferences: ## - [Arterial, Collector, 1/3]
## pairwise: ## - [Arterial, Local, 1/6]
## # preferences are defined pairwise ## - [Collector, Local, 1/2]
## # 1 means: A is equal to B ## children: *alternatives
## # 9 means: A is highly preferrable to B ## Riding Quality:
## # 1/9 means: B is highly preferrable to A ## preferences:
## - [Speed, Safety, 1/2] ## pairwise:
## - [Speed, Environment, 5] ## - [Arterial, Collector, 2]
## - [Speed, Riding Quality, 4] ## - [Arterial, Local, 5]
## - [Safety, Environment, 5] ## - [Collector, Local, 2]
## - [Safety, Riding Quality, 4] ## children: *alternatives
## - [Environment, Riding Quality, 1/4] ## #
## children: ## # End of Goal Section
## Speed: ## ###########################
## preferences: #########
## pairwise: ## Next is to calculate priorities (eigenvalues)
## - [Arterial, Collector, 3] and visualize our data
## - [Arterial, Local, 6] hfcAhp <- Load(ahpFile)
## - [Collector, Local, 3] Calculate(hfcAhp)
## children: *alternatives Visualize(hfcAhp)
## Safety: ## And finally is to present resuls in table
## preferences: AnalyzeTable(hfcAhp)

417

TIS2017_Book.indb 417 2/8/2017 8:05:24 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Efficient practices in railway ballast maintenance and quality

assessment using GPR

L. Bianchini Ciampoli, A. Calvi & A. Benedetto

Roma Tre University, Rome, Italy

F. Tosti & A.M. Alani

University of West London, London, UK

ABSTRACT: The need for effective and efficient railway maintenance is always more demanded all over
the world as the main consequence of aging and degradation of infrastructures. Primarily, the filling of
air voids within a railway ballast track-bed by fine-grained materials, coming up from the subballast layers
by vibrations and capillarity effects, can heavily affect both the bearing and the draining capacity of the
infrastructure with major impacts on safety. This occurrence is typically referred to as “fouling”. When
ballast is fouled, especially by clay, its internal friction angle is undermined, with serious lowering of the
strength properties and increase of deformation rates of the whole rail track-bed. Thereby, a detailed and
up-to-date knowledge of the quality of the railway substructure is mandatory for scheduling proper main-
tenance, with the final goal of optimizing the productivity while keeping the safety at the highest standard.
This paper aims at reviewing a set of maintenance methodologies, spanning from the traditional and most
employed ones, up to the most innovative approaches available in the market, with a special focus on the
Ground Penetrating Radar (GPR) Non-Destructive Testing (NDT) technique. The breakthrough brought
by the application of new processing approaches is also analyzed and a methodological framework is
given on some of the most recent and effective maintenance practices.

1 STATEMENT OF THE PROBLEM layers laying upon the subgrade, namely, ballast
and subballast. In more detail, ballast is a homo-
Freights, bulk goods and commuters travel every geneously hard-rock-derived graded material. Its
day between cities, terminals and production poles main functions are to resist to the stresses imposed
upon ballasted rail tracks, all around the world. by passing loads and transmit the vertical forces,
As a result of society development and techno- properly attenuated, to the subgrade. Thereby,
logical enhancement, the speed of trains is increas- ballast holds the primary roles of retaining the
ing, as well as the requested performance of the track in its correct position, absorbing the acous-
track-bed. On the other hand, to ensure safety of tic waves within the air-filled voids, and ensuring
the transport and highest standards of produc- proper drainage of meteoric water (Benedetto
tivity, the railway network needs to be effectively et al., 2016a).
and timely maintained, due to expected increased Differential track settlements are prevalently
deformation rates by the passing loads. related to the substructure, whose mechanical
A railway track-bed can be broadly divided into response, is the major concern of designers and
superstructure and substructure (Figure 1). The maintenance managers (Indraratna, 2016). Indeed,
former includes steel rails, fastening system and whilst the superstructure components react to
sleepers, whilst the latter is composed of granular forces elastically with negligible deformations, the
stress cycles imposed by the passing loads induce
relevant deformation into the granular layers com-
posing the substructure (Chrismer, 1985). This
cyclic loading determines a breakage of the sharp
corner of ballast grains and a fragmentation of the
weaker particles. Thereby, the grading of the sub-
structure changes (Selig & Waters, 1994; Ebrahimi
et al., 2012) with implications on the mechanical
Figure 1. Typical railway track-bed cross-section. behaviour and drainage capacity of the material.

419

TIS2017_Book.indb 419 2/8/2017 8:05:26 PM

 Furthermore, the air-filled voids can get polluted are carried out all along the track at discrete inter-
by the deposit of concrete and steel dust produced vals. Nevertheless, these methods have several dis-
at the ballast-sleeper and wheel-rail contact (Indrar- advantages. Indeed, visual inspections do not allow
atna et al., 2014), respectively, by the pouring of coal for subsurface investigation and understanding
dust from the passing freights (Huang et al., 2010; of the deep causes of the damage. On the other
Tennakoon et al., 2015), and by the upward migra- hand, drillings and diggings do not provide con-
tion of fine clay particles from the subgrade along tinuous data on the health conditions of the track,
with capillary water (Al-Qadi et al., 2010). Mostly, although they can return highly reliable informa-
regardless of the nature of the process, this filling of tion (Al-Qadi et al., 2008; Al-Qadi et al., 2010).
ballast voids by fine particles is referred to as ‘foul- Furthermore, these are labour intensive and time
ing’. When the ballast is fouled, especially by clay, its consuming (Shao et al., 2010).
internal friction angle is undermined, with serious Thereby, the recent focus is to move towards
lowering of the strength properties and increase of the use of time- and cost-effective technologies,
deformation rates of the whole rail track-bed. capable of performing rapid and non-destructive
In view of the above, a detailed and up-to-date inspection of the track-bed.
knowledge of the quality of the railway substruc- Amongst these technologies, two-dimensional
ture is mandatory for scheduling proper main- (2D) (Cho et al., 2006) and three-dimensional (3D)
tenance, with the final goal of optimizing the (Sun et al., 2014) laser scanners, based on an optical
productivity while keeping the safety at the highest method of investigation, are currently used for rail-
standard. way inspections. In more detail, 2D and 3D imag-
This study evaluates a set of maintenance ing methods have been used for assessing critical
methodologies, spanning from traditional and features such as size and shape of ballast particles.
most employed ones, up to the most innovative In the last decades, GPR has attracted great
approaches available in the market, with a special interest in the assessment of railway ballast condi-
mention to the Ground Penetrating Radar (GPR) tions. To that effect, there is a plenty of literature
Non-Destructive Testing (NDT) technique, whose studies proving the capabilities of such a non-de-
application in railways is increasingly attract- structive tool to overcome the limits of the tradi-
ing the attention of researchers, companies and tional investigation methodologies (Hyslip et al.,
administration. In Chapter 2, an overview of the 2003; Shao et al., 2010).
up-to-date methodologies for maintaining railway
ballast is given; hence, the GPR method, pros and
cons are discussed in Section 3; then, in Section 4, 3 GROUND PENETRATING RADAR
a scheme of efficient maintenance management is
proposed; conclusions and research open issues 3.1 Working principles
are finally presented in Section 5. GPR is a geophysical inspection technique based on
the transmission of Electromagnetic (EM) waves
towards a medium, e.g., soil, and the reception of
2 RAILWAY MAINTENANCE the transmitted or reflected signal. The behaviour
of the traveling wave depends on the dielectric
With regards to the aforementioned railway sub- properties of the medium itself, according to the
structure, ballast plays a crucial role in maintaining EM field theory. In more detail, the wave velocity is
serviceability of the track and ensuring its struc- influenced by the dielectric permittivity ε, whereas
tural capacity (Al-Qadi et al., 2008; Al-Qadi et al., signal loss and attenuation are mainly affected by
2010). The fouling process causes degradation of the electric conductivity σ. The magnetic perme-
the original substructure, and can compromise the ability μ is equal to the free space magnetic perme-
performance of the railway track system. At worst, ability μ0 for all the non-magnetic materials, and
such an occurrence can increase the risk of derail- does not affect the propagation of the EM waves.
ment (De Chiara et al., 2014). In practical terms, when the EM impulse emit-
Thereby, monitoring and maintenance stages of ted by a source encounters a dielectric contrast in
track-beds are relevant to ensure high performance the medium, it generates a partial reflection and
of the railway system over its service time, and transmission of the signal. By collecting such
granting the highest safety standards. At the same reflections through a receiving antenna, it is pos-
time, they are challenging tasks to be carried out sible to image the subsurface features, in both two
due to the significant longitudinal extent of these or three dimensions.
infrastructures. Nowadays, several different GPR systems
Traditional methods for the monitoring and are available in the market. Main differences are
assessment of ballast degradation are visual inves- related to the type of the antenna and centre fre-
tigations, punctual drillings and diggings, which quency of investigation as well as to the working

420

TIS2017_Book.indb 420 2/8/2017 8:05:27 PM

 system. A comprehensive presentation of the theo- By means of complex experimental and theoreti-
retical and practical features of GPR is given by cal efforts, Shao et al. (2010) proposed an automatic
Daniels (2004) and Jol (2009). algorithm for the classification of ballast condi-
tions, based on the application of Support Vector
Machine (SVM) onto processed GPR frequency
3.2 Railway applications
spectra. The method relies on the assumption that
The use of GPR for the investigation of railway the frequency spectrum is affected by the fouling
substructures has mostly developed over the last conditions, in a predictable way, which involves the
20 years. Originally, GPR surveys were carried out possibility to recognize the ballast quality by ana-
using ground-coupled antennas, with centre fre- lyzing the relevant frequency spectrum.
quencies usually below 500 MHz. Although these Fewer laboratory activities can be found in the
antennas do not provide high resolution, they literature, although they provide essential informa-
ensure deep signal penetration. In the last dec- tion to calibrate railway-based models through a
ade, low-frequency antennas have been replaced proper characterization of the surveyed materials
by higher frequency (1 GHz÷2 GHz) air-coupled (Shangguan et al., 2012; De Chiara et al., 2014;
systems, bringing many advantages in terms of bal- Fontul et al., 2014; Benedetto et al., 2016a; Tosti
last degradation assessment (Roberts et al., 2006). et al., 2016).
Indeed, air-coupled antennas are installed on track Also in view of the above studies, a relatively new
inspection vehicles, capable to monitor the track railway-related research branch began to capture
conditions continuously and without any interfer- scientists’ interest over the last years (Benedetto
ence with rail traffic (Al-Qadi et al., 2008; Fontul et al., 2016b), namely, the numerical simulation of
et al., 2014). On the other hand, higher frequencies both physical and dielectric properties of ballasted
provide greater resolutions and they allow detecting track-beds. Indeed, the simulation of the ballasted
changes in voids volume of ballast, which are likely structure in virtual environment leads to undenia-
linked to fouling phenomena (Roberts et al., 2006). ble advantages. One of the most outstanding is the
Different features of the GPR signals collected possibility to replicate and extend the experimental
in railway investigations have been taken into conditions, in terms of both frequencies of investi-
account by experts and related to several substruc- gation and track-bed characterization, by properly
ture quality parameters. In Hugenschmidt (2000), setting the numerical parameters, which is a mas-
a ground-coupled 900 MHz system mounted onto sive save of time and economical resources.
a cart and towed by a locomotive was employed
for surveying a railway section. A comprehensive
3.3 Advantages and limitations
data post-processing allowed the author assessing
the ballast thickness and identifying several zones Although GPR and NDTs in general often allow
of clay migration upwards. Olhoeft & Selig (2002) to overcome the limits of traditional approaches, a
used an instrumented vehicle equipped with sev- deep knowledge about their limitations is required.
eral air-launched antennas with different centre This is a simple rule to avoid overestimation of the
frequencies to perform a visual multi-frequency methods and misinterpretation of the results, which
analysis of GPR signals collected on both the cen- may lead to detrimental effects. Undeniably, the
treline and sides of the track. The authors argued employment of air-coupled system mounted onto
that the highest performances were reached with rail vehicles grants a traffic-speed continuous data
the 1 GHz antenna. They also showed how dif- collection, which involves no disturbance to the
ferent results coming from tests performed at the railway traffic and a comprehensive monitoring of
sides of the track may indicate stability issues. the track. Furthermore, a regularly scheduled GPR
A comparison between data from 1 GHz and survey is definitely much more cost-effective, if
2 GHz air-coupled antennas was performed by compared to traditional tests. It also holds very few
Roberts et al. (2006). It was noticed how the higher concerns about safety issues, which cannot be said
frequency was capable to detect differences in the for traditional track-bed drilling and trenching.
scattering of EM waves, due to the pollution of air In turn, GPR is not a direct method, as it meas-
voids within the ballast. ures the potential effects of fouling on dielectric
More recently, frequency-based studies on rail- properties of the whole track-bed. Since such
way fouling detection can be found in the literature. effects can be produced by multi-sources, special-
In Al-Qadi et al. (2008), the authors presented a ist expertise in the interpretation of the data is
multi-method approach to assess the ballast con- mandatory, even in the case of automatic methods.
dition, based on direct analysis, scattering ampli- Furthermore, as the EM behaviour of the ballast is
tude envelope, and Short Time Fourier Transform highly influenced by size distribution and mineral-
(STFT). A 2 GHz centre frequency antenna was ogy of the aggregates, a proper characterization is
used for the surveys. always required for calibration purposes. In such a

421

TIS2017_Book.indb 421 2/8/2017 8:05:27 PM

 framework, numerical simulation can represent a 2. Numerical simulation. Once the aggregates and
useful mean of integration for GPR. the subgrade are characterized, it is possible to
In general terms, GPR appears as a powerful reproduce the survey environment within a vir-
tool for reducing costs and simultaneously increas- tual scenario, numerically simulated. Thereby,
ing the effectiveness of railway ballast monitoring. synthetic GPR signals related to different foul-
Nevertheless, due to its limitations it cannot be ing conditions can be generated.
employed as a stand-alone technology at the cur- 3. GPR survey. Hence, GPR data can be collected
rent state of the art, and it grants the highest per- along the railway section or network. In view of
formance when integrated with other methods. the above, most effective radar configurations
are equipped with air-coupled antennas, with
centre frequencies ranging between 1 GHz and
4 SCHEME OF EFFICIENT 2 GHz.
MAINTENANCE 4. Signal processing. The collected signal requires
now a proper processing in order to remove
In this Section, a procedural GPR-based scheme noise and clutter and, more in general, to reduce
for performing efficient railway maintenance is the ambiguities and uncertainties. The process-
proposed. ing technique must be chosen with reference to
Once taken into account the complex envi- the specific model that is planned to be used.
ronment of railways and the relevant features 5. Model calibration. The parameters used in
of GPR, a prompt for an early-stage and non- the selected model need to be calibrated with
destructive fouling detection can be listed hereafter regard to the physical properties of the tested
(Figure 2). material.
6. Model application and fouling detection. The
1. Physical characterization of ballast. The first
model can now be applied to the database col-
step necessarily consists in a comprehensive
lected along the track. The expected output is
knowledge of the physical and dielectric proper-
an overall assessment of the track-bed, and the
ties of the rock-derived aggregates. Such infor-
identification of the most fouled areas.
mation can be derived from “ad hoc” laboratory
7. Crosscheck. The unambiguousness of the results
tests performed on materials excavated in situ,
can be significantly raised through a compari-
or it can be provided directly by the railway net-
son between the data collected on-site and the
work administration. In addition, if the minera-
simulated ones.
logical nature of the ballast is known, shape and
8. Definition of the maintenance priorities. If the
size of the aggregates can be assessed by 2D or
results are confirmed by simulation, a map
3D laser scan tests.
of ballast deterioration throughout the whole
track is now available, and therefore it is pos-
sible to prioritize maintenance actions.

5 CONCLUSIONS

This paper aims at reviewing existing practices

for the maintenance of railway assets and it intro-
duces a new methodology scheme for the applica-
tion of effective GPR-based maintenance actions,
based on the most recent and up-to-date process-
ing approaches developed in this area of expertise.
The first Section introduces the problem, which
is tackled more thoroughly in the second Sec-
tion. Insights on the potential of GPR are given
in Section 3, along with a short review of the most
acknowledged scientific evidences by literature. In
view of the above, several assumptions have been
made. Firstly, the most efficient and effective GPR
configuration for railway applications is composed
of air-coupled pulsed systems, mounted onto
rail vehicles and equipped with a high frequency
antenna. Secondly, accuracy and reliability of
Figure 2. Scheme of effective maintenance using GPR. GPR can reach proper standards when integrated

422

TIS2017_Book.indb 422 2/8/2017 8:05:27 PM

 with complementary methods. Accordingly, in the Fontul, S., Fortunato, E. & De Chiara F. 2014. evaluation
maintenance scheme proposed in Section 4, it is of ballast fouling using GPR, In proc. of 15th Inter-
worthwhile noting that the application of a GPR national Conference on Ground Penetrating Radar
model requires proper calibration by a physical (GPR 2014), Brussels, June 30 – July 4.
Huang, H., Tutumluer, E., Hashash, Y., & Ghaboussi, J.
characterization of the materials (e.g., performed 2010. Laboratory validation of coal dust fouled ballast
by laser scanner), as well as a crosscheck by numer- discrete element model, In Paving Materials and Pave-
ical simulation, to validate the results. To that ment Analysis, pp. 305–313.
effect, an integration of different methodologies Hugenschmidth, J. 2000. Railway track inspection using
would seem the best solution. GPR, In Journal of Applied Geophysics, 43, pp.
147–155
Hyslip, J.P., Smith, S.S., Olhoeft, G.R. & Selig E.T. 2003.
ACKNOWLEDGEMENTS Assessment of railway track substructure condition
using ground penetrating radar, in Proc. Annu. Conf.
AREMA, Chicago, NY.
This work has benefited from the network activi- Indraratna, B. 2016). 1st Ralph Proctor lecture of ISS-
ties carried out within the EU funded COST MGE. Railroad performance with special reference to
Action TU1208 “Civil Engineering Applications ballast and substructure characteristics, In Transporta-
of Ground Penetrating Radar.” tion Geotechnics, 7, pp. 74–114.
Indraratna, B., Nimbalkar, S., Coop, M. & Sloan S.W.
2014. A Constitutive model for coal-fouled ballast cap-
REFERENCES turing the effects of particle degradation, In Computer
and Geotechnics, 61, pp. 96–107.
Al-Qadi, I.L., Xie, W. & Roberts, R. 2008. Time-frequency Jol, H. 2009. Ground penetrating radar: theory and appli-
approach for ground penetrating radar data analysis to cations, In Elsevier ed., Amsterdam, The Netherlands.
assess railroad ballast condition. In Research in Non- Olhoeft, G.R. & Selig, E.T. 2002. Ground penetrating
destructive Evaluation, 19, pp. 219–237. radar evaluation of railway track substructure condi-
Al-Qadi, I.L., Xie, W., Roberts, R. & Leng Z. 2010. Data tions, In 9th International Conference on Ground Pen-
analysis techniques for GPR used for assessing railroad etrating Radar (GPR 2002), Santa Barbara, California,
ballast in high radio-frequency environment. In Journal April 29–2 May.
of Transportation Engineering, 136(4), pp. 96–105. Roberts R., Rudy J., Al-Qadi I.L., Tutumluer E. & Boyle
Benedetto, A., Tosti F., Bianchini Ciampoli L., Calvi J. 2006 Railroad ballast fouling detection using ground
A., Brancadoro M.G. & Alani M.A. 2016a. Electro- penetrating radar – a new approach based on scatter-
magnetic characterization of clean and fouled railway ing from voids, In 9th European Conference on NDT.
ballast through gpr signal processing and numerical Selig E. T. & Waters J. M. 1994. Track geotechnology and
simulation. In Construction and Building Materials, substructure management. In Thomas Telford ed.,
In press. London.
Benedetto, A., Tosti, F., Bianchini Ciampoli, L., Pajew- Shangguan, P., Al-Qadi, I.L. & Leng Z. 2012. Ground-
ski, L., Pirrone, D., Umiliaco, A. & Brancadoro, M.G. penetrating radar data to develop wavelet technique
2016b. A simulation based approach for railway appli- for quantifying railroad ballast-fouling conditions, In
cations using GPR. In proc. of 16th International Con- Transportation Research Record, 2289, pp. 95–102.6
ference on Ground Penetrating Radar (GPR 2016), Shao, W., Bouzerdoum, A., Phung, S.L., Su, L., Indratatna,
Hong Kong, June 13–16. B. Rujikiatkamjorn C. 2010. Automatic classification
Cho, G.C., Dodds, J. & Santamaria, J.C. 2006. Particle of ground-penetrating-radar signals for railway-ballast
shape effectson packing density, stiffness and strength assessment, In IEEE Transactions on Geoscience and
of natural and crushed sands. In Journal of Geotech- Remote Sensing Vol. 49(10), pp. 3961–3972.
nical and Geoenvironmental Engineering, 132(5), pp. Sun, Y., Indraratna, B. & Nimbalkar S. 2014. Three-
591–602 dimensional characterisation of particle size and
Chrismer, S. M. 1985. Considerations of factors affecting shape for ballast, In Géotechnique Letters, 4 (3), pp.
ballast performance. In American Railway Engineering 197–202.
Association (AREA), Bulletin 704, AAR Research and Tennakoon, N., Indraratna, B., Nimbalkar, S. & Sloan
Test Department Report No. WP-110, pp. 118–150. S.W. 2015. Application of bounding surface plasticity
Daniels, D.J. 2004. Ground Penetrating Radar, The Insti- concept for clay-fouled ballast under drained loading,
tution of Electrical Engineers, London. In Computer and Geotechnics, 70, pp. 96–105.
De Chiara, F., Fontul, S. & Fortunato E. 2014. GPR labo- Tosti, F., Benedetto, A., Calvi, A., & Bianchini Ciampoli,
ratory tests for railways materials dielectric properties L. 2016. Laboratory investigations for the electromag-
assessment, In Remote Sensing, 6(10), pp. 9712–9728. netic characterization of railway ballast through GPR.
Ebrahimi, A., Tinjum, J.M. & Edil T.B. 2012. protocol for In Proc. 16th International Conference of Ground
testing fouled railway ballast in large-scale cyclic triax- Penetrating Radar (GPR 2016), Hong Kong, June
ial equipment, In Geotechnical Testing Journal, 35 (5), 13–16, 2016.
pp. 1–9.

423

TIS2017_Book.indb 423 2/8/2017 8:05:27 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

BIM experience in infrastructural large projects: Doha Metro—

Al Jadeda Station al Matar B case study

A. Ingletti, M. Scala & L. Chiacchiari

3TI PROGETTI ITALIA SpA, Rome, Italy

ABSTRACT: Civil engineers and architects typically outline projects and construction data using
technical drawings. At the beginning of the XXI Century, the development of new modeling technologies
has led to introduction of a new intelligent approach to 3D design: the Building Information Modeling
(BIM). BIM enables to improve the planning, design, construction and management of infrastructure
or buildings for the entire life cycle. 3TI is involved in Doha Metro Red Line South Underground main
stations project, as consultant in architectural finishes and MEP works of Al Jadeda Station al Matar
B. BIM process has been efficiently applied. This paper wants to underline that all project phases have
been studied using this new technology, from the analysis of the two-dimensional drawing, through the
construction of a 3D model, achieving an high detailed level for every single designed part. All specialists
involved in the project have the knowledge of the process and have to manage their updating in order to
reach a global understanding. Design coordination is the responsibility of all parties (meant as specialists
of different sectors) as an ongoing process until the completion: the outputs are not simple representa-
tions but a set of information. Definitely, this paper discusses the importance of BIM technology in Civil
Engineering and Architectural sectors, as an important mean for the time optimization compared to the
traditional approach and the reduction of costs, thanks to the prevision and the resolution of the issues,
during the planning through the interface checking. The analysis of the case of study highlights as the
coherence of BIM procedure application enables to adopt a maintenance strategy leading to a longer life-
cycle of Al Jadeda Station.

Keywords: BIM process, interface checking, maintenance strategy

1 INTRODUCTION The design and construction practices can be

digitally managed by BIM tools, thus the whole
1.1 BIM System life-cycle of the project can be monitored.
The composition of all the items and data of a
The Building Information Modeling technology
project represented by the System Implementation
represents a powerful process of development and
phase gives an integrated and structured model,
use of a computer software model able to integrate
threedimensional, smart and ultra-detailed, able to
data from several users in one Building Informa-
cover several topics:
tion Model. This “Model Based” methodology,
as called in AGC (2006), encloses a three-dimen- − Evaluation
sional, geometric, object-oriented representation − Collaboration
of the project and has data attributes making the − Implementation strategy
model “intelligent”. − Cost and Time
Collaboration between professionals and all the − Quality and Safety
figures involved in the project is ensured, thanking − Deliverables
to the flexible environment of work. − Design and Documentation
Indeed, design and project data are well-struc- − Contract administration
tured by the use of several software for analysis − Commissioning.
(such as Risa 3D, RAM, STAAD, ETABS), design
(Revit and Architectural Desktop, Dynamo Studio, As highlighted by Succar (2008), BIM process
Bentley’s MicroStation…), coordination, detailing assures interconnections within different fields:
and rendering software (Navisworks, Tekla Struc- procedures in BIM, players and deliverables are
ture, and Graphisoft). interlocked, so that guidelines and best practices,

425

TIS2017_Book.indb 425 2/8/2017 8:05:29 PM

 professionals’ activities and technologies (for soft- Particularly, the consolidated experience of
ware, networks and communication) do not rep- 3TI enables to claim that BIM technology is actu-
resent self-contained compartments, but mutually ally a pervasive focus on information, not limited
enrich each other. to a representation. Thus, as per the Company’s
The above mentioned mechanism of “knowledge understanding, in order to optimize the power of
transferring” within sectors leads to relevant techni- BIM, it is important that model, workflow, family
cal, operational and business advantages, thanks to and design settings are set up at the beginning, in
a greater control on the projects components. Par- order to converge toward the desired central point
ticularly, a correct and clear implementation of all of clear management of the complexity of the
the information ensures a successful development project.
of the project and a reduction of modification and All design phases have to be thought in
reissues, which results in efficiency maximization BIM philosophy and of course, all specialists
and time saving. In this way BIM can be defined involved in the project have to know and manage
a model checker, since the first phases of project BIM technology in order to achieve the best
implementation up to the operational phase, and benefits.
more, as for example manufactured product infor- Rightly related to the latter point, of advantages
mation, can be uploaded into the BIM database can be listed a few:
too, and this allow even to budget and schedule
• designing only once;
maintenance.
• simultaneous updating of plant, sections and
Essentially the BIM is not simple software but
elevations;
a complex management system in planning steps
• functional capability on improving cost analysis;
of a building or infrastructure. The 3D modelling,
• smaller project team necessity;
in fact, includes dimensional, quality, timing and
• advanced understanding of design
coordination data, and along with outputs pro-
constructability;
vided, such as work order flows, costs, accounts,
• interference checking, with finding “mistakes”
and financial elements, it supports the decision-
before they are made;
making processes. Therefore, the key business driv-
• revisions in much less time.
ers can now rely on Just-In-Time (JIT) delivery
of materials and equipment, which is an outcome In this sense, the integration within parties is
realizable by complete coordination prior to the important and 3TI successfully created a collabo-
start of construction, very early during planning rative BIM environment. This approach has been
stage. In addition, the possibility to simulate main- adopted within both the Professionals hierarchy
tenance procedures allows to asset a thought plan and technical staff, enhancing also the staff-tasks
or verify the impact of retrofits or maintenance in using the BIM solution day by day. The design
works in advance, besides aligning the decisions, models have been processed through an iterative
with the goal of wanted high maintainability-levels. activity of model-based design-coordination-
Lifecycle benefits regarding maintenance are guar- training, to insure the constructability with greater
anteed, and a robust management strategy can be focus on clashes.
then construct for the adoption of BIM.
2 DOHA METRO PROJECT
1.2 BIM experience in large projects
2.1 Overview on the project
3TI boasts numerous valuable experiences in the
preparation of projects of major structures and Doha Metro is a singular and prestigious project
infrastructures in BIM system. of the new passenger rapid transit system in
A large project carried out with BIM technolo- Qatar’s capital city, which is actually under con-
gies is the new 520-bed hospital in La Spezia. This struction and in the near future will establish easier
fan-shaped building opens up to the sea, in order and more convenient connections within the city’s
to not interfere with the panoramic view. Public central areas as well as the outskirt areas of Doha
spaces have been designed, along with a building (Figure 1). The aim of the project is to promote
of radiation therapy and about 750 parking spaces public transportation as a valid alternative to pri-
carefully hidden by a series of terraces. vate means of transport.
Furthermore, 3TI has developed five stations The planned construction stage is divided into
within the development plan of the metro lines at two phases: the first will see the construction of
Doha (Qatar), as detailed in the next paragraph. three of the four lines (Red, Gold, and Green line)
On these and other projects, all the complexity and 37 stations. These lines are expected to be open
of BIM technology has emerged, together with its to the public in late 2019. The second phase will be
great functionality. completed by 2026, and will involve the expansion

426

TIS2017_Book.indb 426 2/8/2017 8:05:29 PM

 Every single element, is included in the book
described by means of schedules composed by
designs and detailed descriptions, this finalized
to realize an efficient construction site, with a
high level of quality control. Modular forms are
described, along with all the elements’ features,
such as shelters, ceilings, walls, floors, and the lay-
out configurations, equipment, wayfinding and
furniture.

2.3 The 3TI BIM Design Stations

Figure 1. A panoramic view of Doha (Jon Bowles 3TI has been involved in the project of metro Doha
photo-gallery). for the design of many stations of red line and
green line. The red line, called “Cost Line”, runs
from the north to the south for 40 km, has 17 sta-
of the Phase 1 lines, and the construction of an tions and connects the city center to the Hamad
additional one, the Blue Line. Others 72 stations International Airport. The green line, called “Edu-
will also be built. cation Line” runs from the east to the west, has
Within the Qatar Integrated Railway Project 8 stations and passes through the Education City.
(QIRP) the appointed Architect UNStudio has The Company has developed five stations of
developed the new Doha Metro Network design the Doha metro project, four of the red line (Al
for the 35 stations in package Phase 1, and around Doha Al Jadeda station—AMB, Oqba Ibn Nafie
60 stations in package Phase 2. station—AME, Al Matar station—AMD, Umm
Doha metro project requires a careful planning Ghuwailina—AMC) and one of the green one (Al
to respect the strict timing and to minimize the Mansoura station—Al Khubaib, Figures 2–3). On
environmental impact. Thus, in order to guarantee the basis of stated criteria and defined elements’
flexibility at the project implementation, UNStu- characteristics by means of the Branding Man-
dio has drafted an Architectural Branding Manual, ual, 3TI Progetti has developed phases M2, M3,
which includes a set of design guidelines, architec- M4 for each station, falling within its assignment
tural details and material outlines. The Architect (Tables 1–2).
has produced this manual also for assuring the Al Doha Al Jadeda station is the pilot station
spatial quality and clarity of the network. All the and any other stations have been developed on the
appointed Design&Built contractors have to refer basis of this major project, by reaching in the end
to it, and respect the architectural of the stations, the same level of detail and definition of the M4
which reflects the heritage of the country in shape phase.
and colours.
Likewise, the manual establishes the phased
project timetable, so that the entire Project life
cycle is divided into four stages. Specifically every
phase of the project design is named:
• M1 – Concept Phase
• M2 – Preliminary Design
• M3 – Detailed Design
• M4 – Construction Design.
This staging is reflected on the items of which Figure 2. Doha Metro Project—Red Line.
the each station is composed.

2.2 Branding Manual

As basis for the design phases UnStudio has pro-
vided a detailed manual, which includes guidelines
and practical specifications to be followed and
respected for the project.
The Manual represents the guidance on the pro-
duction of various elements which compose the
station units, up to the more elementary detail. Figure 3. Doha Metro Project—Green Line.

427

TIS2017_Book.indb 427 2/8/2017 8:05:29 PM

 Table 1. Red line stations.

Al Doha Al Jadeda (AMB)

Station
name Item Phase

Landscaping M3–M4
Architectural detailed design M3–M4
and finishes services
Entrances and Pop ups M2–M3–M4
Signage M3–M4
BOQ M3–M4
Video animation
Oqba Ibn Nafie station (AME)
Station
name Item Phase
Landscaping M2
Architectural detailed design M4 Figure 4. The general simplified Station Components
and finishes services (Branding Manual) - Landscape around.
Entrances and Pop ups M2–M3–M4
Signage M3–M4
BOQ M3–M4 2.4 Al Jadeda Station Al Matar B case study
Al Matar (AMD) Al Doha Al Jadeda station is placed on the junc-
Station
name Item Phase tion Al Doha Al Jadeda Street—Bring Road. The
station serves a mixed-use area, that includes a
Landscaping M2 Civil Defence building, a main police station,
Architectural detailed design M4 hotels, retail and residential buildings.
and finishes services 3TI has been awarded for the design finishes,
Entrances and Pop ups M2–M3–M4 structures and Mechanical, Electrical and Plumb-
Signage M3–M4 ing (MEP) Engineering services for items B—En-
BOQ M3–M4 trances and Pop ups; the same services have been
Umm Ghuwailina (AMC) carried out for Item A - Station Box, except struc-
Station tures. In both cases have been developed signage
name Item Phase
and bill of quantity (BOQ).
Signage M3–M4 Al Doha Al Jadeda station is on four levels and
BOQ M3–M4 has four public entrance pods with vertical circu-
Video animation lation, single lift, stairs and two escalators; these
elements connect the station box to the concourse
level, which includes staff and public facilities as
prayer rooms, toilets, etc. and opens vertical cir-
Table 2. Green line stations.
culation by stairs and escalators to the platform
Al Mansoura (Al Khubaib) level. The station box results in a platform level
Station (island type) with single tracks and three sections
name Item Phase of subways.
There are four main entrances. Entrance N°1
Landscaping M2 is adjacent to the station concourse and it is con-
Architectural detailed design M4 nected to Entrance N°2 trough a subway under Al
and finishes services Matar Street to North East, and Entrance N°2,
Entrances and Pop ups M2–M3–M4 in turn, is connected by a subway to North West
Signage M3–M4 to Entrance N°3. The Entrance N°2 has a short
BOQ M3–M4
length and will be connected with Entrance N°4
after a subway extension. This last entrance will
be constructed in a different time, but this solution
For better and more effective management of will not have negative effects on the pedestrian area
projects and design files, each station has been or affect emergency escape conditions.
splitted into 3 items: Item A—Station Box, Item B Materials are fundamental components of archi-
- Entrances and Pop ups and Item C—Landscape, tectural fit-out work materials specified within the
as can be seen in Figure 4. Architectural Branding Manual. Modularization

428

TIS2017_Book.indb 428 2/8/2017 8:05:30 PM

 and prefabrication is the base of each single ele- In addition to technical drawings, many render-
ment defined in the Manual. This strategy aims ing and videos, to easier and clearer explain the
to facilitate construction operations, developing developing of the station project to the Client,
dimensionally and functionally flexible products. have been recorded. These videos were also used to
Prefabrication means to realize functional parts publicize the project to a non-technical audience.
of a work in advance and outside construction
site, giving the opportunity to reduce costs or
2.5 BIM Environment
complexity.
This approach is applied to different stages of Designing in “BIM oriented” way means commu-
the Al Jadeda station project and the BIM tech- nicate, through the exchange of a unique model,
nology represented the best way to design and pre- with colleagues and partners, without loss of
assembly integrated system units. BIM permits to quality. BIM allows indeed collaboration between
create parametric elements to repeat endlessly: the designers and also interoperability between soft-
structural pieces are controlled by the parametric ware tools.
structure within properly called families, allowing In the presented Al Jadeda Station case study
their forms to be automatically generated and later several software for analysis, design, coordination,
fabricated (Figure 5). detailing and rendering (mostly Revit, Naviswork
In Al Jadeda Station, petal panels fill all areas and Dynamo) have been used. By means of these
around the ceilings’ vaults, which are the main tools, it is possible to build a unique big model, by
characteristic elements of the local architecture linking many objects and sources together, also
sinking its roots in the traditional heritage. They attaching attributes and merging different file
are modelled for example by modular elements in types. Follows an in-depth examination about soft-
the shape of petals (Figure 5–6). ware mentioned before.
The architectural design is not the only techni- The entire project has been developed in Revit.
cal sector which has been developed by means of This station is a very complex architectural and
BIM applications. MEP systems in fact, have been engineering object, and designing it using tradi-
designed in BIM: ducts for HVAC systems, piping tional software would have been certainly much
runs for water and gas supply and disposal; routing more complicated in planning, time consum-
trays and control boxes for electrical and commu- ing control, reports production and technical
nication systems and all the related activities have management.
been carry out by BIM technology. Revit is a software to design with paramet-
ric modelling that, unlike Autocad, works not
with simple lines, but with well-defined objects
of complex shapes. It works in three dimensions
and allows multiple users to work simultaneously
on the same document through an internal system
called worksets. This system consists on a server-
based method of work-sharing which allows team
members to work on a local file (authorized based
on its role). Every update of secondary file is syn-
Figure 5. Shape and families characterization (3TI
chronized on the main file. This process allows a
Progetti). parallel constant updating. The whole concept of
BIM includes Revit Architecture for planning and
architectural design, Revit Structure, for structural
management of the project and Revit MEP for the
plants, all the three active simultaneously on the
same project and, therefore, on the same file. Once
an amendment is made on the 3D model, the soft-
ware modifies directly and simultaneously plans,
sections, elevations and details’ sheets. The design
simplification and flexibility is facilitated by using
templates and libraries that are the basis for defini-
tion of an object model and its assembly. “Fami-
lies” are the parametric components that make up
the graphics system for designing complex sets (fur-
niture and interior design) and simple ones (walls
and pillars). Even the printing process of docu-
Figure 6. Doha vault-game (UNStudio website). ments is more simple and rapid (automatic scales

429

TIS2017_Book.indb 429 2/8/2017 8:05:31 PM

 both on the basis of line weights both according to
the scale of views).
The BIM new approach reduces planning man-
agement time during the whole design process.
In this way, areas and volumes, energy analysis,
lighting and automatic material take-offs doable
can clearly be defined at the earliest stages of the
project. The software is intended for all experts
involved in architectural and engineering planning
and technical designers of construction companies,
because its main objective is to control the project Figure 7. Structure-Duct interference detection (3TI
evolution from concept to detailed design, keeping Progetti).
control on time and costs.
Naviswork is then a software able to integrate
the design itself with the as-built world. It enables
to have control of the project from a global point
of view: the holistic view of the project undoubt-
edly allows managing models, identifying and
clashing the models for interference, and virtu-
ally constructing a building using a construc-
tion timeline. Within the many potentialities and
resources of the tool, it can be mentioned that
it offers the possibility of live interactive walk-
throughs, to be used as demonstration for clients Figure 8. Plumbing fixtures (floor drain)-Structural
or contractors. plinth floor clash detection (3TI Progetti).
Al Jadeda Station project has been coordinated
also by means of the use of this software, particu- Dynamo is finally a software able to develop
larly for the revision, simulation and analysis of complex parametric geometries, and to make the
the modelling. In the studied case Naviswork has using of all kind of data-processing more effective
been of great importance in detection, identifica- and accurate than before. It provides faster solu-
tion and management of interferences in a more tion of problems regarding geometry and mod-
effective way. The main file contains external links els, and is able to transform simple drawings into
of all different technical files (drawing files, MEP interoperable workflows for analysis, documents
files, structural files, etc.), on which the experts con- and simulation. This software defines various text-
stantly and simultaneously work. In this way Navis based programming languages, creates relation-
“annotates” every updating of the files and allows ships and sequences of actions that make custom
locating and predicting problems before building algorithms.
(imperfection detected between a structural com- As regards the case study, 3TI has used Dynamo
ponent and airducts can be seen in Figure 7). software to construct the basic geometry of the
When it identifies interference, under a key- modular element “vault”, both in the platform and
parameter finding, a circle in red color is showed in subway level, taking advantage of the software’s
(Figure 8) around the clashing area (also com- ability in processing parametric elements that can
ments and annotations can be added, which will be repeat endlessly. As known, the delivery of the
visible in every file of the projects too). In the Al final documents of a project involves the prepa-
Jadeda Station, for example, Naviswork has helped ration of a huge number of documentations and
to identify interference between the ceiling and an drawings. The use of this software has enabled 3TI
utility duct: during the modelling phase, MEP engi- to significantly reduce time in the preparation of
neers had updated the size of duct in a larger one, title blocks of the drawings, as once defined the
and it was interfering with ceiling at the moment scheme, the patterns, matrices or images positions,
of analysis; consequently the height of ceiling have then automatically and rapidly, drawings’ layouts
been lowered to solve the defect. in all deliverable are completed.
BIM tools have been useful also for determin-
ing the right positions of catwalk accesses for
2.6 BIM—More than 3D
inspection: maintenance of plants is guaranteed
by means of this 90 cm high technical floor, and One of the most important innovations offered by
the panels of finishes are interrupted at wanted BIM in the planning process is the introduction
intervals which respect the fusion of functionality of the “fourth dimension” (4D—the time): with
of Architecture-Structure-MEP. specific tools, in fact, BIM will track all temporal

430

TIS2017_Book.indb 430 2/8/2017 8:05:32 PM

 stages of the project from design to construction,
allowing the interaction with the actual “step”.
Once concluded the phase M4 of 3D model-
ling, up to the level of detail of shop drawings, it
is possible to define times and ways for site and
construction management, by adding to each ele-
ment, already modelled, many detailed parameters,
agreed with the construction company. In this way,
quantity of all elements in the construction site
and costs can be controlled. BIM is able to indi-
cate all information recorded by users and more:
ranging from number of walls to number of fur-
niture, even neon maintenance or bulbs life before
next replacement.
As to each elements of the project it is associ-
ated a precise value, the insertion or deletion of
any object from the model involves a change in the
overall construction cost, which can be controlled
minute by minute in the design phase. Therefore,
BIM includes also a fifth phase: 5D—cost control,
which can be managed before the construction site
setting-up. Figure 9. Conventional CAD vs BIM Approach, Azhar
As regard to Al Jadeda Station project, these et al. (2008).
stages have not started yet.

burden can be reduced due to less linking time

2.7 Benefits and disadvantages
between modelling packages and simulating cal-
As regards the different disciplines, a list of ben- culation programs. The transferring phase of sec-
efits can be analysed, noticing that it is reflected in tions’ characteristics, used materials’ features or
the outcomes of each BIM-structured-model, the manufacturing components units is cancelled, by
high contribution in work effectiveness. eliminating any potential human error, as the user
In the architectural field, BIM users make the dif- always working on updated model.
ference, as the greater efficiency in graphical inter- Another important benefit using BIM is reduc-
face and communication within informatics tools tion of processing time of bill of quantity, as indi-
and areas of work are the winning components. cated in the BIM Handbook (2008). In case of
In order to guarantee that the available functions study from the model tables containing any data
used are fully and correctly tamed, staff has to be of project items like space areas, number of dif-
qualified and able to quickly transfer the model, ferent element and their technical characteristic,
between various tools, without loss of quality. etc… have been taken out automatically. These
For those who still works with “no BIM” com- sheets have hugely reduced time of calculation as
puter packages, separated files (Figure 9) or in well as human errors. Furthermore making use
bulk documentations have to be controlled, each of parametric elements directly linked with their
single element designed: for instance ones useful main outputs in automatic tables, it was possible
for the definition of bill of quantities, ones related to quickly suppose the costs of the other station, in
to the development of photorealistic images, once which the same objects with same technical charac-
for land register interaction, once to carry out the teristic were used.
energetic studies and produce certification activi- Extremely significant after the BIM birth, has
ties etc. been the reduction in time spent for the identifi-
Modeling the Al Jadeda Station directly in 3D cation of complications between networks and
has allowed to have the full awareness of the evolu- plants for services (such as passage of electric
tion of each element of the project and to iden- cables, pipes and ventilation, gas or smoke ducts
tify any design flaws, which have been immediately interfering with structural elements).
resolved. This has been possible thanks to BIM It is clear that designing in BIM means to be
technology, as described at paragraph 2.5, and also expert to control all the details, communicate with
their easier dialogue mode in the integration within colleagues and partners (which potentially are
all the parent tools. In particular, in those models using other software), sharing all data and project
that involve significant calculations, structural information. For this reason the Company needs
analysis, dynamic assessment, the computational more qualified resources with deep knowledge of

431

TIS2017_Book.indb 431 2/8/2017 8:05:33 PM

 this technologies, the ones who can be called bim- • up to 40% elimination of unbudgeted change;
mers. This latter point conduct to the arising of the • cost estimation accuracy within 3%;
production team costs at first step. • up to 80% reduction in time taken to generate a
The direction that is taking the actual design- cost estimate;
ing mode evolution is represented by the adoption • a savings of up to 10% of the contract value
of techniques ever more modern. The evidence of through clash detections;
the quality BIM-based design, and the recognised • up to 7% reduction in project time.
benefits will lead to the standardization of the
processes and technologies. In the meanwhile in
2.9 Conclusions
technical specifications is noticeable that innova-
tive solutions will experience a smooth transition Having gambled in Building Information Mod-
to BIM, as per Art. 23 of the Legislative Decree eling (BIM) as a design tool and management of
18 April 2016, n. 50 “New Code of the public the entire life of an infrastructure, 3TI has been
contracts”. found many advantages in Italy as well as abroad.
“The Italian engineering companies, in fact, in
terms of quality and technical skills (of which it is
2.8 Economic advantages
definitely now part of the BIM), can boast strong
The Designer Company 3TI during the design of competitiveness abroad, and assuming that criti-
Al Jadeda Station (and other stations) have experi- cal mass is sufficiently high for addressing major
enced and appreciated that BIM gives a significant projects, they are able to excel in all areas” (inter-
reduction in designing times and running calcula- view to the 3TI Chairman).
tions. An example is identified in the possibility This paper aims to emphasize how great and
that BIM tools allow to have the bill of quantity as powerful has been found the use of BIM system in
an output of the software: it is precise and correct modelling and design projects, as per 3TI experi-
(provided that all the elements inserted in the model ence, and how it has been needful for the design of
are right and not doubled), not a business of BOQ case study.
Expert anymore (even though the Expert of course Control and validation of models is possible in
controls the output of the software). Furthermore every step of the project: indeed, in the BIM work-
the Design Engineer Company may rely on the flow verification of data does exist consistency,
experience gained over years of working in the BIM whether they come from the same company as well as
Environment particularly for the reduced time of by external professionals working on/with the same
implementation of drawings on only one model model. The interoperability is surely the attribute
platform instead of on a set of different software’s that allows the proper communication among all
decks. Less time is translated into a better undertak- workers involved, avoiding economic waste.
ing and greater productivity from the workers, which Certainly it is essential a control highlighting
lead in minor commitment of human resources. beforehand the design and planning inconsistencies,
Moreover the Designer can guarantee to the as well as real interference between the construction
Construction Company anytime the awareness of components, as has been experienced in the petal
the dimension of the project. Every step of plan- vaults’ ceiling of the Al Jadeda Station design. The
ning process can be monitored and controlled and model checking allows through customized rules
the company can have information at all moments (internally agreed by users), to optimize the process
on technical specifications, sizes and characteris- and make it easily communicable to stakeholders.
tic of each element, cost’s estimate, time sched- Although the advantages offered by a technol-
ule, location of materials’ storage, etc. thanks to ogy such as BIM are obvious, and appear increas-
the use of BIM by the Designer. Obviously cost ingly clear to those who use it, the difficulties to its
benefits for construction company comes after spread are many. For example, the effort of design
phase M4, from phase 4D—site management, as firms or companies for transition from traditional
better anticipated in previous paragraphs. Finally, CAD to BIM is considerable (both in terms of
the most important thing is that the global vision software implementation and in staff’s recruitment
of the project guarantees the higher standard of and training).
security: in an interactive way, banned areas are It is also necessary to say that if it is true that
‘marked out’ by sound emitted from transceivers BIM represents an excellent solution for quality
when a danger is signalled. guarantee, it is also true that the bimmers have to
Stanford University Center for Integrated Facil- produce a clear and “clean” model. So, while a well-
ities Engineering in the studies listed in the annual structured BIM model, created from the beginning
technical report CIFE (2008) based on 32 major following quality standards, would be certainly
projects using BIM, indicates benefits such as stud- manageable in any aspect and without any sig-
ied by Azhar et al. (2008): nificant difficulties, a model poorly controlled will

432

TIS2017_Book.indb 432 2/8/2017 8:05:33 PM

 lead to complications: the re-editing of a model Azhar, S.; Nadeem, A.; Mok, J.Y.N.; and Leung, B.H.Y.
constructed for example with overlapped lines or 2008. Building Information Modeling (BIM): A
components’ layers not right placed, would be New Paradigm for Visual Interactive Modeling and
time-consuming and costly. Simulation for Construction Projects. Proceedings of
the First International Conference on Construction
Progressive use of specific methods and elec- in Developing Countries (ICCIDC-I), August 4–5,
tronic tools, such as electronic modelling and Karachi, Pakistan.
information for the construction industries, infra- Eastman, C., Teicholz, P., Sacks, R. & Liston, K. 2008.
structure companies and engineering firms, is BIM Handbook: A Guide to Building Information
moving the Italian Legislator to insert notes in the Modeling. Canada. John Wiley & Sons
Standards accordingly. Gao J. and Fischer M. 2008. Framework & Case Stud-
ies Comparing Implementations & Impacts of 3D/4D
Modeling Across Projects - Technical Report #TR172
ACKNOWLEDGEMENTS March 2008. Center for Integrated Facility Engineer-
ing (CIFE), Stanford University. Available online at
http://cife.stanford.edu/sites/default/files/TR172.pdf.
The authors are thankful to Arch. Franca Frances- Legislative Decree 50/2016. Regulation of public pro-
cucci for her diverse support and valuable advises, curement regarding works, services and supplies
excellent cooperation in the content of the paper implementing Directives 2014/23/EC, 2014/24/EC and
and suggestions for improvements, and to Eng. 2014/25/EC
Luca Mezzadri for his technical point of view and Paloscia, L., Interview to 3TI Chairman Ingletti
images. Alfredo, available on line at the Strade e Autostrade
on line website: http://online.stradeeautostrade.it/
notizie/2016–05–25/fatturato-in-crescita-per-3ti-
progetti-53179/
REFERENCES Succar, B., Building information modelling framework:
A research and delivery foundation for industry
Associated General Contractors of America (AGC) stakeholders. Automation in Construction, 2009, 18,
2006. AGC Contractors’ Guide to BIM. Available on 357–375.
line at Whole Building Design Guide website: https:// UNStudio website: http://www.unstudio.com/projects/
www.wbdg.org/bim/bim_libraries.php?l = a. qatar-integrated-railway-project-qirp

433

TIS2017_Book.indb 433 2/8/2017 8:05:33 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The pavement management system for improving airports

and roads infrastructures

G. Battiato & L. Cosimi

RODECO Group, Voghera, Italy

C. Stratakos
NAMA LAB S.A., Athens, Greece

ABSTRACT: The PMS (Pavement Management System) provides a set of objectives and well-organized
procedures for determining areas having high maintenance priorities and work plans, assigning resources
and budgeting for pavement M&R (Maintenance and Rehabilitation). It also can be used to quantify
information and furnish detailed recommendations for works required to maintain a set of pavements at
a satisfactory level of service and safety while minimizing the cost of M&R. By the PMS it’s possible not
only to evaluate the actual condition of a pavement network, but also to predict their future condition
through the use of pavement condition parameters, such as residual life and overlay, IRI and PCI ect. Since
the late ‘80 s RODECO Group has developed RO.MA® (Road Management) PMS & Asset Management
software, in order to increase the quality of the Roads and Airports Infrastructures in the short, medium
and long term, for improving pavement durability and safety.

1 INTRODUCTION by users and, in particular, the international experi-

ence gained by RODECO in the field of Pavement
The purpose of a PMS is to improve the quality of Evaluation and Management System.
airport infrastructures and plan maintenance and Particular attention was engaged in software
rehabilitation measures in the short, medium and development, simplification and flexibility of data
long term, maximizing available resources that are entry procedures, in order to reduce the amount
not always adequate to the customer needs. of information necessary to obtain the results of
RO.MA.® PMS satisfies these requirements as the PMS.
asset management tool, easy to use and can be cus- The main requirements of the software are the
tomized following recommendations of the Road following:
and Airport Agency.
• provide a methodology to assess the pavement
The survey of functional and structural charac-
condition (surface and structural parameters),
teristics of the pavements, using high-performance
using the Pavement Quality Index (PQI);
systems, is the basis of PMS: the most important
• provide a device to define an optimum plan of
parameters, measured in the infrastructure and
scheduled maintenance and rehabilitation of
used by RO.MA.® PMS software, are the bearing
airport and road pavements, with a priority list
capacity (measured with HWD) for ACN-PCN
of interventions;
evaluation, roughness and planimetric profiles,
• estimate the optimal time to apply maintenance
surface distresses and PCI (Pavement Condition
measures, using appropriate decay models to
Index), skid resistance.
predict pavement condition in the future;
RODECO Group has recently developed and
• propose planned maintenance activities opti-
introduced new technologies for automatic meas-
mizing the cost/benefit ratio, with some budget
urement of the PCI, IRI and distress; in particular
constraints introduced by the user.
the ADE System (Automated Distress Evaluation)
which allow more extensive and widespread use of Some recent RO.MA.® applications (as Larnaka
high-performance systems for surveying and qual- and Sarajevo Airports) have specifically high-
ity control of road and airport infrastructures. lighted as an efficient PMS system can optimize
The RO.MA.® PMS has been continuously the maintenance on a multiannual basis, substan-
updated relating to the evolution of tially reducing the overall cost of management of
high-performance systems, the new needs expressed airport infrastructure.

435

TIS2017_Book.indb 435 2/8/2017 8:05:35 PM

 2 PAVEMENT EVALUATION

RO.MA.® PMS software was designed to process

data coming from the analysis of the results of the
Pavement Evaluation phases (field tests). For the
evaluation of structural and surface characteristics
of an airport pavement network, there are different
types of high-performance systems, which allow to
quickly record all the parameters required for a
proper evaluation of the PMS.
According to ICAO Annex 14 and ENAC “Ital-
ian Civil Aviation Authority”, the typical pave-
ment surveys, that it’s possible to conduct, are the
following:
• definition of the pavement bearing capacity
with HWD (Heavy Falling Weight Deflectometer) Figure 1. RODECO Laser Profilometer during Malta
and calculation of ACN and PCN parameters; International Airport surveys.
• survey of longitudinal and transverse profiles
using Laser Profilometer; The irregularities of a pavement are the result of
• measurement of pavement surface condition and an infinite number of wavelengths, componing the
evaluation of PCI (Pavement Condition Index); longitudinal profile of a pavement.
• investigations with GPR (Ground Penetrat- The profilometer can detect the actual profile
ing Radar) and coring to define the pavement of the pavement in the XY coordinates (relative),
stratigraphy; where X is the distance measured by the odom-
• survey of the skid resistance. eter and Y represents the elevation profile of the
pavements.
Generally, the system is capable of storing the
2.1 HWD (Heavy Falling Weight Deflectometer)
actual average profile every 100/200 mm of each
The structural characteristic of an airport pave- section.
ment network are analyzed using the Heavy Falling The knowledge of the amplitude to the short-
Weight deflectometer (HWD), which can ade- waves (1–3.3 m), medium (3.3 m - 13 m) and long
quately simulate the load condition of an aircraft. (13 m - 60 m) is very important to identify the
For each HWD measuring point, through the cause of the irregularities.
software RO.ME. (Road Moduli Evaluation, Where the irregularity is related to the short
developed by RODECO), are estimated: waves, the problem can be found in the surface lay-
ers of the pavement (surface distresses), while the
• the values of the E1, E2, E3 moduli (asphalt lay-
irregularities related to the long and medium waves
ers, subbase and subgrade) under test condition;
may be due to problems of subsidence in the bot-
• the value of the E1 modulus (asphalt layer),
tom layers.
referred to 20°C;
The software allows, through simulation, to
• residual fatigue life of the pavement in years;
analyze the filtered results obtaining the values of
• the critical layer;
irregularities at the wavelengths desired.
• the calculated theoretical reinforcement neces-
Starting from the real profile, the following
sary to support the project traffic, in mm;
parameters are estimated for sections of 25 m:
• calculation of ACN/PCN in accordance with
ICAO. • IRI (International Roughness Index) averaged
in mm/m on 25 m sections;
• irregularities filtered to short waves from 1 to
2.2 Laser profilometer
3.3 m, in mm/m, on sections of 25 m;
The Laser Profilometer is used for the survey of • irregularities filtered to medium waves from 3.3
the pavement roughness, longitudinal and trans- to 13 m, in mm/m, on sections of 25 m;
verse profiles and the definition of the index IRI • irregularities filtered to long waves from 13 to
(International Roughness Index), measured in 60 m, in mm/m, on sections of 25 m;
mm/m. • simulation of a 3 m straightedge for calculating
The IRI is a standardized index that contains the maximum deflection, as required by ICAO
the information required to establish the regular- standards;
ity of a pavement surface, as defined by the World • the maximum deflection on sections of 45 m
Bank Technical Paper No. 45. and the number of irregularities that in this

436

TIS2017_Book.indb 436 2/8/2017 8:05:35 PM

 section exceed 20–30 mm as required by ICAO • raveling;
standards; • depressions.
• cross slope (%).
Each type of surface distress should be evalu-
The equipment of RODECO Laser Equipment ated by using 2 indices:
includes:
• quantity index: % of the area affected by the
• various lasers, for a width of about 3 meters, distress;
with a sampling frequency of 5 mm; • quality index: L (low) = low severity—M
• n. 3 accelerometers and n. 2 high-precision (medium) = Moderate severity—H (high) = high
gyroscopes; severity.
• software for raw data processing, to calculate
cross-slope, rutting, longitudinal profiles (IRI) and The images of the pavement, as detected by the
irregularities for short, medium and long waves. ADE system, are post-processed with a specific
software called “Automated Distress Evaluation”,
through it’s possible to recognize and codify, on a
2.3 Distress Survey: PCI (Pavement Condition database, the different types of distress, in a com-
Index) analysis pletely automatic way.
The survey of the surface distress for rigid and From ADE surveys, PCI (Pavement Condi-
flexible pavements is necessary to: tion Index) is calculated for each sample unit, in
accordance with the ASTM D5340-12 (airport
• control the surface degradations and their evo- pavements) and ASTM D6433 (road pavements).
lution over time; Starting from Videocar survey, global quantity
• identify degraded areas to plan emergency of possible preventive maintenance/rehabilitation
actions; is defined, together with needs and priorities of
• provide preventive maintenance and rehabilita- maintenance, like crack sealing or other type of
tion to slow or halt the degradation process and preventive treatments.
prolong the service life of the pavements. The PCI is a numerical index, ranging from 0 for
a failed pavement to 100 for a pavement in perfect
RODECO Group has developed advanced
condition. PCI is divided into three classes:
technologies for the automatic detection of sur-
face distress as the ADE (Automated Distress • 70 ≤ PCI < 100 good
Evaluation). • 55 < PCI < 70 fair
Digital images, using Videocar system, are • PCI ≤ 55 poor
acquired each 4 meters, on all airport pavements
(runways, taxiways, aprons), and simultaneously dif- Using this classification, a detailed map of all
ferent types of surface distress (classification using airport or road pavements is realized, that permits
3 level of severity) are detected each 25 m section. to identify promptly the critical sample unit (red
The main surface distresses that ADE system color in the map).
can process are the following:
2.4 Ground Penetrating Radar (GPR) Survey
• longitudinal cracks;
By GPR technology, it is possible to continuously
• transverse cracks;
detect the pavement stratigraphy and estimate the
• alligator cracking;
thickness of the different layers. The output of this
survey consists of numerical tables and/or graphics
that provide the thickness of the different layers of
the existing airport pavements. The data obtained
from the GPR surveys are used to process the
HWD data, to estimate the elastic moduli of dif-
ferent layers, and for the PMS drawing.
RODECO Group has a radar system, that may
mount up to 3 antennas operating between 200 MHz
and 2 GHz, and that can investigate the pavement
airport infrastructures up to 3 m of depth.
The antennas are installed on a vehicle properly
equipped with the control unit, a computer proc-
ess, a high-resolution color video and an encoder.
The collected data can be displayed on screen
in real time and then processed to produce tables
Figure 2. RODECO Videocar-ADE System. and graphs.

437

TIS2017_Book.indb 437 2/8/2017 8:05:36 PM

 2.5 Skid Resistance Survey • traffic;
• number of layers used for backanalysis.
The skid data collection has a very important rule
to assess the adequacy of air traffic safety level. There are various statistical techniques avail-
The GripTester, approved by ICAO, is a very able for sharing data in a series of homogeneous
simple device, consisting of a trailer towed to a sections. One technique is the cumulative sum
vehicle; the trailer has two side wheels and a central method, introduced in the ‘50 s for the industrial
wheel, braked during movement, used to measure quality control, and resumed in the “AASTHO
the friction coefficient. Guide for Design of Pavement Structures guides”.
The braked wheel is constantly sprinkled with This method is the best among existing in the
water by a distributor during the tests; water comes identification of homogeneous sections, but at the
from a tank installed on the driving vehicle. same time presents a difficult to translate in an
The water flow is regulated with an electronic algorithm. It was therefore used another method,
pump to ensure the desired thickness (eg. 1 mm of inside the RO.MA.® PMS software, based on
water) between the tire and pavement; this water dichotomization of the measures.
thickness don’t depend of the speed measurement. This method considers a probability of change of
During the trailer movement, two longitudinal an r—parameter at a specific significance level; r—
strain gauges measure the strain “Fo” (which is parameter is the ratio between the quadratic distance
opposed to the travel) and the vertical load “Fv”, between two successive data and variance of data.
given by the weight of the trailer; “Fv/Fo” ratio is The method uses the Carré Moyen test of
the Grip Number (GN). consecutive differences for the study of a value’s
The strain gauges are connected to an electronic randomness and Gaussian characteristics in
system that records data value of the friction coeffi- a section. In addition, it uses a dichotomist
cient measured every 10 meters on an onboard com- technique to perform the segmentation. Through
puter. The tests are performed on airport pavements this, the method verifies whether data behaves
according to the requirements of the legislation (ICAO under a normal probabilistic distribution before
Annex 14—Aerodromes Volume I—Aerodrome performing the segmentation.
Design) at a speed of 65 and 95 km/h. The segmentation method searches for the point
Measurements are performed along 4 lines, at where the decision function g(i) is maximized, as
±3 and ±6 meters from the runway centerline. shown in the following equation:
Processed data can be provided in charts and
graphics for every test point values, and the relative ⎡ i ⎤
2
n
average values every third of the runway length, as g (i ) = ⋅ ⎢ ∑(x (x j x )⎥ (1)
required by ICAO standards, are reported. ( n i )i ⎣ j =1 ⎦
The Skid Resistance of road pavements is also
gk x S (i ) (2)
measured by SCRIM equipment. 1≤ j k

where n is the total number of data contained in the

3 DEFINITION AND EVALUATION OF analyzed section, xj is the j-value of the measured
HOMOGENEOUS SECTIONS point and x is the mean of the data in the analyzed
section. The point where g(i) is maximum within
One of the most important step in the interpretation the section, represents the point where the mean
and processing data is related to choice airport or changes. It is also the first element of the second
road sections with similar characteristics; an homo- segment of the same section. Later, the procedure
geneous section is characterized by having virtually is repeated in both segments, obtaining consecu-
constant parameters throughout its considered tive maximum points of g(i). The segmentation
length. The procedure to calculate the homogene- process ends when the g(i) function contains no
ous section permits to define the pavement section more maximum points within either analyzed seg-
having similar features: once identified, they can- ment, or when the homogeneous section length is
not be further divided into subsections having sig- less than a length value fixed by the user.
nificantly different behavior. The subdivision of an
airport or road network element into homogene-
ous sections may be based on various parameters,
4 DESCRIPTION OF RO.MA.® PMS
whose considered are the following:
SOFTWARE
• IRI;
• PCI; The RO.MA.® PMS software, developed for road
• deflection basis parameters (D0, SCI “Surface networks, and specialized for airports, is a tool
Curvature Index” and BCI “Base Curvature that can provide the optimal maintenance and
Index”); rehabilitation strategies, identified through clear

438

TIS2017_Book.indb 438 2/8/2017 8:05:36 PM

 and simple procedures. The software, starting from
a database relating to the pavement condition and ∑w i =1 (3)
its further division into homogeneous sections, eval-
uates alternative strategies in a reference period of The PQI formula, instead, is the following:
10 years, considering the one with the highest effec-
100
tiveness/cost ratio, for each homogeneous section. PQI w1 ⋅ ⋅ ( IRII max IIRI )
The software flow chart starts from the survey IRII max IRI
I min
data (residual life, IRI and PCI), then proceeds 100
to the definition of homogeneous sections, with + w2 ⋅ ⋅ ( PCI C )
CI max − PCI
PCI
CI max PCI
PCI min
the method previous described, for all considered 100
parameters. + w3 ⋅ ⋅ (RL Lmax − RL ) (4)
As the database was created, and enriched over Lmax − RLmin
RL
the time by surveys, the software works follow-
ing four main steps, which will subsequently be Once the database is created, the priorities for
detailed separately: maintenance will be calculated for each section,
needed to define the list of maintenance strategies
• database of pavement parameters referred to the during the PMS time period.
current condition of the pavement, definition of
the homogenous sections and priority levels set
for the network elements. 4.2 Prediction models of pavement parameters
• prediction models of the IRI, PCI, and RL (IRI, PCI and RL) and selection of
(Residual Life) over the time and parameter maintenance level
range for specific maintenance treatments; Prediction models for IRI, PCI and RL param-
• maintenance measures and their characteristics; eters have been provided inside the software as a
• calculation of the optimal strategy for each function of the possible traffic class (of the airport
homogeneous section, considering budget con- or road network). For each parameter, basing on
straints or free budget. technical experience, a decay curve was calculated
for each traffic class. These models depends on
4.1 Definition of database and priority levels time, so the value of a particular parameter can be
calculated at any time during the service life of the
The first step is to create a database, which shall pavement.
contain all the features of the surveyed pavements. For each parameter it is necessary to identify
First, it’s necessary to identify, inside the entire air- different ranges, when a type of maintenance treat-
port or road network, the branches which will be ment can be used, according to the three mainte-
assigned a priority for maintenance management; nance classes:
this priority decreases with the importance of the
branch itself. It will be necessary to report the fol- A. Preventive maintenance: required treatments
lowing data for each branch: that restore mainly the PCI parameters;
B. Partial reconstruction maintenance: required
• ID code; measures that restore PCI, IRI and RL param-
• traffic flow: traffic can be divided into three cat- eters not in a complete way;
egories (high, medium, low), where the lower C. Rehabilitation maintenance: required treat-
priority level is associated with high traffic level, ments that restore PCI, IRI and RL parameters
and so on; to the project condition.
• number of layers of the backcalculated model,
type and thickness, and average moduli E1, E2,
E3 (MPa), respectively, of the asphalt concrete
layer at 20° C, subbase and subgrade;
• average IRI, PCI and RL values, calculated for
the homogeneous sections;
• geometrical characteristic, such as width, length
and area of the homogeneous sections, and
parameters measured by the laser profilometer,
such as cross slope, and straight-edge;
• average value of the PQI parameter (Pavement
Quality Index), which expresses the overall pave-
ment condition. This variable is calculated as the
sum of the various analyzed parameters (IRI,
PCI, and RL); each of them has an associated Figure 3. Prediction model for IRI parameter, depend-
multiplier wi such that: ing on the traffic level.

439

TIS2017_Book.indb 439 2/8/2017 8:05:37 PM

 4.3 Types of maintenance and rehabilitation
treatments and their characteristics
Within the RO.MA.® PMS software, it was decided
to use a set of 9 maintenance treatments, divided
into the three maintenance classes (see classifica-
tion above-mentioned). For each treatment, it was
necessary to define:
• the maintenance class (preventive, partial recon-
struction or rehabilitation);
• what parameters are repaired and how it restores
(the recovery of each parameter is expressed as a
percentage of the existing value); Figure 4. Definition of different strategies for an homo-
• the unit cost (depending of the type of treat- geneous section.
ment);
• the expected life for each maintenance and
rehabilitation treatment (the required time of Each section will have different maintenance
pavement to decay in the condition before applied strategies, because managing authority could
maintenance). decide to take action on the section at different
time, as shown in the Fig. 4: this is linked closely
to the range values of individual parameters repre-
4.4 Definition of the optimum strategy for each senting the pavement condition.
homogeneous section considering budget The next step is to define the effectiveness of each
constraints or free budget strategy, its cost and the effectiveness/cost ratio, in
order to choose which is the optimal strategy for
Once defined maintenance treatment and their
each section. In this case it’s necessary to consider
application range, the last step of the software is to
two possible cases: the case where there is a free
identify the best possible strategy for each homo-
budget, and if there are economic constraints.
geneous section during the PMS time.
Strategy refers to a plan of action that includes
4.4.1 Free budget Analysis
one or more specific maintenance operations
For each strategy it’s necessary to identify:
designed to restore and improve the performance
characteristics of the pavements. • his effectiveness, calculated on the PQI parameter;
Starting from the current value of PQI, the soft- • total costs of maintenance treatments considered.
ware calculates the best possible maintenance strat-
About the calculation of the effectiveness of sin-
egy, combining the various types of treatments.
gle-strategy (E), it will be identified graphically by
This calculation is performed considering the pos-
the included area between the curve that represents
sibility that the strategies will be not applied to the
the strategy after his application and the PQI curve
first year of analysis, but they are deferred over the
before. About the overall costs of each strategy,
time, allowing the decay of all parameters, follow-
they will be referred to the initial year of the analy-
ing their prediction model.
sis period (CCA: actualized construction costs),
This approach is used because some interven-
depending on the discount rate r, that expresses the
tions have a useful life lower than the period of
amount of interest paid/earned as a percentage of
analysis on which the strategy is evaluated, then
the balance at the end of the annual period.
calculating strategies that involve the application
The ratio between the effectiveness of single
of multiple treatments it’s possible to cover the
strategy and total cost (E/CCA) will be calculated
entire period of analysis.
for all possible strategies of each homogeneous
The strategies are established from the decision
section; the strategy, having the higher value of the
tree, starting from the set of maintenance treat-
previous ratio, will be chosen. At the end of this
ments listed in the previous chapter: a strategy will
process, which will be repeated for each homogene-
therefore consist of one or more branches of the
ous section, it will be possible to estimate the total
tree, depending on the duration of the proposed
budget to be invested for the implementation of
maintenance period. At the end of the analysis, the
the best identified strategies.
results, for each strategy, will be a vector having the
following features:
4.4.2 Restricted budget analysis
• ID code, related to the homogeneous section; This method, used to identify the best strategies
• maintenance treatments types and year of when a preset budget is fixed in the time of analysis
application. (10 years), consists of the following steps:

440

TIS2017_Book.indb 440 2/8/2017 8:05:37 PM

 a. definition of the budget to invest; • BCI (Base Curvature Index), which characterizes
b. definition of maintenance strategies during the base layer.
the period of analysis, for all homogeneous
sections; For each section, the following parameters have
c. calculation of effectiveness “E” and total cost been calculated, starting from the backanalysis
“CCA” of each maintenance strategy; process:
d. calculation of the ratio effectiveness/cost of • Modulus of elasticity of the bituminous layer at
each strategy (E/CCA); 20°C (E1), and its thickness;
e. selection of the strategy (referred to a k homo- • Modulus of elasticity of the base layer (E2), and
geneous section) having the higher value of E/ its thickness;
CCA: it will be used as reference for the begin- • Modulus of elasticity of the subgrade layer (E3);
ning of the iterative calculation; • Modulus E0: surface modulus calculated under
f. calculation, for the k section, the marginal cost- the load plate;
effectiveness ratio MCE of all other expected • Residual fatigue life of the most critical layer;
strategies on the same k section. It is expressed • Theoretical overlay required to reinforce the
in terms of effectiveness and cost of each i-th weak pavement;
strategy identified for the section k, as com- • ACN/PCN coefficients.
pared with the selected strategy in step 5;
g. if MCE is negative or Ei < Ek, the i-th strategy The output of the software was the following:
is eliminated and will not be considered in fur- • Identification of the homogeneous sections and
ther steps, otherwise the combination replaces application of the maintenance measures;
the selected strategy in step 5, and the available • Year of the rehabilitation for the weakest homo-
budget is updated; geneous sections;
h. the process is repeated until no other strategy • Type of maintenance;
can be selected for each section in each year of • Total surface area of the rehabilitation;
analysis, and when the allocated budget is fully • Unit cost (€/m2);
committed. • Total Cost of the maintenance for each homoge-
This method allows, in addition to the definition of neous section;
the best strategies to be applied to all homogeneous • Total cost for all the PMS period (there were no
sections, to perform an analysis at different budget budget constraints) and divided by year.
levels: inserting a different value for the estimated The adopted solutions, added to preventive and
budget, strategies choice will change, in accordance curative maintenance, consisting in the application
with the most appropriate capital to invest. of a special fiberglass grid. This type of mainte-
nance has been optimized by the software, capable
5 IMPLEMENTATION OF AIRPORT of detecting the thickness of the pavement to mill,
PAVEMENT MANAGEMENT SYSTEM: strictly related to the depth in which to insert the
THE EXAMPLES OF LARNAKA AND grid; the aim was to reduce the time and costs of
SARAJEVO the intervention, obtaining the same final result of
a new resurfacing.
5.1 Larnaka International Airport
The software, for the Larnaka Airport, was applied
at the project level: the customer has identified dif-
ferent areas where HWD test has been conducted,
in order to investigate the pavements having differ-
ent distresses on the surface.
Tests have been performed with a step of 20 m,
collecting data useful for the realization of a main-
tenance project such as to meet the institution’s
needs and ensure the necessary characteristics of
bearing for the next 15 years.
The various areas have been divided into homoge-
neous sections, based on the following parameters:
• D0, which is the measured deflection at the
center of the loading plate;
• SCI (Surface Curvature Index), which charac- Figure 5. Definition of the homogeneous section for
terizes the surface layers of the pavement; the Larnaka Aiport.

441

TIS2017_Book.indb 441 2/8/2017 8:05:38 PM

 5.2 Sarajevo International Airport The use of the system in all reported cases in
this paper has allowed the optimization of the
RODECO Group performed a series of high
maintenance budgets for all agencies. This was
performance surveys on the Sarajevo Airport
accomplished through the use of RO.MA.® PMS
pavements, in order to plan a pavement manage-
techniques to identify specific areas of the airport
ment system for the next 10 years. Using all the
network that are in need of maintenance; at the
technologies described before, the actual pavement
same time, RO.MA.® PMS software has identified
condition has been detected.
areas of the airport that did not need maintenance.
RO.MA.® PMS software was used to define the
The users have benefitted from this analysis by
strategy of preventive maintenance and rehabilita-
knowing exactly where implementing maintenance
tion, once made the surveys using high-performance
techniques, saving time and money.
systems (IRI, PCI, HWD etc.). The software out-
put was a list of maintenance measures per year
and for each homogeneous section located within REFERENCES
the airport network. In the chart obtained by the
software the following parameters were shown: AASHTO Guide for design of pavement structures, 1993,
Appendix J: J1–J5.
• year of implementation of maintenance meas- ASTM D5340-12, Standard Test Method for Airport
ures belonging to the chosen strategies for each Pavement Condition Index Surveys, 2012.
homogeneous section; ASTM D6433-11, Standard Practice for Roads and Park-
• type of maintenance measures to apply, their ing Lots Pavement Condition Index Surveys, 2011, 1–2.
dimension and location; Battiato, G., 1990. Evaluation of PCN (Pavement Classi-
• unit cost (€/m2) and total cost of maintenance fication Number) of airfields by means of the F.W.D.
per year (€). In Proceedings for Third International Conference on
Bearing Capacity of roads and airfields, Trondheim,
The primary objective of the PMS for this air- Norway, 3–6 July 1990.
port was to plan and especially “optimize” the Battiato, G., 1998. Prove e controlli non tradizionali sulle
pavement maintenance and rehabilitation strategy, pavimentazioni bituminose. La valutazione della qualità
i.e. to maintain the highest level of functional and delle pavimentazioni stradali mediante l’impiego dei sis-
temi ad alto rendimento. In Proceedings for SITEB 7°
structural characteristics of the pavement with corso tecnico di base su leganti e conglomerati bituminosi,
minimal cost, assuring a significant savings for the S. Donato Milanese, Italy, 25–27 November 1998.
managing authority. Battiato, G., 2000. Valutazioni e controlli di qualità delle
The main results of the PMS application on Pavimentazioni Aeroportuali. In Proceedings for Le
Sarajevo Airport was to demonstrate that, even if pavimentazioni aeroportuali: 90–108, Rome, Italy,
the very critical condition of the runway pavement, 30 Novembre 2000.
it was feasible to project a total rehabilitation by a Battiato, G.; Al Emadi K.M.I.; Liberati P, 2003. Descrip-
reconstruction limited to the surface layers, apply- tion et application du systeme de gestion routier (PMS)
ing special materials and technologies, with a huge developpe pour l’Etat du Qatar. In Proceedings for XXII
Congres Mondial Routier—C6 Comite sur la gestion rou-
saving of economic resources and time. tiere, Durban, South Africa, 19–25 October 2003.
Battiato, G.; Larsen B.K., 1990. Description and Appli-
cation of RO.MA.® (Road Evaluation and Pavement
6 CONCLUSIONS Management System). In Proceedings for III Interna-
tional Conference on Bearing Capacity of roads and air-
Some practical examples of the use of the RO.MA.® field, Trondheim, Norway, 3–6 July 1990.
PMS were presented with a brief description of Battiato, G.; Amé E.; Wagner T., 1994. Description and
the pavement evaluation and PMS phases of the Implementation of RO.MA.® for Urban Road and
RO.MA. method. The pertinence of the proposed Highway Network Maintenance. In Proceedings I for
Third International Conference on Managing Pavements,
method has been discussed in terms of technical 43–50, S.Antonio, Texas, USA, May 22–26 1994.
and economic aspects. The implementation of Battiato, G.; Larsen B.K.; Ullidtz P.,1987. Verification of
a database on the user’s PC containing PE and the analytical – empirical method of Pavement Evalu-
PMS data allows the user to prepare more realistic ation based on FWD testing. In Proceedings for 6th
annual maintenance budgets based on a scientific International Conference on Structural Design of Asphalt
approach to maintenance rehabilitation problems. Pavements, Ann Arbor, Michigan, USA, 1987.
Moreover, the use of new software for graphic ENAC Numero: 3/2015-APT, Ed. n. 1 del 1 Ottobre 2015,
representations of the PMS results increases the Airport Pavement Management System - Linee guida
information available for the agency to manage its sulla implementazione del sistema di gestione della
manutenzione delle pavimentazioni.
network in terms of efficiency and economy, and ICAO Annex 14 to the Convention of International Civil
the graphic presentation of the data allows also Aviation – Aerodromes – Volume I – Aerodrome Design
the user to easily understand the results of the and Operations, Sixth Edition, July 2013: ATT A4-A10,
RO.MA. analysis. A28–A29.

442

TIS2017_Book.indb 442 2/8/2017 8:05:38 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The use of porous asphalt for the improvement of the grading plan
geometry and drainage of pavement surfaces on urban roads

V. Ilić, M. Orešković & D. Gavran

Department for Roads, Railroads and Airports, Faculty of Civil Engineering, University in Belgrade, Serbia

I. Pančić
Transportation Institute CIP, Belgrade, Serbia

ABSTRACT: Porous asphalts are applied all over the world to minimize the effect of noise caused
by everyday traffic. Beside the effect of noise reduction, thanks to their open texture and improved
drainage characteristics, porous asphalts decrease the impact of spraying water behind a moving vehicle.
Since drainage capability of porous asphalt as a surface permeable layer is much higher compared to a
conventional dense-graded asphalt surface course, the values of cross grades of pavement wearing course
on the road sections with porous asphalts should be additionally considered. Hence, the primary topic of
this paper is how the application of porous asphalt could affect the geometric design of grading plans of
pavement surface on urban roads, especially in intersection zones, where different longitudinal and cross
grades of intersecting directions have to be mutually aligned. Due to a higher drainage capacity of porous
asphalt layers, layout plan of storm water inlets can be changed if pavement cross grades are reduced.

1 INTRODUCTION lowest elevations of pavement surface so that their

full capacity could be reached.
Correct design and construction of drainage sys- Particularly sensitive areas in terms of storm-
tem of urban roads is one of the most responsi- water drainage of urban roads are intersections,
ble tasks of civil engineers engaged in the design where pavement surfaces of different intersecting
of transport infrastructure in cities. Design of roads have to be accurately graded and fitted into
drainage elements of the largest number of urban each other in geometrical sense. Grading plan of
road sections is based on the city sewage systems, pavement surface is used by hydrotechnical pro-
which, in addition to their drainage function, play fessionals as an initial background drawing for
an important role in protection against pollutions. determining the exact locations of storm water
Due to intensive urban development and construc- inlets.
tion of buildings right next to R.O.W. (right of way) In this paper, the recommendations for proper
boundaries, drainage systems of urban roads have setting of storm water inlets in intersection zones,
to be dimensioned with certain capacity reserves in as well as the analysis of the impact of porous
order to be able to accept rainwater run-on from asphalt application as a wearing course of pave-
neighboring building roofs and facades (Maletin ment structure, are given. In the case study of
2009). real four-leg intersection located in Belgrade it
Key elements of the whole pavement drainage was shown how the reduction of pavement cross
system are storm water inlets which accept rain- grades affects the change of grading plan geom-
water from gutters and divert it to the storm drain etry of pavement surface. Also, through the analy-
pipes. Also, they represent a membrane which sis of simplified cross sections of urban road, the
retains larger debris and waste materials washed impact of modified cross grades on elevations
from the pavement surface (Maletin 2009). Effi- heights of outside pavement edges and curb ele-
cient drainage concept implies that storm water ments was discussed in detail. In additional case
inlets have to be placed in the direction of major it was shown how the grading plan of pavement
flows of rainwater collected along the road curb surface on parking lot is changed due to the appli-
(Despotović 2009). Except directly next to the cation of porous instead of standard dense-graded
curb, storm water inlets should be placed at the asphalt pavement.

443

TIS2017_Book.indb 443 2/8/2017 8:05:40 PM

 2 GENERAL PRINCIPLES OF GEOMETRIC longitudinal grades required for efficient drainage,
DESIGN OF GRADING PLAN or their longitudinal grades have excessive values
due to high superelevation rates of pavement.
In order to design and plot a grading plan of pave- Bearing in mind a simple fact that water always
ment surface at intersection the basic design ele- runs perpendicular to the direction of contour
ments such as longitudinal grades, vertical curves, lines, in order to make a storm water inlet layout
grades of all cross section elements of intersecting plan, it is necessary to determine the specific drain
roads, as well as the methods of attaining superel- area per one storm water inlet on the basis of pre-
evation (superelevation development length, viously collected hydrotechnical parameters such
superelevation rate and axis of pavement rotation), as designed storm event, rainwater runoff condi-
have to be defined. In general, longitudinal grades tions and maximum flow. For the exact definition
of road depend on the rank of a road according to of these parameters a constant monitoring of pre-
functional classification and the objective location cipitations and research in the field are essential.
requirements. All of the above mentioned elements It is very important to avoid the positioning of
must be mutually harmonized in intersection zones, storm water inlets at pedestrian and / or bicycle
in order to get merged and uniquely designed pave- crossings. In addition, storm water inlets should be
ment surfaces in 3D. positioned directly before or after pedestrian cross-
Outside pavement edges, which are usually fol- ing in order to collect as much water as possible,
lowed by the gutters that collect rainwater flowing and prevent rainwater runoff across the pedestrian
along the streets, must be designed with required crossings. In real urban road network, however,
longitudinal grades which guarantee adequate con- is not so rare that storm water inlets are placed
ditions for pavement drainage. For standard con- wrongly, and because of that the required amount
ditions and normal roughness of pavement whose of water can not be cached, or the rainwater that
flatness lies within the construction tolerance the pours down the pavement surface just misses incor-
minimun value of longitudinal grade is iN = 0.5%. rectly installed storm water inlets.
The adoption and implementation of longitudinal
grades below this limit deteriorate drainage con-
ditions, and this requires a detailed analysis of 3 BASIC CHARACTERISTICS OF POROUS
runoff parameters and, in certain circumstances, ASPHALT
the specific allocation of storm water inlets along
urban road section, or even the application of Porous asphalts are bituminous mixtures char-
non-standard drainage elements. Maximum longi- acterized by high percentage of air voids. Pave-
tudinal grade does not depend on drainage require- ments with porous asphalt mixtures must have
ments, because its upper limit is conditioned by adequate bearing capacity and satisfactory drain-
the functional rank of road section within the age of storm water. Infiltrated water further flows
urban road network and the general infrastructure to the layer which contains reservoir for collected
requirements defined in urban development plans rainwater, or runs through all layers of pavement
(Maletin et al. 2010). Also, the maximum values of structure to the poorly compacted sub-grade. Per-
longitudinal grades can be used only on the sec- centage of air voids, even up to 25%, is achieved
tions between adjacent intersections, while in the due to an open gradation curve. During the design
intersection zone a certain reduction of longitudi- process of porous asphalts, it is not enough to take
nal grades is needed. into account only the characteristics of the mate-
Based on a shape and pattern of contour lines rial, traffic load, etc., but also many other param-
drawn in grading plans road designers can perceive eters that can contribute to the reduction of overall
uniformity of grading changes and notice possible noise levels and improve their drainage capabilities
deficiencies of suggested design solutions. This such as the size of aggregate grains, discontinu-
particularly applies to the horizontal distances ity of mixture and porosity. The size of aggregate
between contour lines generated next to the pave- grains in the mixture affects the permeability, dura-
ment edges, so for example, variation of distances bility and stability under the traffic load. By reduc-
between contours lines indicates some irregulari- ing the grain size, noise caused by the interaction
ties, and if contour interval changes inconsistently of tire with the road surface will be reduced. If the
the pavement surface is improperly modeled in 3D. porosity is increased, the absorption of noise and
Those areas on pavement surface, where contours the drainage of water through porous layers are
do not look like smooth lines, require further cor- also increased.
rections. Some elevation corrections of grading plan Due to the high content of voids in the mixture,
are also necessary in cases when outside pavement the use of porous asphalt significantly reduces the
edges followed by the gutters and curbs do not have occurrence of “aquaplaning’’ effect, if these voids

444

TIS2017_Book.indb 444 2/8/2017 8:05:40 PM

 are not clogged (Bendtsen 2002). There seems to be after which a vacuum is needed to remove liquid
a tendency that porous pavements on highways and and retained solid pieces. It is recommended that
rural roads keep the porous structures open dur- first cleaning should be performed three months
ing their lifetime. This is explained by the existence after construction, and then after every six months.
of “self-cleaning mechanism’’, where the vehicle Recent experiences in the field have shown that the
tires press water down in the voids of the porous water pressure around 20 MPa gives satisfactory
pavement under high pressure during the periods results (Fletcher & Theron 2011).
of rain. On the other hand, on low speed urban The biggest drawback of porous pavements is
roads water pressure, due to slower rotation of their high cost and limited durability compared to
vehicle’s wheels, is not high enough to ensure a standard dense-graded asphalt mixtures. Due to
continuous “self-cleaning’’ effect of the pavement, the relatively short exploitation period (maximum
causing the clogging of voids in the pavement on 8 to 10 years) (Ferguson 2005) and high investment
such roads. This was revealed first by the engineers costs, porous asphalts are rarely applied in poor
from Copenhagen in Denmark, after their experi- countries. Also, a major disadvantage of these
mental research was conducted on real test sections pavements is high maintenance costs, especially
(Bendtsen 2005, Bendtsen et al. 2005). during winter.
Friction resistance of porous asphalt is an impor-
tant characteristic, as well as drainage. At lower
4 APPLICATION OF POROUS ASPHALT
speeds, the friction resistance of porous asphalt is
FROM THE ASPECT OF PAVEMENT
smaller in comparison with standard asphalt mix-
DRAINAGE
tures because the exposed surface of aggregate at
the tire pavement contact is smaller. Consequently, One of the most important properties of porous
the microtexture of a surface porous layer and the pavements is their ability of pavement draining,
roughness of exposed aggregate particles become i.e. reduction of surface runoff from pavement sur-
main features. Lack of “friction’’ at lower speeds faces. In order to better understand the hydraulic
is not considered the major disadvantage, since regime, and a new hydrologic balance that is estab-
low friction values are usually dangerous at high lished by the construction of porous pavement,
speeds that are common on motorways. Therefore, all components, inflow (surface run-on, precipita-
the porous asphalts are suitable for urban roads tion), as well as runoff of water in the system have
paving where traffic speed is slower. to be analyzed (Fig. 1). The main inflow (filling)
Uncontrolled traffic growth has a negative sources of system are primarily atmospheric pre-
impact on the environment. Except noise and cipitation (rainwater) and surface run-on from the
harmful exhaust gases, a significant source of adjacent pavement and other surfaces. One part of
pollution is a large amount of particles generated the inflow immediately infiltrates through wearing
by friction and tire wear. Porous asphalts have a course of pavement, and the other part evaporates
positive impact on the improvement of the quality into the atmosphere or further flows along certain
of water that flows on the road surface and infil- hydraulic gradient according to the grading char-
trates through pavement layers. A large amount acteristics of a surface course. A certain portion of
of dangerous particles is retained in the upper the infiltrated water inflow in the pavement can be
layers. Quality control of water that runs through collected by drainage pipes laid on the subgrade
the whole pavement structure was done in several surface (formation level), and all the rest of the
countries around Europe and similar results were
obtained (Gopalakrishnan et al. 2014). The water
that is collected at the bottom of the reservoir is
diverted via a perforated tube out and analyzed.
It has been noticed that the concentration of the
adsorbed metals zinc, lead, and others is only 20%
to 30% relative to their concentration in surface
layers. In order to prevent that dangerous particles
penetrate into natural soil and pollute ground-
water, a regular vacuuming and maintenance of
porous pavements are recommended.
Porous asphalts are prone to clogging if they are
not properly maintained. Proper maintenance of
porous asphalt extends their lifetime and retains
their functional characteristics. Porous pavements Figure 1. Water balance variables for permeable pave-
are cleaned by using water jets under high pressure, ments (source: Eisenberg et al. 2015).

445

TIS2017_Book.indb 445 2/8/2017 8:05:40 PM

 asphalt pavements which are not subjected to
high traffic loads.
2. Porous asphalt layers paved as wearing (top)
course over the previously constructed, water-
proof flexible or rigid pavement. This type of
porous asphalts is often called “Overlay’’ in
the literature, while in the USA “Permeable
Friction Course—PFC’’ is a widely used term
(Fig. 2b). Unlike porous asphalts with the res-
ervoir in the base course, which allows water to
infiltrate into the subgrade, this type of porous
asphalt has entirely different drainage mecha-
nism. All infiltrated water in “Overlay’’ layer is
transversally diverted through voids to specially
designed gutters or hollow curbs laid along
pavement edges. Most problems with the drain-
age of “Overlay’’ layer are caused by the water
that remains trapped at the contact surface
between the new layer of porous asphalt and the
old waterproof pavement course. Trapped water
can cause serious problems during the winter at
low temperatures because there is not enough
Figure 2. a) Porous asphalt paved over water permeable space to expand due to the snow and ice in the
base course with open-graded base reservoir; b) Porous cavities of the porous surface layer.
asphalt paved over the existing pavement—Overlay
(source: Ferguson 2005).
5 MODIFICATION OF THE GRADING
infiltrated inflow is further drained through the
PLAN GEOMETRY AFTER THE
subgrade material into the subsoil.
APPLIACTION OF POROUS
Because of its open texture, porous pavement
ASPHALT—CASE STUDY
reduces the effects of spraying and splashing water,
and this phenomenon was reduced by more than
In order to clearly understand the impact of porous
95% after construction a new porous asphalt layer
asphalt application on the changing of pavement
(Rungruangvirojn & Kanitpong 2009). This feature
cross grades, and consequently on the modification
changes over time due to clogging of the pores.
of grading plan geometry, two practical cases from
Overall, in terms of drainage of porous pave-
site are shown. The first case refers to the standard
ments, two types of pavement structures with wear-
four-leg intersection (crossroad), and the second to
ing porous asphalt layers can be distinguished:
the application of porous asphalt as a surface layer
1. Porous asphalt layers with infiltration base (res- of the porous pavement structure at parking lots.
ervoir). With this type of pavement structure, In Figure 3a the grading plan of the intersec-
porous asphalt layers are constructed over the tion of Severna tangenta and Zage Malivuk streets
permeable base layer, through which the infil- is presented. The plotted drawing of grading plan
trated water inflow is drained to the so-called “res- represents the existing situation where the wearing
ervoir’’ from unbound aggregates, and further to top course of pavement was made from standard
the uncompacted subgrade of pavement struc- dense-graded hot mix asphalt. Severna Tangenta
ture all the way down to the natural ground (sub- Street was considered as a major road (arterial)
soil) (Fig. 2a). A reservoir layer should consist and Zage Malivuk Street as a minor road (cross
of pure, uniformly granulated aggregates with street). All sections where changes of pavement
30%–40% of voids (Eisenberg et al. 2015). The cross grade occur, as well as all vertical points of
thickness of the layer containing reservoir is intersection of longitudinal alignments, for both
determined on the basis of the required capac- major and minor road directions are precisely
ity for storage of infiltrated water, structural labeled. Figure 3a shows that in this case cross
requirements, seasonal changes of high ground- grades of the major road are equal to 2.50% and
water levels in the soil, and the frost depth. have crown shape along the entire intersection
This type of porous pavement is mostly used zone, and that adjacent longitudinal grades of the
for parking areas in residential or commercial cross street are also 2,50% in order to fit into the
complexes, then for the construction of pedes- cross grades of the major road. At the same time,
trian and bicycle paths, driveways, and for other the longitudinal grade of the major road of 1.00%

446

TIS2017_Book.indb 446 2/8/2017 8:05:40 PM

 Figure 3. Grading plan of pavement surface at the four-leg intersection: a) when wearing course is paved with stand-
ard dense-graded asphalt mix; 4b) when wearing course is paved with porous asphalt.

represents the pavement cross grade of the minor other hydrological parameters of the site in Bel-
road in the intersection zone. The shape of con- grade where the intersection is located, the specific
tour lines on the grading plan clearly indicates that drain area for one storm water inlet is estimated
vertical alignment of minor road accurately fol- as 250.00 m2.
lows cross grades of pavement of the major road. The grading plan of the same intersection,
Equidistance or contour interval is 2.50 cm. On with changed values of pavement cross grades
the grading plan of pavement surface every fourth due to laying of porous asphalt as a wearing
contour line is labeled in ascending order, and the course of pavement structure, is plotted again in
positions of pedestrian crossings and storm water Figure 3b. Cross grades of the pavement on the
inlets are displayed too. Taking into consideration major and minor road are reduced from 2.50% to
the designed storm event (designed rainfall) and 1.00%. The longitudinal grade of the major road

447

TIS2017_Book.indb 447 2/8/2017 8:05:41 PM

 remained unchanged, since its correction is very edges decreased. The advantage of this method
rarely, almost never, done. However, due to the of traveled way superelevation is that elevation
reduction of pavement cross grades of the major heights along the outside pavement edges remain
road the longitudinal grades of the minor road the same, and the elevations of existing storm
had to be changed radically. As a consequence water inlets do not change either. In addition,
of reduced longitudinal grades of the minor road this method of attaining superelevation is more
in the intersection zone, the positions of vertical suitable for urban roads that pass through city
points of intersection (VPI) significantly shifted in centers (downtowns) among densely built struc-
the vertical alignment of the minor road compared tures, because the elevations of entrance steps
to the previous case. Although the grading plan in the surrounding buildings do not have to be
in Figure 3b was plotted using the same equidis- changed. Major disadvantage of this superel-
tance as in the grading plan in Figure 3a, due to evation method is that most of the layers of the
lower values of pavement cross grades, horizontal existing pavement structure, including a wear-
distances among contour lines are wider. It is evi- ing course, have to be destroyed and removed
dent that contour lines in the grading plan in Fig- later in order to construct a stable and flat base
ure 3b seem more harmonious and are smoother course for the new layers of porous asphalt. The
compared to the contour lines in the grading plan entire median has to be lowered down vertically
plotted in Figure 3a. which may cause additional problems if some
After the construction of porous asphalt run- underground utility lines have been already
off from the pavement surface reduce substan- installed below the median surface. These com-
tially due to the infiltration of one part of surface plex construction works are inevitably followed
run-on and precipitation through porous layers. by enormous investment costs.
Therefore, the specific drain area, which was used 2. In the second case pavement cross grade is
for determining the number of storm water inlets, reduced by revolving a traveled way about
has to be estimated again. If it is assumed that the median edges. Using this method of attaining
runoff coefficient of porous wearing course is 0.50 supereleva-tion all structures and facilities which
(for dense-graded asphalt mixtures its value is usu- were built next to the pavement edges are func-
ally between 0.90 and 0.95), it can be easily con- tionally disrupted due to the rising of the eleva-
cluded that the specific drain area in comparison tions of side curbs along the outside edges of
with the one which was used for the positioning of traveled way (Fig. 4b). This further leads to the
storm water inlets in Figure 3a, should be increased uplifting of sidewalks and elevations heights of
almost two times, or to the value of approximately entrance steps of neighboring buildings, which
450.00 m2. For this reason, the layout plan of can be very inconvenient for all residents of
storm water inlets in Figure 3b is different from these buildings. Not only functionality of build-
the one which is displayed in Figure 3a. Hence, for ings, but also their aesthetics and architecture
the drainage of the identical pavement surface a suffer because of the entrance steps uplifting.
smaller number of storm water inlets is needed. Also, vertical positions of storm water inlets
The reduction of pavement cross grades of and other elements of drainage system that are
primary urban roads and city arterials with sepa- located right next to the side curbs along the
rate carriageways can be performed in two ways, outside pavement edges have to be corrected.
as shown in simplified drawings of cross sections For primarily ranked urban roads whose edges
in Figures 4a and 4b. In densely populated urban are not encroached with houses and buildings,
areas where buildings are placed right next to the this superelevation method for the decreasing
sidewalk edges, changing of the pavement cross of pavement cross grades represents better engi-
grades can extensively impact on other elements neering solution compared to the supereleva-
of public infrastructure and facilities. Two specific tion method illustrated in Figure 4a. Existing
cases from road design practice are discussed: layers of pavement structure together with the
wearing course of dense-graded asphalt mix can
1. Cross grade decreases by revolving a traveled be kept, but in that case, due to reduced cross
way about the outside lateral edges of the grades, a new wearing course of porous asphalt
traveled way, and the axis of rotation coincides must be paved in a layer of variable thickness.
with the contact edge between the surface wear-
ing course of pavement and adjacent side curbs. Changing of the pavement cross grade at park-
From the cross section drawing in Figure 4a it ing lots whose pavement structure was founded
can be seen that the elevation levels and verti- on permeable subgrade or embankment is much
cal position of outside curb elements next to the easier to implement, as shown in the example in
ends of the traveled way stayed the same, while Figure 5. Since all surface run-on is infiltrated
the elevation heights of the curbs along median through porous layers to the subgrade surface and

448

TIS2017_Book.indb 448 2/8/2017 8:05:42 PM

 Figure 4. Simplified cross sections of urban road witch illustrate the effects of cross grade reduction by revolving a
traveled way: a) about the outside lateral edges of pavement surface; b) about median edges.

Figure 5. Changing of pavement cross grades at parking lot after the construction of wearing course of porous
asphalt over water permeable base with reservoir.

449

TIS2017_Book.indb 449 2/8/2017 8:05:42 PM

 further into natural soil, with this type of park- structure is founded on a permeable subgrade
ing lots there is no need for setting storm water there is no need for setting storm water inlets and
inlets and the construction of the whole system the construction of the whole system for pavement
for pavement drainage, as it is necessary to do for drainage.
parking areas with standard dense-graded asphalt Is it rational to change pavement cross grades
pavement. The biggest drawbacks for a wider due to the implementation of a new porous asphalt
application of this type of pavement structure layer as a wearing course on urban roads remains
for the construction of parking lots are high cost the basic dilemma. If we only take into account
of used materials, very rigorous quality require- increased investment costs then this proves an
ments during construction and relatively expensive unreasonable decision, but after comparing the
maintenance treatments. shapes of contour lines of grading plans plotted for
pavements with and without porous asphalts, the
issue of ride comfort, besides noise and pollution
6 CONCLUSION reductions, as well as the elimination of “splash
and spray’’ effect, are gaining in importance.
The key reason for writing this paper was to inves- Also, it should not be forgotten that application
tigate how the construction of new porous asphalt of porous asphalts, especially on urban roads with
surface layers can affect the existing geometry of higher operating speeds, contributes to traffic
grading plans of urban roads. In order to clearly safety.
understand the impact of porous asphalt applica-
tion on the changing of pavement cross grades, and
consequently on the modification of grading plan REFERENCES
geometry, two practical cases from site are shown:
four-leg intersection and parking lot. Bendtsen, H., Larsen, L.E. & Greibe, P. 2002. Devel-
opment of noise reducing pavements for urban roads,
The reduction of pavement cross grades of pri- Report No. 4. Copenhagen: Danish Transport
mary urban roads and city arterials with separate Research Institute.
carriageways can be done in two ways: by revolv- Bendtsen, H. 2005. The DRI-DWW Noise Abatement Pro-
ing a traveled way about the outside lateral edges gram - Project description. Issue No. 24. Copenhagen:
of pavement surface or about median edges. In the Danish Road Institute & Danish Road Directorate.
first case elevation levels of the outside curb ele- Bendtsen, H. et al. 2005. Two-layer porous asphalt for
ments next to the ends of the traveled way stayed urban roads, Report No. 144. Copenhagen: Danish
the same, while the elevation heights of the curbs Road Institute & Danish Road Directorate, p. 5–16.
along the median edges decreased. The advantage Despotović, J. 2009. Kanalisanje kišnih voda (In Serbian).
Belgrade: Faculty of Civil Engineering University of
of this method of the traveled way superelevation Belgrade.
is that elevation heights along outside edges of Eisenberg, B. et al. 2015. Permeable Pavements. Reston:
pavement remain the same, as well as elevations American Society of Civil Engineers-ASCE, Virginia,
of existing storm water inlets. However, in order USA.
to construct a stable and flat base course for the Ferguson, B.K. 2005. Porous Pavements. Boca Raton:
new layers of porous asphalt most of the layers of CRC Press, Taylor & Francis Group, Florida, USA.
existing pavement structure, including a wearing Fletcher, E. & Theron, A.J. 2011. Performance of open
course, have to be removed. In the second case, by graded porous asphalt in New Zeeland. NZ Transport
revolving a traveled way about median edges, all Agency research report 455. Wellington: NZ Trans-
port Agency.
structures which were built next to the pavement Gopalakrishnan, K., Steyn, W.J. & Harvey, J. 2014. Cli-
edges are functionally disrupted due to the rising mate Change, Energy, Sustainability and Pavements.
of elevations of side curbs along the outside edges Berlin: Springer-Verlag, Germany.
of traveled way. For primarily ranked urban roads Maletin, M. 2009. Gradske saobraćajnice (In Serbian).
whose edges are not encroached with houses and Belgrade: Orion Art.
buildings, this superelevation method for decreas- Maletin, M., Anđus, V. & Katanić, J. 2010. Techni-
ing pavement cross grades is a better solution com- cal guidelines for intersection design (PGS-PR / 07).
pared to the superelevation method applied in the Belgrade: Faculty of Civil Engineering University of
first case. Belgrade.
Rungruangvirojn, P. & Kanitpong, K. 2009. Measure-
Unlike complex modifications of grading plans ment of Visibility Loss due to Splash and Spray: the
of intersections, changing of cross grades of pave- Comparison between Porous Asphalt, SMA, and
ment in parking lots, due to construction of new Conventional Asphalt Pavements. Transportation
permeable porous layers, does not represent a hard Research Board, 88th Annual Meeting. 11-15 January
task for road designers. If new porous pavement 2009. Washington D.C.

450

TIS2017_Book.indb 450 2/8/2017 8:05:43 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Urban tracks for Light Rail Transit (LRT). Dublin experience

M. Corsi
Track Design Manager, Transport Infrastructure Ireland, Republic of Ireland

ABSTRACT: Luas Cross City (LCC) is Dublin new 5.9 km light rail extension (Luas is the commercial
name for Dublin LRT), being built as part of the overall development, since 2001, of a new, modern and
efficient LRT network for the Irish Capital.
One of the major challenges for what is essentially a city central, partly on-street LRT, has been the
design and construction of the embedded track system. This design was influenced by the experiences
of previous Dublin LRT projects and by the outcomes of benchmark analyses with other European net-
works and trial track panels.
TII (Transport Infrastructure Ireland), is delivering LCC through a D & B contract, based on very
strict Employer Technical Requirements. The resulting innovative track design is described in detail in
this paper.

1 LRT EMBEDDED TRACK SYSTEMS 2 THE LUAS (DUBLIN LRT) NETWORK

EMBEDDED TRACK TYPES
Urban light rail track is referred to as “embedded
track” because the rails (generally grooved type) 2.1 Overall Luas network
are embedded into the road surfaces.
The overall Luas network is formed by two lines,
Embedded track international standards do not
Green and Red and it was built over a decade start-
exist, and each Country (and even every city in
ing in 2001, with the northern section of the verti-
the same Country) applies different requirements,
cal line (Green Line) currently under construction,
tolerances, design principles and construction
the “Luas Cross City” (LCC).
philosophies.
Among the various National standards or
guidelines, the French CERTU, and German VDV 2.2 Initial lines A-B-C
would be the most commonly referred to, with
Luas Green and Red Lines (Lines A-C and Line B)
the English ORR and Irish GDRIRS being more
were opened to passenger traffic in 2004. The two
functional requirements.
lines were not connected in the City Centre and
Together with city specific systems, mostly
measured a total length of approx. 25 km.
based on long lasting experience (Zurich, Freiburg
Few sections of these lines were on-street run-
for example), there are off-the-shelf or turn-key
ning, with embedded track built with grooved rail
proprietary systems by the largest track manufac-
embedded in up to three stages concrete structural
turers available.
slab.
Among all of them, the main differentiator param-
The original track system had the following
eters are the form of gauge control (with or without
features:
sleepers, direct fixation, no fixation at all), concrete
levels and reinforcement (low, middle or high con- – RI59 N rails installed with no sleepers (EN 14811
crete, and traditional reinforcement, no reinforce- Railway applications—Track—Special purpose
ment, fibre-reinforcement), the type of construction rail—Grooved and associated construction)
(top-down or bottom-up) rail insulation (rubber – ALH type factory-made rail encapsulation.
profiles, rubber filler blocks, chambers, number – Concrete shoulders 150 mm wide on each side
and materials of the components, etc.), casting sys- of the rails for asphalt or granite setts finish.
tem (precast or in-situ), rail-road joint type, finishes – 150 mm subbase (3% cement bound) +300/
(exposed concrete, asphalt, modular elements, grass). 350 mm reinforced concrete structural slab
Within this frame, the following sections +100 mm reinforced infill slab (or 100 mm
describe the track system specifically developed asphalt or granite sett). Total depth 550/600 mm
and adopted for Dublin Light Rail system. from bottom of subbase.

451

TIS2017_Book.indb 451 2/8/2017 8:05:45 PM

 Figure 3. Embedded Track for line A1 under
construction.

Figure 1. Schematic of the Luas (Dublin LRT) network

at completion of Luas Cross City.

Figure 4. Typical cross sections of Embedded Track for

line B1.

Figure 2. Typical cross sections of Embedded Track for Figure 5. Typical cross sections of Embedded Track for
lines A-B-C. line C1.

– Top down construction with temporary

supports.
– Stray current active protection (ALH encapsu-
lation) and passive protection (current collec-
tion cage around each rail). (EN 50122: Railway
applications—Electric safety, earthing and the
return current)
Figure 6. Typical cross sections of Special Embed-
ded Track for line A1 at road crossings—no concrete
shoulders.
2.3 Luas extensions A1-B1-C1
The two lines were subsequently extended between
2007 and 2011, with the construction of three new
sections, A1-B1-C1 for an additional total length
of approx. 13.5 km.
Each of those extensions was delivered with a
different D&B or Employer’s design contract, with
the result of three new interoperable but conceptu-
ally different embedded track systems, with differ-
ent structural design, sleepers, concrete shoulders
and rail encapsulation systems. Figure 7. Special Embedded Track for line A1 at road
crossings—no concrete shoulders.
2.4 Special type A1
concrete height to 40 mm below the rail head, and
For Line A1 a special type of embedded track was use polymer modified asphalt, of the type TSCS
tested over some of the busiest road junctions, to (Thin Surface Course System), (NRA National Road
experiment the durability of embedded track with Authority Specification for Road Works. Series 900
no shoulders. The selected solution was to limit the Road Pavements—bituminous bound materials).

452

TIS2017_Book.indb 452 2/8/2017 8:05:45 PM

 Also, to seal the joint with the rail encapsulation
at the surface, a hot bituminous filler was poured on
both rail sides (40 mm wide on running side to cover
for the wheel tread width in the worst scenario).
This track has been in operation for approximately
five years and no issues have manifested so far.

3 ORIGINAL EMBEDDED TRACK ISSUES

Figure 10. Concrete shoulders “V” shape initial crack.
3.1 General
In the course of the last 12 year’s operation, the
various track types aged differently. In general
there were no major issues, to the satisfaction of
the Infrastructure Manager and the Maintainer.
Despite a general positive outcome, some recur-
ring defects were noted, as listed below.

3.2 Early Back Flange Contact (BFC)

Figure 11. A and B. Examples of concrete shoulders
Early Back Flange Contact (BFC) appeared on some structural collapse and failed repair works.
of the tighter curves where no gauge control bar or
sleeper was provided (Lines A-B-C). The absence of It was soon evident that the existing track sys-
sleepers and active gauge control may have contrib- tems presented two shortcomings in relation to
uted to gauge spread under lateral load, exacerbated side wear.
by the high lateral flexibility of rail encapsulation The first was the adoption of R260 rail steel
(factory applied rail coating), also due to the sin- hardness for the grooved rail on all alignments
gle encapsulation casting process (same mechanical (EN 13674 Railway applications—Track—Rail
characteristics in both vertical and lateral planes). standards) presenting high carbon medium hard-
Early side wear of the keeper, friction noise, and ness steel, with medium side wear, and low chances
potential for derailment risk were the consequences. of successful gauge corner repair welding at low
temperature.
The second was the challenge associated with
3.3 Rail replacement
rail replacement in a full rail concrete embedment
Guiding rails on tighter curves (R < 50 m) started scenario. The shoulders were initially designed to
reaching side wear maintenance limit in 2013. provide additional rail lateral stability in the lack
of sleepers, a clear demarcation line between road
and track (maintenance responsibility between
LRT operator and City Council), and a protection/
support for rail encapsulation and road surface
In terms of rail replacement, all the “shoulder-
based” designs indicated a proposed “cut line”,
next to each side of the rail to free it up prior to
grouting-in the new encapsulated rail.
As it emerged at first trial, the remaining part of
Figure 8. Shoulder cut-out with holes for the additional
the shoulder would not be structurally sound and
fastenings. New rail yet to be inserted and grouted. therefore the new inserted rail would have to be
chemically anchored in the main slab via an alter-
native fastening prior to re-grouting the missing
part of the shoulder.
This made the rail replacement very complicated
and expensive. Also, the vertical cold joint within
the shoulder made it fragile and prone to failure,
particularly under road traffic in shared running
sections.
Dimension, structural soundness and level of
reinforcement of the shoulders were highlighted as
Figure 9. Wheel-Rail interface on Dublin Luas. critical factors.

453

TIS2017_Book.indb 453 2/8/2017 8:05:48 PM

 As part of this thorough analysis, a 5 m long
single track trial test panel was also built to test
constructability, resistance and durability of four
different track arrangements and finishes. Follow-
ing the outcomes of that real-scale test, a new idea
of embedded track was developed and used for the
construction of the Rosie Hackett Bridge track in
Dublin city Centre.

4.1 Track trial panel

The track trial panel was delivered in 2013, with
the ongoing condition monitoring regime still in
place.
It was divided in two parts, one finished with
Figures 12. A. Absence of rail-road joint and ingress
asphalt, and one with modular elements (granite
of water under the rail foot—imprinted concrete finish
with stains from water pumping. – B. Damages at the setts). The asphalt section was divided in two (one
exposed encapsulation joint, with repair works (cut-out per rail), with one full depth asphalt as required by
slot, filled-in with polymer based pourable joint sealant). the City Council (100 mm in total, 60 binder and 40
wearing), and one thin wearing course with 40 mm
3.4 Concrete shoulders deterioration PMSMA (Polymer Modified Stone Mastic Asphalt).
The setts section was also divided in two, with
Concrete shoulders deteriorated as a consequence and without rail concrete shoulders.
of several different factors. Two different pourable polymer-based joint
External edge of wheel-tread runs over the sealants were also tested, one on each rail.
brittle concrete surface, damaging the joint. Luas While the evaluation of the long term durabil-
wheels are 110 mm wide to negotiate S&C with ity is ongoing, the following initial outcomes were
deep flange-way, and this makes the tread up to noted:
40 mm wider than the rail head.
These cracks developed into deeper failures for – Polymer based joints are working well and hold-
water ingress, rail pumping and winter freezing, ing on to rail and concrete in the “with shoulder
exacerbated by road traffic at shared sections. configuration”.
In 2013, 60% of the embedded track at road – 40 mm thin layer asphalt is also working well
crossings showed structural deterioration with with no joint sealant (asphalt laid against the
18% requiring shoulder replacement (once or even rail) with no sign of deterioration along the rail
twice). or de-bonding from concrete base.
– granite setts in the without-shoulder configura-
tion are getting dislodged for the poor vertical
3.5 Non-sealed rail-shoulder joint stability under traffic loads, due to the ineffec-
With the exception of Luas A1 road crossings, all tive bedding mortar (40 mm thickness) possibly
other solutions had the rail encapsulation exposed due to the poor bedding preparation and water
to the surface with no sealant. content control. This stresses the importance of
That arrangement contributed to the develop- site quality control in setts laying.
ment of several issues for the ingress of water, – Setts along the rail with no shoulder do not
debris, dirt often deteriorating into cracked shoul- present instability issues, but the wider bedding
der and damage to the encapsulation or to its joints problem jeopardized the full panel.
(collars).

4 NEW DEVELOPMENTS

Following the analysis of previous installations

and building on the lessons learned from the main-
tenance outcomes, TII undertook a comprehensive
benchmark analysis of existing European systems
(France, Germany, Italy, UK and Switzerland) to
establish a new embedded track standard for the
future lines in Dublin. Figure 13. Test panel plan and cross sections.

454

TIS2017_Book.indb 454 2/8/2017 8:05:50 PM

 Figure 17. Rosie Hackett Bridge—Embedded Track
Cross Section over the bridge.
Figure 14. Test panel completed.

Figure 18. Rosie Hackett Bridge—Embedded Track

with asphalt finish across the Quays.

Figure 15. Rosie Hackett Bridge—Embedded Track

with granite setts and derailment containment channel.

Figure 19. Rosie Hackett Bridge—Embedded Track

with asphalt finish across the Quays.

The architectural requirement (granite setts fin-

ish) further called for very shallow track slab, with
the adoption of a proprietary slim gauge control
Figure 16. Rosie Hackett Bridge—Plan profile of the
track alignment. system, in the form of flat steel sleepers at 1.5 m
interaxis and structural concrete shoulders for
additional rail lateral support. Additionally, for the
– 100 mm hot rolled asphalt (composed by
first time in Dublin, a “low profile” rubber encap-
60–40 mm courses) is performing well with the
sulation and pourable polymeric sealant was used
exception of the joint with the polymer sealant
on a real project on both rail edges.
showing signs of deterioration along the joint
As part of the bridge works, track was also
itself. These have developed initially in the first
built across both Quays for approximately 25 m
12 months and seem to have stabilized since.
on each side. In those sections, no shallow track
requirement and heavily trafficked road junctions
called for a different track system with no concrete
4.2 Rosie Hackett Bridge track
shoulders, heavier bi-block sleepers, thicker track
The first section of Luas Cross City track (100 m) slab and a polymer based joint sealant.
was installed as part of the construction of a new In this case, while the initial TII recommendation
public transport dedicated bridge over the River was for 40 mm Thin Wearing Course (TWC) sys-
Liffey, the Rosie Hackett Bridge, in 2014. tem, similar to Line A1 road crossings and the most
The bridge was designed and commissioned successful finish of the track trial panel, a strong
by the City Council, while TII had the respon- opposition of a key stakeholder led to a design
sibility for the design, technical specifications change towards the use of the standard 60+40 mm
and site supervision for track and LRT systems binder-wearing Hot Rolled Asphalt (HRA).
installation. Soon after opening the road to traffic, some sec-
River navigability and vertical track alignment tions developed minor rail-to-asphalt joint failures,
led to a very slim deck with extreme span to depth particularly at bus turning movements.
ratio, imposing significant challenges to the struc- A mix of causes appears to have contributed:
tural design of the bridge and the track itself. type of asphalt and large aggregate, lack of proper

455

TIS2017_Book.indb 455 2/8/2017 8:05:52 PM

 compaction along the rail, asphalt milling prior to repair for tight curves), and provide an overall low
sealant laying (dislodging larger aggregate), low Life Cycle Cost.
bitumen quantity, and its evaporation for the use The proposed embedded track had to facilitate its
of a flame torch to dry up the joint prior to sealant installation within constrained central areas and strin-
pouring. gent traffic management conditions, while ensuring
track laying geometry and tolerances (RPA (Railway
Procurement Agency) 2007. Track alignment—Design
Handbook—absolute H&V tolerance 5 mm, relative
5 NEW EMBEDDED TRACK FOR LUAS
3 mm/3 m, gauge +−2 mm, twist at 3 m 1:250).
CROSS CITY (LCC)

5.1 Background 5.3 LCC embedded track design

Following agreements with the various stakehold-
– RI59 N rails with higher steel grade. (EN 13674
ers, all the lessons learned from the previous instal-
Railway applications—Track—Rail standards)
lations were considered for the new track design.
– 18 m long rubber encapsulation profiles (to min-
Essentially, the new track had to fulfill all the
imize joints), low profile, with optimized joints
railway requirements (GDRIRS Guidelines by CRR
to sleeper fasteners interface.
—Irish Commission for Railway Regulation, BoS-
– Double side rail-to-road polymer based pour-
trab SpR, 2004, EN 50122: Railway applications—
able joint filler, (40–25 mm wide).
Electric safety, earthing and the return current),
– Light gauge bar (flat steel sleepers) at 3 m cent-
including rail stability, ease of maintenance, joint
ers to facilitate construction and maintenance
stability, stray current among others, but also the
(rail replacement as described below).
road requirements, as the shared track was part
– Variable height foundation (depending on cap-
of the urban road network (NRA Specification for
ping CBR values, min 200 mm) in C8/C10 pour-
Road Works. Series 700-800-900).
able lean concrete mix + 300 mm reinforced
That led to the track design being structurally
(twin layer plus top mesh for asphalt finish
based on DMRB (Irish Design Manual for Road
only) concrete structural slab + 100 mm infill
and Bridges by National Road Authority HD24-25-
asphalt finish. (EN14227-1 Hydraulically bound
26/10) in what referred to the pavements and struc-
mixtures. Part 1. Cement bound granular mix-
tures (capping, foundation and structural slab).
tures/EN 206-1:2002/A2:2005: Concrete—Part
A typical example is the integration of the con-
1: Specification, Performance, Production and
crete shoulders into what is essentially a semi-rigid
Conformity)
road pavement design, with the requirements in
– For granite setts finish, (setts are 120 mm
terms of skid resistance to be achieved with exposed
deep, 100 mm square, over 40 mm bedding),
aggregate treatment of the shoulder’s surface.
the required insert pocket between shoulders is
This design was carried out in a joint effort
160 mm deep instead of 100 mm, making the
between the awarded D & B Joint Venture (JV) for
main slab only 240 mm thick. This is to keep
LCC and TII, within the strict rules dictated by a D
foundation levels constant irrespective of the
& B contract. Based on these, the Contractor has
track finishes.
full design responsibility within stringent Employer
– Total single track structure width 2200 mm for a
Requirements (ER) produced by the tendering
standard gauge of 1435 mm, with chamfered slab
Authority (TII in this case) and the Authority car-
edges for improved road pavement interface.
ries out compliance checks to verify the adherence
– Track depth 600 mm from bottom of
to the ER.
foundation.
– Top down construction with lateral bracing as
5.2 Design objectives for LCC embedded track temporary supports.
– Stray current active protection (rail encapsula-
Most of the design objectives have already been
tion) and passive protection (reinforcement). In
described in the course of this paper.
excess of 10 Ohm-km measured on site pre and
The fundamental requirement was the capacity
post pour testing.
to withstand to heavy road traffic for a design life
of 60 years in all its components, with the excep-
tion of the rails in curves with radius less than 5.4 LCC embedded track innovations
50 m.
The track had to be highly standardized, light – Higher rail steel grade for curves below 50 m. The
and easy to install, easy and cost effective from 290GHT (heat treated) steel grade offers higher
maintenance viewpoint (like for example allowing wear resistance, longer side wear life, less corru-
easy rail replacement and gauge corner welding gation (longer grinding intervals), vibrations and

456

TIS2017_Book.indb 456 2/8/2017 8:05:55 PM

 ground-borne noise. Reduced carbon content from the rail heads. Sleeper’s fasteners are protected
HB260, facilitates weld-repair at low temperature. with a plastic cap to facilitate future rail replace-
Rail life extended by approx. 7 to 8 years. ment, and their position is indicated by a notch
– Improved rail rubber encapsulation profile with in the rail keeper.
longitudinal joints staggered from the sleeper – Standardized cross section for different finishes
fasteners position (to avoid clamping dam- with standard reinforcement cages and shoulder
ages), low profiles (25 mm below top of rail and stirrups to speed up construction process in nar-
keeper) and optimized top edge overlap with the row sites and between operational traffic lanes.
pourable joint. – Rail concrete shoulders to avoid rail-road joint
issues was not a new feature in Dublin. But to pre-
– Light gauge bar provided at 3 m centers for con- vent shoulder structural issues, two measures were
struction and future rail replacement purposes adopted: more structurally sound design, with a
only. Self-regulating spindles to support track wider footprint (up to 210 mm wide) and higher rein-
during construction and facilitate the lateral forcement (with hook type stirrups every 200 mm
adjustments through push-pull bars anchored to and longitudinal 16 mm bar), and the sealing joint at
the running edge of the rail to prevent wheel contact
and seal rail jackets from water ingress.
– Larger shoulder with the reinforcement set back
40 mm from the rail foot to facilitate rail replace-
ment by saw-cutting at a constant rail offset, yet
leaving its structural vital part intact. Sleeper’s
fasteners re-use for rail replacement.
– High density pourable polymer-based sealing
joint on both rail sides for the reasons men-
tioned above.
Figure 20. Luas Cross City Embedded Track with – Track slab edges chamfered arrangement to
asphalt. mitigate the risk of adjacent flexible road sur-
face reflective cracking and to reduce transversal
footprint of the track while avoiding overlapped
stress-relieving membrane otherwise required by
the DMRB. Track slab width only 2,200 mm, with
significant savings and improved buildability.

Figure 21. Luas Cross City Embedded Track with

granite setts.

Figure 24. Construction detail of the shoulder

reinforcement.

Figure 22. Rail encapsulation made of three profiles

and with low type side profiles with 45 degree upper
edge.

Figure 25. Construction detail of the rail arrangement

with concrete shoulders and joint filler on both sides. To
note the precision in relative levels of the three elements
and the pigmented and exposed aggregate texture of the
Figure 23. Flat steel sleepers. shoulders.

457

TIS2017_Book.indb 457 2/8/2017 8:05:55 PM

 5.5 Construction sequence and photos

Figure 32. Joint sealant application. The joint surfaced

Figure 26. C8/C10 lean concrete foundation. are dried and prepared for better adhesion of the flow-
able polymer based material.

Figure 27. Main slab reinforcement cage and shoulder

stirrups.
Figure 33. Tack coat and base-wearing courses applied
and rolled for asphalt finish sections.

Figure 28. Shoulder stirrups details. Their precise lev-

elling will be completed after the rail-sleeper panel is in
place (sitting on sleeper spindles) and at final levels.

Figure 34. A-B. Setts laying.

Figure 29. Sleepers and rails are laid, assembled over

the reinforcement cage, welded, encapsulation profiles
are joined over the welds and final track alignment
adjustments are made. Formworks are assembled with
chamfered sides.

Figure 35. Track laying in College Green (Trinity Col-

lege in the background).

Figure 30. Main slab concrete pour.

Figure 31. Shoulders are poured between shutters

installed over the slab pour. Timber profiles hanging
from the rail head to form the joint recesses. Retardant is Figure 36. Double track laying in St. Stephen’s Green
sprayed over the shoulder surface to expose aggregate. North.

458

TIS2017_Book.indb 458 2/8/2017 8:05:58 PM

 6 CONCLUSIONS RPA (Railway Procurement Agency) 2007. Track
alignment—Design Handbook.
Following 15 years of track construction experi- RPA (Railway Procurement Agency) 2001. Track
ence in Dublin and benchmark analyses with the alignment—Tramway Clearances.
UIC 719. Earthworks and Trackbed construction for
major European networks, with the additional railway lines.
information gathered from real scale testing, sig- UIC 15273. Railway applications. Gauges.
nificant innovations were introduced in the new NRA (Irish National Road Authority) DMRB (Design
embedded track design for Luas Cross City, as Manual for Road and Bridges). HD25–26/10: Pave-
described in the paper. Monitoring regime is ment and Foundation Design.
already in place to evaluate the benefits of those NRA DMRB. HD24/06: Pavement Design and Mainte-
innovations in terms of rail-road joint, but tram nance: Traffic Assessment.
operation and full monitoring will not start until NRA Specification for Road Works. Series 700 Road
late 2017. It is expected that the new track will Pavements—General.
NRA Specification for Road Works. Series 800 Road
deliver higher performances in terms of noise and Pavements—Unbound and Cement Bound Mixtures.
vibration, corrugation development, rail and road NRA Specification for Road Works. Series 900 Road
surfaces durability, stray current resistance, and Pavements—bituminous bound materials.
overall low maintenance cost. BoStrab SpR, 2004. German Federal Regulations on the
Construction and Operation of Light Rail Transit
Systems, Guidance Regulations (SpR).
REFERENCES IS EN14227-1 Hydraulically bound mixtures. Part 1.
Cement bound granular mixtures.
CRR (Irish Commission for Railway Regulation) — IS EN 206-1:2002/A2:2005: Concrete—Part 1: Specifica-
GDRIRS (Guidelines for the Design of Railway tion, Performance, Production and Conformity.
Infrastructure and Rolling Stock) Section 7 “Tram- IS EN 13674 Railway applications—Track—Rail
ways” (doc. number RSC-G-008-B) standards.
ORR (UK Office of Rail Regulation) “Design Require- IS EN 14811 Railway applications—Track—Special pur-
ments for Street Track”. Tramway Technical Guid- pose rail—Grooved and associated construction.
ance Note 1. EN 50122: Railway applications—Electric safety, earth-
CERTU LROP Tome 1 and 2 “Plate-formes de tram- ing and the return current—Fixed installations. Part 1:
way”. AITF. 2011 and Part 2: 2010.

459

TIS2017_Book.indb 459 2/8/2017 8:06:06 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Study on paving factors of stone path in campus based on characteristics

of pedestrian walking

Yang Yanqun, Huang Qinwei, Xu Meiling & Yu Sheng

College of Civil Engineering, Fuzhou University, Fuzhou, China

ABSTRACT: Stone path is usually seen in places such as campus, park, scenic spots, etc.; And this
study is about stone path with gradient smaller than 1%. The pavement of stone path is closely related to
the characteristics of pedestrian walking. Therefore, the author collects 306 valid samples by studying the
stone paths and the characteristics of the students in a university, and analyze their influencing factors
including the gender, height, single/multi-walker, weight-bearing status, personality analysis, etc., obtain-
ing the reasonable pedestrian stride and its main factors affecting it by variance analysis. According to
the reasonable pedestrian stride, the author also designs the pavement parameters such as the slate width,
spacing of slates, and the center-to-center spacing of two adjacent slates, etc. At last, the author verifies
the conclusion of the paper by Electroencephalograph (hereinafter EEG) experiment, proving that these
pavement parameters are worthy of reference.

1 INTRODUCTION

In university campus, the open spaces between two

adjacent buildings are usually filled with plants for
the greening of environment, and lawn is the most
common one. To prevent the pedestrians between
the two adjacent buildings from trampling on the
lawn, the campus management will usually pave a
path on the lawn. Among various pavement mate- Figure 1. Stone path.
rials, stone path itself is of ornamental functions
and can get people through without affecting the lated any pavement elements such as slate width
landscape (Han, 2014). Therefore, stone path is and the spacing of two slates, etc. The design of
widely used in campus; it can be seen in places like stone path does not take into account the walking
teaching building, library, experimental building characteristics of pedestrians because of the casu-
and small park, etc. alty of the stone path construction, which causes
In the college, as a supplement to cement con- discomfort to the pedestrians walking on it. First
crete road and sidewalk paving road, etc., stone of all, pedestrians will easily miss their steps and
path is basically used by students. Compared to step into the gap between the two slates, causing
the users of the cement concrete road and side- accidental injury and damages to the shoes; sec-
walk paving road, stone path users are often in a ondly, during peak hour of pedestrian traffic flow,
leisurely walking state. Most stone paths in campus the discomfort leads to the bad traffic capacity of
are paved on flat lawn (gradient <1%), and only stone path, increasing the congestion degree rather
seldom of them are paved on area with steep slope, than easing the traffic pressure; and finally, due to
thus this study is about stone path with gradient poor comfort of walking caused by improper pave-
smaller than 1%. ment, some pedestrians may give up walking on it
Due to different causes, the pavement of stone and choose to walk on the roadside lawn, which
path is of comparatively high casualty. In the cur- greatly goes against its original intention. There-
rent code, Except the code in Code for the Design fore, it’s necessary to take walking characteristics
of Urban Green Space (GB50420-2007) that the of pedestrians into the consideration of the pave-
width of green lane should less than 0.8 m, other ment elements of stone path in university campus.
related codes including Code for the Design of Compared with the research in motor vehicles,
Urban Road Engineering (CJJ37-2012) and Code the research in walking characteristics of pedestri-
for the Design of Park (CJJ 48-92) have not stipu- ans is later and more complex. One main cause is

461

TIS2017_Book.indb 461 2/8/2017 8:06:08 PM

 the diverse forms of walking, which brings about People in group A are more confident, aggressive,
no fixed pattern in walking. The earliest relevant easy to be nervous, and have a sense of achievement.
research was conducted by Henderson L F in People in Group B are comparatively at ease
Department of Mechanical Engineering in Sydney and calm, and remain unruffled to anything.
University. After investigating the walking speed of 2. Parameters Measurement of Pedestrians’ Walk-
693 university students, 628 pedestrians in Sydney ing Characteristics
and some children aging from 6 to 8, he put forward Test personnel laid down the tape on the cement
that the walking speed of one single pedestrian is concrete pavement, and invited the participants to
presented as Gauss distribution, the average speed walk on the road paralleled to the tape (walking
of pedestrians is 1.34 m/s with 0.26 m/s standard length is about 20 m). There were 6 video cassette
deviation (Henderson, 1971, 1972, 1974). In 1975, recorders (VCR) next to the tape to make sure they
the research of Palmer A et al. stated clearly that can record the foot path and scale of the tape. The
gender is an influencing factor of walking charac- test personnel timed the walking of participants by
teristics. Variance analysis demonstrated that gender stopwatch; and after the test, they analyzed these
and age are very important independent factors. videos (Fig. 2).
Apart from the study of basic section, in recent The stride frequency is the walking steps of the
years many scholars from home and abroad have subject divided by walking time. The walking speed
studied the walking characteristics in different road is the total walking length divided by walking time.
sections including transportation hubs, pedestrian The stride is obtained by intercepting the stride of
crosswalk, mountain ladder section, etc. (Ma et al, each walking and getting the mean of it. The sub-
2009, Alhajyaseen et al, 2011, Virkler et al, 1994, jects’ height and feet size were registered through
Sisiopiku et al, 2003, Laplante & Kaeser, 2004, inquiry before the experiment.
Zhang et al, 2012, Zhang, 2014, Zhu & Lin, 2015,
Xu, 2013, Yang, 2009, Zhang, 2009, Zhao et al,
2.2 Participants
2006). Meanwhile, they have also studied the walk-
ing characteristics in different areas (Tanaboriboon To ensure the reliability of the experimental result,
et al, 1986, Chen & Dong, 2005, Lam et al, 2002, the experiment invited 310 college students to attend
Rastogi et al, 2013, Laxman et al, 2015). the experiment at random. After collecting data and
To probe into the relations between pedestrian’s eliminating invalid samples, the paper carried out sta-
walking characteristics and the pavement elements tistical analysis of the 306 valid samples. The distri-
of stone path, the author has done researches in the bution of the 306 valid samples is shown in Table 1.
walking characteristics of pedestrians in university
campus in their normal walking conditions, and
analyzed the factors which influence the pedestrian
stride, stride frequency, and the walking speed.

2 MATERIAL AND METHODS

There are many factors influencing pedestrian’s

stride, stride frequency and walking speed, such as
the height, gender, age, etc (Ma et al,2009). Accord-
ing to the make-up of the pedestrian traffic flow in
Figure 2. Date collection of pedestrian walking.
college, the author analyzes pedestrians’ walking
characteristics based on their gender, height, single/ Table 1. The Distribution of 306 samples.
multi-walker, weight-bearing state, characteristics
analysis, etc., obtaining the walking characteristics The number Percentage
of the pedestrians. Then, integrated with the feet Classification of people (%)
size of the students, the author gives suggestions on
the pavement elements of stone path in campus. Gender Male 171 55.88
Female 135 44.12
Height (cm) 155–165 116 37.91
2.1 Design and procedure 165–175 141 46.08
1. Personality Test 175–185 49 16.01
Personality A 134 43.79
The personality of the participant is defined by
test result B 172 56.21
filling out the psychological assessment question-
Single walking 60 19.61
naire before collecting the walking characteristics.
Multi walking 246 80.39
According to the result, the participants will be
Weight-bearing 105 34.31
divided into two group: A and B.

462

TIS2017_Book.indb 462 2/8/2017 8:06:08 PM

 3 RESULTS

3.1 Preliminary analysis

Table 2 lists the statistics of the stride frequency,
walking speed and stride of the 306 valid samples.
Seeing from the statistical analysis, the stride
frequency, walking speed, and stride of university
students are presented in Gaussian distribution.
The main construction elements of stone path
are slate width b, center-to-center spacing of two
adjacent slates c, and the edge spacing of two adja-
cent slates d. In addition, b+d = c (Fig. 6). The main
influencing factor of b is the feet size of students.
According to the investigation data, the mean of Figure 5. Statistical analysis of stride.
feet size of male subjects is 25.95 cm, while female
23.48 cm. Besides, stride fluency has a great impact
on d, while stride frequency and walking speed barely
have relations with b and c. Consequently, this paper
put emphasis on the analysis of the stride.

Table 2. The statistics of stride frequency, speed, and

stride.

Stride Walking speed

frequency (m/s) Stride (m)

Mean 1.76 1.22 0.68

Standard 0.21 0.22 0.12
deviation

Figure 6. Construction elements of stone path.

3.2 Influnce of gender

The distribution figure of male and female stride is
drawn by software Stata. We can see from the Fig.7
that male and female stride are basically presented
in Gaussian distribution.
To explore the influence of gender on pedestri-
ans’ stride and do significance test on the mean of
male and female stride, we build original hypoth-
esis as follow:

Figure 3. Statistical analysis of stride frequency. H0: μa = μb

where a = male; b = female.

The variance analysis of the samples was done
by using software Stata; significant level α = 0.05.
On the basis of the table, we can see that in the
null hypothesis H0, while p < 0.01, it can be con-
sidered that the mean of the two groups are of
significant difference. The result of Bonferroni’s
statistical inspection suggests that the male stride
is observably higher than female stride (p < 0.005).

3.3 Influence of height

Difference heights may also cause the variation
Figure 4. Statistical analysis of walking speed. on walking characteristics. There is a difference

463

TIS2017_Book.indb 463 2/8/2017 8:06:09 PM

 Figure 7. The distribution of male and female stride. Figure 8. Stride distribution in different heights.

Table 3. Variance analysis. Table 4. Bonferroni statistical test table.

Source SS df MS F Prob > F Row mean-Col mean 160 170

Between groups 0.217 1 0.217 51.99 0.000 170 0.48684

Within groups 1.271 304 0.004 0.000
Total 1.488 305 0.004 180 0.75230 0.25646
0.000 0.042

of stride between a tall and big stature pedestrian

and a short and small stature person. It can be
seen from the statistical data of Chen et al (Chen
& Dong, 2005). that foreign pedestrians have a
faster walking speed than Chinese pedestrians due
to height, etc. Integrated with the height of the
students in campus, this paper divides the height
into three groups as 155–165 cm, 165–176 cm, and
175–185 cm.
The distribution of stride in different heights
was drawn by using software Stata. We can see
from the Fig. 8 that the stride in different heights is
presented as Gaussian distribution.
To look into the influence of gender on pedestri-
an’s stride, we had done the significance test on the Figure 9. The stride of single-walker and multi-walker.
mean variation of stride in different heights, and
analyzed the variance of these samples by Stata.
The result illustrates that the stride of 155–165 cm corresponding p tested in the null hypothesis H0 is
is remarkably lower than that of 165–176 cm and less than 0.01, we can regard that the two group of
176–185 cm (p < 0.005), while there is no signifi- means are of great variation. In line with the result
cant difference between the mean of stride of of Bonferroni statistical test, the stride of single-
165–175 cm and that of 175–185 cm. Due to space walker is obviously bigger than that of multi-walk-
limitations, we only list the Bonferroni statistical ers (p < 0.005).
test table.
3.5 Influence of weight-bearing and A&B types
3.4 Influence of single/ multi-walker personalities
Besides the personal and environmental factors, The result of Bonferroni statistical test shows that
the walking speed is also influenced by whether there is no significant difference between A-type
the pedestrian is walking alone or not. As shown personality pedestrians and B-type personality
in Fig. 9. pedestrians (p < 0.005), nor is there any difference
The significance test and variance analysis whether the pedestrian bears weight or not (p <
were done on the difference in stride mean of sin- 0.005). Hence the paper doesn’t develop the expo-
gle/multi walkers. Result indicates that while the sition and argumentation about it.

464

TIS2017_Book.indb 464 2/8/2017 8:06:10 PM

 Table 5. Different classes of KPI in pavement. Unit: 1. The main factors affecting pedestrian’s stride
centimeter. in campus are gender, height and single/multi-
walker;
Class 1 KPI Class 2 KPI 2. Pedestrian’s stride is generally distributed in
Slate width (b) Feet size 0.6–0.8 m/step, thus irrespective of artistic, etc.,
Edge spacing of two slates (c) c = d-b the center-to-center spacing of two slates should
Center-to-center spacing of height conform to pedestrians’ stride range;
two slates (d ) gender 3. Among all the students, the mean of male feet
single/multi—walker size is 25.95 cm, while that of female is 23.48 cm,
the male average stride is 0.70 m/step, while
female is 0–65 m/step.
3.6 The pavement elements analysis of stone path On the basis of the above conclusions, this
based on KPI paper puts forward the suggested value of stone
path pavement elements.
KPI (Key Performance Indicators) is a kind of tar-
We simulated a temporary stone path in cam-
geted quantitative management indicators which
pus according to the suggested value in Table 6. In
measures the process performance through setting
order to explore whether this simulated one is more
values in, sampling, calculating, and analyzing the
comfortable than others, we conducted the verifi-
key parameters of the input and output end of one
cation experiment by using the EEG equipment.
process in an organization (Fu & Xu, 2001). It is
The verification includes experiments on
a tool which can change the strategic target into
15 groups of male and 15 groups of female. Each
an operational vision. Associated with the above
participant conducts 3 sets of comparative experi-
experimental analysis, the author gives the differ-
ments with head-mounted wireless EEG device
ent classes of KPI in stone path pavement.
(E2015U908-14). Participants in the 1st set walk
Feet size affects the slate width, thus the width
on stone path paved according to the parameters
of peddles should at least holds the length of a foot
in Table 6; participants in the 2nd set walk on the
(275 mm). Besides, it should reserve a 10 cm space
real stone path in campus (slate width b = 28 cm,
(Wu, 2002). Meanwhile, considering the cost, the
Edge spacing of two slates c = 27 cm, center-to-
length of the slate should not be too long. After
center spacing of two slates d = 55 cm, all of them
analyzing the samples, the author builds the fol-
are smaller than the suggested value in Table 6);
lowing formula between slate width b and gender:

b = 0.03x + 0.32 Table 6. The suggested value of stone path pavement

elements.
where x is the proportion of male in total number; Pavement site b c d
Unit: meter.
The indicators that influence the center-to- The institute with more male students
center spacing of two slates are height, gender, and (such as civil engineering institute 35 34 69
single/multi-walker. Since the height and single/ and mechanics institute)
multi-walker are hard to estimate in different col- The institute with more female students 33 32 65
leges and using phases, the KPI of them is set as 0. (such as school of arts and humanities
According to the stride data, the author builds the and school of foreign languages)
following formula between center-to-center spac- Institutes of equal proportion of male 34 33 67
ing of two slates d and gender: and female, the library, etc.

d = 0.05x + 0.65

Due to c = d-b, thus the relation of the edge

spacing of two slates c is as follow:

c = 0.02x + 0.33

4 VERIFICATION AND DISCUSSION

According to the analysis of experimental data, we

can receive the following conclusions: Figure 10. The simulated stone path (left).

465

TIS2017_Book.indb 465 2/8/2017 8:06:11 PM

 Figure 11. The real stone path in campus (right).

Figure 13. The mean EEG of female.

The result indicates that both male and female

will have an increase in EEG when they walk on
stone path. But when they walk on a stone path
paved as suggested values and a regular road, the
EEG remains a comparatively stable state. This
explains that there is scarcely any difference between
walking on a stone path built according to the sug-
gested values in Table 6 and walking normally, but
participants might feel uncomfortable when they
walk on the improper stone path. Comparing the
Figure 12. The mean EEG of male. EEG changes of male to that of female, we can see
that although both of them show big wave in EEG
participants in the 3rd set walk normally on the when walking in the real stone path, female’s EEG
ordinary road. The walking distance is 8 m, and par- fluctuation is obviously bigger than male’s. The
ticipants have 2 minutes to rest before they taking probable reason of this is that the stride of female
each experiment, ensuring there is no interference is smaller than male’s, and the feet size of female is
between experiments. After the test, we analyzed also shorter than that of male. Therefore, the stone
the changes in EEG of these participants. path paved in line with the parameters in table 6 is
The means of male and female walking on the more suitable for pedestrians to walk as compared
two different stone paths and walking normally are with other stone paths. These parameters are worth
listed hereinafter: referring to.

466

TIS2017_Book.indb 466 2/8/2017 8:06:11 PM

 5 EXPECTATION Lam, W.H.K. et al. 2002. A study of the bi-directional
pedestrian flow characteristics at Hong Kong sig-
The research results of the paper can provide refer- nalized crosswalk facilities [J]. Transportation
ence to the short stone path in campus. In consid- 29(2):169–192.
Laplante, J.N. & Kaeser, T.P. 2004. The continuing evolu-
eration of the widely usage of stone path and fully tion of pedestrian walking speed assumptions [J]. Ite
taking its advantages, we need do further studies in Journal 74(9).
the following aspects: Laxman, K.K. et al. 2015. Pedestrian Flow Characteris-
1. The selected road section in this experiment is tics in Mixed Traffic Conditions [J]. Journal of Urban
20 m long, which can meet the need of lawn Planning & Development 136(1):23–33.
Ma Y. et al. 2009. Impact Analysis of Pedestrian’s Char-
width linking the two adjacent buildings in the acteristics on Walking Behavior [J]. Traffic & Trans-
college. However, if the stone path is taken as portation: Academic Edition (1).
a long recreational trail, it needs to be further Mavros, P. et al. 2016. Geo-EEG: Towards the Use of
explored and studied integrated with pedestri- EEG in the Study of Urban Behaviour[J]. Applied
an’s physiological and psychological characters Spatial Analysis & Policy 9(2):1–22.
(including EEG, Electrocardiogram, Skin Con- Virkler, M.R. & Elayadath S. 1994. Pedestrian speed-flow-
ductance, etc.). density relationships[J]. In Transportation Research
2. In the analysis, the difference in personalities Record: Journal of the Transportation Research Board
has relatively small impact on the stride, but it 1438 (1):51.
Palmer A. 1975. Sex differences and the statistics of
has great influence on the stride frequency and crowd fluids[J]. Behavioral Science 20(4):223–227.
walking speed. In the research of the pavement Rastogi R. et al. 2013. Pedestrian flow characteristics for
elements of long stone path, further study of different pedestrian facilities and situations[J]. Euro-
stride frequency and walking speed is needed. pean Transporttrasporti Europei (53):5.
3. In the field research of sites like park, etc., we Sisiopiku V.P. & Akin D. 2003. Pedestrian behaviors at
find out that environmental factors such as lon- and perceptions towards various pedestrian facilities:
gitudinal slope, pedestrian volume, weather, etc. an examination based on observation and survey data
will also affect the pavement elements of stone [J]. Transportation Research Part F Traffic Psychology
path. Thus we need to consider the impact of & Behaviour 6(4):249–274.
Tanaboriboon Y. et al. 1986. Pedestrian Characteristics
environment as well in the further study of pave- Study in Singapore[J]. Journal of Transportation Engi-
ment elements of stone path. neering Asce 112(3):229–235.
4. The collected samples of this experiment are Wu T. 2002. Urban Detail: Analysis on Design prin-
only for college students. In the future it can be ciples and Methods (in Chinese) [D]. Chongqing
considered to extend to faculties, other people in University.
parks and communities, etc., so as to provide ref- Xu H. 2013. Research on Pedestrian’s Traffic Character-
erence for the pavement elements of stone path. istics in the Ladder section of Mountain City (in Chi-
nese) [D]. Chongqing Jiaotong University, 2013.
Yang L. 2009. Research on the Traffic Characteristics of
REFERENCES Pedestrian Comprehensive Passenger Transport Ter-
minal (in Chinese) [D]. Jilin University.
Alhajyaseen, W.K.M. et al. 2011. Effects of Bi-directional Zhang G. 2009. Research on the Pedestrian Behavior and
Pedestrian Flow Characteristics upon the Capacity of Microscopic Simulation on the Square of Passenger
Signalized Crosswalks [J]. Procedia - Social and Behav- Transport Terminal (in Chinese) [D]. Beijing Jiaotong
ioral Sciences 16(1):526–535. University.
Chen R. & Dong L. 2005. Observations and Preliminary Zhang J. 2014. Analysis on Traffic Characteristics of
Analysis of Characteristics of Pedestrian Traffic in Pedestrian at Signalized Intersection (in Chinese) [D].
Chinese Metropolis (in Chinese) [J]. Journal of Shang- Chang’an University.
hai University: Natural Science 11(1):93–97. Zhang Y. et al. 2012. Analysis of Walking Road System
Fu Y. & Xu Y. 2001. Performance Management (in Chi- in University Campus from the Perspective of Envi-
nese) [M]. Shanghai: Fudan University Press. ronmental Behavior (in Chinese) [J]. Journal of Xi’an
Han, G. 2014. On the Construction Techniques of Gar- Aviation College (5):60–63.
den Road Pavement (in Chinese) [J]. Architectural Zhao G. et al. 2006. Study on Pedestrian Flow Traffic
Engineering Technology and Design (14). Characters of Special Games (in Chinese) [J]. Road
Henderson, L.F. 1971. The statistics of crowd fluids. [J]. Traffic & Safety (2): 19–21.
Nature 229(5284):381–3. Zhu D. & Lin Y. 2015. Characteristics, Models and
Henderson, L.F. & Lyons, D.J. 1972. Sexual differences in Methods– A Preliminary Exploration of the
human crowd motion. [J]. Nature 240(5380):353–355. Microscopic Simulation of Pedestrian Connection
Henderson, L.F. 1974. On the fluid mechanics of Behavior in Influenced Urban Realm around Rail
human crowd motion[J]. Transportation Research Transit Stations (in Chinese) [J]. Architectural Jour-
8(6):509–515. nal (3):24–29.

467

TIS2017_Book.indb 467 2/8/2017 8:06:13 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Performance evaluation of regenerated HMA with very

high RAP content

N.A. Alvarado Patiño & J. Martinez

Laboratoire de Génie Civil et Génie Mécanique, INSA de Rennes, France

E. Lopez
Fenixfalt SAS, Buros, France

ABSTRACT: This research presents the thermomechanical performances of Hot Mix Asphalt (HMA)
containing conventional and regenerated Reclaimed Asphalt Pavement (RAP). Different mixtures
containing 40 to 70% RAP are evaluated and compared to a control mix (without RAP). The material
properties are determined through an extensive experimental program including workability, rutting
resistance, susceptibility to water, stiffness, fatigue resistance and performances at low temperature.
The results show that, compared to the control mixture, the use of RAP significantly increases the stiff-
ness of the mixes without a significant loss of fatigue resistance; however, RAP does reduce the per-
formance of HMA at low temperatures. The use of an additive allows the mixtures to partially recover
properties such as workability, fatigue resistance and ductility at low temperatures.

1 INTRODUCTION Secondly, a sufficient level of mechanical perform-

ances like cracking at low temperature and fatigue
1.1 Objectives and context resistance must be guaranteed. Current observations
show that when using mixtures with high percent-
The objective of this research work is to increase
ages of RAP, a reduction in workability, an increase
the rate of Reclaimed Asphalt Pavement (RAP) in
in rutting resistance and in elastic stiffness occur.
Hot Mix Asphalt (HMA) without important loss
This is due to the ageing of the RAP binder and the
of the thermomechanical performances by using
consequent hardening of the final binder (Shu et al.
an additive.
2008, Olard et al. 2009, Hajj et al. 2009, Silva et al.
An experimental laboratory program is devel-
2012, McDaniel et al. 2012, Mangiafico et al. 2015).
oped involving two main steps: (i) composition
The effect of RAP content on the fatigue resist-
and production of different mixtures with vari-
ance of the mixtures is observed differently, depend-
able RAP content with and without the additive;
ing on the authors and on the conditions. In some
(ii) testing of the general and thermo-mechanical
cases, up to a certain content of RAP, the fatigue
properties of the mixes. The material considered
resistance of the material is not reduced and is even
is a typical class-4 dense French asphalt known as
sometimes improved (Huang et al. 2005, Tabakovic
grave-bitume (GB4). For all the mixtures, the grad-
et al. 2010, Mangiafico et al. 2015, Perez-Martinez
ing curve of the aggregates and the final binder
et al. 2016). When the RAP content is increased,
content are kept constant and are in accordance
the fatigue resistance may even deteriorate com-
with French standards.
pared to the control mixture without RAP (Shu
et al. 2008, Hajj et al. 2009).
1.2 Previous results
For virgin HMA, the performances at low tem-
Using high rates of RAP in HMA gives rise to dif- peratures are mainly controlled by the type of
ferent limitations. First, the optimization of the binder (Arand 1990, Pucci 2001, Maia et al. 1997,
manufacturing process for mixing RAP and virgin Soenen & Vanelstraete 2003, Olard 2004). Informa-
components is necessary in order to guarantee cor- tion on the cracking resistance of RAP at low tem-
rect homogeneity between virgin and aged binders peratures is relatively recent, and supposes a loss in
(Olard et al. 2008, Navaro 2011, Valdés et al. 2011; performance (Huang et al. 2004, Hajj et al. 2009).
Vislavičius & Sivilevičius 2013). Moreover, differ- In order to compensate for performance loss in
ences between laboratory and plant production RAP mixtures, special binders and additives are
must be controlled (Reyhanegh & Daniel 2016). introduced allowing an increase in RAP content.

469

TIS2017_Book.indb 469 2/8/2017 8:06:15 PM

 As a consequence, an increase in workability has ous according to the RAP classification. Penetra-
been obtained (Silva et al. 2012, Lopez et al. 2015). tion and softening temperatures are representative
Moreover, the additives have enhanced the crack- of an aged binder from surface layers.
ing resistance of RAP (Lopez et al. 2011, Largeaud Different grades of the virgin binder (20/30,
et al. 2015, Porot et al. 2016). 35/50, 50/70 and 70/100) are used to ensure the
same class (20/30) of the final binder for all the
mixtures.
2 MATERIALS AND TEST METHODS The additive employed for this research (REGE-
FALT®) has the function of reducing the oxidized
2.1 Characterization of the material components hydrocarbon components included in the RAP
binder and to facilitate the interpenetration of the
The components are characterized exhaustively
particles of old and new binders, which leads to the
to reduce the heterogeneity of the final mixtures
integration of the components of the new bitumen
(Vislavičius & Sivilevičius 2013). The stocks of RAP
and to an increase in the performances of the mix-
and virgin aggregates are mechanically homogenized
tures (Lopez et al. 2011, Lopez et al. 2015).
in piles. A second laboratory homogenization is
required for each sample (one control test per tonne).
The gradation results are shown in Figure 1. 2.2 Tests performed for characterizing the mixes
All virgin aggregates used in this work come
The manufacturing conditions of the mixtures are
from massive rocks (sandstone quartzite), while
in agreement with standard specifications (NF
the filler is a limestone. The average density of the
EN 12697-35+A1 2007) and constant for all RAP
aggregates is 2630 kg/m3.
The main properties of RAP are summarized in
Table 1. Properties can be considered as homogene- Table 2. Tests performed for the formulation of the
mixes.

Measured properties Test NR**

Workability Gyratory shear press (GSP) 3

Sensitivity to water Duriez 5×2
Rutting resistance Wheel tracking test 2
Complex modulus 2PB-TR 4
Fatigue resistance 2PB-TR 6×3
Low temperature UTST 3×4
resistance* TSRST 3

*According to (NF EN 12697-46 2013).

**Number of replicates.

Figure 1. Grading curves of the different aggregates.

Table 1. Main properties of RAP.

Properties Test results

Type* 12.5 RA 0/10

Petrography Quartzite-eruptive
Density of particles (coated) 2472 kg/m3
Soluble binder content 5.11%
Penetration at 25°C 16.4 (0.1 mm)
Softening point (Ring and Ball) 65.5°C
Category* TL1 − B1 − G1 − R1
Figure 2. Equipment for the low temperature tests in
*According to (NF EN 13108-8 2006). climate chamber.

470

TIS2017_Book.indb 470 2/8/2017 8:06:15 PM

 contents: pre-heating of virgin aggregates and RAP obtain a conventional grading curve of GB4 mate-
at 190°C and 180°C respectively. In order to guar- rial. For all the mixtures, the RAP grading curve is
antee correct homogeneity between components, a considered in total without separation into differ-
mixing duration of 5 mn is adopted. The preheat- ent fractions; thus the constant grading of the dif-
ing temperature of the virgin binder depends on its ferent mixes is obtained by adjusting the fractions
penetration grade. of virgin aggregates (Table 3).
Standard tests performed on the mixes agree Constant total binder content is fixed for all the
with French specifications (Table 2). Three differ- mixtures by modifying the amount of the virgin
ent equipment are used for the compaction of the binder content (Table 3). Similarly, the final binder
samples. The GSP specimens have a diameter of penetration (class 20/30) is obtained by adjustment
160 mm and are tested at a temperature of 180°C. of the virgin binder grade using the usual mixing
The Duriez specimens, of 120 mm diameter, are law (NF EN 13108-8 2006):
compacted statically with application of a verti-
cal force of 180 kN during 5 mn. The specimens a lg pen1 + b pen2 = (a b ) lg penmixi (1)
for the remaining tests are compacted with a plate
compactor in rectangular molds of different sizes where penmix = calculated penetration of the binder
(length and width in mm): 500*180 for the rutting in the mixture with RAP; pen1 = penetration of the
tests and 600*400 for the others. In the rutting RAP binder; pen2 = penetration of virgin binder;
tests all the mixtures are compacted with the same a and b = portions by mass of the RAP binder (a)
energy, as for the following tests, the compaction and of the virgin binder (b). (a + b = 1).
energy is adapted to the required voids content. The additive content Q (‰) is determined from
Two types of low temperature tests are per- the proportion of RAP (R) in the mixture and the
formed using the equipment shown on Figure 2: RAP binder content (X) by the following empiri-
the Uniaxial Tension Stress Test (UTST) at four cal relation: Q = 36 R X.
different temperatures (+20°C, +5°C, −10°C and
−25°C) and the Thermal Stress Restrained Speci-
men Test (TSRST) with an initial temperature 4 GENERAL PROPERTIES
of + 10°C and a temperature gradient of −10°C/h.
The temperature is measured using thermo- All mixtures are manufactured according to the
couples fixed to the surface of the specimen and requirements of French standards for the GB4
both on the surface and inside a control sample for (NF EN 13108–1 2007). The grading curve and the
temperature correction (Figure 2). binder content are verified systematically for each
test.

3 MATERIALS COMPOSITION
4.1 Workability (GSP test)
Three proportions of RAP content are considered The workability tests are performed with the Gyra-
(40%, 55% and 70%, with and without additive) and tory Shear Press (GSP type 2). Figure 3 shows the
a control mixture containing only virgin aggregates results are shown in Figure 3 for the different mix-
and virgin binder. A constant grading curve for all tures in terms of voids content versus the number
the above mixtures is obtained by first mixing the of gyrations. All mixtures are in agreement
70% RAP formula with different proportions of with the French specifications (voids content at
the elementary fractions of virgin aggregates to 100 gyrations V100 < 9%). Compared to the control

Table 3. Main properties of the mixes.

Grading mixes (passing by weight) Virgin

Total Penetration Softening
binder
RAP 0.063 mm 2 mm 6.3 mm 10 mm 14 mm Additive binder at 25 °C point Density
Id (%) (%) (%) (%) (%) (%) (‰) Grade (%) (%) (1/10 mm) (°C) (kg/m3)

0R 0 8.4 30.9 59.6 81.3 97.0 – 20/30 4.90 4.90 29 61.6 2449
40R 40 8.4 31.0 59.2 81.4 97.1 – 35/50 2.85 30 58.0 2461
40R′ 40 8.4 31.0 59.2 81.4 97.1 0.7 35/50 2.85 30 58.0 2461
55R 55 8.4 31.0 59.8 81.9 97.1 – 50/70 2.09 29 58.6 2465
55R′ 55 8.4 31.0 59.8 81.9 97.1 1.0 50/70 2.09 29 58.6 2465
70R 70 8.4 31.0 59.3 81.8 97.1 – 70/100 1.32 24 60.9 2470
70R′ 70 8.4 31.0 59.3 81.8 97.1 1.3 70/100 1.32 24 60.9 2470

471

TIS2017_Book.indb 471 2/8/2017 8:06:17 PM

 Figure 3. Workability of the mixes (GSP test). Figure 4. Compressive strength ratio (i/C) and voids
content (Duriez test).

mix (0R), the mixtures containing RAP become

less workable. The additive, however, increases the
workability, especially for the mix with the high-
est rate of RAP (70R’), producing a voids content
equal to that of the control mix.

4.2 Sensitivity to water (Duriez test)

Susceptibility to water is determined using the
Duriez test which measures the ratio i/C of the
compressive strength in dry and saturated condi-
tions at 18°C. A slight reduction in the i/C ratio
is observed for the mixtures containing RAP
(Figure 4). Nevertheless, the ratio remains above
95%, which is far higher than the specifications
for this type of material (i/C > 70%). Moreover,
the voids content is observed to increase with the
ratio of RAP, so too does the favorable effect of
the additive: as in the GSP test.

4.3 Rutting resistance Figure 5. Initial voids content and rut depth measured
The rutting resistance is measured by the mlpc® at 30,000 cycles.
wheel tracking device at a temperature of 60°C.
The voids content V of the compacted plates fol- 5 STIFFNESS AND FATIGUE RESISTANCE
low the same variations displayed in the results
above, showing the favorable effect of the addi- The stiffness and the fatigue resistance are meas-
tive (Figure 5). The values are in agreement with ured on standard trapezoidal specimens at 4.5%
French specifications (8% < V < 5%). The rutting voids content, using strain-controlled two-point
percentages at 30,000 cycles (P30,000) decrease with bending tests (2PB-TR). Stiffness tests are per-
an increase in RAP content, and are slightly higher formed on four samples at a temperature of 15°C,
when using the additive, but the variations remain a frequency of 10 Hz and a strain level of about
low (about 1%). For all the mixes the rutting val- 39 μm/m. Fatigue tests are carried out according
ues stay small compared to the allowable values to the standards (10°C; 25 Hz) with three levels of
(<10%). deformation and six samples per level.

472

TIS2017_Book.indb 472 2/8/2017 8:06:17 PM

 5.1 Complex modulus and phase angle
Due to the hardness of the aged binder, the
mixes exhibit a complex modulus that increases
with the RAP content and is much higher (about
50%) than that of the control mix (Figure 6). For
the same reason, the phase angle decreases with
the RAP content (from 13 to 7 degrees). Similar
results have been obtained by Mangiafico et al.
(2015) and by Perez-Martinez et al. (2016) signi-
fying that RAP increases the elastic component
of the stiffness and reduces the viscous compo-
nent. On the other hand, the additive very slightly
reduces the complex modulus without affecting
the phase angle.

5.2 Fatigue resistance

The fatigue results (Figure 7 and Table 4) show Figure 7. Fatigue resistance of the mixes (10°C; 25 Hz).
that for all the mixtures, the fatigue resistance is
quite high (ε6 > 147 μm/m) and that, compared to Table 4. Synthesis of stiffness and fatigue results.
the control mix, the addition of RAP leads to a
small degradation of ε6 (less than 8 μm/m). E* φ ε6 Δε6
Moreover, an increase of ε6 is observed with the Mixture (MPa) (°) (μm/m) (μm/m) b
additive, especially for the mix 70R’, which is above
0R 12,285 13.1 154.8 5.6 −0.14
the control mix 0R. On the other hand, the scat-
ter Δε6 is slightly higher when using additive but 55R 17,751 8.9 147.2 3.9 −0.14
it remains of the same order as that of the control 55R’ 17,105 8.9 149.5 6.0 −0.18
material. Another effect of the additive which was 70R 19,093 7.3 150.4 3.6 −0.12
observed is the decrease in the slope b of the Wöh- 70R’ 18,844 7.3 163.0 5.6 −0.17
ler curve compared to the mixes without additive.
This leads the life duration of the pavement to be
6 LOW TEMPERATURE PERFORMANCES
less susceptible to the variations of the strain level,
when using the additive.
The specimens are cored from plates compacted
in the laboratory and have a cylindrical shape:
160 mm height by 50 mm diameter (Figure 2), with
a voids content of 5.5 ± 0.5%.

6.1 UTST results

The strain rate of the test (4,2 ± 0.8 10−6 s−1) is set at
a lower value to the one recommended by the stand-
ards, but this parameter has been shown to have
no effect on the results at low temperatures (Olard,
2004, Steiner et al. 2016). Typical stress-strain
curves show two types of fracture (Figure 8): frag-
ile fracture for low temperatures (−25°C; −10°C)
and ductile for higher ones (+5°C; +20°C). In the
case of ductile fracture, the tensile strength βt and
the failure strain εfailure are taken at maximum stress
according to the standards.
For all the mixtures the tensile strength is maxi-
mum for intermediate temperatures (Figure 9 and
Table 5) corresponding to the fragile-ductile transi-
tion temperature Tfde (Olard 2004). Compared to
the control mix, this temperature is increased by
Figure 6. Complex modulus and phase angle of the the presence of RAP: from −15°C for 0% RAP
mixes (15 °C; 10 Hz). up to 0°C for 70% RAP. The effect of the binder

473

TIS2017_Book.indb 473 2/8/2017 8:06:18 PM

 hardness on the tensile strength at low tempera-
tures has been shown by Olard (2004). For the
mixtures containing 70% RAP, the additive has a
favorable effect as it allows the enhancement of the
transition temperature (from 0°C to −3°C). In the
ductile domain the tensile strength increases sig-
nificantly with RAP content. This effect is due to a
higher mobilization of the aged binder through its
elastic stiffness (Olard 2004, Arand 1990).
The failure strain (Figure 10 and Table 5) dec-
rea-ses with RAP content due to the ageing of
the RAP binder, and the differences increase with
the temperature. On the other hand, the additive
increases the failure strain, allowing the recovery
of some of the ductility.

6.2 TSRST results and strength reserve

Figure 8. Typical stress-strain response from UTST The TSRST test simulates the thermal stress σcry
tests 40R’. produced in a longitudinally restrained pavement
caused by an environmental temperature drop.

Figure 9. Tensile strength versus temperature. Figure 10. Tensile failure strain versus temperature.

Table 5. Synthesis of UTST/TSRST results.

UTST TSRST Reserve

βt (at T °C) εfailure (at T °C) Δβt (at T °C)

−25 −10 +5 +20 −25 −10 +5 +20 Tfde σcry,failure Tfailure −20 +10
Mix (MPa) (MPa) (MPa) (MPa) (‰) (‰) (‰) (‰) (°C) (MPa) (°C) (MPa) (MPa)

0R 4.28 5.32 2.57 0.57 0.166 0.382 1.899 4.854 −15.3 3.74 −28.2 2.83 1.60
40R 3.52 5.46 4.79 1.14 0.134 0.272 1.068 3.343 −6.1 3.52 −24.0 1.34 3.95
40R’ 4.00 5.34 4.66 1.03 0.143 0.265 1.549 4.236 −6.3 3.67 −26.5 1.86 3.86
55R’ 4.01 5.21 4.87 1.14 0.155 0.261 1.503 3.821 −3.9 3.82 −24.0 1.15 4.16
70R 3.66 5.18 5.60 1.87 0.148 0.231 0.803 2.299 +0.3 3.66 −21.8 0.57 4.94
70R’ 3.92 5.61 5.52 1.65 0.161 0.263 0.963 2.971 −2.9 3.95 −23.9 1.02 4.68

474

TIS2017_Book.indb 474 2/8/2017 8:06:19 PM

 The test results are characterized by the maxi- the TSRST tests (NF EN12697–46 2013). It cor-
mum stress reached σcry,failure (cryogenic failure responds to the allowable stress afforded to traffic
stress) and by the corresponding temperature when the pavement is submitted simultaneously to
Tfailure. For all the tests, the initial temperature cryogenic stress.
(T0 = + 10°C) is chosen such that it has no effect on Figure 12 shows that both the maximum
the final result (Tušard et al. 2014). strength reserve and the corresponding tempera-
The results (Figure 11 and Table 5) show that the ture increase with RAP content.
failure stress does not depend on the RAP content, Moreover, the reserve curves allow us to differ-
but that the failure temperature increases signifi- entiate two domains depending on temperature
cantly, which leads to a reduction in the cracking (Table 5).
resistance of RAP mixtures at low temperatures. For temperatures higher than −5°C the RAP
Similarly to the UTST results, the additive allows mixtures exhibit significantly higher strength
the decrease of the failure temperature. reserves than the control mixture (up to + 110%).
The strength reserve Δβt(T) at a given tem- Inversely, for temperatures lower than −5°C the
perature T is defined as the difference between strength reserve of the control mix is higher than
the tensile strength βt(T) and the cryogenic stress that of the RAP mixtures. In this domain, the addi-
σcry(T) derived respectively from the UTST and tive increases the strength reserve up to 0.6 MPa.

7 CONCLUSIONS

The extensive set of tests performed on a dense HMA

with different RAP contents and the presence or
absence of an additive lead to several conclusions.
The increase in RAP content in the mixes shows
a reduction in the workability and a small effect on
the sensitivity to water and on the rutting resist-
ance. The elastic stiffness is strongly increased and
the fatigue resistance is slightly reduced. In terms
of behavior at low temperatures, the performances
are significantly reduced.
The use of the additive allows the partial or total
recovery of the performances deteriorated by the
RAP; such as workability and fatigue resistance.
Furthermore it has only a small effect on the elas-
tic stiffness. At low temperature, an increase of the
thermomechanical properties is observed when
Figure 11. Cryogenic stress from TSRST test.
using the additive.

ACKNOWLEDGEMENTS

This research project is part of the Ph.D. thesis pre-

pared by the first author under a contract CIFRE
2013/0334 between the French ANRT (Agence
Nationale de la Recherche et de la Technologie)
and the Fenixfalt company. The authors gratefully
acknowledge the following companies for their par-
ticipation in the project: Le Foll TP, Roger Martin
and Braja-Vésigné. The authors are also grateful
to C. Garand from the LGCGM laboratory of
INSA—Rennes, for his help in the layout and in
the organization of the low temperature tests.

REFERENCES

Arand, W. 1990. Behaviour of asphalt aggregate mixes

Figure 12. Tensile strength reserve of the mixes. at low temperatures. Mechanical Tests for Bituminous

475

TIS2017_Book.indb 475 2/8/2017 8:06:20 PM

 Mixes, Proceedings of the 4th RILEM Symposium: Olard, F., Bonneau, D., Dupriet, S., Beduneau, E. &
67–81. Budapest. Seignez, N. 2009. Formulation spécifique en laboratoire
Hajj, E.Y., Sebaaly, P.E., & Shrestha, R. 2009. Laboratory des enrobés à très fort taux de recyclés (jusqu’à 70%), à
Evaluation of Mixes Containing Recycled Asphalt chaud ou en technique E.B.T.®. Revue générale des routes
Pavement (RAP). Road Materials and Pavements Design (RGRA) (872): 62–66.
10(3): 495–517. Olard, F., Le Noan, C, Bonneau, D., Dupriet, S. & Monnier,
Huang, B., Kingery, W. & Zhang, A. 2004. Laboratory Y. 2008. Le recyclage à très fort taux (>50%) des enrobés
Study of Fatigue Characteristics of HMA Mixtures chauds et des enrobés tièdes pour une construction routière
Containing RAP. International Symposium on Design durable. Revue générale des routes (RGRA) (866): 56–62.
and Construction of Long Lasting Asphalt Pavements: Perez-Martinez, M., Marsac, P., Gabet, T., Hammoum,
501–522. Auburn. F., de Mesquita Lopes, M. & Pouget, S. 2016. Effects
Huang, B., Li, G., Vukosavljevic, D., Shu, X. & Egan, B.K. of Ageing on Warm Mix Asphalts with High Rates of
2005. Laboratory Investigation of Mixing Hot-Mix Reclaimed Asphalt Pavement. In Chabot et al. (eds),
Asphalt with Reclaimed Asphalt Pavement. Transporta- Proc. 8th RILEM International Conference on Mecha-
tion Research Record (1929): 37–45. nisms of Cracking and Debonding in Pavements, Nantes,
Largeaud, S., Faucon-Dumont, S., Eckmann, B., Hung, 7–9 June 2016: 113–118. Springer.
Y., Lapalu, L. & Gauthier, G. 2015. Caractérisation Porot, L. & Vuillier, B. 2016. SYLVAROAD™
du comportement à basse température des liants bitu- RP1000 Additive, un régénérant BIO pour AE à forts
mineux. Revue générale des routes et de l’aménagement taux. Revue générale des routes et de l’aménagement
(RGRA) (928): 70–77. (RGRA) (933): 54–57.
Lopez, E., Gasca-Allue, C., Chifflet, P. & García-Serrada, Pucci, T. 2001. Approche prévisionnelle de la fissuration
C. 2011. La nouvelle vie du bitume. Une innovation par sollicitation thermique des revêtements bitumineux.
récente de régénération des bitumes vieillis des enrobés. Thèse de doctorat: École Polytechnique Fédérale de
Congrès mondial de la Route (AIPCR). Mexico. Lausanne.
Lopez, E., Willem, M. & Mabille, C. 2015. Bétons bitu- Reyhanegh, R.-R. & Daniel, J.S. 2016. Mixture and Pro-
mineux recyclés et régénérés à fort taux d’AE. Revue duction Parameters Affecting Cracking Performance
générale des routes et de l’aménagement (RGRA) (928): of Mixtures with RAP and RAS. In Chabot et al. (eds),
22–27. Proc. 8th RILEM International Conference on Mecha-
Maia, A.F., Marciano, Y., Archimastos, L., Corté, J.-F., nisms of Cracking and Debonding in Pavements, Nantes,
Gourdon, J.-L., Piau, J.-M. & Faure, M. 1997. Low- 7–9 June 2016: 307–312. Springer.
temperature behaviour of hard bitumens: experiments Shu, X., Huang, B. & Vukosavljevic, D. 2008. Laboratory
and modeling. Mechanical Test for Bituminous Mixes evaluation of fatigue characteristics of recycled asphalt
(MTBM), proceeding of the 5th. lnternational RILEM mixtures. Construction and Building Materials 22(7):
Symposium. Lyon. 1323–1330.
Mangiafico, S., Sauzéat, C., Di Benedetto, H., Olard, F., Silva, H., Oliveira, J.R. & Jesus, C.M. 2012. Are totally
Pouget, S, Dupriet, S. & Van Rooijen, R. 2015. Effet de recycled hot mix asphalts a sustainable alternative for
l’ajout d’agrégat d’enrobé sur le module complexe et road paving?. Resources, Conservation and Recycling 60:
la résistance en fatigue des enrobés bitumineux. Revue 38–48.
générale des routes et de l’aménagement (RGRA) (928): Soenen, H. & Vanelstraete, A. 2003. Performance indicators
35–40. for low temperature cracking. 6th International RILEM
McDaniel, R., Shah, A., Huber, G.A. & Copeland, A. 2012. Symposium on Performance Testing and Evaluation of
Effects of reclaimed asphalt pavement content and virgin Bituminous Materials: 458–464. Zurich.
binder grade on properties of plant produced mixtures. Steiner, D., Hofko, B., Dimitrov, M. & Blad, R. 2016.
Road Materials and Pavements Design 10(3): 495–517. Impact of Loading Rate and Temperature on Tensile
Navaro, J. 2011. Cinétique de mélange des enrobés recyclés Strength of Aspahalt Mixtures at Low Temperatures.
et influence sur les performances mécaniques. Thèse de In Chabot et al. (eds), Proc. 8th RILEM International
doctorat: École Nationale Supérieure d’Arts et Métiers Conference on Mechanisms of Cracking and Debonding
ParisTech. in Pavements, Nantes, 7–9 June 2016: 69–74. Springer.
NF EN 12697-35+A1 2007. French standard. Bituminous Tabakovic, A., Gibney, A., McNally, C. & Gilchrist, M.D.
mixtures. Test methods for hot mix asphalt. Part 35: 2010. Influence of recycled asphalt pavement on fatigue
Laboratory mixing. AFNOR. performance of asphalt concrete base courses. Journal of
NF EN 12697-46 2013. French standard. Bituminous Materials and Civil Engineering 22(6): 643–650.
mixtures. Test methods for hot mix asphalt. Part 46: Low Tušar, M., Hribar, D. & Hofko, B. 2014. Impact of char-
temperature cracking and properties by uniaxial tension acteristics of asphalt concrete wearing courses on crack
test. AFNOR. resistance at low temperatures. Transport Research
NF EN 13108-1 2007. French standard. Bituminous Arena. Paris.
mixtures. Material specifications. Part 1: Asphalt Valdés, G., Pérez-Jiménez, F., Miró, R., Martínez, A. &
Concrete. AFNOR. Botella, R. 2011. Experimental study of recycled asphalt
NF EN 13108-8 2006. French standard. Bituminous mixtures with high percentages of reclaimed asphalt
mixtures. Material specifications. Part 8: Reclaimed pavement (RAP). Construction and Building Materials
Asphalt. AFNOR. 25: 1289–1297.
Olard, F. 2004. Comportement thermomécanique des Vislavičius, K. & Sivilevičius, H. 2013. Effect of reclaimed
enrobes bitumineux à basses températures. Relations asphalt pavement gradation variation on the homoge-
entre les propriétés du liant et de l’enrobé. Thèse de doc- neity of recycled hot-mix asphalt. Archives of Civil and
torat: Institut National des Sciences Appliquées de Lyon. Mechanical Engineering 13(3): 345–353.

476

TIS2017_Book.indb 476 2/8/2017 8:06:21 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Finite element modelling of flexible pavement reinforced with geogrid

G. Leonardi & R. Palamara

“Mediterranea” University of Reggio Calabria, Reggio Calabria, Italy

ABSTRACT: The always increasing patterns of vehicle traffic on roads, in all weather, sets that the
maintenance and rehabilitation of infrastructures among the engineering key tasks for transport infra-
structures. Furthermore, the need to guarantee high performance wilts increasing pavement life, has led
research and industry to focus a greater attention on the use of pavement reinforcements. In recent years,
numerous kinds of geosynthetic grids have been introduced on the market, which can be used for pave-
ment reinforcement, extending extend pavement life, guaranteeing high performance and reducing costs
of service and maintenance. The use of geosynthetics can produce several benefits, such as drainage,
reinforcement, filter, separation and proof. In this paper, the effectiveness of geogrids as reinforcement
was investigated. The study proposes a numerical investigation using a three-dimensional Finite Element
Method (FEM) to analyze the importance, in terms of rutting at the top of pavement system, of geogrids
in the behavior under wheel traffic loads on pavement. In the model, a multi-layer pavement structure was
considered with a geogrid reinforcement and the model dimensions, element types and meshing strategies
are taken by successive attempts to obtain desired accuracy and convergence of the study. FEM results
show that a geogrid reinforcement can provide lateral confinement at the bottom of the base layer by
improving interface shear resistance and reduce rutting at the top of pavement system.

1 INTRODUCTION Herbst et al. [10] illustrated an interesting set of

data from an experimental site in Austria, where
In recent years, the need to increase pavement the comparative benefits of geogrids and geotex-
service life and to guarantee high performance has tiles could be directly assessed. Penman & Hook
turned a greater attention on the use of pavement [14] described how glass-fiber based geogrids had
reinforcements. During the life period of pavement successfully used as interlayers to extend the design
structure, it is vulnerable to different kinds of dis- life of asphalt pavements on airport runways, taxi-
tresses. In fact, under the traffic load the asphalt ways and aprons.
mixtures are submitted to elastic, plastic, visco- The results obtained by Chehab et al. [15]
elastic and visco-plastic deformations. Permanent revealed a partial improvement in reflective crack-
deformation (rutting) is one of the serious distresses ing resistance due to the incorporation of fiber-
in which pavement structure may be involved. reinforced interlayer. It is clear from these studies
Rutting can be defined as the permanent defor- that many practitioners see significant benefit in
mation of pavement along the wheel path caused using asphalt geogrid reinforcement.
by load traffic. To estimate permanent rut depth of Designing a flexible pavement reinforced with
the asphalt layer, in general nonlinear visco-plastic glass fiber grid and evaluating the effectiveness
Finite Element Analysis (FEA) is commonly per- of reinforced pavement performance is a complex
formed using different programs like ABAQUS, problem requiring considerable research and study
ANSYS or PLAXIS. A lot of research has been [Buonsanti et al, 2012; Leonardi, 2015]
conducted to reduce the pavements rutting phe- In this study, a three-dimensional finite element
nomenon. One of the recent methods is related to model, using ABAQUS software, was developed to
reinforcing pavement structures by means of the investigate the importance of geogrids reinforce-
use of geogrids and in general by the use of differ- ment, in terms of rutting. Furthermore, the influ-
ent geosynthetic types. ence of the position of geogrid reinforcement has
The use of geogrid reinforcement in the construc- been investigated considering three different posi-
tion of road pavement started in the 1970s. Then, tions within the pavement. The geogrid reinforce-
the technique of geogrid reinforcement has been ment layer was first placed at the HMA—base
increasingly used and many experimental and ana- interface, then it was placed in the middle of the
lytical studies have been performed to assess geogrid base course and finally at 2/3 depth of the pave-
behavior in the flexible pavement [Howard, 2009]. ment surface.

477

TIS2017_Book.indb 477 2/8/2017 8:06:22 PM

 2 ASPHALT REINFORCEMENT capacity, is achieved by shifting the failure surface of
AND PERMANENT DEFORMATION the pavement system from the relatively weak sub-
grade to the relatively strong base course material.
2.1 Flexible pavement improvement by As the of permanent deformations increase
geosynthetics: functions and mechanisms the tension membrane mechanism develops. If
the geosynthetic has a sufficiently high tensile
The geosynthetics used in flexible pavement are essen-
modulus, tensile stresses will be mobilized in the
tially fiber glass grids and geogrids. Geogrids have
reinforcement, and a vertical component of this
traditionally been used in three different pavement
tensile membrane resistance will help to support
applications: (a) mechanical sub-grade stabilization,
the applied wheel loads.
(b) aggregate base reinforcement, and (c) asphalt con-
crete (AC) overlay reinforcement. Geogrids used like
reinforcement in transport application perform two 2.2 Permanent deformation (Rutting)
primary functions: separation and reinforcement, in
which the geogrid improves the mechanically prop- Permanent deformation in the form of rutting is
erty of the pavement system [Cardile et al 2014, one of the most important failure mechanisms in
Calvarano et al. 2016]. Three fundamental reinforce- asphalt pavements.
ment mechanisms have been identified involving Rutting is a serious issue for road users for two
the use of geogrids to reinforce pavement materials: reasons: driving in the rut is difficult and the water
lateral restraint (Fig. 1), improved bearing capacity collecting in these depressions cannot drain freely
(Fig. 2), and tensioned membrane effect (Fig. 3). off the pavement surface and may cause aquaplan-
Lateral restraint refers to the confinement of the ing, making rutting a potential safety hazard. Rut-
aggregate material during loading. The geogrid mini- ting in flexible pavement develops gradually with
mizes lateral movement of aggregate particles and increasing numbers of load applications, usually
increases the modulus of the base course, which leads appearing as longitudinal depressions in the wheel
to a wider vertical stress distribution and consequently paths accompanied by small upheavals to the sides.
to a reduction of vertical deformations [Calvarano In addition, as a consequence of the increased tire
et al. 2016]. The second mechanism, improved bearing pressure and axle load, the surfacing asphalt layer
is subjected to increased stresses, which result in
permanent (irrecoverable) deformations. The per-
manent deformation accumulates with increas-
ing number of load applications. The permanent
deformation in the surface layer, thus, accounts
for a major portion of rutting on flexible pave-
ments subjected to heavy axial loads and high tire
pressures.
The main cause of rut initiation is shear strains
Figure 1. Lateral restraint. in asphalt. Generally, there are three causes of rut-
ting in asphalt pavements: accumulation of per-
manent deformation in the asphalt surfacing layer,
permanent deformation of subgrade, and wear of
pavements caused by studded tires. These three
causes of rutting can act in combination, i.e., the
rutting could be the sum of permanent deforma-
tion in all layers and wear from studded tires. In the
past, subgrade deformation was considered to be
the primary cause of rutting and many pavement
Figure 2. Bearing capacity increase restraint.
design methods applied a limiting criterion on ver-
tical strain at the subgrade level. However, recent
researches indicate that most of the rutting occurs
in the upper part of the asphalt surface layer.

3 METHODOLOGY

3.1 Finite Element Model

The present work aim to model reinforced and
Figure 3. Membrane tension support. unreinforced flexible pavement using 3-D analysis

478

TIS2017_Book.indb 478 2/8/2017 8:06:23 PM

 with ABAQUS software, by modelling its dimen-
sions, layers’ characteristics and different mesh-
ing. The objective is to compare an unreinforced
flexible pavement section with a reinforced sec-
tion in terms of surface rutting, considering for
the last case different positions in the system of
layers.
Despite the fact that it requires considerably
more computational time and computer memory,
the 3D analysis is still considered superior to the
2D. The necessity for adopting the 3D analysis
arises from the following advantages:
1. it better reflects the complex behavior of the
composite pavement system materials under
different configurations traffic loads,
2. it is preferred when verifying numerical model
results with laboratory or field test results,
3. it allows the simulation of the loaded wheel rec- Figure 5. Open meshes dimension.
tangular footprint [Bassam Saad, 2005].

All pavement layers were modelled by using 3D

3.2 Model and materials
deformable solid homogeneous elements except for
The flexible pavement system (Fig. 4) used in 3-D the geogrid, modelled like a membrane.
ABAQUS software consisted of asphalt concrete The materials data required for the simulations
(HMA) surface layer, base layer, sub-base layer have been assumed to represent realistic material
and sandy subgrade layer, subjected to traffic load properties published in previous studies.
with 0.01 second loading period. An elastoplastic material model for the asphalt
To model the flexible pavement 5 parts were concrete layer is necessary to allow this layer to
instanced, simulating 5 m wide and 5 m long lay- deform permanently. Permanent deformations in
ers with variable thickness, in particular: HMA layers with only elastic properties are zero when
surface course thickness is 0.076 m, base course the load is removed. Based on previous analyses,
thickness is 0.216 m, sub-base course thickness creep, Drucker Prager, and modified Drucker
is 0.254 m and sandy sub-grade layer thickness is Prager material models have been used for HMA,
1.454 m. The geogrid has been modelled as 5 m base, and subgrade layers, respectively.
wide layer with a thickness of 0.003 m and with For example, the model used in the study con-
open meshes (Fig. 5). ducted by Zaghloul and White (1993) employed 3D
Model dimensions have been selected to reduce dynamic finite element was a visco-elastic model
any edge effect error, keeping the elements’ sizes for the asphalt concrete, an extended Drucker-
within acceptable limits. Prager model for granular base course and Cam
Clay model for the clay subgrade soils.
Zaman et al. (2003) successfully developed a vis-
coelastoplastic creep model representing the time-
dependency of asphalt mixtures to evaluate their
rutting potential and to identify factors having a
significant effect.
Therefore, as regards the HMA layer a mechani-
cal model of Creep type has been chosen, for the
geogrid and the base layer a linear elastic behavior,
for the remaining layers a mechanical elastic-elastic
behavior, according to the Drucker-Prager model
of hyperbolic [Leonardi, 2015]. Table 1 shows the
mechanical characteristics of the various layers.
After defining the geometry and the mechanical
properties of the various parts the features inter-
face assignment and the mesh creation have been
done. The interface mechanical characteristics are
Figure 4. Flexible pavement system. very important because they significantly affect

479

TIS2017_Book.indb 479 2/8/2017 8:06:23 PM

 Table 1. Mechanical characteristics.

Landslide
susceptibility HMA Base Subbase Subgrade Geogrid

Modulus of 1 e9 316 e6 41 e6 10 e6 28 e9
elasticity
[Pa]
Poisson’s ratio 0.35 0.35 0.3 0.3 0.3
Friction angle 15 10
[°]
Initial tension 10000 10000
[kPa]

A 9 e9
m 0.67
Figure 6. Mesh model.
n −0.5

the response of the model depending on the type

of contact that is considered.
As regards the behavior at the interface between
the geogrid and the layer of HMA, the interaction
is constituted by the tangential and normal com-
ponent. For the normal component, the hard con-
tact type behavior has been considered while in the
tangential direction a penalty type behavior with a
coefficient of friction equal to 1.5 has been chosen.
The same features have been used for the geogrid
and the base layer interface while, to simplify the
model, the other contact surfaces has been defined
as tie constraint, able to limit both the displace- Figure 7. Geogrid mesh model.
ments and the relative rotations.
The mesh model dimensions have been selected
so to reduce any edge effect errors. The generated
mesh has been designed to give an optimal accuracy.
To improve the convergence speed, linear brick ele-
ments 8 nodes reduced integration (C3D8R) have
been used for all layers (Fig. 6), whereas for the
geogrid the mesh membrane type M3D4 (4-node
quadrilateral membrane) has been used (Fig. 7).

3.3 Load and boundary conditions

The traffic load is considered as the main factor in
designing the pavement system. This includes axial
loads, configuration of axles, tire contact areas,
number of load repetitions, vehicle speed. Heavy
vehicles, such as trucks, significantly influence Figure 8. Load area and boundary conditions.
pavement distresses and failure. For loading area,
wheel load can be applied on the wheel area, on the For the loading duration equivalent, pulse, and
entire wheel path, or on an equivalent circular/semi- moving loading are the most common methods.
circular area. In this study the traffic load has been In this analysis, the application of the load
applied in correspondence of the central part of the impulse time, modelled as a truck speed of about
model, simulating the passage of a twin wheel of a 50 km/h, has been considered and calculated using
heavy vehicle having an axle of 80 kN. The load has the formula proposed by Brown (1973):
been simulated as a pressure of 550 kPa acting on a
17 × 19 cm rectangular area (Fig. 8). log t . h − 0.2 − 0.94 ⋅ log V

480

TIS2017_Book.indb 480 2/8/2017 8:06:24 PM

 where t is the time in seconds of application, h is The pavement behavior has been simulated
the thickness of the surface layer in meters, V is the and the displacements have been considered as a
speed in km/h. response of traffic load applied. The final magni-
Since the boundary conditions have a significant tude of the displacements, U, evaluated at the load
influence in predicting the response of the model center at the end of loading, resulted to be affected
and given the size adopted, the model has been con- by the position of the geogrid in flexible pavements.
strained at the bottom (U1 = U2 = U3 = UR1 = UR The final magnitude of the surface displacements
2 = UR3 = 0); X-Symm (U1 = UR2 = UR3 = 0) U was 1.76 mm (Fig. 12) for the first configuration,
on the sides parallel to y-axis; and Y-Symm 1.75 mm (Fig. 13) for the second configuration and
(U2 = UR1 = UR3 = 0) on the sides parallel to 1.4 mm (Fig. 14) for the third configuration, while
x-axis. for the unreinforced test section the surface dis-
placement was 2.01 mm (Fig. 15).
4 ANALYSIS AND RESULTS
4.1 Geogrid position
The present study has been undertaken to investigate
the optimum position of the geogrid in a flexible
pavement system. Therefore, a series of finite element
simulations have been carried out to evaluate the ben-
efits of the different geogrid positions. Three different
configurations have been studied, namely the HMA-
base layer interface (Fig. 9), in the middle of base Figure 12. Displacements for the first configuration
course (Fig. 10) and inside the HMA layer at a height (Fig. 10).
of 1/3 of its thickness from the bottom (Fig. 11).

Figure 13. Displacements for the second configuration

(Fig. 11).
Figure 9. First configuration: geogrid at the HMA-
base layer interface.

Figure 10. Second configuration: geogrid inside the Figure 14. Displacements for the third configuration
base layer. (Fig. 12).

Figure 11. Third configuration: geogrid inside the Figure 15. Displacements for the unreinforced test
HMA layer. section.

481

TIS2017_Book.indb 481 2/8/2017 8:06:26 PM

 Figure 16. Rutting comparison.

Table 2. Displacements and improvement.

Displacements Improvement
Section [mm] [%]

reinforcement 1.40 30.00

inside the HMA
layer at a 1/3
reinforcement in 1.76 12.40
HMA-base layer
interface Figure 17. Displacement in correspondence of the
reinforcement in 1.75 12.90 geogrid.
the middle of
base course
Unreinforced test 2.01
section
5 CONCLUSION

Several researches have been conducted to reduce

The examination of the permanent surface
the rutting phenomenon in flexible pavements.
deformation after the traffic load showed that the
One of the prevention methods is the geosynthetic
results for the geogrid-reinforced test sections are
reinforcement. The analysis conducted in this
lower than the results obtained for unreinforced
paper with regards to the flexible pavement rein-
test section (Fig. 16).
forcement, showed that the geogrid reinforcement
The results show the improvement offered by
positively influences the permanent deformation
the reinforcement in flexible pavement at the dif-
of asphaltic surface; in fact, it may be observed
ferent geogrid positions. The presence of the geog-
a significant decrease in the vertical displacement
rid implies an improvement in term of surface
obtained for the reinforced pavement section com-
displacement reduction, corresponding to the
pared to the unreinforced pavement section.
12.4%, 12.9% and 30% for test section with
Based on the 3-D ABAQUS software outputs
the geogrid placed at HMA-base layer interface,
applied on pavement structure to evaluate the ben-
in the middle of base course and inside the HMA
efits of reinforcing pavement with geogrid at three
layer at a distance of 2/3 from the bottom respec-
positions, the following conclusions can be drawn:
tively (Table 2). Results confirm that there is higher
efficiency when the geogrid is placed inside the 1. A significant improvement in the pavement
HMA layer at a distance of 2/3 from the bottom. behavior is obtained by placing the geogrid
In this test session, the displacement of the rein- layer inside the HMA layer at a distance of 2/3
forcement geogrid following the application of the of thickness. Vertical displacement and effec-
traffic load was equal to 1.3 mm (Fig. 17). tive stress responses are significantly lower for

482

TIS2017_Book.indb 482 2/8/2017 8:06:29 PM

 reinforced pavement system in comparison with Chehab, G., Chaignon, F., Thompson, M., and Palacios,
unreinforced pavement. C. (2008), Evaluation of fiber reinforced bituminous
2. Moderate improvement in pavement system interlayers for pavement preservation. In: Pavement
behavior is gained by adding geogrid at the top Cracking, CRC Press.
Herbst, G., Kirchknopf, H., and Litzka, J. (1993), Asphalt
of sub-base layer. overlay on crack-sealed concrete pavements using stress
3. The best location of adding geogrid results to distributing media. In: 2nd Int. RILEM Conf. Reflec-
be in the vicinity of the point of tire pressure tive Cracking in Pavements, Liege, CHAPMAN &
application within the asphalt concrete layers. HALL.
4. The use of geogrid significantly enhances the Hook, K. and Penman, J. (2008), The use of geogrids to
asphalt concrete resistance to the deformation retard reflective cracking on airport runways, taxiways
and to the rutting failure development. and aprons. In: Pavement Cracking, CRC Press.
Howard, I. L. and Warren, K. A. (2009). Finite-element
modelling of instrumented flexible pavements under
stationary transient loading. J. Transportation Eng.
REFERENCES ASCE, 135 (2): 53–61.
Leonardi, G., (2015), Finite element analysis for airfield
Buonsanti, M., Leonardi, G., and Scopelliti, F. (2012), asphalt pavements rutting prediction. Bulletin of the
Theoretical and computational analysis of airport Polish academy of Sciences: Technical Sciences, Vol-
flexible pavements reinforced with geogrids. RILEM ume 63, Issue 2, 1 June 2015, Pages 397–403.
Bookseries, Volume 4, 2012, Pages 1219–1227. Saad, B., Mitri, H., and Poorooshasb, H., (2005),
Calvarano, L.S., Gioffrè, D., Cardile, G., Moraci, N., Three-Dimensional Dynamic Analysis of Flexible
(2014), A stress transfer model to predict the pullout Conventional Pavement Foundation. In Journal of
resistance of extruded geogrids embedded in com- Transportation Engineering.
pacted granular soils, in proceeding of 10th Inter- Zagh, S., and White, T., (1993) Use of a Three-
national Conference on Geosynthetics ICG 2014, Dimensional, dynamic finite element program
Berlin. for analysis of flexible pavement, Transportation
Calvarano L. S., Palamara R., Leonardi G., Moraci N. Research Record.
(2016). Unpaved road reinforced with geosynthetics. Zaman, M., Pirabaroban S., and Tarefder, R., (2003)
Procedia Engineering, 158 (2016) pp 296–301, ISSN: Evaluation of rutting potential in asphalt mixes using
1877–7058- DOI:10.1016/j.proeng.2016.08.445, Ed. finite element modelling, in Annual Conference of the
Elsevier. Transportation Association of Canada.

483

TIS2017_Book.indb 483 2/8/2017 8:06:30 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Critical success factors for PPP infrastructure projects in India

R.P. Pradhan, A. Verma, S. Dash, R.P. Maradana & K. Gaurav

Indian Institute of Technology Kharagpur, West Bengal, India

ABSTRACT: Public Private Partnership (PPP) mode is a strategic step to boost infrastructure develop-
ment in an economy. However, the achievement of this strategy depends upon the stress of its Critical
Success Factors (CSFs). This research investigates the CSFs for PPP infrastructure projects in India. Using
Principal Component Analysis (PCA), the study finds the factors- lack of community participation, lack of
political will and absence of good governance- that delay the projects and make them financially unviable.

1 INTRODUCTION 1999; Boynton and Zmud, 1984; Brodie, 1995; Chan

et al., 2010; Doloi, 2009; Grant, 1996; Li et al., 2005;
Infrastructure is a key to development. However, Mohr and Spekman, 1994; Ng et al., 212; Qiao et al.,
the association between these two varies from place 2001; Sanvido et al., 1992; Scandizzo, 2010; Tiong,
to place, time to time, and sector to sector. India is 1996; Verma et al., 2016; Verma, 2014; Yeo, 1991;
no exception to this issue. Infrastructure problems Zhang, 2005). All these factors are categorically
in India are very acute and range from poor condi- placed in four heads, namely technical, financial &
tion of roads to shortage of energy. The airports economic, political & legal, social & risk-oriented
as well as seaports of India need to be upgraded to factors. We develop a questionnaire to know the
meet international standards. With rapid urbani- importance of these factors in each category. About
zation for its ever-growing population, there is an four hundred respondents were contacted to gather
additional demand for all kinds of infrastructure, this information. The respondents are mostly execu-
such as electricity, drinking water, transportation tives of PPP infrastructure projects in India.
and other basic amenities. The growth of these The study deploys Principal Component Analy-
infrastructure requires enormous investment in the sis (PCA) to identify the most critical success fac-
Indian economy. However, due to lack of budg- tors. PCA is a data driven technique and is mostly
etary finance, government encourages the Public used to reduce the original factors into a relatively
Private Partnership (PPP) mode of infrastructure lower number which, can be used to represent rela-
development. Still after more than two decades tionships among sets of interrelated variables. It
of globalization and privatization, the PPP mode can be done by employing any one of the following
of infrastructure development is not so common rotations: orthogonal rotation (e.g. varimax), or
in India. Consequently, the objective of this paper oblique rotation (e.g. promax).
is to examine the critical success factors for PPP
infrastructure projects in the Indian economy.
The remainder of this paper is organized as fol- 3 EMPIRICAL RESULTS
lows. Section 2 describes methodology. Section 3
describes the empirical results. Section 4 discusses Before proceeding with PCA results, the appropri-
the results. The final section concludes with policy ateness of this analysis was evaluated by employ-
implications. ing sampling adequacy tests, Kaiser-Meyer-Olkin
(KMO) test, Bartlett’s test, and correlation matrix
of responses. We do not report these results due to
2 RESEARCH METHODOLOGY space restriction. We report the PCA results below in
the form of component extraction and component
The investigation of this problem is a two-step proc- loadings.
ess. First, the identification of Critical Success Factors Four components are extracted in this analy-
(CSFs) for PPP infrastructure projects; and second, sis and they explain 51% of total variance in the
pointing out the most suitable CSFs for successful original dataset. This is further verified by analyz-
implementation of PPP infrastructure projects. In ing component correlation matrix. Correlations
the first phase, on the basis of existing literature, we among all the four components are found to be low
identify twenty six critical success factors for PPP (i.e., below 0.3) and hence, this justifies the selec-
infrastructure projects (Akintoye et al., 2003; Birnie, tion of four components (see Table 1).

485

TIS2017_Book.indb 485 2/8/2017 8:06:32 PM

 In response to above identified factors, we use To avoid repetition in loadings, a choice was made
scree plot (see Fig. 1) to authenticate these results. between items of highly similar content on the
Table 2 provides the loadings of all four factors, basis of loading size.
which have been obtained on the basis of PCA. From PCA results, the 26 CSFs are finally
reduced to 22 and these are categorically clus-
Table 1. Component correlation matrix. tered into 4 major factors. These include factor
Component C1 C2 C3 C4
1- stable project environment-political and social,
factor 2- shared role of public and private parties
C1 1.000 in project execution, factor 3- favorable environ-
C2 0.051 1.000 ment for service delivery by private consortium,
C3 0.049 0.070 1.000 and factor 4- government control over risk allo-
C4 −0.027 0.145 0.039 1.000 cation. The sub-success factors under each of
these four factors (Factors 1–4) are illustrated in
Note: C1-C4 are components of CSFs. Table 3.

Table 3. Final component loadings.

Factors and description

Component F1 F2 F3 F4

P3 0.826
R5 0.659 Stable
project
S4 0.857 Environ-
Figure 1. Scree Plot of Eigen Values over Components. mental-
Table 2. Pattern matrix with component loadings. S1 0.635 Political
and
Component F1 F2 F3 F4 P5 0.597 Social
F4 0.712 Shared
P3 0.826 role of
F5 −0.763 public
S5 −0.738 −0.301 and
R5 0.659 T2 0.690 private
parties
S4 0.857
in project
S1 0.635
T3 0.648 execution
P5 0.597 0.398
P4 0.632
R6 −0.546 0.451
R1 0.542
F4 0.712
T1 0.539
T2 0.396 0.690
F6 0.539
T3 0.648
S3 Favorable for 0.613
P4 0.632 Environment
R1 0.542 0.340 P1 Service 0.610
T1 0.318 0.539 0.311 Delivery
F3 0.456 −0.415 by Private
F6 0.358 0.539 R4 Consortium 0.578
S3 −0.320 0.613 F1 0.561
P1 0.610 P2 0.474
R4 −0.405 0.578 F2 0.448
F1 0.313 0.561 R2 Government 0.702
P2 0.474 −0.438 Control
F2 0.448 over
R2 0.702 R3 Risk 0.616
R3 −0.508 0.616 Allocation
T4 0.549 T4 0.549
S2 −0.395 R6 0.451

Note: F1-F4 are factors of component matrix; and P3-S2 Note: F1-F4 are factors of component matrix; and P3-S2
are corresponding components for factor loadings. are corresponding components for factor loadings.

486

TIS2017_Book.indb 486 2/8/2017 8:06:32 PM

 4 DISCUSSION 4.2 Significance index and agreement analysis
A significance index is deployed to analyze the
We intend to highlight the key critical success fac-
CSFs’ ranking across various attributes. Responses
tors that can enhance the performance of these
from different groups, viz, government sector, pri-
four attributes. We deploy a few additional tools,
vate player, consultants are separately indexed and
such as rank agreement factor, maximum rank
their results are compared to each other to identify
agreement factor, percentage disagreement and
CSFs from each respondent group. Agreement anal-
percentage agreement factors, to examine the rela-
ysis is applied to check the rank agreement among
tive importance of these critical success factors and
various respondents. The results of this analysis are
attributes. This study brings out the relative impor-
presented in Tables 5–6. A strong Percentage of
tance index and significance index to address these
Agreement (PA) has been observed between govern-
CSFs in the Indian economy.
ment and private players over CSFs. The analysis
identified five key success factors amongst a total of
4.1 Relative importance index 26 CSFs. The basic characters of a PPP infrastruc-
ture project, such as large sunk investment, high
We have used the Relative Importance Index (RII) to sensitivity to demand variations, great exposure to
get the ranks of these four attributes in terms of their financial markets and vulnerability to political insta-
criticality as perceived by respondents. This is based bility are to be kept in mind before analyzing the key
on the findings of Iyer and Jha (2005) and Verma success factors as well as proposing relevant mitiga-
et al. (2016). The study declares five most important tion measures for it. Indian PPP scenario needs to
factors for successful implementation of PPP infra- be redefined with a more mature and strong insti-
structure projects in India. These include need for tutional environment with a willingness and com-
community participation, undue political interruption mitment to implement the partnership. Each CSF is
during project implementation, experience of govern- further analyzed and recommendations are made to
ment in packaging PPP projects, financial viability of formulate an institutional environment which shall
project, and flexibility of contract for change in output allow for better implementation of PPP infrastruc-
specification and renegotiations (see Table 4). ture projects in the Indian economy.

Table 4. Relative importance index.

Component Success attributes RII Ranking

S2 Need for community participation 0.863 1

P2 Political interruption & pressure 0.833 2
T1 Government sector experience 0.757 3
F6 Financial viability 0.749 4
T3 Contract flexibility 0.742 5
S4 Acceptable toll 0.737 6
P1 Stability in political system 0.709 7
P4 Favorable legal framework 0.678 8
F2 Adequate local financial market 0.676 9
F3 Promising economy 0.653 10
F4 Availability of long-term cash flow 0.643 11
S1 Community response during implementation 0.643 11
R5 O & M agreement risk allocation 0.643 11
S5 Job opportunities for local community 0.641 12
P3 Environment sensitivity of project 0.618 13
T2 Private consortium capability 0.613 14
P5 Statutory and institutional arrangements 0.605 15
R4 Supply agreement 0.597 16
R1 Concession agreement risk allocation 0.582 17
F1 Long-term demand 0.575 18
T4 clarity in service quality delivery 0.562 19
F5 Competition from other PPP projects 0.542 20
R2 Shareholder agreement risk allocation 0.542 20
R3 Loan agreement/insurance agreement 0.494 21
R6 Guarantees/comfort letters/support 0.484 22
S3 Reliable service delivery 0.473 23

Note: S2-S3 are components of critical success factors; and RII is relative importance index.

487

TIS2017_Book.indb 487 2/8/2017 8:06:32 PM

 Table 5. Ranking and agreement between government and private sector.

Critical success factor Government index Sector rank Private index Sector rank Agreement analysis

T3 88.18 1 88.57 1 RAF = 06

S2 84.55 2 79.05 3 RAF
Max = 3.4
F6 83.64 3 76.19 4 PA = 82.3%
T1 80.91 5 73.33 6
P2 80.00 6 73.33 6

Note: F1-F4 are factors of component matrix; and P3-S2 are corresponding components for factor loadings.

Table 6. CSF rankings and agreement between consultants to government and private sector.

Critical success factor Government index Sector rank Private index Sector rank Agreement analysis

S2 88.18 1 86.67 1 RAF = 2

P2 87.27 2 86.67 1 RAF
Max = 3
T1 72.73 5 74.29 2 PA = 33.33%
F6 69.09 7 70.48 3

Note: F1-F4 are factors of component matrix; and P3-S2 are corresponding components for factor loadings.

4.3 Important factors for successful PPP projects which is followed in the Latin American countries
implementation under Economic & Financial Re-equilibrium
model, is that a set of priority rules are established
In this section, we highlight the importance of the
in the contract, which provide a guideline for the
previously mentioned five attributes in order to
negotiation process. Renegotiations, an inherent
justify their implementation in PPP infrastructure
feature of PPP contracts, can be facilitated as per
projects.
the changing project environment. Instead of pre-
paring a rigid guideline for project delivery, a set of
4.3.1 Flexibility in contracts strong conditions for monitoring of project should
Flexibility in contractual arrangement calls for be fixed and initial design should be allowed to
trade-off between capturing efficiency gains and adapt as it fits the actual scenario. There should be
contractual rigidity. It is generally infeasible to ample space for modification in scope, investment
prepare a contractual document between parties plan, price or condition of service provided to
to the PPP transaction that could comprehensively the concessionaire by government and in price of
mitigate all project risks—present and future— services charged by concessionaire. Furthermore,
and thereby be used to govern a PPP project until agreements can have a clause to modify financing
completion. It is rather suggested that PPP con- structure and specific conditions for refinancing
tracts attempt to be more flexible and dynamic as and/or conversion of debt into equity. Flexibility
opposed to being comprehensive and static (Orr, in technology selection to private party without
2005). A contract can go under renegotiations in compromising on quality of service delivery should
two ways, viz., discretionary and contractual. The also be given. Lastly, the consortium should note
former one is followed in Indian industry where that provision of flexibility should be at all levels
there is no prior established rule or provision for of project implementation, i.e., strategic, tactical
negotiation in the contract and both the parties and operational.
try to establish a common understanding on new
terms. This method is vulnerable to transparency 4.3.2 Need for community participation
issues like lobbying. For any successful PPP project, it is recommended
To tackle the drawbacks, the consortium facili- that the public authority shall consult with all
tates availability of objectives, restrictions and key stakeholders at the initial stages of project
outcomes of the renegotiation process, to the pub- preparation and project structuring. These stake-
lic. It can also conduct public hearings before new holders include public or project users, commu-
terms are signed. Second method of renegotiations, nity groups and associated Non-Governmental

488

TIS2017_Book.indb 488 2/8/2017 8:06:32 PM

 Organization (NGOs), private operators, financial structuring of project needs to be done. Addition-
institutions, political representatives, and other ally, optimal tariffs and concession periods can be
government organizations (ICRA, 2011). designed to replace high tolls over a long period by
Absence of a buy-in from the people at large has ensuring better monitoring projects and holding
led to significant hurdles at various stages of a PPP contractors accountable. Cost recovery mechanism
project in recent years. For instance, disturbance should also be strengthened instead of subsidizing
caused by displaced people during land acquisi- the services for political benefits. Strong institu-
tion; or resistance to collection of revenues in the tions can be set-up which focus on lending to dif-
form of toll, charges or tariffs during operational ferent phases of a project based on the risk profile
stage. So, public support is critical to any project. to help projects continually restructure themselves
This helps in sustaining PPPs through innovative such that the risk of their debt matches with the
organizational arrangements and incentives. The risk profile of the project at any stage (Mahalin-
local community participation can be encouraged gam and Kapur, 2009).
by involving users directly or indirectly into legal
and organizational arrangements for supervis- 4.3.4 Government experience in packaging
ing PPP infrastructure projects. During the initial PPP projects
stage, proper communication of project benefits Good governance is the key requirement to provide
to various stakeholders helps in mobilizing public transparency, fair treatment and open competition.
support. Structured ICE (information, commu- Absence of governance or weak authority makes
nication, and education) and public consultation the potential investors and lenders apprehensive
activities also help to bring out the concerns, appre- about the increasing cost of money. Thus, many
hensions and acceptance of various stakeholders good players are lost, which reduces competitive
on the project. The public or private party can pressure on bidders, resulting in increased costs
involve local NGOs for this purpose. The NGOs and reduced quality. Cost and time over-runs are
can use existing contacts with the municipality and also caused due to negative attitude of rent seek-
other influential bodies to ensure maximum sup- ing and other forms of corruption (Delmon, 2011).
port of the people for unhindered execution of Good governance requires putting into place the
project (ICRA, 2011; Verma et al., 2016). enabling institutions, procedures and processes
surrounding PPPs in order to fully benefit from the
4.3.3 Financial viability arrangement. Authorities should take initiatives
Infrastructure projects are usually financed by a for better structuring of PPP projects. Technical
combination of various forms, like private equity, and regulatory capacity should be added to public
debt, governmental funds and user charges. In institutions and officials, to manage the PPP proc-
Indian market, there are usually no user charges ess so as to maximize returns for all shareholders.
and the services are subsidized for political reasons. There is an absence of institutionalized mechanism
Hence, the PPP gets more exposed to the volatility to do traffic volume forecasting in India, not only
of capital market. The financial meltdown and scar- for 25 years, but also for 5–10 years. To meet this,
city of credit availability may impose significant there can be a framework within government body
financial risk. To overcome it, innovative financing to carry out forecast studies, so that cost-recover-
schemes can be applied which will provide support ing risks can be mitigated at pre-bid stage. This will
for project development and structuring arrange- attract parties to submit higher bids for feasible
ments that stand the test of time. This shall render projects. Dedicated and cross-sectoral professional
the project less dependent on volatile markets by units need to be installed to support project imple-
cutting government financial support in the form mentation (Nataraj, 2007). Streamlining of approv-
of Viability Gap Fund (VGF) or Project Develop- als and clearances is the most critical requirement
ment Fund (PDF). Such innovations in financing to expedite project implementation. Reforms in the
can be done by employing cost effective technical functioning of land, labor, and financial markets,
measures to generate funds and effective resource as well as the removal of restrictive regulation are
utilization to avoid time and cost over-run. required which will enable private sector to become
Concessionaire can utilize the annuity pay- a capacity builder (Verma, 2013).
ments received during operations stage from
public authority by securitizing it and thus gener- 4.3.5 Political interruption and pressure
ating more funds for future. Some projects like the The primary objective of PPP can only be reached
Vadodara-Halol toll road have used deep discount when government agencies are a willing partner
bonds, take-out financing, cumulative convertible with active participation to make the project viable
preference shares and long term loans to avert for private players. The main reasons of private
financial risks. In order to convince Indian users sector failure to deliver are lack of law and order
to pay for the services, an efficient marketing and support from administration side leading to delay

489

TIS2017_Book.indb 489 2/8/2017 8:06:32 PM

 and unwanted hurdles in project execution. There ACKNOWLEDGEMENTS
is a lot of political syndicate pressure to give con-
tracts to influential elements. Moreover, the prices We thank the RCG School of Infrastructure
are invariably based on political pressures or con- Design and Management and Vinod Gupta School
siderations. Absence of credible government com- of Management at Indian Institute of Technology
mitment and presence of bureaucratic red tape Kharagpur (India) to carry out this research. We
makes capital more expensive resulting in higher also grateful to the anonymous referees and con-
tariffs. To address issues of transparency and cor- ference convener for their constructive comments
ruption, the government should instill reforms and suggestions.
to structure and support PPP arrangements by
publishing mandatory disclosures and fair prac-
tices to be followed by all projects. The possibil- REFERENCES
ity of setting up a web based PPP market place
can be explored to add transparency. Dedicated Akintoye, A.; Hardcastle, C.; Beck, M.; Chinyio, E.;
grievance redressal cell/dispute resolution mecha- Asenova, D. 2003. Achieving Best Value in Private
nism should be set-up. A single window clearance Finance Initiative Project Procurement, Construction
for high priority projects should be done to safe- Management and Economics, 21(5): 461–470.
Birnie, J. 1999. Private Finance Initiative (PFI)—UK
guard private players from red-tape delays. New Construction Industry Response, Journal of Construc-
market based products can be developed such as tion Procurement, 5 (1): 5–14.
independent pre-bid rating, to assist investors in Boynton, A.C.; Zmud, R.W. 1984. An Assessment of
identifying well-structured PPP projects. Most Critical Success Factors, Sloan Management Review,
importantly, management information systems 25 (4): 17–27.
should be employed to continuously monitor the Brodie, M.J. 1995. Public/Private Joint Ventures: The
performance of the PPP projects over the project Government as Partner—Bane or Benefit? Real Estate
life cycle. MIS can help in evaluation of PPP Issues, 20 (2): 33–39.
projects by tabulating and summarizing various Chan, A.; Lam, P.; Chan, D.; Cheung, E.; Ke, Y. 2010.
Critical Success Factors for PPPs in Infrastruc-
stage experiences as a database for future (see, for ture Developments: Chinese Perspective, Journal of
more details, Verma et al., 2016). Construction, Engineering & Management, 136 (5):
484–494.
Delmon, J. 2011. Public-Private Partnership Projects
5 CONCLUSION in Infrastructure, Cambridge University Press,
Cambridge.
This paper provides an overview of critical suc- Doloi, H. 2009. Analysis of Pre-qualification Criteria
cess factors of PPP infrastructure projects in India. in Contractor Selection and their Impacts on Project
On the basis of principal component analysis, we Success, Construction Management and Economics,
27 (12): 1245–1263.
recognize four factors, namely stable project envi- European Investment Bank. 2000. The European Invest-
ronment-political and social, shared role of public ment Bank and Public Private Partnerships. The
and private party in project execution, favorable Newsletter of the International Project Finance Asso-
environment for service delivery by private consor- ciation, 1: 3–4.
tium, and government control over risk allocation. Grant, T. 1996. Keys to Successful Public-Private Part-
The success of these factors are further predicted nerships, Canadian Business Review, 23(3): 27–28.
by some sub-factors. These include community Hambros, S.G. 199). Public-Private Partnerships for
participation, undue political interruption during Highways: Experience, Structure, Financing, Applica-
project implementation, experience of government bility and Comparative Assessment, Canada.
ICRA 2011. Toolkit for Public Private Partnership
in packaging PPP projects, financial viability of the Frameworks in Municipal Solid Waste Management
project, and flexibility of contract for change in out- Volume I –Overview and Process. Public Private Part-
put specification and renegotiations. It can be noted nerships Knowledge Series, Government of India,
that this study has been principally focused towards New Delhi.
overall PPP infrastructure scenario in the Indian Iyer, K.C.; Jha K.N. 2005. Factors Affecting Cost Per-
economy. There can be further research to examine formance: Evidence from Indian Construction
this issue sector-wise, such as housing, transporta- Projects, International Journal of Project Manage-
tion, water supply and policy recommendations for ment, 23 (4): 283–295.
each infrastructure sector can be made. This study Li, B.; Akintoye, A.; Edwards, P.J.; Hardcastle, C. 2005.
Critical Success Factors for PPP/PFI Projects in the
does not provide any meaningful outcome in terms UK Construction Industry, Construction Manage-
of understanding the clustering effects of the simi- ment and Economics, 23 (5): 459–471.
lar attributes and their predictive capacity. This can Mahalingam, A.; Kapur, V. 2009. Institutional Capacity
be possible by using some advanced statistical meth- and Governance for PPP projects in India. Lead 2009
ods, which is beyond the scope of this research. Conference, CA, USA.

490

TIS2017_Book.indb 490 2/8/2017 8:06:32 PM

 Mohr, J.; Spekman, R. 1994. Characteristics of Partner- Scandizzo, P.L.; Ventura, M. 2010. Sharing Risk through
ship Success: Partnership Attributes, Communication Concession Contracts, European Journal of Opera-
Behaviour, and Conflict Resolution Techniques, Stra- tional Research, 207 (1): 363–370.
tegic Management Journal, 15 (2): 135–152. Tiong, R.L.K. 1996. CSFs in Competitive Tendering
Nataraj, G. 2007. Infrastructure Challenges in South and Negotiation Model for BOT Projects, Journal of
Asia: The Role of Public-Private Partnerships. Dis- Construction, Engineering and Management, 122 (3):
cussion Paper, No. 80. Asian Development Bank 205–211.
Institute, Tokyo. Verma, A. 2013. Critical Success Factors in Indian Sce-
Ng, S.T.; Wong, Y.M.W.; Wong, J.M.W. 2012. Factors nario for PPP Projects in Infrastructure Development,
Influencing the Success of PPP at Feasibility Stage- A M. Tech. Thesis, Indian Institute of Technology
Tripartite Comparison Study in Hong Kong, Habitat Kharagpur, India.
International, 36 (4): 423–432. Verma, A.; Pradhan, R.P.; Bele, S.K.; Gaurav, K. 2016.
Orr, R. 2005. Proceedings of the First General Counsels Critical Success Factors for PPP Projects in Infra-
Roundtable, Collaboratory for Research on Global structure Development: The Indian Scenario, Indian
Projects, Working Paper Series, Stanford, CA, USA. Journal of Regional Science, (forthcoming).
Qiao, L.; Wang, S.Q.; Tiong, R.L.K.; Chan, T.S. 2001. Yeo, K.T. 1991. Forging New Project Value Chains- a
Framework for Critical Success Factors of BOT Paradigm Shift, Journal of Management in Engineer-
Projects in China, Journal of Project Finance, 7 (1): ing, 7 (2): 203–211.
53–61. Zhang, X. 2005. Critical Success Factors for Public-
Sanvido, V.; Grobler, F.; Parfitt, K.; Guveris, M.; Coyle, M. Private Partnerships in Infrastructure Development,
1992. Critical Success Factors for Construction Journal of Construction, Engineering and Manage-
Projects, Journal of Construction, Engineering and ment, 131 (1): 3–14.
Management, 118 (1): 94–111.

491

TIS2017_Book.indb 491 2/8/2017 8:06:32 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

An on-ramp metering and speed control application on O1 freeway

in Istanbul

H.G. Demir & Y.K. Demir

Faculty of Engineering, University of Niğde, Niğde, Turkey

ABSTRACT: In this study, to make improvements in freeway traffic flow, speed limits and ramp meter-
ing have been applied on on-ramps of the European side of 2.5 km Istanbul O1 Freeway during peak
hours. Speed limits from 10 to 70 km per hour in multiplies of 10 km per hour and local ramp metering
control ALINEA were applied on on-ramps of the study area. Simulation software CORSIM was used to
test control results. The comparison of the results with the current condition shows that 20 km/h constant
speed limit that was applied to on ramps of Beşiktaş reduce the average travel time per vehicle and the
effect of shock waves on the study area. It shows as an alternative for local ramp-metering. The total travel
time of vehicles on mainstream results %5 decrease.

1 INTRODUCTION

Traffic management measures can increase the effi-

ciency of road transport infrastructure. These appli-
cations are made to regulate traffic and improve
traffic flow. Freeway management application are
also made for these reason. With the development of
traffic technology and control theory freeway man-
agement are also developed alongside. Ramp meter-
ing and variable speed limits are proven technique to
improve traffic flow on freeways. Ramp metering is
a equipment which use a traffic signal placed at the
end of a ramp to allows traffic to enter the freeway
under control. Variable Speed Limit Signs (VSL)
systems consist of Variable Message Signs (VMS)
equipped along a freeway to display the current Figure 1. Working Zone of O-1 freeway (Demir, 2012).
speed limits advisory or regulatory under prevail-
ing traffic condition, weather conditions, construc-
tion or maintenance activities and other factors. 2 CORSIM
In this study, speed limits and ramp metering were
applied on on-ramps. The aim of this application For testing the estimated control effects on the
was to enter the vehicles from on ramp to main road zone, CORSIM (McTrans, 2012) microscopic sim-
at a lower rate. In this way, it is expected to reduce ulation software was used. CORSIM RTE has the
congestion at the entrance on the mainstream. capability to intervene in the system with the VSL
With the reduction of congestion, it is expected to control device. As part of the way for the creation
prevent the formation of shock waves and queues. of CORSIM simulation model, the road was coded
With the application of ramp metering, the effects into CORSIM network format by aid of the maps
of speed limit application were compared. Controls (Google, 2007). Traffic volume data are obtained
were applied to the on ramps of 2.5 km segment of by video recording for calibration of CORSIM
O1 Freeway, which is located on the European side model. The record was taken from 16:23 to 17:27
of Istanbul, near the Bosphorus Bridge (Figure 1). for evening peak period in 16 October 2007. Then
The study segment has two on ramps Beşiktaş and for examining the records, observation stations
Balmumcu. Congestion occurs during evening rush (OS1 and OS2 in Fig. 2) were set using the maps
hours in direction from Europe to Anatolian side. and on site measurement by GPS devices. Speed
Geometrical irregularities are available such as a and volume data were measured in one- minute
horizontal curve and a bottleneck. interval. Total 65 minutes of surveys were used for

493

TIS2017_Book.indb 493 2/8/2017 8:06:35 PM

 et al. 1991). ALINEA has had multiple success-
ful field applications (Paris, Amsterdam, Glas-
gow, and Munich) (Bogenberger and May 1999).
ALINEA algorithm use the system output as input
for the next iteration. Ramp flow is regulated to
keep the downstream measured occupancy (Oout)
below critical occupancy (Ocr). Figure 4 illustrates
the sketch of an on ramp. Occupancy “O” is the
portion of time that a vehicle is detected at a cer-
tain location. It has a unique value for traffic con-
Figure 2. Observation stations (OS1 and OS2) in the dition, as seen in Figure 5. That has an advantages
working zone.
for the ALINEA metering algorithm. Papageor-
giou (Papageorgiou et al. 1991) calculated ramp
metering rate, r(k) (veh/h) in discrete time interval
k = 1, 2, ... as shown Equation 1 below:

r(k) = r(k – 1) + KR [O – Oout (k–1)] (1)

The metering rate r(k–1) at previous time interval

k–1, KR>0 is a regulator parameter. Acceptable
Figure 3. Schematic Figure of of the study site includ- result could be found if this value was selected
ing two on on-ramps and detectors (Demir, 2012).
70 veh/hour. This value is not sensitive (Papageor-
giou et al. 1991). O is desired occupancy which is
calibration (Demir, 2012). Calibration was con- selected slightly lower than the critical occupancy
firmed by Theil’s test comparing observed speed Ocr (occupancy corresponding to the capacity
and flow with CORSIM outputs at two observa- flow). Oout: measured occupancy measured down-
tion stations (Demir, 2012). stream on the on ramp during time interval k, r(k)
metering rate is updated between 40 s and 5 min
(Chu and Yang 2003). Metering rate is calculated
3 APPLICATION OF SPEED LIMITS ON
ON-RAMPS

Because of disturbing effects of the on ramps on

main flow, variable speed limits were applied from
10 km to 70 km per hour in multiplies of 10 km per
hour and in matrix form on amp of Beşiktaş and
Balmumcu (Figure 2). The study site of freeway
segment was divided into six sections (Figure 3)
so that the effects on the on-ramp traffic could be
analyzed more efficiently. After simulation of con-
trol application it could be seen that for 20 km per
hour speed limit on Beşiktaş on ramp minimum Figure 4. Sketch of an on-ramp.
queue was built (R02 represents 20 km per hour
speed limit applied on Beşiktaş on ramp). For the
control application control tools in CORSIM soft-
ware was accessed with the intervention of RTE
that was written with Visual C.

4 RAMP METERING CONTROL

APPLICATION FOR ON RAMPS
OF BESIKTAS USING ALINEA
ALGORITHM

ALINEA (Asservissement Lineaire d’entree Au-

toroutiere) was the first local ramp metering control
strategy to be based on straightforward application
of classical feedback control theory (Papageorgiou Figure 5. Fundamental diagram (May, 1990).

494

TIS2017_Book.indb 494 2/8/2017 8:06:36 PM

 by measuring the difference between the measured Figure 6, it can be seen that an upstream proro-
occupancy, and the desired occupancy. gating shockwave starting on Link-6 at 35 minutes.
The metering rate r(k) is between the range [rmin, Table 1 summarize speed variation before and
rmax]. Where rmax is the flow capacity of on ramps after shock wave on links of study site. Average
(for single lane uncontrolled on ramp equal between speed and standard deviation of speed are the two
1800 and 2200 veh/hour). rmin>0 is minimum admis- parameters pointed in Table 1. After the occupa-
sible ramp flow (200 400 veh/hour) (Papageorgiou tion of shockwave, standard deviations of speed
et al. 2007). ALINEA control was applied on on Link-1, Link-2 and Link-3 increased. Contras-
Beşiktaş on ramp (A01 represent ALINEA control tively that, standard deviations of speed on Link-5
application). KR value was selected 70 veh/hour. and Link-6 decreased. But in total, standard devia-
Desired occupancy value was selected %20. Meas- tion of speed increased in the study zone.
ured occupancy value was measured on the loop Table 2 lists average travelling time for each
detector by CORSIM which was placed 40 meter links before and after shockwave occurred. Total
far. average travelling time for study zone increased
from 102.41 seconds to 365.55 seconds after for-
mation of shock- wave. The highest increments in
5 THE EFFECTS OF CONTROL travelling time is observed on Link-3 and Link4.
APPLICATIONS ON WORKING ZONE This increments are result of a bottleneck effect.
Because of high demand of vehicles joined from
For determine the effects of control application on Beşiktaş ramp, density increased in Link-5.
freeway segment, graph and charts have been gen- The shockwave formation can be seen in Fig-
erated. To evaluate of performance criteria, mean ure 7. The Figure points out that densities fluc-
speed, mean travel time, queue and supply of work- tuations along simulation on Link-4 and Link-5.
ing area had been considered. The necessary data Especially after 50th minutes density spreads
for the generation of chart and graphs were taken upstream. Figure 8 represents vehicles discharge
from CORSIM. With using of a function inside the against the total demand. After 30th minutes, the
CORSIM RTE interface the detector and link data discharging curve diverges from demand curve,
for the study site were recorded in a database every although discharge flow is 7560, demand is 8362
ten seconds. At the end of simulation which was at 120th minutes of simulation. End of the simu-
120 min data was converted into chart and graphs. lation, study zone could discharge 14,076 vehicles
The graphs and charts which were used described
below:- Speed Graphs: This graphs shows the
speed variation on the links according to time.
Shock waves can be easy observed. Speed is repre-
sented in gray color on the graph. The dark color
defined low speed whereas the light color defined
high speed level.- Density Graphs: This graphs
shows the variation of density according to time.
The density value are derived from link values. On
this graphs lighter colors represent lower density
whereas darker color defined higher density value. Figure 6. R00 speed graph.
Demand-Supply Graphs: On the horizontal axis,
the time is shown, and on the vertical axis, the vehi- Table 1. R00 mean speed before and after the shock
wave.
cles (included on ramps) demand on working zone
against vehicles outflow in link 6 are given.
- Average Travel Time Table: This represent the
average travel time per vehicle before and after
shock wave occurred. The average travel time is the
sum of the vehicles travel time exit on each link per
10 sec interval divided by the number of total vehi-
cles per 10 sec interval exit on each link. The last Table 2. R00 mean travel time before and after the
columns on table represent variation of the aver- shock wave.
age travel time of the vehicles according to current
situation (%). Queue Graphs: This graph shows
the vehicles waiting on the virtual link that could
not entry to the working zone.
R00 symbolizes the current state of the study
zone. When current situation is examined in

495

TIS2017_Book.indb 495 2/8/2017 8:06:37 PM

 totally. The inadequate discharging causes queu- 5.1 Application 1 (R02)
ing on entering sections as seen on Figure 9. End
With R02 control strategy 20 km per hour speed
of the simulation, number of vehicles waiting for
limit was applied on Beşiktaş ramp. R02 raised aver-
entering working zone, at main road (O1), Bal-
age speed of whole simulation up to 3.18% relative
mumcu on ramp and Beşiktaş on ramp were 400,
to uncontrolled state (Table 3). Also standard devia-
500 and 150 vehicles respectively. The inadequate
tion of speed and average travelling time (Table 4)
discharging causes queuing on entering sections as
decreased 1.19% and 5.06% respectively relative to
seen on Figure 9. End of the simulation, number of
uncontrolled state. This control created a speed pat-
vehicles waiting for entering working zone, at main
tern as shown Figure 10. It explains that, shockwave
road (O1). Balmumcu on ramp and Beşiktaş on
propagation could not be prevented but postponed
ramp were 400, 500 and 150 vehicles respectively.
to 40th minutes of the simulation. Figure 11 shows
that the control succeeded reducing and postpon-
ing queuing at O1 main road and Balmumcu on
ramp. At the end of the simulation, 1000 vehicles
were observed at on ramps and O1 entrance. The
number of vehicles discharged from working zone
along whole simulation is 14,070 vehicles.

Table 3. R02 mean speed before and after the shock

wave.
Figure 7. R00 density graph.

Table 4. R02 mean travel time before and after the

shock wave.

Figure 8. R00 demand-supply graph.

Figure 10. R02 speed graph.

Figure 9. R00 queue graph. Figure 11. R02 density graph.

496

TIS2017_Book.indb 496 2/8/2017 8:06:38 PM

 5.2 Application 2 (A01) Change on standard deviation of the mean speed
during total simulation time. According to current
Figure 11 and Figure 13 clearly show that ALINEA
condition R02 application decreased the travel
control strategy prevents congestion on main road
time, shifted shock wave, decreased queues on O1
which free flow condition was prevailed. Hence
main road and Balmumcu on ramp, but increased
average speed increased while its standard devia-
queue on Beşiktaş on ramp. A01 ALINEA appli-
tion decreased over whole simulation time. With
cation on Beşiktaş on ramp increased the mean
improvement of speed, average travelling time
speed on main road during simulation time and
drops from 288 second to 96 seconds (Table 6). The
increased the travel time after shock waves. It has
adverse effect of this control was long queues that
been seen at the 50th min. queued up 1500 vehicles,
was 3950 vehicle observed on Beşiktaş on-ramp.
at the end of simulation 3950 vehicles on Beşiktaş
From the result of AO1 application it can be said
on ramp. R02 application was effective in reduc-
that Beşiktaş on ramp cause traffic congestion on
ing the travel time and shifting the shock wave and
main road. Ramp metering control on AO1 did
could be considered as an alternative.
not cause effective solution for traffic jam. The
number of vehicles discharged from working zone
along whole simulation is 11671 vehicles which is
6 EVALUATION AND RESULTS
the lowest value of the application.
In this study 2.5 km of O1 Freeway, disturbing
effects of the ramps on main flow speed limits
and ALINEA algorithm were applied at on ramp
to improve freeway traffic flow at peak hour. For
testing the estimated effects on the zone, COR-
SIM microscopic simulation software was used.
The looking at the current state of the traffic flow
within the working zone, at the 35th minutes of
the simulation time, shock wave was emerged
propagating from link 5 to upstream. After the
Figure 12. A01 speed-time graph. formation of the shock wave, the average speed
reduce from 94 km/h to 37 km/h, mean travel time
per vehicle increase from 102 seconds to 365 sec-
onds. From the results of simulation, it can be
seen that the traffic problem arise from the bot-
tleneck due to high demand of Beşiktaş on ramp.
The VSL impacts were criticized under different
indicators such as travel times per vehicle, shock
wave effects, mean travel speed, standard devia-
tion of mean speed. R02 has the same properties
with the current applications before shockwave
Figure 13. A01 density-time graph. occupation. After formation of the shock wave it
can be seen that the mean speed increase accord-
Table 5. A01 mean travel time before and after the ing to current condition. There was no with
shock wave. 20 km per hour. It should be emphasized that
vehicle speed up on ramps as joining on the main-
road. However, this type of control application
needs testing before. Travelling speed with 20 km
per hour is thought to be possible with intelligent
vehicle and transportation technology. The provi-
sion of speed is the subject of a different investi-
gation. However, it is thought that the spread of
Table 6. A01 mean travel time before and after the electric vehicles in the future, more precise speed
shock wave. limit could be implemented. The primary results
show that Application 1 is capable of reducing
the total travel time per vehicle as %5 percent
against increasing queue length on Beşiktaş. But
total queue is 50 vehicles less than uncontrolled
case. In addition it shifted the beginning of the
shock wave without making changes on aver-

497

TIS2017_Book.indb 497 2/8/2017 8:06:39 PM

 age speed. It can be seen that there is a trade-off Chu, L. & X. Yang (2003). Optimization of the alinea
between main-stream and ramps queues. For this ramp-metering control using genetic algorithm with
reason the control problem should be optimized. micro- simulation. Transportation Research Board
Dynamic speed limits application on ramps can 82nd Annual Meeting, Washington, DC.
Demir, H.G. (2012). Flow Control in Urban Freeways by
be suggested as further work because of changing Applying Variable Speed Limits in A Model-predictive
condition of the freeway traffic flow. Control Framework. Ph. D. thesis, Istanbul Technical
University, Istanbul, Turkey.
Google (2007).Satellite data. Web Page. http://www.
ACKNOWLEDGMENT googleearth.com.
Papageorgiou, M., H. Hadj-Salem, & J. Blosseville (1991).
This study was generated from the thesis of Hatice Alinea: A local feedback conrol law for on-ramp
Göçmen Demir which completed on the structure metering. In Freeway operations, highway capacity,
of thesis by Istanbul Technical University. Authors and traffic flow, pp. 58–64. National Research Coun-
cil, Washington, D.C.
wish to express a sincere thanks to all person and Papageorgiou, M., H. Hadj-Salem, & F. Middelham
companies for their contributions. (1991). Alinea local ramp metering- summary of field
results. In Freeway operations, highway capacity, and
traffic flow, pp. 58–64. National Research Council,
REFERENCES Washington, D.C.: TRB.
Papageorgiou, M., E. Kosmatopoulos, L. Papamichail,
Bogenberger, K. & A. May (1999). Advanced coordi- & W. Yibing (2007). Alinea maximises motorway
nated traffic responsive ramp metering strategies. throughput:an answer to flawed criticism. Traffic
Working Paper. Engineering and Control 48, 271–276.

498

TIS2017_Book.indb 498 2/8/2017 8:06:40 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Transport infrastructure capacity calculations using green

travelling planner

I. Celiński, M. Staniek & G. Sierpiński

Faculty of Transport, Silesian University of Technology, Katowice, Poland

ABSTRACT: The article provides a discussion on the chosen approaches to the calculation of the
traffic capacity of road network cross-sections and movements at intersections in line with the idea of
sustainable development of transport. The methods of calculating traffic capacity used until now have
emphasised exclusively the physical aspect of the respective measure. The said measure is related to the
number of vehicles covering a specific road cross-section over a unit of time. Under such an approach,
the characteristics being significant for individual means of transport are mainly their physical properties,
rather than their potential in terms of allowing transport to develop in a sustainable manner. The authors
argue that a certain specific parity should be taken into account in calculations of road network cross-
section capacity, ensuring that such cross-sections are used efficiently from the point of view of transport
sustainability. The said parity regulates the manner in which individual means of transport access the
network. The article presents a methodology for calculation of the parity based on the features of a tool
known as Green Travelling Planner (trip planner). The discussion is supplemented with findings of origi-
nal research into multimodal cross-sections in the Upper Silesian urban area in Poland. There is a perfect
correspondence between the methodology discussed in this article and the extensive framework of actions
undertaken by the EU with regard to transport (European Commission. 2011, Janic M. 2016).

1 INTRODUCTION road (transport) networks. It needs to be pointed

out that the capacity of road network cross-
Within the road network, the theoretical traffic sections and of movements at intersections is a
capacity of a road cross-section or movement at an certain time- and space-limited resource. As in the
intersection depends on a large number of variable case of any resource that is limited, certain con-
parameters. The characteristics taken into account ditions need to be determined for its utilisation.
in the evaluation of the traffic carrying capacity of Road traffic is a self-organising process, so the
infrastructure elements are the following: average related resource is utilised, as it were, naturally on
vehicle speed, level of service, safe gap between vehi- a random basis. This randomness can be assumed
cles, other distances between vehicles, directional to represent an unfavourable phenomenon. Traf-
distribution of traffic, share of heavy goods vehi- fic management can determine the conditions of
cles, type of intersection, geometrical parameters of utilising capacity only to a limited extent. Combin-
the road, and many others. Essentially, the capacity ing traffic organisation and control in an accurate
calculation procedures are related to a single, glo- manner may help regulate to the full extent the way
bal measure expressed by the number of vehicles in which the resource is utilised.
travelling through a specific road cross-section The authors argue that a specified parity should
over a unit of time. Traffic capacity is measured in be taken into account in the calculations of capac-
vehicles per time unit. The measurement is usually ity of transport network cross-sections. This parity
carried out at intervals of 15, 20 or 30 minutes, or should be understood as a reasonable division of
of one hour. If traffic is highly heterogeneous, the the capacity resource into appropriate parts, sec-
measurement interval is reduced to an order of tens tioned off and allocated on an exclusive basis to
of seconds. In this approach, the capacity calcula- different means of transport, in such a way as to
tion procedures are of a technical nature, taking assure their efficient utilisation in terms of sus-
into account exclusively the physical aspect of the tainable development of transport. The process of
traffic processes taking place within the transport sectioning off and allocating the resource may take
network. Above all, however, they fail to take into place in the same way as in ICT networks.
account sustainable development, which is a criti- In this article, the said parity will be described as
cal aspect when it comes to the correct shaping of a rationally determined proportion of the amount

499

TIS2017_Book.indb 499 2/8/2017 8:06:42 PM

 of traffic in private transport to the amount of traffic) determined for them. In that case, the
traffic in public transport (PrT: PuT, Private Trans- unused resource can be resold to private vehicle or
port to Public Transport). The parity components heavy goods vehicle users in quantities that do not
are referred to respectively as PPrT and PPuT. The put transport sustainability at risk.
parity component related to public transport vehi- The methodology of calculating capacity in road
cles indicates, in the given network cross-section, network cross-sections accounting for the parity
how important the relevant means of transport discussed in this article is relevant mainly for trans-
is from the perspective of transport sustainability port networks characterised by strong diversity of
(PPuT). At the same time, it indicates the value of means of transport and roads, and ones in which
the parity component for private transport. This roads have minimum two lanes in each direction
approach has already been implemented to a cer- within the network area. Moreover, it is valid for
tain extent in transport networks, for instance, networks with a large number of cross-sections,
by the provision of bus lanes for public transport where more than one means of transport oper-
vehicles. The problem in this respect consists in the ates. Taking the foregoing into account, the article
determination of reasonable and clear criteria for defines the multimodal cross-section and describes
the division of traffic carrying capacity as a highly research into road assets of this kind, performed in
limited resource. This would make it possible to set the Upper Silesian conurbation in Poland.
forth precise conditions for the provision of bus As described in this article, the parity component
lanes and other similar solutions. Furthermore, for Public Transport (PPuT) was calculated using
one needs to bear in mind that capacity is a highly a specialised travel planner known as Green Trav-
heterogeneous resource across the entire transport elling Planner (GT Planner) (Esztergár-Kiss D. &
network. The drivers of change in traffic stream Csiszár Cs. 2015, Sierpiński G. et al. 2014, Sierpiński
distribution include mainly suburbanisation proc- G. et al. 2016, Sierpiński G. 2017). The PuT value
esses (Aguiléra, A. et al. 2009, Handy, S. et al. 2005, can be determined by way of basic calculations
Khattak, A. J. & Rodriguez, D. 2005, Limanond T. using data acquired from GT Planner. The data
et al. 2011, Lu, X. and Pas, E. I. 1999, Millward, H. stored in GT Planner describe actual and planned
& Spinney, J. 2011, Sáez A. E. & Baygents J. C. transport behaviours of the inhabitants of the area
2014). Consequently, the criteria for the division covered by GT Planner’s operation. GT Planner is
of capacity must be universal for the entire area it usually built for the entire transport network area
concerns, and apply 24 hours a day. of a single urban agglomeration.
Following the ICT network analogy, a reason- The approach suggested in the article also
able capacity division might lead to a greater diver- requires a modification of the measures used in
sity of rates charged for access to the road network, the calculations of capacity, from object-related
also for different public transport operators. What ones (vehicles) to subject-related ones (number of
this means is that the quality of transfer within the travelling persons).
transport network for various means of transport
is proportional to the rights to use it which one has
acquired. If this approach were adopted, private 2 MULTIMODAL CROSS-SECTION
users would be able to acquire rights to use bus
lanes at prices making it possible to bring the idea The share of vehicles of a specific type in a traf-
of transport sustainability into life. fic stream determines the traffic structure, which
The fundamental issue addressed in this article is in turn influences capacity (measured as before, in
related to the method of calculating the parity com- vehicles). In classical methods of capacity calcula-
ponents (PPuT, PPrT = 1-PPuT) for any road net- tion, this influence results, however, only from the
work cross-section or movement at an intersection. physical properties of infrastructure elements and
The capacity measure should take into account the vehicles (lane width, vehicle dimension, vehicle
value of the parity for the given means of transport speed etc.). This approach to capacity calculation
in the given road network cross-section, if calcula- usually fails to take into account the differences in
tion of the components becomes technically fea- vehicle occupancy rates within the specific cross-
sible. Consequently, the parity and capacity thus section. A bus carrying 50 passengers on average
calculated should impose the methods of traffic is the equivalent of only two passenger cars, more
management and control within the specific road or less, from the point of view of classical capac-
network cross-section. Any restrictions imposing ity measures. In reality, from the perspective of both
certain forms and methods of traffic management transport sustainability and the size of the physical
and control resulting from the parity adopted can road transport processes, the minimum difference in
be bypassed only if operators of public transport share for this resource corresponds to 8:50, while the
services are unable (for various reasons) to take maximum one to 2:50 (for different passenger vehi-
advantage of the parity (component, share in cle occupancy rates). Therefore, if the share in the

500

TIS2017_Book.indb 500 2/8/2017 8:06:42 PM

 resource in the road network cross-section is calcu- makes it possible to determine the parity compo-
lated in travels, each two private vehicles registered nents for public transport. Moreover, knowledge
in the road network cross-section cause a theoreti- of the modal split makes it possible to build better
cal loss of at least 42 travels. Estimating the loss of traffic models, streamline transport task alloca-
cross-section capacity, it is assumed that 2 passenger tion, plan lines and perform routing for means of
cars occupy approximately the road surface, which public transport, set up PuT stops in appropriate
can by use by one bus. It was also assumed equal places, parameterise the prioritisation process in
50 passengers in the bus at rush hour. Number of road traffic control systems, and sometimes also
the passengers in each of the two cars was set as 4, make rates for transport services uniform.
although based on extensive research authors in the Figure 1 expresses the idea of sustainable devel-
Silesian conurbation in 2009 it was 1.4. In this man- opment of transport in the form of a parity point
ner specified minimum loss of capacity resulting (PPuT). In this case, it consists in the balancing
from the existing (observed) modal split. Estima- of the share of private vehicles and other vehi-
trion was realized for rush hours due to the fact that cles in the traffic structure. If the traffic carrying
during this period traffic volume is close to capac- capacity measures are expressed in travels rather
ity. In realistic situation, the variance filling buses than in vehicles, the largest capacity reserves will
in rush hour on different lines modifies these losses be available with regard to the number of public
(from 42 trips), but according to the authors mostly transport vehicles that can be allowed to enter
up (>>42 trips). In cross-sections where the average the road network. Currently, this reserve is hid-
daily traffic is around 50,000 vehicles per 24 hours, den in the number of private vehicles within the
this means a loss of around 1,050,000 travels per transport network used by travelling persons capa-
day, per road network cross-section. These figures ble of changing their transport-related behaviour
show the essence of the problem related to the sus- patterns in a conscious manner. As it has already
tainable development of transport. been mentioned, the new procedures suggested
Traffic structure is usually studied within a in the article for the calculation of capacity in
precisely determined road cross-section making it the road network are reasonable in cross-sections
possible to measure the value effectively. The per- in which more than one means of transport has
centage share of individual means of transport in been identified. These are referred to as multimo-
the network cross-section is referred to as modal dal cross-sections characteristic from the point of
split. This modal split can be expressed with the view of the entire road network (Celiński, I. et al.
following equation: 2015). Due to the temporal and spatial irregulari-
ties in road traffic distribution, the authors of the
Scs = STM
T STM
STM :,..,: STM
T n (1) article suggest that such cross-sections should be
1 2
determined deliberately in the space of the road
network analysed (Figure 2 shows a heterogene-
where MS Scs = modal split in the cross-section
ous spatial distribution of those cross-sections).
TM = share of the ith means of trans-
(relation); STM
i The criterion used to determine whether a cross-
port; n – number of means of transport;
section is multimodal, which the authors adopted
arbitrarily, is that it must be possible to register at
n least four different types of means of transport in
TM = 100%
∑ STM (2) that cross-section. This criterion means that cars,
i
i =1 buses, heavy goods vehicles and trams need to be
registered in the modal split. The research con-
The value expressed in the form of equation (1) ducted by the authors in the road network of the
can be defined very accurately only for individual
traffic streams (RMS ≅ 0). Except in the case of
individual traffic streams, the modal split structure
changes to a greater or smaller extent as a result of
changes in the traffic flow management plan, dif-
ferent movements of vehicles within the road net-
work, and local interaction between vehicles.
Figure 1 illustrates the dynamics of the
changes of modal split in one of the measurement
cross-sections in the Upper Silesian conurbation
(Poland). The figure shows that the process is
highly heterogeneous. Nevertheless, even approxi-
mate determination of modal split with regard to its
distribution in time and space of the road network Figure 1. Modal split dynamics and parity point value.

501

TIS2017_Book.indb 501 2/8/2017 8:06:42 PM

 Upper Silesian urban area shows that the number ment by the individual means of transport (2). This
of points meeting the above criterion is not very is analogous to the division of bandwidth in ICT net-
high. For instance, in the main city of the urban works with unrestricted access for all users with equal
area (Katowice), the number of such cross-sections (random) access rights. The modal split shown in Fig-
ranges at around 20 measurement points. ure 3 is a certain random set of values imposed by
Figure 2 shows the spatial distribution of the network structure and its organisation, and con-
100 points where multimodal cross-sections are sequently expressed with demand channelled through
located, studied by the authors in the area of the actual use. In the authors’ opinion, a different modal
Upper Silesian conurbation in Poland. split value (obligatory parity) can be determined in
The pilot study has made it possible to under- transport network cross-sections. Obligatory parity, as
stand how important for road network organisa- opposed to conventional parity, pursues goals related
tion it is to take into account various types of modal to sustainable development of transport. Obligatory
split in capacity calculations: observed modal split parity is a set of values for individual means of trans-
(presented in the pie charts in Figure 3) and hidden port taking into account suppressed demand in the
modal split, discussed further on in the article. transport network (Szarata, A., 2010a, 2010b, 2012,
Figure 3 shows selected results of the measure- 2013). Consequently, parity components can be
ment discussed in this article in multimodal cross- determined for the individual means of transport. A
sections. A significant heterogeneity of the modal parity determined in this way will make it possible to
split can be observed in the transport network manage and control road networks in a spirit of trans-
space. The points where multimodal cross-sections port sustainability (Beckerman, W. 1994., Ionescu G.
have been determined should account for cases of 2016., Kanister, D. 2008, Monzon-de-Caceres A. &
cross-sections of roads with different numbers of Di Ciommo F. 2016, Okraszewska, R. 2008, Stanley,
lanes. If trams are excluded from the criterion, the J. & Lucas, K. 2014).
number of multimodal cross-sections increases.
Knowledge of the traffic structure in multimodal
cross-sections allows one to determine the actual 3 CROSS-SECTION PARITY
breakdown of capacity (conventional parity, observed
parity, modal split) for a specific infrastructure ele- One method of determining the parity component
for public transport and its other types in a road
network cross-section or intersection movement
involves using a travel planner, i.e. GT Planner
(Fig. 4). This planner makes it possible to repro-
duce the transport behaviour patterns of people
living in a specific area.
The application shown in Figure 4 allows users
to plan their travels. The starting point and the des-
tination need to be indicated on the map, and the
means of transport should be chosen together with
several other significant travel parameters. Con-
sequently, a transfer route is mapped out in the
transport network covered by GT Planner opera-
tion. The route can be parameterised in the form
Figure 2. Multimodal cross-section in the Upper Sile-
of data describing transport-related behaviour:
sian urban area.

Figure 3. Modal split in cross-sections in the Upper Figure 4. GT Planner—demand parametrisation

Silesian urban area (POLAND). option.

502

TIS2017_Book.indb 502 2/8/2017 8:06:43 PM

 B = (GP, AP,T ,T
TB TP, Δt, R, MOT )
TM ,TP (3) the specific road network cross-section and with
i regard to specific means of transport.
Figure 5 illustrates the idea of how to pro-
where TBi = transport behaviour of the ith user; ceed when determining obligatory parity in road
GP = travel generation points set; AP = travel network cross-sections based on modal split cal-
absorption point set; T—start time; TM—means culated for known characteristics of observed
of transport; TP—indirect point set, dt—start demand (referring to demand observed in network
time shift window; R—space resistance; MOT— cross-sections in the form of moving vehicles or
travel motivation. GT Planner data) and suppressed demand (travels
The set of transport behaviours determined in recorded only in GT Planner). The figure shows a
accordance with equation (3) should constitute a multimodal cross-section in the transport network
certain sample of behaviours observed in the actual of the city of Katowice (Poland), in the direct
road network. This sample, after the data have vicinity of a large transport hub (with buses, trams
been processed, can provide a good estimate of the and trains being the available modes).
values of modal split distribution in actual road It is clear that three out of the six travels
network cross-sections. In GT Planner, however, planned between the same point of start and desti-
it is possible to store all route planning activities, nation (neighbouring cities of Mikołów and Kato-
also those which will not be converted into actual wice) are not present in the relevant measurement
travels. For instance, a GT Planner user may plan cross-section (marked with the horizontal red line).
a travel which is impossible at the given time given There may be various reasons for this particular
the set of data included in the GTFS (no bus or distribution of travel in the network’s space. Trav-
train is about to depart at the moment, or the wait- els that should appear in a single measurement
ing time to change to another means of transport cross-section spread out over a much wider area
is too long; no public transport line is available of the transport network for instance as a result of
between the starting point and the destination, or the failure to match the public transport offering
the combination of multimodal connections does to suppressed (hidden) demand, or of poor access
not fit within the travel time determined earlier; one to the network. Demand for a specific transport
cannot leave one’s urban bike at the point of desti- network cross-section is therefore composed of
nation, etc.) Therefore, in each transport network actual and suppressed transport behaviours:
cross-section, a value can be defined determin-
ing the number of planned transport behaviours m l
that were impossible for various reasons and con- Dcs = ∑ TBB + ∑ STBTB (5)
i j
sequently were not converted into actual behav- i =1 j =1
iours. The said value describes what is referred to
as transport-related preferences, which to a certain where Dcs = overall demand in the cross-section;
extent constitute what is referred to as suppressed TBi = transport behaviour of the ith user actu-
demand in the transport network. Knowledge of ally displayed (observed) in the cross-section;
the suppressed demand characteristic and of actual STBi = suppressed transport behaviour of the jth
transport behaviours makes it possible to determine user (not observable in any real cross-section of the
the parity based on the demand characteristic in the road network—they can be observed in GT Plan-
specific transport network cross-section instead of ner); m—number of actual transport behaviours,
the observed parity (actual modal split): l—number of suppressed travel behaviours.

| cs cs | ≥ 0 (4)

where DMSMScs = desired modal split;

The difference in value described by equation
(4) indicates how different traffic in the specific
transport network cross-section (specified parity)
is from the idea of transport sustainability (obliga-
tory parity). The said value describes the relation-
ship between satisfied and unsatisfied demand.
Balancing demand at that point also consists in
an attempt to channel suppressed demand. At the
same time, when switching to capacity measures
expressed in travels rather than vehicles, applying
the obligatory parity makes it possible to indicate Figure 5. Comparison between real modal split and
by how much the capacity value can increase in suppressed modal split (data from GT Planner).

503

TIS2017_Book.indb 503 2/8/2017 8:06:44 PM

 The gap in Figure 5 left by the three travels where Ccs = overall classical capacity in the
which did not take place using means of public cross-section; C = constant input capacity;
0
transport is probably filled immediately by private (v/c)i = volume-to-capacity ratio for level of service
means of transport. i; fd = adjustment factor for directional distribution
of traffic; fw = adjustment factor for narrow lanes
and restricted lateral clearance; fHW = adjustment
4 PARITY-BASED CALCULATION factor for the presence of heavy vehicles in the traf-
fic stream.
In line with the idea of transport sustainability and Using the above equation, one can determine
with the methodology suggested by the authors, separately (in a breakdown) the capacity for pri-
capacity in a road network cross-section can be vate vehicles and public transport vehicles:
expressed/calculated taking the obligatory parity
for access to that cross-section into account. The CcsPPuT C 0 * (v / c )i * fd * fw * fHW * ( PPuT / 100 ) (8)
relevant parity can be calculated using a tool such
as GT Planner. Therefore, the general capacity of where C PPuT = public transport share of classical
a road network cross-section can be expressed in HCM capacity in the cross-section based on the
the form of a sum: obligatory parity component for PuT;
O
= C 0 * (v / c i * fd * fw * fHW * (1 − PPuT / 100 ) (9)
C 'CS = PPuT * CCS (1 PPuT ) * CCS (6) Ccs

where C O = non-public transport share of classi-

where C 'cs = sustainable cross-section capac-
cal capacity in cross-section based on obligatory
ity; PPuT—cross-section public transport parity
parity component for other vehicles;
component; Ccs —simple capacity calculation;
Capacity expressed in the form of equations (8)
(1-PPuT) = modal split for other vehicles, parity
and (9) can be written as a unit convenient from
component.
the point of view of transport sustainability analy-
The expressions written as equation (6) make
sis, in the following form:
it possible to divide capacity in the given cross-
section of the road network into the part intended
for private vehicles and freight-carrying vehicles, C SD = C PPu
PPuT
T O
cs n C cs* p (10)
and the part intended for public transport vehicles cs
(the sum can be composed of parts related to all
means of transport). Moreover, expression (6) can where C SD —sustainable development capacity
be calculated, as far as the measures are concerned, in the cross-section; n —average occupancy of
on the basis of the number of passengers (or into means of public transport; p —average occupancy
travels), rather than of the number of vehicles. In of other vehicles. If the expressions describing the
this case, capacity for the given road cross-section components of the sum total of equation (10) are
is expressed with the number of travels broken defined as capacity for travels made by means of
down into two kinds of means of transport. If public transport C PPuT (tripi ) and capacity for travels
there is more than one lane in the given cross-sec- using other means of transport, C O(tripi ) is a com-
tion of the road, traffic can be managed reason- parison of these values makes it possible to make
ably in the lanes. If the number of travels taking reasonable changes to the traffic distribution by
place in the given road cross-section using public lanes in the given road network cross-section. The
transport is close to or larger than 50%, it needs to relation, written as the following inequality:
be assumed that a single lane should be sectioned
off exclusively for the purposes of means of public C PPuT (tripi )
≥1 (11)
transport. The key aspect consists in answering the C O (tripi ) ⋅ λ
question about the minimum percentage of these
travels that should lead to a bus lane for PuT being where λ = sustainable development capacity
sectioned off. The obligatory parity is calculated scale indicator, means that a reasonable criterion
taking into account the suppression of some trav- is obtained for sectioning off one or more lanes
els and acknowledging the fact that they need to be exclusively for the purposes of public transport
shifted to other transport network cross-sections. (e.g. bus lanes).
The capacity of a road with two lanes can be calcu- The scale indicator λ should be defined as 2
lated for instance using the HCM equation: (physical proportion between private vehicles and
public transport).
Ccs = C0 * (v / c * f * fw * f (7) In the case of capacity calculations for move-
i d HW ments at point elements of infrastructure, the

504

TIS2017_Book.indb 504 2/8/2017 8:06:46 PM

 methodology followed will be similar. The HCM the transport network in the given cross-section,
equation for the determination of capacity in a the authors expect high consistency of the results.
minor stream is as follows: In the case of suppressed travels, the reliabil-
ity of the parity component may be much lower.
v * t / 3600 This results from the fact that travels that were
e c routed (i.e. attempted) using the planner account
C =V (12)
r v * t / 3600 for an undetermined part of the general number
1− e f
of suppressed travels. The assumed high consist-
ency between the actual transport behaviours in
where V = overriding (major) stream volume; tc—
the transport network and those planned results
critical gap in the major stream [sec], tf—follow up
from the fact that according to the authors’ data,
time in the minor stream [sec].
the group that uses planners is representative of
According to the transformations applied in
the society at large. For the purposes of studying
equations (7) and (8), the sustainable capacity of a
transport behaviour patterns, this number exceeds
stream at a road intersection will be written in the
by 500% the required representative sample sizes
following manner:
in large agglomeration areas. The inhabitants like
to plan their travels, just like 42% of the general
v * t / 3600
e c public (Gretzel, U. et al. 2007).
C SD = V * ( PPuT / ) (13)
r v * t / 3600
1− e f
6 HIDDEN RESERVE OF CAPACITY AND
where: C SD —sustainable stream capacity for PuT PARITY TRADE
transport; V r = overriding stream volume;
The fact that a conventional parity is observed
in road network cross-sections resulting from the
self-regulation of road traffic streams leads to
5 RELIABILITY OF THE PARITY VALUE hidden capacity reserves appearing in each cross-
section, expressed with the number of travels that
The value of the obligatory parity component for can be translated into means of public transport.
means of public transport calculated in the man- These hidden reserves are visible only when capac-
ner described above is obtained on the basis of ity is converted from units expressed in the number
data stored in the GT Planner archive (concerning of vehicles into the number of travels using the
actual and suppressed transport behaviours). The parity component for means of public transport.
reliability of these data can be estimated on the Each replacement of a private vehicle with a pub-
basis of measurement data obtained from record- lic one leads to an increase in capacity of the rel-
ers (visual, inductive and others) installed in the evant transport network cross-section, expressed
transport network cross-sections. The proportion with the number of travels. This change, however,
expressed as equation (13) determines, therefore, should not exceed the number of travels by public
the level of accuracy of mapping demand to the transport declared by the previous users. A similar
transport network in different groups of means of aspect concerns switching between means of public
transport. transport operating in the specific network. In this
case, the prevalence of one operator may disrupt
t the sustainable development of transport.
r
CR = (14) In the authors’ opinion, there may be cases in
PPuT t
p which selected transport network cross-sections,
with a determined obligatory parity for the indi-
where CR RPPuT = reliability of the parity compo- vidual types of means of transport, might remain
nent value; tr—number of actual travels in the unused. If the PPuT parity component is unused
observed cross-section; tp number of planned/ by public transport operators, a part of it can be
requested travels in the cross-section;. sold to other transport network users. It must
The relationship expressed in the form of equa- be assumed that in transport networks, includ-
tion (13) is not linear across the whole range of the ing dense ones, every part of the resource will be
variable on the x-axis, therefore the behaviour of utilised. It is highly likely that particularly afflu-
the relationship needs further studies. This, how- ent users of private means of transport and fleet
ever, can only happen after GT Planner has been operators will be interested in resources from the
openned for public use (at present, it is at the stage transport network being resold so that they can
of functional tests). As for the calibration of the use them for their own vehicles. Nevertheless, these
number of planned travels with those observed in resources should be guaranteed to public transport

505

TIS2017_Book.indb 505 2/8/2017 8:06:48 PM

 operators. The total obligatory parity in a road parity can be supplemented with education regard
network cross-section can be written as the follow- to the culture of mobility (Okraszewska R. et al.
ing equation: 2014).

TPTN = ( PPuT ) P Pr T + PFHV + ... (15)

REFERENCES
where: TPNT = total parity in the transport net-
work; P P T = parity component for individual Aguiléra, A., Wenglenski, S., Proulhac, L. 2009. Employ-
users in the transport network; PFHV = par- ment suburbanisation, reverse commuting and travel
ity component for freight veh. in the transport behaviour by residents of the central city in the Paris
network. metropolitan area, Transportation Research Part A 43:
685–691.
The parity component for public transport is Beckerman, W. 1994. Sustainable Development: Is it a
composed of the following elements: Useful Concept? Environ. Values.
Celiński, I. Sierpiński, G., Staniek, M. 2015. Przekroje
PPuT = ∑ PPuT
TAop1 + ∑ PPuT op 2
TAP + ∑ PPuT
TTop 3 .... (16) multimodalne w sieciach transportowych. Logistyka 4:
2714–2723.
where: ∑ PPuT op1 = parity component for op1 Esztergár-Kiss D., Csiszár Cs. 2015. Evaluation of multi-
A modal journey planners and definition of service lev-
number operator in buses in the transport net-
els. International Journal of Intelligent Transportation
work; ∑ PPuT = parity component for op2
op1
Systems Research 13: 154–165.
AP
number operator in articulated buses. European Commission. 2011. White Paper: Roadmap to a
Single European Transport Area — Towards a competi-
tive and resource efficient transport system. Com(2011)
7 CONCLUSIONS 144.
Gretzel, U., Yoo, K.H., Purifoy, M. 2007. Online travel
Reviever Study, Laboratory for Intelligent Systems in
Using a travel planner with the features GT Plan- Tourism: 4–7.
ner offers makes it possible to determine in a Handy, S., Cao, X., Mokhtarian, P. 2005. Correlation
reasonable manner the various components of or causality between the built environment and travel
the obligatory parity in transport network cross- behavior? Evidence from Northern California, Trans-
sections. This comment also concerns the constitu- portation Research Part D 10: 427–444.
ents of that parity with reference to various types Ionescu G. 2016. Transportation and the Environment:
of means of transport and to the operator. Assessments and Sustainability. CRC Press Taylor &
The problem addressed in the article is simple Francis Group.
only in relation to the idea of founding a manda- Janic M. 2016. Transport Systems: Modelling, Planning,
and Evaluation. CRC Press Taylor & Francis Group.
tory modal split in selected sections of the trans- Kanister, D. 2008. The sustainable mobility paradigm.
port network and the determination on the basis Transport Policy 15: 73–80.
of their capacity. In fact, calculations of the value Khattak, A.J., Rodriguez, D. 2005. Travel behavior in
of the parity problem is very complex. Evidence of neo-traditional neighborhood developments: A case
this even the fact that there are currently no tools study in USA, Transportation Research Part A 39:
to implement the proposed concept. According to 481–500.
the authors, such as GT Planner tool enables closer Limanond T., Butsingkorn T., Chermkhunthod Ch. 2011.
parity optimal value from the point of view of sus- Travel behavior of university students who live on
tainable transport development. This value should campus: A case study of a rural university in Asia,
Transport Policy 18: 163–171.
also take into account the conditions of transport Lu, X., Pas, E.I. 1999. Socio-demographics, activity par-
infrastructure and its organisation in the various ticipation and travel behavior, Transportation Research
sections of the transport network. These issues are Part A 33: 1–18.
described in various Authors publications. Millward, H., Spinney, J. 2011. Time use, travel behav-
Another issue for discussion is the question of ior, and the rural–urban continuum: Results from the
whether the current method for capacity calculat- Halifax STAR project, Journal of Transport Geogra-
ing are adequate for appearing in the road net- phy 19: 51–58.
work, significant percentages of public transport. Monzon-de-Caceres A., Di Ciommo F. 2016. CITY-
The authors, in the Silesian conurbation identified HUBs: Sustainable and Efficient Urban Transport
Interchanges. CRC Press Taylor & Francis Group.
sections where this share is approaching to 25%. Okraszewska R. 2008. Przestrzenne sytuacje konf-
Assuming further steps towards the sustainable liktowe wywołane rozwojem systemu transportowego w
development of the transport this share should warunkach równoważenia rozwoju, Gdańsk University
continue to grow. of Technology.
The actions addressed here with regard to the Okraszewska R., Nosal K., Sierpiński G. 2014. The Role
determination of the components of the obligatory of the Polish Universities in Shaping A New Mobil-

506

TIS2017_Book.indb 506 2/8/2017 8:06:49 PM

 ity Culture — Assumptions, Conditions, Experience. Silva Cruz, I., Katz-Gerro, T. 2015. Urban public trans-
Case Study of Gdansk University of Technology, port companies and strategies to promote sustainable
Cracow University of Technology And Silesian Uni- consumption practices, Journal of Cleaner Production,
versity of Technology. Proceedings of ICERI2014 http://dx.doi.org/10.1016/j.jclepro.2015.12.007
Conference: 2971–2979. Stanley, J., Lucas, K. 2014. Workshop 6 Report: Deliver-
Sáez A.E., Baygents J.C. 2014. Environmental Transport ing sustainable public transport, Research in Transpor-
Phenomena. CRC Press Taylor & Francis Group. tation Economics 48: 315–322.
Sierpiński G., Staniek M., Celiński I. 2014. Research Szarata A. 2010a. Modelowanie symulacyjne ruchu
And Shaping Transport Systems With Multimo- wzbudzonego i tłumionego, Transport Miejski i
dal Travels — Methodological Remarks Under The Regionalny 3:14–17.
Green Travelling Project. Proceedings of ICERI2014 Szarata A. 2010b. Modelowanie ruchu tłumionego w
Conf.: 3101–3107. ujęciu symulacyjnym, Modelowanie podróży i progno-
Sierpiński G., Staniek M., Celiński I. 2016. Travel behav- zowanie ruchu 94: 169–282.
ior profiling using a trip planner. Transportatrion Szarata A. 2012. Badania ankietowe dotyczące zjawiska
Research Procedia 14C: 1743–1752. ruchu wzbudzonego w podróżach transportem zbi-
Sierpiński G. 2013. Revision of the MoCHCH05dal Split orowym, 9th Scientific-Technical Conference Logi-
of Traffic Model. Activities of Transport. Commu- trans, Szczyrk.
nications in Computer and Information Science 395: Szarata, A. 2013. The simulation analysis of suppressed
338–345. traffic. Advances in Transportation Studies: 29: 35–44.
Sierpiński G. 2017. Technologically advanced and respon- Transportation Research Board. 2010. Highway Capacity
sible travel planning assisted by GT Planner. Lecture Manual.
Notes in Network and Systems 2: 65–77.

507

TIS2017_Book.indb 507 2/8/2017 8:06:50 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The influence of road marking, shape of central island, and truck apron
on total and truck accidents at roundabouts

Jwan Kamla, Tony Parry & Andrew Dawson

Faculty of Engineering, Nottingham Transportation Engineering Centre (NTEC), University of Nottingham, UK

ABSTRACT: For in-service roundabouts where traffic volume has changed since design, road marking is
an important factor affecting the safety and capacity of roundabouts. The principal aim of this study is to
investigate the influence of road markings, truck apron, and shape of the central island on total and truck
accidents. The results indicate that, the highest rate of total and truck accidents for the selected rounda-
bouts were recorded for concentric-spiral marking, followed by spiral, partial- concentric, concentric, and
roundabouts with no markings. The majority of the selected roundabouts have oval shape and they have
higher rate of total accidents and truck accidents, than circular shape roundabouts. In addition, only three
of the selected roundabouts have truck aprons and the rate of truck accidents are high in these locations.

1 INTRODUCTION roundabouts mainly include accommodating

trucks within the available geometry.
Road transport authorities are responsible for pro- For a well-designed roundabout with balanced
moting the safety of road networks, and their aim traffic movement and efficient operations, no addi-
is to reduce fatalities and injuries arising from acci- tional road markings at approaches and within the
dents on road networks. The main cause of many circulatory carriageway are required. For in-service
accidents is driver error, and trucks in particular roundabouts where traffic volume has changed
are a type of vehicle whose effect on the safety of since design, road marking is considered to be an
the road network should be taken into account, important factor affecting the safety and capacity
as they cause many fatalities and serious injuries, of roundabouts. When the circulatory carriageway
because of their size, the freight they carry, and is wide, this may confuse drivers if there are no
the different and difficult maneuvers that they markings within the circulatory lanes. Weber et al.
require compared to cars and other types of vehi- (2009) indicated that bigger roundabouts are bet-
cles (Carstensen et al. 2001). Roundabouts have ter for trucks and other large vehicles. They stated
become popular in developed countries; in the that the use of road markings within the circula-
United Kingdom (UK) roundabouts are widely tory makes the roundabout safer for truck driv-
used instead of other junction types. Roundabouts ers, as they can stay in their own lane. Using road
are considered safer than other intersection types markings at roundabouts will reduce three types of
because; the number of conflict points decreases, accidents: “side to side collisions on the circulatory
they lead drivers to reduce their speeds, they regu- lanes, drivers being forced on to the central island,
lates turning movement of other vehicles, and they and collisions between entering and circulating
are considered to give better operational perform- vehicles” (DMRB TA 78/97 1997, p. 2/1). Using
ance (Kennedy 2007; Design Manual of Roads and road markings within the circulatory lanes at
Bridges (DMRB) TD 16/07 2007). grade-separated roundabouts increases the efficient
Geometric layout, operational analysis, and use of the circulatory lanes, as drivers can choose
safety evaluation are significant requirements the right path (DMRB TD 16/07 2007). According
for the roundabout design process. Small modi- to DMRB TA 78/97 (1997) there are four types of
fications in geometry can lead to considerable markings: concentric, partial-concentric (for wide
changes in the safety and/or operational perform- circulatory width), concentric-spiral, and spiral
ance of roundabouts. Any sudden change in geo- (more suitable for large roundabouts).
metric design leads the roundabouts to be less safe Regarding the shape of the central island, the
(Kennedy 2007). Truck rollover accidents are com- majority of the roundabout design guidelines
mon at roundabouts (Kemp et al. 1987). Weber advise the use of circular roundabouts; the major-
et al. (2009) stated that issues with trucks at ity of other types of intersections are converted

509

TIS2017_Book.indb 509 2/8/2017 8:06:52 PM

 to non-circular roundabouts and these intersec- – Lanes
tions can have poor accident records (Kennedy • 2 lanes = 39 and
2007). Alphand et al. (1991b) illustrated that oval • 3 lanes = 31,
roundabouts have higher accident rates than circu- – Grade type
lar ones, however Rodegerdts et al. (2010) reached • At grade = 19 and
the conclusion that the latter are safer because they • Grade separated = 51.
encourage a constant speed with the circulatory
Despite these differences, the selected rounda-
lanes, while oval roundabouts increase the speed in
bouts are similar with respect to design standards
the straight line then induce speed decrease when
(all are on modern motorways) and traffic levels.
the vehicle approaches the arc, which precipitates
Individual lane widths, traffic signage and mainte-
loss of control accidents within circulatory lanes.
nance regime will all be similar and other features
Another geometric parameter that can be used
such as sight lines will meet the minimum required
within the small roundabout is a truck apron,
standards.
which is “an over run area (a raised low profile area
around the central island)” which can be important
for trucks using small roundabouts (DMRB TD 2.2 Accident data
16/07 2007, p.7/4). Gingrich and Waddell (2008)
For the selected locations STATS19 data (UK acci-
in a study of a roundabout during morning and
dent statistics) were acquired within a 350 m radius
evening peak, recorded that 624 trucks travelled
from the center of the roundabouts, for 11 years
within the roundabout, 77% of which did not use
(2002–2012). The accident data from STATS19
the truck apron, and of those that did, 67% did
includes information about collision circumstances
so to prevent other cars in the adjacent lane from
(accident reference code, year, accident severity,
travelling beside them.
time and date, weekday, weather and lighting condi-
The principal aim of this paper is to investi-
tion, road type and road surface condition, vehicle
gate the influence of road markings, truck apron,
and casualty numbers, easting/northing coordi-
and shape of the central island on total and truck
nates, etc.), vehicle details (accident reference, type
accidents.
of vehicle, etc.) and reported contributory factors.

2 DATA SOURCES AND DESCRIPTION 2.3 Traffic data

Average Annual Daily Traffic (AADT) was
2.1 Description of the selected locations
acquired from the Department of Transport, Traf-
For this study, 70 roundabouts in the UK includ- fic Counts (DFT 2016).
ing 284 approaches were selected. The selected
roundabouts comprise nine roundabouts on the
2.4 Geometric information
motorway M1, ten roundabouts on the M6, six
roundabouts on the M5 and nine roundabouts on Road marking, shape of central island, and truck
the M4, with the others located on different motor- apron information, was acquired from aerial pho-
ways and A-class roads. They include signalized, tographs of an on-line mapping site.
un-signalized, or partially signalized roundabouts.
They are are either two lane, or three lane, and
the majority of roundabouts are grade-separated. 3 METHOD
Note that in this study, a roundabout is consid-
ered partially signalised when one or more of the In order to identify the locations that triggered
approaches and circulatory lanes are signalised, accidents, the Earth Point program, Excel to KML,
but not all. was used, which displays Excel files on Google
The characteristics of the whole 70 roundabouts Earth and can be found at (Earth Point 2016).
include: Firstly, the accident position coordinates were
– Legs converted from grid easting/northing to latitude/
• 3 legs = 12, longitude using Grid InQuest Version 6.6.0 available
• 4 legs = 39, as free download from Ordnance Survey Ireland:
• 5 legs = 12, and “The Grid InQuest software provides a means for
• 6 legs = 7, transforming coordinates between The European
– Traffic signals Terrestrial Reference System 1989 (ETRS89) and
• Yes = 20, the National coordinate systems of Great Britain,
• No = 28, and Northern Ireland, and the Republic of Ireland”
• Partial = 22, (Quest Geo Solutions Ltd 2004). Then the converted

510

TIS2017_Book.indb 510 2/8/2017 8:06:52 PM

 points were uploaded to Google Earth and checked 4.2 General total and truck accident trends
to see whether they were located in the selected study
According to STATS19 data, 5,520 casualties
roundabouts. Once they had all been checked, then
in all categories of accidents around the selected
the number of accidents were counted manually.
locations (entry, exit, and circulatory lanes) were
In order to separate truck accidents from total
recorded during the 11 years from 2002 to 2012.
accidents, so as to upload them separately to Google
Of these, 26.6% of accidents include trucks. It was
Earth, the STATS19 Excel sheet for vehicle details
found that more fatalities occur in truck accidents
was used, one of the columns of which indicates the
(2.10%) than in accidents involving only other
types of vehicles. Each truck accident has a refer-
types of vehicle (1.7%). This shows that truck
ence code, which was used in casualty details to find
accidents are more dangerous because of their
and highlight truck accidents. With truck accidents
size, weight, and manoeuvrability as identified by
highlighted, they were then filtered from other casu-
previous studies (DFT 2014; Trucks V 2013; US
alty details, then easting and northing, by the same
Department of Transportation 2014; Carstensen
process, was converted to latitude and longitude and
et al. 2001; Grygier et al. 2007; Kennedy 2007).
uploaded to Google Earth. After uploading them,
It was found that 60% and 57% of total and of
the numbers of truck accidents for each circulatory
truck accidents occurred within 100 m of the entry
and approach of the roundabouts were counted
line, 32% and 36% within the circulatory lanes,
manually. Note that the definition of truck accidents
and 7% of total and truck accidents happened on
in this study is any accidents involving a truck.
approaches at a distance greater than 100 m from
Note that DMRB TD 16/07 (2007) uses 100 m
the entry line. The locations that recorded higher
from the entry line as a measurement guide for the
number of accidents beyond 100 m distance are
design of roundabouts including speed limit within
very busy and big roundabouts. In Italy, Montella
100 m on approach, for maximum flare length, and
(2007), in a study for 15 urban roundabouts (55
for maximum exit kerb radius. Therefore, the ruler
approaches) of three- and four-leg types, found
from Google Earth was used to compute the percent-
that 65% of accidents occurred at roundabout
age of total and truck accidents within 100 m dis-
approaches, with 15% and 20% in the circulatory
tance, within the circulatory lanes, and beyond 100 m
and exit lanes, respectively.
distance, so as to understand how far the accidents
occurred away from a given approach’s entry line.
Road markings, shape of central island, and truck
4.3 Road marking
aprons for the selected locations were investigated
using Google Earth. Then for each type of marking, Accident rates were identified for each type of
and for each shape of roundabout central islands, marking and the results are reported in Table 1.
the number of accidents was recorded and the rate It was found that the highest rates of total and
of accidents per roundabout was identified. truck accidents were recorded in roundabouts hav-
ing concentric-spiral markings; followed by con-
centric, partial-concentric, spiral, and roundabouts
with no markings.
4 RESULTS
Five of the roundabouts with concentric mark-
ings are five and six-arms, with accident rates of
4.1 Overview
78, and the other eleven roundabouts with con-
In the UK, roundabouts are commonly used as centric marking are three and four-arm rounda-
a high traffic volume junction. This has led to bouts with accident rates of 56. And because
the construction of large roundabouts with high concentric-spiral and spiral marking is more
Inscribed Circle Diameter (ICD) and results in a
high circulating speed. Two points are important Table 1. Road marking type and the rate of total acci-
and should be considered by design organiza- dent, and truck accidents for the selected locations.
tions during roundabout rehabilitation and safety
improvement (DMRB TD 16/07 2007): Total Truck
accident accident
• the need to consider the geometry of each part, Roundabout
and Marking type no. No. Rate No. Rate
• the need to review the existing roundabout
marking Concentric 16 1006 63 218 14
Partial-concentric 15 916 61 188 13
Therefore, for the selected roundabouts accident Concentric- spiral 16 1420 89 403 25
rate based on different geometric characteristics Spiral 11 676 61 150 14
were identified and the results are given in the fol- None 12 216 18 28 2
lowing sections.

511

TIS2017_Book.indb 511 2/8/2017 8:06:52 PM

 suitable for big roundabouts (DMRB TA 78/97 truck accidents per roundabouts). In addition, they
1997), it is necessary to re-assess the big rounda- have higher AADT as reported in Table 2; oval
bouts that have concentric markings and consider shape roundabouts have higher AADT by 41%
changing these markings in order to make the relative to circular shape roundabouts. This result
path within the roundabouts more efficient for is in line with Alphand et al. (1991b) although their
the users, and thereby reduce accidents that might result was for total accident rates.
occur because of insufficient marking within the
roundabouts.
4.5 Truck apron
Roundabouts with concentric spiral markings
are associated with a higher rate of total and truck The set of roundabouts considered in this study
accidents. The rates of total and truck accidents for included 19 at-grade roundabouts and they were
concentric and spiral type marking are similar, fol- considered small compared to grade-separated
lowed by partial concentric marking. Note that five roundabouts. For the 19 selected locations, only
of the roundabouts that have no markings within the three roundabouts have truck aprons. In these
circulatory lanes are grade-separated and the rate of locations, the rates of total and truck accidents
total and truck accidents are 25 and 3.4, respectively; are 90, and 20, respectively; while for the other 16
while in the other seven at-grade locations with no locations without truck apron the rate is 16, and
marking the rate of total and truck accidents is lower 2.0, respectively. This indicates that availability
(13 and 1.57, respectively). This indicates that these of truck aprons in the three small roundabouts
grade-separated roundabouts may benefit from has not reduced accidents to the general level and
marking within the circulatory lanes. could even be somewhat causative of the much
Regarding the roundabouts that have spi- higher accident rates observed; given that both
ral markings, four of them are small at-grade truck and all-vehicle accident rates are increased
roundabouts. Based on DMRB TA 78/97 (1997) relative to other roundabouts. Note that the per-
spiral markings are suitable for big roundabouts. centage of trucks in these locations is high (9–10%)
Comparing these three at-grade locations to the and, probably because these locations have enough
other eight roundabouts with spiral marking, a space (circulatory width of >10 m) for trucks to
similar rate of total accidents (59 relative to 62) negotiate the roundabouts, they may not use the
was recorded; and higher truck accident rate (16 truck aprons. Gingrich and Waddell (2008) have
relative to 13) was recorded. These three at-grade found that during morning and evening peak peri-
roundabouts probably require re-assessment. ods 77% of trucks did not use truck aprons.

4.4 Shape of central island

5 CONCLUSIONS
In this study, there are 43 oval-shape central island
roundabouts, and 27 circular shape roundabouts. This paper reported the total and truck accident
Table 2 reports the rates of total accidents, truck numbers within 70 roundabouts in the UK, and
accidents per roundabout, and AADT with respect considered them in light of their road markings.
to oval shape and circular shape roundabouts. It is Based on DMRB RD 16/07 (2007) big rounda-
clear that the rates of total and truck accidents are bouts require spiral and spiral-concentric markings,
higher in oval shaped rather than circular shaped while small roundabouts require concentric mark-
roundabouts. The probable reason for having this ings. In this study regarding road marking, some
result is 42 out of 43 of the oval shape roundabouts big roundabouts have simple concentric marking
are grade-separated. The rate of total and truck and recorded high total and truck accident rates.
accidents at grade-separated roundabouts is higher In addition, spiral markings are recommended for
relative to these rates at at-grade roundabouts (a big roundabouts but some of the small roundabouts
rate of 71.9 relative to 29.9 for total accidents per have spiral markings and recorded high rates of total
roundabouts, and a rate of 17.3 relative to 5.4 for and truck accidents and these locations may require
a re-assessment regarding marking in order to make
Table 2. Central is land shape type. the movement path easier for the roundabout users
and provides safety at these roundabouts.
Total Truck Oval shaped roundabouts have a higher rate of
Accident AADT
accident total and truck accidents relative to those of cir-
Shape Roundabout per cular shape but all the oval shapes are big rounda-
type no. No. Rate No. Rate roundabouts bouts with higher AADT.
Oval 43 3315 77 811 19 60543
For the selected locations, only three of the
Circular 27 919 34 176 7 35388
roundabouts have truck aprons and the rate
of truck accidents are high in these locations.

512

TIS2017_Book.indb 512 2/8/2017 8:06:52 PM

 Addition of truck aprons in these roundabouts has Study of heavy truck air disc brake effectiveness on
not reduced accident rates to the general level. the national advanced driving simulator. In: The 20th
ESV Conference Proceedings.
Kemp, R., Chinn, B. and Brock, G. 1978. Articulated
vehicle roll stability: methods of assessment and
REFERENCES effects of vehicle characteristics. In Transport and
Road Research Laboratory: TRRL Laboratory Report
Alphand F, Noelle U and Guichet B, 1991b. Rounda- 788.
bouts and road safety; state of the art in France. Inter- Kennedy, J. 2007. International comparison of roundabout
sections without traffic signals II. Proceedings of an design guidelines, TRL.
International Workshop 18–19 July 1991 in Bochum, Montella, A. 2007. Roundabout in-service safety reviews:
Germany, pp126–140. safety assessment procedure. Transportation Research
Carstensen A Gitte, Hansen W, Hollnagel V, Højgaard H, Record: Journal of the Transportation Research Board,
Jensen I, Kines P, Klit L, Kofoed P, Mikkelsen J, and 2019, 40–50.
Petersen O, 2001. Analysegruppen for Vejtrafikuheld Quest Geo Solutions Ltd. 2004. Grid InQuest DLL User
(AVU), Lastbiluheld- en dybdeanalyse af 21 uheld. Manual (Version 6).
Department For Transport (DFT) 2014. Reported road Rodegerdts, L., Bansen, J., Tiesler, C., Knudsen, J.,
casualties in Great Britain, Annual Report, Statistical Myers, E., Johnson, J., Moule, M., Persaud, B., Lyon,
Release. C., Hallmark, S., Isebrands, H., Crown, R., Guichet,
Design Manual for Roads And Bridges, TA 78/97, 1997. B., and O’Brien, A., 2010. NCHRP Report 672:
Design of Road Markings at Roundabouts, London, Roundabouts: An informational guide, Transporta-
UK. tion Research Board, Washington, DC.
Design Manual for Roads And Bridges, TD 16/07, 2007. Trucks, V. 2013. European Accident Research And Safety
Geometric Design of Roundabouts, London, UK. Report. Available at: http://www.volvotrucks.com/
DFT 2016. Department for Transport, Traffic Counts. SiteCollectionDoments/VTC/Corporate/Values/
Available at: http://www.DfT.gov.uk/traffic-counts/ ART%20Report%202013.pdf/[Accessed 29 Sep.
area.php/ [Accessed 10 Aug. 2016]. 2016].
Earth Point 2016. Excel To KML - Display Excel files on US Department Of Transportation 2014.”Traffic Safety
Google Earth. Available at: https://www.earthpoint.us/ Facts”, National Highway Traffic Safety Administra-
ExcelToKml.aspx/ [Accessed 10 Aug. 2016]. tion, DOT HS 812 101, Washington, DC 20590.
Gingrich, M. and Waddell, E. 2008. Accommodating Weber, M. Button, N. 2009. Accommodating small and
trucks in single and multilane roundabouts. In: Trans- large users at roundabouts. In: Annual Conference of
portation Research Board, National Roundabout Con- the Transportation Association of Canada: Sustain-
ference, Kansas City, 12–18. ability in Development and Geometric Design for Road-
Grygier, P.A., Garrott, W.R., Salaani, M.K., Heydinger, ways, Vancouver, British Columbia, Canada.
G.J., Schwarz, C., Brown, T. and Reyes, M. 2007.

513

TIS2017_Book.indb 513 2/8/2017 8:06:52 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A runway veer-off risk assessment based on frequency model:

Part I. probability analysis

L. Moretti, G. Cantisani, P. Di Mascio & S. Nichele

Dipartimento di Ingegneria Civile, Edile e Ambientale. Sapienza, University of Rome, Rome, Italy

S. Caro
Departamento de Ingenieria Civil y Ambiental, Universidad de Los Andes, Bogota, Colombia

ABSTRACT: This two-part paper presents a comprehensive airport risk assessment methodology for
veer-off accidents of aircraft runways. Veer-off is a lateral runway excursion, and it occurs when an air-
craft leaves the runway during a movement. The most frequent causes for this type of accidents are inap-
propriate pilot performance and aircraft condition; while other contribution factors include transversal
wind, contamination of the runway, and poor visibility. The consequences of these events involve both
damage to the aircrafts and potential health effects on passengers and crew members. Statistical data
collected throughout the world by the authors were used as a reference to conduct frequency and risk
assessments of veer-offs. The objective of the first part of the paper is to assess the probability of a veer-
off accident. To accomplish this objective, a cumulative probability distribution was used to represent
the phenomenon. It was found that the exponential curve described by the Poisson distribution could be
properly used to describe this type of accident. The results show that the average frequency of a veer-off
accident is 1.44 in ten million movements for commercial flights over 30 Mg (ton) and that veer-offs are
more frequent to occur during landing than during take-off procedures. The proposed analysis permits to
compute the probability of veer-offs at any airport, after considering its specific conditions (e.g. number
and type of plane, kind of movement, bearing capacity of the subgrade, etc.).

1 INTRODUCTION which includes to calculate both the probability of

occurrence of certain hazard, as well as the mag-
Risk management represents a fundamental aspect nitude of the consequences of such an occurrence
of Air Traffic Management (Eurocontrol 2001). (ISO 1995, International Civil Aviation Organiza-
Several studies reported in the literature have iden- tion 2010).
tified some of the main factors involved in air- A runway veer-off is an operational risk present
port accidents: weather; crew technique decision in airports, defined as the excursion in which an
or performance; and airport system factors (e.g. aircraft leaves the side of the runway. ICAO Annex
malfunctioning of assistance instruments) (Flight 14 provides recommendations for acceptable con-
Safety Foundation 2000, Flight Safety Foundation sequences in case that an aircraft runs off the
2009). side of the runway. Specifically, the Aerodrome
The concept of safety requires a detailed analy- Design Manual Part 1. 5.3.22 specifies the follow-
sis accounting for different perspectives of the sub- ing regarding the function of runway strips “…., it
jects involved. According to Kirkland et al. (2004), should be graded in such a manner as to prevent the
there are several objective or subjective types of collapse of the nose landing gear of the aircraft. The
safety levels of a system: the real level of safety, surface should be prepared in such a manner as to
the target level of safety and the achieved or the provide drag to an aircraft and below the surface, it
deemed level of service. A good-quality safety man- should have sufficient bearing strength to avoid dam-
agement design for airport infrastructures needs an age to the aircraft…..” (International Civil Aviation
accurate methodology to evaluate the risk and pre- Organization 2013a).
vent accidents (Attaccalite et al. 2012, Di Mascio In many countries, national specifications
et al. 2012, Canale et al. 2005, Wong et al. 2009a, b, require special risk evaluations (Moretti et al. in
Kirkland et al. 2004, Di Mascio & Loprencipe in press). In Italy, for example, the licensing regula-
press). Therefore, to reduce risk, service providers tions of aerodromes, issued by the Italian Civil
should carry out a risk assessment (Spriggs 2002), Aviation Authority, ENAC, have transposed the

515

TIS2017_Book.indb 515 2/8/2017 8:06:54 PM

 ICAO standard and, in case of insufficient bear- that involved several types of flights (passenger
ing strength of the strip, ENAC requires a run- aviation, cargo, training, military and agricultural
way veer-off risk assessment (Ente Nazionale per flights) were analysed.
l’Aviazione Civile 2014). At last, this study used a reduced number of
It should be noticed that lateral offset from the records to get a more homogeneous database.
travel lanes represents a critical operational aspect This is due to the fact that accident reports of gen-
not only in airport fields. Downgrade roadways, eral aviation are not always complete and clearly
for example, are required to have adjacent truck recorded in the literature, so the overall database
escape ramps. These ramps act as emergency areas would not have been consistent. Thus, the analysis
and provide a location for out-of-control vehicles was conducted on 301 events (accidents or serious
(National Cooperative Highway Research Pro- incidents), related to passenger/cargo flights within
gram 1992, American Association of State High- the period 1953–2015 (Fig. 1). The number of
way and Transportation Officials 2001). As for events reported until the 1970s was relatively low,
the road context, a systematic methodology has most likely due to underreporting.
determined the need for truck escape ramps and The date of the accidents allowed calculat-
their locations (Abdelwahab et al. 1996): likewise, ing the average interval between two consecutive
airport design should reduce the risk of veer-off events. Considering all accidents involving pas-
events to its least. For this reason, the materials senger and cargo flights from 1953 to 2015, the
and geometrical configuration used in the lateral number of days between two successive events is
airport runway safety areas should guarantee tech- 78. However, since data corresponding to acci-
nical, environmental and economic sustainability, dents that occurred many decades ago are scarcely
in order to balance the stakeholders’ competing available, as observed in Figure 1, if we consider
interests (Di Mascio et al. 2012, Di Mascio et al. only the events since the 1980s, the frequency of
2014, Moretti 2013, Moretti et al. 2014, Transport the recorded veer-offs levelled off at approximately
Canada Civil Aviation 2011). In the case of a veer- 8.6 events per year, resulting in an average number
off event, the materials used for the runway strip of days between two veeroff events of 42. Figure 2
should contribute to reduce the risk of damage to represents the frequency of the observed times
an aircraft, the health impact on the occupants of between two consecutive events.
the airplane and the air traffic interruption (Trans-
port Canada Civil Aviation 2011).
This research assesses the risks of veer-offs by
evaluating their likelihood and severity. The anal-
ysis takes into account existing methodologies
to assess risk of plane crashes (Wong et al. a, b,
Kirkland et al. 2004), and it considers that the level
of risk has to be As Low As Reasonably Practi-
cable (ALARP) (Airport Cooperative Research
Program 2011, Airport Cooperative Research Pro-
gram 2014, UK Civil Aviation Authority 2014).
This analysis considers available statistical data
and it proposes a new reference system to describe
this type of accident.
Figure 1. Number of veer-offs by year.

2 METHODOLOGY

The present research relies on statistical data

obtained from several international sources
(Agenzia Nazionale per la Sicurezza del Volo
2016, Airdisaster.com 2016, Airsafe 2013, Aus-
tralian Transport Safety Bureau 2012, Aviation
safety network 2005, Boeing 2000, 2002, 2003,
2006, 2010, 2012, 2013, 2014, 2015, 2016, General
Aviation Manufacturers Association 2005, Flight
Safety Foundation 2014, International Civil Avia-
tion 2013b, National Transportation Safety Board
2014). Based on this information, 3500 veer-off Figure 2. Frequency of observed time between two
records that occurred between 1953 and 2015 and consecutive events.

516

TIS2017_Book.indb 516 2/8/2017 8:06:54 PM

 Taking into account data collected over the last Table 1. Common causes of veer-offs.
years, the authors estimated the frequency of these
accidents against the total number of movements Frequency
over the same period of time.
Causes %
The analysis considered the frequency of veer-
offs that occurred to commercial flights over 30 Pilot performance 63
Mg (ton). The results from this analysis show that Wind characteristics 25
the average veer-off probability is equal to 1.44 ⋅ Aircraft landing gear system 14
10−7, whereas the average frequency of an aircraft Aircraft maintenance 7
accident is 30 ⋅ 10−7. The order of magnitude of Aerodrome management 6
the veer-off rate computed taking into account the Weather conditions 6
evaluated databases coincides to the one reported Runway surface (e.g. presence of 6
in the “Final Report on the Risk Analysis in Sup- contaminants)
port of Aerodrome Design Rules”, prepared by Engine failure 6
the Norwegian Civil Aviation Authority (AACN), Collision with obstacle(s) 3
which reports a frequency of occurrence of a veer- Other 9
off of 2.2 ⋅ 10−7 (Norwegian Civil Aviation Author-
ity 2001).
A runway veer-off may occur during four dif- Table 2. Causing factors classified by category.
ferent phases of a flight: 1) landing, 2) take-off, 3)
Frequency
taxiing, or 4) during other operational condition.
Ac-cording to the collected data, 98% of veer-offs Factors %
occurred during landing and take-off, and veer-offs
during landings were found to be approximately 3 Aircraft 46
times more likely to happen than during take-offs Pilot performance 39
(223 veer-offs during landings, 78 during take- Weather conditions 31
offs), as confirmed by AACN (2001). The ratio of Runway surface characteristics 17
incidence between the two types of accidents is 2.8 Airport management 11
(Fig. 3).
Table 1 summarizes ten common causes of veer-
off events with their associated frequencies for the The Ishikawa diagram in Figure 4 categorizes all
same set of data. As shown, the total of all per- potential causes that were found as a result of this
centages exceeds 100 percent because of the mul- analysis.
tiple contribution factors associated with the same Consequences of a veer-off include damage to
veer-off event. the aircrafts and potential health impact on occu-
The causes for veer-offs were categorized pants. Regarding mechanical damages, Table 3
into five general classes: 1) aircraft conditions, summarizes eight types of possible consequences,
2) pilot performance, 3) weather conditions, 4) with their associated frequencies, which were
conditions of the runway surface, and 5) airport obtained from the analysed data.
management. Statistics on health effects show that there were
A transfer-in procedure within a fault tree anal- 383 fatalities that occurred in 32 veer-offs. The
ysis (FTA) was applied to calculate the likelihood average number of fatalities per veer-off with
of occurrence of each factor, as listed in Table 2 deaths was 11.90; whereas the average number of
(Ale 2002). deaths per veer-off was 1.27 (calculated consider-
ing 301 veer-offs collected by the authors). The
analysis of the available data also shows that the
accidents recorded in the case of passengers and
cargo flights were more serious but less frequent
than in other types of flights.
Three categories of accidents were considered in
this study:
• accidents that resulted in deaths,
• accidents that caused injuries but did not involve
deaths,
• accidents that did not cause injuries or deaths.
Figure 5 shows the distribution of health effects
Figure 3. Percentage of veer-offs by phase of flight. caused by veer-offs.

517

TIS2017_Book.indb 517 2/8/2017 8:06:55 PM

 3 RESULTS

The collected records highlight that the severity of

a veer-off accident largely depends on the type of
flight. As a matter of fact, injuries suffered by occu-
pants due to veer-offs are more likely to occur in
commercial aviation (passenger and freight flights)
than in other types of flights. Indeed, for veer-offs
occurring in these flights, the ratio between fatali-
ties and occupants on board is equal to 36%, which
proves the severity of these events.
The values in Tables 4 to 6 correspond to the sta-
tistical parameters pi, calculated for each j category
of the total accidents, according to the following
equations:
nevents , j
p1, j = (1)
Nevents ,TOTAL
n pax , j
p2, j = (2)
N pax ,TOTAL
ndeaths
p3 = (3)
N pax ,TOTAL
ndeaths
p4 = with j = 1 (4)
N pax , j
ninjuries
p5, j = with j = 1, 2
j ,j
Figure 4. Ishikawa diagram for veer-off accident. (5)
N pax ,TOTAL
Table 3. Veer-off consequences.
ninjuries
p6, j = with j = 1, 2
j ,j
(6)
Frequency N pax , j
Consequences %
Table 4. Accidents with deaths.
Mechanical damage 50
Collision with obstacle(s) 32 Amount
Damage to landing gear system 32
Indicator %
Crash into an embankment or in a 13
drainage channel p1 11
Fire 11 p2 25.5
RWY, taxiway (TWY) or apron crossings 5 p3 9.2
Crossing of airfield perimeter fence 4 p4 36
Other 3 p5 0.1
p6 1.3

Table 5. Accidents with injuries

(without deaths).

Amount

Indicator %

p1 18
p2 25.5
p5 2.6
Figure 5. Percentage of veer-offs by health effect on air- p6 10.5
craft occupants.

518

TIS2017_Book.indb 518 2/8/2017 8:06:55 PM

 Table 6. Accidents without injuries function of the type of movement. According to
or deaths. the collected data, the frequency of occurrence as
a function of the instrumentation of the runway
Amount and the type of movement was calculated, obtain-
Indicator % ing the following results:
• instrument landing: 1.37 ⋅ 10−7,
p1 71 • non-instrument landing: 9.72 ⋅ 10−7,
p2 49 • instrument take-off: 6.96 ⋅ 10−8,
• non-instrument take-off: 6.82 ⋅ 10−9.
where nevents,j is the number of observed accidents The statistical analysis of the veer-off events
with j category, j is a code to representing the type show that these accidents could be properly de-
of accident (with fatalities j = 1, injuries j = 2 or scribed using a Poisson distribution because of
without them j = 3), Nevents,TOTAL is the total number the following characteristics of the phenomenon
of observed accidents, npax,j is the number of pas- (Scozzafava, 1995):
sengers involved in accidents with j category,
• the event is something that can be counted in
npax,TOTAL is the total number of passengers involved
whole numbers,
in veer-offs, and ndeath is the total number of fatali-
• occurrences are independent, so that one occur-
ties involved in veer-offs.
rence neither diminishes nor increases the chance
As observed, fatal accidents account for approx-
of another,
imately 11% of the total amount of accidents, but
• the events occur randomly in a given interval of
involve 25.5% of the total people on board of the
time (or space),
flights evaluated. Also, the overall percentage of
• the probability of occurrence of an event in a
deaths of occupants during veer-offs was found to
small time interval Δt is proportional to Δt and it
be 36%. Considering all the passengers involved in
can be estimated as proportional to the product
all the examined veer-offs, the percentage of deaths
λΔt, where λ is the mean number of events per
was 9.2% (Table 4).
interval Ta (i.e., the average interval between two
The cases of accidents with injuries, but without
consecutive events), as expressed by Equation 7,
deaths, account for approximately 18% of the total,
and
and involve 25.5% of people on board (Table 5).
Finally, the results show that there were no
1
health effects in 71% of the events, which involve λ= (7)
49% of the total amount of people on board of the Ta
affected planes (Table 6).
As mentioned previously, veer-offs may be due • it is possible to count how many events have
to two main reasons: a misalignment at the begin- occurred, but it is meaningless to ask how many
ning of a take-off or landing procedure, or a loss of such events have not occurred yet.
of the directional control by the pilot during the
Equation 8 gives the probability of observing x
landing, after touchdown, or during the take-off
events within a time interval t:
roll.
As reported in the literature, accident rates also
depend on the type of movement, as demonstrated ( λt ) x − λt
px P (X λ x) = e (8)
in Figure 3. Indeed, the accident rate of veer-off x!
depends strongly on the instrumentation available
during the movement: on non-instrument runways where λ is the mean, variance and prevision values
or with non-precision instruments, the frequency of distribution.
of occurrence is higher than the one recorded on Veer-off accidents have been modelled also by
runway with precision instruments. With refer- evaluating the path related to the event. Regarding
ence to landing veer-offs, the analysis conducted the spatial occurrence of the veer-offs, the initial
by AACN found that 59% of the events occurred configuration representing the departure of the
after a precision approach, while the remaining air-craft in the runway affects the outcome and
41% occurred as a result of a non-instrument or the final stopping location of the aircraft. This
non-precision instrument approach (Boeing 2015). research evaluated the veer-off path by defining
With respect to take-off veer-offs, the ratio between three geometrical characteristics: 1) the local drift
flights with precision instruments and flights with angle, 2) the average drift angle, and 3) the drift
no-instruments or with non-precision instruments radius. The initial two cases refer to a polar coor-
is on average 10.2. This value allowed calculating dinate system with a pole located at the runway
the frequency of occurrence of a veer-offs as a threshold, whereas the former is determined by

519

TIS2017_Book.indb 519 2/8/2017 8:06:57 PM

 a pole at the initial point of the veer-off path, as At the end of a veer-off accident, there are three
shown in Figure 6. main scenarios depending on the final position of
The data herein evaluated show that in more the aircraft:
than 66% of veer-offs the average drift angle was
• veer-off over the runway,
up to 5 degrees, as shown in Figure 7.
• veer-off over the Cleared and Graded Area
The obtained average drift angles are repre-
(CGA), with the external gear over the CGA,
sented in Figure 8.
but the wing tip inside the strip,
The final point of the drift path leads to deter-
• veer-off over the CGA, with the wing tip over
mine the lateral probability distribution, as depicted
the strip.
in Figure 6: it describes the probability that at the
end of a veer-off the aircraft travels beyond a cer- The final position of the aircraft allows calculating
tain distance from the centreline of the runway. the transversal and longitudinal distribution probabil-
The curve trends of the final positions resulted ity curves related to the centreline of the runway. The
in a negative exponential distribution. They were position of the aircraft after the veer-off from the lon-
compared with the curves obtained by the Norwe- gitudinal and transverse edges of the runway could be
gian Civil Aviation Authority in a veer-off event described by a distribution function of the cumulative
risk analysis (Norwegian Civil Aviation Authority probability. Considering the final position and the size
2001) and it is possible to conclude that the results of the runway, the possibility of describing the phe-
are comparable. nomenon using an exponential distribution function
was examined. The probability distribution depends
on a 2D geometrical configuration and, in order to
simplify the model, the problem was divided in two 1D
scenarios mutually perpendicular. The overall curve
of probability consists of four separate branches.
Two branches have the abscissa axis orthogonal to
the longitudinal axis of the runway and two branches
Figure 6. Location of accidents system.
have the abscissa axis coincident with the longitudinal
axis of the runway. The two transversal branches of
the cumulative probability distribution are not neces-
sarily symmetrical in relation to the geotechnical and
anemometry local conditions. Likewise, the two longi-
tudinal branches with the origin in the midpoint of the
runway are not necessarily symmetrical in relation to
the runway length required by traffic.
Figure 9 represents the average cumulative prob-
ability distributions related to veer-off events during
both take-off and landing procedures. As observed,
the data does present an exponential trend.
Taking into account the movement type, the
lateral deviation was found to be larger during
take-offs than during landing procedures consider-
ing equal levels of veer-off probabilities (Fig. 10).
Figure 7. Observed distribution of average drift angles.

Figure 9. Frequency distribution of the lateral distance

Figure 8. Observed distribution of drift radius. of the planes after veer-off events.

520

TIS2017_Book.indb 520 2/8/2017 8:06:57 PM

 take-off procedures, as mentioned previously, but
the average final position of the aircraft during
take-off is more distant from the runway centreline
than during landing.
The probability distributions of the final trans-
versal location of the aircrafts after an incident
were obtained from the analysis of the veer-off
reports. The proposed probabilistic model for the
frequency of the accidents uses a multivariate
logistic regression according to the Poisson distri-
bution. This probabilistic density distribution was
observed to properly describe the phenomenon,
since it has an exponential qualitative trend and
Figure 10. Cumulative probability of veer-off. because 90% of the veer-off events are included
within the strip width.
Also, the analysis show that the final spatial loca-
Results of this research are close to the ones tion of an aircraft involved during veer-off may be
obtained by AACN. represented through a polar coordinate system.
The drift angle and the drift radius allow identify-
ing the risk areas in and around airport runways.
4 CONCLUSIONS The findings obtained in the study demonstrate the
importance of veer-offs and the need for mitigat-
The first part of this paper presents the methodo- ing these accidents.
logical advancements achieved in the development
of veer-off accident frequency models.
The authors worked using a single comprehen-
REFERENCES
sive database of relevant accidents to quantify the
probability of veer-off events during take-off and Abdelwahab, W.M. & Morral, J.F. 1996. Determining
landing, both instrument and non-instrument or need for and location of truck escape ramps. Journal
with non-precision instrument operations. Veer- of Transportation Engineering, Vol. 123, No. 5, pp.
offs were clustered according to their frequencies, 350–356. doi: http://dx.doi.org/10.1061/(ASCE)0733-
causes and con-sequences. Inappropriate pilot per- 947X(1997)123:5(350).
formance, aircraft damage, meteorological condi- Attaccalite, L., Di Mascio, P., Loprencipe, G. &
tions and runway contamination were determined Pandolfi, C. 2012. Risk Assessment Around Airport.
to be the main causes of the accidents that involved Procedia: Social & Behavioral Sciences, pp. 852–861,
53. doi:10.1016/j.sbspro.2012.09.934.
areas lateral to the run-way and, rarely, also areas
Agenzia Nazionale per la Sicurezza del Volo 2016. Avail-
across the boundaries of the airports. Statistics of able from www.ansv.it.
accidents and operations show that since 1990 the Airdisaster.com 2016. Available from www.airdisaster.
frequency of a veer-off is 1.44 in ten million move- com.
ments, whereas the average frequency of an aircraft Airport Cooperative Research Program (ACRP) 2011.
accident is 30 in ten million movements. ACRP Report 50: Improved models for risk assess-
According to the available data, it was concluded ment of runway safety areas. Transport Research
that 97% of the total veer-offs occurred during Board.
landing and take-off procedures, and that veer-offs Airport Cooperative Research Program 2014. ACRP
Report 107: Development of a runway veer-off loca-
during landing are approximately 2.8 times more
tion distribution risk assessment model and reporting
likely to occur than during a take-off. Concerning template. Transport Research Board.
the consequences of a veer-off, it was determined Airsafe 2013. Available from www.airsafe.com.
that the mechanical damage of the aircraft is the Ale, B.J.M. 2002. Risk assessment practices in The Neth-
most likely consequence (50%). On the contrary, erlands. Safety Science, vol. 40, pp. 120–126.
the least likely consequence is the crossing of the American Association of State Highway and Transporta-
airfield perimeter fence (5%). Regarding the health tion Officials 2001. A policy on geometric design of
effects of these accidents on occupants, statistics highways and streets. Green book, pp. 259–269.
show that at least one fatality occurred in 11% of Australian Transport Safety Bureau 2012. Available from
www.atsb.gov.au.
the cases, whereas injuries were reported in 18% of
Aviation safety network 2005. Airliner accident statistics
the total veer-offs. Besides, there were not health 2004. Statistical summary of fatal multi-engine air-
effects in 71% of the events. liner accidents in 2004.
With respect to the movement, this type of acci- Boeing 2000. Statistical summary of commercial jet air-
dent is more frequent during landing than during plane accidents - worldwide operations 1959–1999.

521

TIS2017_Book.indb 521 2/8/2017 8:06:58 PM

 Boeing 2002. Statistical summary of commercial jet Flight Safety Foundation 2014. Available from www.
airplane accidents - worldwide operations -2001. aviation-safety.net.
Boeing 2003. Statistical summary of commercial jet General Aviation Manufacturers Association 2005. Gen-
airplane accidents - worldwide operations 1959–2002. eral aviation statistical databook 2004.
Boeing 2006. Statistical summary of commercial jet International Civil Aviation Organization 2010. Annex
airplane accidents - worldwide operations -2005. 13: Incident reporting, data systems and information
Boeing 2010. Statistical summary of commercial jet exchange.
airplane accidents - worldwide operations 1959–2009. International Civil Aviation Organization 2013a. Annex
Boeing 2012. Statistical summary of commercial jet 14: Aerodrome Design and Operations.
airplane accidents - worldwide operations 1959–2011. International Civil Aviation Organization 2013b. Safety
Boeing 2013. Statistical summary of commercial jet Management Manual, SMM. ICAO Doc 9859.
airplane accidents - worldwide operations 1959–2012. International Air Transport Association 2015. Safety
Boeing 2014. Statistical summary of commercial jet report.
airplane accidents - worldwide operations 1959–2013. ISO 1995. ISO 8402:1995/BS 4778. Quality management
Boeing 2015. Statistical summary of commercial jet and quality assurance. Vocabulary.
airplane accidents - worldwide operations 1959–2014. Kirkland, I.D.L., Caves, R.E. & Humphreys, I.M. 2004.
Boeing 2016. Statistical summary of commercial jet An improved methodology for assessing risk in air-
airplane accidents - worldwide operations 1959–2015. craft operations at airports, applied to runway over-
Canale, S., Distefano, N. & Leonardi, S. 2005. A risk runs. Safety Science 42, pp. 891–905.
assessment procedure for the safety management of Moretti, L., Di Mascio, P. & D’Andrea, A. 2013. Envi-
airport infrastructures. Proceedings at III Interna- ronmental Impact Assessment of Road Asphalt Pave-
tional Congress of Società Italiana Infrastrutture ments. Modern Applied Science, 7(11), 1–11. http://
Viarie, SIIV. dx.doi.org/10.5539/mas.v7n11p1.
Di Mascio, P., Cardi, A., Di Vito, M. & Pandolfi, C. Moretti, L. 2014. Technical and economic sustainability
2012. Distribution of Air Accidents Around Runways. of concrete pavements. Modern Applied Science, 8(3),
Social and Behavioral Sciences, vol. 53, pp. 862–871, 1–9. DOI: 10.5539/mas.v8n3p1.
Proceedings, V International SIIV Congress Sustain- Moretti, L., Cantisani, G. & Caro, S. (in press). Air-
ability of Road Infrastructures, Rome, Italy, 29th-31st port veer-off risk assessment: an Italian case study.
October 2012. doi: 10.1016/j.sbspro.2012.09.935. ARPN Journal of Engineering and Applied Sciences.
Di Mascio, P., Moretti, L. & Panunzi, F. 2012. Economic (1819–6608).
Sustainability of Concrete Pavements. Procedia - National Cooperative Highway Research Program 1992.
Social and Behavioral Sciences, vol. 53, pp. 125–133, Truck escape ramps. NCHRP, n. 178.
Proceedings, V International SIIV Congress Sustain- National Transportation Safety Board 2014. Available
ability of Road Infrastructures, Rome, Italy, 29th-31st from www.ntsb.gov.
October 2012. doi: 10.1016/j.sbspro.2012.09.866. Norwegian Civil Aviation Authority 2001. Final report on
Di Mascio, P., Loprencipe, G. & Moretti, L. 2014. Com- the risk analysis in support of aerodrome design rules.
petition in rail transport: methodology to evaluate Rules, AEAT/RAIR/RD02325/R/002 Issue 1, 2001
economic impact of new trains on track. ICTI2014 – Scozzafava, R. 1995. Primi passi in probabilità e statis-
Sustainability, Eco-efficiency and Conservation in tica. Zanichelli Editore.
Transportation Infrastructure Asset Management, Spriggs, J. 2002. Airport risk assessment: examples, mod-
669–675, Losa & Papagiannakis (Eds)-Taylor & Fran- els and mitigations. 10th safety-critical systems Sym-
cis Group. phosium, Southampton.
Di Mascio, P. & Loprencipe, G. in press. Risk analysis Transport Canada Civil Aviation 2011. Airport pave-
in the surrounding areas of one-runway airports: a ment bearing strength reporting. Advisory Circular
methodology to preliminary calculus of PSZs dimen- AC302-011.
sions. ARPN Journal of Engineering and Applied Sci- UK Civil Aviation Authority, UK CAA 2014. Licensing
ences (ISSN 1819-6608). of aerodromes.
Ente Nazionale per l’Aviazione Civile (ENAC) 2014. Wong, D.K.Y., Pitfield, D.E., Caves, R.E. & Appleyard,
Regolamento per la costruzione ed esercizio degli A.J. (2009a). The development of a more risk-sensi-
aeroporti. tive and flexible airport safety area strategy: Part I.
Eurocontrol 2001. ESARR 4 – Risk assessment and miti- The development of an improved accident frequency
gation in ATM. model. Safety Science 47, pp. 903–912. doi:10.1016/j.
European Aviation Safety Agency 2014. Annual Safety ssci.2008.09.010.
review 2013. Wong, D.K.Y., Pitfield, D.E., Caves, R.E. & Appleyard,
Flight Safety Foundation 2000. FSF ALAR Briefing A.J. (2009b). The development of a more risk-sensitive
note 8.1 - Runway excursions and runway overruns. and flexible airport safety area strategy: Part II. Acci-
Available from www.aviation-safety.net. dent location analysis and airport risk assessment case
Flight Safety Foundation 2009. Reducing the risk of run- studies. Safety Science 47, pp. 913–924. doi:10.1016/j.
way excursions. Available from www.aviation-safety.net. ssci.2008.09.011.

522

TIS2017_Book.indb 522 2/8/2017 8:06:58 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A runway veer-off risk assessment based on frequency model:

Part II. risk analysis

L. Moretti, G. Cantisani, P. Di Mascio & S. Nichele

Dipartimento di Ingegneria Civile, Edile e Ambientale. Sapienza, University of Rome, Rome, Italy

S. Caro
Departamento de Ingenieria Civil y Ambiental, Universidad de Los Andes, Bogota, Colombia

ABSTRACT: This two-part paper presents a comprehensive airport risk assessment methodology for
aircraft runway veer-off accidents around the world. Veer-off is a lateral runway excursion, and it occurs
when an aircraft gets out the runway during a movement. The most frequent causes for this type of
accidents are inappropriate pilot performance and aircraft condition; while influential factors include
transversal wind, contamination of the runway, and poor visibility. The consequences of these events
involve both damage to the aircrafts and potential health effects on passengers and crew members. Statis-
tical data collected by the authors were used as reference for conducting frequency and risk assessments
of veer-offs. A successful probabilistic model of veer-off accidents was developed in the first part of the
paper. This second part assesses the level of safety guaranteed by the Cleared and Graded Area (CGA),
which is a part of the runway strip that enhances the deceleration of an aircraft in case of an excursion.
This area has to be dimensioned according to the requirements provided by the International Civil Avia-
tion Organization Standards (ICAO) and the regulations issued by the Italian Civil Aviation Authority.
Therefore, a model of veer-off accidents was developed to quantify the risk of an event. The results from
the risk analysis showed that there are some cases for which the Real Level of Safety (RLS) was higher
than the proposed Target Level of Safety (TLS), meaning that the level of damage expected is higher than
the proposed one. An application example of the model complying with the requirements of the ICAO
Regulation for non-instrument Code A runways is also presented. Overall, the results obtained from this
study show the proposed model contributes to the implementation of Safety Management Systems at
airports, as stipulated by the ICAO.

1 INTRODUCTION lateral runway surfaces (International Civil Avia-

tion Organization, 2013). Available specifications
A high-quality safety management design of airport vary according to the different designations of run-
infrastructures needs an accurate methodology to way (i.e. instrument or non-instrument runways).
evaluate the risk and prevent accidents (Attaccal- The methodology proposed in this study is a
ite et al. 2012, Wong et al. 2009a,b, Kirkland et al. useful tool to evaluate the veer-off risk of a spe-
2004, Di Mascio & Loprencipe in press). cific air-port (Moretti et al. in press) and, as such,
The initial portion of this paper focused on the it offers the airport management body the possibil-
development of a statistical model to assess the ity of implementing appropriate measurements to
frequency of aircraft veer-offs during landings achieve specific safety requirements.
and take-offs, including both instrument and non-
instrument or non-precision movements. Data avail- 2 METHODOLOGY
able since 1990 for commercial flights over 30 Mg
(ton) showed that the average frequency of a veer- The probability that an aircraft travels beyond
off accident was 1.44 in ten million movements. certain distance from the runway centreline was
Also, data collected by the authors suggested defined by combining the probability of veer-off
that veer-offs are more frequent to occur during a events and the probability of the lateral location
landing (probability dl equal to 73%) than during a distribution after the accident. Statistics show that
take-off (probability dto equal to 27%), as confirmed veer-offs are more frequent during non-instrument
by the Norwegian Civil Aviation Authority (2001). runway operations, although these movements
Currently, the reference code of runway plays a are less than 10% of all operations. Indeed, the
decisive role in the definition of the extension of probability of a veer-off in a non-instrument oper-

523

TIS2017_Book.indb 523 2/8/2017 8:07:00 PM

 ation, pnon-I, is 0.09, which results in a probability of
precision operation, pp, of 0.91.
D=
∑ event,j
,
(Nfatalities + Ninjuries )
⋅100 (5)
In order to represent more effectively the actual
frequencies of veer-off events, the authors divided
∑ event,j
,
Nooccupants
the veer-off probability distribution into four differ-
ent situations, taking into account landings and take- For each mechanical or physical available conse-
offs, both on instrument or non-instrument runways. quence, a statistical analysis was performed using
Using the statistical analysis of the data presented in the veer-off data presented in the first part of the
the first part of this article, the following distribu- paper. Table 2 shows the average frequencies for
tions were defined (Equation 1, 2, 3 and 4): damage D obtained.
Substantial health damages, on the other hand,
f 37 10 −7 ⋅ e −0,0219 x
11.37 (1) are related to the ramming through the airfield
perimeter fence (D = 63.1%), whereas minor dam-
f 9.72 ⋅10 −7 ⋅ e −0,0219 x (2) ages were observed to occur when the landing gear
f 6.96 ⋅10 −8 ⋅ e −0,0143 x (3) system crashes or when the airplane crashes into
f 6.82 ⋅10 −9 ⋅ e −0,0143 x (4) an embankment aside the runway (D = 5.5%); the
average value of D is equal to 18%.
where Eq. 1 is valid for landings on instrument run- Then, according to the levels of D, five catego-
ways, Eq. 2 is valid for landings on non-instrument ries of severity were defined (Table 3).
runways, Eq. 3 is valid for take-offs from instru-
ment runways, Eq. 4 is valid for take-offs from Table 2. Health effects, D, for each consequence.
non-instrument runways, and f is the frequency
Damage
of an aircraft running beyond a certain distance x
measured from the runway (RWY) centreline. Consequence D%
Six cumulative frequencies were defined by
calculating the expected average interval between Crash against an embankment 5.5
two subsequent events, and by associating the con- or against a drainage channel
sequences of an event with its frequency. Highly Landing gear damage 5.6
frequent events are associated with mechanical TWY, RWY, apron crossings 18.8
damages of the aircrafts, whereas unlikely events Mechanical damages 20.7
cause the aircraft to crash through the airfield Collision with obstacle(s) 22.7
perimeter fence, as observed in Table 1. Fire 36.6
On the other hand, the consequence of a veer- Crossing of airfield perimeter fence 63.1
off event with reference to the potential health
effects on people is defined through the coefficient Table 3. Categorization of severity.
D. This quantity corresponds to the ratio between
the number of fatalities and injuries and the total Health effects
number of occupants involved in the veered off air- on occupants
craft (Equation 5):
Severity Injury Fatality D % Consequence

1 Negligible No No 0 No damage
Table 1. Consequences based on the likelihood of 2 Minor Yes No 0<D Landing gear
occurrence of veer-off events. and damage;
D < 15 Crash
Probability Observations Consequences against
an obstacle
A Highly < 3 months Mechanical damages 3 Major Yes No 15 < D TWY, RWY,
frequent and Apron
B Frequent > 3 months Landing gear damage D < 25 crossings;
Collision with Collision
obstacle (s) with
C Occasional > 2 years Fire Crash against an obstacle (s)
embankment or 4 Hazardous Yes Yes 25 < D Fire
against a drainage and
channel D < 50
D Remote > 4 years RWY, taxiway (TWY) 5 Catastrophic Yes Yes 50 < D Crossing of
crossing and the airfield
E Improbable > 10 years Apron crossings D < 100 perimeter
F Extremely > 20 years Crossing of airfield fence;
improbable perimeter fence hull loss

524

TIS2017_Book.indb 524 2/8/2017 8:07:00 PM

 3 RESULTS The iso-risk lines for each consequence, listed in
Table 4, are represented in Figure 1. These results
Using the definition of the frequency interval and highlight the need for actions that could help
the categories of damage, it is possible to compute reducing the risk in certain conditions.
the level of risk as “the combination of the harm Combining the values in Table 4, it is possible to
probability category with the consequence of a spe- calculate the Real Level of Safety (RLS). This quan-
cific hazard being realized” (Airport Cooperative tity is obtained by combining the mean current veer-
Research Program, 2014), which is described by the off probability (1.44 ∙ 10−7) and the weighted sum of
elements of the risk assessment matrix (Equation 6): recorded damages for each consequence, according
to the following expression (Equation 8):
R ji p j ⋅ Di (6)

RLS pveer − offfff ⋅ Daverage = 1.44 ⋅10 −7 ⋅

∑D p
j event j
Thus, for each kind of potential consequence,
the cumulative frequency of a particular veer-off
∑p event j

is the product of the probability of veer-off events (8)

and the j damage of those veer-off events, as
expressed by Equation 7: For the purposes of this research, a Target Level
of Safety (TLS) equal to 1.26 ∙ 10−8 was assumed.
pj pveer − offff ⋅ pevent j 1 44 10 −7 pevent j (7) This value implies to have a veer-off with the mean
current veer-off probability and the damage D
where pevent,j derives from the data presented in the equal to half of the actual level of 9%. This value
first part of this paper. of damage is reasonably achievable with the appli-
Figure 1 shows the computation of the current cation of requirements for safety areas that are
risk levels of aircraft damage with respect to each both width and the bearing capacity of the soil
result and category. Each curve is a series of points located in the CGA. These requirements guaran-
of equal risk and it represents, for each damage tee a stopping performance of the CGA, as they
category, its level of risk. have been tested and validated for various weather
For each consequence and corresponding damage conditions (Australian Government Civil Aviation
level listed in Table 2, the authors calculated the actual Safety Authority, 2003).
level of risk. The results are presented in Table 4. According to the statistical data collected by the
authors, it was found that the RLS is higher than
TLS, as showed in Figure 2.
According to Figure 2, some safety actions
should be taken in airports in order to comply with
the mini-mum desired TSL. As for some of the
consequences (involving collision with obstacle(s)
and mechanical damages), the level of risk exceeds
the imposed threshold. In these cases, a proper
re-design of the runway strip is necessary to close
gaps and to accomplish the imposed requirements.
Available actions involve, among others, dimen-
sional and geotechnical features on CGA and on
the strip.

Figure 1. Iso-risk lines for each consequence.

Table 4. Level of risk for each consequence.

Consequence Actual level of risk

Crossing of airfield 0.273 ⋅ 10−9

perimeter fence
Crash against an embankment 0.840 ⋅ 10−9
or against a drainage channel
TWY, RWY, apron crossings 0.948 ⋅ 10−9
Landing gear damage 2.42 ⋅ 10−9
Fire 5.59 ⋅ 10−9
Collision with obstacle(s) 11.5 ⋅ 10−9
Mechanical damages 15.6 ⋅ 10−9
Figure 2. Comparison between RLS and TLS curves.

525

TIS2017_Book.indb 525 2/8/2017 8:07:01 PM

 RLS and TLS should be compared not only The ICAO regulation states that the CGA should
considering the consequences of veer-offs, but be “(…) constructed as to minimize hazards arising
also assessing these values for each runway code. from differences in load bearing capacity to aeroplanes
Reported results have shown that the problem of which the runway is intended to serve in the event of
veer-offs prevention has geometrical implications an aeroplane running off the runway” (International
to be considered. For each type of runway code Civil Aviation Organization 2013, 2006). Many
(RWcode) defined by the ICAO Annex 14 (Interna- national aeronautical authorities have interpreted
tional Civil Aviation Organization 2013), the risk of this statement by defining a minimum value of a
the main landing gear departing beyond the CGA bearing capacity index of the soil composing the
(Ente Nazionale per l’Aviazione Civile 2014) (RCGA_ CGA, and a maximum value of the subsidence of
GEAR) was calculated using equations (1), (2), (3) and the soil. For example, a minimum CBR (Californian
(4). The obtained results are listed in Table 5. Bearing Ratio) of 15% and a subsidence less than
The results show that the RLSCGA_GEAR 15 cm are currently required in Italy (Ente Nazi-
level of safety, which considers the geometrical onale per l’Aviazione Civile 2008). If these values
configuration of the CGA, is typically lower than are not achieved, the Italian Authority states that a
the pro-posed TLS, although for more than 25% specific veer-off risk assessment is necessary (Ente
of the runway runway codes the RLS is higher Nazionale per l’Aviazione Civile 2008). However, no
than TLS. indication about the allowable risk level is provided.

Table 5. Veer off risk of the main landing gear departing beyond the CGA with D = 18%.

RCGA_GEAR

Landings on precision Landings on non- Take-offs from Take-offs from non-

RWcode instrument runways instrument runways instrument runways instrument runways

1A 1.09E-08 9.58E-08 7.35E-09 8.31E-10

2A 1.09E-08 7.70E-08 7.35E-09 7.20E-10
3A 2.62E-09 3.58E-08 2.90E-09 4.37E-10
1B 1.10E-08 9.74E-08 7.42E-09 8.40E-10
2B 1.10E-08 7.83E-08 7.42E-09 7.28E-10
3B 2.66E-09 3.64E-08 2.93E-09 4.41E-10
1C 1.14E-08 1.01E-07 7.59E-09 8.58E-10
2C 1.14E-08 8.09E-08 7.59E-09 7.44E-10
3C 2.75E-09 3.76E-08 2.99E-09 4.51E-10
4C 2.75E-09 3.76E-08 2.99E-09 4.51E-10
3D 2.90E-09 3.97E-08 3.10E-09 4.67E-10
4D 2.90E-09 3.97E-08 3.10E-09 4.67E-10
4E 2.90E-09 3.97E-08 3.10E-09 4.67E-10

Table 6. Veer off risk of the main landing gear departing beyond the CGA with D = 9%.

RCGA_GEAR

Landings on precision Landings on Take-offs from Take-offs from

RWcode instrument runways non-instrument runways instrument runways non-instrument runways

1A 9,00E-09 7,94E-08 6,08E-09 6,88E-10

2A 9,00E-09 6,38E-08 6,08E-09 5,96E-10
3A 2,17E-09 2,96E-08 2,40E-09 3,62E-10
1B 9,14E-09 8,07E-08 6,15E-09 6,96E-10
2B 9,14E-09 6,48E-08 6,15E-09 6,03E-10
3B 2,20E-09 3,01E-08 2,43E-09 3,65E-10
1C 9,45E-09 8,34E-08 6,28E-09 7,11E-10
2C 9,45E-09 6,70E-08 6,28E-09 6,16E-10
3C 2,28E-09 3,11E-08 2,48E-09 3,73E-10
4C 2,28E-09 3,11E-08 2,48E-09 3,73E-10
3D 2,40E-09 3,29E-08 2,57E-09 3,87E-10
4D 2,40E-09 3,29E-08 2,57E-09 3,87E-10
4E 2,40E-09 3,29E-08 2,57E-09 3,87E-10

526

TIS2017_Book.indb 526 2/8/2017 8:07:02 PM

 Therefore, this study could be used to provide Many national aeronautical authorities demand
reference regarding allowable values of risk. For a risk assessment when the minimum value of the
ex-ample, the current average risk level might be bearing capacity of the soil composing the CGA
considered the maximum allowable risk, so that the is too low, but generally there is not any indication
values listed in Table 5 could be used as references. regarding allowable levels of risk. Within this con-
The authors also calculated the level of risk text, this study might be used to provide a refer-
RCGA_GEAR with a damage D equal to half of the cur- ence for allowable risk values. For example, current
rent damage (D = 9%). Table 6 shows the results. or target average risk values might be considered as
In this case, for some cases of RWcode the results the maximum allowable risk levels.
of RCGA_GEAR are higher than the selected TLS value.
are higher than the selected TLS value. Also, as ob-
served, the most critical results were obtained for
landings on non-instrument runways. These results REFERENCES
may depend on the geometrical characteristics of
the CGA, mostly for 1 and 2 non-instrument run- Airport Cooperative Research Program (ACRP) 2011.
way categories, whereas the width of the CGA of ACRP Report 51: Risk Assessment Method to Sup-
instrument runways can guarantee the required port Modification of Airfield Separation Standard.
Transport Research Board.
TLS. The enlargement of the CGA should not be Attaccalite, L., Di Mascio, P., Loprencipe, G. & Pan-
overlooked as a protection strategy. dolfi, C. 2012. Risk Assessment Around Airport. Pro-
cedia: Social & Behavioral Sciences, pp. 852–861, 53.
doi:10.1016/j.sbspro.2012.09.934.
4 CONCLUSIONS Australian Government Civil Aviation Safety Authority
2003. CASR 139 Manual of Standards.
This study has focused on the development of a Di Mascio, P. & Loprencipe, G. in press. Risk analysis
risk-based methodology to mitigate the severity of in the surrounding areas of one-runway airports: a
runway veer-offs. It involved the analysis of selected methodology to preliminary calculus of PSZs dimen-
sions. ARPN Journal of Engineering and Applied Sci-
events that occurred between 1953 and 2015 through- ences (ISSN 1819-6608).
out the world. Collected data included frequency of Ente Nazionale per l’Aviazione Civile (ENAC) 2008.
the events and causes and consequences of these Circolare 07/02/2008 Linee guida per l’adeguamento
accidents. The proposed risk analysis considered delle strip aeroportuali.
commercial flights over 30 Mg (ton), and it demon- Ente Nazionale per l’Aviazione Civile (ENAC) 2014.
strated that the veer-off accident requires a rigorous Regolamento per la costruzione ed esercizio degli
and thorough evaluation of the current operative aeroporti.
conditions to mitigate these type of accidents. International Civil Aviation Organization 2006. Aero-
The statistical analysis of the available data drome Design Manual – part I.
International Civil Aviation Organization 2013.
showed that the current veer-off probability is equal Annex 14 – Volume I Aerodrome Design and
to 1.44 ∙ 10−7 and that the damage (defined as the Construction.
ratio between the number of fatalities and injuries Kirkland, I.D.L., Caves, R.E. & Humphreys, I.M. 2004.
that occur to the occupants of the veering off aircraft An improved methodology for assessing risk in air-
with respect to the total number of occupants) is craft operations at airports, applied to runway over-
equal to 18%. Consequently, the Real Level of Safety runs. Safety Science 42, pp. 891–905.
(RLS) was calculated to be equal to 2.6 ∙ 10−8. Moretti, L., Cantisani, G. & Caro, S. in press. Air-
Using the statistical analysis of the data pre- port veer-off risk assessment: an Italian case study.
sented in the first part of this article, the frequency ARPN Journal of Engineering and Applied Sciences.
(1819–6608).
curves of an aircraft running beyond a certain Norwegian Civil Aviation Authority 2001. Final report
distance x measured from the runway centreline on the risk analysis in support of aerodrome design
were computed, and the level of risk at different rules. Rules, AEAT/RAIR/RD02325/R/002 Issue 1,
distance from the runway was determined. For 2001.
each runway reference code, the level of risk was Wong, D.K.Y., Pitfield, D.E., Caves, R.E. & Apple-
computed considering both the RLS and the Tar- yard, A. J. 2009a). The development of a more risk-
get Level of Safety, TLS. In the analysis, the value sensitive and flexible airport safety area strategy: Part
of TLS was assumed to be equal to half of the cur- I. The development of an improved accident frequency
rent damage (i.e. for the damage D = 9%, the TLS model. Safety Science 47, pp. 903–912. doi:10.1016/j.
ssci.2008.09.010.
was equal to 1.26·10−8). The most critical results Wong, D.K.Y., Pitfield, D.E., Caves, R.E. & Appleyard,
from these results were found for landings on non- A. J. 2009b). The development of a more risk-sensitive
instrumented runways, where the risk of the main and flexible airport safety area strategy: Part II. Acci-
landing gear departing beyond the CGA presented dent location analysis and airport risk assessment case
the higher values (RCGA_GEAR equal to 1.01·10−7 dur- studies. Safety Science 47, pp. 913–924. doi:10.1016/j.
ing landings on non-instrument 1C runways). ssci.2008.09.011.

527

TIS2017_Book.indb 527 2/8/2017 8:07:02 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Risk factors that increased accident severity at US railroad crossings

from 2005 to 2015

E. Dabbour
Abu Dhabi University, Abu Dhabi, United Arab Emirates

ABSTRACT: This study identifies and quantifies the effects of different risk factors that increase acci-
dent severity at railroad-highway grade crossings. The research is based on utilizing binary logit to analyze
all vehicle-train accidents that occurred at railroad-highway grade crossings in the United States from
2005 to 2015. The study investigates the temporal stability of the identified risk factors throughout the
analysis period to identify the most significant risk factors that are temporally stable. Age and gender of
the driver were found to be the most-significant temporally-stable risk factors identified. It was also found
that darkness and adverse weather conditions may reduce accident severity, but they were not temporally
stable throughout the analysis period. Accidents related to young drivers were found to be less severe. The
findings of this research have the potential to help decision makers develop policies and countermeasures
that reduce the severity of injuries at railroad-highway grade crossings by focusing on risk factors that
consistently exhibit significant effects on the severity of accidents.

1 INTRODUCTION of drivers’ injuries at railroad crossings in USA

where they found that female and older drivers
Accidents at railroad-highway grade crossings were more likely to be involved in severe accidents
(or simply railroad crossings) are usually among as compared to male and younger drivers, respec-
the most severe accidents due to the large ratio tively. Eluru et al. (2012) developed a latent class
between the weight of a typical train and that of model to identify risk factors that may increase
a typical motor vehicle. That ratio is estimated to the severity of driver injuries at highway–railway
be approximately 4,000 to 1 (Hao & Daniel 2016; crossings by using accident data from 14,532 cross-
Yan et al. 2010). According to data obtained from ings in USA for the period from 1997 to 2006. The
the Federal Railway Administration (FRA 2016), factors that were found to significantly influence
there was a total of 25,556 reported accidents at driver’s injury severity include driver’s age, time
railroad crossings in the United States during of the accident, and weather conditions. Hu et al.
the period from 2005 to 20015. Those accidents (2010) used logit modeling to identify risk factors
included 9,727 serious accidents that resulted in that increase the severity of accidents at railroad
3,204 fatalities and 10,685 injuries. crossings using data from Taiwan for the period
Several research studies attempted to identify from 1995 to 1997. Variables such as daily number
risk factors that increase the frequency of acci- of trains, daily number of trucks, and the use of
dents at railroad crossings so that countermeas- obstacle-detection devices were found to influence
ures may be implemented to reduce the number the severity of accidents. McCollister and Pflaum
of those accidents (Saccomanno et al. 2007; Oh (2007) used binary logit to develop models that
et al. 2006; Saccomanno et al. 2004; Austin and predict the probability of accidents, injuries, and
Carson 2002). Other studies focused on analyzing fatalities at railroad crossings. The risk factors that
risk factors that increase accident severity at rail- were found to increase injury severity include train
road crossings, given that an accident has already speed, number of trains, traffic volume, and acci-
occurred. Hao et al. (2016) used ordered probit dent history. The proportion of heavy vehicles in
modeling to analyze all accidents that occurred at the traffic volume and the proximity of traffic con-
railroad crossing in USA during the period from trol devices were also found to impact the severity
2002 to 2011 where they found that injury severity of accidents.
increases during AM and PM peak hours as well It must be noted that most of the previous studies
as during PM off-peak period. Hao and Daniel that investigated the factors influencing the severity
(2013) used ordered probit modeling to determine of drivers’ injuries at railroad crossings were based on
the significant factors influencing the severity compiling data from different years into one dataset

529

TIS2017_Book.indb 529 2/8/2017 8:07:04 PM

 that was used for the analysis. This approach disre- 3 METHODOLOGY
gards the fact that the factors identified in those stud-
ies might be temporally unstable so that they might The most common modeling methods used in
have been significant during the period being ana- investigating the factors influencing injury severity
lyzed and then their significance might change. The in traffic accidents include:
change in the significance over time might be due to
a. Binary logit modeling (e.g. Wang et al. 2016; Ye
the ongoing changes in vehicle and train technologies,
et al. 2016; Wong and Chung 2008; Chang and
drivers’ attitudes, traffic volumes, road conditions,
Yeh 2006);
conditions of railroad crossings, and policy imple-
b. Ordered logit, or probit, modeling (e.g. Hao
mentation. Understanding the temporal trends of the
et al. 2016; Hao and Daniel 2014; Haleem and
factors affecting injury severity would help researchers
Abdel-Aty 2010; Gray et al. 2008); and
assessing the effectiveness of implementing different
c. Mixed logit modeling (e.g. Naik et al. 2016; Wu
safety treatments so that they could identify whether
et al. 2016; Behnood & Mannering 2015; Kim
any safety improvements are attributed to the specific
et al. 2013).
treatment or simply attributed to the temporal insta-
bility of the factors being addressed. Therefore, the In binary logit modeling, the response variable
purpose of this research study is to investigate tem- must be dichotomous (e.g., injury accidents vs. no-
poral stability of the factors affecting the severity of injury accidents). Ordered logit, and probit, mod-
drivers’ injuries resulting from accidents at railroad eling allows for ordinal response variables (e.g.,
crossings. The study is based on using binary logit different severity levels of accidents), but it is based
modeling to analyze all accidents that occurred at on the assumption that the impacts of contributing
railroad crossings in USA for the period from 2005 to factors on injury severities are consistent, or pro-
2015. A separate model is developed for each analysis portional, across all observation in all the ordered
year, and the models are compared together to deter- classes. This assumption may not reflect the het-
mine whether the effects of the identified factors are erogeneity nature found in the data being analyzed.
temporally stable throughout the analysis period that Mixed logit models have the flexibility to account
extends for 11 years from 2005 to 2015. The findings for individual heterogeneity of the impacts of differ-
of this analysis have the potential to assist decision ent factors on injury severities by allowing param-
makers identifying the most significant risk factors eters to follow a distribution and to randomly vary
that are temporally stable so that different resources across observations. However, they may disregard
may be allocated to reduce the impacts of those risk the ordinal nature of the outcome variable.
factors. This research study also provides more-recent Based on the above brief discussion, binary logit
quantitative measures that may be used to determine modeling was selected as the tool utilized in this
the feasibility of implementing different counter- research study to analyze accident data. Binary
measures in reducing the severity of accidents at rail- logit is a generalized linear model that predicts the
road crossings. probability of occurrence of an event by fitting
data to a logit function in the form of (McCullagh
and Nelder, 1989):
2 DATA COLLECTION
f(z) = ez/(1 + ez) (1)
The dataset used for the research is obtained from
the Federal Railway Administration (FRA 2016) where z is the logit function, which is a measure of
and it covers all accidents that occurred at railroad the total contribution of all the explanatory vari-
crossings in the United States for eleven years start- ables (risk factors) used in the model; and ƒ(z) is
ing from 2005 until 2015. Incidents related to sui- a dichotomous variable that is assumed to follow
cide, or attempted suicide, were excluded from the Bernoulli distribution. It represents the probability
analysis since they are actually deliberate actions of an accident to be serious (i.e. resulted in either
and not accidents. Using accident records that a fatality or an injury), given that the accident
extend for eleven years provides the benefits of has already occurred. Based on that, the variable
capturing year-to-year temporal changes over that ƒ(z) takes the value of “1” if the accident is seri-
period so that countermeasures may be developed ous, or “0” if the accident is minor (i.e. resulted in
to reduce the impacts of risk factors that are found property-damage only with no injuries). Selecting
to be temporally stable. During the period from this binary structure for the outcome variable (i.e.
2005 to 2015, there was a total of 25,556 reported modeling the accident as either severe or minor)
accidents at railroad crossings, from which 23,767 has the potential to overcome the inconsistency
accidents were between trains and motor vehicles. in reporting accident severity. For example, an
The remaining 1,789 accidents were between trains injury accident in the United States is defined as
and either pedestrians or non-motorized vehicles. an accidents that resulted in an injury for a person

530

TIS2017_Book.indb 530 2/8/2017 8:07:04 PM

 (e.g. vehicle driver) that required medical treatment A Wald test is used to test the statistical signifi-
(FRA 2016). In Australia, according to the Austral- cance of each of the estimated coefficients (β1, β2,
ian Transport Safety Bureau (ATSB 2012), an injury β3, ..., βk). The Wald statistic of asymptotic chi-
accident is defined as an accident that resulted in square distribution is the squared value of the Z
an injury that required hospitalization for less than statistic and is computed as:
30 days. Using binary logit modeling will overcome
the inconsistency in the thresholds used to define Wald [[βˆi / ( βˆi )]4 (3)
injury accidents by simply classifying an accident to
either severe (that resulted in death or any injury) or
where βˆi is the ith estimated coefficient and SE
minor (that did not result in any injury).
(βˆi ) is the standard error of that coefficient. In this
The logit function has the following form:
study, the Odds Ratio (OR) is used to interpret the
significance of different explanatory variables (risk
z = β0 + β1x1 + β2x2 + β3x3 + ⋅⋅⋅ + βkxk (2)
factors) where an estimate of the odds ratio of a
certain explanatory variable is exp (βˆi) while hold-
where β0 is the intercept and (β1, β2, β3, ..., βk) are the
ing the other explanatory variables unchanged.
regression coefficients of the explanatory variables
Finally, the 95% confidence interval (CI) is also
(x1, x2, x3, ..., xk), respectively. The explanatory vari-
used in this study to describe the upper and lower
ables (risk factors) that were investigated include:
limits of the odds ratio with 95% confidence level
d. Heavy vehicle (HV): takes a value of “1” if the and is given by | βˆi Z 00.95
.95 ( βˆi ) |.
vehicle that collided with the train was a heavy
vehicle (i.e. a truck or a bus), and a value of “0”
otherwise; 4 RESULTS AND DISCUSSION
e. Freezing air temperature (FR): takes a value
of “1” if the air temperature at the time of the 4.1 Descriptive statistics
accident was below the freezing point (0°C or
The distribution of all accidents between trains
32°F), and a value of “0” otherwise;
and motor vehicles at railroad crossings, for the
f. Darkness (DA): takes a value of “1” if the cross-
period from 2005 to 2015, is shown in Table 1. For
ing was not lighted at the time of the accident
the purpose of this study, accidents are classified
(by either daylight or illumination), and a value
as either minor (no injury) accidents or serious
of “0” otherwise;
(injury or fatal) accidents. As shown in the table,
g. Adverse weather (AW): takes a value of “1”
there was a total of 23,767 accidents between
if the weather at the time of the accident was
trains and motorized vehicles during the analysis
adverse (i.e. rain, snow, sleet, or fog), and a
period. They include 15,462 minor accidents, 6,394
value of “0” otherwise;
injury accidents, and 1,911 fatal accidents. The
h. Obstructed view (OB): takes a value of “1” if
mean and standard deviation values of all catego-
the view of the vehicle driver was obstructed
ries are also shown in the table. It is noted that the
(by a natural object, a built object, or another
percentage of serious accidents during that period,
train on an adjacent track), and a value of “0”
as compared to all accidents, is 34.94% with an
otherwise;
average mortality rate of 0.1348 deaths per acci-
i. Female driver (FE): takes a value of “1” if the
dent. When comparing this mortality rate with the
vehicle driver was female, and a value of “0”
overall mortality rates for all vehicular accidents at
otherwise;
highway intersections in USA, as obtained from
j. Younger driver (YO): takes a value of “1” if the
the National Highway Traffic Safety Administra-
age of the vehicle driver was below the 21 years,
tion (2016), it was found that the mortality rate for
and a value of “0” otherwise; and
vehicle-train accidents is significantly higher than
k. Older driver (OL): takes a value of “1” if the age
that for vehicle-vehicle accidents. This is due to the
of the vehicle driver was above 65 years, and a
significantly large weight of a typical train as com-
value of “0” otherwise.
pared to the weight of a typical motor vehicle (Hao
The speed of the train (for accidents when a and Daniel 2016; Yan et al. 2010).
train hits a vehicle) and the speed of the vehicle Table 1 also shows a significant decrease in the
(for accidents when a vehicle hits a train) were not number of total accidents between the period from
investigated since they are obvious risk factors that 2005 to 2008 and the period from 2009 onward. By
have been investigated in an enormous number of applying statistical significance two-tail t-test for
previous research studies (e.g. Hao et al. 2016; Hal- the difference between the two means, at 5% sig-
eem and Gan 2015; Liu et al. 2015; Hao & Daniel nificance level, the t-stat is found to be 6.83 [df = 9
2014; Fan & Haile 2014; Russo & Savolainen 2013; and p < 0.001]. The decrease in the number of
Eluru et al. 2012; Miranda-Moreno et al. 2009). serious accidents was also highly significant with

531

TIS2017_Book.indb 531 2/8/2017 8:07:04 PM

 Table 1. Number (and percentages) of accidents at railroad crossings in USA from 2005 to 2015(a).

Serious Accidents

Year No Injuries Injury Fatal Total Total Accidents Total Injuries(b)

2005 1,966 (67.51%) 710 (24.39%) 236 (8.10%) 946 (32.49%) 2,912 (100%) 1,293
2006 1,892 (67.43%) 666 (23.73%) 248 (8.84%) 914 (32.57%) 2,806 (100%) 1,343
2007 1,755 (66.73%) 658 (25.02%) 217 (8.25%) 875 (33.27%) 2,630 (100%) 1,292
2008 1,483 (65.53%) 608 (26.87%) 172 (7.60%) 780 (34.47%) 2,263 (100%) 1,148
2009 1,184 (66.37%) 454 (25.45%) 146 (8.18%) 600 (33.63%) 1,784 (100%) 880
2010 1,164 (62.18%) 559 (29.86%) 149 (7.96%) 708 (37.82%) 1,872 (100%) 991
2011 1,181 (62.79%) 549 (29.18%) 151 (8.03%) 700 (37.21%) 1,881 (100%) 1,166
2012 1,119 (61.99%) 534 (29.59%) 152 (8.42%) 686 (38.01%) 1,805 (100%) 1,095
2013 1,172 (61.91%) 573 (30.27%) 148 (7.82%) 721 (38.09%) 1,893 (100%) 1,075
2014 1,362 (65.89%) 556 (26.90%) 149 (7.21%) 705 (34.11%) 2,067 (100%) 975
2015 1,184 (63.86%) 527 (28.43%) 143 (7.71%) 670 (36.14%) 1,854 (100%) 1,106
Total(c) 15,462 6,394 1,911 8,305 23,767 12,364
Mean 1,406 581 174 755 2,161 1,124
SD(d) 305.16 70.05 37.97 105.18 406.32 138.20
(a)
For accidents between trains and motor vehicles after excluding suicide/attempted suicide.
(b)
Including fatalities.
(c)
For 11 years (2005 to 2015).
(d)
Standard deviation for the population.

t-stat 5.84 [df = 9 and p < 0.001]. Furthermore, the highway accidents (Teftt, 2008; Williams and
decrease in the total number of injuries and fatali- Shabanova, 2003).
ties was significant with t-stat value of 3.91 [df = 9 Another significant risk factor is associated
and p = 0.004]. with female drivers (FE). This risk factor is sig-
nificant in 9 years out of the 11-year analysis
period. More important, it has consistently been
4.2 Binary logit models significant since 2011 onward. This finding is also
The binary logit coefficients (and standard consistent with other previous research studies
errors) of the risk factors that were investigated (Haleem and Gan 2015; Zhao & Khattak 2015;
are shown in Table 2. The corresponding odds Hao and Daniel 2013; Eluru et al. 2012). This
ratios, along with their 95% confidence intervals, finding may be attributed to one of the following
are shown in Table 3. In both tables, risk factors two explanations:
that are statistically significant (with p < 0.05) are
a. The differences between male and female driv-
shown in bold. As shown in the two tables, the
ers in terms of body size and structure; or
most significant risk factor that is temporally sta-
b. Possible cognitive differences between male and
ble throughout the entire analysis period is when
female drivers in taking last-second corrective
the driver is over the age of 65 years (OL). This
measures to reduce the severity of an accident
finding may be explained by the possible deterio-
(given that the accident is unavoidable).
ration in health conditions of older drivers, which
make them more susceptible to injuries. Further- To test the two hypotheses shown above, a
more, older driver may also have slower reaction binary logit analysis was conducted on a sample
times and therefore may be slower than other driv- data of all drivers involved in single-vehicle colli-
ers in reacting to emergency situations by taking sions in North Carolina between 2007 and 2013,
the necessary actions needed to reduce the severity where the outcome was whether the collision led
of an accident, such as steering, accelerating, or to rollover (Dabbour 2012). Based on that analy-
braking. This finding is consistent with the find- sis, it was found that driver gender was not a sig-
ings of other research studies that found that older nificant factor with p-value found to be 0.127.
drivers are more likely to be inured in accidents at Since avoiding rollover in a single-vehicle collision
railroad crossings (Zhao & Khattak 2015; Haleem usually requires last-second corrective measures, it
and Gan 2015; Hao and Daniel 2013; Eluru et al. can be concluded that female drivers do not lack
2012). This finding is also consistent with other those last-second corrective measures; and there-
research studies where older drivers were found fore, the increased probability for female drivers to
to be more likely to be injured in vehicle-vehicle be injured in an accident at railroad crossing may

532

TIS2017_Book.indb 532 2/8/2017 8:07:05 PM

 Table 2. Binary logit coefficients (and standard errors) (a).

Year HV FR DA AW OB FE YO OL

2005 0.078 −0.281 −0.275 −0.539 −0.002 0.180 −0.412 0.429

(0.079) (0.121) (0.097) (0.141) (0.161) (0.094) (0.093) (0.137)
2006 −0.027 −0.112 −0.394 −0.253 −0.125 0.333 −0.512 0.777
(0.081) (0.151) (0.099) (0.137) (0.196) (0.096) (0.098) (0.135)
2007 0.007 −0.201 −0.279 −0.265 −0.040 0.386 −0.619 0.670
(0.083) (0.124) (0.103) (0.140) (0.181) (0.095) (0.104) (0.140)
2008 −0.082 −0.249 −0.226 −0.083 −0.342 0.305 −0.628 0.537
(0.089) (0.129) (0.106) (0.138) (0.193) (0.103) (0.112) (0.148)
2009 0.167 −0.048 −0.393 −0.131 −0.408 0.268 −0.891 0.416
(0.100) (0.138) (0.126) (0.159) (0.222) (0.115) (0.133) (0.161)
2010 0.201 −0.191 −0.632 −0.567 0.097 0.175 −0.711 0.371
(0.097) (0.133) (0.120) (0.158) (0.209) (0.111) (0.127) (0.147)
2011 −0.034 −0.180 −0.387 −0.677 −0.039 0.345 −0.970 0.892
(0.095) (0.137) (0.114) (0.166) (0.204) (0.111) (0.129) (0.154)
2012 0.037 −0.154 −0.369 −0.440 −0.772 0.471 −0.704 0.444
(0.097) (0.182) (0.112) (0.179) (0.266) (0.111) (0.131) (0.154)
2013 −0.091 −0.132 −0.173 −0.100 −0.067 0.317 −1.011 0.410
(0.095) (0.133) (0.104) (0.149) (0.202) (0.107) (0.138) (0.147)
2014 −0.072 0.011 −0.231 −0.284 −0.114 0.412 −0.929 0.489
(0.093) (0.121) (0.102) (0.142) (0.197) (0.104) (0.133) (0.142)
2015 −0.102 −0.243 −0.018 −0.452 −0.462 0.285 −0.857 0.562
(0.097) (0.148) (0.106) (0.157) (0.230) (0.111) (0.130) (0.147)
(a)
Numbers in bold indicate significant risk factors.

Table 3. Odds ratios (with 95% lower and upper Confidence Intervals) (a).

Year HV FR DA AW OB FE YO OL

2005 1.08 0.76 0.76 0.58 1.00 1.20 0.66 1.54

(0.93–1.26) (0.60–0.96) (0.63–0.92) (0.44–0.77) (0.73–1.37) (0.99–1.44) (0.55–0.79) (1.17–2.01)
2006 0.97 0.89 0.67 0.78 0.88 1.39 0.60 2.18
(0.83–1.14) (0.67–1.20) (0.56–0.82) (0.59–1.02) (0.60–1.29) (1.16–1.68) (0.49–0.73) (1.67–2.83)
2007 1.01 0.82 0.76 0.77 0.96 1.47 0.54 1.95
(0.85–1.19) (0.64–1.04) (0.62–0.93) (0.58–1.01) (0.67–1.37) (1.22–1.77) (0.44–0.66) (1.48–2.57)
2008 0.92 0.78 0.80 0.92 0.71 1.36 0.53 1.71
(0.77–1.10) (0.61–1.01) (0.65–0.98) (0.70–1.21) (0.49–1.04) (1.11–1.66) (0.43–0.66) (1.28–2.29)
2009 1.18 0.95 0.68 0.88 0.67 1.31 0.41 1.52
(0.97–1.44) (0.73–1.25) (0.53–0.86) (0.64–1.20) (0.43–1.03) (1.04–1.64) (0.32–0.53) (1.11–2.08)
2010 1.22 0.83 0.53 0.57 1.10 1.19 0.49 1.45
(1.01–1.48) (0.64–1.07) (0.42–0.67) (0.42–0.77) (0.73–1.66) (0.96–1.48) (0.38–0.63) (1.09–1.93)
2011 0.97 0.84 0.68 0.51 0.96 1.41 0.38 2.44
(0.80–1.17) (0.64–1.09) (0.54–0.85) (0.37–0.70) (0.64–1.44) (1.14–1.75) (0.29–0.49) (1.80–3.30)
2012 1.04 0.86 0.69 0.64 0.46 1.60 0.49 1.56
(0.86–1.26) (0.60–1.23) (0.56–0.86) (0.45–0.91) (0.27–0.78) (1.29–1.99) (0.38–0.64) (1.15–2.11)
2013 0.91 0.88 0.84 0.91 0.94 1.37 0.36 1.51
(0.76–1.10) (0.68–1.14) (0.69–1.03) (0.68–1.21) (0.63–1.39) 1.11–1.69 (0.28–0.48) (1.13–2.01)
2014 0.93 1.01 0.79 0.75 0.89 1.51 0.39 1.63
(0.78–1.12) (0.80–1.28) (0.65–0.97) (0.57–0.99) (0.61–1.31) (1.23–1.85) (0.30–0.51) (1.24–2.15)
2015 0.90 0.78 0.98 0.64 0.63 1.33 0.42 1.75
(0.75–1.09) (0.59–1.05) (0.80–1.21) (0.47–0.87) (0.40–0.99) (1.07–1.65) (0.33–0.55) (1.31-2.34)
(a)
Numbers in bold indicate significant risk factors.

be mainly attributed to the differences between explanation provided in another research study
male and female drivers in terms of body size and (Eluru et al. 2012) where it was concluded that the
structure. This explanation is consistent with the higher physiological strength of a male driver (as

533

TIS2017_Book.indb 533 2/8/2017 8:07:05 PM

 compared to a female driver) might result in a less be injured in vehicle-train accidents at railroad
severe injury for male drivers. crossings. This factor is temporally stable through-
Darkness conditions at railroad crossings (DA) out the analysis period. This finding is consistent
were found to reduce the severity of collisions at with the findings of previous research studies that
railroad crossings in 9 years out of the 11-year analyzed accident severity at either railroad cross-
analysis period. However, the effect was not tem- ings or highway intersections (Haleem and Gan
porally stable in the most recent years since it was 2015; Eluru et al. 2012; Wong & Chung 2008). No
not significant in years 2013 and 2015. A possible temporally-stable significance was found for the
explanation for this finding is that drivers are usu- effects of freezing temperature (FE), heavy vehi-
ally more alert in darkness conditions and there- cles (HV), and obstructed view of the driver (OB).
fore they may take immediate evasive measures Therefore, it may be concluded that those factors
to reduce the severity of an accident with a train have no temporally-stable correlation with the
given that such an accident was unavoidable. This severity of accidents at railroad crossings.
finding is consistent with the findings of other
research studies (Yau 2004; Krull et al. 2000)
where it was found that darkness conditions may 5 CONCLUSIONS
reduce the severity of traffic accidents at highway
intersections. However, previous research studies In this paper, binary logit modeling was used to
found that darkness conditions may contribute to investigate the temporal stability of the factors
increased severity of accidents at railroad crossings that increase accident severity at railroad-highway
(Hao et al. 2016; Zhao & Khattak 2015; Hao & grade crossings (or simply railroad crossings). This
Daniel 2014). Other research studies found that objective was achieved by analyzing all accidents
darkness conditions may also increase the severity that occurred in USA at railroad crossings during
of vehicle-vehicle accidents at highway intersec- the period from 2005 to 2015. It was found that the
tions (Kim et al. 2013; Zhang et al. 2011; Abdel- most temporally-stable factor was the old age of
Aty 2003). In a research study by Plainis and vehicle driver (if the age of the driver is above 65
Murray (2002) it was suggested that the reaction years). This may be explained by the possible dete-
time for drivers generally increases when driving at rioration in health conditions, which make older
low luminance levels; and therefore those drivers drivers more susceptible to injuries. Furthermore,
would have less time to take any evasive measures older drivers may also have slower reaction times
to reduce the severity of accidents. Further research and therefore may be slower than other drivers in
may be needed to clarify those conflicting findings reacting to emergency situations by taking the nec-
regarding the effect of darkness conditions on the essary actions needed to reduce the severity of an
severity of accidents. accident, such as steering, accelerating, or braking.
Adverse weather conditions (AW) were found Female drivers were also found to be more likely
to be a significant factor in reducing the severity to be injured in accidents at railroad crossings,
of collisions at railroad crossings in only 6 years which is mainly attributed to their smaller body
out of the 11-year analysis period; and therefore size and more fragile body structure.
this factor may be considered as temporally unsta- Darkness and adverse weather conditions were
ble. However, several research studies found that both found to reduce the probability of being
adverse weather conditions may reduce the sever- injured in an accident at a railroad crossing, which
ity of accidents, either at railroad crossings or at may be explained by the extra attention paid by
highway intersections (Behnood and Mannering, drivers under those conditions so that drivers
2015; Haleem and Gan 2015, Fan & Haile 2014; may take the immediate actions needed to reduce
Behnood et al., 2014; Krull et al., 2000). A possible the severity of an accident at a railroad crossing.
explanation is that drivers are usually more alert However, both factors were not temporally stable
during adverse weather conditions and therefore since darkness was significant in only 9 years out
they would travel at lower speeds and would also of the 11-year study period, and adverse weather
be ready to take immediate evasive measures to was significant in only 6 years out of the 11-year
reduce the severity of an accident with a train given study period. Furthermore, the findings of previ-
that such an accident was unavoidable. However, ous research studies regarding these conditions
few other studies suggested that adverse weather are mixed since some research studies confirm the
conditions may actually increase the severity of findings provided in this research study while other
accidents at railroad crossings (Hao et al. 2016; previous research studies suggest that darkness and
Hao and Daniel 2014). Further research may be adverse weather conditions actually increase acci-
needed to clarify those conflicting findings. dent severity at railroad crossings. Further research
An interesting finding is that young drivers may be needed to investigate the effects of these two
(YO) were found to be consistently less likely to factors on accident severity at railroad crossings.

534

TIS2017_Book.indb 534 2/8/2017 8:07:05 PM

 It was also found that drivers below the age of logit modeling. Proceedings of 93rd Annual Meeting,
21 years are less likely to be injured in accidents at Transportation Research Board, Washington, D.C.
railroad crossings. Finally, it was found that freez- Federal Railway Administration (2016). Highway-rail acci-
ing temperature, heavy vehicles, and obstructed dents statistical data. Federal Railway Administration,
Office of Safety Analysis. Available online from https://
view of the driver are not significant factors in safetydata.fra.dot.gov/OfficeofSafety/publicsite/query/
increasing accident severity at railroad crossings. query.aspx (date accessed September 23, 2016).
It must be noted that train and vehicle speeds Gray, R.; Quddus, M.A.; and Evans, A. (2008). Injury
were not investigated since these two risk factors severity analysis of accidents involving young male
have been previously investigated in an enormous drivers in Great Britain. Journal of Safety Research
number of research studies and they were found to 39: 483–495.
be obvious risk factors that significantly increase Haleem, K. & Abdel-Aty, M. (2010). Examining traf-
accident severity at railroad crossings. fic crash injury severity at unsignalized intersections.
Decision makers may use the findings of this Journal of Safety Research 41: 347–357.
Haleem, K. & Gan, A. (2015). Contributing factors of crash
research study to identify the more temporally- injury severity at public highway-railroad grade cross-
stable significant risk factors that increase the ings in the US. Journal of Safety Research 53: 23–29.
severity of accidents at railroad crossings so that Hao, W. & Daniel, J. (2016). Driver injury severity related
they can allocate the resources needed to reduce to inclement weather at highway–rail grade crossings
the effects of those risk factors. Furthermore, the in the United States. Traffic Injury Prevention 17(1):
binary logit models developed in this paper provide 31–38.
more-recent quantitative measures that may be Hao, W. & Daniel, J. (2014). Motor vehicle driver injury
used to determine the feasibility of implementing severity study under various traffic control at highway-
different countermeasures in reducing the severity rail grade crossings in the United States. Journal of
Safety Research 51: 41–48.
of accidents at railroad crossings. Hao, W., & Daniel, J. (2013). Severity of injuries to motor
vehicle drivers at highway-rail grade crossings in the
United States. Transportation Research Record: Journal
of the Transportation Research Board 2384: 102–108.
REFERENCES Hao, W., Kamga, C., and Wan, D. (2016). The effect of
time of day on driver’s injury severity at highway-rail
Abdel-Aty, M. (2003). Analysis of driver injury severity grade crossings in the United States. Journal of Traffic
levels at multiple locations using ordered probit mod- and Transportation Engineering 3(1): 37–50.
els. Journal of Safety Research 34: 597–603. Hu, S.R., C.S. Li, and C.K. Lee (2010). Investigation of
Austin, R.D., and J.L. Carson (2002). An Alternative Key Factors for Accident Severity at Railroad Grade
Accident Prediction Model for Highway–Rail Inter- Crossings by Using a Logit Model. Safety Science
faces. Accident Analysis and Prevention 34(1): 31–42. 48(2): 186–194.
Australia Transport Safety Bureau (2012). Australian Kim, J., Ulfarsson, G., Kim, S., Shankar, V. (2013).
Rail Safety Occurrence Data: 1 July 2002 to 30 June Driver-injury severity in single-vehicle crashes in Cali-
2012. Final Report RR-2012-010. Australia Transport fornia: A mixed logit analysis of heterogeneity due to
Safety Bureau, Canberra, Australia. age and gender. Accident Analysis and Prevention 50:
Behnood, A. & Manne ring, F. (2015). The temporal 1073–1081.
stability of factors affecting driver-injury severities in Krull, K., Khattak, A., Council, F. (2000). Injury Effects
single-vehicle crashes: Some empirical evidence. Ana- of Rollovers and Events Sequence in Single-Vehicle
lytic Methods in Accident Research 8: 7–32. Crashes. Proceedings of the 80th Annual Meeting,
Behnood, A.; Roshandeh, A.; and Mannering, F. (2014). Transportation Research Board, Washington, D.C.
Latent class analysis of the effects of age, gender, and Liu, J., Khattak, A., Richards, S., and Nambisan, S.
alcohol consumption on driver-injury severities. Ana- (2015). What are the differences in driver injury out-
lytic Methods in Accident Research 3–4: 56–91. comes at highway-rail grade crossings? Untangling
Chang, H. & Yeh, T. (2006). Risk Factors to Driver Fatal- the role of pre-crash behaviors. Accident Analysis and
ities in Single-Vehicle Crashes: Comparisons between Prevention 85: 157–169.
Non-Motorcycle Drivers and Motorcyclists. Journal Miranda-Moreno, L.F., L. Fu, S.V. Ukkusuri, and D.
of Transportation Engineering 13(3): 227–236. Lord (2009). How to Incorporate Accident Severity
Dabbour, E. (2012). Using Logistic Regression to Iden- and Vehicle Occupancy into the Hot Spot Identifica-
tify Risk Factors Causing Rollover Collisions. Inter- tion Process? Transportation Research Record: Journal
national Journal for Traffic and Transport Engineering of the Transportation Research Board 2102: 53–60.
2(4): 372–379. McCollister, G.M., and C.C. Pflaum (2007). A Model to
Eluru, N., M. Bagheri, L.F. Miranda-Moreno, and L. Predict the Probability of Highway Rail Crossing Acci-
Fu (2012). A Latent Class Modeling Approach for dents. Journal of Rail and Rapid Transit 221: 321–329.
Identifying Vehicle Driver Injury Severity Factors at McCullagh, P. & Nelder, J. A. (1989). Generalized linear mod-
Highway–Railway Crossings. Accident Analysis and els, 2nd Edition. Chapman and Hall, New York, NY.
Prevention 47: 119–127. Naik, B.; Tung, L.; Zhao, S.; and Khattak, A. (2016).
Fan, W. & Haile, E. (2014). Analysis of severity of vehicle Weather impacts on single-vehicle truck crash injury
crashes at highway-rail grade crossings: multinomial severity. Journal of Safety Research 58: 57–65.

535

TIS2017_Book.indb 535 2/8/2017 8:07:05 PM

 National Highway Traffic Safety Administration (2016). Williams, A. & Shabanova, V. (2003). Responsibility of
Fatality Analysis Reporting System (FARS). Avail- drivers, by age and gender, for motor-vehicle crash
able online at http://www-fars.nhtsa.dot.gov (date deaths. Journal of Safety Research 34(5): 527–531.
accessed: September 15, 2016). Wong, J. & Chung, Y. (2008). Comparison of method-
Oh, J.; Washington, S.; and Nam, D. (2006). Accident ology approach to identify causal factors of accident
Prediction Model for Railway–Highway Interfaces. severity. Transportation research Record: Journal of
Accident Analysis and Prevention 38(2): 346–356. the Transportation Research Board 2083: 190–198.
Plainis, S. & Murray, I. (2002). Reaction times as an index Wu, Q.; Zhang, G.; Zhu, X.; Liu, X.; and Tarefder, R.
of visual conspicuity when driving at night. Ophthal- (2016). Analysis of driver injury severity in single-
mic and Physiological Optics 22(5): 409–415. vehicle crashes on rural and urban roadways. Accident
Russo, B. & Savolainen, P. (2013). An examination of fac- Analysis and Prevention 94: 35–45.
tors affecting the frequency and severity of crashes at Yan, X., Han, L., Richards, S., and Millegan, H. (2010).
rail-grade crossings. Proceedings of 92nd Annual Meet- Train–vehicle crash risk comparison between before
ing, Transportation Research Board, Washington, D.C. and after stop signs installed at highway–rail grade
Saccomanno, F., Fu, L., and Miranda-Moreno, L. (2004). crossings. Traffic Injury Prevention 11(5): 535–542.
Risk-Based Model for Identifying Highway–Rail Yau, K. (2004). Risk factors affecting the severity of sin-
Grade Crossing Blackspots. Transportation Research gle vehicle traffic accidents in Hong Kong. Accident
Record: Journal of the Transportation Research Board Analysis and Prevention 36: 333–340.
1862: 127–135. Ye, X., Poplin, G., Bose, D. Forbes, A. Hurwitz, S.,
Saccomanno, F., Park, P., and Fu, L. (2007). Estimating Shaw, G., Crandall, J. (2016). Analysis of crash
Countermeasure Effects for Reducing Collisions at parameters and driver characteristics associated
Highway–Railway Grade Crossings. Accident Analy- with lower limb injury. Accident Analysis and Pre-
sis and Prevention 39(2): 406–416. vention 83: 37–46.
Tefft, B (2008). Risks older drivers pose to themselves Zhang, Y., Li, Z. B., Liu, P., & Zha, L. (2011). Exploring
and to other road users. Journal of Safety Research contributing factors to crash injury severity at freeway
39(6): 577–582. diverge areas using ordered probit model. Procedia
Wang, X., Liu, J., Khattak, A., Clarke, D. (2016). Non- Engineering, 21: 178–185.
crossing rail-trespassing crashes in the past decade: A Zhao, S. & Khattak, A. (2015). Motor vehicle drivers’
spatial approach to analyzing injury severity. Safety injuries in train–motor vehicle crashes. Accident Anal-
Science 82: 44–55. ysis and Prevention, 74: 162–168.

536

TIS2017_Book.indb 536 2/8/2017 8:07:05 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Development of a road asset management system in Kazakhstan

G. Bonin & S. Polizzotti

SPT Srl, Rome, Italy

G. Loprencipe
Sapienza Università di Roma, Rome, Italy

N. Folino & C. Oliviero Rossi

Università della Calabria, Rende (CS), Italy

B.B. Teltayev
KazDorNII (Kazakhstan Highway Research Institute), Almaty, Kazakhstan

ABSTRACT: The paper describes the development of a custom Road Asset Management System
(RAMS), that will take care of about 23’500 km of roads that make the Kazakhstan main network of
roads (Republican roads), that also includes 6 international corridors. The system has two main roles,
(1) organizing the asset information (road cadaster or “road passport” according to the Kazakh standard)
in a modern digital database and (2) managing the maintenance of the network, optimizing through
economic analysis the budget allocation for maintenance works. The system development takes also care
of organizing the data collection procedures for both roles, that will be done using automated devices
installed on mobile laboratories. The system will also use data from other sources, such as the growing
Intelligent Transport System (ITS) equipment (mainly weather stations, cameras, Weigh In Motion (WIM)
devices and traffic counters for the purposes of this system). The system is organized as a web based
service and it is accessible through any internet connected device, offering the operators the possibility
to browse the database or update it in any place with an internet connection available. One of the key
element of the system is its ability to make analysis and forecasts: the system is developed to measure
periodically condition data across all the network, to have a clear understanding and control on the
status of the roads. This module uses Highway Development and Management Model (HDM-4) to make
pavement maintenance analysis and optimization of resources. The system will start its operation with the
first complete data collection, that will be calibrated over the years by the repetition of condition analy-
sis, allowing to improve reliability and quality of analysis forecasts. The system will also serve for other
analysis, such as the control of Asset Value, analysis on the effect of new road projects over the network.

1 INTRODUCTION of cheap and powerful computing tools allowed to

efficiently manage the problem.
Road networks may be still considered, as a whole,
one of the more expensive assets for each Country
to develop and maintain; nevertheless, there is usu- 2 ASSET MANAGEMENT SYSTEMS
ally not enough attention to the management and
optimization of the expenditures related to it. Roads assets include a wide range of items, whose
The reason for this comes from multiple factors, most valuable is usually the road pavement, fol-
including the higher attention to the development lowed by the artificial structures and tunnels
of the road networks (opposed to rational man- (depending on terrain characteristics): these assets
agement and maintenance) and to the overall com- are managed through dedicated systems named,
plexity of the technical and economic evaluation. according to the core item Pavement Management
In the past 40 years, several attempts to ration- Systems (PMS), Bridge Management Systems
alize the management of such complex systems (BMS) and Tunnel Management Systems (TMS).
(asset management) has been started, but only in Modern road network systems are getting more
the last 20 years, with the widespread availability and more complex, in particular higher categories

537

TIS2017_Book.indb 537 2/8/2017 8:07:07 PM

 highways that nowadays include many Intelligent
Transport System (ITS) systems, as well as road
furniture (e.g. roadside safety devices and vertical
road signs) that require dedicated and continuous
monitoring and maintenance. The management
of such complex system require a new holistic
approach, that should take into account all the
parts of the road: these new systems go under the
name of Global Management Systems (GMS)
(Loprencipe et al., 2012, Bonin et al., 2007).

3 INFORMATION ON KAZAKHSTAN Figure 1. Kazakhstan main road network and neigh-

bour Countries.
Kazakhstan became an independent republic after
the fall of Soviet Union, in 1991; over the last 25 Table 1. Kazakhstan Republican Road network pave-
years the Country progressed steadily reaching the ment type.
upper-middle income status.
Kazakhstan has a wide territory (2’724’900 sq. Network
km) and a very small density (6.5 people/sq.km Pavement type km share, %
with a population of less than 18 million peo- Asphalt concrete 10’001 42.6%
ple); this is mostly due to its location and climatic Warm Mixed Asphalt (WMA) or
conditions. Hot Mixed Asphalt (HMA)
Kazakhstan is located in the middle of Central Cold Mixed Asphalt (CMA) 11’491 48.9%
Asia, just South of Siberia; the climate is continen- Cement concrete 97 0.4%
tal but, due to its size, it may be further divided Gravel and cold asphalt (transition) 1’797 7.6%
into 4 climatic zones. In most of its Central-North Unpaved 107 0.5%
and North-Eastern regions it suffers from very Total 23’493 100%
cold winters, where the soil may be continuously
frozen for more than 4 months.

The Republican road network consists of about

3.1 Road network
23’500 km of roads and is the main network of
Kazakhstan, due to its position in the Central Asia roads in the Country; it includes (i) international
and to its mostly flat land, it is a natural location corridors (ii) main national road network (iii) the
for international routes, such as the corridors that toll road network.
connects South (Kyrgyzstan, Uzbekistan and oth- Most (91.5%) of the Republican roads have an
ers through these) and North (Russian Federa- Asphalt concrete pavement, although about 8.1%
tion) and mostly East (China) and West (towards are gravel or unpaved roads (see Table 1).
Europe through Russian Federation).
For this reason, the World Bank funded a large
3.2 Review of the current situation—road
project on the “Western Europe-Western China
organizations
international transit corridor (CAREC1 1B and
6B)”, worth more than US$ 2 Billion (in the refer- Kazakhstan is still moving in its transition from
ences). This large project started in 2009 and mainly the old centrally planned economy to the mar-
targets the construction or improvement of these ket economy: while many transformations have
international corridors, but also includes a specific already been made, these are not yet fully consoli-
fund for the improvement of Public Administra- dated and many changes in the road organizations,
tion and Management procedures, including the management procedures and standards are still in
development of a Road Asset Management Sys- progress.
tem for the Republican Roads of Kazakhstan. The Road Asset Management System is a
modern tool that may serve to support the transi-
tion, not necessarily toward the market economy,
1
CAREC: Central Asia Regional Economic Program
but certainly to a more modern and efficient
(http://www.carecprogram.org/), a partnership of cen- management.
tral Asian Countries and multilateral institutions (funds The road sector is now controlled by a govern-
and investment banks) to promote development through mental authority, the Committee of Roads (CoR)
cooperation. that is under the Ministry of Investment and

538

TIS2017_Book.indb 538 2/8/2017 8:07:07 PM

 Development (MID); the CoR is surrounded by passport”. The contents of this “road passport”
some other publicly owned road organizations: is completely defined within the standard and
represents the official inventory of road assets,
• Road manager
which goes under the name of “road cadastre”
National Company (NC) KazAutoZhol is a
in other Countries.
Joint Stock Company (JSC) that has been estab-
• PR RK 218–27–14 (currently updating) on
lished recently (resolution 79/2013/CoR and
Diagnostics of road assets, that includes the
Government Resolution of December 26, 2013
standards and procedures for the testing of
1409).
roads assets, mainly road pavement and road
The CoR realizes some of its main objectives
structures (bridges and culverts).
through KazAutoZhol, such as (i) realization
• PR RK 218–19–01 on Spring and Autumns
of a state policy on development of a network
survey, that describes the procedures for the
of highways and (ii) maintenance of roads and
semi-annual road condition assessment and the
their structure for ensuring uninterrupted and
procedures for the calculation of the synthetic
safe traffic.
Global Road Quality Score.
• Maintenance operator
Republican State Enterprise (RSE) KazakhAv- Other standards define specific tasks and topics,
toDor is the road organization that, through its such as traffic measurements, material specifica-
local regional branches, conducts routine repair, tions and testing, quality control.
maintenance and landscaping of the republican
road network. The Committee for Roads acts as
a client for such works through KazAutoZhol. 3.4 Current maintenance and repairs estimates
Wider scope repairs are usually performed It is important to notice how the terminology
through bidding processes opened to private slightly differs in Kazakhstan from the interna-
contractors, although (Oblast) branches of tional common practice: this may lead to incon-
KazakhAvtoDor may participate, as a rule, for sistencies and misunderstandings in the analysis
republican roads sectors that belong to their and reports when international consultants oper-
region. ate in the Country, however the terminology in use
• Research, technical and scientific advisory is rather clear and consistent if it is well exposed
JSC KazDorNII is the Kazakh road research and clarified.
institute, whose stocks are owned by MID. Under the term “maintenance”, in the Kazakh
The most general direction of the institute’s sci- practice, are included: winter maintenance, land-
entific activity is enhancement of roads opera- scaping (mowing, trimming of vegetation), drain-
tion reliability, such as (i) investigating road age cleaning and the current maintenance (other
safety conditions in order to improve normative small maintenances excluding road repair).
base of roads designing and reconstruction and The repairs on road pavement or structures are
(ii) investigating opportunities to use traditional divided into:
materials as well as local and new materials,
including composite and industrial waste mate- • “current repairs”, with smaller scope, including
rials for construction of roads. patching potholes, crack sealing, repair of edge
KazDorNII has also a full set of testing equip- breaks, etc. (performed by public maintenance
ment for road materials as well as special testing company KazakhAvtoDor);
vehicles (Mobile Laboratories, ML) for the data • “mid-term repairs”, with intermediate scope,
collection of roads. including any pavement overlay—that is not
common practice in Kazakhstan—or recy-
cle/reconstruction of top layer of pavements,
3.3 Review of the current situation—road
under 50 mm (performed by public or private
standards and practice
contractor);
The current Kazakh standards are developed on • “major repairs” or “reconstruction”, with the
the basis of the Soviet Union standards, continu- wider scope: the definition differs mainly on
ously updated and coordinated with the other being the “reconstruction” a set of road pavement
Countries of the Commonwealth of Independent works that involve an upgrade of road category.
States (CSI).
The current practice for road maintenance is
Regarding the management of road assets,
driven by the “spring and autumn survey” stand-
3 standards are mostly relevant to the process:
ard, that is a road maintenance assessment that is
• PR RK 218–28–03 (currently updating) on Cer- performed twice a year, using the autumn survey
tification of roads (Road Passport), that includes results as a basis for the budget estimate for the
technical details for the definition of the “rad maintenance and repairs.

539

TIS2017_Book.indb 539 2/8/2017 8:07:07 PM

 There are, although, several shortcomings in the The main goal of the project described in this
current practice: paper is to address the shortcomings identified in
the current process for road maintenance manage-
1. Subjective assessment: the autumn and spring
ment in Kazakhstan.
surveys on the road conditions are carried on
This is largely supported by international organ-
by a commission that includes inspectors from
izations and funding agency, because it has been
road manager organization, road works/repair
identified as a key element to provide an assess-
organization and Police; the reports are based
ment on the real conditions of roads, to evaluate
on visual assessment and expert’s judgement of
the best strategies to solve the issues and to opti-
the inspectors rather than quantitative, objec-
mize the available budget allocated for the purpose
tive, measurements.
of improving road transport.
2. Cost estimate without prediction of pavement
deterioration: he current method, regarding the The RAMS main achievements will be:
road pavement maintenance, is essentially a cost
• The creation of a digital inventory, through
estimate of the repair works, based on the pre-
a structured database, of data regarding the
viously detailed visual assessment results. Such
Republican road network, including asset data,
cost estimate, (a) being the assessment made
conditional data (mainly regarding road pave-
during the autumn and (b) without the use of
ment), information regarding road maintenance
a road pavement deterioration model, always
expenditures and other data.
results in out of date cost estimates, when they
• The establishment of improved methodologies
are applied: the repair works, after the discus-
for the collection of data (road asset inventory
sion and approval process inside the Author-
and road pavement condition), through the use
ity, usually start more than 18 months after the
of advanced testing vehicles (mobile laborato-
assessment.
ries), whose availability in the Country, already
3. Optimization (i): reports are further used to
sufficient to start the activities, will be further
calculate, by a set of formulas that “weighs”
extended through a parallel project.
the asset conditions to calculate a global qual-
• The establishment of a new budget estimate
ity index, used to assign a “score” to each road;
and optimization procedure, through the use of
such score is used to prioritize the budget and
HDM-4 analysis. This will also allow the possi-
to represent the status to the Authorities (min-
bility to run iterative budget analysis, to expand
istries) involved in the budget allocation for the
the sensibility on the optimal budget allocation
next years. The synthetic single score including
by considering different scenarios.
all the maintenance needs may result in unclear
• The new HDM-4 based analysis allows also to
identification of the issues.
integrate the maintenance analysis with other
4. Optimization (ii): the current cost estimate
projects, such as new roads or reconstruction:
process does not use any optimization strat-
specific longer term analysis will allow to tune
egy and this is made worse by the fact that the
the budget allocated for maintenance with future
budget allocation are usually lower than the
new expansion of the network.
necessary need; the latter, that is likely in any
Country, without a budget reallocation strat- Following this, it is to remark that a complete
egy through optimization, causes maintenance Road Asset Management System is not only the soft-
inefficiencies and may carry also to an overall ware and equipment used to collect, store, analyse
lower care of the cost estimate process (there is and deliver information and results, but it is, indeed,
a general lack of confidence/interest into mak- extended to the whole process of managing the Road
ing it accurate, since no one will take advantage Asset, including, to a higher level, the road sector
of it). organization and, to a more practical level, all the
5. New technologies: this lacking process of rules (laws, technical standards, operating methodol-
assessment/analysis/optimization/control makes ogies) and the experience of all the people involved.
it very difficult to take advantage of new tech- The completion of such complex system will
nologies, either on the materials, machinery, take some years and will need adjustments and
management and execution of works. adaptation, as well as continuous support from the
government, the stakeholders and the operators.
4 DEVELOPMENT OF THE KAZAKH ROAD The transition towards the completion of the sys-
ASSET MANAGEMENT SYSTEM (RAMS) tem and the reform of the supporting standards and
maintenance approaches will be done in 3 stages:
Some earlier projects identified the need to support
reorganization of the road sector in Kazakhstan 1. First stage: foundation/development
with the development of a modern Road Asset includes the development of the core sys-
Management System (RAMS). tem (database, software and Information

540

TIS2017_Book.indb 540 2/8/2017 8:07:07 PM

 Technology (IT) architecture), the arrangement (such as Microsoft Word), making it very unpracti-
of equipment for data collection and definition cal for any reuse.
of methodologies and operating procedures. Kazakhstan has a dedicated organization and
During this phase it should be also detailed the some State protocols to enhance the use of dig-
strategic plan for the further implementation of ital technologies that goes under the name “Dig-
the system, that should also include the support ital Kazakhstan 2020”; part of this program, that
to Performance Based Maintenance Contracts. is now in an advanced status, is the transfer of all
2. The second stage: initialization the Governmental Servers and Systems to a mod-
It is foreseen an initial minimum 3/5-year time ern Cloud based infrastructure, that will assure the
frame for this, that will focus on completing best level of security for data, maintenance stand-
the reference data collection, the adjustment ards for hardware and availability (in terms of fast
of standards, methodologies and procedures communication channels and redundancy/failure
for the system operations. This will be done tolerance).
through the feedback from the actual use of the The Kazakhstan RAMS, that is a natural can-
system. During this stage the system manage- didate to became a Governmental system, also
ment will be maintained within the Ministry, uses this cloud based architecture, with a dedi-
with the direct involvement of the Road Man- cated MS SQL server for the DBMS (Data Base
ager, to allow a more comfortable transition to Management System). This assures the availabil-
the new procedures. ity of data (accessible through dedicated Virtual
3. The third stage: system complete (fully Private Network (VPN) channels or through
operational) Internet, depending on the level of access) and
This starts when (i) the system is ready enough a reliable and powerful platform for the analysis
to complete every year the cycle of data collec- needs.
tion-analysis-budget forecast for all the Repub- The database is structured according to the data
lican road network predicting reliably the road needed, that is divided into 3 main categories:
maintenance expenditures and (ii) the new
1. Asset data (road passport data in the local
performance standards and all the related pro-
standard).
cedures are set properly and all the stakehold-
2. Road condition data (mainly through the mobile
ers are ready to take full advantage of the new
testing vehicles)
procedures.
3. Other data, including those transferred from
The readiness in stage 3 represents the maturity, other systems (such as traffic data, weather con-
intended as the capacity to provide reliable fore- dition data, etc.) and data input by road organi-
casts, as well as allowing further improvements zations (operations data—road open/closed/
through research. roadworks, emergency data—road accidents/
The improvements include the whole process, severe weather conditions, management data—
starting from the general rules and standards, the e.g. actual expenditure data for roadworks and
correct setup of the road sector management proc- maintenance).
ess, the data collection and analysis procedures, and
Data needed for the yearly analysis through
also to take full advantage of any new technology.
HDM-4 (that comes from all the 3 categories) is
a subset of all the available data fields (in particu-
4.1 RAMS database lar regarding 1) asset data and 2) road condition
data): the local technical standards foresees a wider
The RAMS database is the core of the system, as it group of data, that is also included in the RAMS
stores all the data in a structured digital database. database and may be useful for additional detailed
This, itself, represents a great step forward for any analysis.
administration that is still based on traditional
practices, often relying on paper documents and
reports. 4.2 RAMS workflow
This situation is also common, yet, in many The general workflow of an Asset Management
Countries. The typical situation is the “incom- system, shown in Figure 2, has been used as a ref-
plete transition” to modern, fully digital systems: erence for the development of Kazakhstan RAMS;
although data are exchanged daily through the it shows clearly that there are several cycles to
offices using email or by exchanging data files, the control the implementation and the efficiency of
data are usually not structured or stored, in the the system, that shall be carefully enforced and
final revision, in some easy to reuse and reanalyse managed.
format, such as any database format. In addition to the items in the diagram, it is
Too often the final elaboration of the data are important to pursue the continuous improvement
stored in a PDF format or as a text documents of the system, that includes:

541

TIS2017_Book.indb 541 2/8/2017 8:07:07 PM

 IRI and rutting measurement systems, cam-
eras, Global Positioning System (GPS) and fully
automated Laser Cracking Measurement Sys-
tem (LCMS®), manufactured from Dynatest®.
• A vehicle with cameras and GPS, with specific
software dedicated to asset recognition and
positioning (Road Passport data collection),
manufactured from NPO Region®.

KazDorNII usually serves as the reference sci-

entific institute for the RoK, including the prep-
aration of technical standard and advisory for
technical and quality control issues. Their vehicles
are currently used regularly for the quality control
of the newly built roads. KazDorNII will likely be
directly involved in the data collection of RAMS
Figure 2. General workflow of an Asset Management
and as technical and scientific advisor for the
system (FHWA Asset Management Primer, 1999). RAMS management.
The OblZhollaboratories are the regional local
units directly controlled by the MID, consisting in
• calibration of the analysis tool (HDM-4), 14 Oblast (regional) units, plus 2 newly established
through the availability of feedback data from for the main cities of Kazakhstan and their metro-
the implementation of the maintenance works politan region, Astana and Almaty.
and comparison with the previous predictions; OblZhollaboratories will be the main source of
• refinement in data collection procedures, that pavement condition data, due to the large avail-
may be enhanced from availability of new equip- ability of equipment and to its dislocation in
ment and technologies, but it is mainly due to the Country: the MID is completing the fleet of
the increased experience of the operators, the 16 mobile testing vehicles (one per each branch),
data quality assurance procedures and the better including the refurbishment of some previously
understanding of the road network; owned vehicles.
• critical analysis of the whole process perform- These vehicles are fully equipped with Falling
ance is necessary, to identify shortcomings and Weight Deflectometer (FWD), Ground Penetrat-
inefficiencies; this may be done through an ing Radar (GPR), International Roughness Index
internal evaluation of the system performance, (IRI) and rutting measurement systems, cameras,
with the improvement of regulatory documents GPS and a semi-automated pavement defect meas-
and standards and through an expert’s external urement system.
supervision and dedicated research projects. The data will be collected during the end of
summer or in autumn, before the winter snow; it is
foreseen that, given the availability of vehicles, the
4.3 Data collection surveys will be repeated every year during the ini-
tial phase of RAMS operation. This will assure a
Although the system is flexible to store data for
repeated feedback from the roads and the improve-
different needs, it is important to clarify that the
ment of the analysis prediction accuracy. When the
main scope of this Project, the development of the
system will be mature enough, in terms of road
Kazakh RAMS, is to assure the possibility to store
condition prediction accuracy, it will be possible to
information and execute analysis for the strategic
reduce the frequency of surveys.
budget allocation.
Survey data, after initial processing with on
For this purpose, the Ministry of Investment
board equipment and software, will be trans-
and Development (MID) of the Republic of Kaza-
ferred (stored) to the RAMS database for further
khstan (RoK) invested in the purchase of highly
processing.
specialized mobile testing vehicles for its National
The data processing is performed by trained
Research Institute for Roads (KazDorNII) and for
operators and, currently, is greatly affected by the
the OblZhollaboratory road organization.
recognition of asset data and from the analysis
KazDorNII has 2 mobile testing devices for this
of the pavement defects, that are not fully auto-
purpose:
mated. Additional statistic strategies (such as
• A fully equipped Multi Function Vehicle those described in Krawczyk B., Szydło A, 2013)
(MFV®) with Falling Weight Deflectometer are adopted to reduce the amount of surveyed area
(FWD), Ground Penetrating Radar (GPR), (subsample) to be processed; this approach is also

542

TIS2017_Book.indb 542 2/8/2017 8:07:07 PM

 confirmed by other pavement distress methods 4.4 Analysis
(such as the Pavement Condition Index (PCI) sur-
The Performance modelling and Alternative eval-
vey methodology described in the ASTM-D6433).
uation shown in Figure 2 are performed in the
This approach will be further improved through
Kazakh RAMS through HDM-4.
specific research projects, when data will be avail-
HDM-4 (Highway Development and Manage-
able, and improved based on the availability of new
ment Tools collectively referred to as HDM-4), is
software to assist the image recognition process.
the last revision of the Highway Design and Main-
Before the data will be fully published in the
tenance Standards Model (HDM), developed by
RAMS, therefore made available to the users and
the World Bank’s Transportation Department for
for the analysis, it will be performed a quality
the analysis, appraisal and optimization/prioritiza-
assurance procedure.
tion of road management and investment alterna-
Since the main data available through the use of
tives. HDM-4 is currently distributed and updated
the mobile testing vehicles are those regarding the
by HDMGlobal, an international consortium of
pavement condition and considering that the input
academic and consultancy.
data for the analysis tool (HDM-4) also includes
The software includes a Pavement Management
other data, it is necessary also to provide them.
System (PMS) based on a Pavement deterioration
HDM-4 does not allow a detailed analysis of
model (engineering module) and an Economic
assets like artificial structures (bridges, culverts,
model, which are used to perform an integrated
retaining walls, tunnels etc.), drainage and tech-
analysis.
nological systems (such as Intelligent Transporta-
The pavement performance models used in
tion System (ITS) devices), so the conditional data
HDM-4 are based on a deterministic approach
regarding these asset categories is collected using
(se details in Watanatada et al., 1987; ISOHDM,
other methods, mainly through specific surveys
1995a,b).
and through the analysis of the reports form the
The input parameters include, traffic, climate
road manager (KazAvtoZhol) and the road main-
conditions, vehicle fleet consistency, road geom-
tenance road organization (KazakhAvtoDor). It is
etry data, safety information (accident rates) and
possible that the RAMS will be further expanded
details on the road pavement, that is the main engi-
with additional modules dedicated to these asset
neering item. The pavement is detailed through its
categories (Bridge and Systems in particular).
structural and construction characteristics (layer’s
The drainage conditions, from the engineering
thicknesses), conditional data (roughness (in terms
point of view, are particularly important in some
of IRI), cracks, potholes, edge breaks, etc.) and sur-
areas, because the flat terrain and the thawing of
face characteristics (friction and texture depth).
winter snow and ice causes big problems, with per-
The analysis consists in a two integrated sub-
sistent floodings and soil softening issues.
analysis:
The other main categories that have to be taken
into account are: • an evaluation of the pavement performance, pre-
dicting its deterioration over time, with or with-
• climate data: this has to be assessed and moni- out the application of maintenance strategies
tored periodically, as it is not necessary to track • an economic analysis, performed through the eval-
the continuous temperature changes for the uation of the Vehicle Operating Costs (VOC).
analysis, but the Country has severe climate con-
ditions, with hot summers and extreme winters, With this approach, it is possible to consider
resulting in high annual excursion (may result in the effects on the users, rather than the sole agency
much more than 100°C if measured on the road cost and asset value.
surface, close to 90°C as standard air tempera- The agency costs, essentially represented in the
ture, (see KazHydroMet.kz) and daily excur- following Figure 3 as Road Work Costs (RWC)
sions (typical around 20°C as air temperature). may be added to the User’s Cost (UC, or VOC) to
• Traffic data and vehicle fleet composition: this is calculate the Total Road Transport Cost and select
one key parameter for any pavement analysis; this the best option.
is currently measured using traditional methods, This tool will greatly enhance the analysis capa-
but there is an increasing number of ITS devices bilities: in the current maintenance evaluation proc-
that will allow continuous monitoring of traffic ess, the budget allocation decision is based on the
and axle loading data (trough Weigh In Motion Cost Estimates calculated using the data collected
(WIM) ITS devices): data from the continuous using by mean of the visual assessment method
monitoring will allow better understanding of described before; the current process, also, is unable
the specific loading conditions, that shall be per- to predict the pavement condition over time, there-
formed as a separate, analysis before the input in fore the budget allocations are decided on the basis
HDM-4. of conditional data that are usually more than

543

TIS2017_Book.indb 543 2/8/2017 8:07:08 PM

 Table 2. Strategic, network, and project-level decisions
(Zimmerman et al., 2011 and FHWA, 1999).

Decision Types of decisions/

level Decision maker Activities

Strategic Legislator Performance targets

Commissioner Funding allocations
Chief Engineer Pavement
Council Member preservation
strategy
Network Asset Manager Project and
(Tactical) Pavement treatment
Management recommendations
Figure 3. Total cost, Road User’s Cost and Road Works Engineer for a multi-year
Cost. District plan
Engineer Funding needed
to achieve
1 year old (including 2 winter seasons) when the performance
maintenance and repair works actually start. targets
The HDM-4 allows the execution of project Consequences
level analysis, road work programming under con- of different
strained budgets, and for strategic planning of long investment
term network performance and expenditure needs. strategies
It is designed to be used as a decision support tool Project Design Engineer Maintenance
(Operational) Construction activities for
within a road management system.
Engineer current funding
This allows to serve Strategic and Network (tac- Materials year
tical) decision levels, according to the definition Engineer Pavement
described in the previous Table 2. This will allow a Operations rehabilitation
direct involvement in the analysis of the high level Engineer thickness design
decision makers (usually Ministries) and mid-high Material type
level decision makers (typically the road manage- selection
ment agency, KazAvtoZhol in RoK). Life cycle
The analysis, after its completion, validation costing
and acceptance, it is loaded into the RAMS data-
base and made available to the authorized users.
Direct browsing of data may be performed on
4.5 Reporting and information to stakeholders the raw data (tables, data fields) or through the
GIS mapping tool. This will also allow further
A very important function of the RAMS is the development of the system, by allowing the profes-
possibility to provide information to the stake- sional road operators to browse the system on the
holders and to the road users, with specific tools road through mobile devices.
and contents. The GIS mapping tool is also the main medium
The system, through its data base core, may of data publishing to the road users, the popula-
directly provide data, through reports (predefined tion. The data will be prepared through the main
or custom made), queries and through a GIS GIS tool within the RAMS, then the data will be
interface. published on an additional dedicated service (web
The data are provided to each user category on server) on the basis of public mapping services
the basis of the user’s needs and according to the (such as Google maps or Yandex maps). This will
RAMS manager authorization. allow the integration of proprietary data (such as
The higher level of authorization is granted to a synthetic 3 level pavement condition index or
the Authorities and to the system administrators. the information on the road service—road open/
Reports are usually targeted to specific needs, closed/with reduced lanes) with publicly available
including periodic reports for official purposes, data through an interface familiar to users.
evaluation of performance, verification.
Queries may be used to perform specific data
analysis, including specific asset verification; the 5 CONCLUSIONS
large amount of structured data will allow research
projects, that are encouraged to further improve The development of a new Road Asset Manage-
the system. ment System is a complex task, that involves the

544

TIS2017_Book.indb 544 2/8/2017 8:07:08 PM

 knowledge of the state of the art for the equipment Deterioration and Maintenance Effects in HDM-4.
and procedures, as well as the understanding of the Asian Development Bank, Manila, Philippines.
stakeholders’ needs. ISOHDM. 1995b. International Study of Highway Devel-
It is very important to understand the current opment and Management Tools Modelling Road Users
Effects in HDM-4 (Asian Development Bank, Manila,
procedure in use and to be able to make the system Philippines).
“useful” to all the involved players, to make its use Krawczyk, B., & Szydło, A. (2013). Identification of
really fruitful. homogeneous pavement sections. Roads and Bridges-
It is, indeed, really important to communicate Drogi i Mosty, 12(3), 269–281.
the system capabilities and needs, to involve all the ISO 690.
stakeholders to use it for their everyday duties and, KazHydroMet.kz, historical records of temperatures in
therefore, continuously improving it. the main cities of Kazakhstan see http://www.kazhy-
The biggest advantages of the use of the RAMS dromet.kz/en/inforeg/
in the RoK will be the availability of a structured, Loprencipe, G., G. Cantisani & P. Di Mascio 2014.
Global assessment method of road distresses, Fourth
fast, always available source of data for the road International Symposium on Life-Cycle Civil Engineer-
assets and the possibility to produce more reliable, ing IALCCE 2014.
detailed and useful analysis, including the optimi- Pavement Design, Construction and Management: a dig-
zation of budget allocations, that are nowadays ital Handbook, 2015. AASHTO. Washington, DC.
not possible. South West Roads: Western Europe-Western China inter-
national transit corridor (CAREC 1B and 6B) Project.
World Bank loan no. 7681-KZ, retrieved on http://www.
worldbank.org/projects/P099270/south-west-roads-
REFERENCES western-europe-western-china-international-transit-
corridor-carec-1b-6b?lang = en&tab = overview
ASTM D 6433–11. 2011. Standard Practice for Roads Watanatada, T., Harral, C., Paterson, W., Dhareshwar,
and Parking Lots Pavement Condition Index Surveys. A., Bhandari, A. & Tsunokawa, K. 1987. The
West Conshohocken, PA, USA. Highway Design and Maintenance Standards
Bonin, G., Cantisani, G., Loprencipe, G., Ranzo, A. Model—Description of the HDM-III Model. The
(2007). Dynamic effects in concrete airport pavement World Bank, Washington, DC. Vol. 1.
joints [Effetti dinamici nei giunti delle pavimentazioni Zimmerman, K.A., Wolters A.S., K. Schattler,
aeroportuali in calcestruzzo]. Industria Italiana del A. Rietgraf. 2011. Implementing pavement manage-
Cemento, 77 (834), pp. 590–607. ment systems for local agencies.
FHWA Asset Management Primer, 1999.
ISOHDM. 1995a. International Study of Highway
Development and Management Tools Modelling Road

545

TIS2017_Book.indb 545 2/8/2017 8:07:08 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Building Information Modeling (BIM): Prospects for the development

of railway infrastructure industry

M. Leone, A. D’Andrea, G. Loprencipe & G. Malavasi

Dipartimento di Ingegneria Civile, Edile e Ambientale. Sapienza, University of Rome, Rome, Italy

L. Bernardini
Italferr S.p.A., Gruppo Ferrovie dello Stato Italiane (Italian State Railways Group), Rome, Italy

ABSTRACT: The last decade has seen a profound change of information representation tools with
which the Infrastructure Manager must necessarily interact with the other parties involved, to explore
all the essential decision-making processes for railway development and operation. In all phases of infra-
structure life (project, construction, operation, adjustment, etc.), the digital representation can greatly
facilitate the organizational model that always puts the information content of all railway components
(geometry, size, location, materials, design and proper maintenance, stress state, etc.) at the center of
any decision-making process. On account of its strong multidisciplinary nature, a railway infrastructure
causes numerous interactions that need to be coordinated properly to meet the performance levels now
required in any transportation system. In this sense, the development of BIM processes for punctual
building has now reached a level of maturity able to satisfy also the recent guidelines of the European
Standards (2014/24 / EU) adopted to promote, first of all, the transparency of public work contracts
and design contests. However, as regards linear construction, the full application of these BIM processes
with an adequate level of maturity still remains rather complex considering that the current processes
representation in use has achieved an adequate level of development. In this case study, applications have
been developed in BIM mode with the aim of highlighting the main potential of this innovative process
including some improvements in order to promote the full implementation of the BIM technology for rail
infrastructure.

1 INTRODUCTION telecommunication, etc. In order to avoid dis-

cordant dynamics and a contrasting technical
Currently railways, especially High-Speed trains, project development it is fundamental to bring
are considered one of the best nationwide means in a technical coordination of the various stages
of transport to solve the problem of mobility. belonging to railway construction projects (Project
However, building, maintaining and operating Manager as regards project planning/design).
HSR lines is expensive; it involves a significant Furthermore, decisions as to how and to what
amount of sunk costs and may compromise the end construction stages are handled, may have
transport policy of a country. to be revised constantly. This may even concern
The full lifecycle development of a railway infra- already well-established information.
structure includes several separate main stages, As a matter of fact, the transition from the
each supervised by the technical expert in charge. construction stage to that of maintenance often
A railway construction project is made up of the turns out to be conflicting or unrelated. This is
following parts: due to the use of two different responsible control
bodies as well as two software computer programs
• Concept;
for data processing with a consequential loss and
• Design/Planning;
discordance of information.
• Construction;
In addition, software undergoes a continuous
• Maintenance;
and unforeseeable upgrading. This makes it almost
• Adjustment/Implementation;
impossible to maintain its application the same
• Disposal (demolition or other use).
throughout the lifecycle of a railway network.
The planning and construction of railway lines Nowadays the means of controlling the entire
involve a number of operating branches. For exam- process is offered by BIM. The common definition
ple: civil engineering, electrification, signaling, of BIM is: “the use of shared digital representation

547

TIS2017_Book.indb 547 2/8/2017 8:07:10 PM

 of a built object (including buildings, bridges, roads, The concept of integration reveals having two
process plants, etc.) to facilitate design, construc- different aspects: the first one regards people and
tion and operation processes and to form a reliable processes and the second one concerns the integra-
basis for decisions” (ISO 29481-1:2016). tion of interoperability systems.
On these grounds the EU has introduced the The integration of some stages and disciplines
norm 2014/24/EU art. 22 c.4 for the use of BIM is extremely useful from an early stage of project.
when constructing public buildings and transport This allows the virtual construction of the work in
systems. For public work contracts and design its definitive configuration and the possibility to
contests, Member States may require the use of see the integration of more disciplines.
specific electronic tools, such as of building infor- Still the BIM process seems to be too compli-
mation and electronic modeling tools or similar. cated for many and initially adopted in a limited
So far the BIM has often been applied for built manner (Howard & Björk, 2008).
objects like buildings and bridges. Instead projects of Nevertheless, literature is full of examples deal-
infrastructural networks have found very little BIM ing with stage integration and the coexistence of
application both in literature and construction. various disciplines (Bonin et al., 2009).
Despite this, there are some case studies using The evolution from design 2D to 3D has
the BIM mode for a railway line section with the improved the structure work and the conflict
aim of highlighting the main potential of this management of structural aspects and Mechani-
innovative process. cal, Electrical and Plumbing (MEP). Moreover,
This was made possible by the cooperation be- the use of 4D has brought together the modeling
tween Italferr S.p.a. (Italian State Railways Group) of the stages concerning the construction site like
and University of Rome with the developing of planning and scheduling, data analytics and man-
some master degree theses. agement (time related information) and those
regarding safety. Lifecycle integration represents a
further important element dealt with by the BIM.
2 REVIEW This allows its modeling from the very beginning
of the project. The emerging advantages can be
The intended function of Digital and Information summarized in the following three points (Aranda-
technology, when employed by BIM, is to integrate Mena et al., 2009):
all activities concerning projects and constructions
of public works. • the possibility of exchanging models with the
On account of this, scientific literature (Nurain, technical project advisors (technical ability);
2013) evidences three main categories: • the ability to project in 3D during the entire
design process (operative ability);
• theoretical perspectives; • the ability to complete important projects of big
• integration; dimensions more and more efficiently (business
• implementation. ability).
Theoretical perspectives explore studies that
A further outstanding aspect of integration con-
have offered key theoretical perspectives on digital
cerns the information exchange among the various
technology use.
technicians belonging to the planning and con-
What is more, researchers have explored the
structing process, thus forming real centralized 3D
application of digital technology in the Architec-
data bases. In spite of the many proposals present-
ture, Engineering and Construction (AEC) dis-
ing standard models and data service, interoperabil-
ciplines. Researchers described their behavior as
ity (integration of systems) is still a problem of the
regards their development and use (Taylor & Bern-
BIM building process (Grilo & Jardim-Goncalves,
stein, 2009). This takes into account the four main
2010). There are exchange systems like the Industry
elements of BIM:
Foundation Classes (IFC), the most advanced, that
• visualization; allow the exchange of information and models.
• coordination; There are other exchanges that adopt interoper-
• analysis; ability only in some sectors. A great effort has to
• integration. be made to find a single standard exchange suitable
for all operators and digital platforms. However,
Instead, visualization, integration and auto-
there are different types of implementation:
mation are the three stages belonging to virtual
planning/design and development/construction. • referring to BIM technology with procedures
Finally, when proceeding with BIM, priority already carried out;
is given to automation, information and • including the main advantages and problems
transformation. dealt with;

548

TIS2017_Book.indb 548 2/8/2017 8:07:10 PM

 • with the users’ perceptions and expectations; objects. The studied track presents 24 interfer-
• with the suggested technical standards for ences, (18 box culverts, two road underpasses, one
implementations; electrical substation, one bridge crossings a chan-
• with activated ecological engineering. nel, two box structures for crossings on foot and
by bike). In fact, this section represents a railway
The approach (Arayici et al., 2011) of network project for densely populated urban cent-
bottom-up implementation results being more ers with a lot of interference. This project has used
efficient for the BIM implementation when linked the support of the integration software owned by
with a pull-technology strategy. The latter reduces the Bentley family.
the resistance to change and the risk of substitut- MICROSTATION: It is a 3D modeler used by
ing the already existing well performing processes. most software of this platform. It allows modeling
Positive aspects can include: any kind of frame in a parametric mode provid-
• the opportunity to produce design and documen- ing the user with immediate programmable report,
tation directly from the BIM model (technical for example the estimate. In addition, this program
result); allows assigning the kind of material and other
• 3D design during the entire process (operative specific properties (service, mechanical capacity,
result); general hyperlink, etc.). It also allows the publica-
• minor risk of wrong interaction between tion of the I-model to be used during the whole
structural aspects and MEP (business results). project by all final users and types of computer
even without the Bentley software.
Other studies show how the BIM process can POWERCIVIL: It is modeler that makes plan-
upgrade all application disciplines and that users ning linear infrastructure easier for civil linear
recognize improvement in all construction indi- project. It can be used to elaborate a Digital
cators (Singh et al., 2011). Expectations, instead, Terrain Model (DTM), horizontal and vertical
concentrate on the development of norms, con- road alignment and their coordination. It is also
tents, software, formation and certification as able to model box structures, pipes, retaining walls,
well as the use of the BIM in ecological projects etc. Furthermore, it can create libraries for the
(Young et al., 2008, Cantisani et al., 2011, Moretti parametric and programmable cross sections to
et al., 2016). Major problems concern integrated solve interference with DTM and other punctual
digital and technological implementations, costs, built object.
more time and effort for 3D models, resources and POWER RAIL TRACK: It is a specific railway
proper coordination. The guidelines for implemen- project modeler that allows to insert cants, transi-
tation focus on the following themes: tion curves and rail switches and to develop railway
• project making; station layouts.
• development and respect for consolidated PROJECTWISE: It is the managing software. It
practices; coordinates all software and users of the project.
• BIM monitoring and support supply for users. It saves all project information and authorizes
modification and visualization by technicians.
Other recommended rules address the matter of Modifications are communicated to other users by
technical evolution, procedure and organization. The alert notice. In addition, this software memorizes
BIM can play an important role in eco-sustainable all project changes and the users who made them.
projects (Zhu, 2010) by identifying its best option Basically, it is the managing platform for user
beforehand. However, some positive examples interoperability in design stage and it guarantees
aside, the ECO-BIM is still underestimated. operation simultaneity.
This railway line project has been developed and
3 DESCRIPTION OF CASE STUDY elaborated in the following three steps:

The project refers to a quadruple-track railway 1. The available cartographic data have been used
belonging to the Milano Rogoredo—Pavia net- for a DTM to insert the above-given railway sec-
work. Its construction consists in two 26 km new tion. Before starting, a preliminary operation of
rail tracks alongside the two already existing rails data reprocessing and cleaning is always needed
tracks from Pieve Emanuele to Pavia. to obtain a good definition of the terrain model
In this case the BIM process has been applied (Fig. 1).
for a 6 km long section starting from km 2. The realization of the railway solid in 3D. This
21+646.880 to km 27+241.060. By doing so, it has point has three steps.
been possible to verify the full potential of BIM a. The horizontal alignment design (Fig. 2).
application development. This regarded the pros This was done by following in parallel the two
and cons of layout interference with other built already existing railways, imposing a distance

549

TIS2017_Book.indb 549 2/8/2017 8:07:10 PM

 the tracks, requiring it to maintain a constant
vertical distance between the old and the new
railway. The design of the vertical curves was
based on the project of railway speed.
c. The setting-out of the cross section in case of
either embankment or trenches and its closure
rules with the terrain, as well as the extrusion
of this section on the alignments. (Fig. 4).
The railway 3D solid (Fig. 5) needs minor and
major built objects to solve interference with the
terrain. The railroad equipment (ballast, rails, and
sleepers) and other civil and technologic railway
installations complete the infrastructure (electrical,
traction, signaling, cable duct, etc.).
This part of the case study describes the build-
Figure 1. Digital Terrain Model (DTM).
ing of two objects used to avoid interference
with the territory. The first one is a box culvert
structure to restore the natural water flow when
interrupted by the embankment construction. The
second one is a rail crossing solved by a new road
underpass.
The box culvert structure design needed a pre-
liminary horizontal placement (Fig. 6) and the
choice of the reference height and the longitudinal
slope (Fig. 7).
In this case, the extrusion of the section along
the design axis considers the subtraction of the

Figure 2. Plane view of the horizontal alignment.

Figure 4. Cross section of the project: definition of all

rules.

Figure 3. Vertical view of the elevation profile.

between track centers of 7.60 m. The layout

was then continued by using the new coor-
dinates and referring only to the odd track
(Italian practice). Its geometry consisted of a
straight (L = 4500 m) and two opposite hori-
zontal curves (R1 = 904 m; R2 = 1500 m).
b. The determination of the elevation profile and
the insertion of the gradients and the vertical
curves (Fig. 3) considered the quadrupling of Figure 5. The railway 3D solid.

550

TIS2017_Book.indb 550 2/8/2017 8:07:10 PM

 to draw up an estimate subsequently. This method
enables the complete control of costs and tim-
ing during design phase as well as the workflow
regarding the construction stages.
The modeling of the road underpass follows
the same operating procedure used for other built
objects. This means definition of road horizontal
layout first and vertical profile after, followed by
cross section design with which the solid of the built
object has been obtained. Also in this case, the sub-
traction of the embankment volume concerning the
overlapping part of the underpass is indispensable to
avoid a wrong estimate and visualization problems.
The underpass design requires the definition
Figure 6. Box culvert structure horizontal placement. of various cross sections that vary throughout the
corridor (Fig. 9).
Fig. 10 shows some of the defined cross sections
used for the underpass modeling.
The underpass has been completed with road pav-
ings and road furniture (parapets, barriers and so on.)
and draining pipes for rain water discharge (Fig. 11).
The hydraulic system has been placed in the fit-
ting spaces of the cross section. Additionally, this
3D multi-disciplinary project works together with

Figure 7. Box culvert structure vertical placement.

Figure 9. The 3D model of the underpass.

Figure 8. The railway 3D solid in correspondence of

the box culvert structure.

embankment volume to avoid having a volume

redoubling (Fig. 8).
This process makes it possible to assign the Figure 10. The defined cross sections used for the
characteristic properties to each material in order underpass modeling.

551

TIS2017_Book.indb 551 2/8/2017 8:07:13 PM

 order to find the most suitable solution among
those offered by computer technology.
The above-cited applications met the expected
and acknowledged design requirements of the
engineering company in charge as well as the tech-
nical specification required by the customer. Fur-
thermore, the models developed for this case study
can be reused with minor modifications for other
projects benefitting from one of the major advan-
tages of the BIM process. Finally, the BIM process
has the potential to overcome procedural delays
that have slowed down the construction industry
development in many countries especially in Italy,
due to shortage of the design. In conclusion, the
implementation of the BIM process in railway
infrastructures needs steady improvement. This
regards library development and available models
Figure 11. The 3D model of the underpass completed
with road pavings and street furniture. for all users so as to encourage the development
of this methodology and therewith its usability of
information throughout the lifecycle of an infra-
structural work.

ACKNOWLEDGEMENTS

We thank our colleagues from Italferr SpA (Italian

State Railways Group), Eng Tascione P. and from
Bentley SpA Eng. Cristallini C. and Magalotti E.
who provided insight and expertise that greatly
assisted the research.

REFERENCES

Figure 12. The railway 3D solid in correspondence of Aranda-Mena G., Crawford J., Chevez A. and Fröese T.
the underpass and the noise barrier. (2009). “Building information modelling demystified:
Does it make business sense to adopt BIM?” Interna-
tional Journal of Managing Projects in Business, 2(3),
a unified and integrated view to limit design errors. 419–434.
In case of road underpasses, design errors may give Arayici Y., Coates P., Koskela L., Kagioglou M., Usher
rise to high-risk situations in the event of adverse C. and O’Reilly K. (2011). “Technology adoption in
climatic conditions (flooding). the BIM implementation for lean architectural prac-
Fig. 12 represents the complete underpass model tice”. Automation in Construction, 20(2), 189–195.
placed in the railway project. Bonin, G., Cantisani, G., Ranzo, A., Loprencipe, G., &
Besides, in correspondence to the under- Atahan, A.O. (2009). Retrofit of an existing Italian
bridge rail for H4a containment level using simula-
pass (Fig. 12) there is a noise barrier model. On tion. International Journal of Heavy Vehicle Systems,
account of its 3D modeling and BIM process It 16(1–2), 258–270.
has been possible to place it correctly in the most Cantisani, G., Loprencipe, G., and Primieri, F. (2011).
suitable foundation system for the given substrate The integrated design of urban road intersections:
(foundation piles for the embankment, lag bolts A case study. In The International Conference on
in correspondence of built objects) this choice of Sustainable Design and Construction (pp. 722–728)
design is directly defined in the model in terms of ICSDC 2011: Integrating Sustainability Practices in
materials, technical characteristic, geometry, etc. the Construction Industry, Kansas City, Missouri.
Directive 2014/24/EU of the European Parliament
and of the Council of 26 February 2014 on public
procurement and repealing Directive 2004/18/EC.
4 CONCLUSIONS Grilo A. and Jardim-Goncalves R. (2010). “Value propo-
sition on interoperability of BIM and collaborative
All BIM mode applications developed in this case working environments”. Automation in Construction,
study have been the result of intensive research in 19(5), 522–530.

552

TIS2017_Book.indb 552 2/8/2017 8:07:16 PM

 Howard R. and Björk B.-C. (2008). “Building information Singh V., Gu N. and Wang X.Y. (2011). “A theoretical frame-
modelling - experts’ views on standardisation and work of a BIM-based multi-disciplinary collaboration
industry deployment”. Advanced Engineering Infor- platform”. Automation in Construction, 20(2), 134–144.
matics, 22(2), 271–280. Taylor J.E. and Bernstein P.G. (2009). “Paradigm
ISO 29481-1:2016. Building information models— trajectories of building information modeling practice
Information delivery manual—Part 1: Methodology in project networks”. Journal of Management in
and format. Engineering, 25(2), 69–76.
Moretti, L., Cantisani, G., & Di Mascio, P. (2016). Man- Young Jr. N.W., Jones S.A. and Bernstein H.M. (2008).
agement of road tunnels: Construction, maintenance “Building information modelling (BIM): Transform-
and lighting costs. Tunnelling and Underground Space ing design and construction to achieve greater industry
Technology, 51, 84–89. productivity.” McGraw Hill Construction.
Nurain Hassan Ibrahim (2013) “Reviewing the evidence: Zhu S.T. (2010). “Sustainable building design based on
use of digital collaboration technologies in major BIM”. Paper presented at the EBM 2010: International
building and infrastructure projects” Journal of Infor- Conference on Engineering and Business Management,
mation Technology in Construction. Vol 1–8.

553

TIS2017_Book.indb 553 2/8/2017 8:07:17 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Dissipated energy approach for the fatigue evaluation of RAP

asphalt mixtures

M. Pasetto
University of Padua, Padua, Italy

N. Baldo
University of Udine, Udine, Italy

ABSTRACT: The results are discussed of a laboratory investigation on the fatigue resistance of Base
Courses Asphalt (BCA) mixes, for road pavements, evaluated by means of the four-point bending test.
The experimental analysis was performed on bituminous mixtures with Reclaimed Asphalt Pavement
(RAP), up to 40% by weight of the aggregate. Both conventional and polymer modified bitumen was
used in the investigation. The mix design was based on volumetric (gyratory) and indirect tensile tests, on
both wet and dry samples. The fatigue behaviour was evaluated at 20°C and 10 Hz, under strain control
mode, by means of the conventional approach, based on the reduction in the initial stiffness modulus,
as well as using a dissipated energy method focused on the Plateau Value concept. Between the different
approaches used for the fatigue data analysis, it has been verified a qualitative consistency in the mixtures
ranking, with a higher fatigue resistance presented by the polymer modified mixtures produced with RAP
aggregates.

1 INTRODUCTION Rejuvenator agents represent also an effective

option useful to improve the fatigue response of
In the last decades one of the most relevant mixtures containing RAP materials (Celauro et al.
research themes within the road scientific com- 2009, Huang et al. 2005). However, an increment
munity has been the reuse of discarded materials of the fatigue resistance of RAP mixtures was
in the pavement structure, in order to improve the documented by few researchers who used a low
sustainability of the road constructions (Kavussi penetration grade bitumen, by means of a proper
& Qazizadeh 2014, Oluwasola et al. 2015, Pasetto mix design method (Miró et al. 2011, Valdés et al.
& Baldo 2012, 2013a, 2014a, Xue et al. 2009, Wu 2011). Eventually, a comparable fatigue resistance
et al. 2007). between RAP hot mix asphalts and high modulus
Among the several type of wastes considered, bituminous mixtures has been recently observed.
the Reclaimed Asphalt Pavement (RAP) aggregate Therefore, the fatigue behaviour of RAP mixes
is one of the most investigated, due to its strate- represents a topic which has to be furtherly studied,
gic importance in the road construction industry in order to improve the knowledge of the mechani-
(Celauro et al. 2009, Oliveira et al. 2013, Pereira cal response of such materials.
et al. 2004, Silva et al. 2012, Stimilli et al. 2016). In this experimental study, the fatigue behav-
Nevertheless, the influence of the RAP aggregate iour of polymer modified asphalt concretes, for
on the fatigue life of the asphalt concretes is still base courses of road pavements, prepared with
not fully clarified. Reclaimed Asphalt Pavement (RAP), was inves-
Some studies reports a lower fatigue resistance tigated using the four-point bending test (4PBT),
(Colbert & You 2012, Miró et al. 2011, Pereira under strain control mode.
et al. 2004, Silva et al. 2012, Stimilli et al. 2016), The conventional stiffness reduction method
primarily due to a brittle response of the mix, as and a dissipated energy approach were used in
a consequence of the stiff aged binder of the RAP order to analyse and elaborate the fatigue data.
particles (Colbert & You 2012). In order to mini- The main goal of the study was to design high
mize such problem, a soft bitumen may be used performance recycled bituminous mixtures, com-
as new binder to be added to the RAP mixtures bining RAP aggregates and polymer modified
(Celauro et al. 2009, Oliveira et al. 2013, Pereira bitumen, in order to improve the fatigue behaviour
et al. 2004, Silva et al. 2012). of conventional polymer modified mixes.

555

TIS2017_Book.indb 555 2/8/2017 8:07:19 PM

 2 MATERIALS Table 1. Bitumen characterization.

2.1 Aggregates and bitumen RAP Conventional Polymer

Properties binder bitumen modified
The RAP considered in the investigation derives
from the demolition of highway asphalt pavements, Penetration (dmm), 18 52 43
characterized by a similar composition and located 100 g, 5 s at 25°C
in the Northern Italy. Conventional crushed lime- Softening point (°C), 64 48 98
stone, sand and filler were also utilized in order R&B method
to produce the bituminous mixtures studied in Fraass breaking - −18 −18
point (°C)
the experimental trials. Two different binders have
Viscosity at 160°C 0.256 0.13 1.33
been used: a bitumen modified with SBS (styrene–
butadiene–styrene) polymers and a conventional
one. All the materials have been supplied by private 2.2 Hot mix asphalts
companies located in the North-eastern Italian
Six asphalt concretes have been analysed and opti-
area. The conventional aggregates (LS—limestone)
mized in the experimental trials, with regards to the
were made available in four different fractions: 0/4,
acceptance requisites of bituminous mixtures for
4/8, 8/12 and 12/20 mm. The grading curves of the
road base courses.
natural aggregates (EN 933–1), along with that of
The conventional bitumen and the modified
RAP are illustrated in Figure 1. The RAP aggre-
binder were utilized for the preparation of two dif-
gate has been tested for the grading curve determi-
ferent types of asphalt concretes, associated to the
nation, after the extraction of the bitumen (white
code AC (Asphalt Concrete) and PMA (Polymer
curve, centrifugation method; EN 12697–1).
Modified Asphalt), respectively.
A bitumen content of 5.3% by weight of the
For each one of the two types of hot mix asphalts,
aggregate was determined, by means of the cold
three different bituminous mixtures, characterized
extraction test performed on the RAP. The bitu-
by an increasing RAP content and therefore by a
men of RAP was recovered through the ABSON
different lithic matrix, were considered. The mix-
method (EN 12697–3); the extracted binder has
tures AC/R0, AC/R2, AC/R4 have been prepared
been subsequently tested following the conven-
with standard bitumen and a RAP content of 0%
tional test methods: penetration at 25 °C (EN 1426),
(R0), 20% (R2) and 40% (R4), respectively. Simi-
softening point (Ring & Ball Method; EN 1427),
larly, the mixtures PMA/R0, PMA/R2 and PMA/
Fraass breaking point (EN 12593) and viscosity
R4, have been produced with a RAP quantity of
at 160°C (EN 13702–2). The results are reported
0%, 20% and 40%, but using the modified binder.
in Table 1; the experimental data determined for
In other laboratory researches it has been veri-
the conventional bitumen and the polymer modi-
fied the possibility to optimize asphalt concretes
fied binder are also presented. On the basis of
with higher RAP contents, up to 50% (Celauro et al.
the experimental data shown in Table 1, the RAP
2009, Colbert & You 2012, Miró et al. 2011, Pereira
binder results really aged and hard. A lower pen-
et al. 2004) 60% (Valdés et al. 2011) and 100%
etration, as well as higher values for the softening
(Oliveira et al. 2013, Silva et al. 2012); however,
point and the viscosity, have been observed for the
rejuvenator agents were used in order to improve
polymer modified binder with respect to the con-
the mechanical response of such mixtures.
ventional bitumen.
In the present experimental study, such agents were
not considered, in order to satisfy a direct demand of
the local road agencies (North-eastern area of Italy);
hence, the maximum quantity of RAP was assumed
equal to the 40% by weight of the aggregate.
However, in this research the attention was pri-
marily focused on the improvement of the fatigue
resistance of asphalt concretes produced with sig-
nificant quantities of RAP, rather than on pushing
the RAP dosage up to very high percentages.

3 METHODS

3.1 Mix design phase

A trial and error method (Pasetto & Baldo 2014a)
Figure 1. Grading curves of the aggregates. has been adopted to design the grading curves of

556

TIS2017_Book.indb 556 2/8/2017 8:07:19 PM

 the asphalt concretes. The different particle sizes of the residual air voids content were checked at
of the aggregates have been combined assuming as three different compaction conditions (10, 100 and
a reference grading envelope the one utilized in the 180 revs); instead, the mechanical strength has
North-eastern Italian area, for road base courses been evaluated at the design number of revolutions
(Veneto Strade 2012). Figure 2 illustrates the design (100 revs).
grading curves of the asphalt concretes and the The assumption of total blending between the
reference envelope adopted. The control asphalt binder of RAP and the virgin bitumen was taken in
concretes (AC/R0 and PMA/R0) were made only consideration in the mix design method (Al-Qadi
with limestone aggregates, to properly compare the 2007). The OBC corresponds to the virgin binder
mixtures prepared with RAP material. identified in the asphalt concrete optimization;
Cylindrical specimens (150 mm diameter and hence, for the mixtures produced with RAP mate-
60 mm height), have been produced using a gyra- rials, the total binder content has to be computed
tory compactor, in order to support the mix design as the sum of the OBC and the binder of RAP.
method. The fundamental gyratory parameters, The RAP aggregates were preheated for 2 hours
that are pressure, speed and angle of rotation, were at 90°C before the mixing phase. On the basis of
assumed equal to 600 kPa, 30 revs/minute and 1.25°, the CIRS specifications (CIRS 2001), the mixing
respectively. The ITS tests have been conducted at temperatures have been fixed at 150°C and 170°C
25°C, following the principal specifications of EN for the conventional binder and the hard modified
12697–23 Standard. In order to evaluate the mois- one, respectively.
ture resistance of the mixes, both dry and wet spec-
imens have been tested. The wet specimens have
3.2 Fatigue tests
been submerged for 15 days in water, by means of
a thermostatic bath. The ratio between the ITS of The four-point bending apparatus has been used
wet (ITSwet) and dry (ITSdry) samples expresses the for the fatigue tests, following the most relevant
Tensile Strength Ratio (TSR); such parameter was specifications reported in Annex D of the Euro-
determined for each asphalt concrete considered. pean EN 12697–24 Standard; hence, a continuous
On the basis of the CIRS design approach (CIRS sinusoidal waveform, with a frequency of 10 Hz,
2001), the volumetric and mechanical acceptance was used for all the trials. The bending tests were
thresholds presented in Table 2 have to be accom- carried out under constant strain conditions,
plished to design the optimum bitumen percentage. adopting three different strain values, within the
The volumetric characteristics, expressed in terms range 200–600 μm/m, to adequately capture the
fatigue response of each mix. The fatigue study
has been performed at 20°C, that is a significant
temperature for the fatigue evaluation of hot mix
asphalts in the north-eastern part of Italy. The
fatigue data, namely stress and strain, the phase
angle and dissipated energy, were obtained by the
apparatus for each loading repetition. For each
hot mix asphalt, slabs (300 × 400 × 50 mm) were
compacted by means of a laboratory compact-
ing roller, according to the main specifications of
the EN 12697–33 Standard. The beam specimens
(400 × 50 × 60 mm), to be used in the fatigue trials,
have been obtained from the cut of the slabs.
Most of the times, the fatigue resistance of
asphalt concretes is evaluated by means of the
Figure 2. Design grading curves of the mixtures.
classical stiffness reduction method (Celauro et al.
2009, Miró et al. 2011, Oliveira et al. 2013, Pereira
et al. 2004, Silva et al. 2012, Valdés et al. 2011).
Table 2. Mix design acceptance requisites. Such method is based on the empirical assump-
tion that at a 50% reduction of the initial stiff-
Parameter Threshold ness modulus, the beam specimen achieves the
Va at 10 revs 10–14% structural failure. In the current investigation the
Va at 100 revs 3–5% fatigue resistance was evaluated using such empiri-
Va at 180 revs >2% cal method, but also applying the dissipated energy
ITSdry >0.6 MPa method originally introduced by Carpenter & Shen
TSR >75%
(2005, 2006), which has been developed by means
of more rational hypotheses.

557

TIS2017_Book.indb 557 2/8/2017 8:07:19 PM

 3.3 Stiffness reduction approach
For a fatigue test carried out under the strain con-
trol mode, the strain is maintained at a constant
value; as a consequence, the stress decreases until
the failure of the specimen, that is conventionally
defined by a 50% reduction of the initial stiffness
modulus. Such criterion has been assumed also
in relevant experimental investigation reported in
the literature (Ghuzlan & Carpenter 2003, Khalid
2000).
In the present laboratory study, the conventional
fatigue curves have been developed combining the
initial value of strain ε0 and the number of loading
cycles Nf, at which a 50% reduction of the initial Figure 3. Determination of PV for mix AC/R2 at
stiffness is achieved. According to previous stud- 400 μm/m.
ies (Artamendi & Khalid 2005), the strain value
observed at the 100th cycle was used as initial
strain (EN 12697–24, Annex D). in such zone the quantity of input energy which
For the interpolation of the fatigue data, a undergoes a transformation into internal damage
power law model has been used: remains constant. In the third and last phase, the
RDEC increases fastly and eventually the specimen
ε 0 = aN
N bf (1) achieves the failure condition. The RDEC value
that remains constant during the second phase,
has been named Plateau Value (PV) of RDEC
where a and b represent the interpolation coeffi- (Carpenter & Shen 2005, 2006). Such value results
cients; their values depend on the type of asphalt really useful for a comparison of the fatigue resist-
concretes. ance of different hot mix asphalts, from an energy
based point of view; basically, the lower the PV, the
3.4 Dissipated energy approach greater the fatigue resistance of the mixture will be
The fatigue response were also analysed with the (Carpenter & Shen 2005, 2006, Pasetto & Baldo
energy method of Carpenter & Shen (2005, 2006), 2013b, 2014b, Shu et al. 2008, Yoo & Al-Qadi
recently also utilized in other investigations (Pasetto 2010). Figure 3 illustrates an example of the deter-
& Baldo 2013b, 2014b, Yoo & Al-Qadi 2010). In mination of PV for the mix AC/R2 at 400 μm/m.
such method, the Plateau Value (PV) of the Ratio In this research, different strain levels have
of Dissipated Energy Change (RDEC) represents been used in order to obtain the damage curves
the most important parameter for a rational evalu- (PV curves—PVC), expressed as a function of Nf
ation of the fatigue behaviour of asphalt concretes. (loading cycles involved for a 50% stiffness reduc-
The RDEC is computed using the formula: tion) and PV.
The interpolation of the PV—Nf data pairs, has
E n +1 − DE n
DE been performed using a power law model:
RDEC = (2)
DE En PV = aN
N bf (3)

where DEn and DEn+1 represent the dissi-

where a and b are regression coefficients; their val-
pated energy produced in load cycle n and n+1,
ues depend on the type of hot mix asphalt.
respectively.
It was argued that the RDEC is not influenced
from other dissipated energy types, related to
4 RESULTS AND DISCUSSION
mechanical work or heat development (Carpenter
& Shen 2005, 2006). Hence, the RDEC can be con-
4.1 Mix design results
sidered a rational parameter, useful to characterize
the internal damage developed within the asphalt For both the binder utilized in the trials, three dif-
concrete, during the fatigue trial. ferent aggregate structures were optimized. Each
The trend of RDEC in a fatigue test, can be asphalt concrete was designed with an equal quan-
subdivided in three different phases. During the tity of the coarser limestone material, that is the
first phase there is a sharp reduction of RDEC, fraction 12/20 mm. Conversely, for the other two
whereas a plateau zone, characterized by a con- limestone fractions (4/8 and 12/20 mm), the quanti-
stant RDEC value, is typical of the second phase; ties were different, in relation to the RAP content.

558

TIS2017_Book.indb 558 2/8/2017 8:07:19 PM

 The aggregate structure of the control asphalt CIRS thresholds. Furthermore, the wet ITS values
concretes (AC/R0, PMA/R0), was characterized by resulted also higher than the dry ITS threshold for
the 47% of sand, 7% of the limestone fraction 4/8, as all the mixes; hence, it has been verified the excel-
well as the 7% of the fraction 8/12, plus 5% of filler. lent moisture resistance of the asphalt concretes
The asphalt concretes containing 20% RAP investigated.
(AC/R2, PMA/R2) were made with the 25% of The total binder percentages identified by the
sand, 9% of both the limestone fractions 4/8 and mix design method were within the range typically
8/12. The filler was fixed at 3%. adopted for the base courses, that is 4.5–5.5% by
The introduction of 40% RAP in the aggre- weight of the aggregate (CIRS 2001), for all the
gate composition (mixes AC/R4 and PMA/R4), mixes. According to the results outlined in previ-
allowed to reduce the filler and the sand contents ous investigations, reported in the literature (Al-
to 1% and 8% respectively; instead, for the lime- Qadi et al. 2007, Ghuzlan & Carpenter 2003),
stone, the fraction 4/8 was utilized at 12% and the the new bitumen content necessary to satisfy the
8/12 at 5%. acceptance requisites resulted a decreasing func-
As it was expected, the greater the RAP content, tion of the RAP quantity; the greater the RAP
the lower the filler amount required to design the content, the lower the virgin binder to be added to
mixtures. the mixes. Such experimental evidence is congruent
Tables 3–4 report the results of the mix design with the assumption of a total blending between
phase. For each mix the most important design RAP binder and new bitumen.
properties have been reported: OBC (by weight No workability issues were observed during
of the aggregate); Air Voids (Va) at 10, 100 and the mixing of the components, namely bitumen
180 revs; bulk density at 100 revs; Indirect Ten- and aggregates, neither for the compaction of the
sile Strength (ITS) for dry and wet conditions, at mixes produced with RAP, for both the types of
100 revs. bitumen.
The bitumen content reported in Table 3 is the
virgin binder that resulted to be necessary for the
4.2 Fatigue analysis based on the stiffness
optimization of the asphalt concretes.
reduction approach
For all the hot mix asphalts, the air voids thresh-
olds fixed in the CIRS mix design method (CIRS Figure 4 shows the conventional fatigue curves,
2001), were completely accomplished, depending obtained by means of the stiffness reduction
on the particular OBC. Each of the asphalt con- method, for the asphalts concretes considered.
crete studied has presented strength properties, Table 5 reports the interpolation coefficients
namely dry ITS and TSR, much higher than the along with the coefficient of determination R2.
The tensile strain ε (106) was determined follow-
ing the specifications reported in the Standard EN
Table 3. OBC and volumetric properties.
12697–24, Annex D, for a fatigue life of 1,000,000
Asphalt OBC Va, 10 revs Va, 100 revs Va, 180 revs loading repetitions.
concretes % % % % Accordingly to the results presented in Table 5,
the polymer modified mixes were characterized
AC/R0 5.00 13.9 4.9 3.4 by a greater fatigue resistance, compared to the
AC/R2 3.94 13.6 4.5 3.3 asphalt concretes produced with the conventional
AC/R4 2.88 11.2 3.5 2.3 bitumen, in terms of tensile strain ε (106). This was
PMA/R0 5.00 14.0 5.0 3.8
PMA/R2 3.94 13.8 4.8 3.7
PMA/R4 2.88 11.3 3.8 2.6

Table 4. Physical and mechanical properties.

Asphalt Bulk density ITS dry ITS wet TSR

concretes Mg/m3 MPa MPa %

AC/R0 2.457 1.04 0.89 86

AC/R2 2.473 1.54 1.36 88
AC/R4 2.484 1.95 1.78 91
PMA/R0 2.451 1.62 1.46 90
PMA/R2 2.465 2.08 1.92 92
PMA/R4 2.477 2.32 2.16 93
Figure 4. Fatigue curves of the mixes investigated.

559

TIS2017_Book.indb 559 2/8/2017 8:07:20 PM

 Table 5. Regression coefficients of the fatigue curves
and ε (106) values.

a b ε (106) R2
Mixes μm/m – μm/m –

AC/R0 3687.1 −0.248 120 0.9972

AC/R2 2606.3 −0.209 145 0.9991
AC/R4 1501.8 −0.117 298 0.9378
PMA/R0 2655.9 −0.173 243 0.9562
PMA/R2 1779.3 −0.125 316 0.9665
PMA/R4 4009.9 −0.170 383 0.9911

verified also for the mixtures with RAP, especially Figure 5. Damage curves of the mixes investigated.
for the greater RAP content (40%). Indeed, the
fatigue curves of the polymer modified mixtures
are shifted towards higher loading cycles values; Table 6. Regression coefficients of the damage curves
moreover, they present a lower slope (Table 5), with and PV6 values.
respect to the control mixes. The improvement of
the fatigue performance due to the polymer modi- a b PV6 R2
fied bitumen was clearly expected, whereas the Mixes μm/m – – –
effect of the RAP on the fatigue life represents a
AC/R0 3.3201 −1.278 7,13101E-08 0.9976
new contribution on the characterization of the
AC/R2 3.5992 −1.306 5,25056E-08 0.9587
RAP mixes. Moreover, according to the results of
AC/R4 2.0827 −1.289 3,84261E-08 0.9999
Table 5, it has been verified that the combined use
PMA/R0 1.9145 −1.234 7,55189E-08 0.9984
of a polymer modified bitumen along with RAP
PMA/R2 0.6936 −1.203 4,19864E-08 0.9999
materials allows to achieve the highest fatigue life,
namely 383 μm/m for HM/R4, with an increment PMA/R4 1.0138 −1.231 4,16824E-08 0.9793
of 58% with respect to the control mix prepared
with limestone aggregates (HM/R0). It is worth of
mentioning that the increase in the fatigue life, due and the virgin binders, which probably allows the
to the use of RAP materials, resulted nonlinear development of a tough adhesion between the thin
with the RAP quantity; indeed, the increment was film of the virgin binders and the RAP grains.
higher with the 40% of RAP rather than the 20%. From another point of view, the thin film of aged
binder that covers the RAP particles could be con-
4.3 Fatigue analysis based on the dissipated sidered as an intermediate stiff layer between the
energy approach stiffer mineral grains of the RAP and the softer vir-
gin bitumen, so admitting a layered configuration
Figures 5 shows the damage curves, whereas within the RAP asphalt concretes. On the basis of
Table 6 reports the coefficients of interpolation and such structural configuration (Huang et al. 2005;
determination (R2); the fitting of the experimental Oliveira et al. 2013), the binder of the RAP con-
data resulted very good for all the mixes studied. tributes to reduce the stress concentration within
In similarity to the conventional approach, the PV the hot mix asphalt, with a consequent beneficial
for a 1,000,000 loading cycles (PV6) has been deter- effect on the fatigue behaviour.
mined, in order to synthetically quantify and com- However, the lowest PV6 values were obtained
pare the fatigue performance of the mixes. for PMA/R4 and PMA/R2, namely the mixes
With respect to the control asphalt concretes produced with both RAP and polymer modified
(AC/R0 and PMA/R0), the mixes made with RAP bitumen. The improved fatigue response shown by
aggregates were characterized by lower PV6 values, the polymer modified asphalt concretes with RAP,
namely, by a damage reduction. Depending on the could be due to the flexibility associated to the
RAP content, the PV6 reductions achieved the 42% polymers of the modified binder.
for the mixes prepared with conventional bitumen Currently, for road base asphalt concretes, the
and the 36% in case of PMB. Italian technical specifications (CIRS 2001) pre-
It has to be observed that the higher the RAP scribe a maximum RAP amount equal to 30%;
content, the better the fatigue resistance, for both furthermore, for such mixes, the use of polymer
the binders. Such results could be justified by the modified binders is not allowed. Hence, the feasibil-
chemical affinity between the bitumen of the RAP ity to improve the fatigue performance of polymer

560

TIS2017_Book.indb 560 2/8/2017 8:07:20 PM

 modified asphalts characterized by a conventional Carpenter S.H., Shen S. 2006. Dissipated energy
aggregate structure, by means of the integration of approach to study hot-mix asphalt healing in fatigue.
RAP, up to 40%, allows to meet the demand for Transportation Research Record 1970: 178–185.
sustainability and high performance of the road Celauro C., Celauro B. & Boscaino G. 2009. Production
of innovative, recycled and high-performance asphalt
constructions. for road pavements. Resources, Conservation and
Focusing the attention on the relative ranking Recycling 54(6): 337–347.
of the mixes (Table 6), the analysis performed with CIRS – Ministero delle Infrastrutture e dei Trasporti.
the dissipated energy approach, confirms the trend 2001. Capitolato speciale d’appalto tipo per lavori
already outlined with the conventional method. stradali. Italy (in Italian).
However, the dissipated energy approach relies on Colbert B. & You Z. 2012. The determination of
rational principles and not on an empirical failure mechanical performance of laboratory produced hot
assumption as the conventional method. mix asphalt mixtures using controlled RAP and virgin
aggregate size fractions. Construction and Building
Materials 26(1): 655–662.
5 CONCLUSIONS Ghuzlan K.A. & Carpenter S.H. 2003. Traditional
fatigue analysis of asphalt concrete mixtures. Proc.
Transportation Research Board 2003 Annual Meeting,
The experimental study described and discussed in
Washington DC, 12–16 January 2003.
this paper concerns the fatigue performance evalu- Huang B., Li G., Vukosavljevic D., Shu X. & Egan B.K.
ation of hot mix asphalts for road base courses, 2005. Laboratory investigation of mixing hot-mix
produced with RAP, up to 40% by the weight of asphalt with reclaimed asphalt pavement. Transporta-
the aggregate, and polymer modified bitumen. tion Research Record 1929: 37–45.
The fatigue data obtained by four point bend- Kavussi A. & Qazizadeh M.J. 2014. Fatigue charac-
ing fatigue tests, carried out at 10 Hz and 20°C, terization of asphalt mixes containing electric arc
have been analysed using the conventional stiff- furnace (EAF) steel slag subjected to long term aging.
ness reduction approach and the dissipated energy Construction and Building Materials 72: 158–156.
method, originally proposed by Carpenter. Khalid H. 2000. A comparison between bending and
diametral fatigue tests for bituminous materials.
With respect to the reference mixture, produced Materials and Structures, 33: 457–465.
using only natural aggregates, all the hot mix Miró R., Valdés G., Martínez A., Segurac P. & Rodríguez
asphalts with RAP materials have shown a better C. 2011. Evaluation of high modulus mixture behav-
fatigue behaviour; the improvements were even iour with high reclaimed asphalt pavement (RAP)
greater for the mixes prepared with polymer modi- percentages for sustainable road construction.
fied bitumen, for both the fatigue data analysis Construction and Building Materials 25: 3854–3862.
method adopted. Oliveira J.R.M., Silva H.M.R.D., Jesus C.M.G. & Abreu
Based on the results obtained with this experi- L.P.F. 2013. Pushing the Asphalt Recycling Technol-
mental study, a positive effect of the RAP materi- ogy to the Limit. International Journal of Pavement
Research Technology 6(2): 109–116.
als on the fatigue resistance of the hot mix asphalts Oluwasola E.A., Hainin M.R. & Aziz M.M.A. 2015.
considered has been outlined. Evaluation of asphalt mixtures incorporating electric
In qualitative terms, the comparative ranking of arc furnace steel slag and copper mine tailings for road
the mixes identified by the empirical approach and construction. Transportation Geotechnics 2: 47–55.
that established by the dissipated energy method Pasetto M. & Baldo N. 2012. Fatigue Performance of
was totally analogous. Asphalt Concretes with RAP Aggregates and Steel
The results obtained in the present paper repre- Slags. In: Scarpas A. et al. (ed.), Proc. 7th RILEM
sent a further validation of the dissipated energy intern. Conference on Cracking in Pavements, RILEM
method based on the Plateau Value concept and Bookseries, 4: 719–727, Delft, 20–22 June 2012.
Dordrecht: Springer Netherlands.
contribute to clarify the importance of a rational Pasetto M. & Baldo N. 2013a. Resistance to permanent
approach for the fatigue evaluation of road bitu- deformation of road and airport high performance
minous mixtures. asphalt concrete base courses. Advanced Materials
Research 723: 494–502.
Pasetto M. & Baldo N. 2013b. Fatigue Performance of
REFERENCES Asphalt Concretes made with Steel Slags and Modi-
fied Bituminous Binders. International Journal of
Al-Qadi I.L., Elseifi M. & Carpenter S.H. 2007. Reclaimed Pavement Research and Technology 6(4): 294–303.
Asphalt Pavement—A Literature Review. Technical Pasetto M. & Baldo N. 2014a. Influence of the aggre-
Report FHWA-ICT-07-001, Federal Highway Admin- gate skeleton design method on the permanent defor-
istration (FHWA), Washington, DC, USA. mation resistance of stone mastic asphalt. Materials
Artamendi I. & Khalid H. 2005. Characterization Research Innovations 18(3): S96-S101.
of fatigue damage for paving asphaltic materials. Pasetto M. & Baldo N. 2014b. Fatigue performance and
Fatigue & Fracture of Engineering Materials & stiffness properties of Stone Mastic Asphalts with
Structure 28: 1113–1118. steel slag and coal ash. In: Kim Y.R. (ed.), Asphalt

561

TIS2017_Book.indb 561 2/8/2017 8:07:21 PM

 Pavements—Proc. of the intern. Conference on Asphalt and Characterization of Sustainable and Innovative
Pavements, ISAP 2014, 1(82): 881–889, Raleigh, 1–5 Bituminous Materials, RILEM Bookseries, 11: 927–
June 2014. Boca Raton: CRC Press, Taylor & Francis 939, Ancona, 7–9 October 2015. Dordrecht: Springer
Group. Netherlands.
Pereira P.A.A., Oliveira J.R.M. & Picado-Santos L.G. Valdés G., Pérez-Jiménez F., Miró R., Martínez A. &
2004. Mechanical characterisation of hot mix Botella R. 2011. Experimental study of recycled
recycled materials. International Journal of Pavement asphalt mixtures with high percentages of reclaimed
Engineering 5(4): 211–220. asphalt pavement (RAP). Construction and Building
Shen S., Carpenter S.H. 2005. Application of the dissi- Materials 25: 1289–1297.
pated Energy concept in fatigue endurance limit testing. Veneto Strade. 2012. Capitolato speciale d’appalto tipo
Transportation Research Record 1929: 165–173. per la manutenzione e la costruzione delle infrastrut-
Shu X., Huang B. & Vukosavljevic D. 2008. Labora- ture stradali. Italy (in Italian).
tory evaluation of fatigue characteristics of recycled Wu S., Xue Y., Ye Q. & Chen Y. 2007. Utilization of steel
asphalt mixture. Construction and Building Materials slag as an aggregate for stone mastic asphalt (SMA)
22: 1323–1330. mixtures. Building and Environment 42: 2580–2585.
Silva H.M.R.D., Oliveira J.R.M. & Jesus C.M.G. 2012. Xue Y., Wu S., Hou H., Zhu S. & Zha J. 2009. Utiliza-
Are totally recycled hot mix asphalts a sustainable tion of municipal solid waste incineration ash in stone
alternative for road paving?. Resources, Conservation mastic asphalt mixture: pavement performance and
and Recycling 60: 38–48. environmental impact. Construction and Building
Stimilli A., Virgili A., Giuliani F. & Canestrari F. 2016. In Materials 23: 989–996.
plant production of hot recycled mixtures with high Yoo P.J. & Al-Qadi I.L. 2010. A strain-controlled hot-
reclaimed asphalt pavement content: A performance mix asphalt fatigue model considering low and high
evaluation. In Francesco Canestrari & Manfred N. cycles. International Journal of Pavement Engineering
Partl (ed.): Proc. 8th RILEM intern. Symp. on Testing 6: 565–574.

562

TIS2017_Book.indb 562 2/8/2017 8:07:21 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Correlation between risk level and operating speed of road section

D. Bellini & M.C. Iaconis

Regione Toscana—Progettazione e Realizzazione Viabilità Regionale Pisa-Siena-Pistoia, Italy

M. Rossi
Dipartimento di Ingegneria Civile e Industriale, Università di Pisa, Pisa, Italy

ABSTRACT: The use of accident data has always been a challenge for the management of road
networks in Italy. On one hand, there were many attempts to find theoretical models to interpret the
data of accidents, while on the other arose the difficulty of finding quality data for targeted studies,
especially for the rural road networks. In recent years however, the possibility of finding accident data
was used to develop a useful procedure at a planning level to identify the road sections on which further
investigation should be carried on, in order to improve road safety conditions. The results obtained with
the application of this procedure to some roads under the jurisdiction of the Tuscany Region Road
Authority made possible to build some bubble diagrams, which allow an immediate and complete view
of all homogeneous sections of the road network. For those sections for which was obtained a high
risk level, deeper investigations were performed by measuring the operating speed, in order to evaluate a
relationship between those detected discrepancies between the design conditions and the real operating
condition of those homogeneous sections (correlated with the perception of the infrastructure by drivers)
and the risk of these sections themselves, allowing the planning of those actions necessary in order to
increase road and traffic safety level.

1 INTRODUCTION as the construction of a road graph for homogene-

ous road sections, made it possible to determine the
The use of accident data has always been a difficult average annual rate of accidents for each homoge-
challenge for the Road Authorities in Italy. On the neous section, on the basis of some indicators, as
one hand many attempts to find theoretical mod- described in the Final Report “Criteria for the clas-
els to interpret the data of accidents were carried sification of the network of existing roads”, drawn
on, on the other it came up the difficulty of find- up and approved by the CNR with DP CNR n.
ing quality data for targeted studies and, especially 13465 of 11/09/1995 and published on 13/03/1998,
for the secondary road networks, the difficulty to as described in the following paragraphs.
obtain updated data on traffic flows. With all these data available, a correlation
In recent years however raised up the possibility between risk level and speed data of homogeneous
to obtain a set of data that, combined, permitted section was found.
to define a useful procedure in the planning level
in order to identify those road sections on which
2 ROAD SAFETY CONDITION ANALYSIS
further investigation must be carried on, in order
AT A NETWORK LEVEL
to improve road safety conditions, as also stated
into the Italian Road Safety Guidelines (D.M. n.
2.1 C.N.R. procedure to evaluate accident rate
182/2012), from which the entire study moved on.
and to classify each homogeneous sections
This was possible thanks to SIRSS project: it
within a road
consists in the normalization of information about
the location of the accidents in accordance with a The accident rate analysis is part of the characteriza-
road graph, in addition to the georeferencing of tion of each road within a network in terms of road
each accident. safety, which enables the identification, in relation
The analysis here performed is the starting point to a statistically significant comparison threshold,
in the developing of a new procedure: data along the accident rate involving the entire road.
14 homogeneous section on regional road S.R.68 The analysis was here performed for 14 homo-
“della Valdicecina” within the Province of Pisa geneous section on regional road S.R.68 “della
were available. Georeferencing of accidents, as well Valdicecina” within the Province of Pisa. In addition

563

TIS2017_Book.indb 563 2/8/2017 8:07:22 PM

 to calculating the accident rate following the CNR Each homogeneous section is classified as with
report quoted above, the analysis was deepened low, medium or high accident rate on the bases of
with the calculation of the density of accidents. the Ti value evaluated with the equation (1) com-
In order to classify the entire road on the base pared with the thresholds evaluated with equations
of the accident rate, the procedure indicated in the (2) and (3).
CNR report requires the following information: The obtained results made it possible to build
some bubble diagrams, which allow an immediate
− georeferenced accidents data, including the date,
overview on each homogeneous section within a
number and type of vehicles involved; only acci-
road or an entire network, highlighting those with
dents with people injured were considere. All
higher risk level. These diagrams have, for each
these data were provided by SIRSS in a specific
homogeneous sections, on the abscissa the values
database;
of the accident density Di of each homogeneous
− division into homogeneous section of the entire
section, measured in number of accidents per km
road;
per year, and in ordinate the Vi values of the speed
− TGM of each hgomogeneous section, referred
(operating speed of project speed) of each homo-
to each specific year.
geneous section, while the area of the bubbles is
The accident rate of each homogeneous section the accident rate Ti of the homogeneous section.
is evaluated as the number of accident in relation to The choice to build this type of representation
one million vehicles per kilometer, and is given by: raised by the necessity to evaluate the risk level of each
infrastructure which the homogeneous section belongs
106 ⋅ Ni apart from the TGM, that represents the road users
Ti = (1)
365 ⋅ li ∑ t TGM i ,t risk level component. In particular, with this type of
representation it was tried to summarize the informa-
tion regarding the contribution of inhomogeneity of
with: radius of curvature within paths with many bends.
Ni number of accidents with injuries On each diagram it is possible to identify 4 differ-
occurred in the tth year within the ent areas, on the base of the speed of 70 km/h (speed
analysis period considered on the ith over which the italian standard D.M. 2367/2004
homogeneous section; confirms the mandatory installation of crash bar-
Li length (km) of the ith homogeneous riers, identifying this speed value as a discriminat-
section; ing value above which it can be considered that the
TGMi,t Average daily traffic on the ith damage resulting from an impact is more than pro-
homogeneous section in the tth year portional to the speed itself) and on the base of the
within the analysis period. average accident density Dm evaluated on a reference
base (national, regional, network, single road):
The CNR report provides also instructions
about the identification of a statistically significant − Low risk level area: Vi ≤ 70 km/h, Di ≤ Dm
reference threshold in order to classify homogene- − Medium risk level area: Vi > 70 km/h, Di ≤ Dm
ous section as, low, medium and high accident rate, − High risk level area: Vi ≤ 70 km/h, Di > Dm
based on the results obtained by the accident rate − Very high risk level area: Vi > 70 km/h, Di > Dm
evaluation as stated above:
The bubbles area is proportional to the accident
Tm 1 rate of the road section: bigger is a bubble, higher
Tin*f Tm − K ⋅ − (2) is the accident rate of the section to which bubble
Mi 2 Mi refers.
It was also necessary to identify three different
Tm 1
Tin*f Tm + K ⋅ + (3) ranges for the bubbles area, to make it easier to
Mi 2 Mi read the graph by assigning to each of them a dif-
with: ferent color.
In the following Figure 1 an example of the bub-
K Poisson’s probability distribution constant, bles graph is reported.
assumed with the value 1.645 in order to have This division is based on the evaluation of the
an error probability less the 10%; average accident rate Tm on the same basis used for
Mi traffic momentum of the ith homogeneous the evaluation of Dm:
section within the entire analysis period, eval-
− Low accident rate: GREEN color, Ti ≤ 2 ⋅ Tm
uated as:
− Medium accident rate: YELLOW color,
2 ⋅ Tm < Ti ≤ 3 ⋅ Tm
Mi li ⋅ ∑ t TGM
M i ,t (4)
− High accident rate: RED color, Ti > 3 ⋅ Tm

564

TIS2017_Book.indb 564 2/8/2017 8:07:22 PM

 This research focused to find a possible correla-
tion between the accident rate and some charac-
teristics (geometrical, mostly) of the road path: in
this way, it could be possible to plan some design
changes in the road path and to evaluate how much
the safety condition will improve.
The research starts from the assumption that the
right perception of the path (and then the speed
maintained by users along the road) is recognized
as one of the main reason at the base of high acci-
dent rates.
It was assumed that each user drive along a
bend at a constant speed, and he accelerates and
decelerates along the following straight and before
the following bend as he recognizes the layout of
Figure 1. Example of a bubbles graph.
the road path.
Then, a correlation between the accident rate
and the difference between the speed theoretically
The analysis of the bubbles diagrams make it attainable at the end of a straight and the design
possible to identify those situations marked by speed of the following bend, under the hypothesis
higher anomaly than the average of the entire net- of linear deceleration assumed equal to 0.8 m/s2
work (or single road) analyzed. as stated in the Italian Standard D.M. 6792/2001
In most case, the anomaly higher than the aver- (“Norme funzionali e geometriche per la costruzi-
age of the entire network is caused by the wrong one delle strade”).
perception of the path by the road users, with the The first step was to find a model to evaluate
results that it is hard for them to drive at a “safe” the design speed of each bend. This is a theoretical
speed along the entire road or on its some specific speed and it can be evaluated by taking in account
section. what is reported in the Italian Standard D.M.
For these situations it is necessary to deeper the 6792/2001 (“Norme funzionali e geometriche per
investigation in order to identify those factors, such la costruzione delle strade). It can be evaluated for
as the geometry of the homogeneous section, the each bend on the road path from:
composition of traffic flows, environmental condi-
tions prevailing, etc. responsible of the anomaly. Vp2
The benefit is twofold: not only to go to pri- =q fT (5)
R ⋅ 127
marily analyze situations with higher absolute
numbers regarding the amount of people involved with:
but, among these, the ones that have the greatest
chances to be improved because suffering from q transversal slope
abnormalities of some kind. fT maximum transversal grip coefficient, inversely
In particular, it seems useful to consider prima- proportional to the design speed
rily the situations relating to red bubbles in the very R radius of curvature
high risk level area of the diagram as representa- VP design speed.
tive of road sections characterized by high number
Knowing the radius of each bend and the length
density of accidents, high accident rates and high
of each straight along a road path, both evaluable
speed.
from technical maps C.T.R. or using special vehicles,
In general, for the yellow and red bubbles some
the theoretical speed Vend at the end of each straight
actions must be planned in order to improve the
was evaluated using the equations of uniformly accel-
road safety conditions, like reinforcement of road
erated motion and under the following hypothesis:
signs, control activities by traffic police organs,
awareness campaigns, new design of the road, etc. − if the stop distance LSTOP (evaluated with the pro-
cedure reported in “Il tracciato stradale: dinam-
iche di percorrenza e visibilità”, F.S.Capaldo,
2.2 Correlation between accident rate and
2005) is lower than the length LST of the straight,
geometrical characteristics of the road path
it was assumed that the driver will start to brake
Once the accident rate analysis is completed and when he will be at a distance of LSTOP from the
bubbles graphs are built, it is possible to identify following bend;
those sections for which some action should be − if the stop distance LSTOP is higher than the
planned. length LST of the straight, it was assumed that

565

TIS2017_Book.indb 565 2/8/2017 8:07:23 PM

 the driver will start to brake when he will be at a son correlation coefficient r (where the X series was
distance of LST from the following bend. the ΔV, and Y series the accident rates), it was found
a value of 0.317. The critical value under two vari-
Once the speed Vend at the end of each straight
ables and n-2 degrees of freedom, at a significance
is evaluated, it can be compared with the speed of
level of 95%, is 0.532: this means that it seems no
the following bend:
correlation is existing between ΔV and accident rate.
Then a second step in the analysis was per-
ΔV Va − Vend (6)
formed. Since 6 homogeneous sections were newly
designed and built in the period 2004–2009, a new
with: analysis was performed considering two different
Va design speed of the bend following the groups: one with those sections newly designed,
straight and one with all the other sections (where only
Vend speed evaluated at the end of the straight. ordinary maintenance was performed).
An F-test was run considering the accident rate
From the comparison, three different cases are of the two different groups, to investigate if the two
possible: groups must be treated as different independent
groups of if they can be treated as a single group.
1. The theoretical speed at the end of the straight
The result obtained was Feval = 101.382.
is lower than the design speed of the following
Entering the Fcrit critical value table for two data
bend: in this case, the user has the time to drive
series, each one with n-1 degrees of freedom and
along the straight for a certain time before start-
with a significance level of 95%, it was found the
ing to decelerate, in order to enter the following
value of 4.88. Since Feval > Fcrit, it can be affirmed
bend at the right speed;
that the two groups are totally independent (at
2. The theoretical speed at the end of the straight is
least from the point of view of accident rate), and
equal to the design speed of the following bend
they must be treated and analyzed independently.
was looked: it is a limit condition, where the user
Then, the Pearson correlation coefficient r was
must start to decelerate immediately after he enters
evaluated again separately for both groups, even
on straight, otherwise he will enter the following
considering the ΔV as the X series and the acci-
bend at a speed higher than the design speed;
dent rate as the Y series. The results obtained are
3. The theoretical speed at the end of the straight
reported here below:
is higher than the design speed of the following
bend: it is the worst condition, because even if − rND: 0.81
the driver start to decelerate just after he enters − rold: 0.29
on the straight, he will not have the possibility
where rND is the coefficient evaluated for the group
to enter the following bend at a safe speed.
constituted by the newly designed sections and rold
Of course, the case n. 3 is the worst and the most the group of all the other sections.
risky, and it is the one that should be analyzed and The results show that there is a really good cor-
correlated with accident rate. relation between ΔV and the accident rate for those
The research of the correlation started from the newly designed sections: within this group, at lower
analysis of available data. Actually, accident rate ΔV values correspond lower accident rate values.
in the period 2010–2013 data and speed data on Some comments on this results are reported in
S.R.68 “della Valdicecina” (within the Province of chapter 3.
Pisa) are available.
Previously from analysis were excluded those
homogeneous section within urban areas: here, 3 CONCLUSIONS AND FURTHER
there are too many variables to be considered to DEVELOPMENT
evaluate a risk level, as pedestrian crossings, pri-
vate access on the road, public lighting, etc. In this study authors tried to provide a procedure
At the end, data on 24 homogeneous section to manage accidents data available (and georefer-
were available: excluding the urban section, the enced) in order to make it a useful instrument to
analysis was performed on 14 sections, where the plan actions along specific sections of a road (or a
accident rate was evaluated. network), in order to consistently improve the road
Then, according to what stated above, all the safety condition; from another point of view, such a
positive speed differences (previous point n. 3) procedure could make it also available to forecast the
were summarized along a homogeneous section entity of the road safety condition improvement.
and normalized it on standard length of 1,000 m. The authors investigate the correlation between
In the first step, it was investigated the correlation geometrical characteristics of a homogeneous
considering all the 14 sections: evaluating the Pear- section along a road path (here taken in account

566

TIS2017_Book.indb 566 2/8/2017 8:07:24 PM

 by considering a normalized speed difference ΔV traffic volume is, indeed, a key factor contributing
between the speed theoretically attainable at the to accident rates according to the statistical CNR
end of a straight and the design speed of the fol- model.
lowing bend) and the accident rate evaluated on However, to find a mathematical relationship
the same section, in order to develop a possible able to forecast the lowering of accident rate as a
procedure to help road authorities to focus and to consequence of a new design, on the basis of the
plan on actions on some homogeneous sections of correlation found here, it is necessary to deep the
their network. The correlation, was investigated by research by developing a new step: indeed, it is
analyzing data available for the 14 homogeneous necessary to extend the analysis on more samples,
sections of S.R.68 “della Valdicecina”. and—more important—also to develop a model
The result from an F-test, run onto accident rate able to consider all those parameters listed above.
data, shows and confirm that those sections newly Once such a model will be available, it will be a
designed and built (according the Italian Standard) powerful help for public administration to under-
must be treated as independent group. stand how much specific actions could lower the
The Pearson correlation coefficient values, eval- accident rate on road sections, or how much spe-
uated for both groups of sections (newly designed cific action could raise the safety condition.
and old), are here reported:
− rND: 0.81
− rold: 0.29 REFERENCES
where rND is the coefficient evaluated for the group Bellini D. et al. (2016), Migliorare la sicurezza delle reti
constituted by the newly designed sections and rold extraurbane secondarie, Strade & Autostrade n. 116,
the group of all the other sections. Gennaio/Febbraio 2016.
These results shows that many parameters affect Bordin M. & M. Stefanutti (1997), Analisi dinamica delle
the accident rate within a certain road section: norme tecniche CNR sulle curve stradali, Strategie e
indeed, the low rold value means that there is no cor- strumenti dell’ingegneria delle infrastrutture viarie—
Convegno SIIV, Pisa.
relation between ΔV and accident rate within old Busi R. & L. Zavanella (2002), La classificazione funzion-
sections of S.R.68, or from another point of view, ale delle strade, EGAF, Forlì.
the accident rate must be correlated also to some CNR (1998), Criteri per la classificazione della rete delle
other characteristics (e.g. visibility distance, road strade esistenti ai sensi dell’art. 13, comma 4 e 5 del
path altimetry, the presence of side entrances on nuovo codice della strada, Roma.
the road, the presence of dangerous intersections, Figueroa Medina A. (2004), Reconciling Speed Limits
etc.). Instead, the high rND value means that there with Design Speeds, JTRP Technical Reports, INDOT
is a good correlation between ΔV and accident rate Division of Research, West Lafayette, Indiana.
within newly designed sections of S.R.68, because Francesco Saverio Capaldo (2005), Il tracciato stradale:
dinamiche di percorrenza e visibilità.
of all other “dangerous” characteristics are no Giuffrè O. et al. (1997), Contributo alla definizione dei
more present along the road path (the aim of a criteri di classificazione della viabilità esistente, Strat-
new design of one or more road section should egie e strumenti dell’ingegneria delle infrastrutture
be to remove as much as possible all the danger- viarie—Convegno SIIV, Pisa.
ous characteristics listed above): according to the Marchionna A. et al. (2011), Application of design con-
results obtained, on newly designed road section sistency evaluation tools for two-lane rural roads:
only geometrical data are correlated to accident a case study from Italy, Transportation Research
rate data. Board 90th Annual Meeting, Compendium of Papers
It is also relevant to note that the accident rates No. 11-0689, Washington.
Ministero delle Infrastrutture e dei Trasporti (2012),
used within this research are data that already take Monitoraggio del PNSS: Linee Guida per la valutazi-
account of the traffic volume, major factor in acci- one dei risultati degli interventi di Sicurezza Stradale, a
dents government. The main factor contributing cura di L. Persia, D. Usami, Direzione Generale per la
to accident rates, the traffic volumes, is therefore Sicurezza Stradale, Roma.
already considered within the parameter chosen Rossi M. (2014), Key elements to draw up a traffic plan
for analysis, for which the highest correlation in the for rural roads, PhD thesis, Pisa.
accident rate is certainly with the volume of traffic:

567

TIS2017_Book.indb 567 2/8/2017 8:07:24 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport systems

TIS2017_Book.indb 569 2/8/2017 8:07:26 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Implementing innovative traffic simulation models with

aerial traffic survey

A. Marella, A. Bonfanti & G. Bortolaso

Trafficlab (Progectolab Group), Alba, Italy

D. Herman
RCE systems s.r.o., Brno, Czech Republic

ABSTRACT: Traffic data and human driver behaviours simulation are two of the most important
parameters for the proper implementation of a traffic simulation. In our case study, we implemented
an innovative method to obtain both a complex set of data of OD matrix and detailed human driver
behaviours data in order to set a specific scenario simulation. Firstly, we recorded an aerial video and,
subsequently, we conducted an advanced traffic analysis of aerial video data using DataFromSky service.
The result is a complex complete set of traffic parameters. Data collection is based not only on classic
OD matrix (volume data, vehicles classification, hourly rates, etc.), but also on innovative dynamic vehicle
database parameters (speeds, lateral and tangential accelerations, travel times and distances, trajectories,
etc.). All of these parameters are supplied to each vehicle detected. For this case study, our team provided
as well a set of specific simulation parameters, such as gap time and follow-up time, for each entry/exit
analysed. It is clear that most of this data were used to the calibration and improvement of our simulation
network. In particular, we inserted specific values for each vehicle class (dimensional and dynamic values)
and a detailed calibration of our behavioural models in Krauss settings. The resulting traffic simulation
scenario shows the highest correlation value between real and simulated driver behaviours, probably, never
obtained before.

1 INTRODUCTION separated by a painted traffic divider between

them, slightly curved in the middle of the two
The dataset used for this case study come from a straight segments. The survey area was recorded
survey that took place in Sheffield (UK), on the by means of a UAV by Vertex Access, for an inter-
A6178, called Sheffield Road, that connects the val time of nearly 10 minutes, on a segment of
city with Rotherham (Figure 1). The road within almost 250 meters placed between Temple Road
the investigated section is one lane per direction, and Grange Lane enter onto the main road. In
fact, the actual distance that was considered for the
data extraction is less than that, since it depends on
the exact place where the entry and exit gates were
positioned. In particular, the examined length was
about 229 meters.
Once the video has been recorded from above
the study area, DataFromSky team allows to
collect a large amount of data for each of the
detected vehicle passed within it: this tool allows
not only to know exactly how many vehicles
entered and exited the area, but also at which
time and how many seconds it needs to cover
the distance, and so the travel time. The dataset
doesn’t restrict to that, providing also the speeds
and accelerations measured for each timestamp
(from second up to milliseconds) between an
entry and an exit gate.
The entire set of parameters become useful in
Figure 1. Study area localization. order to calibrate and implement the behavioural

571

TIS2017_Book.indb 571 2/8/2017 8:07:26 PM

 car-following model, that defines the speed of a Unmanned Aerial Vehicle (UAV) or an aerostatic
vehicle in relation to the vehicle ahead: a modifi- balloon, or simply shot from high-rise.
cation of the Krauss model is used as default by Thanks to long-term development of specialized
the simulation software SUMO (Open Source algorithms for detection and tracking, DataFromSky
Licensed). The idea on which the models work, is able to provide comprehensive information about
both the original and the modified one, is letting the behaviour of road users merely from aerial video
vehicles drive as fast as possible while maintaining data taken by an ordinary camera. Indeed, unlike in
perfect safety, being always able to avoid a collision many commonly used systems for traffic monitor-
if the leader starts braking. ing, the only requirements for detailed analysis of
Through the software SUMO the network were road user behaviour by DataFromSky tool is an aer-
loaded from OpenStreetMap, and four detectors ial video and a description of the recorded scene.
were placed on lanes in the same position as the Our detection algorithms are capable of recog-
ones used with DataFromSky, so that the com- nizing vehicles as small as 16 × 16 pixels to allow for
pared trajectory, the real and the simulated ones, coverage of a large area by one device only (more
have the same length. The comparison among dif- than 30,000 m2). As high-quality video stabilization
ferent scenarios was implemented by the analysis algorithms are an integral part of DataFromSky
of the travel times, that provide an overall measure tool, the sensing platform is not required to be
of vehicles, alone or in a group, driving into the 100% stable during the recording.
network. The term of comparison (scenario no.0) The overall system can be divided into three
was represented by the real investigation of the main parts:
travel time, provided by the automatic calculation
of DataFromSky. − Pre-processing
Finally, two different simulations were devel- − Vehicle detection
oped with SUMO—Simulation of Urban Mobility: − Tracking
the scenario no.1, with the default values assigned
In the pre-processing step, the acquired image
by the simulator to the ‘car’ vehicle (only this type
gets undistorted and geo-registered against a user-
of vehicle has been considered in this case study),
selected reference frame. The efficiency and robust-
and the scenario no.2, where the aggregated values
ness of vehicle detection methods in aerial images
of speed and acceleration got from the DataFrom-
have been addressed several times in the past.
Sky analysis was used to better define the vehicle
DataFromSky developers designed a highly opti-
driving behaviour, showing improvement in the
mized multi-layer detector and trained it on a hand
output simulation.
annotated training dataset with 80,000 positive and
80,000 negative samples. The detector is capable of
learning from its errors and improving over time.
2 WHAT IS AND WHAT DOES
The tracking part is an extremely challenging task
DATAFROMSKY
due to the presence of noise, occlusion, dynamic
and cluttered backgrounds, and changes in appear-
Every year, traffic jams and accidents inflict huge
ance of the target. A subtype of sequential impor-
costs worldwide. The most effective strategies for
tance sampling filter has been employed to deal
dealing with these notorious problems include
with these challenges. Due to the outstanding per-
improvements in traffic management, signalizing
formance of detection and tracking algorithms,
systems and the quality of road network. However,
DataFromSky team can provide high-quality data
to design and implement the changes successfully,
and guarantee a hit rate of over 96%.
a large amount of data is needed. Conventional
Each vehicle is labelled by a unique ID and all
approaches to data collection offer simple meas-
its positions, speeds, and accelerations are recorded
urements of multiple quantities, often disjointed.
during its passage through the monitored section.
DataFromSky offers an innovative approach to
Thanks to extraction of spatiotemporal informa-
the traditional problems—the use of aerial video
tion about each vehicle in the scene, DataFromSky
which brings numerous advantages.
offers unlimited possibilities in data interpretation
DataFromSky (Apeltauer et al., 2015) is a service
and inspection.
based on automatic extraction of time-space trajec-
The service offered by DFS is thus articulated
tories of vehicles from aerial video. A wide range of
into the following steps:
traffic parameters, such as speed, accelerations and
gate counting, are derived through an advanced − The data collection is performed by external
computer vision analysis of the image. There is no partners operating drones/UAVs commercially
need for expensive equipment, sophisticated sensor in the target areas.
and time-consuming installation. The video can be − The recorded videos are automatically
collected using any small aerial platform, be it an analysed by the Extractor tool developed by

572

TIS2017_Book.indb 572 2/8/2017 8:07:27 PM

 Figure 3. DataFromSky output video data example.

Figure 2. DataFromSky analysis steps.

Figure 4. DataFromSky analysis data example.
DataFromSky. It is used to extract trajectories
and classifications.
− The output is delivered to the client together package including the simulation application itself
with software for further analysis and data inter- as well as supporting tools, mainly for network
pretation—DataFromSky Viewer. import and demand modelling. Through a three
− DataFromSky developers can modify the soft- steps elaboration process, the micro simulator
ware according to any needs and requirements. reproduces and then forecasts how a given traffic
demand distributes and behaves within a road net,
DataFromSky also includes a tool for advanced on the basis of the geometric characteristics and
data analysis. A virtual counting gate may be intersection regulation of the roads, considering
defined anywhere in the input video to establish the the interaction among different types of vehicle.
number of vehicles in this place. Further, analysists An example of the procedural phases needed
can also obtain detailed information about traffic to perform a traffic simulation is shown in the fol-
lane utilization, sectional measurements, average lowing graph (Figure 5): the road network can be
speeds, observed anomalies, etc. (Figures 3 and 4). either generated using netconvert or netgenerate,
The tool also provides advanced functions for traf- whereas od2trips, duarouter, dfrouter, jtrrouter are
fic engineers such as Tg (time gap) and Tf (time the applications that allow to describesontoment
to follow) estimations, and computation of vehicle of vehicles, each of which is defined at least by a
lateral acceleration. unique identifier, the departure time and the vehi-
Thanks to the advanced technology of cle’s route through the network.
DataFromSky, application and research fields are Additional variables may add further devices
very wide. and components to the simulation, or more accu-
rately define specific part in order to refine and
3 WHAT IS AND WHAT DOES SUMO improve the quality of the simulation.
By means of these inputs, the application sumo/
“Simulation of Urban Mobility” (SUMO) (Jakob sumo-gui performs a time-discrete simulation,
et.al. 2012) is an open source traffic simulation whose model is space-continuous and where each

573

TIS2017_Book.indb 573 2/8/2017 8:07:27 PM

 ⎡ v f (t ) + vl (t ) ⎤
a = vl (t )
vsafe [ gn (t ) vl (t ) ] / ⎢ + τ ⎥;
⎣ 2b ⎦
gn (t ) = xl (t ) x f (t ) mi G Gap

vsafe = safety speed

gn = distance between follower and leader
b = deceleration
xl = leader position
xf = follower position
Figure 5. Common Network and Demand generation The main input parameter for the Krauss
procedures. model’s specification in SUMO, associate to every
defined type of vehicle, are:

vehicle’s movement through the network is com- − Acceleration [m/s2]: the acceleration ability of
puted using a car-following model. vehicles of this type that a driver chooses;
Car-following models usually compute a vehi- − Deceleration [m/s2]: the deceleration ability of
cle’s speed by looking at the maximum allowable vehicles of this type that a driver chooses;
speed, its distance to the leading vehicle and the − σ: the driver imperfection (between 0 and 1);
leader’s speed. In particular, SUMO uses an exten- for values above 0, drivers with the default car-
sion of the stochastic car-following model devel- following model will drive slower than would be
oped by Krauss (1998) as default, because of its safe by a random amount;
high execution speed and simplicity. − τ [s]: the driver’s reaction time; drivers attempt
Indeed, speed, acceleration and deceleration of to maintain a minimum time gap of τ between
every single vehicle can’t be the ones desires, as if the rear bumper of their leader and their own
each of them was alone and free to choose its own front-bumper + minGap;
free flow travel speed. On the contrary, the behav- − minGap [m]: empty space after leader;
iour of each driver is influenced by the other’s driv- − maxSpeed [m/s]: the vehicle’s maximum speed;
ing, even more as the flow increases, and most of − SpeedFactor and SpeedDev: the first defines the
all by the leader’s. vehicles expected multiplicator for lane speed
Assuming two cars following each other on the limits, whereas the second represent its relative
same road, the acceleration profile of the following standard deviation, so the ratio of the standard
car is just a function of the relative speeds between deviation to the mean; they are used to sample
the two cars, the follower (f) and the leader (l): a vehicle specific chosen speed from a normal
distribution, with mean the one calculated with
∂v f (t ) the SpeedFactor and a deviation speedDev. A
= λ ⋅ [(v f (t ) v f (t )] vehicle keeps from the distribution its chosen
∂t SpeedFactor for the whole simulation and mul-
tiplying it with edge speeds can compute the
vf = follower’s speed actual speed for driving on each edge. Thus vehi-
vl = leader’s speed cles can exceed edge speeds, although vehicle
λ = sensibility speeds are still capped at the vehicle type’s max-
τ = reaction time Speed. Using speed distributions is highly advis-
t = time of analysis able to achieve realistic car following behaviour:
In the modified Krauss model working on if all vehicles have the same maximum speed on
SUMO the correlation among vehicles is defined any given road, they will not be able to catch up
by considering for the following vehicle the maxi- with their lead vehicle causing unrealistic large
mum allowable speed, always ensuring it to avoid headways.
collision with the vehicle ahead, even though in
case of hard braking. The desired speed describes
the will of a vehicle to reach it, while still taking 4 ANALYSIS
into account the safety limits.
One of the main problem in comparing simulated
to real data is the collection of the data themselves.
⎡ v f (t ) + vl (t ) ⎤
a = vl (t )
vsafe [ gn (t ) vl (t ) ] / ⎢ + τ ⎥; As a matter of fact, it’s generally not easy to be
⎣ 2 b ⎦ able, in terms of cost, devices and concrete capabil-
ity, to extract some sort of data from the real study
where: environment.

574

TIS2017_Book.indb 574 2/8/2017 8:07:28 PM

 The service proposed by DataFromSky can over- gated value needed to represent the vehicle char-
come lots of limitation in the road traffic field, but acteristic within the SUMO network, shown in
not this only, providing sufficient amount of data Fig. 7. In particular, two types of vehicle have
to enable vehicle location and movement moni- been defined, one per direction, and the following
toring. From here we started to develop our data: Krauss parameter of Table 4 have been adopted
having a video from above the study area (watch it instead of the default ones, reported in Table 5.
at https://vimeo.com/117601820), we automatically In particular, Speed Factor represent the ratio
analysed the traffic activity within an entry and between the actual speed on the road, taken the
an exit gate we set as shown in Fig. 6. Moreover, 85th percentile of the mean speeds of all vehi-
tables Table 1, Table 2 and Table 3 exhibit an exam- cles as reference, and its speed limit, whereas the
ple of the main part of the whole extracted dataset Speed Deviation comes from the calculated stand-
needed for the following processing steps. ard deviation over the mean speeds on the study
For each of the two directions, data coming from segment.
table Table 1 have been reported into a graph, with The maximum acceleration and deceleration
entry time [s] on the x-axis, and actual travel time taken for the simulation correspond to the 95th
on the y-axis, calculated as the difference between percentile of the actual maximum values (positive
the exit and entry time. Although not many fac- and negative ones) for each vehicle, as shown in
tors can influence the short chosen road section Fig. 8 (a) and (b).
(only 229 m), since the covered distance is almost On the contrary, the following default values are
straight and there are no turning vehicles and no commonly used in SUMO
intersection, except for two accesses that were not The simulations with the two different defini-
used within the recorded time interval, the curve tions of vehicle, the default one for scenario no.1
is not flat at all. This means, as we expected, that and the one coming from DFS for the scenario
even if the network is very simple, maybe overly, no.2, were run. In both case, vehicle were intro-
drivers have different behaviour: the curves have duced at the beginning of the road at the time
mean travel time equal to 16,7 s and 17,1 s, in the corresponding to time recorded data from Data-
East-to-West and West-to-East direction respec- FromSky, so that at the entrance were almost equal
tively, but the curves trend vary between 13 s and the distances, in time and space, between following
21 s in the first case, 11 s and 21 s in the second. vehicles.
On the other hand, both Table 1 and Table 2 In order to evaluate the quality of the simula-
were useful for investigate and collect the aggre- tion done with SUMO software and, in particular,
to do a benchmarking of the analysis performed
with default and real parameters, in comparison
with the observed behaviour, data sets have been
shown on the graphs below, and the mean values
reported on Table 6.
It can clearly be seen that the results obtained
from the scenario no.1 can’t describe with suffi-
cient degree of accuracy the reality of the phenom-
enon, since they do not provide a variance in the
simulation. The curve appears rather flat, which
means that the travel time doesn’t significantly
change and every vehicle driving on the network
spend almost the same time between the entry and
Figure 6. Position of the entry and exit gates with DFS the exit of the lane. Besides, the mean value around
on the examined lanes. which the curve is built differs from the mean of

Table 1. Example of DataFromSky output data: id, type, gate, time, distance, speed.

Vehicle Entry Entry Exit Exit Travelled Avg.

Vehicle id type gate time (ms) gate time (ms) dist. (m) speed (m/s)

24 Car 14 8320 20 26440 228,6 12,61

28 Car 14 40120 20 57360 228,2 13,24
29 Car 14 54640 20 69040 228,5 15,87
… …

575

TIS2017_Book.indb 575 2/8/2017 8:07:29 PM

 Table 2. Example of DataFromSky output data: tan- Table 4. Krauss Model real car parameters (scenario2).
gential acceleration for every second.
Speed Speed a max d max
Vehicle id n cars factor dev. (m/s2) (m/s2)
Travel time (s) 24 28 29 … East to West 30 1.13 0.11 1.85 –0.90
West to East 48 1.11 0.13 1.12 –3.27
1 0.10 −0.50 0.97 …
2 0.66 0.34 0.22 …
3 0.77 0.38 −0.26
4 0.69 0.43 0.01 Table 5. Krauss Model default car parameters
(scenario1).
5 0.27 −0.02 −0.38
6 0.19 0.29 −0.87 speed speed a max d max
7 0.19 0.38 0.07 n cars factor dev (m/s2) (m/s2)
8 0.24 −0.17 −0.23
9 −0.01 −0.16 −0.50 East to West 30 1.0 0.0 2.9 −7.5
10 −0.03 −0.22 −0.58 West to East 48 1.0 0.0 2.9 −7.5
11 0.11 0.32 −0.32
12 −0.10 0.06 0.19
13 −0.13 −0.03 −0.32
14 −0.30 −0.16 −0.31
15 −0.21 −0.19 −0.31
16 −0.19 −0.21 0.00
17 −0.31 −0.21 0.00
18 −0.28 −0.44 0.00
19 −0.25 0.00 0.00
20 0.00 0.00 0.00
a max [m/s2] 0.77 0.43 0.97
a min [m/s2] −0.31 −0.50 −0.87

Table 3. Main aggregated output values from

DataFromSky.

Mean Mean 85ile mean Mean

distance speed St. dev. speed travel
n cars (m) (m/s) (σ) (m/s) time (s)

East to 30 228.41 13.86 1.46 15.69 16.65

West
West to 48 235.97 13.99 1.76 15.41 17.10
East

Figure 8. Example cumulative distribution function

of acceleration (a) and deceleration (b) for West to East
direction.

the scenario no.0: the first is 1,4 s higher in the

East-to-West direction, and 1 s higher in the West-
to-East direction (Figures 9 and 10).
On the contrary, for both considered directions,
it is clearly visible the greatest similarity of the
Figure 7. Sumo network representation with Instant scenario no.2 to the observed data set’s curve. The
Induction Loops. divergence between their relative mean is limited

576

TIS2017_Book.indb 576 2/8/2017 8:07:30 PM

 Table 6. Comparison between travel time on different observed to do it, but the simulation keep random-
scenarios. ness in choosing its dynamic parameter among
the distribution we set. That is the reason why in
Mean travel Mean travel Mean travel time
time (s) Scenario time (s) Sce- (s) Scenario no.2 this type of analysis it is not necessary to have the
no.0 (with nario no.1 (with SUMO complete overlaying of the curves, since that’s not
DataFromSky (with SUMO survey values) what we expected from the simulation.
survey) default values)

East to 16.65 18.06 16.52

5 CONCLUSIONS
West
West to 17.10 18.09 16.82
East The value of the current work should be identi-
fied not only in the appreciable precision degree
achieved through the processing, but primarily
in the methodological approach proposed, that
starting from the existing models has turned to a
formulation that fits the particular dataset consid-
ered, with an elaboration and a contextualization
of the parameters.
It represents only a preliminary attempt to
implement a commonly used simulation tool, with
the aid of a full and not easily obtainable data set,
in order to better reproduce what is the common
approach of vehicle in that type of environment,
with that type of drivers.
In future work all the aspect that have been
Figure 9. Comparison between observed and simulated considered should be increased, from the network
Travel Times, direction East to West. to number and types of sampling vehicles over
which extract the information. That upper analy-
sis should even more highlighted the differences
between the default set of vehicles parameters and
the one coming from a real analysis of the drivers’
attitude in a specific context.

REFERENCES

Apeltauer, J. & Babinec, A., & Herman, D. & Apeltauer,

T. 2015. Automatic vehicle trajectory extraction for
traffic analysis from aerial video data. The Interna-
tional Archives of the Photogrammetry, Remote Sens-
Figure. 10. Comparison between observed and simu- ing and Spatial Information Sciences, Volume XL-3/
lated Travel Times, direction West to East. W2, 2015 PIA15+HRIGI15 – Joint ISPRS conference
2015, 25–27 March 2015, Munich, Germany.
Jakob, E. & Behrisch, M. & Bieker, L. 2012. Recent Devel-
to 0,13 and 0,28 seconds, and the trend fits very opment and Applications of SUMO - Simulation of
well with the one of the scenario no.0. There is still Urban MObility; Daniel Krajzewicz. International
lack of precision in the position of the maximum Journal On Advances in Systems and Measurements,
and minimum peaks of the curve, but this is due 5 (3–4):128–138.
to the fact that we implemented an overall type of Krauß, S. 1998. Hauptabteilung Mobilität und Sys-
temtechnik des DLR Köln 1998. Microscopic Mod-
vehicle, got by aggregate parameter consideration eling of Traffic Flow: Investigation of Collision Free
rather than the single vehicle dynamic characteris- Vehicle Dynamics ISSN 1434–8454.
tics, just like commonly happens when compute a Vertex Access, 296 Sheffield Road, Rotherham, S60 1DX
traffic simulation. By means of that, vehicles were Phone: +44 (0) 1709 379 453 Web: www.vertexaccess.
entered the network at the same time they were com

577

TIS2017_Book.indb 577 2/8/2017 8:07:31 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Social exclusion and high-speed rail: Some evidence from three

European countries

F. Pagliara, L. Biggiero & F. Menicocci

Department of Civil, Architectural and Environmental Engineering—University of Naples Federico II,
Naples, Italy

ABSTRACT: Very few contributions are present in the literature dealing with the issue of social exclu-
sion related to High-Speed Rail (HSR) services. The objective of this study is to compare the current
situation of non-HSR users in Italy, Spain and England and the factors preventing them from choosing
this service. Three countries with very different HSR networks have been chosen in order to get evidence of
the different behaviour. For this purpose, three surveys have been delivered to Italian, English and Spanish
users of the transport systems for long distance journeys. Data about their socioeconomic characteristics
and their perception of social exclusion have been collected based on seven principles. The main result of
these surveys has been that a relationship between social exclusion and HSR is evident, especially in terms
of economic and geographical exclusion. Moreover, these factors are perceived differently based on the
different “shape” of the HSR networks and the service provided.

1 INTRODUCTION the Spanish, English and Italian HSR systems.

The methodology is proposed in section 4. Con-
According to Levitas et al. (2007) social exclusion clusions and further perspectives are presented in
is “The lack or denial of resources, rights, goods section 5.
and services, and the inability to participate in the
normal relationships and activities, available to the
majority of people in a society, whether in economic, 2 SOCIAL EXCLUSION AND HSR
social, cultural or political arenas. It affects both the SYSTEMS
quality of life of individuals and the equity and cohe-
sion of society as a whole.” Among the very few studies present in the litera-
In the literature different approaches have been ture on the link between HSR systems and their
proposed to solve the problem of social exclusion impacts on social exclusion, the statistical analysis
related to transport systems. Indeed, transport of surveys carried out by Cass et al. (2005) reports
systems planning should be integrated with the interesting results. It indicates that HSR has both
urban and social policies. One first step towards positive and negative social impacts. The positive
the reduction of social exclusion might be that of social impact is represented by the increased acces-
promoting activities to increase accessibility. sibility and activities for commuting HSR users.
The objective of this paper is to analyse whether The concept of accessibility represents the rela-
High Speed Rail (HSR) systems can increase tionship between the system of activities in a given
social exclusion for long-distance trips, taking territory and the transport system serving it. The
into account that other transport alternatives are study carried out in Spain by Monzón et al. (2010)
available to users. Specifically, considering that the shows the role played by the selection of the com-
trend in future transportation systems investments mercial speed. Indeed, an increase from 220 km/h
is represented by these services, authors would like to 300 km/h in a given corridor results in signifi-
to analyse this phenomenon and its impacts on cant negative impacts on spatial equity between
social exclusion. For this purpose, three case stud- locations with and without a HSR service.
ies have been analysed, i.e. Spain, UK and Italy. The same authors propose an assessment meth-
This paper is organised as follows. Section 2 odology for HSR projects following a twofold
presents a literature review on the link between approach, i.e. addressing issues of both efficiency and
HSR and social exclusion. Section 3 introduces equity. The procedure uses spatial impact analysis

579

TIS2017_Book.indb 579 2/8/2017 8:07:34 PM

 techniques and is based on the computation of 3 HIGH SPEED RAIL SYSTEMS IN SPAIN,
accessibility indicators. Efficiency impacts are UK AND ITALY
evaluated in terms of increased accessibility result-
ing from the HSR project, with a focus on major Since January 2016, Spain has the world´s sec-
urban areas. Likewise, spatial equity implications ond longest high-speed network, after China, and
are derived from changes in the distribution of the longest in Europe (Pagliara et al. 2016a), with
accessibility values among these urban agglomera- around 3,100 km of HS lines in operation. The serv-
tions (Monzón et al. 2013). ice of HSR in Spain—known as AVE, Alta Veloci-
The paper by Chen & Wei (2013) reports the dad Española—is operated by RENFE Operadora,
case study of Hangzhou East Rail station in the Spanish national railway company. Since 2005,
China. This area is undergoing a rapid industri- AVE trains run on a HSR network owned and
alization and thus workers’ incomes are increasing managed by ADIF, the public company in charge
significantly. However, HSR is still not affordable of the management of most of the Spanish railway
for the majority of the population. In the contri- infrastructure. Although RENFE Operadora is the
bution by Shi & Zhou (2013), the aim is that of only company operating the high-speed trains now-
analysing transportation equity change in terms of adays, private companies may be allowed to oper-
accessibility change experienced by cities served by ate trains in the future, in accordance with the EU
the HSR line in China. The main research findings, legislation. It is envisaged that the Madrid-Valencia
from the equity assessment, reveal that investments corridor will be the first case to introduce competi-
in HSR systems do not have a strong impact in fos- tion in the HSR services in the country.
tering social exclusion in terms of being excluded The first HS line was opened in 1992, connecting
from the use of the new high speed infrastructure. the cities of Madrid, Cordoba and Seville. It was
From these contributions it is clear that elements designed according to the technical standards of
of social exclusion from HSR are represented by the French high-speed TGV. In the following years,
the accessibility and the income. Therefore high the network was extended towards the northern
income users can afford HSR but also a high acces- part of the country, with the aim to create a con-
sibility to HSR station can contribute to increase nection to France and thus to the European HSR
its choice. network (see Figure 1).

Figure 1. Spain’s HSR network as of November 2015 (Source: www.adif.es).

580

TIS2017_Book.indb 580 2/8/2017 8:07:34 PM

 Great Britain has been one of the last nations
to introduce HSR systems. High Speed 1 was
inaugurated in 2007 and it connects London with the
Channel Tunnel. The travel time between London
and Paris is only two hours and a half.
At its opening the line was positively welcomed
by travelers and this was one of the first reasons for
the design of another line, the High Speed 2, whose
works should start in 2017. It will link London
Euston to Birmingham (2026), and then, following
a ‘Y’ shape, it will connect Manchester (2027) on
the west and Leeds (2033) on the east. The line will
be 540 km long, with speeds of almost 400 km/h
(see Figure 2).
In Italy, the first HSR line was inaugurated
in 1992 between Florence and Rome with the so
called “Direttissima”, which allowed trains to run
at 230 km/h covering the 254 km between Rome
and Florence in about two hours. However, this
project dated back to 1970.
The new generation of HSR (i.e. with trains run-
ning at 300 km/h) started in December 2005 between
Rome and Naples and Milan and Bologna. Later,
in December 2009, the project was extended with
Figure 3. Italy’s HSR network as of 2014 (Source:
the Milan-Turin and the Bologna-Florence lines,
Delaplace et al. 2014).

as well as with the urban penetration into the cit-

ies of Rome and Naples. In 2015 the Italian HSR
network was operational for more than 1400 km
(see Figure 3).
He national Italian network and operations are all
owned by FS (State Railway) Holdings, a fully gov-
ernment owned company. It has three key operating
subsidiaries: Trenitalia operates all freight and pas-
senger trains, including the high-speed trains, RFI
(Rete Ferroviaria Italiana) manages the infrastruc-
ture, and TAV (Treno Alta Velocità SpA) is respon-
sible for the planning and construction of the new
HS infrastructure.
The introduction of the new private operator
Nuovo Trasporto Viaggiatori (NTV) in April 2012,
competing with the Trenitalia on the national HSR
network, represents the first case of competing
HSRs operating on the same line (i.e. multiple oper-
ators on a single infrastructure). NTV represents
the first private society to benefit of the European
liberalisation of the High-speed train networks.

4 THE METHODOLOGY

This contribution is based on the framework of

factors that may limit the mobility of socially
excluded people, proposed by Church et al. (2000).
In their paper, the categories of exclusion con-
nected to transport, applied for urban trips, are the
Figure 2. UK’s HSR network as of June 2016. following ones:

581

TIS2017_Book.indb 581 2/8/2017 8:07:35 PM

 1. Physical exclusion: physical barriers, i.e. lack of Table 1. Socioeconomic and trip characteristics for the
disabled facilities or timetable information, lim- three countries.
iting accessibility to transport services.
2. Geographical exclusion prevents people from Spain UK Italy
non- non- non-
accessing transport services, especially those liv- HSR HSR HSR
ing in rural or peripheral urban areas. users users users
3. Exclusion from facilities, concerning the low Characteristics Levels [%] [%] [%]
accessibility connected with facilities, like shops,
schools, health care or leisure services. Age 18–23 13.23 13.41 10.40
4. Economic exclusion represents the high mon- 24–34 12.88 16.61 22.22
etary costs of travel preventing or inhibiting 35–55 32.91 34.06 44.47
access to facilities or employment and thus hav- >55 40.98 35.92 22.21
ing an impact on incomes. Gender M 48.02 50.02 58.11
5. Time-based exclusion refers to other demands F 51.98 49.98 41.89
on time, like combined work, household and Nationality National 97.21 56.61 99.90
child-care duties, reducing the time available for Other 2.79 43.39 0.10
travel. Education Degree or more 73.48 38.11 46.11
6. Fear-based exclusion concerns to the fears for Other 26.52 61.89 53.89
personal safety precluding the use of public Occupation Full time/ 56.71 33.64 62.59
spaces and/or transport services. part time
worker
7. Space exclusion is the security or space manage-
Student 12.50 20.12 15.10
ment preventing given groups having access to
Unemployed 2.00 0.84 3.70
public spaces, like first class waiting rooms at
Freelance 1.97 10.11 6.87
stations.
Retired 26.83 27.64 11.45
Other – 7.65 0.29
In this paper, these categories have been adapted Monthly <€1500 24.90 12.42 48.20
to medium-long distance trips. income
Starting from this premise three Revealed Pref- €1500–€3000 53.16 13.06 28.62
erence (RP) surveys were carried out. The ques- >€3000 21.94 74.52 23.17
tionnaires were created on web platforms and 414 Trip purpose Work 4.81 14.01 30.39
useful ones were collected in Spain, 359 in the Study 0.94 9.66 13.23
United Kingdom, 968 in Italy. Due to the survey Holiday 33.00 18.38 42.25
method used, based on the web platform, the sam- Personal 61.24 57.94 14.13
ple needed to be weighted (Pagliara et al. 2016a, b, activities
Pagliara & Biggiero 2016). The percentages of gen- Travel type Alone 31.67 43.29 39.25
der and age classes, based on the Spanish census Partner 39.58 18.21 13.03
(INE 2015), the UK census (ONS 2014) and the Colleagues – 8.39 14.68
Italian census (ISTAT 2015) have been considered Friends 15.07 8.30 16.15
to adjust the sample. Then, those observations with Relatives 13.68 21.83 16.89
a trip length lower than 60 km have been removed
from the sample since they typically correspond to
regional trips, not operated by HSR services. In partial or full time job, even if in UK there are a lot
this case authors tried to avoid any bias present in of students and retired. In Italy and in Spain almost
the data set used to make inferences. In Table 1 the 80% of non-HSR users have a household monthly
socioeconomic and trip characteristics of the non- income less than 3,000€ while the opposite can be
HSR users for the three cases are reported. found in UK (this result can be justified consider-
It is interesting to notice some differences and ing that non-HSR users represent the majority of
similarities among the case studies. Concerning the the sample (see Figure 4). Almost 60% of non-HSR
age, there are not significant differences, just for the users in Spain and in UK have travelled for personal
case of Italy it is possible to see almost 80% of non purposes, while in Italy only 14% have travelled for
HSR-users are less than 55 years old. This result the same purposes. On the contrary in this coun-
can be justified considering economic factors or try non-HSR users have travelled for holidays and
work or study engagements. Spain and Italy have a work purposes. Almost 40% of non-HSR Spanish
similar trend concerning the non-HSR users nation- users have travelled with the partner, while the same
ality (almost 100% of the sample), while in UK non- percentage in UK and in Italy have travelled alone.
HSR users are a bit more than 50% and this can be The mode choice is represented in Fugure 4.
supported by the multiethnic image of the country. It is evident in Italy and in Spain that HSR is the
Concerning the occupation, non-HSR users have a most chosen transport alternative for long-distance trips.

582

TIS2017_Book.indb 582 2/8/2017 8:07:35 PM

 Table 2. Church et al.’s categories of social exclusion
(non-HSR users)—Comparison.

Case studies

Categories of Spain UK Italy

exclusion [%] [%] [%]

Economic 39.85 48.73 27.20

Time-based 10.59 11.07 6.30
Spatial 6.27 1.38 –
Fear-based 1.23 – 0.90
Geographical 23.87 24.04 58.70
Physical 9.75 9.90 5.60
Facilities 8.44 4.88 0.30
Total 100.00 100.00 100.00

Figure 4. Transport mode choice (all samples).

Figure 6. Church et al.’s categories of social exclusion

(non-HSR users)—Comparison.

Even if the presence of two companies compet-

ing on the same HSR line has reduced travel costs,
there are still a lot of users who cannot afford it.
In UK, on the other hand, the first two factors of
Figure 5. Transport mode choice for non-HSR users.
exclusion are inverted. Geographical exclusion is
less felt with respect to the economic one. The lim-
ited extension of the HSR network in UK could
Car has been chosen by 32%, 30% and 24% in lead to think that geographical exclusion could
Spain, UK and Italy respectively. In UK, as stated have been highly felt. However this has not been the
before, HSR has been chosen by the 11% of the result, considering that a limited network with low
respondents, the same for the bus, while 48% of the accessibility to the stations does not make HSR a
respondents have chosen train and 30% car. real alternative with respect to the conventional rail
Non-HSR users are represented in Figure 5. system, which is capillary and efficient. For Spain,
In Table 2, the choice among the seven Church it is possible to note that the result is very similar
et al.’s categories of social exclusion has been ana- to that obtained in UK, even for different reasons.
lysed and summarized in Figure 6. Indeed, the Spanish network, as reported before,
It follows that in Italy, geographical exclusion is is very extended but the economic factor is still an
the most perceived, followed by the economic and element of exclusion.
time-based. This is justified by the fact that the It is interesting to highlight that in all the three
HSR system connects only big cities and therefore cases time-based exclusion is the third factor. The
it is not considered as a real alternative transport other elements of exclusion are less perceived by
mode. non-HSR users.

583

TIS2017_Book.indb 583 2/8/2017 8:07:36 PM

 In the following Tables, from 3 to 8, the factors Table 6. Church et al.’s categories of social exclusion vs
of exclusion are reported on the basis of the trip monthly income (non-HSR users)—UK.
purpose and the hosuehold monthly income. In
Tables 3 and 4 the results for Spain are reported. Monthly income
It results that Spanish non-HSR users perceiv- Low-
ing the factors of exclusion are those who have Categories medium High Total
travelled for personal purposes and have a low- of exclusion [%] [%] [%]
medium income. Specifically, those travelling for
personal purposes are free to plan the trip, taking Economic 53.32 46.68 100.00
Time-based 21.39 78.61 100.00
Table 3. Church et al.’s categories of social exclusion vs Spatial 47.35 52.65 100.00
trip purpose (non-HSR users)—Spain. Fear-based – – –
Geographical 53.88 46.12 100.00
Trip purpose Physical 52.53 47.47 100.00
Facilities 46.31 53.69 100.00
Personal
Categories Work Study Holiday activities Total
of exclusion [%] [%] [%] [%] [%] Table 7. Church et al.’s categories of social exclusion vs
trip purpose (non-HSR users)—Italy.
Economic 4.47 35.23 1.00 59.29 100.00
Time-based – 11.69 2.85 85.47 100.00 Trip purpose
Spatial – 24.07 3.01 72.92 100.00
Fear-based – 32.41 – 67.59 100.00 Personal
Geographical 5.49 44.51 – 50.00 100.00 Categories Work Study Holiday activities Total
Physical – 23.50 2.65 73.85 100.00 of exclusion [%] [%] [%] [%] [%]
Facilities – 22.84 3.06 74.10 100.00
Economic 34.16 10.90 36.05 18.90 100.00
Time-based – 4.05 91.22 4.73 100.00
Table 4. Church et al.’s categories of social exclusion vs Spatial – – – – –
monthly income (non-HSR users)—Spain. Fear-based 88.89 – 11.11 – 100.00
Geographical 29.72 10.34 40.06 19.88 100.00
Monthly income Physical 9.21 15.79 55.26 19.74 100.00
Facilities – – 100.00 – 100.00
Low-
Categories medium High Total
of exclusion [%] [%] [%] Table 8. Church et al.’s categories of social exclusion vs
monthly income (non-HSR users)—Italy.
Economic 77.96 22.04 100.00
Time-based 93.01 6.99 100.00 Monthly income
Spatial 67.40 32.60 100.00
Fear-based 57.23 42.77 100.00 Low-
Geographical 73.85 26.15 100.00 Categories medium High Total
Physical 69.37 30.63 100.00 of exclusion [%] [%] [%]
Facilities 55.52 44.48 100.00
Economic 70.06 29.94 100.00
Table 5. Church et al.’s categories of social exclusion vs Time-based 97.30 2.70 100.00
trip purpose (non-HSR users)—UK. Spatial – – –
Fear-based 88.89 11.11 100.00
Trip purpose Geographical 80.12 19.88 100.00
Physical 38.16 61.84 100.00
Personal Facilities 100.00 – 100.00
Categories Work Study Holiday activities Total
of exclusion [%] [%] [%] [%] [%]
into account the economic and time availability,
Economic 11.25 11.09 18.45 59.21 100.00
but also the trip origin and the destination.
Time-based 31.25 1.45 10.74 56.57 100.00
In Tables 5 and 6 the results for UK are
Spatial 13.55 11.69 24.04 50.73 100.00
reported.
Fear-based – – – – –
Geographical 9.43 17.04 11.06 62.47 100.00
The same considerations made for Spain can be
Physical 9.64 7.45 41.40 41.52 100.00
adapted to the case study of UK. Indeed, also in
Facilities 36.40 – – 63.60 100.00 this case, those suffering of social exclusion re the
ones travelling for personal purposes.

584

TIS2017_Book.indb 584 2/8/2017 8:07:36 PM

 Concerning the income, those suffering of geo- Indeed, a quantitative approach should be taken
graphical and economic exclusion are the ones into account with the aim of better evaluating
of low-medium income. On the other hand high the perception of social exclusion related to mode
income English non-HSR users highly perceive the choice.
time-based exclusion.
In Tables 7 and 8 the results for Italy are
reported. REFERENECES
The latter are very similar to the case study of
Spain with the difference that the only category of Cass, N., Shove, E., Urry, J. 2005, Social exclusion, mobil-
exclusion is represented by those who have travelled ity and access. The social review 3: 539–555.
for holidays. Also in this case, this category was the Chen, C-L., Wei, B. 2013, High-Speed Rail and urban
most represented in the sample, confirming that trips transformation in China: the case of Hangzhou east
rail station. Built Environment 39: 385–398.
fo toursim purpose, but also for personal purposes, Church, A., Frost, M., Sullivan, K. 2000, Transport
are represented by a highly elasticity of choice (e.g. and social exclusion in London. Transport Policy 7:
destination, time, budget etc.). Therefore the same 195–205.
comments can be applied to this case study. Delaplace, M., Pagliara, F., Perrin, J. e Mermet, S. 2014,
Can High Speed Rail foster the choice of destination
for tourism purpose? Procedia Social and Behavioral
5 CONCLUSIONS AND FURTHER Sciences 111: 166–175.
PERSPECTIVES Levitas, R., Pantazis, C., Fahmy, E., Gordon, D., Lloyd, E.,
Patsios, D. 2007. The Multi-dimensional Analysis of
In this paper the relationship between HSR and Social Exclusion. London: Department for Communi-
ties and Local Government (DCLG).
social exclusion has been analysed. Following Monzón, A., Ortega, E., López, E. 2010. Social impacts
the framework proposed by Church et al. (2000). of high speed rail projects: addressing spatial equity
The results of three Revealed Preference surveys effects. Proceedings of the 12th WCTR, July 11–15,
have shown that only some criteria are perceived Lisbon, Portugal.
by the users and among them only some of them Monzón, A, Ortega, E., López, E. 2013. Efficiency and
are the most relevant. For those who have not cho- spatial equity impacts of high-speed rail extensions in
sen HSR, the main reason for that, is the economic urban areas. Cities 30: 18–30.
exclusion, i.e. the cost of the HSR ticket. It follows Pagliara, F., Menicocci, F., Gomez, J., Vassallo, J.M.
the geographic exclusion, i.e. the low accessibility to 2016a, Economic, geographical and time-based exclu-
sion as main factors inhibiting Spanish users from
the departure/arrival station. The fact that both cri- choosing High Speed Rail. Paper submitted to Trans-
teria are greatly perceived by low income classes can port Policy.
be interpreted by the location of residences of these Pagliara, F., De Pompeis, V., Preston, J. 2016b, Travel
classes of travellers. For the higher cost connected cost: not always the most important element of social
with the use of the residences, it is clear that those exclusion. Paper submitted to Transport Policy.
having higher incomes live in city centres, which, Pagliara, F., Biggiero, L., 2016, Social exclusion from
in general, are served by a good public transport High Speed Rail systems: An exploratory study. Paper
and by taxi as well. Indeed a good public transport accepted for publication by Ingegneria Ferroviaria.
system can allow an easy access to the departure/ Shi, J., Zhou, N. 2013. How Cities Influenced by High
Speed Rail Development: A Case Study in China,
arrival station. Journal of Transportation Technologies 3: 7–16.
Further perspectives will consider the collection http://www.adif.es
of a larger data set which can support these find- http://www.ine.es
ings and the specification and calibration of some http://www.istat.it
mode choice models. https://www.ons.gov.uk

585

TIS2017_Book.indb 585 2/8/2017 8:07:36 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Modeling airport noise using artificial neural networks

and non-linear analysis

M. De Luca
University of Naples “Federico II”, Naples, Italy

D.C. Festa & G. Guido

University of Calabria, Rende, Italy

ABSTRACT: The noise generated by air traffic is an issue debated very much in the scientific community.
It is a long history of consciousness, because it is not an easy topic to understand. From 1966 in Europe,
this matter was recognized as a source/cause of pollution, whose effects are dangerous to human health
(particularly about hearing and extra-hearing damage). In this work two models are built to estimate
the level of aircraft noise (Lva) around populated areas. In particular the study evaluates the effects of
this phenomenon on people and buildings. Data of the level of noise pollution (Lva), were detected
on the airport of Lamezia Terme (IATA: SUF, ICAO: LICA), during the period 2006–2008, through 7
receptors located close to the airport. To the air traffic, reference was made to the data provided by the
“post-holder” office at the airport of Lamezia Terme. The data were processed with the Multivariate
Analysis and Artificial Neural Network technique. Two models were obtained: Model 1 (Model MVA)
and Model 2 (Model ANN). Both models showed good predictive ability in terms of Lva. In particular,
model 2 (ANN model) was better than model 1 (MVA Model) in terms of residual. In addition, to test
the simulative capabilities of the two models (MVA model and ANN model), an experiment is conducted
on a sample data not used for the construction of the two models.The comparison showed that the ANN
model is the most reliable because it has the lowest residual.

1 INTRODUCTION airport; the land-use cost, noise reduction charges

and other risks of the offshore airport were also
The noise generated by air traffic is an issue lesser; the creation of the offshore airport may be
debated very much in the scientific community. It’s more favorable to the city’s development.
a long history of consciousness, because it is not Sadr et al. (2014) studied land use planning
an easy topic to understand. Indeed, since 1966 in around airports, by employing Remote Sensing
Europe, this matter was recognized as a source/ (RS) and Geographic Information Systems (GIS),
cause of noise/pollution, whose effects are danger- in conjunction with an optimization algorithm
ous to human health (particularly about hearing using an Integrated Noise Model (INM) software,
and extra-hearing damage). In the last years, many to establish the potential effects of aircraft noise
researchers studied this issue, providing support at Imam Khomeini International Airport (IKIA)
tools for the management and control of this phe- in Tehran (Iran). The results indicated that devel-
nomenon, especially in places close to populated oping IKIA together with the residential develop-
areas, to control the effects on people and on the ment will increase airport noise.
buildings (buildings, artifacts, etc.). Maldonado et al. (2013), studied airport noise
Ozkurt et al. (2015), studied the results of noise exposure around Viracopos International Airport
in Izmir Adnan Menderes Airport (Turkey). The quantifying the proportion of highly annoyed people
results showed that about 2% of the resident popu- in surrounding zones using simulations, integrated
lation was exposed to noise levels of 55 dB(A) or noise models and geographic information systems.
higher during day-time in İzmir. Licitra et al. (2014), simulated the present and
Hua-Kun Yan, et al. (2013), in the Dalian Inter- future scenarios and they evaluated the health end-
national Airport compared the aircraft noise pollu- points produced by Galileo Galilei Airport of Pisa
tion and the cost-risk effects. The findings showed (Italy).
that the aircraft noise pollution of the offshore air- Sari et al. (2014), developed an aircraft noise
port was lesser than that of the expanded inland model for the İstanbul Atatürk Airport (Turkey).

587

TIS2017_Book.indb 587 2/8/2017 8:07:38 PM

 It was found that 1.2% of the land area of İstanbul 1
City exceeded the threshold of 55 dB(A) during uj = (2)
1 + e (∑( ) ij )
− iw Ui + J
daytime.
Heleno et al. (2014), made an analysis of airport
noise based on LAeqD and LAeqN methodology. where θj is the bias unit uj (i.e. the degree of sensitiv-
The goal of the study was to make an analysis ity of uj when it receives an input signal from ui).
of airport noise through the LAeq metric and to
propose new alternatives based on Day Equiva- 2.2 Multi Layer Perceptron (MLP) and the Back
lent Sound Level (LAeqD) and Night Equivalent Propagation (BP) algorithm
Sound Level (LAeqN) noise metrics.
Black et al. (2007) made a study in residential In this study, a neural network with MLP architec-
neighborhoods near Sydney Airport with high ture was used. Training was carried out using the
exposure to aircraft noise. Noise aircraft measure- Back Propagation (BP) algorithm. The neurons
ments were analyzed. The study showed that those (or units) that comprise this type of network are
who have been chronically exposed to aircraft noise organized into layers: an input layer, an output and
were more likely to report stress and hypertension. a number of intermediate layers between input and
output referred to as hidden, defined by the user.
Initially the weights are assigned random values
2 TECHNIQUES USED IN normalized in the range [0,1] or [−0.5, +0.5] and
DATA ANALYSIS initially there is a pattern “p” input to the network:
Xp = (X0, X1, X2, ..., Xn − 1) with X0 = 1 and a
Two different types of techniques are used for the vector consisting of the output values Tp = (T0, T1,
analysis in this study: MultiVariate Analysis (MVA) T2, ..., Tm − 1). In this way, the network will consist
and Artificial Neural Network (ANN). For the of (n − 1) input neurons and (m − 1) output neu-
first the description is omitted because it is present rons. The “weighted sum of the inputs” for each
for many years in the technical literature. For the layer is calculated using Eq. 1 and its value of acti-
second, much more recent, the following basic vation, i.e. output, using Eq. 2. Then, the weights
principles are described (Čokorilo et al, 2014). must be changed so that the output of the network
(i.e. the output of the last layer of neurons) increas-
2.1 The Artificial Neural Network multilayer ingly approximates the target set by the user.
approach It was defined a function error (Eq. 3) propor-
tional to the square of the difference between the
Inspiration for the structure of the ANN is taken output and target for all output neurons:
from the structure and operating principles of the
( )
human brain. It is made up of neurons linked by 1
∑ Tpjj − Opj
2
connections that represent the connections between Ep (3)
2 j
the biological synoptic neurons. The function of a
biological neuron is to add its input and produce
Subsequently, Back Propagation is applied i.e. the
an output. This output is transmitted to subsequent
weights are varied so that error Ep tends towards zero
neurons, through the synoptic joints, only if the
(starting from the last layer to the first). We define,
transmitted signal is high (i.e., greater than a prede-
for the current pattern p, a variation Δwij of weight
termined value), otherwise, the signal is not trans-
wij between the neuron i and j that given by (Eq. 4).
mitted to the next neuron. In the network, therefore,
a neuron calculates the weighted sum, using (Eq. 1),
∂E p
(considering the input xi and weights wi) and com-
pares it with a threshold value; if the sum is greater
Δ p wij α
∂w ijp
(
β Δp )w i
ij (4)
than the “threshold” value, the neuron “lights up”
and the signal is transmitted. Otherwise, the neuron where α is the learning coefficient (learning rate), β
does not turn on and the flow stops. is momentum, and Δp − 1wij is the variation of the
same weight calculated according to the previous
n
model. The new weights are given by Eq. 5.
I ∑w ⋅ x
i =1
i i (1)
w new
ij = w old
ij + Δ p ⋅ w ij (5)
The activation value “ui” rather than “uj”, con-
nected to weight Wij, is a function of the weighted The variation of the weights is calculated starting
sum of the input. This function may take various from the layer of output neurons and backward
forms. In this study, a function of type (Eq. 2) was toward the first hidden layer. The derivatives can
used. be calculated using Eq. 6.

588

TIS2017_Book.indb 588 2/8/2017 8:07:38 PM

 Δp A i δ j β Δ p −1 ⋅ w ij (6) There are 7 survey stations, whose location
and georeferencing are reported in Table 1 and
Figure 2.
where Ai is the value of the i-th neuron of the layer Specifically, in phonometric reference (Fig. 3)
being considered; δj is given by Eq. 7 if we are con- we use the Lva parameter, whose expression is
sidering the output layer. shown below (see equation 9), as recommended by
European Directive 2002/30/CE on noise monitor-
δ j = (Tj O j ) ⋅ O j ( Oj) (7)
ing resulting from air traffic).

It is given by Eq. 8 for all other intermediate ⎡1 N ⎤

layers. L va = 10 log ⎢ ∑10 Lvaj /10 ⎥ (9)
⎣ N j=1 ⎦
δj j ( j )∑ jk δk (8)
k
Table 1. Location of “Sound level meter”.
To train a network, this process must be run many
times (at least 1,000) with different patterns, each Label “Sound Distance
of which features a different weight. This process level meter” runway
station Latitude Longitude center [m]
is performed until the error is less than a predeter-
mined value (the value is set by the user). When the REC-1 16°15,940’00” 38°54,161’00” 1687
process converges, the network is ready to classify a REC-2 16°14,915’00” 38°54,168’00” 382
new input with an unknown target. The parameters REC-3 16°13,531’00” 38°54,450’00” 1634
α and β are chosen by the user with values between REC-4 16°14,133’00” 38°54,317’00” 707
0 and 1; in the present study α was assumed equal to REC-5 16°14,724’00” 38°54,571’00” 470
0.5 and β equal to 0.4. In particular, α is linked to the REC-6 16°15,239’00” 39°54,426’00” 735
convergence of the network (De Luca et al., 2016). REC-7 16°13,917’00” 38°54,282’00” 238

3 DATA COLLECTION AND

INSTRUMENTATION USED
IN MONITORING

The investigation interests the runway of the inter-

national airport of Lamezia Terme (IATA: SUF,
ICAO: LICA) whose layout is shown in Fig. 1.
In the last few years the airport has highly incre-
mented passengers traffic and handling activities
(De Luca & Dell’Acqua, 2013). Airports belonging
to a local air transportation system where compe- Figure 2. Location of “Sound level meters”.
tition is strong exploit their inputs lesser than air-
ports with local monopoly power. In 2011, more
than 2 million of passengers were recorded at the
airport of Lamezia Terme.
The sound level measurements are performed fol-
lowing the European Directive 2002/30/CE, and they
are measured with the instrument shown in Figure 1
from October 1 2006 to October 2, 2008 h 24.

Figure 1. Overview of Lamezia Terme airport. Figure 3. Instrumentation.

589

TIS2017_Book.indb 589 2/8/2017 8:07:39 PM

 Table 2 shows data collection; in particular for In addition (for each movement, from October 1
each “sound level meter” (indicated by acronym 2006 to October 1 2008) in Table 3 other informa-
SL) the daily maximum values of noise levels tion is reported (De Luca et al, 2016).
detected are illustrated. In particular, the Vmax (maximum speed) and the
Thrust T are defined by the aircraft manufacturers.
Table 2. Lva detected by Sound level meters. The takeoff speed (Vtakeoff), i.e. the speed with maxi-
mum energy production of noise on the airport struc-
SL1 SL2 SL3 SL4 SL5 SL6 SL7 ture, is assumed equal to 0.285*Vmax. The Power P
Lva Lva Lva Lva Lva Lva Lva is derived using the following equation (10).
Date [db] [db] [db] [db] [db] [db] [db]

01/01/2006 98 90 67 65 80 89 131 P T × Vtakeofff (10)

02/01/2006 99 90 67 65 80 89 131
03/01/2006 100 90 67 65 81 89 131 Subsequently data contained in Tables 2
04/01/2006 100 91 67 65 81 89 131 and 3 are aggregated into classes of Lva (class
05/01/2006 101 91 68 65 81 89 132 amplitude equal to 10 dB). For other variables,
06/01/2006 102 92 68 65 81 90 132 see Table 4, the average value in each class is
07/01/2006 102 92 68 66 81 90 132 considered.
08/01/2006 103 92 68 66 81 90 132
09/01/2006 103 93 68 66 82 90 132
.... .... .... .... .... .... .... .... 4 MVA APPLICATION (MODEL 1)

Table 3. Data collection. Model 1 (Eq. 11) was obtained using MVA (shown
in Chapter 2) on data set contained in Table 4; the
T, P, structure of the model and the variables used are
Type Thrust Vmax Vtakeoff Power the following:
Date Company Aircraft [N] [km/h] [km/h] [Kw]
L va a1 ⋅ P1/ 2 a2 ⋅ d2 (11)
1/1/2006 Helvetic F 100 67 845 241 16159
1/1/2006 Austrian CRJ 78 860 245 19062
1/1/2006 A.Berlin F 100 67 845 241 16159
Lva, (dependent variable);
1/1/2006 Airone B 737 64 946 270 17280
D, Distance (Predictor);
1/1/2006 Alitalia M 82 186 811 231 42938 P, Power (Predictor).
1/1/2006 Airone B 737 64 946 270 17280 Model 1 (Eq. 11) is characterized by a coeffi-
1/1/2006 Airone B 737 64 946 270 17280 cient of determination ρ2 = 0.88 and a significance
1/1/2006 Alitalia A 319 115 850 243 27899 greater than 95% (see Table 5).
…… ……... ….. …… ……. …….. …… It is immediate to observe that:
• a1 presents positive value therefore if the Power,
Table 4. Aggregated/Organized data in Lva classes. P, increases Sound Level (Lva) increases;
• a2 presents negative value therefore if the
Lva classes [dB] D, Distance [m] P, Power [kW]
distance, D, increases the Sound Level (Lva)
20–30 1122 619001 decreases.
30–40 965 574497
40–50 866 640848
50–60 934 566705 5 ANN APPLICATION (MODEL 2)
60–70 864 503595
70–80 950 564299 Model 2 is obtained using the ANN technique
80–90 828 503451 shown in Chapter 3 using the same variables in the
90–100 884 547246 previous Chapter (MVA application). In particular,
100–110 549 568255
110–120 607 489310
Table 5. Model parameters.
120–130 439 486237
130–140 358 460735 Interval
140–150 377 553864
150–160 389 620240 Std. Lower Upper
160–170 470 652417 Par. Est. Error Bound Bound Sign.
170–180 470 698177
180–190 470 739234 a1 0.232 0.010 0.210 0.254 >95%
190–200 470 704856 a2 −1.31E-4 1.28E-5 −1.58E-4 −1.04E-4 >95%

590

TIS2017_Book.indb 590 2/8/2017 8:07:41 PM

 Model 2 is obtained using 70% of dataset for train- 6 ANN VERSUS THE MVA
ing and 30% for testing. Fig. 4 shows the ANN
architecture while the parameters of the ANN Table 7 and Figure 5 show the comparison
estimates are shown in Table 6. between the two models and they denote that
Model 2 is better than Model 1 because the resid-
ual has a lower total sum (see last two columns
of Table 7).
To test the simulative capabilities of the two
models (MVA model and ANN model), an experi-
ment is conducted on a sample data not used for
the construction of the two models. Table 8 shows
the results obtained.
Even in this case, the ANN model is more
reliable than the MVA model. Table 9 and
Figure 6 show the results of the comparison
Figure 4. ANN Architecture. (Dell’Acqua, 2016).

Table 6. Parameters of the ANN estimates.

Predicted

Hidden Hidden Output

Layer 1 Layer 2 Layer

Predictor H (1:1) H (1:2) H (1:3) H (2:1) H (2:2) Lva

Input (Bias) −0.200 −0.287 −0.287

Layer D −0.352 −0.713 −0.713
P −0.100 0.718 0.718
Hidden (Bias) 0.724 −0.249
Layer H (1:1) 0.227 −0.837
1 H (1:2) 0.293 2.244 Figure 5. Graphical Comparison: ANN Vs MVA.
H(1:3) 0.316 −1.044
Hidden (Bias) 0.738
Layer H (2:1) 2.495 Table 8. Data used to validation models.
2 H (2:2) −5.480
P D Lva
Power Distance Observed
Table 7. Comparison between ANN and MVA and
Residual. 476200 734 80
395404 470 100
Lva Lva Lva D P Res. Res.
Observed MVA ANN Distance Power 1 2
402424 237 120
521579 470 140
25 17 36 1122 619001 8 11 672155 470 160
35 54 50 965 574497 19 15 888617 470 180
45 87 65 866 640848 42 20 952087 470 200
55 60 55 934 566705 5 0
65 67 78 864 503595 2 13
75 56 53 950 564299 19 22
85 75 83 828 503451 10 2 Table 9. Comparison between ANN and MVA (valida-
95 69 65 884 547246 26 30 tion phase).
105 135 116 549 568255 30 11
115 114 114 607 489310 1 1 P D Lva Lva Lva
125 137 125 439 486237 12 0 Power Distance Observed MVA ANN
135 141 129 358.6 460735 6 6
145 154 132 377 553864 9 13 476200 734 80 89 83
155 163 155 389.4 620240 8 0 395404 470 100 117 116
165 158 164 470.3 652417 7 1 402424 237 120 140 125
175 165 181 470.3 698177 10 6 521579 470 140 139 132
185 170 185 470.3 739234 15 0 672155 470 160 161 164
195 166 182 470.3 704856 29 13 888617 470 180 190 181
Total Residual 257 165 952087 470 200 197 185

591

TIS2017_Book.indb 591 2/8/2017 8:07:42 PM

 Dalian International Airport, Dalian, China. Trans-
portation Research Part D: Transport and Environment,
24: 37–43; doi: 10.1016/j.trd.2013.05.005.
Sadr, M. K., Nassiri, P., Hosseini, M. & Monavari, M,
2014. Assessment of land use compatibility and noise
pollution at Imam Khomeini International Airpor.,
Journal of Air Transport Management, 34: 49–56;
doi:10.1016/j.jairtraman.2013.07.009.
Maldonado F., Heleno, T. & Slama, J. 2013. Noise
mitigation action plan of Pisa civil and military
airport and its effects on people exposure. Journal
of Air Transport Management, 31: 15–1; doi: 10.1016/j.
jairtraman.2012.11.001.
Licitra, G., Gagliardi, P., Fredianelli, L. & Simonetti,
Figure 6. Graphic Comparison between ANN and D. 2014. Applied Acoustics, 84: 25–36; doi:10.1016/j.
MVA (Validation phase). apacoust.2014.02.020.
Sari, D., Ozkurt, N., Akdag, A., Kutukoglu, M. &
Gurarslan, A. 2014. Modeling of noise pollution
7 CONCLUSIONS and estimated human exposure around İstanbul
Atatürk Airport in Turke., Science of The Total
In this work, two models are built to estimate the Environment, 482–483: 486–4921; doi:10.1016/j.scito-
level of aircraft noise (Lva) around populated areas. tenv.2013.08.017.
Data of the level of noise pollution (Lva) are Tarcilene A., Slama J. & Maldonado, F. 2014. Analy-
detected on the airport of Lamezia Terme (IATA: sis of airport noise through LAeq noise metrics.
SUF, ICAO: LICA), during the period 2006–2008, Journal of Air Transport Management, 37, Pages 5;
through 7 receptors (SL) situated close to the air- doi:10.1016/j.jairtraman.2014.01.004.
Black, D. A., Black, J. A., Issarayangyun T. & Samuels
port. The traffic data, for the same period, are pro- S.E.2007. Aircraft Noise Exposure and Resident’s
vided by the “post-holder” office of the airport of Stress and Hypertension: A public Health Perspec-
Lamezia Terme. tive for Airport Environmental Management. Journal
The data organized in Table 4 are processed of Air Transport Management, 13: 264–276. http://
through the MVA technique and ANN technique dx.doi.org/10.1016/j.jairtraman.2007.04.003.
(see Chapter 2) and two models have been obtained: Čokorilo, O., De Luca, M. & Dell’Acqua, G. 2014.
Model 1 (MVA model) and Model 2 (ANN model). Aircraft safety analysis using clustering algo-
Both models show good predictive ability in terms rithms. Journal of Risk Research, 17(10). DOI:
of Lva. In particular, model 2 (ANN model) is bet- 10.1080/13669877.2013.879493.
De Luca, M. 2015. A Comparison Between Prediction
ter than model 1 (MVA Model), in terms of resid- Power of Artificial Neural Networks and Multivariate
ual (see Table 8). Analysis in Road Safety Management, Transport, 1–7,
For a first procedure of validation, the mod- ISSN: 1648–4142, DOI: http://dx.doi.org/10.3846/164
els are tested on a sample of data that have not 84142.2014.995702.
been used to construct the two models. The sample De Luca, M.; Abbondati, F.; Pirozzi, M.; Zilioniene, D.
comes from the same airport (Lamezia Terme) and 2016. Preliminary study on runway pavement fric-
the same period (2006–2008). Even in this case, tion decay using data mining. Transportation
both models show a good predictive capacity and Research Procedia, 14: 3751–3670. DOI: 10.1016/j.
the comparison between them denotes that Model trpro.2016.05.460.
De Luca, M. & Dell’Acqua, G. 2013. Runway surface
2 is better than Model 1. friction characteristics assessment for Lamezia Terme
airfield pavement management system. Journal of Air
Transport Management, 34: 1–5, ISSN: 0969–6997,
REFERENCES doi: 10.1016/j.jairtraman.2013.06.015.
De Luca, M., Abbondati, F., Yager, T.J. & Dell’Acqua,
Ozkurt, N., Hamamci, S., F. & Sari, D. 2015. Estima- G. 2016. Field measurements on runway friction decay
tion of airport noise impacts on public health. A case related to rubber deposits, Transport, 31( 2): 177–182.
study of İzmir Adnan Menderes Airport. Transporta- DOI: 10.3846/16484142.2016.1192062.
tion Research Part D: Transport and Environment, 36: Dell’Acqua, G., De Luca, M., Prato, C.G., Prentkovskis,
152–159; doi: 10.1016/j.trd.2015.02.002. O. & Junevičius, R. 2016. The impact of vehicle move-
Hua-Kun Yan, Nuo Wang, Liao Wei & Qiang Fu, 2013. ment on exploitation parameters of roads and runways:
Comparing aircraft noise pollution and cost-risk a short review of the special issue. Transport 31(2):
effects of inland and offshore airports: The case of 127–132, DOI: 10.3846/16484142.2016.1201912.

592

TIS2017_Book.indb 592 2/8/2017 8:07:43 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A new safety performance index for speed-related crashes

G. Guido, V. Astarita, A. Vitale & V. Gallelli

University of Calabria, Rende, Italy

F.F. Saccomanno
University of Waterloo, Waterloo, Canada

ABSTRACT: The use of GPS-equipped smartphones by drivers can facilitate the tracking and moni-
toring of vehicle operations in real-time. This kind of solution would be implemented to obtain estimates
of individual vehicle operating speeds from the traffic stream. However, before GPS data can be used
in safety performance analysis, tracking errors caused by en-route satellite signal disruptions need to be
taken into account to accurately reflect real-world traffic conditions. The main aim of this paper is to
demonstrate how, when adjusted for errors, GPS-based estimates of instantaneous speeds can be used
to highlight locations where safety is compromised due to poor road geometry. A vehicle-specific Track
Safety Performance Index (TSPI) is developed to measure the difference between individual vehicle oper-
ating speed and the design speed for the location. The results from two case studies support the use of
error-adjusted GPS probe data for identifying sites where safety at a given location is compromised by
driving too fast for the underlying geometric restrictions. Locations with higher TSPI values were found
to correspond closely to sites with a higher number of speed-related crashes as reported over a period of
five years. The results from the two case studies were found to be consistent in linking GPS probe esti-
mates of TSPI to locations of higher crash risk.

1 INTRODUCTION characteristics (type of road, road surface, geomet-

ric features, etc.), traffic conditions (volume, speed,
One of the most common methodologies to density, etc.), vehicle attributes (maneuverability,
estimate safety makes use of inferential statis- braking capability, stability, etc.), and environment
tics applied to crashes databases therefore being (weather conditions, light conditions, etc.). Among
considered a reactive approach to the problem. these factors speed is universally recognized as the
Although this method seems to intuitively link the main factor for the risk of accidents. The World
causes to effects, a good knowledge of the dynam- Health Organization states that “speed has been
ics of the events preceding the crash may provide identified as a key risk factor in road traffic inju-
a more useful support to the implementation of ries, influencing both the risk of a road crash as
appropriate countermeasures. well as the severity of the injuries that result from
Safety performance indicators represent a use- crashes”.
ful tool for evaluating road safety conditions on According to the AASHTO Green Book
the basis of objective parameters deducible from (AASHTO 2001), one of the most important fac-
the vehicle kinematics. Most of these indicators tors in roadway design is the design speed. When
provide a causal or mechanistic basis for explain- roads are designed encouraging drivers to travel
ing interactions between different pairs of vehicles at higher desirable speeds several safety problems
belonging to the traffic stream that can compro- may occur if the differences between vehicle design
mise safety (Hayward 1971, Minderhoud & Bovy speeds, operating speeds and speed limits are con-
2001, Huguenin et al. 2005, Saccomanno et al. siderable at a given location (Stamatiadis et al.
2008, Guido et al. 2011, Astarita et al. 2011). 2009).
However, some accidents, as well as some poten- A number of empirical studies have explored the
tial conflicts, are caused by the aggressive behav- relationship between vehicle operating speeds and
ior of users who adopt speeds unsuitable for the crash occurrence (Aarts & van Schagen 2006, Tay-
road geometry causing isolated vehicle crashes. lor et al. 2000, Hassan & Abdel-Aty 2011, Nilsson
Indeed, safety is influenced by a number of traffic 2004). The American Association of State High-
and geometric factors, such as driver features and way and Transportation Officials Green Book
conditions (experience, stress, tiredness, etc.), road (AASHTO 2011) recommends different “safe

593

TIS2017_Book.indb 593 2/8/2017 8:07:45 PM

 design speeds” or standards for different types of 2 SAFETY PERFORMANCE INDEX
roads and traffic conditions, taking into account FORMULATION
features such as, super elevation, radii of curva-
ture, sight distances, lengths of crests and sags, etc. Safety performance in this paper for a given
Higher design speeds reflect safer road geometries, vehicle is expressed as the difference between
such as, sweeping curves, steeper banking, longer instantaneous vehicle operating speed and rec-
sight distances, and more gentle hills and valleys. ommended safe design speed for the location.
These permit drivers to increase their operating For uncongested conditions instantaneous vehi-
speeds without incurring an increased crash risk. cle operating speeds are assumed to reflect driv-
To fully appreciate the speed-safety relationship, ers “desired” speed at a given location subject
we need to understand how individual drivers at a to prevailing weather conditions (in this paper
given location select their operating speeds based assumed ideal). Equation 1 provides a math-
on what they perceive to be safe (Walton & Bathurst ematical formulation of a Track Safety Perform-
1998, Haglund & Åberg 2000, Sümer et al. 2006, ance Index (TSPI) for an individual vehicle at a
Wallén Warner & Åberg, 2008, Mannering 2009). specific point in its trajectory (or distance incre-
Hence, setting operating speeds below safe design ment i), such that:
thresholds ensures a degree of safety for individual
vehicles in the traffic stream along a given route. ⎛ v op − vt ⎞
One of the challenges in this analysis is to obtain TSPI = α ⋅ ⎜ i t i ⎟ (1)
⎝ vi ⎠
accurate empirical estimates of individual vehicle
operating speeds over time. GPS smartphones
offer a potentially low-cost means to acquire time- where:
dependent estimates of instantaneous speeds in a viop = mean operating speed of sample of vehi-
nonintrusive reliable manner, without the need for cles at location i.
stringent setup controls for videotaping and image vit = safe design speed at location i.
detection (Guido et al. 2012, Guido et al. 2013, α = scaling factor.
Guido et al. 2014, Bierlaire et al. 2010, Herrera vit in Equation 1 represents the safe design
et al. 2010). The focus of this research is on the standard for the geometric conditions at location i,
use GPS probe vehicle estimates of instantaneous as provided in the AASHTO Green Book. viop rep-
speeds for input into a Track Safety Performance resents the mean instantaneous operating speed of
Index (TSPI) for a given location. a sample of GPS vehicle probes traversing location
Two case study applications are introduced to i, such that for a sample of m0 probes:
demonstrate how uncertainty in GPS-based speed
estimates can affect the resultant measure of safety m0
performance. When estimates of speed in TSPI are
properly adjusted for GPS errors, the case studies
v op i ∑v
m =1
i ,m / m0 (2)

indicate that they can provide a reliable method for

highlighting potential safety problems along the In Equation 2, vi,m is the instantaneous speed of
route. Logically we would expect that if crashes the mth vehicle traversing location i, as extracted
occur consistently at a given site that their location from GPS output.
along the route would correspond to sites with high The underlying premise in the study is that each
TSPI measures. Factors affecting the accuracy of driver sets his/her operating speed based on a per-
the GPS-based TSPI are explored empirically for ception of what is safe for the location. An over-
the case study highway segments. estimation of “what is perceived to be safe” will
The paper is organized according to the follow- inevitably contribute to increased crash risk, espe-
ing structure. Section 2 provides the formulation cially as it relates to single vehicle speed-related
of the safety performance index introduced in crashes. Safe design speed is essentially established
this work. Section 3 describes the model adopted from geometric restrictions on mobility at a given
for the safe design speed used as benchmark in location. For this paper, we have assumed that all
safety evaluation. In section 4 a brief description drivers in the probe sample have the same instan-
of a benchmarking exercise is presented that is taneous operating speed (sample mean) and that
addressed to estimate the accuracy of instantane- they face the same safe design speed for the loca-
ous operating speeds obtained from GPS probes. tion vit.
Section 5 introduces a case study application. Sec- TSPI from Equation 1 is location specific, but
tion 6 provides the results for a second case study it can also be expressed for the entire segment by
application highlighting the transferability of the weighting each distance increment (or location) i
methodology. Section 7 summarizes the major by the mean speed of vehicles in the sample at that
conclusions of the study. location, such that:

594

TIS2017_Book.indb 594 2/8/2017 8:07:45 PM

 ⎡ where:
n
di ⎤
∑ ⎢TSPII ⋅ v i ⎥ Lcr is the length of circular curve (m),
TSPI =
i =1 ⎣ i ⎦ Lc11 and Lc12 are the lengths of clothoids preceding
n (3)
di and succeeding the circular curve (m),
∑v
i =1
R is the radius of circular curve (m),
i
L = Lcr+Lc11+Lc12 (km).
Equation 3 reflects the safety performance
index of the entire route segment based on dif- 4 UNCERTAINTY IN GPS PROBE SPEEDS
ferences between mean vehicle operating speeds
and safe design standards. Two inputs in Equa- The major problem with estimating instantaneous
tions 1 and 3 need elaboration: 1) procedure for operating speeds from GPS probes is their lack of
establishing the safe design speed vit along the precision as affected by real-world road and traf-
route, and 2) procedure for treating uncertainty fic conditions commonly encountered along the
in the GPS speed estimates at location i for vehi- route (Chalko 2007). This concern was addressed
cle m (vi,m). experimentally by Guido et al. (2014), with respect
to a GPS benchmarking traffic study. This study
yielded GPS confidence intervals for instantaneous
3 SAFE DESIGN SPEED MODEL vehicle operating speeds, subject to varying satel-
lite signal disruptions en-route. A brief description
The AASHTO Highway Safety Manual protocols of this benchmarking exercise has been presented
define safe design speeds to consider “… a number in this section of the paper.
of important factors, such as, underlying topogra- The level of precision associated with GPS
phy, anticipated operating speed of traffic stream output is related to the number of satellites in
over the longer tract of road, adjacent land use view and their relative position. This is normally
and development density, and the functional clas- reported by the GPS receiver in terms of its Cir-
sification of the highway” (AASHTO 2010). cular Error Probable (CEP) range. A device cali-
In this paper, the design speed is calculated based bration exercise is carried out to establish CEP in
on the V85 model introduced by Lamm (2002). terms of different radii centered on a known fixed
For a two-lane highway and uncongested traffic ground control point. Since this is a result of a
conditions, the safe design speed was established “fixed” ground control test, CEP may not be able
by Lamm by first estimating the V85 speed along to account for the myriad of factors affecting error
the route. V85 reflects the operating speed that is under dynamic road and traffic conditions.
exceeded by 15% of vehicles in the traffic stream, Guido et al. (2014) obtained a sample of smart-
thus providing a measure of the upper range of phone probe vehicle operating speeds for vary-
“risk behavior” in the driver sample. For two-lane ing satellite signal reception and compared these
undivided highways, Lamm suggested a V85 safe speeds to benchmark values obtained using a high
design speed of 105.3 Km/h for tangents under frequency video V-Box (Race Logic ®) mounted
ideal uncongested conditions. on the same vehicle. Lack of precision in the speed
The tangent design speed is reduced accord- estimates was expressed as the signal Root Mean
ingly to account for restrictions to mobility caused Square Percentage Error (RMSPE).
by sharper horizontal curvature. For such curves,
Lamm suggested a safe design speed expression of 2
the form: ⎡ Si Sˆi ⎤
N

∑ ⎢ ˆ ⎥
i =1 ⎣ Si ⎦
RMSPE = (6)
v t ( km / h ) = 105.31 + 2 ∗ 10 5 ∗ CCR
CCRS2 − 0.071 ∗ CCRS (4) N

Where CCRs reflects the curvature change rate where

and is used to tailor the safe design speed for tan- Si = speed of probe,
gents to account for differences in horizontal align- Ŝi = V-Box benchmark speed,
ment. For circular and transitional curves, CCRS N = number of probes.
is expressed as a function of curve length and
radius, such that: In order to obtain estimates of RMSPE, a rela-
tionship was established between the reported
signal CEP range and its corresponding RMSPE
⎛ Lc11 Lcrcr Lc12 ⎞
+ + ∗ 63.7 for a sample of probe vehicle passes. Instanta-
⎝ 2R R 2R ⎠ neous speeds obtained from four GPS-enabled
CRS =
CCR (5)
L smartphones were compared to benchmark values

595

TIS2017_Book.indb 595 2/8/2017 8:07:45 PM

 As illustrated in Figure 2, the segment of the
SS106 used for this case study consists of a 2.4 Km
tangent, bounded by two circular curves of radius
660 m (for Curve 1) and 325 m (for Curve 2),
respectively.
Figure 3 illustrates the mean operating speeds
along the SS106 case study segment as obtained
from GPS output, with corresponding 95% con-
fidence intervals reflecting uncertainty in the GPS
signal from Guido et al. (2014). These profiles are
juxtaposed on the safe design speed of the segment
Figure 1. RMSPE mean and 95th percentile confidence
intervals for different CEP values.
from Lamm’s formulation for tangents and curves.
The mean instantaneous operating speeds were esti-
mated for 32 GPS vehicle passes along the segment
obtained from the high frequency V-Box receiver. in increments of 100 m. About 3,500 of instantane-
Confidence intervals for probe operating speeds in ous speed data were acquired by the probe vehicles
the sample were subsequently obtained experimen- (GPS) along the highway stretch. Mean operating
tally for varying road and traffic conditions. speed of this sample of vehicles was then calculated
Figure 1 illustrates the relationship between for each 100 meters segment as in Equation 2. Sites
mean RMSPE and CEP, along with corresponding where operating speeds exceed safe design thresholds
RMSPE confidence intervals (95%). can be viewed as posing potential safety problems.
From Figure 1, we note that the width of the For this case study, the mean operating speeds
95% confidence interval becomes decidedly more were found to be close to the safe design thresholds
pronounced with increasing CEP values, suggest- for curve 1 and tangent portions, but exceeding these
ing increased uncertainty in the GPS speed estimate thresholds in the sharper curve 2 portion of the
with increased loss of satellite signal strength. segment. The upper bound speed profiles (dotted)
A series of CEP-related speed adjustments for suggests that there is a 2.5% chance that operating
mean and 95% confidence intervals (“lower bound” speeds will exceed safe standards as recommended
and “upper bound”) were assessed based on this by AASHTO based on the GPS sample. This sug-
analysis and then applied to the case study exercise. gest that although “on average” operating speeds can
be viewed as being safe, these is a chance that these
results may be biases due to errors in the GPS receiver.
5 ANALYSING SAFETY PERFORMANCE This gives a kind of statistical “safety margin” that
FOR CASE STUDY HIGHWAY SEGMENT we may wish to consider as a basis for possible inter-
vention along the case study highway segment.
A case study based on GPS-based instantaneous
speeds is used to yield the TSPI for a two-lane rural
highway segment. This case study seeks to provide
insights into several practical safety questions:
Which sites along the highway segment are more
prone to crashes? How well do these sites corre-
spond to locations of reported crashes (total and
single vehicle speed-related)? How is the assessment
of safety along the highway affected by uncertainty
in GPS instantaneous speed estimates? Do GPS-
Figure 2. Case study highway segment.
based instantaneous speeds provide a reliable met-
ric for assessing road safety performance?
The SS106 is a two lane undivided highway
in southern Italy serving about 5,000 to 20,000
vehicles per day (De Luca 2015). Historically the
number of crashes along this highway have been
high and it’s considered to be one of the most dan-
gerous two-lane roads in the country. The crash
mortality rate for this highway is also high, averag-
ing 8.5 fatalities per 100 crashes, as compared to a
national average of 4.9 fatalities per 100 crashes.
The major cause of crashes along this highway is
attributed vehicles travelling too fast for the condi- Figure 3. 95% confidence interval for the vop speed pro-
tions (geometric). file and safe design speed threshold.

596

TIS2017_Book.indb 596 2/8/2017 8:07:46 PM

 Figure 4 illustrates the upper bound profile for Conceptually crashes can be viewed from two
TSPI for the SS106 case study, based on higher risk perspectives: a) crashes that result from safety
driver behavior. Let’s consider two safety interven- problems at a given location (consistent), or b)
tion thresholds (A and B). Threshold A suggests random “acts of God” or crashes that result
intervention for elements where TSPI ≥ 2 and B from difficult-to-predict events. It is reasonable to
(less tolerant of risk) where TSPI is greater than assume that if crashes occur with consistency, their
zero. Threshold A would suggest intervening in location would correspond to sites with higher val-
the sharper Curve 2 areas; while Threshold B rec- ues of TSPI or risk. As defined in this paper, TSPI
ommends intervention along the entire case study focuses on deviations between vehicle operating
segment (curves and tangent). This latter threshold speeds and safe design thresholds. This measure
would obviously come at a higher intervention cost. of TSPI reflects risk of crashes related to exces-
Presumably, threshold B would be adopted for a less sive speed, normally involving a single vehicle. Pre-
restrictive safety budget and a reduced tolerance to sumably other types of crashes would be picked up
crash risk. A similar comparison can be carried out by different forms of TSPI. Hence, for the index
considering other driver risk sensitivities, as would introduced in this paper, our intent is to compare
be reflected in a different TSPI measure. the location of single vehicle speed-related crashes
The safety index TSPI can be used directly to to TSPI as presented in Equation 1 (i.e. difference
guide intervention strategies along the case study between instantaneous operating speed and safe
segment. For this application TSPI has been aggre- threshold). If crashes are consistent, we would
gated for the tangent and curve elements, and the expect the location of single vehicle speed-related
results are summarized in Table 1 confirming the crashes to correspond to sites of higher TSPIs.
previous results that the major safety concern along The Centro di Monitoraggio Sicurezza Stradale
the case study segment is in Curve 2. For this por- in the Province of Crotone reported a total of 40
tion of the case study, TSPI has been found to be crashes for the SS106 case study segment over the
significantly higher than either Curve 1 or the tan- period 2007–2011. These resulted in 4 deaths and
gent. Based on these results, we could recommend 111 injuries. Of the 40 crashes, 19 (47.5%) took
Curve 2 be given priority in safety intervention. place on sites with high TSPI values (> 2), and
TSPI has been normalized using a feature scal- 21 (52.5%) on sites with moderate TSPI (between
ing of the form: 1–2). As illustrated in Figure 5, twenty-eight (28)
of these crashes (70%) were single vehicle and

t=
(x x ) (7)
attributed by the police to excessive speed, 57.1%
of these occurring on sites with TSPI > 2. This
(x x ) suggests that for the first SS106 case study seg-
ment, speed-related crashes took place at locations
where t is the normalized TSPI value, x is the with higher TSPIs, where drivers presumably over-
observed TSPI value, and xmin and xmax are the min- estimated safe design speeds for the underlying
imum and maximum TSPI values, respectively. road geometry.
The kinds of intervention that are likely to be
most effective in reducing crashes are improve-
ments in highway geometry along the curves, or
an improvements in driver speed advisory and
enforcement such that drivers are more aware of
geometric/traffic mobility restrictions along the
segment. The nature of the relationship between
different treatments and changes in TSPI and the
number of likely crashes is not considered to be
within the scope of this paper.

Figure 4. Track safety performance indices that exceed

thresholds for intervention.

Table 1. TSPI for curve 1, tangent section and curve 2.

Length Radius TSPI

ID Section (m) (m) (U.B.) t

1 Curve 1 850 660 1.714 0.450

2 Tangent 2,400 – 0.973 0.256
3 Curve 2 600 325 3.384 0.890 Figure 5. Locations of crashes due to speeding on the
SS106 segment.

597

TIS2017_Book.indb 597 2/8/2017 8:07:46 PM

 6 TRANSFERABILITY OF RESULTS

One of the major issues in using mobile GPS probe

data for input into safety performance analysis is
the transferability of the results. To address this
issue a second 10 Km stretch of two-lane highway
(SS107) was selected for analysis (Fig. 6).
For this application 85th percentile of TSPI pro-
files were obtained based 45 mobile GPS probes. Figure 6. Locations of speed-related and totally crashes
As illustrated in Figure 6, the SS107 case study for 10 Km SS107 case study route.
Table 2. TSPI and crash rate for the SS107 segment (2009–2013).

ID Curve or Tangent Length (m) Radius (m) TSPI t Crashes1 Rate2

1 C1 48 220 −0.114 – 0 0
2 T1 86 0 −0.122 – 0 0
3 C2 58 186 −0.126 – 0 0
4 T2 94 – −0.149 – 0 0
5 C3 54 141 −0.174 – 0 0
6 T3 74 – −0.168 – 0 0
7 C4 320 117 −0.120 – 0 0
8 T4 299 – −0.052 – 0 0
9 C5 355 181 −0.030 – 0 0
10 T5 194 – −0.090 – 0 0
11 C6 236 211 −0.060 – 0 0
12 T6 195 – −0.023 – 0 0
13 C7 73 171 −0.031 – 0 0
14 T7 116 – −0.029 – 0 0
15 C8 403 408 −0.033 – 0 0
16 T8 156 – −0.052 – 0 0
17 C9 578 618 0.033 0.246 1 0.173
18 T9 337 – 0.122 0.911 3 0.889
19 C10 220 920 0.086 0.638 2 0.908
20 T10 205 – 0.089 0.665 0 0
21 C11 170 658 0.134 1.000 2 0.076
22 T11 393 – 0.115 0.856 3 0.764
23 C12 358 287 0.084 0.628 2 0.559
24 T12 128 – 0.055 0.408 0 0
25 C13 73 428 0.008 0.063 0 0
26 T13 316 – 0.029 0.213 1 0.316
27 C14 372 169 −0.133 – 0 0
28 T14 195 – −0.041 – 0 0
29 C15 366 184 −0.015 – 1 0.273
30 T15 138 – −0.058 – 0 0
31 C16 246 243 −0.038 – 1 0.407
32 T16 604 – −0.092 – 0 0
33 C17 286 176 −0.269 – 0 0
34 T17 376 0 −0.308 – 0 0
35 C18 122 529 −0.101 – 0 0
36 T18 453 0 0.077 0.574 2 0.442
37 C19 122 224 0.084 0.626 0 0
38 T19 189 0 0.078 0.579 0 0
39 C20 100 172 0.051 0.377 0 0
40 T20 132 0 0.067 0.496 0 0
41 C21 434 187 0.017 0.126 1 0.230
42 T21 274 0 0.008 0.056 0 0
1
Number of speed related crashes; 2Crash rate per 100 m.

598

TIS2017_Book.indb 598 2/8/2017 8:07:47 PM

 route consists of 42 distinctive geometric sections location of speed-related crashes along the SS107
(21 curves and 21 tangents). case study route. This supports the findings that
GPS probe operating speeds were adjusted for were obtained from the first case study the SS106.
satellite signal reception errors as discussed above The extraction of GPS mobile probe data from
from the previous benchmark exercise. Upper the SS106 and SS107 has demonstrated a signifi-
bound speed profiles were compared to the design cant degree of consistency for estimating instan-
thresholds for the route to produce 42 mean sec- taneous operating speeds and TSPI values. Both
tion-specific TSPI values. The design speeds for case studies provide a convincing link between
each section were obtained based on Lamm (2002), GPS-based TSPI values and speed-related crashes,
as discussed previously. i.e. the higher the TSPI, the higher the number of
A total of 64 crashes were reported by police for speed-related crashes.
these road sections over the period 2009–2013, of
which 30% (13) were single vehicle crashes attrib-
uted by the police to excessive speed. The location 7 CONCLUSIONS
of both speed-related and total crashes are shown
in Figure 6. For each of the 42 geometric ele- This paper has presented a road safety perform-
ments, estimates were obtained for TSPI, number ance study in which safety is analyzed comparing
of speed-related crashes and crash rate and, these average operating speed profiles to a recommended
have been summarized in Table 2. The crash rate safe speed thresholds along a given road seg-
is expressed in terms of the number of crashes per ment. Operating speeds have been obtained from
100 m over the 2009–2013 period. a smartphone probe vehicles sample, while safe
The purpose of this case study is to test the trans- speed thresholds are based on v85 Lamm model
ferability of the previous SS106 results, i.e. that for mainly accounting for geometric conditions. The
the SS107 higher TSPIs correspond to sites with a assumption is made that each driver sets his indi-
higher number of reported speed-related crashes. vidual operating speed based on his perception of
A close match would attest to the consistency of the safe design speed. The safe design speed can be
the findings between the SS106 and the SS107, and overestimated by drivers, and this poses potential
provide stronger evidence supporting the use of crash risks at specific road locations. Hence, safety
mobile GPS probes for targeting potential speed- performance is expressed for a given road segment
related safety problems along a highway. as the degree to which operating speeds at a given
Given that TSPI and crash rates may not be point in time exceed safe design standards.
normally distributed, a nonparametric Spearman A safety performance index (track safety per-
rank correlation was obtained to test similarities formance index or TSPI) was formulated to assess
between TSPI values and reported speed-related lack of safety over the case study segment, account-
crash rates for the 42 geometric elements along the ing for the difference between the probes speed
SS107. A Spearman ranking index is of the form: profiles and selected safe design thresholds. The
TSPI metric introduced in this paper only reflects
n
speed-related crashes caused by restrictive road
6 ⋅ ∑ di2 geometry. Other safety concerns could be consid-
ρs = 1 − i =1
(8) ered using different expressions for TSPI from that
n (n − ) introduced in this paper.
Two case studies were analyzed for two-lane
where: highways in which instantaneous operating vehicle
di = rank difference between two samples. speeds were obtained from a GPS equipped smart-
n = sample size (no. of sections). phones. Operating speed profiles from the probes
Higher values of ρs reflect reduced differences sample were then adjusted to account for uncer-
between TSPI and speed-related crash rates. tainty in the speed estimates under dynamic road
As summarized in Table 3, the correlation and traffic conditions. The results of this analysis
parameter (ρs) in the Spearman test was found to show that TSPI is affected by the uncertainty in
be significant at the 5% level. Hence it can be con- speed estimation, as well as by the thresholds used
cluded that TSPI provides a good metric for the to guide intervention.
The case study applications have confirmed that
GPS mobile probe measures of TSPI closely match
Table 3. Spearman test for the SS107 sections. sites of single vehicle speed-related crashes for sev-
eral years of reported data. The two case studies
N ρs d.o.f. t p-value
yielded consistent results concerning the relation-
42 0.60 40 4.694 <0.001 ship between TSPI and speed-related crashes, i.e.
as the TSPI at a given location is increased the

599

TIS2017_Book.indb 599 2/8/2017 8:07:47 PM

 observed number of speed related crashes at this Guido G., Vitale A., Saccomanno F.F., Festa D. C.,
location is also increased. Astarita V., Rogano D., & Gallelli V. 2013. Using
Notwithstanding the preliminary limitations smartphones as a tool to capture road traffic attributes.
of the smartphone sample used in this study, the Applied Mechanics and Materials, 432: 513–519.
Guido, G., Gallelli, V., Saccomanno, F.F., Vitale, A.,
proposed methodology has yielded some promising Rogano, D., & Festa, D. 2014. Treating uncertainty
results. The approach has been shown to yield relia- in the estimation of speed from smartphone traffic
ble traffic (instantaneous speed) attributes for indi- probes. Transportation Research Part C, 47: 100–112.
vidual vehicles in the traffic stream. These can be Guido, G., Vitale, A., Astarita, V., Saccomanno, F.F.,
used as inputs into safety assessment and in devel- Giofré, V.P., & Gallelli, V. 2012. Estimation of safety
oping cost-effective safety treatments. Of course, performance measures from smartphone sensors. Pro-
uncertainty in safety performance indicator needs cedia Social and Behavioral Sciences, 54: 1095–1103.
to be taken into account to ensure that decisions Haglund, M. & Åberg, L. 2000. Speed choice in rela-
concerning safety performance from smartphone tion to speed limit and influences from other drivers.
Transportation Research Part F, 3: 39–51.
probe estimates are not based on incorrect unreli- Hassan, H. & Abdel-Aty, M. 2011. Exploring visibility-
able measures of speed and other traffic attributes. related crashes on freeways based on real-time traffic
A comparison among TSPI measures along flow data. Proceedings of the 90th Annual Meeting of
the segment closely reflects historical crashes sites the Transportation Research Board, Washington, D.C.
along the case study segment. This provides an Hayward, J. 1971. Near misses as a measure of safety
interesting dimension to the analysis. If historical at urban intersections. PhD Thesis, Department of
crashes occur where risks have been identified by Civil Engineering, The Pennsylvania State University,
TSPI as being high, then we would expect these Pennsylvania.
crashes to exhibit a high degree of consistency and Herrera, J.C., Work, D.B., Herring, R., Ban, X., Jacob-
son, Q., & Bayen, A. 2010. Evaluation of traffic data
hence reflective of potential safety problems. On obtained via GPS-enabled mobile phones: the Mobile
the other hand, if historical crash sites don’t match Century field experiment. Transportation Research
high TSPI sites, then crashes along the segment Part C, 18: 568–583.
can be viewed as being essentially random events Huguenin, F., Torday, A., & Dumont, A. 2005. Evaluation
and difficult to predict and resolve. of traffic safety using microsimulation. Proceedings
of the 5th Swiss Transport Research Conference –
STRC, Ascona, Swiss.
REFERENCES Lamm, R., Psarianos, B., & Cafiso, S. 2002. Safety
Evaluation Process for two-lane rural roads: a 10 year
Aarts, L., & van Schagen, I. 2006. Driving speed and the review. Transportation Research Record, 1796: 51–59.
risk of road crashes: A review. Accident Analysis and Mannering, F. 2009. An empirical analysis of driver per-
Prevention, 38: 215–224. ceptions of the relationship between speed limits and
AASHTO 2011. A Policy on Geometric Design of High- safety. Transportation Research Part F, 12: 99–106.
ways and Streets. 6th Edition. Minderhoud, M. & Bovy, P. 2001. Extended time to colli-
AASHTO, 2001. A Policy on Geometric Design of High- sion measures for road traffic safety assessment. Acci-
ways and Streets. 4th Edition. dent Analysis and Prevention, 33: 89–97.
AASHTO, 2010. Highway Safety Manual. First Edition. Nilsson, G. 2004. Traffic safety dimensions and the power
Amundsen, F.H. & Hyden, C. 1977. Proceeding of First model to describe the effect of speed on safety. Bulletin
Workshop on Traffic Conflicts. Institute of Transport 221, Lund Institute of Technology, Lund.
Economics, Oslo/Lund Institute of Technology, Oslo, Saccomanno, FF., Cunto, F., Guido, G., & Vitale, A.
Norway. 2008. Comparing safety at signalized intersections
Astarita V., Giofré V., Guido G., & Vitale A. 2011. Inves- and roundabouts using simulated traffic conflicts.
tigating road safety issues through a microsimulation Transportation Research Record, 2078: 90–95.
model. Procedia Social and Behavioral Sciences, 20: Stamatiadis, N., Grossardt, T., & Bailey, K. 2009.
226–235. How driver risk perception affects operating speeds.
Bierlaire, M., Chen, J., & Newman, J. 2010. Modeling Route Advances in Transportation Studies an International
Choice Behavior From Smartphone GPS data. Report Journal, 17(A): 17–28.
TRANSP-OR 101016, Transport and Mobility Labo- Sümer, N., Özkan, T., & Lajunen, T. 2006. Asymmetric
ratory, Ecole Polytechnique Fédérale de Lausanne. relationship between driving and safety skills. Acci-
Chalko, T.J. 2007. High accuracy speed measurement dent Analysis and Prevention, 38: 703–711.
using GPS (Global Positioning System). Melbourne, Taylor, M.C., Lynam, D.A., & Baruya, A. 2000. The
Australia. effects of drivers’ speed on the frequency of road
De Luca, M. 2015. A Comparison Between Prediction crashes. TRL Report, No. 421, Transport Research
Power of Artificial Neural Networks and Multivari- Laboratory TRL.
ate Analysis in Road Safety Management. Transport, Wallén Warner, H., Haglund, M., & Åberg, L. 2008.
1–7, DOI: http://dx.doi.org/10.3846/16484142.201 Drivers’ beliefs about exceeding the speed limits.
4.995702 Transportation Research Part F, 11: 376–389.
Guido G., Astarita V., Giofré V., & Vitale A. 2011. Safety Walton, D. & Bathurst, J. 1998. An exploration of the
performance measures: a comparison between micro- perceptions of the average driver’s speed compared
simulation and observational data. Procedia Social to perceived driver safety and driving skill. Accident
and Behavioral Sciences, 20: 217–225. Analysis and Prevention, 30(6): 821–830.

600

TIS2017_Book.indb 600 2/8/2017 8:07:48 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Monitoring the transport sector within the sustainable energy action

plan: First results and lessons learnt

I. Delponte
DICCA, Department of Civil, Chemical and Environmental Engineering, University of Genoa, Genoa, Italy

ABSTRACT: The Sustainable Energy Action Plan (SEAP), promoted by the Covenant of Mayor
(CoM), is a key tool for policies aimed at reducing fossil fuel consumption and GHG emissions, in accord-
ance with the Kyoto protocol and its updates. It is implemented at the municipal level and is constituted
by a Baseline Emission Inventory (BEI) and a set of measures to be implemented (Action Plan, AP). In
particular, the paper will investigate the “weight” of the transport sector in terms of emissions within the
BEI and the relevance and the variety of its actions inserted in the AP.
To achieve an actual implementation of the SEAP and to obtain its expected targets, monitoring is a
crucial component. SEAP monitoring has to look at both the progress of each single action and its global
environmental effect, which requires more than one level of evaluation and control. Economical crisis and
conjunctural effects mainly influenced the transport Co2 decrease, making AP less important, or quite
indifferent in terms of reducing. Is the plan still effective? Or a change of approach can be provided?
According to these statements, the author proposes an analysis of the case of Genoa SEAP (2010) and
its monitoring process, recently concluded (2015), particularly focusing on transport sector and its goals.
The activity implemented in the city of Genoa is representative of that challenge and is a helpful test case
for other cities addressing the same issues.

1 INTRODUCTION achieving the score of reducing energy demand.

In the same year, the so-called SET plan (Strategic
1.1 General background Energy Technology plan) was promoted, a strat-
egy dealing with the energy field new technologies
Climate change concern has brought several differ-
to be implemented. With the 2008, the engage-
ent approaches to manage and reduce greenhouse
ment of EU achieved meaningful pillars by means
gas emissions connected with energy generation
of fundamental instruments, such as the first
and consumption, at both global and local scales
was the Climate Action, which promotes strate-
(Wilbanks & Kates 1999, GHG Protocol 2015). In
gies for all the involved stakeholders at long and
this trend, a leading role has been certainly played
short terms. Afterwards we have the 2nd Strategic
by the European Union, which from the first years
Energy Review, which introduces the well-known
of this century has been implementing environ-
“20–20–20” strategy.
mental policies to face climate change scenarios
More recently, the European Commission pre-
and favorite low emission actions (Mertens 2011,
sented the “Roadmap for moving to a low-carbon
EU Climate, 2015).
economy in 2050” (2015). This Roadmap aims at
Summing up briefly the steps of the engagement
a reduction of GHG emissions in the EU 27 by at
process by EU in the energy sector, a particularly
least 80% in 2050 vis-a-vis emissions in 1990.
meaningful moment was when, in 2005, was explic-
itly expressed the need of a shared policy at the UE
level around these topics. The first result was the
1.2 The Sustainable Energy Action Plan (SEAP)
publication, in 2006, of the Green Paper Energy
“A European Strategy for Sustainable, Competi- After the European Directives, the Member
tive and Secure Energy”, anticipating the exigent States adopted the targets, drawing up national
of a common planning on energy efficiency and Action Plans for the emissions reducing, since
RES exploitation. the first Two-Thousands. But in consequence
In 2007 the Action Plan for Energy Efficiency of the adoption of the Renewable Energy and
for the 5-years period 2007–2012 was drawn up, Climate Change Package in 2008, the European
containing the targets of 20% of reducing and Commission reckoned to launch, at local entities
the definition of the fields of intervention for scale, the initiative of the Covenant of Mayors

601

TIS2017_Book.indb 601 2/8/2017 8:07:49 PM

 (CoM), with the aim of sparking and support accurate, complete and, most of all, available in
the efforts by Municipal Administrations, as the future to be compared in the following years
the basic unit of the public administration and during the monitoring activities through 2020. In
citizens, in the process of actualization of energy order to quantify the CO2 emissions, EC in the
and climate change policies (Alberti & Mar- SEAP guidelines gives two methodologies. Each
zluff 2004, Derissen et al. 2011). In this way, the local authority can select between the “Standard”
decisive role of municipalities in the mitigation emission factors method, in line with the IPCC
of the main causes (and consequent effects) of (Intergovernmental Panel on Climate Change)
climate change was acknowledged, above all tak- principles which consider the CO2 production
ing into account that the 80% of the energy con- due only to energy consumption within the terri-
sumptions and production of CO2 is associated tory of the Municipality, or the LCA (Life Cycle
to urban activities. Assessment) approach in order to assess, on the
The CoM initiative, launched on 29th Janu- other side, the impact of the whole cycle process
ary 2008 by the European Commission, and the of energy production.
planning tool it promotes, the Sustainable Energy After the data collection and the compilation
Action Plan (SEAP), is located within this frame- of the BEI the analysis of the results leads to the
work and aimed to promote the implementation identification of the main criticalities and the main
of EU commitments on the Kyoto Protocol, by areas by which improvements are to be attained.
means of an unilateral and voluntary participation In response to those, the Action Plan (AP) is
of European cities. the operative tool by which getting the definition
The SEAP is based on the results of the “Base- of sustainable development strategies, regulations,
line Emission Inventory” (BEI), which quanti- and actions, in accordance with the environmental
fies the energy consumption of the territory for policy adopted by the Municipality.
the adopted reference year, and identifies several Hence, the aim of the Action Plan is to set out
long (LT) and short term (ST) actions in differ- actions and to define priorities in respect to the
ent priority areas, in order to get the expected different fields for the reduction of CO2 emissions,
GHG reduction. Drawing up, implementation both in public and in private sector. Actions are
and monitoring are the three integrated phases outlined with different time-horizon: short-term,
by which the goals of the SEAP can be achieved, for intermediate targets within at most 4/5 years
through a coordinate initiative at Municipal level from the commencement of the SEAP, and long-
involving public institutions, private stakeholders term, referring to the strategy towards 2020.
and simple citizens. In this way, the SEAP turns The priority setting process to rank the foreseen
out to be a key document for the authorities that measures in order to arrange their relevance has to
defines the municipal energy policies and, at the be defined by means of different criteria, involv-
same time, a dynamic tool to be upgraded and ing the stakeholders so as to sort the most effec-
optimized on the obtained results and the benefit/ tive actions. In this document every single action
cost ratio of every action in view of the compli- is to be precisely described and, moreover, all its
ance of the EU objectives. details must be clearly defined (such as objectives,
In order to quantify the emission reduction tar- expected saving and CO2 emission reduction as
gets and the achievements of the SEAP actions well as the responsible in charge for the action,
within 2020, an overall view as a reference situa- timetables and deadlines, involved players and
tion in terms of final energy flows and CO2 emis- stakeholders, budget and risk analysis) in a form:
sions inside the Municipality territory is needed. this precise work reveals to be fundamental in the
In this way, the Baseline Emission Inventory tool following implementation of the SEAP, since the
has the function to quantify the amount of CO2 forms become a key reference document during all
emitted due to energy consumption in the territory the subsequent phases of the CoM.
of the Covenant signatory in a given baseline year. It is a shared idea that only through accurate
Thus, the Baseline Emission Inventory is consid- monitoring activities and tracking progress can
ered a strategic instrument because it provides real SEAP implementation can be achieved. In
the quantification and the sources areas of main 2014, the Covenant of Mayors Office (CoMO),
CO2 emissions to set the priorities of the future in collaboration with the Joint Research Centre
reduction interventions and to assess the impact (JRC) of the European Commission, released the
and the progress of the SEAP actions. The final “Reporting Guidelines on Sustainable Energy
energy consumptions are divided in two sectors: Action Plan and Monitoring” document a tool
“Buildings, equipment/facilities and industry” aimed to control and check the progress of SEAPs.
and “Transport”. The baseline data collection is For the implementation of a Sustainable Energy
to be implemented in order to be representative Action Plan in fact, after planning, has to take
of the considered local situation; the BEI must be into account the changing and updating needs, the

602

TIS2017_Book.indb 602 2/8/2017 8:07:49 PM

 knowledge scenario and the related administration 2 THE CASE OF GENOA
initiatives; simultaneously, the territory feedback
and the economic and regulatory framework also 2.1 The SEAP of 2010
need to be considered. In this sense, monitoring
The Municipality of Genoa joined the initiative
activities are supposed to be the way to control
on February 2009 and its Sustainable Energy of
processes and to recalibrate objectives and instru-
Action Plan (SEAP), with the aim of a 23,7% CO2
ments of implemented measures. The assessment
reduction by 2020, has been the first to be officially
phase deriving from the monitoring should be able
published by the EU Commission.
to refine the approach in light of the needs and dif-
For the Municipality of Genoa, the reference
ficulties. Thus, through a multi-stage strategy, we
baseline year was set in 2005 (first year with the
are able to develop virtuous tools for the imple-
whole data available) and the CO2 inventory was
mentation of actions.
based on the overall consumptions of energy, except
The SEAP, thus, is a concrete working tool to
those coming from the industrial sector. The “Stand-
be implemented and, meanwhile, strictly moni-
ard” method was used to evaluate GHG emissions,
tored, through which CoM’s Signatories are
by following IPCC guidelines and particularly the
committed to approach and to achieve the final
“bottom-up” approach, based on the fuel end-use
objectives established in the strategic vision. In
in the several fields. The results of the data analysis
this sense, a forward-looking vision of the local
for year 2005 of the energy flows in Genoa are clas-
administrations is a key-point to a successful
sified as required in the JRC template not including
approach to the SEAP, like to the CoM from
energy consumption data of industries and of long-
which it follows.
distance transports (railway, highways, sea and air
From the operative standpoint, as stated in
transport). The total CO2 emissions have been calcu-
SEAP Guidelines, CoM signatories are commit-
lated by means of the National/European Emission
ted to producing two documents after the SEAP
Factors as the total of the contributions from each
submission. The first one, to be submitted every
energy source. In compliance with SEAP guidelines,
two years, is an implementation report containing
the presented data do not reflect the whole energy
qualitative and quantitative information on inter-
flow of the city of Genoa but only the emissions
ventions to evaluate, monitor and verify the status
generated by the sectors addressed by the SEAP,
of the Action Plan (SEAP Implementation Status)
namely the civil sector and local transports. Energy
and its effect; the second one is an update of the
consumption scenario in Genoa shows that the civil
CO2 emission inventory, named the Monitoring
sector represents the by-far prevailing use of energy
Emissions Inventory (MEI), to be compared with
(77%) in respect to the transport sector. Particu-
the previous BEI for monitoring the progress in
larly, natural gas is revealed as the main fuel source
terms of emission reductions every four years.
being used in Genoa city and electricity consump-
CoM provides a monitoring template for the
tion results significant, both of them primarily used
SEAP Implementation Status, in which every
in tertiary and residential sectors.
measure presents new fields to be filled in such as
Within the two sectors the SEAP focuses on -
staff capacity allocation, overall budget spent so
civil field and local transport-, CO2 emissions from
far and, where possible, main barriers encountered
electricity consumption (35%), from natural gas
during SEAP implementation.
consumption (35%) and from fuel (gasoline + Die-
From the operative viewpoint, MEI appears
sel) consumption for local transports (27%) result
to be an updated version of the BEI referring to
to be very similar. From the BEI, the overall CO2
a specific monitoring year in order to easily com-
emissions in the baseline year 2005 resulted of
pare the two documents and to track the progress
2.142.484 tons: this value has been considered as
of SEAP implementation. As the Baseline Emis-
the 100% reference target to be compared in order
sions Inventory, the MEI calculates the current
to evaluate and quantify the effectiveness of the
amount of final energy consumption and the asso-
mitigation SEAP actions (Schenone et al, 2015).
ciated CO2 emissions in terms of energy carrier
and sector such as building equipment facilities,
industries, transport, agriculture, forestry and fish-
2.2 The monitoring of 2015
eries. Indicating the fuel emission factors makes it
then possible to automatically evaluate the associ- Due to the delay in publishing the monitoring tem-
ated CO2 emissions. Furthermore, information on plate by CoM, original deadline for submitting the
energy supply (municipal green energy purchases, Monitoring Template will be set in 2015 (instead
local/distributed electricity and heat/cold produc- of 2014), in order to become familiar with the
tion generated from renewable energy sources and reporting framework.
CHP plants) must be included in the related parts Considering that Genoa was the first city to
of the MEI template. attain the SEAP approval, it was among the firsts

603

TIS2017_Book.indb 603 2/8/2017 8:07:49 PM

 as well (with Glasgow and Goteborg for instance, Table 1. BEI 2005 and MEI 2011 GHG emission in
to cite a few with similar characteristics) to submit Genoa Municipality territory (tCO2).
the report, so it is quite impossible to make in-
depth comparisons with other results coming from SECTOR BEI 2005 MEI 2011
the monitoring phase. A wide literature on the field Municipality 114.801 100.889
is supposed to be ready and consistent in a couple Tertiary 606.851 609.117
of years. Residential 934.340 831.458
By the way, as a frontrunner in the drafting Public Lighting 18.257 17.293
of the Monitoring Emission Inventory, Genoa Transport 468.235 422.226
Municipality used data collection and processing TOTAL 2.142.484 1.980.982
tools according to a logic of governance and coor-
dination with local organizations and individuals
to seize opportunities resulting from initiatives of On the other side, the environmental monitor-
local, regional and international level. ing concerned the estimated energy savings and
Just as with the BEI, the MEI compilation is also the associated reduction of CO2 to 2015 due to
quite complex because of the difficulty in collect- each action, as required by the JRC Monitoring
ing consistent and coherent data. In many cases, in Guidelines. In addition, where possible, values of
fact, the availability of complete data sources with energy production from renewable sources have
the same level of granularity (or aggregation) is been shown.
almost impossible, thus making statistical process- It should be underlined that the percentage allo-
ing or adoptions of other indicators necessary. cation of CO2 savings was conferred by assessments
Based on these considerations and on the gov- differing from case to case, in order to preserve the
ernance process, it was decided to report the MEI uniqueness of cases and the specificity conducted
to 2011, the year for which the Regional Environ- estimates. This allowed highlighting the presence
mental Information System of the Region of Lig- of critical issues affecting interventions perform-
uria (where Genoa is located) is able to produce ance and, by means of the comparison between
complete energy balances at the regional, provincial the progress and the environmental monitoring, to
and municipal levels, and from which it is possible better understand the nature of the criticalities and
drawing information on final energy consumption then identify possible corrections or incentives.
of Genoa territory. The variety of monitored actions can be attrib-
The sector with the highest energy consumption uted in general to some typical situations. In some
and from where the most of the GHG emissions cases, actions progress can appear at an advanced
derives is the civil one (public administration, com- stage, but the CO2 reduction imputation is still
mercial, residential and public lighting), represent- equal to 0, because of the expected benefits will be
ing the 77% of the total, against the 23% incidence activated with the commissioning. This is typical
of the transport sector. of infrastructural interventions, which also require
In the civil sector the leading cause of fuel con- the convergence of other actions to work at full
sumption and emissions is due to the residential performance. Another case is when CO2 allocation
area affecting the total with 42% in the case of is corresponding to the progress percentage, i.e.,
emissions; then the tertiary, with a weight of 31%, with regard to plants intervention actions, where
the public administration (5%), and public light- the implementation immediately leads to the sav-
ing (4%). ings achievement. Moreover, in the case of actions
Comparing with the BEI, for the consumption to be redefined, points not equal to 0 have been
of 2011 it was observed a reducing effect of −9,5% assigned in the progress status where some alter-
thanks to the civil sector and −2,5% to transport, native hypotheses have already emerged in the
which produced an average decrease of CO2 emis- internal comparison to the administration. This
sions of the −9,4%. has been done in the belief that preserving the sus-
After calculating the MEI, the SEAP actions tainability objective of the plan, it is possible and
monitoring has been conducted with an approach adequate, in some cases, proceed with a relocation
designed to define both the progress of single of the initiatives on the occasion of funding or new
actions and its environmental monitoring. In par- perspectives not yet present in the early stages of
ticular, the action progress analysis consisted in for the SEAP preparation.
the improvement verification updated to 2015 of It is noted that the allocation of the CO2 sav-
the SEAP interventions, both qualitatively through ing, in many cases, is the result of deterministic
the definition of eight classes (not started yet, in algorithms (even if calculated from estimates and
definition phase, started, in progress, advanced, approximations), in others a mixture of deter-
completed, postponed, canceled) and quantita- ministic methods, also through tools simulation,
tively through progress percentages. and in others on the basis of examples from the

604

TIS2017_Book.indb 604 2/8/2017 8:07:49 PM

 literature or derived from observation of urban Transport Energy Environment (TEE, a GIS-
governance dynamics. based decision support system software), the plan
According to the innovative nature of the SEAP included several actions addressed to sustainabil-
tool, the carried on monitoring activities stands ity, strictly verified for their estimation.
out for its experimental nature, requiring a con- Regarding private and commercial vehicular
tinuous monitoring in terms of scientific content circulation, two types of estimation were executed,
and methods proposed. one related to the urban fleet and the other to
Regarding the progress of Genoa’s SEAP traveled paths. In the first case, compared with
actions, almost all of the interventions have been the percentage distribution by national estimate,
completed, started or, at least, defined. During we found that the fleet was not far from the read-
the monitoring activities critical issues emerged, ings, although some categories were more precise
related to interventions no longer valid due to because of the obvious standardization of national
updates of administrative policies or extreme dif- conditions. Some algorithms were run to configure
ficulty in the implementation, making necessary actual metropolitan fluxes (Delponte et al. 2013).
for some actions to be deleted. Plus, four actions After extracting the Genoa fleet and fluxes, to every
related to local electricity production at the end of vehicular typology was assigned features regarding
2015still not started mainly due to technological fuels of the circulating fleet (parameters of con-
hurdles and realization extreme difficulties (Del- sumption owing to Euro 0,1 and other categories).
ponte et al. 2013). Using not average but actual travel patterns (and
their lengths) avoids an estimation error about 25%
of emissions caused by approximation, owing to a
3 FOCUS ON TRANSPORT downscaling effect that overestimates current met-
ropolitan conditions relative the rest of the region.
The SEAP comprises numerous municipal com- The simulation also determines foreseen actions
petencies to attain a meaningful reduction of CO2 concerning the public sector, in which it is conven-
emissions and an energy policy of strong impact ient to invest because of great difficulties in using
directly addressed to citizens and urban transport private automobiles (lack of parking and space
being one of these. To reduce urban GHG emissions devoted to infrastructure). In the development of
from transport, it is important to examine all param- the tool, it was planned to implement a system of
eters contributing to emissions, which are usually very urban mobility to enable easier access to and move-
strictly related to city shape and settlement location ment around the city, thanks to alternative means
(Kennedy et al. 2009). There are many actions that of transport, surface and underground local public
could be copied from one city to another to optimize transport, cycle paths, pedestrian isles, intermodal
urban movements. However, the majority of these use of public elevators, funiculars, and introduction
actions must be tailored to the geomorphology and of increased water-based transport (Navebus, as in
transportation attitude of citizens (Van de Coever- Table 2). The program regarding rationalization of
inga and Schwanenb 2006, Rickwood et al. 2008). the municipal fleet and greater use of a car-sharing
The complex nature of connections between system for internal needs were also sorted for attain-
land-use transformation, climate change, and ing the residual percentage of reduction. This reduc-
energy requires new investigations from global to tion is not significant from a numerical standpoint
local levels. but is important to the leading role of public admin-
This paper mainly focuses on the third sec- istration as an example for citizen behavior.
tor concerning municipal transport measures, Actions for private and vehicular circulation
although all of the above-mentioned aspects are were simulated together and then assessed for the
closely interrelated and involve other urban plan- pursuit of the general target. Establishing a unique
ning issues, such as density, the presence of an simulation, the foreseen percentages were recalcu-
appropriate mix of activities (reducing local mobil- lated with consideration of the multiplier effect of
ity), and the role of green areas (mitigating sum- contemporaneous realization of the interventions.
mer heat) (Steemers 2003, Wilson 2006, Banister The concurrent actions were allowed in such a way
and Anable 2009, Wilson and Piper 2010, Romero- to maximize repercussions on the integrated trans-
lankao 2012, Pasimeni 2014). port system.
Geographic peculiarities, combined with the From the quantitative point of view, the more
role of Genoa as a port and seafaring city, has effective actions related to transport emissions
influenced both its urban development and infra- reducing are the new parking system provided
structure (which is typically linear along the coast). (high tariffs for the city-centre and an annual fee
Thanks to a simulation using the software tool for residents in the boroughs) and promotion of
VISUM (for preliminary studies) and MTCP cycling and pedestrian modes (environmental
(for plan scenarios) and calculation support from islands).

605

TIS2017_Book.indb 605 2/8/2017 8:07:49 PM

 Table 2. List of local transport reduction expectations Table 3. BEI 2005 and MEI 2011 GHG emission due to
on the total BEI (in percentage, ST short term, LT long transport in Genoa Municipality territory (tCo2).
term).
BEI 2005 MEI 2011
Actions Reducing
Municipality 9.838 8.836
Municipal Fleet Public transport 32.765 43.595
Rationalization of use of the 0,036% Private transport 425.631 369.795
municipal fleet_ST TOTAL 468.235 422.226
Renewal of the municipal 0,010%
fleet_LT
Local public transport
Eco-friendly fleet transition 0,20% first considered. The second step was taking into
plan_ST the account the divergence on traffic fluxes. The
Eco-friendly fleet transition 0,18% traffic congestion monitored along the incurred
plan_LT years showed a reduction of −5,6% on the urban
Strengthening of the local 0,50% area, while the renewal of the vehicular fleet is
railway system_LT
not so far from the National statistics (Euro 0 and
Elevators and funiculars_ST 0,20%
Euro 1, most pollutant, have almost disappeared).
Elevators and funiculars_LT 0,40%
Matching the two previous aspects and con-
Navebus_LT 0,03%
sidering also that the network was not strongly
Protected axes_ST 0,60%
changed, the results showed 369.795 tCO2 for the
Protected axes_LT 0,80%
Extension of the Subway 0,30%
private transport sector, namely a −13% in respect
line_ST of the BEI.
Extension of the Subway 0,30% As far as the public side was concerned, what
line_LT already said about privates fits as well, regard-
Private and commercial transport ing general criteria. Nevertheless, in those six
Resident parking policy: 4,20% years, simultaneously there was an introduction
Blue Areas_ST of new more performing public means as well
Resident parking policy: 4,20% as an increase of energy consumptions. But this
Blue Areas_LT result was due only to the more frequent use of
Environmental islands_ST 1,40% the air conditioning on board and the relevance
Environmental islands_LT 1,40% of less pollutant vehicles has already starting to be
Interchanging hubs_ST 0,30% observed since 2013.
Interchanging hubs_LT 0,30% Looking at the monitoring phase, after 5 years,
Infrastructures_ST 3,00% Transport Actions (TRA) generally present a dis-
Large-scale infrastructures_LT 1,00% crete progress (Figure 1): most of the actions are
City Logistics_LT 0,40% in fact started, ongoing, advanced and completed
Other (above all the short-term ones –ST- which conclu-
Car sharing_LT 0,40% sion was expected for 2015). Three actions have
Soft mobility – Cycling 0,10% been postponed and two deleted, because of the
facilities_ST
extreme difficulties in the realization of the inter-
Soft mobility – Cycling 1,50%
facilities_LT
ventions. It should however be specified that the
total mobility system emitted, in recent years,
smaller quantities of carbon dioxide in respect of
foreseen trends and also less than implied on the
4 LESSONS LEARNT FROM MONITORING basis of actions contents implementation.
As shown by Genoa’s SEAP general progress in
As showed in Table 3, transportation weighted 2015, in terms of both action progress in percent-
totally 468.235 tCo2 in the 2005, with a certain age and CO2 emissions reduction, very few sectors
decrease in 2011. Going deeply, in the transport correspond to the expected values.
sector the percentage of consumption related to In absolute terms, TRA and EDI (Buildings)
private transport (89%) is higher with respect to are running well, but considering that they are the
public transport (9%) and municipal utilities (2%); decisive ones, the most efforts must be devoted to
this percentage results being almost unchanged them. This acknowledgement should determine
also within CO2 emissions. a consequent governance strategy for Genoa in
For analyzing the consumptions’ trends related implementing “core” actions. Is this already the
to metropolitan travels, the differences (2005–2011) case? Are they easy to implement, and what is
about the performances of the private fleet were at the level of feasibility of the interventions they

606

TIS2017_Book.indb 606 2/8/2017 8:07:49 PM

 – Moreover, in Genoa, an increasing percentage
of people uses bike for daily commuting; the
modal shift is partly due to the municipal ini-
tiatives, but above all for a higher sensibilization
towards sustainable alternatives in favour of less
pollutant and cheaper means of transport, as an
indirect effect of the crisis (and of the increasing
cost the the fossil fuels);
– At the end, the remote accessibility to public
services can be considered as a life-style change
in the urban areas: where ICT’s offer prevents
from physical moving, it has a notable influence
on mobility phenomena.
Some other conclusions, with a general meth-
Figure 1. Genoa’s SEAP actions progress in 2015 (in odological significance, can be drawn up as well.
percentage) compared to the expected reduction by 2020
for each sector (focus on TRA action progress).
– By the comparison between BEI and MEI, it
comes out that the reducing, due the field of trans-
port, registered in 6 years is about −2,5% of CO2
promote? Are there any “enhancement factors” tons on the total of BEI, but is quite −10% taking
that can be crucial for their realization (which into account only the comparison BEI-MEI in the
implies a large part of the overall efficiency)? sector. It means that the decrease of 2005–2011
Hence, the main outcomes here shortly repre- emissions is more relevant in the case of transpor-
sented underline a certain decrease in transporta- tation than for the other sectors of SEAP.
tion consumptions. They were also compared with – Regarding the Action Plan, the percentage of
not local trends: differences and analogies were initiatives’ realization is about 35%; the result is
verified in order to demonstrate the features of the not so satisfactory, considering that the moni-
Genoa case. This could significantly due to: toring was made in the middle of the validity of
the SEAP (the expected outcomes for 2015 were
– The bigger performances of the fleet represents
about 55% of realization, see Fig. 1).
a generalized environmental advantage, which
– But, at the same time, the emissions of the MEI
reflects upon consumptions and emissions, likely
show that the sector is reducing its potential
the international trends;
pollution of almost 10%, that corresponds to
– As come up from the Italian economical frame-
the half of the predicted reduction for all the
work of those years, after decreasing the citizens’
actions’ realization (and the trend for the upcom-
buying power, the mobility demand and the
ing years, so far, is not foreseen moving away).
resources engaged on transport sector diminish
(fuel litres, travels, fluxes...); This is due, as told before, to conjunctural
– Economical crisis (which decreasing trends are aspects that are not related to the soundness of
getting more and more visible in the years after municipal initiatives, but, largely, to vehicles per-
2011) influenced a lot the population life-style formances and reduction of mobility demand.
and, specifically in transportation, was decisive This is good for the air we inhale, because a lesser
for the ownership of the second-cars (fleet is cur- pollutant transport system is good for everybody,
rently decreasing only for cars, not for other types but a concluding remark can be pinpointed about
of vehicles) and regarding the use, less intensive, the incisiveness of municipal planning actions in
of the existent vehicles (fluxes in decrease, too). the urban transportation field; their importance in
In many cases, a mentality change clearly shifts the agenda can be absolutely underestimated in the
to more sustainable habits (for instance the dou- future, if results even better can be achieved by no-
bling of the car-sharing system members, even local effects. The Genoa case shows that a certain
if also that has been influenced by a shortage of efficacy in carrying out governance actions and
travels); achieving environmental goals has been registered
– About local context, huge efficacy has been for parking tariffs actions (which weakens more
registered for the new parking system, because and more the mobility demand): others are more
the introduction of the tariffs in the city-centre difficult to be realized, from the spatial planning
and the annual fee foreseen for residents disin- point of view, but at the same time, they are less
centivized the use of individual car for daily and incisive for the partial contribution they bring.
short travels; (and, so, reducing kilometres and Moreover, many recent reports say that the
fluxes) transport sector is one of the most difficult to be

607

TIS2017_Book.indb 607 2/8/2017 8:07:50 PM

 treated in order to reach a substantial “energy tran- EU Climate. Available online: http://ec.europa.eu/dgs/
sition”, because of the strong dependence on fossil clima/acquis/index_en.htm.
fuels, which in other sectors is decisively reducing. Genoa Sustainable Energy Action Plan Summary Avail-
So, transport planning has to regain its legitimacy able online: http://www.urbancenter.comune.gen-
ova.it/sites/default/files/archivio/allegati/SEAP%20
whereas it does not seem to be an evident urban summary_0.pdf.
priority. International Local Government Greenhouse Gas Pro-
Urban areas are worldwide responsible of tocol. Available online: http://www.ghgprotocol.org/
almost 80% of the total GHG emissions but, at city-accounting.
the same time, municipalities can be considered as Kennedy, C., Steinberger, J., Gasson, B., Hansen, Y.,
favourite areas that play a crucial role in the safe- Hillman, T., Havranek, M., Pataki, D., Phdungsilp,
guard of the environment and in the setting out A., Ramaswami, A., and Mendez, G.V, 2009. Green-
of more resilient initiatives. So, the Covenant of house gas emissions from global cities, Environmental
Mayor represents a relevant part of this strategy: Science & Technology 43, 7297–302.
Mertens K., 2011. Recent Developments of EU Envi-
through the related commitments and the SEAP, ronmental Policy and Law J. Eur. Env. Plan. Law, 8,
cities are called to work hard for facing the chal- 293–298.
lenge of GHG reduction. At this point, future Pasimeni, M.R., Petrosillo, I., Aretano, R., Semeraro, T.,
research should investigate the role of the trans- De Marco, A., Zaccarelli, N., Zurlini, G., 2014. Scales,
port sector in the SEAP tool, namely how to deal strategies and actions for effective energy planning: A
with it, respecting the identity and the general aim review. Energy Policy 65, 165–174.
of the plan and considering the increasing munici- Reporting Guidelines on SEAP and Monitoring Avail-
pal role. Maybe that lack of solutions by transport able online: http://www.covenantofmayors.eu/
planners observed in many cases can underline, by Library,84.html.
Rickwood, P., Glazebrook, G., and Searle, G., 2008.
the contrary, the opportunity to involve not only Urban Structure and Energy. A Review, Urban Policy
environmental and technological experts in solving and Research 26 (1), 57–81.
problems, but also sociological ones, taking more Roadmap for moving to a low-carbon economy in 2050.
into account urban livability’s complexity. A more Available online: http://ec.europa.eu/clima/policies/
comprehensive approach could give nourishment roadmap/index_en.htm.
to the current impasse of ideas. Romero-lankao, P., 2012. Governing Carbon and Cli-
mate in the Cities: An Overview of Policy and Plan-
ning Challenges and Options. European Planning
REFERENCES Studies, 20(1), 7–26.
Schenone, C., Delponte, I., Pittaluga, I., 2015. The prep-
Alberti M.; Marzluff J., 2004. Ecological Resilience in aration of the Sustainable Energy Action Plan as a
urban ecosystems: Linking urban patterns to human city-level tool for sustainability: The case of Genoa.
and ecological functions. Urban Ecosystems, 7, Journal of Renewable and Sustainable Energy, 7 (3).
241–65. Steemers, K., 2003. Energy and the city: density, build-
Banister, D., & Anable, J., 2009. Transport Policies and ings and transport. Energy and buildings, 35(1),
Climate Change. In S. Davoudi, J. Crawford, & A. 3–14.
Menhood (Eds.), Planning for Climate Change. Strat- Van de Coevering, P., and Schwanenb, T., 2006. Re-
egies for Mitigation and Adaptation for Spatial Plan- evaluating the impact of urban form on travel pat-
ners, 55–69. London: Earthscan. terns in Europe and North-America, Transport Policy
Delponte, I., Pittaluga, I., and Schenone, 2013. C., 13 (3), 229–239.
Develop, implementation and monitoring of the Sus- Wilbanks J., Kates R.W., 1999. Global changes in local
tainable Energy Action Plan of the Municipality of places: How scale matters. Climate Change, 43,
Genoa, Proceedings of 31st UIT Heat Transfer Con- 601–28.
ference, 599–608. Wilson, E., 2006. Adapting to Climate Change at the
Derissen S., Quaas M.F., Baumgärtner S., 2011. The Local Level: The Spatial Planning Response. Local
relationship between resilience and sustainable devel- Environment, 11(6), 609–625.
opment of ecological-economic systems. Ecological Wilson, E., & Piper, J., 2010. Spatial Planning and Cli-
Economics 2011, 70, 1121–8. mate Change. Routledge, Ed., Oxon.

608

TIS2017_Book.indb 608 2/8/2017 8:07:50 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A model to evaluate heavy vehicle effect in motorway traffic flow

R. Mauro, S. Cattani & M. Guerrieri

DICAM, University of Trento, Trento, Italy

ABSTRACT: Highway operating conditions are also determined by interactions between passenger
cars and heavy vehicles in the flow. Therefore, a suitable model which describes and predicts the heavy
vehicle effect can be useful to achieve more realistic criteria to evaluate the Level of Service (LOS) for
infrastructures in the uninterrupted flow. After reviewing briefly the models available in literature to
take account of the effects on operating conditions due to heavy vehicles, the paper examines some traf-
fic data analysis of the Italian A22 motorway and later presents a closed-form traffic model for heavy
vehicles via the queuing theory. Some applications of this model are displayed for operational purposes
and a critical evaluation of the results is also made in comparison with some experimental evidence.

1 INTRODUCTION Units (PCUs). Therefore, PCEs have an impor-

tant role in highway design and functional analysis
The plano-altimetric geometry, the vehicular flows (Webster & Elefteriadou 1999; Elefteriadou 2014).
and the heterogeneity in the vehicular traffic mix Nowadays many different methods are available
composing the traffic streams are the most impor- to estimate the PCEs as function of one or more
tant factors affecting the Level of Service (LOS) of traffic flow parameters (Shalin & Kumar 2014):
motorways and highways.
• PCEs based on Flow Rates and Density;
Despite the proportion of heavy vehicles per-
• PCEs based on Headways;
centage (trucks, buses, recreational vehicles, etc.) is
• PCEs based on Speed;
generally smaller with respect to the total flow, its
• PCEs based on Queue Discharge Flow;
impact on highway performances is remarkable.
• PCEs based on Delays;
The effect of heavy vehicles on the MOE (meas-
• PCEs based on volume-to-capacity ratio;
ures of effectiveness) is mainly ascribed to two
• PCEs based on Vehicle-Hours;
factors: their larger dimensions and the lower per-
• PCEs based on Travel Time;
formance (acceleration, maximum speeds, etc.)
• PCEs based on HCM method.
compared to passenger cars (Sun, et al., 2008).
In addition, the performances of heavy vehicles Although the evaluation of LOS is generally
may vary under different traffic and geometric con- carried out by means of HCM 2010 methodology,
ditions, and are related to the following factors: many other alternative methodologies can be used
for this purpose (Wu, 2016).
• traffic regime: unsaturated or saturated
Furthermore, the HCM 2010 doesn’t allow eval-
conditions;
uating the LOS in the case of particular types of
• traffic flow rate for unsaturated conditions;
traffic control strategies, implemented to increase
• terrain type: level, rolling and mountainous.
safety and/or capacity, and cannot also be used for
Since the first versions the Highway Capacity highway performance analysis in real-time.
Manual (HCM) has provided the methodology to In the light of previous critical issues, this
evaluate the Passenger Car Equivalents (PCEs) for research introduces a novel closed-form traffic
the capacity and Level of Service (LOS) analysis. model, based on the P-K theory (Pollaczek 1930;
PCEs were first formally introduced by the Kleinlock 1975), for the evaluation of MOE in
HCM (1965) but the algorithms used for their real-time.
evaluation have been modified in the following ver- The proposed model does not need to estimate
sions of the manual. the PCEs.
The HCM 2000 defines PCE as “the number of First applications were done to evaluate the
passenger cars displaced by a single heavy vehicle effect of trucks’ Overtaking Prohibition (OP) in
of a particular type under specified roadway, traffic terms of LOS.
and control conditions”. By means of PCEs, traffic The OP is a very common strategy utilized to
flows can be expressed in terms of Passenger Car increase the passing lane capacity and motorway

609

TIS2017_Book.indb 609 2/8/2017 8:07:52 PM

 safety, although it can generate the formation of In which δ is the perception-reaction time
platoons of heavy vehicles. In addition, increases (δ = 1 s)
of rear-end collisions may occur in the right lane 3. in the event of Erlang headway distribution
(Mauro & Guerrieri 2016). (with a parameter K, Mauro & Branco 2012)
The case study of the Italian motorway A22 for vehicles travelling on the overtaking lane,
(Autostrada del Brennero) was examined. any service time s and vehicle headways, accord-
The sampling of traffic flows, platoons, speeds ing to P-K relationships, the mean b of the serv-
was performed in four observation sections. The ice time s, b = E[s], and the variance σ2(s) can be
proposed traffic model has been used to evaluate calculated as follows:
delays, queues and level of service in real-time.

e kQT − ∑
k
(kQT )i
i!
2 NOVEL CLOSED-FORM TRAFFIC b E[s] T + i=0
(2)
MODEL FOR EVALUATING
k −1
Q∑
(kQT )i
HIGHWAY MOE i=0 i!

The queuing theory can be used as the basis for ⎡ k +1

(kQT )i ⎤
considering the influence of heavy vehicles on the (k 1) ⎢e kQT − ∑ ⎥
⎢⎣ i=0 i ! ⎥⎦
+ ( E [ s ] − T ) (3)
2
level of service of highway segments. σ 2( ) =
The novel closed-form traffic model presented
kQ 2 ∑
k −1
(kQT ) i

in this section is based on the following general i!

i=0
assumptions under which the model works:
• heavy vehicles travel only on the right lane, with 4. the time spent in queuing E[w]:
speed vvl (as in the case of trucks’ overtaking
prohibition); Qvl (b + ( s ))
• the passenger cars travel with a speed v, except E [w ] = b + (4)
2 ( − Qvl b )
for the case in which they must queue behind
heavy vehicles before starting overtaking maneu-
vers (while in waiting for overtaking maneuvers 5. the average speed of passenger cars (Vpc) is
the speeds of light and heavy vehicles are the ⎛ ⎛ v + vvl ⎞
same v = vvl); Vpc β E w ⋅ vvl + β ⋅ ta ⋅ ⎜ +
• in the right lane the passenger cars, before reach- ⎝ ⎝ 2 ⎟⎠
⎛ v + vvl ⎞ ⎞
ing heavy vehicles, decelerate with a constant value
of deceleration d. Instead, overtaking maneuvers
(tr ta − td ) v + β ⋅ td ⋅ ⎜
⎝ 2 ⎟⎠ ⎟⎠
/( ⋅ + )
take place with a constant acceleration a; (5)
• denoting with γ the percentage of heavy vehicles
γ = Qvl/Qtot (Qvl = traffic flow of heavy vehicles and where:
Qtot = total flow) and with β the percentage of
passenger cars travelling along the right lane vvl
(Qmn = β( Qtot-Qvl)). The passenger cars travelling tr = (6)
on the overtaking lane are: Q = Qtot-Qvl- Qmn; (v vvl ) ⋅ Qvl
• for the passenger cars that have to make the
overtaking manouvres, the right lane is regarded is the time taken by a passenger car with a speed v
as a hypothetical acceleration lane in which the to travel in the space comprised between two suc-
service time w is equal to the interval time spent cessive heavy vehicles, whose speed is vvl;
in queuing before the overtaking manoeuvres;
• the proposed model defines LOS for motorway v vvl
and highway as a function of the average speed. ta = (7)
a
In the light of the above assumptions the average
speed can be estimated with the following steps: is the time taken by a passenger car to accelerate
from vvl to v (equal to speed of passenger cars in
1. estimation of γ, β, Qtot, Qmn, Qvl (by means of the overtaking lane)
sampling traffic flows);
2. calculation of critical gaps by means of the fol- v vvl
td = (8)
lowing relationships: d
v vvl is the time required by a passenger car to decelerate
T= + 2δ (1)
2a from v to vvl.

610

TIS2017_Book.indb 610 2/8/2017 8:07:52 PM

 Figures 1÷3 summarize the correlation between 3 THE A22 MOTORWAY CASE STUDY
average speed Vpc and total flow Q (Vpc = Vpc (Q))
for different traffic conditions and for γ = 10% (see The A22 motorway, “Autostrada del Brennero”,
Fig. 1), γ = 20% (see Fig. 2), γ = 30% (see Fig. 3). links the Po Valley and the A1 Freeway with Aus-
tria and Germany. The A22 is one of the principal
axes of the Italian highway network and belongs to
the TEN-T Network (Trans-European Road Net-
work, corridor Helsinki—La Valletta).
The A22 overall length is about 313 km.
Fig. 4 shows the layout and overtaking prohibi-
tions along the A22. Instead, Fig. 5 shows the typi-
cal semi-cross section.
The overtaking prohibition for heavy vehicles is
managed as follows:
• vehicles with weight > 7.5 t, from hour 0:00 to
24:00, from 0 to 85 km;

Figure 1. Vpc = Vpc (Q) relationships for different

β (a = 0.6 m/s2, d = 2 m/s2, v = 130 km/h, δ = 1 s, vvl
= 90 km/h, γ = 10%).

Figure 2. Vpc = Vpc (Q) relationships for different β

(a = 0.6 m/s2, d = 2 m/s2, v = 130 km/h, δ = 1 s, vvl = 90 km/h,
γ = 20%).

Figure 3. Vpc = Vpc (Q) relationships for different β.

(a = 0.6 m/s2, d = 2 m/s2, v = 130 km/h, δ = 1 s, vvl = 90, Figure 4. A22 motorway layout and overtaking
γ = 30%). prohibitions.

611

TIS2017_Book.indb 611 2/8/2017 8:07:53 PM

 Table 1. Observation sections.

Vertical
Observation Horizontal alignment;
section Location alignment slope (%)

Kofler 063+500 Tangent Curve: R = 10,000 m;

slope = 0.41%
S. Michele 123+960 Tangent Tangent, slope =
0.03%
Portale Affi 205+500 Curve R = Curve: R = 10,000 m;
1000 m slope = 0.23%
Mantova 271+900 Tangent Curve: R = 15,000 m;
Figure 5. A22 Typical semi—cross Section. Sud slope = 0.00%

• vehicles with weight > 12 t, from hour 6:00 to

22:00, from 85 to 313 km. Table 2. PCEs relationships.
The Annual Average Daily Traffic (AADT) is Passing
between 41,907 vehicles per day (section “Brenne- p Right lane lane Carriageway
ro—Vipiteno”) and 62,646 vehicles per day (sec-
tion “Ala Avio—Affi”). >0 PCE = 1.166⋅(γ%)−0.126 2 1.126⋅ (γ%)−0.205
>0 PCE = 1.35 2 1.220⋅ (γ%)−0.132
3.1 Estimation of capacity and distributions of ≈0 PCE = 1.186⋅ (γ%)−0.112 2 1.161⋅ (γ%)−0.172
heavy vehicles between lanes
The capacity analysis was done for two observa-
tion periods: 5÷11 May 2014 and 8÷14 Decem- 1
− ⋅( k )2
ber 2014. The observation sections are given in v vf ⋅ e 2 k jjam (10)
Table 1.
The macroscopic flow parameters (i.e. flow “q”, where vf is the free-flow speed and kjam the jam
speed “v” and density “k”) have been calculated for density. The typical fundamental diagrams (with
intervals of 5 minutes and 15 minutes. LOS limits, see HCM 2010 values) for the right
Traffic flows were estimated in terms of passen- lane, passing lane and carriageway are depicted
ger car unit (PCU), by means of the following PCE in Figure 6 (Guerrieri & Mauro 2016). Instead,
relationship: Table 3 illustrates the flow relationship parameters
for the years 2003 (Mauro et al. 2013) and 2014.
(1 − γ ) ⋅ ((d
d AL + d LA − d ) + γ ⋅ d LL The Level of service has been calculated for
PCE = (9)
d AA all the sections (considering an hourly volume
VHP = 0.1⋅AADT, see Table 5). Figures 7–8 show
In which: the frequency distribution histograms of the per-
γ percentage of heavy vehicles
• γ: centage of heavy vehicles in the traffic stream on
• d AA : average headways of couples of passenger the right lane and on the overpassing lane at week-
cars; days (section San Michele). It is immediately veri-
• d AL : average headways of couples of vehicles in fiable that overtaking prohibitions are generally
which the leader vehicle is a passenger car and respected by heavy vehicles over 7.5 t, though this
the follower vehicle is a heavy vehicle; leads to the formation of many platoons.
• d LL : average headways of couples of heavy Figure 9 shows the flow distribution between
vehicles; lanes in the northbound roadway for the section
• d LA : average headways of couples of vehicles in San Michele.
which the leader vehicle is a heavy vehicle and The analysis of the speed and flow processes
the follower vehicle is a passenger car. shows that on the A22 motorway the observed
average speed of passenger cars (passing lane)
PCEs relationships have been tabulated as func- is close to 130 km/h (see Fig. 10), instead the
tion of the slope p (see Table 1) in Table 2. observed average speed of heavy vehicles (vvl) is
In all, N5’ = 24,192 couples (v; k), (q; k), (v; q). close to 80 km/h (right lane).
have been carried out. In addition, sampling of platoons has been car-
For each section the flow diagrams are obtained ried out at observation sections shown in Table 2
by means of the Drake model (May 1990; Wang et al. on 7th May 2014 (Wednesday). The flow param-
2009) in which the relationship between speed (v) eters of platoons, with 2÷20 heavy vehicles for each
and density (k) is given by the following equation: platoon, have been investigated by collecting data

612

TIS2017_Book.indb 612 2/8/2017 8:07:55 PM

 Figure 7. Frequency distribution of the heavy vehicle
percentage. Section San Michele, northbound roadway,
right lane, weekdays.

Figure 8. Frequency distribution of the heavy vehicle

percentage. Section San Michele, northbound roadway,
passing lane, weekdays.

Figure 6. Speed—Flow diagrams for the right lane

(red), passing lane (b) and carriageway (c), A22, north-
bound roadway.

Table 3. Traffic flow parameters (A22).

Year 2014

Lane/ vf kjam [pcu/ C vjam

carriageway [km/h] lane/km] [pcu/h] [km/h]

Right lane 106 24 1552 65 Figure 9. Relationship between lane flow rate and total
Passing lane 128 25 1916 77 flow rate Qt (San Michele, northbound roadway).
Carriageway 115 47 3254 70
Year 2003

Lane/ vf kjam [pcu/ C vjam

carriageway [km/h] lane/km] [pcu/h] [km/h]

Right lane 107 24 1534 65

Passing lane 130 25 1983 79
Carriageway 118 49 3490 71

regarding vehicle speeds, headways and mass. The

following parameters have been analyzed: Figure 10. Observed average speed of passenger cars
and flow processes. Section S. Michele, northbound road-
• frequencies of the number of platoon sizes of way (weekday).
“i” heavy vehicles (i = 1,2,…20);
• minimum headway (τmin), mean headway (τmean), • minimum speed (Vmin), mean speed (Vmean),
and maximum headway (τmax) between vehicles and maximum speed (Vmax) of vehicles of each
of each platoon; platoon;

613

TIS2017_Book.indb 613 2/8/2017 8:07:56 PM

 Table 4. Number of platoons whose vehicles travel with
headways comprised within classes of headway. Section
San Michele, northbound roadway.

τi [s] τj [s] τ* = (τi + τj )/2 [s] N Vm [km/h]

0 1 0.5 15 87.28
1 2 1.5 90 85.07
2 3 2.5 225 84.6
3 4 3.5 179 85.51
4 5 4.5 109 86.25
Figure 11. Frequency of platoons as function of
5 6 5.5 71 88.06
the number of vehicles for each platoon. Section San
6 7 6.5 33 90.45
Michele, northbound roadway.
7 8 7.5 27 92.45
8 9 8.5 18 89.86
9 10 9.5 13 94.04
10 11 10.5 6 90.31
11 12 11.5 6 87.36
12 13 12.5 4 94.25
13 14 13.5 5 89.47
14 15 14.5 1 81.5
15 16 15.5 3 88.22
16 17 16.5 2 90.92
17 18 17.5 0 –
Figure 12. Relationship headway—platoon size (red
line: τmin; blue line: τmean). Section San Michele, north- 18 19 18.5 0 –
bound roadway. 19 20 19.5 0 –

s2; d = 2 m/s2; δ = 1 s, v = 130 km/h, vvl = 90 km/h.

The mean value of the coefficients β is 0.58, instead
γ = 11%–18% depending on the section.
The first results of the traffic model are given
in Fig. 14; further phases of the research require a
more accurate calibration of the model in order to
reduce the deviations between observed and esti-
mated values (Fig. 15).
Figure 13. Relationship maximum headway–platoon
size. Section San Michele, northbound roadway. However, in the hypothesis of uninterrupted
flows, the proposed model allows the real-time esti-
• the number of platoons whose vehicles travel with mation of the motorway density (k) and Level of
headways comprised within predetermined values Service (LOS), as shown in Fig. 16 for the case of
(from 0 < τ ≤ 1 seconds, to 1 < τ ≤ 20 seconds). the right lane. Finally, the speed levels carried out
by means of the model presented above can be used
Fig. 11 shows the relative frequencies of the as input data for the calculation of the reliability φ.
number of platoons with “i”. In all the sections of Αs well known, the flow instability is a random
the A22 the size of platoons rarely exceeds ten vehi- event whose occurrence depends on the speed level
cles (i >10). process. The probability that no instability will
Fig. 12 and Fig. 13 show respectively the trends occur in the interval time T defines the reliability
of minimum headway (τmin), mean headway (τmean) φ of the vehicle flow (for the period T). The φ value
and maximum headway (τmax). for a lane traffic flow can be estimated with the
The mean headway is characterized by low fluc- expression (Ferrari 1988; Mauro et al. 2013):
tuations; instead, the minimum headway increases
monotonously as function of the platoon size. 8,82
⎛ Q ⎞
The number of platoons whose vehicles travel φ = 1 − 19, 80 ⋅ ⎜ ⋅ T 1,983 ⋅ M 2 (11)
with headways comprised within predetermined ⎝ 10, 000 ⎟⎠
classes (τI ÷ τj) are given in Table 4.
Through the traffic model explained in Section 2 where Q = flow rate constrained along the off-
the mean of passenger cars speed has been estimated, side lane (veh/h); T = period of time (minutes);
in real-time, using the traffic sampling of 7 May M = absolute value of “b”, angular coefficient of
2014 (Wednesday). In order to implement the algo- relationship σv2 = a+b⋅ln(k) (m2⋅km⋅s–2). The coef-
rithm, the following input data were used: a = 0.6 m/ ficient σv2 is the standard deviation of speed level.

614

TIS2017_Book.indb 614 2/8/2017 8:07:58 PM

 Table 5. Level of service estimation for each segment of A22.

AADT (Carriageway) VHP

Initial Final Section Name [veh/day] [veh/h] LOS

0 + 000 15 + 870 Brennero—Vipiteno 23,049 2305 C

15 + 870 38 + 030 Vipiteno—Bressanone 24,287 2429 C
38 + 030 47 + 600 Bressanone—Bress.ne Z.I. 26,613 2661 D
47 + 600 53 + 070 Bressanone Z.I.—Chiusa 25,793 2579 C
53 + 070 77 + 470 Chiusa—Bolzano nord 25,181 2518 C
77 + 470 85 + 330 Bolzano nord—Bolzano sud 26,828 2683 D
85 + 330 101 + 800 Bolzano sud—Egna Ora 31,764 3176 E
101 + 800 121 + 450 Egna Ora—S. Michele 32,346 3235 E
121 + 450 131 + 440 S. Michele—Trento nord 31,831 3183 E
131 + 440 136 + 460 Trento nord—Trento centro 28,855 2886 D
136 + 460 142 + 000 Trento centro—Trento sud 29,081 2908 D
142 + 000 157 + 850 Trento sud—Rovereto nord 33,586 3359 F
157 + 850 166 + 740 Rovereto nord—Rovereto 33,201 3320 F
166 + 740 179 + 125 Rovereto sud—Ala Avio 34,141 3414 F
179 + 125 206 + 670 Ala Avio—Affi 31,323 3132 E
206 + 670 225 + 370 Affi—Verona nord 26,519 2652 C
225 + 370 228 + 000 Verona nord—int. aut. A4 33,237 3324 F
228 + 000 243 + 670 int. aut. A4—Nogar. Rocca 34,471 3447 F
243 + 670 256 + 180 Nogar. Rocca—Mantova nord 33,580 3358 E
256 + 180 265 + 000 Mantova nord—Mantova sud 33,658 3366 F
265 + 000 276 + 710 Mantova sud—Pegognaga 33,918 3392 F
276 + 710 285 + 630 Pegognaga—Reggiolo Rolo 30,050 3005 D
285 + 630 302 + 175 Reggiolo Rolo—Carpi 30,492 3049 D
302 + 175 312 + 150 Carpi—Campogalliano 32,817 3282 F
312 + 150 313 + 085 Campogalliano—Autosole 33,388 3339 F

Figure 14. Average speeds (Vpc) of the passenger cars (blue line) in the right lane as function of the flow rates (Q)
on the overpassing lane (red line), calculated with the traffic model explained in the section 2. Section San Michele
northbound roadway (7/5/2014).

Figure 15. Estimated and observed values of the average speeds of passenger cars on the right lane.

615

TIS2017_Book.indb 615 2/8/2017 8:07:59 PM

 Figure 16. Density (k) and LOS estimated with the proposed traffic model, Section San Michele northbound road-
way (7/5/2014) (right lane).

4 CONCLUSIONS REFERENCES

The presence of heavy vehicles is particularly Elefteriadou, L. 2014. An Introduction to Traffic Flow.
Springer.
interesting in traffic study, and more generally Ferrari, P. 1988. The reliability of motorway transport sys-
in Highway Engineering, with reference to many tem. Transp. Res. Part B, 22(4), 291–310.
technical applications, especially to topics like Guerrieri, M. & Mauro, R. 2016. Capacity and safety
capacity, level of service, road safety and air pol- analysis of Hard-Shoulder Running (HSR). A motor-
lution emissions. way case study. Transportation Research Part A: Pol-
The research proposes a closed-form traf- icy and Practice, Vol. 92, pp. 162–183.
fic model based on the P-K queuing theory that HCM (1965). TRB, National Research Council, Wash-
allows to estimate in real-time (under predeter- ington, DC.
mined assumptions) many traffic flow param- HCM (2000). TRB, National Research Council, Wash-
ington, DC.
eters, like the mean speed of the passenger cars HCM (2010). TRB, National Research Council, Wash-
travelling in the right lane (Vpc), the lane density ington, DC.
(k) and the Level of Service (LOS). The novel Kleinlock, L. 1975. Queuing Systems, I. New York: John
traffic model does not require the use of Passen- Wiley and Sons.
ger Car Equivalents (PCEs) with regard to LOS Mauro, R. & Guerrieri, M. 2016. Safety and capacity
analysis. analysis of A22 motorway (in Italian). Internal Report,
The model has been employed to investigate the DICAM, Unitn.
functionality of the Italian motorway A22 (Auto- Mauro, R., Branco, F. 2012. Two vehicular headways time
strada del Brennero). Along this motorway, heavy dichotomic models. Modern Applied Science, 6 (12), pp.
1–12.
vehicle overtaking prohibition is in force. The first Mauro, R., Giuffrè, O., Granà, A. 2013. Speed stochastic
part of the research was devoted to the calculation processes and freeway reliability estimation: Evidence
of the macroscopic flow parameters. The traffic from the A22 freeway, Italy. Journal of Transportation
survey has shown that overtaking prohibitions are Engineering, 139 (12), pp. 1244–1256.
generally respected by heavy vehicles; this implies May, A.D. 1990. Traffic Flow Fundamentals, Prentice-Hall.
that many platoons travel on the motorway right Pollaczek, F. 1930. Über eine Aufgabe der Wahrscheinli-
lanes. chkeitstheorie. Mathematische Zeitschrift.
In order to estimate the average passenger cars, Shalini, K., Kumar, B. 2014. Estimation of the Passen-
this research has employed the following values ger Car Equivalent: A Review. International Journal
of Emerging Technology and Advanced Engineering
of acceleration (a), deceleration (d), perception- Volume 4, Issue 6, pp. 97–102
reaction time, (δ), speeds (v and vvl): a = 0.6 m/s2, Sun, D., Lv, J., Paul, L. Calibrating Passenger Car Equiv-
d = 2 m/s2, δ = 1 s, v = 130 km/h, vvl = 90 km/h. alent (PCE) for Highway Work Zones using Speed
In the hypothesis of uninterrupted flows, the and Percentage of Trucks. 2008. TRB 2008 Annual
proposed traffic model allows to estimate the Meeting CD-ROM.
motorway density and level of service in real time Wang, H., Li, J., Chen, Q. Y,. Ni, D. 2009. Speed-Density
(the right lane was examined). In addition, the Relationship: From Deterministic to Stochastic. TRB
results can be used as input data for the calculation 88th Annual Meeting at Washington D. C.
of the reliability of the vehicle flow φ, in real-time. Webster, N; Elefteriadou, L.. 1999. A Simulation Study
of Truck Passenger Car Equivalents (PCE) on Basic
Although the new traffic model requires a more Freeway Sections. Transportation Research. Part B:
accurate calibration, the first results show that the Methodological 33(5), 323–336.
procedure is characterised by a reliable and rapid Wu, N. 2016. New Features in the 2015 German Highway
analytical approach in passenger cars speed and in Capacity Manual. Transportation Research Procedia
LOS evaluation. 00 (2017) 000–000.

616

TIS2017_Book.indb 616 2/8/2017 8:08:00 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Quality and energy evaluation of rail-road terminals by microsimulation

A. Carboni & F. Deflorio

Department of DIATI, Politecnico di Torino, Torino, Italy

ABSTRACT: The aim of this paper is to present a method to evaluate the quality and energy perform-
ance of inland freight terminals, using a quantitative approach based on traffic microsimulation models.
The transport of goods should meet requirements of environmental sustainability and combined trans-
port can be an eco-friendly option for medium/long distance connections, in which the railway mode
can provide its efficiency if properly accessible. The analysis of the terminal aims at quantifying Key
Performance Indicators (KPIs) by traffic microsimulation. The relevant features of the typical phases of
the internal process are represented and the traffic flow data of arrivals are disaggregated by specific serv-
ice needs. The model is used to compare the chosen indicators in different scenarios, varying the arrival
rates for trucks and reducing the duration of check-out and cranes operations. Results of quality and
energy are also reported disaggregated for the different types of service lines.

1 INTRODUCTION multimodality has to become economically attrac-

tive for shippers. Thirty per cent of road freight
The intermodal transport is defined as a tranship- over 300 km should shift to other modes such as
ment of goods through the same transport unit rail or waterborne transport by 2030, and more
using two or more transport means and without than 50% by 2050, facilitated by efficient and green
the manipulation of the freight themselves (UNI/ freight corridors.
CE). To be sustainable from the environmental and Figure 1 shows that only in Estonia, Latvia and
economic points of view the intermodal transport Switzerland the road transport mode has a con-
must be efficient for all the actors involved in the tained share in refer to total percentage of freight
process. The crucial point is to take advantage of transport, whereas in the other countries it still
the operational benefits—as cost, capacity and largely, probably due to different geographical con-
flexibility—of transport modes, then merging texts or policies adopted.
them into a single transport chain increasing the The transport by railway, sea or inland water-
attractiveness of rail freight transport. ways is preferred on long distances for economic
The Directive 92/106/CEE defines the “com- scale reasons to reduce the impact of road trans-
bined transport” as “the transport of goods port, whereas the initial and final haulage are
between Member States where the lorry, trailer,
semi-trailer, with or without tractor unit, swap
body or container of 20 feet or more uses the road
on the initial or final leg of the journey and, on the
other leg, rail or inland waterway or maritime serv-
ices where this section exceeds 100 km as the crow
flies and make the initial or final road transport leg
of the journey”.
The UE, in the White Paper, has reiterated the
need of reduce drastically the greenhouse gas emis-
sions worldwide with the goal of maintain the glo-
bal warming under 2°C. In total, by 2050, Europe
must reduce emissions by 80–95% compared to
1990 levels (European Commission 2011). The
White Paper further states that freight transport
over short and medium distances (roughly below
300 km) will continue to be carried, in large meas- Figure 1. Inland freight transport modal share (from
ure, with trucks. In longer distances, options for EU transport in figures 2014, elaboration by Castaldo
road decarbonisation are more limited, and freight (2014)).

617

TIS2017_Book.indb 617 2/8/2017 8:08:02 PM

 managed in road transport mode, because it pro- even in special areas in case of technical stops or
vides more flexibility and accessibility (Carreira in parking lots for semitrailers.
et al. 2012). • Check-out operations for the trucks leaving the
The transfer between the two modes is oper- terminal.
ated in intermodal terminals that deal both of the
Although there are other procedures inside the
transfer operations and the network change. More
terminal, these specifically involve operations from
specifically, the International Union of Railways
the railway side and therefore not directly involved
defines the rail-road terminals as interchange hubs
in this study. The main elements included in rail-
between rail and road traffic. They are fitted with
road terminal are: rail siding for wagon both to
all the equipment required to handle and trans-
transhipment that manoeuvres and other opera-
ship loading units from one transport mode to the
tions, buffer lanes for ITUs, loading and driving
next in a rapid and efficient manner: gantries and
lanes for the trucks, gates and internal road net-
mobile cranes, modern computer systems integrat-
work (Ballis & Golias 2004). In general, the ter-
ing tracks, storage areas, transshipment areas and
minal operations in real-life environment are very
connections to roads and motorways.
complex and their detailed description is out of the
The intermodal transport chain must guaran-
scope of this paper.
tee a good level of speed, efficiency and safe tran-
The focus of this research is on the road part
shipment of ITUs (Intermodal Transport Units,
of the terminal and its functional interactions with
such as containers, semitrailers and swap bodies)
rail, crane sub-systems and gate operations. The
between the transport modes, also to consider the
traffic microsimulation model later described in
specific features of the territory. In fact, in UE the
the paper is then built to represent these features
concentration of intermodal terminals is not uni-
of the terminal.
form, with a relevant number of infrastructures
in the north regions and around mountain (Alps,
Pyrenees). 2.2 Modal transfer energy
The energy analysis in intermodal transport for
2 STATE OF THE ART freight regards principally the following aspects
(Zumerchik et al. 2011):
2.1 Principal equipment and processes in the high
• Line Haul Energy is the fuel/energy needed
capacity terminals
to transport goods from origin to destination
In Figure 2 there is a synthetic scheme of a typi- through different modes.
cal rail-road combined transport terminal with the • Modal Transfer Energy is the fuel/energy used in
main processes, in detail: the terminal for modal transfer by cranes, dray-
age trucks, yard tractors, service vehicles, as well
• Check-in operations for the trucks incoming at
as energy use for switching.
the terminal, including the inspection proce-
• Storage Energy refers to deposits and storing.
dures and the documents management for goods
and drivers. These two operations can also be The interest of this paper is on the second point,
performed in two distinct phases and places. therefore the energy spent to transhipment opera-
• Loading or unloading operations under cranes, tions inside an inland terminal. In particular, the
from truck to railway wagon or vice versa, or focus is on the energy consumption by trucks
within the terminal on the road side, which is in
charge of the various transport operators using
the terminal. This estimation has been performed
with a traffic microsimulation approach, where
each vehicle can be monitored along the simulation
period with a sub-second step time resolution. The
other modal transfer energy components are not
considered for the following reasons. The energy
use due to the crane operations for the ITUs han-
dling depends on the type of crane, ITU weight
and dimension, and these data are not available,
although they can be easily measured by the termi-
nal operators. For containers operations, an aver-
age value of 4.4 kWh/TEU is considered in IFEU
et al. (2014). The energy profile of the internal road
Figure 2. Key elements in a typical inland terminal. tractors cannot be independent by the operations

618

TIS2017_Book.indb 618 2/8/2017 8:08:03 PM

 scheduled in the terminal. Therefore many vehicle • Reliability, measure the congruency of effective
random interactions can be detected, during the transport service and the scheduled time.
terminal operations and at the queues in relevant • Flexibility, i.e. as the system is able to respond
phases of the process. Also, this contribution can to contingencies without changing the level of
be measured by the terminal operators, since usu- service.
ally they manage the fleets. • Qualification, the terminal ability of responding
to more complex logistical requirements.
• Safety and Security, related to the loss or dam-
2.3 KPIs age ITUs and their documents.
Ballis & Golias (2004) underline that the impor- • Accessibility, definable through the actual work-
tance of the quality of service offered by the ing hours of the terminal.
combined transport terminals is recognized as an
important factor associated to the attractiveness
of the combined transport.
3 METHODOLOGY
Different actors are involved in the terminal proc-
ess and this creates several points of view to evaluate
A typical terminal for the transhipment of ITUs
the efficiency of terminal services and its perform-
between railway and road mode is simulated tak-
ance. In this study two main stakeholders are taken
ing into account the specific road infrastructure
into account: users who want to ship goods and
and vehicles, as well as their interaction at relevant
terminal operators who have to manage the proc-
points of the internal process. Also, the energy
ess. The first one can be considered those customers
consumption estimation for vehicles using the ter-
who use the terminal, then in practice represented
minal is considered. All of these requirements can
by truckers. Their main benefits are the energy sav-
be modelled by simulation. Although the relevant
ing inside the terminal and total time spent between
events of the whole process could be represented by
the check-in and check-out operations (turnaround
a discrete-event simulation model, such as in Ricci
time). Both of these aspects can be considered part
et al. (2016), if the focus is on the traffic interactions
of the quality assurance of the terminal, which is in
along the connecting roads, vehicle queues and
charge of the second actor involved.
their energy consumptions, microsimulation tools
The Key Performance Indicators must be clear,
can provide a more effective modelling. They are
coherent, compatible, controllable, complete,
based on a time-sliced approach and widely used
pertinent and feasible. The intermodal terminals
in traffic engineering studies. The terminal layout
have a fundamental role, therefore an impact, on
was then modelled with Aimsun® and applied to a
entire logistic chain of combined transport both
large-sized Italian terminal (Hupac Busto Arsizio)
on economic, quality and efficiency point of view.
on the basis of the Open Street Map information.
Currently the European intermodal terminals are
not subject to standard quality evaluations or
some Level of Service (LOS) as in other transport 3.1 Model and hypotheses
sectors.
The typical activities in a rail-road combined trans-
At global level the performance measure of rail-
port terminal, by a truck driver point of view are
road terminal essentially takes place by means of
the following:
the turnaround time, which is the full time between
the entrance of a truck in the terminal for the deliv- • physical check-in (code, damages, labels,
ery or pickup of ITU and its exit from the same, irregularity)
but also the degree of utilization and productiv- • documents exchange and verification
ity. The last one is function of the average time • loading and unloading operation with cranes or
of ITU permanence in intermodal terminal and it semi-trailers
has obviously relevant implications on transport • check-out.
efficiency: the lower the residence time, the greater
The first two sub-processes may take place
productivity (Zumerchik et al. 2011).
simultaneously or in two different areas, manually
Ballis (2003) has correlated standard values of
or automatically, for example with particular Opti-
selected quantifiable indicators with six possible
cal Character Recognition Cameras, or Barcodes,
Level of Service, letter A represent the excellence,
RFID or other security technologies. In the sce-
the ideal situation, whereas letter F is for an unac-
nario these operations are performed in different
ceptable state. The set of selected indicators are the
places manually.
following:
The trucks arrival can be divided into three cat-
• Waiting time for users, it is the turnaround time egories, based on their mission: delivery, collection
included the initial queue. or both.

619

TIS2017_Book.indb 619 2/8/2017 8:08:03 PM

 The terminal layout is assumed with check-in fication doubles to take into account the two
and check-out connected to the same external road activities requiring separate procedures.
infrastructure. Internal roads are composed by sec-
The reference scenario (S0) is set assuming the
tion, nodes and roundabouts with specific, lanes,
assignment of the 8 lines to the 6 cranes as reported
traffic rules. In Figure 3 the basic terminal layout
in Table 1 and an average service time duration for
is shown with the truck activities locations used for
any crane equal to 3 minutes, so the unitary service
the simulation; in particular the screenshot repre-
rate μ is equal to 20 vehicles per hour. To consider
sent the path of line 1 (described in more detail
the random phenomena linked to this operation,
below) in which the codes indicate:
due to the position, the type of ITU or weather
• Ck_in_phy1 = physical check-in conditions. A normal distribution is assumed with
• Ck_doc5 = documents exchange and verification 60 s for the standard deviation. The total average
• Barrier = automatic barrier for the entrance to service rate is then 120 vehicles per hour, if we con-
the terminal sider all the six cranes operating.
• Crane 1 = loading and unloading operation To consider the system in under-saturation
• Ck_out = check-out. conditions, for this reference scenario, the arrival
rate has been set subsequently. The frequency of
The method proposed in this paper consists in
the 8 lines is then calculated assuming a number
the traffic simulation of the trucks flow as public
of “equivalent lines” equal to 12, because 4 lines
transport lines with specific stops along the route
are for a single service (delivery or collection) and
inside the terminal. Any stop model one of the
4 lines with both of them. The single arrival rate λ
activities listed above, characterized by different
was considered roughly equivalent to 10 vehicles
times due to the usually expected operations that
per hour and consequently the mean time interval
can take place.
between departures was set to 6 minutes.
In particular, eight service lines were chosen to
Each line has its timetable, such as reported for
create possible situations based on the characteris-
the line 8 in Figure 4, and the frequency has been
tics of the mission:
assumed identical for all the lines. To consider
• LINE 1–2, the vehicles should carry out the random phenomena, also the headway has been
delivery operation. They would have to do the modeled with a normal distribution with 30 sec-
physical check-in at the entrance and then will onds of standard deviation. To avoid the con-
be served by a crane; centration of vehicles in short time intervals the
• LINE 3–4, these vehicles should do only the col- lines departures are scheduled every 45 seconds,
lection, this means that they would not have to by dividing equally the mean headway among the
the physical check-in, so can go directly to the 8 lines.
documents point. One crane tranships the unit The stops are fundamental elements for the
and then the activity of check-out would require building of the model in which the vehicles per-
more time in order to verify the load; form some typical activities in an inland ter-
• LINE 5–8, dedicated both for delivery and collec- minal. As aforementioned, each line includes
tion which are supposed at two different cranes. specific stops during the route that may also
The duration for documents exchange and veri- vary their duration depending on the character-

Table 1. Operation order* for cranes and relation with

lines.

Crane
Line 1 2 3 4 5 6
1 I
2 I
3 I
4 I
5 II I
6 I II
7 II I
8 I II

Figure 3. Positions of truck activities and service line 1. * I = first; II = second.

620

TIS2017_Book.indb 620 2/8/2017 8:08:03 PM

 E n (t ) = max[[ E f f2vn (t f3vn (t )2 +
(1)
f4 an (tt f5an (t )2 + f6vn (t )an (t )]

where vn(t) and an(t) are the instantaneous speed and

acceleration of vehicle n at time t. E0 is a lower limit
of emission (g/s) specified for each vehicle and pol-
lutant type, and f1 to f6 are emission constants spe-
cific for each vehicle and pollutant type determined
by the regression analysis. To derive the emission
functions, in Panis et al. model, for heavy vehicle
the measurements were carried out from two type
of vehicles (Iveco Eurocargo and Volvo FH12–420)
Figure 4. Example of timetable for the line 8.
respectively in number of 1638 and 4514.

3.3 Performance indicators

istics of the service line itself. The time required
for the documents exchanging and verification is The KPIs chosen for this paper can be classified in
set at 100 s (standard deviation = 30 s) for each two following groups: quality and energy.
operation, whereas 200 s is for lines with double As for the quality, the delay and queue length
service (delivery and collection). The six cranes are useful from the terminal management point of
were represented as fixed stops in specific zones view in order to identify possible bottlenecks of
of the road sections. This is an assumption which the system. Furthermore, the turnaround time is
relates to the way the cranes are used for opera- important for several actors. The truck driver and
tions and it would need to be confirmed case by the client are interested because is the total time
case. In fact, even with electric gantry cranes on spent for terminal operations. The terminal opera-
fixed tracks, they can cover larger areas, maybe tor can use this value to be more competitive. The
overrunning the action space of the nearby crane, policy maker can chose this to compare similar
if required. terminals for quality evaluation. Finally, less time
The vehicles type that use the lines previously is required in inland terminals, means less time in
described are trucks. In particular their character- the logistic chain, increasing the competitiveness
istics are defined with some dimension variability of the rail-road combined transport.
of to distinguish the possible different truck types As for the energy estimation, the fuel consump-
(e.g. trailer or articulated trucks) and the related tion refers to the road vehicles travelling inside
space used on the road. the terminal from the first road section before
the check-in point and the final section after the
check-out point, during the simulation period. The
3.2 Emission and energy consumption model fuel consumption, in our case is estimated from the
The emission model used and also implemented emission data, provided by the Panis et al. model,
in Aimsun® was propose by Panis et al. (2006). as explained in §3.2.
It is chosen in this study for its rich data set used
during the calibration process. The traffic emis-
sions are estimated on an instantaneous emission
model integrated in a microscopic traffic simula-
tion model. These vehicle emissions are calculated
in relation to the type, the instantaneous speed and
acceleration. Panis et al. (2006) have focused on
some pollutants, for their potential health impacts
and external costs: Nitrogen Oxides (NOx), Volatile
Organic Compounds (VOC), and Particulate Mat-
ter (PM). Carbon dioxide (CO2) is also estimated
for its effect on global climate change and can be
useful if converted in fuel: 1 kg of CO2 corresponds
to 0.4 l of fuel, with negligible differences between
diesel and petrol.
The emission functions for each vehicle are
derived using non-linear multiple regression tech- Figure 5. Screenshot of vehicle simulation along roads
niques (Eq. 1). inside the intermodal terminal.

621

TIS2017_Book.indb 621 2/8/2017 8:08:04 PM

 4 APPLICATION AND RESULTS expected, the higher values are for the double serv-
ice trucks. Also, a different value is observed for
4.1 Baseline scenarios and experimental collection and delivery services, since the first one
conditions requires less time at the physical check-in point.
The light increasing trend over time for all the val-
The scenario S0 was used as baseline to build the
ues shows that the system is just over the equilibrium.
model with characteristics described in the sec-
This is also confirmed by the values in Table 3.
tion 3.1. In this scenario an equilibrium condition
Indeed, observing the values related to the sce-
has been set for services and arrivals.
nario 1, obtained increasing the time intervals for
This baseline scenario has been tested with two
arrivals from 6 to 7 minutes, they are quite con-
complementary scenarios (S1 and S2), simulating
stant, as should be in stationary conditions.
respectively the decrease and increase of the arriv-
On the other hand, when the arrivals increase, the
als at terminal. For sake of simplicity a uniform
quality performance of the terminal, operated in sim-
variation of the headways for arrivals has been
ulation with the same service level, is affected more
applied for all the lines, assuming a value of 5 min-
dramatically and the model estimates, in the worst
utes for scenario S1 and 7 minutes for scenario 2.
case, a turnaround time of 45 minutes (Table 4).
Any microsimulation experiment is composed
by ten replications for a 1 hour simulation period
and the final indicators are evaluated on the basis Table 2. Turnaround time for service line in S0.
of the average values. A warm up period equal to
30 minute before the simulation period has been Turnaround Time [min] 6.15 6.30 6.45 7.00
also used to obtain significant statistical indicators,
avoiding the empty network state at the initial time. Line 1 Delivery 1 17,0 19,7 23,0 25,2
Line 2 Delivery 6 17,4 18,7 21,7 24,4
Line 3 Collection 2 12,1 13,5 14,4 17,3
4.2 Results for quality indicators in Line 4 Collection 5 12,3 14,0 14,2 16,1
baseline scenarios Line 5 D+C 20,5 23,7 26,6 29,2
A global view of the terminal performance can Line 6 D+C 24,8 27,3 30,0 33,0
be observed in Figure 6 by the queue indicator at Line 7 D+C 23,2 26,8 29,2 31,6
the various sections during the whole simulation Line 8 D+C 23,2 26,8 28,7 32,4
period, including the six cranes where the value is
approx. 1 and the check in/out points with a high
value at the exit. Table 3. Turnaround time for each service line in S1.
The travel time for the 8 service lines provides
Turnaround Time [min] 6:15 6:30 6:45 7:00
a disaggregate estimation of the turnaround time,
highlighting the different quality levels experi- Line 1 Delivery 1 14.4 15.6 15.6 16.6
enced in the terminal, which is in a range between Line 2 Delivery 6 15.4 16.7 16.3 16.7
approximately 12 and 32 minutes (Table 2). As Line 3 Collection 2 11.8 13.0 13.3 13.3
Line 4 Collection 5 11.7 12.2 13.2 13.4
Line 5 D+C 20.1 21.0 21.0 22.3
Line 6 D+C 22.3 22.6 23.6 24.3
Line 7 D+C 22.2 23.4 25.9 25.0
Line 8 D+C 23.5 23.9 23.9 23.5

Table 4. Turnaround time for each service line in sce-

nario S2.

Turnaround Time [min] 6.15 6.30 6.45 7.00

Line 1 Delivery 1 20,6 25,6 30,8 33,2

Line 2 Delivery 6 21,2 26,3 32,1 35,6
Line 3 Collection 2 13,3 16,0 18,9 23,0
Line 4 Collection 5 13,0 15,2 18,4 21,3
Line 5 D+C 23,3 29,4 33,6 38,5
Line 6 D+C 27,8 34,9 39,8 45,3
Line 7 D+C 25,6 30,7 35,2 39,9
Figure 6. Maximum queue at various section of the Line 8 D+C 26,0 32,0 37,6 41,6
terminal.

622

TIS2017_Book.indb 622 2/8/2017 8:08:05 PM

 4.3 Results for quality indicators in Table 7. Turnaround time for each service line in sce-
improved scenarios nario S5.

To test the model in assessing possible improve- Turnaround Time [min] 6:15 6:30 6:45 7:00
ments of the terminal operations, two further sce-
narios have been simulated (S3 and S4). Line 1 Delivery 1 17.7 21.2 24.5 27.4
In S3 a reduction of the service time in the last Line 2 Delivery 6 18.0 22.2 25.9 30.1
phase of the process is generated, whereas the arriv- Line 3 Collection 2 8.8 9.0 10.3 12.8
als are as in S0. This can be obtained by modifying Line 4 Collection 5 9.6 9.9 10.3 12.2
the check-out operations for the loaded vehicles with Line 5 D+C 19.6 23.5 26.2 29.7
a collected ITU, with the support of security technol- Line 6 D+C 22.7 26.4 30.4 35.3
ogies. In Table 5 the higher variations are observed Line 7 D+C 21.7 26.7 29.1 33.2
for collection vehicles and at the end of the period, Line 8 D+C 20.6 26.3 28.9 33.7
when the interactions between vehicles are more rel-
evant. However, also for vehicles not involved in the
improvement at check-out point, there is an observed Table 8. Summary of the scenarios explored.
reduction of the turnaround time, since there is a
unique exit gate and interaction phenomena occur. Service Service Average
Rate of rate for rate for turnaround
In S4 the simulation is focused to assess a pos-
Scenario Arrivals cranes check-out time [min]
sible improvement of the quality level of the S2,
by assuming a better positioning system for the S0 BASE BASE BASE 22.4
crane areas, which may generate a reduction of S1 − BASE BASE 18.7
the service time. In particular, this value was set up S2 + BASE BASE 28.3
to 2 minutes, in place of 3 minutes, to balance the S3 BASE BASE + 20.3
increased time interval of arrivals. S4 + + BASE 27.3
The technical solution tested in S4 is not suffi- S5 + + + 22.0
cient to reduce the turnaround time for each line.
Therefore, a further scenario (S5) is simulated to
combine the decrease of the cranes service time
combination of these two improvements better
and the introduction of security technologies in
responds to the increase in demand.
the check-out area (as S3). Table 7 shows that the
Before reporting the energy results, a summary
Table 5. Turnaround time variation [%] for service line of the explored scenarios and their average turna-
in S3. round time is reported in Table 8 to compare the
assumptions and the actions simulated.
Turnaround Time 6:15 6:30 6:45 7:00 The global results on turnaround time indicate
that the increase of demand simulated in S2 and
Line 1 Delivery 1 −5% −9% −15% −18% the consequent decrease of the quality level can be
Line 2 Delivery 6 −5% −2% −11% −14% managed improving cranes and check out opera-
Line 3 Collection 2 −10% −8% −18% −25% tions. However, to maintain the quality level as
Line 4 Collection 5 −15% −15% −16% −19% observed in S0, both of the improvements should
Line 5 D+C −3% −8% −7% −11% be applied as in S5, since the only improvement of
Line 6 D+C −5% −8% −6% −10% the crane performance is not enough (S4).
Line 7 D+C −6% −7% −7% −6%
Line 8 D+C −5% −10% −8% −12%
4.4 Energy results
Table 6. Turnaround time for each service line in sce- A disaggregate estimation of the energy used in the
nario S4. terminal by road vehicles has been collected dur-
ing simulation for each line (Table 9). As expected,
Turnaround Time [min] 6:15 6:30 6:45 7:00 their variation is in line with the level of conges-
Line 1 Delivery 1 20.4 25.3 31.0 34.8
tion within the terminal. However the variation is
Line 2 Delivery 6 20.3 27.2 31.2 36.2 not linearly dependent with the traffic: at the same
Line 3 Collection 2 12.2 15.7 19.2 23.4 headway variation for arrival (S1 and S2) a different
Line 4 Collection 5 12.7 15.6 18.0 23.3 variation of fuel consumption is estimated, respec-
Line 5 D+C 22.4 26.7 33.1 37.4 tively −16% and +19%. In S3 the improvement
Line 6 D+C 24.3 30.1 34.8 39.7 of the quality has a related reduction in energy,
Line 7 D+C 24.6 29.9 34.8 39.6 although its value is not relevant (−4%). The S4 and
Line 8 D+C 24.9 29.8 34.4 39.2 S5 scenarios show the expected reduction in total
fuel used respect their reference scenario (S2).

623

TIS2017_Book.indb 623 2/8/2017 8:08:06 PM

 Table 9. Total fuel [l] for the different investigated on the current state of the system and explore pos-
scenarios. sible improvements aimed at improve its efficiency,
for example by removing bottlenecks. All data used,
S0 S1 S2 S3 S4 S5 including the terminal layout, although describe
Line 1 24.0 20.1 29.7 22.8 28.8 27.2 realistic scenarios, are assumed mainly to test the
Line 2 25.1 20.6 30.4 23.5 29.9 27.3 ability of this method to reproduce the required
Line 3 21.2 17.7 24.5 20.2 23.6 20.5 process operations and not to provide an appraisal
Line 4 22.1 18.3 25.5 20.4 24.6 22.7 of the terminal performance. In this first phase, the
Line 5 26.3 21.3 31.7 25.3 30.5 28.7 main focus was then to evaluate the effectiveness of
Line 6 34.5 28.9 40.3 33.7 38.3 36.0 the simulation model, by monitoring the variation
Line 7 33.0 29.1 39.3 31.9 37.3 35.7 of the chosen indicators and their consistency with
Line 8 28.8 25.1 35.1 27.8 33.4 30.3 expectation for the established scenarios.
Total 215.1 181.0 256.5 205.6 246.3 228.5 Further research can be carried out to apply the
−16% 19% −4% 15% 6% model with real data provided by a terminal opera-
tor and better tune the parameters used during
simulations. Also, pollutant emissions inside the
Table 10. Time and fuel estimation in idle phases for terminal can be estimated with the simulation tool
S2.
to investigate on possible actions aimed at improv-
Travel Stop Total Fuel % ing the air quality at specific critical areas. A future
S2 Time [s] Time [s] % Time Fuel [l] Idle [l] Fuel research can be devoted to study other alternative
simulators to compare their ability in reproducing
Line 1 932 648 70% 29,73 7,10 24% the required performance for terminals.
Line 2 978 657 67% 30,37 6,90 23%
Line 3 772 489 63% 24,52 4,41 18%
Line 4 757 436 58% 25,45 4,20 17% REFERENCES
Line 5 1268 985 78% 31,74 8,36 26%
Line 6 1392 911 65% 40,29 7,26 18% Ballis, A., 2004. Introducing Level of Service Standards
Line 7 1445 964 67% 39,32 7,27 18% for Intermodal Freight Terminals. In TRB. Washing-
Line 8 1421 1097 77% 35,10 8,37 24% ton: Transportation Research Record: Journal of the
68% 21% Transportation Research Board, pp. 79–88.
Ballis, A. & Golias, J., 2004. Towards the improvement of
a combined transport chain performance. European
These values can be converted in energy per Journal of Operational Research, 152(2), pp.420–436.
vehicle and are in a range between 2 to 6 litres. Available at: http://www.sciencedirect.com/science/
This estimation being related to the global article/pii/S0377221703000341 [Accessed February 8,
process in the terminal, include also the fuel con- 2016].
sumption during the stop phases of the vehicles, Carreira, J.S., Santos, B.F. & Limbourg, S., 2012. Inland
intermodal freight transport modelling. In Euro-
assuming the engine is on. An approximate estima- pean Transport Conference 2012. Glasgow: ETC
tion of the fuel consumed during these phases can Proceedings.
be performed calculating the running time of the Castaldo, M., 2014. È sufficiente ridurre i costi di produz-
vehicle travel time. Table 10 shows the comparison ione per rilanciare il trasporto ferroviario? In Mercin-
between the variation in terms of time and fuel treno 2014. Roma: Trenitalia.
for the most congested scenario (S2). The average European Commission, 2011. WHITE PAPER,
contribution of idle phases to the global time spent Bruxelles.
inside the terminal is 68% in terms of time, which IFEU et al., 2014. EcoTransIT World (Ecological Trans-
is higher than the 21% of total fuel. port Information Tool for Worldwide Transports).
Methodology and Data Update., (December 2014),
p.106. Available at: http://www.ecotransit.org/down-
load/ecotransit_background_report.pdf.
5 CONCLUSIONS Panis, I.L., Broekx, S. & Liu, R., 2006. Modelling instan-
taneous traffic emission and the influence of traf-
The model is built to simulate the terminal with a fic speed limits. Science of the Total Environment,
traffic microscopic approach and has provided an 371(1–3), pp.270–285.
estimation of the required indicators for its qual- Ricci, S. et al., 2016. Assessment methods for innovative
ity, expressed as the turnaround time, and energy, operational measures and technologies for intermo-
expressed as the total fuel consumed by the road dal freight terminals. Transportation Research Pro-
cedia, 14, pp.2840–2849. Available at: http://dx.doi.
vehicles. Also the flexibility of the simulated termi- org/10.1016/j.trpro.2016.05.351.
nal can be easily assessed, since the demand varia- Zumerchik, J., Sr, J.L. & Rodrigue, J., 2011. Incorporat-
tion for arrivals is modifiable in the model. ing Energy-Based Metrics in the Analysis of Intermo-
The model can then be used as a decision sup- dal Transport Systems in North America. Journal of
port tool for terminal operators to provide a view Transportation Research Forum, pp.97–112.

624

TIS2017_Book.indb 624 2/8/2017 8:08:06 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

New features of Tritone for the evaluation of traffic safety

performances

V. Astarita, V.P. Giofrè, G. Guido, A. Vitale, D.C. Festa, R. Vaiana, T. Iuele, D. Mongelli,
D. Rogano & V. Gallelli
Dipartimento di Ingegneria Civile, University of Calabria, Rende (CS), Italy

ABSTRACT: Recent research papers have confirmed that traffic simulation can identify near crashes
events and establish a good base for the estimation of real crashes risk. The objective of this paper is
to present the new features of a microsimulation model originally developed to estimate road safety
performance.
The presented microsimulator has many new features that can be useful to engineers and researchers
such as:
– Dynamic calculation of traffic and road safety indicators.
– Simulation of satellite location data obtained by GPS and smartphones.
– Simulation of adaptive traffic lights activated by FCD data.
– 20 different acoustic emission models.
– Possibility of taking into account “instrumented” vehicles to assess new Intelligent Transportation
System performances.
The paper describes the above listed new features of TRITONE that combined with the calculation of
road safety indicators evaluations allow planners to benefit from the availability of a useful and innovative
tool for traffic simulation.

1 ORIGINS OF TRITONE TRAFFIC This is because in addition to being functionally

SIMULATOR efficient and user-friendly and allowing the user to
evaluate directly and dynamically many road safety
Currently there are several commercially micro- parameters such as Deceleration Rate to Avoid the
simulation packages for road traffic, which provide Crash (DRAC) and Time to Collision (TTC) indi-
detailed information on its likely evolution. How- cators (Guido et al., 2011b), TRITONE has been
ever, only recently some simulators have evolved to developed with new features that are described in
reproduce road safety performance indices (Guido the following parts of this paper.
et al., 2011a, Astarita et al, 2011). For this rea- The equations describing these two indica-
son, in 2009 the University of Calabria developed tors are determined by the kinematic interactions
TRITONE, a traffic simulation package (Astarita between Following Vehicle (FV) and Lead Vehicle
et al., 2012) able to provide, not only the evolution (LV) and are the following:
of road traffic and all usual performance indices
(VFV (t ) −VLV (t ))
2
but also road safety performance indices. In detail,
DRAC
CFV (t t) =
2 * ⎡⎣(X LV (t ) − X FV (t )) − LLV (t ) ⎤⎦
TRITONE is able to represent accurately the pos-
sible areas of a road network that have a higher risk
of accidents. TRITONE can, in fact, reproduce
dynamic traffic evolution evaluating risky situa- CFV (t
TTC t) =
(X LV (t) − X FV (t)) − LLV (t)
tions by taking into account the geometrical aspects VFV (t ) − VLV (t )
of road infrastructure and the actual behavior of
drivers considering in detail all the characteristics where,
of vehicles and drivers. Road safety parameters t = time interval (s)
indicators produced by TRITONE allow the pre- X = position of the vehicles (m)
diction of dangerous situations and plan preventive L = vehicle length (m)
solutions already in the design phase. V = speed (m/s)
TRITONE was conceived with the aim to Moreover, TRITONE was conceived as a
become a leading tool in the field of road safety. university research tool, and is therefore widely

625

TIS2017_Book.indb 625 2/8/2017 8:08:09 PM

 used in this area since it is available to research-
ers and scholars in the field. It allows researchers
to validate their mathematical simulation models
through appropriate standardized libraries or
direct implementations.

2 NEW FEATURES OF TRITONE

In recent years, many new features have been added

to this traffic simulation package. In most cases
new algorithms have been added that reflect cur-
rent development trends of road traffic processes.
The new features that this paper intends to
present are:
• Tools for the dynamic calculation of road safety
indicators.
• Simulation of GPS sensors (smartphones or Figure 1. Road sections with a high probability of
other equipment) in vehicles accidents.
• Simulation of adaptive traffic lights triggered by
mobile devices.
on which link of the urban road network the prob-
• Noise pollution rating.
ability of accident occurrence is higher.
• Auto-generation of road networks from GIS
TRITONE, unlike the Surrogate Safety Analy-
maps
sis Model project (SSAM) from the Federal High-
way Administration (FHWA, 2008), calculates and
evaluates the safety indicators (DRAC and TTC),
3 NEW FEATURES IN THE DYNAMIC dynamically during the simulation without post-
CALCULATION OF TRAFFIC AND processing.
ROAD SAFETY INDICATORS TRITONE identifies the risk areas of the road
network by imposing thresholds to DRAC (<
TRITONE can be used to find solutions to a large 3.40/3.35 m/s2) (AASHTO, 2004, Archer, 2005)
number of issues, TRITONE can, in fact, carry and TTC values (> 1.5 s) (Van der Horst, 1991)
out the traditional dynamic analysis on the evolu- measured on link.
tion of traffic circulation allowing users to com- In each simulation time step, TRITONE
pare their various design scenarios (which may dynamically calculates the DRAC and TTC values
include normal intersections, roundabout inter- for each vehicle. These are then averaged over time
sections, signalized intersections, junctions split- and network links. In a specific interval, each link
levels, etc.). Moreover, it can be used to evaluate has an average value of the DRAC and TTC that is
and optimize road networks in cases of exceptional dynamically compared with a given thresholds. In
events (emergencies, construction sites, accidents, this way TRITONE has the possibility of pinpoint-
evacuation plans, etc.). In all cases, it is possible to ing critical areas or links during the simulation.
use a network of roads corresponding as much as With a few steps, it is possible to model the road
possible to reality (lane width, slopes, narrow pas- network, adding the traffic demand and after cal-
sages, etc.) allowing users. that accept and use this culation of the circulating traffic identify the net-
tool. to obtain reliable results. work links that present critical values.
The original purpose of TRITONE was to eval-
uate in advance the risk of road accidents occur-
rence, this task can be performed now with new 4 SIMULATION OF SATELLITE
multiple representation tools offered by new ver- LOCATION DATA OBTAINED BY GPS
sions of TRITONE. New tools allow the user to AND SMARTPHONES
locate easily and quickly the critical safety issues
that arise on the network. Figure 1 shows an exam- GPS technologies are now available for most of
ple of road sections where it is possible to pinpoint the vehicles on our roads. These sensors are often
possibly unsafe situations with a study of road supplied by the insurance companies to verify the
safety indicators, such as DRAC and TTC. good behavior of the drivers, by security companies
DRAC and TTC indicators are evaluated to prevent theft or simply through smartphones or
directly in TRITONE allowing the user to identify any other mobile communications device in the

626

TIS2017_Book.indb 626 2/8/2017 8:08:10 PM

 vehicle. Satellite location technology has allowed
the Intelligent Transportation System (ITS) indus-
try to provide a myriad of new systems that are
based on smartphones or other instrumentation in
the vehicle.
Many papers have been recently presented on
smartphone applications as a solution to many
traffic problems (Astarita et al., 2014, Festa et al.,
2013), moreover, there have been many researches
on obtaining traffic data from instrumented vehicles
as traffic probes (Floating Car Data), among them
Astarita and Florian, (2001), Rose (2006), Astarita
et al., (2006), Sohn and Hwang (2008), Dailey and
Cathey (2002), Axer and Friedrich (2014). Some
researchers have been studying in particular traffic Figure 2. Simulation of GNSS error.
safety issues: Biral et al. (2005), Vaiana et al. (2014),
Guido et al. (2014), Kerner et al. (2015).
The use of smartphones has been studied to 5 SIMULATION OF ADAPTIVE TRAFFIC
verify the penetration rate required to obtain LIGHTS ACTIVATED BY FCD DATA
positive outcome in traffic information systems:
Ygnace et al., (2000) and Ferman et al., (2003). The article “A Cooperative Intelligent Trans-
Some researchers have been exploring directly on portation System for Traffic Light Regulation
the field such as in the Mobile Century Project Based on Mobile Devices as Floating Car Data
directed by Alex Bayen (Herrera et al.,2010) and in (FCD)” (Astarita et al., 2016) for the first time
Guido et al. (2012, 2013). introduces the possibility of regulating traffic
There are now many systems that exploit light systems through the use of FCD coming
these devices, but they have a big drawback, as from common smartphones that are on board
they can only really be useful (and tested) after vehicles circulating on the road network. A sim-
a substantial percentage of users adopts them. In ple explanation of how it works is contained in
fact, the costs for testing some of these systems Gordon (2016).
in the design phase on a whole road network has Previous academic works have been dedi-
a very high cost. For this reason in TRITONE a cated only to priority control systems of traffic
new module has been developed that can repro- lights for emergency vehicles or public transport
duce the presence of GPS sensors or other Global (Ekeila and other, 2009), dedicated radio links or
Navigation Satellite System (GNSS) on simulated dedicated localization systems and/or extraction
vehicles. of information from FCD. No other academic
TRITON is able to simulate Floating Car Data seminal paper on using a smartphone applica-
(FCD). TRITONE can reproduce the presence tion data to regulate traffic lights is known to the
of a GPS signal in a vehicle, whether it originates authors.
from the insurance black box or from a smart- TRITONE, having the ability cited above to
phone. This makes it possible to study in detail the simulate FCD has allowed the development and
evolution and develop algorithms for complex ITS testing of algorithms useful to obtain a better
systems on entire urban networks. TRITONE per- regulation of adaptive traffic lights. In fact, with
forms this task with an algorithm that can simulate TRITONE it has been possible to evaluate algo-
both the presence of the devices and the position rithms with different values of traffic demand and
error in GNSS (Fig. 2). Moreover, a map-matching different percentages of vehicles equipped with
algorithm is added to obtain data that are similar GNNS. Figures 3–4 show how the arrival of an
to those that would be obtained in the real imple- instrumented vehicle, with a GPS sensor, near a
mentation of such systems. With this methodology traffic light, changes the intersection rules. The
(Lou et al., 2009) it is possible to assess the per- adaptive traffic light gives priority to this vehicle
formance of ITS systems that are based on Float- (always ending the current phase with a smooth
ing Car Data (FCD). transition of green, yellow and then red). The sys-
At the end or during a simulation it is possible to tem estimates the time required for this vehicle to
know the location track of vehicles carrying a GPS perform the maneuver and adjusts the green phase
(GNSS) sensor, this allows the behavior of users in in a congruent manner.
an ITS system to be predicted and/or the perform- Subsequently, at the end of the maneuver of this
ance of the entire simulated road network under a vehicle the traffic light cycle returns to standard
certain ITS control strategy to be evaluated. values always respecting the phase rotation.

627

TIS2017_Book.indb 627 2/8/2017 8:08:10 PM

 mainly average values of traffic flows, both light
and heavy vehicles.
Many models have been developed for traffic
noise prediction (see Barone et al. 2011, Crocco
et al. 2011 and Mongelli 2013 for more details).
There are now more complex solutions that assess
amplification and attenuation of sound also taking
into consideration additional features such as the
type of road surface, the topography of the terrain,
the elements present on the roadside such as anti-
noise walls or weather conditions, etc.
TRITONE also has been expanded to take into
account the calculation of noise emissions in a more
detailed way. The introduction of a noise emission
module in TRITONE has provided multiple for-
mulations to be coupled with the microsimulation
tools, currently in TRITON there are included 20
different noise simulation models, based on the fol-
lowing formulations:
• CSTB in non-built-up areas (NMPB-Routes-96,
Figure 3. Example of precedence of the green traffic 1995)
light. • CSTB in built-up areas (NMPB-Routes-96,
1995)
• Josse (Quartieri et al., 2009)
• Alexandre (Quartieri et al., 2009)
• Burgess (Burgess, 1977)
• Griffiths and Langdon (Griffiths & Langdon,
1968, Bertoni et al., 1988)
• Garcia and Bernal (Grippaldi et al., 199)
• Johnson and Saunders (Johnson & Saunders,
1968)
• OMTC (Quartieri et al., 2009)
• Lamure (Lamure, 1965)
• Corriere and Lo Bosco (Corriere & Lo Bosco,
1991)
• CoRTN (Anon, 1975, HMSO, 1988)
• EMPA (Quartieri et al., 2009)
• C.N.R. (Canelli et al., 1983, Cocchi et al., 1991)
• RLS 90 (RLS, 1990, RLS, 1981)
• C.E.E. (Quartieri et al., 2009)
• CETUR (CETUR, 1980)
• Brambilla
• NMPB (NMPB-Routes-96, 1995, NMPB-
Figure 4. Example of precedence of the green traffic
light.
Routes-2008, 2009, Directive 49/EC, 2002,
SETRA, 2009)
6 EVALUATION OF NOISE POLLUTION The implementation of most current models of
noise emission calculation, allows users to have a
To cope with increasing noise pollution many wide choice for various traffic simulation scenar-
nations like the United States (FHWA), the EU ios. TRITONE allows users also to perform easily
(2002/49/EC), Russia (ODM 218. 013–2011), Japan a comparison between different sound emission
(ASJ RTN-Model 2008) and China (HJ2 4–2009) models. The module structure of the TRITONE
have started to impose clear limits for the protec- simulator is described in Figure 5. It is necessary
tion of human health. Over the years, studies have first to build the network with geometric and
been devoted to the possibility of coupling noise other properties of the roads and the definition
emissions models for road traffic with the current of the O/D matrices, a choice then can be made
micro-simulation techniques. The development of on the behavioral models for drivers (car follow-
noise models began in the 50 s, taking into account ing, lane changing, etc.), it is then possible also to

628

TIS2017_Book.indb 628 2/8/2017 8:08:10 PM

 choose the sound emission model that has to be
used. The simulator will simulate the movement
of vehicles on the net and show, in addition to the
traditional output graphs common of all micro-
simulators also those of sound levels in thematic
maps.
For each instant of simulation, TRITONE
provides the calculation of traffic noise levels as a
function of the noise emission model chosen and
the traffic flow present in each moment on each arc
of the road network.
The simulation then allows the variation of the
Figure 6. Result of sound micro-simulation for experi-
noise level with time to be obtained, for example
mental site.
assessing both what happens in the hours in which
there is intense traffic and in those in which there
is a scarce presence of vehicles. Also it allows the noise emission models with microsimulation it
making of design decisions based on the actual seems possible to analyze driving behavior, traf-
behavior of the vehicles with respect to changes in fic flows and their effects on the road adjacent
the type of infrastructure. areas.
To assess the quality of the TRITON simula-
tor, some tests were carried out in different parts
of the urban area of the city of Cosenza. In the 7 AUTO-GENERATION
real site, for comparison with the simulated values, OF TRANSPORTATION NETWORKS
traffic noise over time also was detected by a sound FROM GIS
level meter. An example of the results obtained are
shown in Figure 6. To obtain a good quality in the results from traf-
These experimental results show that it is pos- fic simulations a high quality of input parameters
sible to extend the functionality of existing road is necessary. For this reason, to avoid mistakes
traffic microsimulators by adding more and more and to increase the quality of input information
features that might be required by international in TRITONE a module has been developed that
standards. By coupling results of researches on can automatically extract simulation parameters
from GIS systems. This module of TRITONE
allows the reconstruction of the transportation
network of an entire city automatically, taking
into account also rail lines. The input data can be
taken from one of the most common GIS systems
on the Internet (OpenStreetMap). Data are then
filtered, according to the user’s needs, and used by
TRITONE to reconstruct the roadway or railway
network.
The roads are drawn precisely, taking into
account details of curved sections. The module
can assign to every road automatically its geomet-
ric (number of lanes, width, etc) and functional
characteristics (speed limit, etc.).
One example of a full network reconstructed
with this module is the network of Sioux Falls
(South Dakota—USA), shown in Figure 7.
As can easily be seen in the picture the access
ramps are modeled precisely allowing to have reli-
able results on the movements of vehicles and their
position in space (feature that can be useful in ITS
modelling).
This new feature results in increased accuracy of
the coded road information, but can also still result
in significant savings of human resources associated
with network coding tasks. For example, a person
Figure 5. TRITONE structure. needs around 35 hours to model the 4105 road sec-

629

TIS2017_Book.indb 629 2/8/2017 8:08:11 PM

 Astarita, V. and Florian, M. 2001. The use of mobile phones
in traffic management and control. In Intelligent Trans-
portation Systems Proceedings IEEE, 2001, pp. 10–15.
Astarita, V., Bertini, R., D’Elia, S. and Guido G. 2006.
Motorway traffic parameter estimation from mobile
phone counts. European Journal of Operational
Research, Vol. 175 (3), pp. 1435–1446.
Astarita, V., Festa, D.C., Mongelli, D. W. E., Tassitani,
A. 2014. New methodology for the identification of
road surface anomalies, Service Operations and Logis-
tics, and Informatics (SOLI), 2014 IEEE International
Conference.
Astarita, V., Giofrè, V. P., Guido, G., Vitale, A. 2011.
Investigating road safety issues through a microsimula-
tion model. Procedia-Social and Behavioral Journal,
Vol. 20, pp. 226–235.
Astarita, V., Giofrè, V. P., Vitale, A. 2016. A Coopera-
tive Intelligent Transportation System for Traffic Light
Regulation Based on Mobile Devices as Floating Car
Data (FCD), 2016, American Scientific Research
Journal for Engineering, Technology, and Sciences
(ASRJETS), Vol. 19–1, pp. 166–177.
Astarita, V., Guido, G., Vitale, A., Giofrè, V. P. 2012. A
new microsimulation model for the evaluation of traf-
fic safety performances, European Transport, 2012, pp
1–16, Paper N° 1, ISSN 1825–3997.
Axer, S. and Friedrich, B. 2014. Level of service estima-
tion based on low-frequency floating car data. Trans-
Figure 7. Example of GIS reconstruction. portation Research Procedia, Vol 3, pp 1051–1058.
Barone, V., Crocco, F., Mongelli, D.W.E. 2011. A math-
tions of the Sioux Falls network (about 116 km2), ematical model for traffic noise prediction in an urban
while with this feature needs only 5.20 minutes. area in Proceedings of the 4th WSEAS International
Conference on Urban Planning Transportation (UPT
’11), pp. 405–410, Corfu Island, Greece.
Bertoni, D., Franchini, A., Magnoni, M. 1988. Il rumore
8 CONCLUSION
urbano e l’organizzazione del territorio, Pitagora Edi-
trice, Bologna.
In conclusion, the idea of pairing the common Biral, F., Lio, M. D., Bertolazzi, E. 2005 Combining
standardized simulation techniques to new con- safety margins and user preferences into a driving cri-
cepts such as Safety Performance evaluations or terion for optimal control-based computation of refer-
the simulation of ITS systems based on satellite ence maneuvers for an ADAS of the next generation.
localization has led to the creation of a useful and In Proceedings of Intelligent Vehicles Symposium,
innovative tool for traffic simulation. IEEE,, pp. 36–41.
New future development of TRITONE, to keep Burgess, M.A. 1977. Noise prediction for Urban Traffic
Conditions, Related to Measurement in Sydney Met-
pace with future technologies in the transportation
ropolitan Area, Applied Acoustics, vol. 10, pp 1–7.
field, include the real time use of TRITONE for the Canelli, G. B., Gluck, K., Santoboni, S. A. 1983. A mathe-
real time implementation of traffic control strategies. matical model for evaluation and prediction of mean energy
level of traffic noise in Italian towns, Acustica, 53, 31.
CETUR, 1980. Guide du bruit des transports terrestres,
REFERENCES fascicule prévision des niveaux sonores.
Cocchi, A., Farina, A., Lopes, G. 1991. Modelli matem-
American Association of State Highway and Transpor- atici per la previsione del rumore stradale: verifica ed
tation Officials (AASHTO), 2004. A Policy on Geo- affinazione del modello CNR in base a rilievi sperimen-
metric Design of Highways and Streets, AASHTO, tali nella città di Bologna, Acta of 19° Convegno Nazi-
Washington, D.C. onale AIA, Naples 10–12 April, 1991.
Anon 1975. Calculation of Road Traffic Noise, London, Corriere, F., Lo Bosco, D. 1991. Valutazione previsionale
United Kingdom Department of Environment and dell’inquinamento acustico nella variabilità urbana,
welsh Office Joint Publication, HMSO. Autostrade, n. 1.
Archer, J., 2005. Methods for the assessment and pre- Crocco, F., D’Elia, S., Mongelli, D. 2011. An integrated
diction of traffic safety at urban intersection and prediction model to define the level of noise in urban
their application in microsimulation modeling. PhD areas, in Proceedings of the 4th WSEAS International
Thesis, Department of Infrastructure, Royal Institute Conference on Urban Planning Transportation (UPT
of Technology, Sweden. ’11), pp. 399–404, Corfu Island, Greece.

630

TIS2017_Book.indb 630 2/8/2017 8:08:11 PM

 Dailey, D. J and Cathey, F. W. 2002 “Virtual speed sensors Kerner, B.S., Demir, C., Herrtwich, R. G., Klenov, S. L.,
using transit vehicles as traffic probes. In Proceedings Rehborn, H., Aleksić, A. and Haug A. 2015 Traffic
of The IEEE 5th International Conference on Intel- state detection with floating car data in road networks.
ligent Transportation Systems, pp. 560–565. In Intelligent Transportation Systems, 2005. Proceed-
Directive 2002/49/EC of the European parliament and of ings. pp. 44–49.
council of June 25 2002 relating to the assessment and Lamure, C. 1965. Niveaux de bruit au voisinage des
management of environmental noise, official journal autoroutes, Proc. Fifth International Congress on
of the European communities, l189/12–25, 18.7.2002. Acoustics.
Ekeila, W., Sayed, T, Esawey, M., 2009. Development and Lou, Y., Zhang, C., Zheng, Y., Xie, X., Wang, W., Huang,
comparison of dynamic transit signal priority strate- Y. 2009. Map-matching for low-sampling-rate GPS tra-
gies, Transportation Research Record Journal of jectories, Proceedings of the 17th ACM SIGSPATIAL
the Transportation Research Board 2111(2111):1–9, International Conference on Advances in Geographic
December. Information Systems. ACM.
Ferman, M. A., Blumenfeld D. E. and Dai X., 2003. A Ministère de l’Ecologie, de l’Energie du Développe-
simple analytical model of a probe-based traffic infor- ment durable et de l’Aménagement du terroire, 2009.
mation system. In IEEE Transactions on Intelligent Road noise prediction, 2 -Noise propagation computa-
Transportation Systems,Vol. 1, pp. 263–268. tion method including meteorological effects, Meth-
Festa, D.C., Mongelli, D.W.E., Astarita, V., Giorgi, P. odologic guide (NMPB2008). Service d’études sur les
2013. First results of a new methodology for the identi- transports, les routes et leurs aménagements Éditions
fication of road surface anomalies, Service Operations Sétra.
and Logistics, and Informatics (SOLI), 2013 IEEE Mongelli, D. W. E. 2013. An integrated prediction model
International Conference. for traffic noise in an urban area, Service Operations
FHWA, 2008. Surrogate Safety Assessment Model and Logistics, and Informatics (SOLI), 2013 IEEE
(SSAM), HWA-HRT-08-049. International Conference.
Gordon J. 2016. Talking to traffic signals. Traffic technol- NMPB-Routes-96 (SETRA-CERTU-LCPC, CSTB), the
ogy international, October/November 2016, pp52–55. French national computation method, referred to in
Griffiths, I.D., Langdon, F.J. 1968. Subjective Response ‘Arrêté du 5 mai 1995 relatif au bruit des infrastruc-
to road traffic noise, Journal of Sound and Vibration tures routières, Journal Officiel du 10 mai 1995, Arti-
8, 16–32. cle 6’ and in the French standard ‘XPS 31–133’
Grippaldi, V., Barbaro, S., Cosa, M. 1996. Modelli di pre- Quartieri, J., Mastorakis, N. E., Iannone, G., Guarnac-
visione per il rumore da traffico in ambiente urbano: un cia, C., D’Ambrosio, S., Troisi A. and Lenza, T.L.L.
quadro di confronto. 2009. A Review of Traffic Noise Predictive Models,
Guido, G., Astarita, V., Giofrè, V. P., Vitale, A., 2011a. Recent advances in applied and theoretical mechanics,
Safety performance measures: a comparison between ISBN: 978–960–474–140–3, pp. 72–80.
microsimulation and observational data. Procedia- Rose, G. 2006. Mobile phones as traffic probes: practices,
Social and Behavioral Journal, Vol. 20, pp. 217–225. prospects and issues. Transport Reviews, Vol. 26(3),
Guido, G., Saccomanno, F. F., Vitale, A., Astarita, V., pp. 275–291.
Giofrè, V. P., 2011b. Un algoritmo per l’acquisizione RLS, 1981. Richtlinien für den Lärmschutz an Strassen.
delle traiettorie veicolari da immagini video. Strade & BM für Verkehr, Bonn.
Autostrade, 2011, Vol. 1, pp. 132–135. RLS, 1990. Richtlinien für den Lärmschutz an Strassen.
Guido, G., Vitale, A., Astarita, V., Saccomanno, F, BM für Verkehr, Bonn.
Giofrè, V. P. and Gallelli, V. 2012. Estimation of SETRA, 2009. Prévision du bruit routier, 1 - Calcul des
safety performance measures from smartphone sensors. émissions sonores dues au trafic routier, juin 2009,
Procedia-Social and Behavioral Sciences, Vol. 54, pp. 124p, 978-2-11-095825-9.
1095–1103. Sohn, K and Hwang K. 2008. Space-based passing time
Guido, G., Vitale, A., Saccomanno, F, Festa, D.C., Astar- estimation on a freeway using cell phones as traffic
ita, Rogano D. and Gallelli, V. 2013 Using smartphones probes. Intelligent Transportation Systems, IEEE
as a tool to capture road traffic attributes”. In Applied Transactions on, Vol. 9(3), pp. 559–568.
Mechanics and Materials, Vol. 432, pp. 513–519. Ygnace, J. L., Drane, C., Yim Y. B. and De Lacvivier R.
Guido, G., Gallelli, V., Saccomanno, F., Vitale, A., 2000. Travel time estimation on the san francisco bay
Rogano, D. and Festa, D. C. 2014. Treating uncer- area network using cellular phones as probes. Cali-
tainty in the estimation of speed from smartphone traf- fornia Partners for Advanced Transit and Highways
fic probes. Transportation Research Part C: Emerging (PATH).
Technologies, Vol. 47, pp. 100–112. Vaiana, R., Iuele, T., Astarita, V., Caruso, M. V., Tas-
Herrera, J. C., Work, D. B., Herring, R., Ban, X. J., Jacob- sitani, A., Zaffino C. and Giofrè, V. P. 2014 Driving
son Q. and Bayen, A. M. 2010. Evaluation of traffic data behavior and traffic safety: an acceleration-based safety
obtained via GPS-enabled mobile phones: The Mobile evaluation procedure for smartphones. Modern Applied
Century field experiment. Transportation Research Part Science, Vol. 8(1), 88.
C: Emerging Technologies, 18(4), 568–583. Van der Horst, A. R. A., 1991. Time-to-collision as a
HMSO Department of Transport, 1988. Calculation of Cue for Decision-making in Braking. In: Brown, I. D.,
Road Traffic Noise, United Kingdom. Haselgrave, C.M., Moorhead, I. and Taylor, S. (eds),
Johnson D.R. & Saunders E.G., 1968. The evaluation of Vision in vehicles 3: 19–26.
noise from freely flowing road traffic, J. Sound. Vib.,
7(2):287–309.

631

TIS2017_Book.indb 631 2/8/2017 8:08:12 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Genetic algorithm-based calibration of microscopic traffic simulation

model for single-lane roundabouts

O. Giuffrè, A. Granà & M.L. Tumminello

DICAM, University of Palermo, Palermo, Italy

A. Sferlazza
DEIM, University of Palermo, Palermo, Italy

ABSTRACT: A calibration procedure for microscopic simulation models based on a genetic algorithm
is proposed. Focus is made on single-lane roundabouts for which many random factors such as gap-
acceptance affect operations. A comparison is performed between the capacity functions based on a meta-
analytic estimation of critical and follow up headways and simulation outputs of a roundabout built in
Aimsun microscopic simulator. Aimsun parameters were optimized using the genetic algorithm tool in
MATLAB® which automatically interacted with Aimsun through a Python interface. Results showed that
applying the genetic algorithm in the calibration process of the microscopic simulation model, a good
match to the capacity functions was reached with the optimization parameters set. By this way, automa-
tion of the calibration process results effective for analysts which use traffic microsimulation for real world
case studies in the professional sphere.

1 INTRODUCTION Results of several applications, as the technical

literature in the road engineering sector refers, show
1.1 Background that simulation-based optimization methods and,
among these, the genetic algorithms can be usefully
Microscopic traffic simulation models have become
applied in the calibration process of microscopic traf-
increasingly useful tools for the advanced analy-
fic simulation models (see e.g. Chiappone et al. 2016,
sis of transport systems and have proven to be
Essa & Sayed 2016, Rubio-Martín et al. 2015, Nay-
an active field of research in computer science
eem et al. 2014, Menneni et al. 2008). Incorporating
and transportation engineering (Barceló 2010).
the optimization problem within the model calibra-
Advances in research and current application to
tion, the iterative process of manually adjusting the
road and highway planning and design over the
model parameters, that users and practioners are
last few years have outlined their great potential to
required to perform, is now automatized. However,
assess operational performances and safety effects,
in order to automate the iterative process of manu-
since they can support the evaluation of road policy
ally adjusting the model parameters, microscopic
and infrastructure changes before implementing
traffic simulation models have to be still enhanced
them in the real world (see e.g. Papathanasopou-
with custom models and/or have be equipped with
lou et al. 2016, Giuffrè et al. 2016a, Vasconcelos
various APIs (when available) to remotely control
et al. 2013). Differently from analytical approaches,
the simulation. Now for practical implications, the
microscopic traffic simulation models capture
question is how to provide integrated software solu-
road traffic interactions through a combination
tions, user-friendly for practioners and transporta-
of complex algorithms which take into considera-
tion engineers which use microsimulation for real
tion car following, lane changing and gap accept-
world case studies in the professional sphere.
ance describing real-world driving behavior. Thus,
Despite several encouraging results already
microsimulation enables the analyst to develop
obtained from studies on the specific topic, it is note-
increasingly higher levels of complexity and uncer-
worthy that commercial traffic simulation software,
tainty in operations of road networks and single
specially formulated on the operating principles of
installations (Ayyub 2011). However, concerns are
genetic algorithms are not yet available. Thus engi-
often expressed by practioners about the possible
neers often need an analytical aid to develop their
misuse of traffic microsimulation. In simulation
own codes and adapt the objective function to the
studies, indeed, model calibration is a very crucial
specific requirements of the problems that are often
task, since reliable results must be obtained from
encountered in professional practice. However, it is
the analysis that we made.

633

TIS2017_Book.indb 633 2/8/2017 8:08:14 PM

 now possible to use software packages with specific Single-lane roundabouts are designed for low-speed
optimization tools as for example MathWorks’s® operations and represent one of the safest treat-
proposes. On this regard, Chiappone et al. (2016) ments for at-grade intersections (see Rodegerdts
applied the genetic algorithm tool of MATLAB® et al. 2010). Figure 1 exhibits the sketch of the at-
to reach convergence of the outputs from Aimsun grade single-roundabout, features of which comply
microscopic simulator to the empirical data (that is with the instructions by the Italian standards on
to minimize the differences between the field meas- geometric design of compact roundabouts (2006).
urements observed in the speed-density diagram The geometric design of the roundabout
and the simulator’s outputs obtained for a selected included: a) the outer diameter (that is the basic
freeway segment); the automatic interaction with parameter used to define the size of a roundabout
Aimsun software was achieved through an external and is measured between the outer edges of the
Python script, so that the data transfer between the circulatory roadway) of 39,0 m; b) the circula-
two programs could automatically happen. tory roadway width of 7,0 m; c) the entry width of
The theory and background of the genetic algo- 3,75 m; d) the exit width of 4,50 m; e) the length
rithms are not presented in this paper since they of legs reaching the roundabout of 35 m with-
are not the objective of this work; instead, we are out parking possibilities for vehicles from 20 m
interested on the applications of genetic algorithms up to the approach zones. By this way, based on
to the calibration of traffic simulations with Aim- the selected roadway widths, the resulting speeds
sun. However, the reader is referred to the large resulted moderate enough to accommodate mixed
number of online manuals for further knowledge traffic flows of passenger cars and heavy vehicles
of genetic algorithms as for example the online in whatever context, also in the built environment.
MathWorks’s® website proposes.
2.2 Behavioural parameters and capacity model
1.2 The objectives of the paper
Traffic operations at roundabouts are typically ruled
The paper proposes a calibration procedure for by the gap acceptance process, which specifically for
microscopic simulation models based on a genetic single-lane sites is also facilitated by speeds mod-
algorithm. Focus is made on single-lane roundabouts erated by the particular geometric design. In gap
for which many random factors such as vehicles acceptance process the critical headway and the fol-
arriving and gap acceptance can affect operations. low-up headway are two key factors in determining
Capacity functions were calculated by specifying the entry capacity which, in turn, depends on circulating
well-known Hagring model for a single-lane rounda- flow under a specified arrival headway distribution
bout (Hagring 1998). Based on the results obtained (see Mauro 2010). Many methods have been devel-
in a previous work by Giuffrè et al. (2016b), a meta- oped to estimate these parameters from gap data; as
analytic estimation of the critical and the follow up a matter of fact, technical literature proposes several
headways was used to calculate the capacity func- empirical research studies focusing on estimation of
tions; they represented the empirical target values of critical and the follow-up headways at roundabouts.
entry capacity to which we compared the simulated However, the decision whether a hypothesis is valid
output. For pursuing the above stated objective, or a statistical method is appropriate for the estima-
the calibration parameters—that cause the model tion of the population parameters, or in general for
to best reproduce empirical capacities—were pre- the study to be performed, cannot be based on the
liminarily identified using sensitivity analysis for a
single-lane roundabout built in Aimsun. A compari-
son was performed between the empirical capacity
functions and the simulation outputs obtained with
the default values for the selected parameters of
Aimsun. The genetic algorithm tool in MATLAB®
was then applied in order to minimize the differ-
ences between the two data set and the automatic
interaction with Aimsun was implemented through
an external Python script that we wrote.

2 PRELIMINARY ANALYSIS

2.1 The roundabout geometric configuration

Based on current design applications, a scheme of
single-lane roundabout was selected for this study. Figure 1. The sketch of the single-lane roundabout.

634

TIS2017_Book.indb 634 2/8/2017 8:08:14 PM

 results of a single empirical research, since results can tion (Cowan 1975) was assumed for the circulating
vary from one research to the next. Rather, analysts stream and the meta-analytical values estimated for
need a mechanism to synthesize data across stud- the critical and follow up headways were used or:
ies (see Borenstein et al. 2009). In this view—finding
that the application of quantitative methods to sum-
exp ⎡⎢ − c (T − )⎤⎥
marize the results of several empirical studies is now Q
widespread in several research fields (see e.g. Elvik ⎛ Δ Qc ⎞ ⎣ 3600 ⎦ (1)
Ce = Qc 1 −
⎝ 3600 ⎠
1 − exp ⎛ − Tf ⎞
2011)—in a previous work, we performed a meta- Qc
analysis, that is the analysis where each (primary) ⎝ 3600 ⎠
study yields an estimate of statistical mean values of
the critical headway and the follow-up headway (i.e.
the effect sizes), we assessed the dispersion in these where Ce = entry-lane capacity [pcu/h]; Qc = circu-
effects and then we computed a summary effect for lating traffic flow [pcu/h]; T = critical headway [s];
each parameter. The reader is referred to Giuffrè Tf = follow –up headway [s]; Δ = minimum headway
et al. (2016b) for a more effective presentation of of major flow [s] that, according to literature, was
the results of the meta-analysis of effect sizes, per- assumed equal to 2.10 s.
formed as part of the literature review through the Thus the entry capacities derived from the
random-effects model. Hagring model for each entry of the single-lane
Table 1 shows the summary effects (i.e. the ran- roundabout, were used in the following as refer-
dom quantitative meta-analytic estimate) for each ence for calibration purposes.
parameter, the 95% lower and upper limits for each
summary effect, the results of the Cochran’s Q test 2.3 Aimsun modelling
(or a measure of heterogeneity that is the sum of
the squared deviation of each effect size from the The microscopic simulation package AIMSUN
mean, weighted by the inverse-variance for each (version 8.1) was used for microscopic modelling
study) and the Higgin’s index I 2 (the ratio of true of the single-lane roundabout; no priority was cre-
heterogeneity to total observed variation). Accord- ated for the legs approaching the roundabout, but
ing to Borestein et al. (2009) the p-value close to priority to vehicles moving anticlockwise on the
zero and I 2 less than 25% for both headways, con- ring was established. Detectors were located so that
firmed the absence of heterogeneity in the data set they could replicate the possible location of field
examined for the single-lane roundabout. At last, detectors, that is at each entry/exit and upstream/
the meta-analytic estimates for the two headways downstream of each entry. The model building
were found nearly consistent across all studies and, produced a single-lane roundabout scheme that
compared to the single studies, provided a more was reproducible and verifiable; at the same time,
reliable result for the parameters of interest. this scheme reproduced as accurately as possible
Further activities involved the construction the geometric intersection pattern of a possible real
of empirical functions for entry capacity; for this world counterpart (Dowling et al. 2004). Before
purpose, the general Hagring’s model for multi– applying the GA-based procedure that automati-
lane intersections was used (Hagring 1998). The cally runs to test the model validity, traffic con-
Hagring’s model was specified for the single-lane ditions on the roundabout were reproduced in
roundabout where a Cowan’s M3 headway distribu- Aimsun with default settings; indeed, we consid-
ered the default parameters of Aimsun version 8.1.
As far as the traffic demand, we used O-D matri-
Table 1. The meta-analytic estimates for critical and ces with due consideration to the direction of turn.
follow-up headways at single-lane roundabouts (Giuffrè Passenger cars were only considered at this stage;
et al. 2016a). the attributes of passenger cars were the default
values given by Aimsun. In order to reproduce the
Headway [s] traffic demand and represent realistic traffic condi-
tions on the roundabout, we assigned O/D matrices
Statistics Critical Follow-Up
such as reproducing the circulating flow Qc, from 0
Random estimate 4.27 3.10 veh/h to 1400 veh/h with a step of about 200 veh/h;
Standard error 0.11 0.07 these matrices guaranteed saturation conditions at
95% lower limit 4.05 2.96 entries, so that the number of vehicles entering the
95% upper limit 4.49 3.25 roundabout was the capacity for the specific entry
Z-value 37.46 41.82 each time considered. It should be noted that at the
p-value 0.00 0.00 time when the analysis presented in this research
Q 23.28 33.90 was carried out, we found that no single-lane round-
I2 1.19 20.37 about having a modern design was installed in our
City. However, different existing unsignalized and

635

TIS2017_Book.indb 635 2/8/2017 8:08:15 PM

 signalized intersections, having size comparable to will be specified for the problem under study in the
the roundabout that we built, had been identified next section); g(⋅) = either the square or the abso-
in the urban road network and some of them were lute-value function; yi,k = the experimental value of
likely to be converted into the single-lane rounda- the i-th variable at the instant k; ŷi,k(uk,β) = the
bout as that chosen for calibration purposes (see e.g corresponding simulated value that is a function
Giuffrè et al. 2011; Giuffrè et al. 2014). of the input u and the parameter vector β.
Calibration parameters were preliminarily identi- The solution of the calibrating problem will be
fied by using a sensitivity analysis; it required that the parameter vector β* that minimize the objec-
some model parameters were changed and adjusted tive function (see Equation 2), or:
in an iterative way until model outputs were close
to empirical data, based on a predefined level of
agreement between the two data sets. The compari- β* β (β ) (3)
son between the set of model parameters manually
derived for calibration purposes and the best combi- Problem expressed in Equation 3 can be solved
nation of the values for the GA-optimized parameters iteratively; however, two problems have to be solved.
is reported in sections 3 and 4. Thus in the context of The first problem involves the initial condition to be
microsimulation based modelling, the capacity func- chosen, whereas the second problem is represented
tions derived from Equation 1 were compared to the by the stopping criteria. According to Davis (1991),
simulation outputs obtained using Aimsun. the problem of the initial condition should not be
In order to automate the iterative process of undervalued. It should be noted that most algorithms
manually adjusting the model parameters, the cali- only search for local minima; when we face multiple
bration of the microscopic traffic simulation model minima (non-convex problem), the algorithm usually
was then formulated as an optimization prob- converges only if the initial guess is already somewhat
lem which searches for an optimum set of model close to the final solution. However, this problem is
parameters through an efficient search method as avoided if genetic algorithms are used, since they are
will be explained in the following section 4. evolutionary optimization algorithms robust with
respect to the initial condition (Davis 1991).
The second problem, instead, can be easily
3 STRUCTURE OF THE GA-BASED solved by selecting a maximum number of itera-
METHOD tions or, the algorithm can be stopped when:

Now we propose the formulation and the solution j ( β )k j ( β )k − 1

of the calibration problem: the formal interpreta- <ε (4)
tion of the problem is given and, subsequently, j ( β )k − 1
the solution by applying genetic algorithms is
described.
where ε = the error stop quantity; j(β)k and
Let {uk}k =1…N and {yk}k =1…N be two input-output
j(β)k−1 = the values of j(β) computed at the itera-
sequences of observed data acquired during suit-
tions k and k−1, respectively. This stopping criteria
able traffic measurements; they are the “experi-
means that the algorithm will be stopped when the
mental surveys”. We want to reproduce the same
objective function variation, between two consecu-
output sequence corresponding to the same input
tive instants, is less than a quantity ε freely chosen.
sequence by simulation. In order to obtain the sim-
The reader is referred to the online MathWorks’s®
ulated output, denoted with {ŷk}k=1…N, the model
website, or to the large number of online manuals,
has to be calibrated; this means to find values
for further details on genetic algorithms and how
for the model parameters such that the simulated
to perform genetic algorithms optimization.
output {ŷk}k=1…N is as close as possible the observed
In our application the “experimental surveys”
output {yk}k =1…N given the same input {uk}k=1…N.
consist of entry capacity values calculated by using
The cost function is defined as:
the Hagring model as a function of the meta-
analytic estimation of the critical and the follow
j (β )
1 q
( )
w ⎡ N g yi , k yˆi , k ( k , β ) ⎤
N i =1 i ⎣ k =1 ⎦
up headways (see previous subsection 2.2). The
estimated output was generated by means of Aim-
(2) sun software which ran with a fixed model cor-
responding to the model under study, and tuned
where k = the discrete time instant; N = the number with a suitable set of parameters. It is obvious that
of measures (each one at each time instant); q = the if the selected parameters are incorrect, then the
number of outputs considered for the identifica- estimated capacity values do not coincide with the
tion procedure; wi = the weight associated with the experimental survey. For this reason, let us select
error on the i-th variable (the generic i-th variable the objective function in Equation 2 as follows:

636

TIS2017_Book.indb 636 2/8/2017 8:08:15 PM

 1 N ⎡ as a restriction of the deceleration component in
J (β ) = ∑ k =1 (C k Cˆ k ( β )) ⎤
2
(5)
N ⎣ ⎦ the car following model and applies this constraint
before updating the position and the speed of the
vehicle (i.e. leader) respect to its follower.
where N = 32, since we have considered observa-
At last the speed acceptance (Sacc ≥ 0) represents
tions (see subsection 2.3) distributed over eight
the level of goodness of the drivers or the degree of
hours, in each of them we ran four simulations
acceptance of speed limits: Sacc ≥ 1 means that the
(one every fifteen minutes).
vehicle will take, as maximum speed for a section,
In the case of Aimsun traffic simulation model,
a value greater than the speed limit, whereas Sacc ≤
the model behaviour is depending on a wide vari-
1 means that the vehicle will use a lower speed limit.
ety of model parameters. In general, if we consider
In the Aimsun car-following model the speed accept-
the model to be composed of entities, (i.e. vehicles,
ance, together with other parameters such as the tar-
sections, intersections, and so on), each of them
get speed and the section speed limit, helps to define
described by a set of attributes, (i.e. parameters of
the desired speed for each vehicle on each section.
the car following, the lane change, gap acceptance,
The speed acceptance may, moreover, influence the
speed limits and speed acceptance, and so on), the
behaviour of the gap-acceptance model; several vehi-
model behaviour is determined by the numerical
cle parameters (i.e. speed acceptance, turning speed,
values of these parameters.
desired speed and so on), influencing all vehicles of
As part of the calibration process we evaluated
a particular type when driving anywhere in the net-
beforehand the calibration parameters by using a
work, also have an influence on the output of the
sensitivity analysis, or recalculating the simulation
model. Now using Equations 5 and 6, the problem in
outputs under different alternative assumptions
Equation 3 can be solved implementing the genetic
to determine the impact of the explored variables,
algorithm tool in MATLAB®. This tool has been
and then we were able to establish for the particu-
applied in order to minimize the differences between
lar model the influence of some important param-
the two set of empirical and simulation data, and
eters on the capacity of the entries. In this way,
the automatic interaction with Aimsun was imple-
we pursued the objective of finding the values of
mented through an external Python script.
the parameters which were able to produce a valid
Starting from a generic initial condition, the
model. Thus, we choose as parameters for the opti-
genetic algorithm generates a set of parameters β,
mization the following:
and then the Aimsun software is running with the
parameters β. Aimsun is attached to MATLAB®
β =[ ] (6) via a python subroutine that allows the data trans-
fer between the two programs. Thus Aimsun pro-
where RT = the reaction time; Hmin = the minimum vides a set of estimated outputs (one for each β)
headway; Sacc = the speed acceptance. and the algorithm computes the objective func-
More specifically the driver’s reaction time or tions 5 associated with each β. At last, the algo-
more easily the reaction time, as used in the car- rithm selects the best parameter β and generates
following model of Aimsun, is the time in seconds a new set of parameters β that is the new genera-
that it takes a driver to react to speed changes in tion. The cycle goes on and on until the specified
the preceding vehicle. The reaction time assigned stopping criterion occurs. We chose the stopping
to a vehicle, moreover, is a global parameter of criterion with a specified, fixed, maximum number
Aimsun, that means during each trip is constant. of iterations (50 generations). In our case, moreo-
In each simulation run, we set the parameter ver, the initial population used to seed the genetic
as fixed and equal to simulation step, that is the algorithm is given by 20 individuals, by using
same value for all vehicles. Moreover, the reaction the default setting of Aimsun as first individual.
time may influence the computing performance Based on these choices, the computational time is
and some simulation outputs, such as the section about 4 hours by using an Intel(R) Core (TM) 2
capacities: in general, the lower the reaction time Quad CPU Q9300 2.50GHz and 8Gb of RAM.
is, the higher capacity values can be obtained. The We stopped the algorithm after 50 iterations so
reason for this is that the drivers are more skilful, that the value of the cost function in Equation 5
as they have shorter reaction times; they can drive reached a steady-state and the algorithm could be
closer to the preceding vehicles, they can find gaps stopped. An upper bound β '' and a lower bound β '
more easily, they have more opportunities to enter for β were introduced to restrict the search domain
the network, etc. The minimum headway is prima- and the condition β '≤ β≤ β '' was then established
rily a lane-changing model parameter; setting this to avoid that parameters without a physical mean-
parameter ensures the minimum headway (mini- ing were generated (i.e. negative reaction time, neg-
mum time in seconds) between the leader and the ative distance among vehicles, etc...). The best β*
follower. Aimsun traffic simulation model includes obtained by solving the optimization problem can
the minimum headway between leader and follower only represent the best value of reaction time, the

637

TIS2017_Book.indb 637 2/8/2017 8:08:15 PM

 Figure 3. Entry capacity: simulation vs Hagring data.

from manual calibration (“Aimsun parameters”),

and the empirical data which were based on the
Hagring model where the meta-analytical estima-
tion of the behavioural parameters was introduced
Figure 2. Block diagram representation of the GA. for capacity calculation. We can see in Figure 3 that
the simulation which used the best parameter set,
minimum headway and the speed acceptance such as returned by GA application, generated a better
that the simulated capacity values track as well as match to Hagring data.
possible the empirical ones. This gives an efficient In addition to the better match to the empirical
automated calibration procedure for simulation data obtained for the simulation with the optimiza-
with Aimsun. tion parameters as shown before, we also analysed
Figure 2 shows the outline of GA which sum- the positive effects of the proposed optimiza-
marizes how the genetic algorithm works and exe- tion procedure by calculating the cost function in
cutes the processing steps, whereas the results are both cases, i.e. for simulation with the default and
presented in the next Section. optimization parameters. Figure 4 shows the val-
ues of the fitness function—that is the cost func-
tion J(β) in Equation 5 - during the optimization
4 SIMULATION RESULTS period, and the corresponding value of the same
function computed with the default parameters.
The optimization problem was solved by apply- The cost function resulted equal to J(β) = 126.8
ing the algorithm illustrated before. According to for optimized parameters and equal to J(β) = 135
the convergence condition, after 50 generations for parameters derived from manual calibration. We
(approximately 4 h of computing time), the algo- want to highlight that the benefit in tuning the model
rithm converged on the optimal solution. The parameters would have been greater if we had used
best combination of the values for the simulation Aimsun default parameters as the initial condition
parameters included: (J(β) = 159.04 for Aimsun default parameters).
This figure shows the positive effect derived from
− the reaction time of 0.86 s instead of the Aim-
the application of the proposed genetic algorithm-
sun default value of 0.80 s;
based calibration, since a better matching to the field
− the minimum headway of 1.58 s instead of the
data is clear in comparison with the simulation that
Aimsun default value of 0 s;
used the default parameters. The graphs in Figure 4
− the speed acceptance of 1, instead of the Aim-
also plots the mean score of the population at every
sun default value of 1.1.
generation: the mean best fit is the mean value of
These values were different from values derived the cost functions calculated for all individuals of
from manual calibration or: the reaction time of the same generation, whereas the best fit is the cost
0.85 s; the minimum headway of 1.6 s, the speed function of the best individual within the generation.
acceptance of 1. Because the genetic algorithm function minimizes
The resulting parameter set returned by the the fitness function, the best fitness value for a popu-
genetic algorithm with the least error was chosen lation is the lowest fitness value for any individual
as the best parameter set (or the best combination in the population. To further test the model valid-
of the model parameters), and then used for pre- ity despite it would have been accepted, we used the
dicting the simulated capacity. GEH index as criterion for acceptance, or otherwise
Figure 3 shows the Aimsun simulation out- rejection, of the model. This index is a global indica-
puts which were gained by using both the best tor widely used in practice for validating traffic micro-
output parameters and the parameters derived simulation models, especially when only aggregated

638

TIS2017_Book.indb 638 2/8/2017 8:08:16 PM

 5 CONCLUSIONS

This paper proposed a procedure based on a genetic

algorithm to calibrate a microscopic traffic simula-
tion model. Microscopic traffic simulation models
are now widely used for real-world case studies in
transportation engineering, since they are faster
and safer than actual on-field implementations.
However, without a proper calibration the results
of the analysis could be misleading and unreliable.
We were interested on the applications of the
genetic algorithm-based calibration to traffic simula-
tions with Aimsun. Focus was made on roundabouts
Figure 4. Values of the cost function J(β) during the for which many random factors such as gap-accept-
optimization period. ance can affect operations. In order to automate the
iterative process of manually adjusting the model
parameters, the calibration of the microscopic traf-
fic simulation model was formulated as an optimi-
zation problem which searches for an optimum set
of model parameters through an efficient search
method. Calibration parameters were preliminar-
ily identified using sensitivity analysis, whereas the
optimum values for these parameters were obtained
by minimizing the error between the simulated and
empirical capacity by using a genetic algorithm.
The proposed calibration procedure was, indeed,
applied to the Aimsun software by using a genetic
algorithm to systematically modify the model
Figure 5. Simulated and empirical relationship with parameters and then to fit the capacity values
95% confidence limits. obtained from simulations to the capacity func-
tions based on a meta-analytic estimation of the
values are available as flow counts at detection sta- critical and the follow up headways. The genetic
tions aggregated to the hour and entry capacity val- algorithm tool in MATLAB® was applied in order
ues (see Barceló 2010, pp. 46–47). Since the deviation to minimize the differences between the two data
of each simulated value with respect to the measure- set and the automatic interaction with Aimsun was
ment for each entry was lower than 5 in 88% of the implemented through an external Python script
cases (however, lower than 8 in the remaining cases), that we wrote so that the data transfer between the
the model could be considered “calibrated” in terms two programs could automatically happen.
of its ability to reproduce the empirical entry capaci- The genetic algorithm-based approach appeared
ties at the examined single-lane roundabout and has to be effective in the calibration of Aimsun; indeed,
been accepted as significantly able to reproduce local the results showed a better match to the empiri-
conditions and traffic behaviour in a satisfactory cal data than simple manual calibration. Despite
way. Note that for the examined roundabout the root this research is a first exploratory analysis for
mean squared normalized error—which provides roundabouts, it should be noted that the compari-
information on the magnitude of the errors rela- son between the empirical data and the simulation
tive to the average measurement—resulted less than results obtained with the default and optimized
0.10, whereas the mean absolute percent error—also parameters, gave insights into the performance of
calculated as supplemental parameter to measure the the calibration procedure that we proposed. The
size of the error in percentage terms—resulted less sensitivity analysis performed to identify the cali-
than 5%. Figure 5 depicts an example of scattergram bration parameters with greater effect on output,
analysis developed to compare empirical versus sim- indeed, confirmed to us that no additional ben-
ulated capacities at entries of the examined single- efits to the calibration process would be given by
lane roundabout. The regression line of empirical a higher number of calibration parameters. Thus,
versus simulated capacity was plotted along with the given the number of model parameters here used,
95% Prediction Interval (95% PI). Based on the R2 we are confident that the procedure can be applied
value of 0.9919 and the fact that most of points were to other case studies. Based on the results of this
within the confidence band of the regression lines, research—summarized by the values of the cost
we reached the conclusion which the model could be function J(β) during the optimization period in
accepted as significantly close to the reality. Figure 4—it is evident that benefits resulting from

639

TIS2017_Book.indb 639 2/8/2017 8:08:16 PM

 the use of a genetic algorithm can compensate the Ghods, A.H & Saccomanno, E.F. 2010. Comparison of
computational efforts deriving from the applica- carfollowing models for safety performance analysis
tion of an optimization technique which auto- using vehicle trajectory data. Annual Conference of the
mates the iterative process of manually adjusting Canadian Society for Civil Engineering 2010 (CSCE
2010), Winnipeg, MB, Canada, June 9–12, 2010. Vol-
the simulation parameters. Anyhow, if we had ume 2: 13391348.
selected more parameters (a situation which can Giuffrè, O., Granà, A., Roberta, M. & Corriere, F. 2011.
occur when more complex problems have to be Handling underdispersion in calibrating safety per-
managed), probably we would have had to deal formance function at urban, four-leg, signalized inter-
with other issues such the nature of the obtained sections. Journal of Transportation Safety and Security
optimum (i.e. if it is a local minimum or the abso- 3(3): 174–188.
lute minimum) and/or how well the local minimum Giuffrè, O., Granà, A., Giuffrè, T., Marino, R. & Marino, S.
approximates the absolute one, and so on. 2014. Estimating the safety performance function for urban
At last, it should be said that the importance and unsignalized four-legged one-way intersections in Pal-
ermo, Italy. Archives of Civil Engineering 60 (1): 41–54.
the advantages of this procedure are also to improve Giuffrè, O., Granà, A., Marino, S. & Galatioto, F. 2016a.
the efficiency of the model to predict different sce- Microsimulation-based passenger car equivalents for
narios (that may occur in reality) by varying differ- heavy vehicles driving turbo-roundabouts. Transport
ent initial conditions. Moreover, the application of 31 (2): 295–303.
this procedure to more complex scheme of multi- Giuffrè, O., Granà, A. & Tumminello, M.L. 2016b. Gap-
lane roundabouts should be designed to address acceptance parameters for roundabouts: a systematic
further problems that practioners and transporta- review. European Transport Research Review March
tion engineers, which use microsimulation for real 2016, 8:2.
world case studies in the professional sphere, are Hagring, O. 1998. A further generalization of Tanner’s
formula. Transportation Research Part B: Methodo-
often called to solve. For these future developments, logical 32(6): 423–429.
more consideration is needed in order to specify the Mathew, T.V. & Radhakrishnan, P. 2010. Calibration of
calibration procedure of microscopic simulation microsimulation models for nonlane-based hetero-
models through automated processes which incor- geneous traffic at signalized intersections. Journal of
porate the latest optimization techniques. Urban Planning and Development 136 (1): 59–66.
MathWorksR, 2016. Online documentation on GA,
URL: https://www.mathworks.com/help/gads/geneti-
REFERENCES calgorithm.html
Mauro, R. 2010. Calculation of Roundabouts. Berlin Hei-
Ayyub, B.M. 2011. Vulnerability, Uncertainty, and Risk: delberg: Springer.
Analysis, Modeling, and Management. Reston, Vir- Menneni, S., Sun, C., & Vortisch, P. 2008. Microsimula-
ginia: American Society of Civil Engineers (ASCE). tion calibration using speed-flow relationships. Trans-
Barceló, J. 2010. Fundamentals of Traffic Simulation. portation Research Record 2088(1): 1–9.
Londono: Springer, 442 p. Nayeem, M.A., Rahman, Md.K. & Rahman M.S. 2014.
Borenstein, M., Hedges L.V., Higgins, J.P.T. & Rothstein, Transit network design by genetic algorithm with elit-
H.R. 2009. Introduction to Meta-analysis. Chichester, ism. Transportation Research Part C: Emerging Tech-
UK: John Wiley & Sons, Ltd. nologies 46: 30–45.
Chiappone, S., Giuffrè, O., Granà, A., Mauro, R. & Sfer- Norme funzionali e geometriche per la costruzione delle
lazza, A. 2016. Traffic simulation models calibration intersezioni stradali [Functional and geometric stand-
using speed-density relationship: An automated pro- ards for the construction of road intersections]. 2006.
cedure based on genetic algorithm. Expert Systems Ministero Infrastrutture e Trasporti [Minister of
with Applications 44: 147–155. Infrastructure and Transport], in Italian.
Cowan, R.J. 1975. Useful Headway Models. Transporta- Papathanasopoulou, V., Markou, I. & Antoniou C. 2016.
tion Research 9(6): 331–375. Online calibration for microscopic traffic simulation
Davis, L.D. 1991. Handbook of genetic algorithms, 1st and dynamic multi-step prediction of traffic speed.
edition. New York: Van Nostrand Reinhold. Transportation Research Part C: Emerging Technolo-
Dowling, R., Skabardonis, A. & Vassili, A. 2004. Traf- gies 68: 144–159.
fic Analysis Toolbox Volume III: Guidelines for Apply- Rodegerdts, L. et al. 2010. Roundabouts: An Informa-
ing Traffic Microsimulation Software. Report No. tional Guide, 2nd edition. NCHRP REPORT 672.
FHWAHRT-04−040. Washington DC: Federal High- Washington DC: Federal Highway Administration-
way Administration-United States Department of United States Department of Transportation.
Transportation. Rubio-Martín, J.L., Jurado-Piña, R., Pardillo-Mayora,
Elvik, R. 2011. Publication bias and time-trend bias in J.M. 2015. Heuristic procedure for the optimization
meta-analysis of bicycle helmet efficacy: a re-analysis of speed consistency in the geometric design of single-
of Attewell, Glase and McFadden, 2001. Accident lane roundabouts. Canadian Journal of Civil Engineer-
Analysis and Prevention 43(3): 1245–1251. ing 42(1): 13–21.
Essa, M. & Sayed T. 2016. A comparison between PAR- Vasconcelos, L., Seco, Bastos Silva, A. 2013. Safety analy-
AMICS and VISSIM in estimating automated field- sis of turbo-roundabouts using the SSAM technique.
measured traffic conflicts at signalized intersections. CITTA 6th Annual Conference on Planning Research:
Journal of advanced transportation 50: 897–917. 1–15.

640

TIS2017_Book.indb 640 2/8/2017 8:08:17 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

More accessible bus stops: Results from the 3iBS research project

M.V. Corazza & A. Musso

Sapienza University of Rome, Rome, Italy

M.A. Karlsson
Chalmers University of Technology, Gothenburg, Sweden

ABSTRACT: Although often perceived as poorly attractive and unreliable, buses are central in the
development of sustainable mobility options in urban areas. Therefore, 3iBS - the Intelligent, Innovative,
Integrated Bus Systems (a research project funded by the European Commission) promoted the research
on a new generation of vehicles and facilities. Specific emphasis was placed on the design of bus stops as
crucial elements to improve the quality and accessibility for all of bus services. The paper describes the
conflicts to solve and stresses the need of univocal directions to design inclusive bus facilities; specific
recommendations for implementing innovations in this field are eventually provided.

1 INTRODUCTION stressed the importance of ease through the con-

cept of immediacy, since accessibility is defined
Buses are central in the everyday life of many as “the general proximity in terms of time of all
citizens. Nevertheless they are often perceived points … to a given kind of activity or facility”
as unappealing (Stradling et al 2007, Beirao & (1995:49); he also identified three sub-dimensions
Sarsfield Cabral 2007, NRTBI 2009, Dobbie et al. of accessibility: diversity (of things to be accessed),
2010). Therefore, 3iBS - the Intelligent, Innovative, equity (of access for the different social groups)
Integrated Bus Systems, a research project funded and control (over the access systems) (1981). For
by the European Commission (EC), moved from Grava (2003), accessibility is also a measure of the
this assumption to propose solutions to revamp quality and operational effectiveness of a commu-
this mode of transport across Europe. More spe- nity. Therefore, the acknowledgment that transit
cifically, the project aimed to promote the research has to serve all leads to the concept of fully acces-
on bus systems with a special focus on a number sible transit (Vuchic 2005). Hence, it is not uncom-
of key topics and, among these, the improvement mon to assess the accessibility of a given transit
of accessibility for all. The results evidenced how facility according to the level of ease by which one
accessibility has become a mainstream issue on can reach it, typically by location-based criteria,
the agendas of transit stakeholders. Therefore, or more generally short distance. For example,
this paper describes the main outcomes in this Banister observes that transport’s “primary aim is
field, with a special focus on problems related to to maintain a high level of accessibility with trip
bus infrastructures, resulting from a specific activ- lengths being as short as possible” (2005:10). This
ity dedicated to issue guidelines for accessibility shifts the focus on how to measure accessibility in
and safety concepts in the future design of bus general (Handy & Niemeier 1997, Geurs & van
systems. Eck 2001, Church & Marston 2003). Accessibility
can be also associated with terms such as “within
walking distance” or “walkable”. Linking transit
2 DEFINING ACCESSIBILITY accessibility with walking can be, on the one hand,
AND DISABILITY a way to emphasize the (supposed) modest physical
efforts required to reach a given destination, and
Accessibility can be broadly defined as the ability to on the other a way of underestimating the efforts
travel between different activities (Vuchic 1999), it of those who are not able to autonomously walk.
can be linked to convenience, as the ability to reach This may also imply that spatial accessibility is
goods, services, destinations (Littman 2014), or the equal to temporal accessibility, but efforts spent to
ease to do it (Sinha & Labi 2007). Engwicht defined reach a given destination may differ according to
accessibility as “the ease with which exchange the travellers’ physical and perceptive conditions.
opportunities can be accessed” (1999:167). Lynch For examples, the “walking distance” to a bus stop

641

TIS2017_Book.indb 641 2/8/2017 8:08:18 PM

 can be less challenging for an elderly person than basic travel information (i.e timetables, variable
the time spent standing in line, waiting for board- messages, announcements, etc.) a demanding task
ing, with no possibility to rest. for dyslexic travellers who develop creative solu-
Accessibility can be also linked to perception. tions such as the use of satellite navigation systems
Hansen (1959) pioneered this concept by defin- (Lamont 2010).
ing accessibility as the “intensity of possibility Disability is therefore difficult to define from
of interaction”. Ropoport observed that: “people a design perspective. However, Carmona et al.
react to environments in term of the meanings observed that there are two main models of disa-
the environments have for them” (1990:13). A bus bility: the “medical model” and the “social model”.
stop and its surroundings can become the weakest The former defines disability in terms of medical
link in the journey chain if they provide users with conditions, and the impairment factors are placed
negative meanings (too distant, uncomfortable, solely on the individual; the latter concerns the bar-
unsafe, etc.). riers “imposed by a disabling society/environment
The approach within the 3iBS project was based unable to make adjustments: this rather than per-
on the notion that “a Public Transport (PT) system sonal impairment, is the disability factor. In this
is accessible to people who are able to use it”. The model, people have impairments, environments
emphasis is placed on the ability of a given PT sys- are disabling” (2003:127). Therefore, shifting the
tem to provide appropriate meanings or, in other focus from single individuals with disabilities to
words, equal conditions of exchange by meeting places and the need to shape them according to
the requirements of all, rather than on the specific everybody’s requirements, represents, in terms of
abilities of the users. If a transit system is not, or design, a more productive basis to start from.
poorly, accessible this means that a number of fac-
tors prevent passengers from using it, not the pas-
sengers themselves. Preventing factors may result 3 THE REGULATORY FRAMEWORK
from inappropriate design criteria for vehicles, TO PROMOTE ACCESSIBILITY FOR ALL
infrastructures and communications. They occur
because they fail to meet users’ physical and/or According to all of the above, accessibility and
cognitive requirements. However, users’ needs are disability seem to be concepts under continuous
many and even indistinct, and so also the definition development, and so are the legal and regula-
of disability. Actually, the line between a disabled tory frameworks behind and the related financial
individual and a non-disabled one is rather thin, supports.
since a person performing everyday activities (walk-
ing with prams, suitcases, shopping trolleys, etc.)
3.1 The European strategy and regulations
requires the same space as does a wheelchair
user. At the same time, other sensory or cognitive Since the 1980s concepts such as accessibility and
impairments do not entail such kind of spatial req- safety have been progressively included in many
uisites. One might argue that the so-called “able- regulations in this field, both at supranational and
bodied” person does not even exist. national levels, although with different approaches.
Therefore, physical, sensory, cognitive or devel- The European institutions have continuosly devel-
opmental impairments give rise to a vast range of oped strategies to increase inclusiveness and equity
requirements, in some cases contrasting, and solu- according to general principles; currently, the most
tions to meet them all can only result in a kind of important course of actions are the Disability
“relative optimum” for a majority of passengers. Action Plan 2006–2015 (CoE 2006) and Euro-
Ramps are a typical example of such a contrast: pean Disability Strategy 2010–2020 (European
designed for wheelchair users, they can be chal- Commission 2010), the EU Disability Action Plan
lenging for blind users, who cannot perceive the 2003–2010 (CSES 2009), all in line with the 2006
length, unless trained or helped, and therefore United Nations Convention on the Rights of Per-
prefer grades, easily detectable by walking sticks sons with Disabilities. Strategic sectors contem-
(Lauria 1994). plated in these documents such as medical care,
The design of inclusive solutions for transport education, employment, and social safeguard are
infrastructure is based on the knowledge of users’ considered key areas to promote participation
needs, but given the variety of disabilities, some of and equality of rights for people with disabilities.
them are still neglected. An example is dyslexia: Accessibility is, thus, meant to ensure full fruition
although 8% of the world population is affected of goods, services and assistive devices, facilities,
by this impairment, dyslexia is far from being con- and communication technologies, in the same way
sidered in the provision of information in public as for “able-bodied”. The importance of having
areas. Difficulties in processing numbers, read- full access to the built environment by applying
ing, and/or spelling, make the comprehension of Universal Design criteria is also stressed, as well

642

TIS2017_Book.indb 642 2/8/2017 8:08:19 PM

 as the need to provide fully accessible transport “adaptation to special needs” in the attempt to
services. These comprehensive directions are pri- increase the passive safety of users with disabili-
marily meant to steer national regulations towards ties, whereas it is a concept to shape the built envi-
a common vision and therefore can be considered ronment according to the needs of all and with the
expressions of general principles in safeguarding aim to improve their active and autonomous par-
people with disabilities. ticipation in social life. Adaptive design is, there-
More specific design criteria can be found at fore, justifiable and/or admissible in historic or
national levels; as already stated, from the 1980s’ natural environments only (cultural heritage sites,
onwards, European Countries have progressively green areas), where physical obstacles are irremov-
enforced more and more specific design criteria able, thus restricting the access options.
to create barrier-free environments, with a special Needless to say, to overcome these problems,
focus on dwellings. Initially, design criteria con- the criteria of Design for All and AT technologies
cerned mainly dimensions and layouts to accom- must be applied throughout the standardisation
modate physically challenged users, and especially process and mandatorily included in the design of
those using wheelchairs. The philosophy behind new facilities and in the rehabilitation and retrofit-
such an approach was strongly affected by the ting of those already operational, according to the
General Product Safety Directive n. 92/59/EEC, lesson learnt from the Americans with Disabilities
according to which products available for the mar- Act (ADA). ADA enforces a comprehensive regu-
ket should be safe, so as to prevent risks for the latory system, complemented by a vast range of
end- users: but even though this principle is imper- technical specifications and standardized solutions
ative, its misinterpretation can lead to design con- including detailed norms for transportation facili-
cepts merely based on passive safety rather than ties, as it considers standards as a mean to over-
active involvement of people with disabilities. The come discrimination and exclusion.
repetition of this approach, i.e. conceiving barrier-
free environments mostly for wheelchair users
3.2 Political support
resulted in a number of problems.
First, it cast a shadow on the requirements of the The regulatory framework described earlier is
remaining part of the population with disabilities, affected by the political support. Unlike the US,
which still affects the way in which many transport in many European Countries minorities started
facilities are designed or upgraded. It is undeniable rather recently to lobby against exclusions, mostly
that especially for what concerns cognitive disabili- thanks to the commitment of associations such
ties, accessibility is far from being fully addressed. as the European Disability Forum, the European
As a result, although ramps and lifts abound now- Blind Union, or the European Union of the Deaf,
adays, provisions for visually or hearing-impaired just to mention a few of the most influential. Liais-
still do not, and even less do the aids for “less con- ing with the European Union institutions is one
sidered” disabilities. of the tasks of these bodies, in order to keep the
Secondly, designing barrier-free environments awareness high at political level.
mostly for wheelchair users, being mainly focused Along with such poltical commitment, at the
on barriers which can be overcome by support beginning of the 2000s, the European istitu-
devices or equipment (typically, walking aids), tions promoted a number of studies in this field.
affected the standardization programs towards the More specifically, the Conférence Européenne des
developments of specifications of single components Ministres des Transports (CEMT) and the EU
rather than to develop common process (Criteria developed several comparative analyses: on dif-
and procedures) to apply univocally and system- ferences in regulations enforced (Législation pour
atically across Europe. Currently, standardization améliorer les moyens d’accès. rapport CEMT/CS/
covers a wide range of Assistive Technology (AT) TPH(2000)7/REV1), on the best practice (Amél-
products; for instance, the European Commit- iorer l’accessibilité des transports—Rapport
tee for Standardization (CEN) issued specifica- CEMT/CM(2004)27), and on the the quality of
tions on: technical aids for disabled persons (CEN transit services in 49 cities (Citizens Network
2007); Man-Machine Interface for users with Benchmarking Initiative—2004), so to have a
special needs (CEN 2012); and medical informat- sound knowledge basis and raise the awareness
ics (CEN 2010), but those which can evolve into a on the problems faced by people with disabilties
standardization of inclusive solutions for the built (Thomas et al. 2007). It is rather difficult to assess
environment are still in the making (as CEN/CLC/ to what degree the needs and expectations of the
JWG5 and 6 respectively on Design for All, and people with disabilities have been met over the
Accessibility in the Built Environment). years, but since the proposal for a European Acces-
Eventually, this approach foster again and sibility Act (EAA) is still to be published by the
again the misconception that inclusive design is an European Commission, more can be achieved in

643

TIS2017_Book.indb 643 2/8/2017 8:08:19 PM

 this field. The lesson learnt from ADA on stand- layouts, materials, textures, colours, etc, according
ardization as a mean to improve social inclusion to their own funding availability. Again, common
clearly shows the importance of enforcing a tool universal design standard criteria and accessibil-
as the EAA and the need for a paradigm shift: it ity master plans, when mandatory across Europe,
is necessary not only to have common mandatory would facilitate rehabilitation processes compelling
inclusive design standards but also to enforce com- local operators to adopt common solutions. Exam-
pulsory accessibility master plans (as for land use) ples of common funding encompassing more local
to have a common quality of barrier-free environ- areas are few but an interesting one is represented
ments across Europe. by the German Lower Saxony administration
However, at a European level, an indirect, but that, in 2013, launched a 400,000 Euro program to
strong, political support is represented by the renovate bus stops in different cities, planning to
many research projects that the European Com- upgrade 50 stops yearly (Vogt 2013).
mission have funded over the years to improve the Different players and areas also make difficult
quality of life of people with disabilities, in line to prioritize plan interventions and above all raise
with the CEMT approach. The outcomes of these appropriate funds. Usually, funding opportunities
projects constitute a comprehensive set of guide- for upgrading transit facilities come from public,
lines, recommendations and best practice examples local (municipal) investment programs, seldom
in the fields of safety and accessibility, but are far including some earmarked funds for improving
from being applied as consolidated practice across accessibility and even more seldom from private
Europe, which again leads to contemplate the need financing. One exception is the German Pro-
for a tool as EAA and its implementation Europe- gramme 233 and 234—Barrierearme Stadt which
wide. enables banks to fund investments for barrier-free
infrastructure, and especially transit facilities. This
is a good example of the successful involvement
3.3 Funding to improve accessibility
of private bodies in a public accessibility improve-
Aside from those granted by the European ment processes, thus supplementing public fund-
Union, specific funds are usually linked to retro- ing with private financing.
fit programs to refurbish major facilities, typically
railways stations, and to comply with what the
national regulations mandate. A good example is 4 THE BUS STOP: A MULTI-
the “Access for All” station improvement scheme, REQUIREMENT ENVIRONMENT
launched recently in the UK. According to the
national Railways Act 2005, it allowed more than “Bus stops are often dreary because they are set
80 facilities to be rehabilitated and an additional down independently, with very little thought given
fund was announced to have more stations trans- to the experience of waiting there, to the relation-
formed by 2019. It is also estimated that for every ships between the bus stop and its surroundings”
1 GBP spent in this rehabilitation program 2.92 (Alexander 1977:452). Not much has changed
GBP is actually gained from additional revenues since this remark. Therefore, the emphasis placed
coming from the enlarged group of passengers within 3iBS on bus stops, as elements to redesign
(DfT 2012). to meet the requirements of all is two-pronged.
Other major railways operators in Europe have On the one hand, according to the definitions
adopted similar initiatives, often issuing their of accessibility and disability reported earlier, it
own guidelines (for example RFI 2012) to reno- stresses the crucial role played by the environment
vate buildings and rolling stock, and solve typical when addressing problems related to access and
problems, such as the access to platforms or the impairment. On the other hand, it highlights the
refurbishment of rest and waiting areas. Neverthe- need for constant, extra funding and strong politi-
less, while this might result in improved accessibil- cal support to increase momentum on these issues.
ity of a majority of larger urban stations, mainly The former calls for a common vision shared by
for physically-challenged and visually/hearing planners, designers, operators and the largest part
impaired passengers, the access continuity with the as possible of passengers; the latter requires a com-
surrounding environment is usually not ensured, mon regulatory framework to develop univocal
or of poor quality. design criteria and standards across Europe.
In the case of railways, one operator in charge This is not an easy task, since a bus stop
of managing one national network is a happen- environment concentrates multiple functions,
stance that facilitates this kind of rehabilitation requirements, performance levels and planning
process. But different, local, smaller operators who and design criteria. In fact, the problem is multi-
manage bus fleets and facilities in just one area faceted. Since a majority of the different phases
are more likely to adopt their own solutions for of a journey may take place at the bus stop

644

TIS2017_Book.indb 644 2/8/2017 8:08:19 PM

 (Figure 1), the design of this facility must provide 5.1 Multiplicity of actors
different solutions to meet a number of common
In terms of management of bus stops, a number
requirements and functions (walkability; com-
of different situations may occur: a) one operator
fort while waiting or boarding/alighting; ability
managing the whole transit network and facilities;
to autonomously perform other travel functions
b) several operators, each managing a part of the
as purchasing tickets, getting information, rest-
local transit network; c) several operators, each
ing; readability and comprehension of signs
managing either the transit network or the ground
and directions; feeling of inclusiveness, etc.); the
facilities.
travel experience may occur under different cir-
For cases b) and c), the earlier mentioned lack
cumstances (good vs adverse weather, daytime
of compulsory guidelines result in a lack of com-
vs nighttime, peak vs off-peak times, frequent
mon vision which becomes a hindrance in the
vs novice travellers, solo travelling vs accompa-
development of common design criteria for fully
nied, beginning vs end-leg of a journey, etc.), and
accessible stops for the whole urban area. Moreo-
take place in different environments (outdoor vs
ver, when bus stops are managed by a third party
indoor, secluded vs frequented, friendly vs hos-
such as advertising companies, matching commer-
tile—perceived, etc.).
cial functions with the needs of all may become dif-
Therefore, it is relevant to conceive the bus
ficult. For instance, excessive advertisement on the
stop as a system in which the main activity, i.e.
shelter walls may result in reduced space for infor-
boarding/alighting, is strictly interrelated with
mation for visually-impaired users (who need areas
additional or associated activities occurring in the
to accommodate large prints/tactile information),
surroundings; each activity has its own meaning
and for the hearing-impaired ones (who need full
for the users, which makes them perceive the bus
visibility of the approaching buses). Poor availabil-
stop as a specific environment and react differ-
ity of transparent surfaces may induce a percep-
ently to it.
tion of reduced security since the principle of “see
and being seen” is strongly diminished, especially
when inside the shelter.
5 FROM CONCEPTS TO PROBLEMS
The problem is not just a matter of conflicts
between accessibility and commercial purposes.
Once users’ needs are analysed according to good
Other factors may contribute to inappropriateness.
practice and the vast scientific literature, and bus
For instance, O’Neill & O’Mahony (2005) reported
stops are designed consequently, the requirements
how, according to the Disabled Persons Transport
for accessibility and appropriate meaning should
Advisory Committee in the United Kingdom, dis-
be met. But, even when urban transit operators
abled people were more likely to rely on taxi serv-
issue their own inclusive design guidelines for
ices. However, in the case study analyzed, 42% of
the network facilities, full accessibility across the
the people with disabilities were happy with buses.
whole transit network is a goal far from being
Although a useful travel option, taxis may not be
accomplished. Unlike the railways sector, plan-
the best way towards inclusiveness, as it: a) misses
ning accessibility for bus systems results in differ-
the opportunity to make conventional transit and
ent outcomes due to a series of factors, which can
its facilities more accessible and b) increases some-
be synthesized into: too many managing actors, too
how the dependency on private travels. As a result,
many facilities and too many end-users in different
if it is assumed at political level that disabled people
environments.
are more likely to be satisfied with taxis, no atten-
tion will be paid to make bus stops more accessible
for all.
An additional factor is the dominating attention
paid to products rather than processes. Specific
products for specific disabilities improve con-
stantly, but Carmona et al. stress that: “Inclusive
design is about attitudes and processes as much as
about products” (2003:127). It is indisputable, then,
that even the most advanced products (as commer-
cial shelters), if not included in a process based on
the integration of the needs of all, do neither “pro-
duce” inclusiveness, nor contribute to create less
disabling environments. This may affect decisions
concerning other elements of the travel chain or
the decision process. For example, the conversion
Figure 1. Main activities of a bus journey.

645

TIS2017_Book.indb 645 2/8/2017 8:08:19 PM

 of obsolete bus fleets into more environmentally-
friendly ones, with low-floor vehicles and on-
board equipment to accommodate the largest part
of passengers with special needs may be thwarted
if the environment surrounding the bus stop is not
accessible, not to mention that the related invest-
ment for such eco-conversion can, then, appear as
an effort for inadequate return. Therefore system
approaches might not, per se, result in integrated
solutions.

Figure 2. A bus stop environment.

5.2 Limits in the feasibility
Among the different management situations, in the
case of situation a), i.e. one operator for the whole to accommodate a wheelchair user and a visually
network, the task of a total rehabilitation might be impaired one).
too demanding. Single operators face the problem Still this approach becomes more difficult to
to upgrade too many facilities, i.e. a huge number adopt in historic centers, where appropriate loca-
of stops with different functions and size: from tion can be demanding, due to narrow sidewalks.
simple bus markers to shelters, hubs, and termi- As a result layouts such that of Figure 2 are unfea-
nals. The approach in this case is rather common, sible. In this case the built environment becomes
since the upgrading process is linked to capacity: disabling and the conflict has virtually no resolu-
priority is given to larger facilities (hubs, terminals) tion: adaptive planning and design remain the only
for which a wide range of solutions is deployed, resources.
to remove what are considered the most common Likewise, allocating the proper space require-
barriers (grades and gaps); then a minor provision ment for each user may be challenging and gener-
of equipment is assigned to shelters (usually to ally regulations for street furniture and for the use
improve ease of use such as armrests at seats, real of sidewalks (typically Highway Codes) increase
time information displays and occasionally tactile the amount of restrictions. Specifications for street
aids), whereas stops with simple bus markers are furniture usually associate size of objects/obstruc-
left not upgraded. However, basic adaptations of tions (benches, bollards, vendors, etc.) to a given
bus markers are not difficult (Lauria 1994; Rickert amount of pavement areas (GMUAM 2001, Win-
2007). Along with different ways of management nipeg Transit 2006), but seldom include LOS and
and vast amounts of facilities to deal with, a third areas to accommodate users with special needs.
element of conflict is the different end-users that The approach followed by Highway Codes and
each stop facility attracts and the difference in regulation on street functions is quite the reverse:
the surrounding environments where it is located. LOS and service areas are provided, but for a very
Guidelines for bus stops design are available in restricted group of street furniture. Usually, none
grey and scientific literature (some of which are includes systematic specifications for the areas
referenced in this paper) all providing well-known required to accommodate users with special needs,
criteria for locating (usually commercial) shelters or recommends functional layouts or adaptations
and their basic equipment in indistinct urban criteria to the type/quality of the environment. The
environments. lesson learnt is that there is no general “recipe” to
This approach may apply to the construction design a bus stop or to rehabilitate this kind of
of a new infrastructure of urban relevance, such facility when already existing in a sensitive context
as a light railway, with dedicated platforms, which (historic or consolidated areas) to improve acces-
entails a total rehabilitation of the streetscape, sibility; planning and design must equally consider
including additional areas to accommodate pedes- requirements of all and constraints from the envi-
trians. Not the same can be obtained when locating ronment, in the awareness that only that kind of
a basic bus stop (typically a shelter) on the side- “relative optimum” can be obtained.
walk, where passenger flows accessing/egressing
the waiting area cannot be separated by the main
pedestrian flows on the sidewalk and the space to 6 RECOMMENDATIONS FOR MORE
accommodate passengers with disabilities is scarce ACCESSIBLE BUS STOPS
(as in Figure 2, where a layout designed for all is
described, including typical pedestrian movements Conflicts and limits reported can be of no easy
around a bus stop, according to different Level resolution but certainly can goad on to progress
of Services—LOS, and some basic requirements towards more inclusive adaptation of the built

646

TIS2017_Book.indb 646 2/8/2017 8:08:20 PM

 environment. Therefore, within 3iBS, Key Direc- Such a network-scale plan calls for a thorough
tions (KD) for the implementation of innovations assessment of the available resources (KD j) to refur-
regarding accessibility for future bus systems have bish facilities according to the enforcing standards,
been provided (Karlsson 2014). These KDs are: along with the application of a specific business
plan targeting changes in perception of the service
a. Process approach for implementing innovations;
by the enlarged community of users (thus includ-
b. Initial assessment of the local level of
ing those previously excluded). All of the above
accessibility;
is possible only if secure and long-term financing
c. Long-term vision;
is available (KD i). Funding is crucial: due to the
d. Strong leadership;
multiplicity of managing actors and involved third
e. Organizational culture;
parties, contributions from all of them are required
f. Political and public support;
and fundraising may involve non-conventional
g. Consideration of regulatory aspects;
bodies such as sponsors, donors, and communities,
h. Possibilities of innovation procurement;
etc. The possibility to use innovation procurement
i. Secure and long-term financing;
(KD h) to support the introduction of innovative
j. Sufficient and relevant resources;
devices (e.g. AT products for advanced communi-
k. Full analysis of existing solutions;
cation systems at bus stops) must be contemplated.
l. Stakeholder analysis;
Needless to say, plans and funding rely on long-
m. Cooperation and collaboration;
term visions (KD c) and strong and durable politi-
n. User studies;
cal support (KD f), at supranational and national
o. Evaluation;
levels for what concerns the issue of standardized
p. Market of the innovation
solutions and innovations, and at local levels to
The translation of such recommendations into launch such plans.
more accessible future bus stops involves a new
vision based on the in-depth comprehension of
6.2 Inclusion of less-studied disabilities
what makes an environment “disabling”, and where
innovation of system and products are aimed at All of the above relies on an in-depth knowledge
increasing inclusion, according to three main lines of the requirements of the community to serve.
of action: 1) Assessment of resources to increase Traditionally, studies on disabilities have focused
accessibility; 2) Inclusion of less-studied disabilities on walking, hearing and visually-impaired users
and 3) Integrated proactive design for all. respectively and as a result, transit facilities solu-
tions have been usually (re)designed to meet the
requirements of each category, separately. How-
6.1 Assessment of resources to increase
ever, transit is used by passengers with all kinds
accessibility
of health conditions and personal travel experi-
The previously reported analyses of current politi- ences and often their impairments do not relate
cal support and availability of funding evidenced to physical barriers as grades or gaps. Meeting
the inadequacy of resources to progress towards requirements associated with mental or intellec-
more inclusive transit environments. Required tual impairments, specific conditions due to age
resources must be regulatory and economic. For or personal perception helps reduce vulnerability,
the former, specific universal design criteria for disadvantage and exclusion. More specifically, for
transit facilities must be standardized at EU-level the design of the bus stops this calls for continu-
and enforced at national level. Consideration of ous analyses of the demand to serve (KD n), both
innovative legal and regulatory aspects to support ex ante and ex post (KD o), including demands of
the standardization process and its implementation users with “conventional” as well as “less known”
(KD g) and especially, innovative, specific techni- impairments. Prior to the upgrading process, stud-
cal solutions for the design of new bus stops will ies of the community should be aimed at assessing
help transit companies to initially assess the status how the requirements are met and the compatibil-
of the facilities (KD b) and solutions (KD k), and ity of the proposed solutions among the different
plan the upgrading process, according to a proc- categories of passengers. Conflicts may arise and
ess approach (KD a). The analysis of the existing the more the requirements to meet, the more likely
solutions is of the utmost importance, as it indi- this is going to happen. Moreover, ex ante assess-
cates the compatibility between the status quo and ments can highlight specific incompatibilities (if
the enforcing standards, and therefore the entity any) between the enforcing standards and specific
of the upgrading process. This entails a systematic disabilities to deal with (footnote: standards are
enforcement of the previously mentioned stand- based on the uniformity of technical criteria and
ards across the whole bus stop network, starting the possibility of missing some specific require-
from the very basic facilities. ment must be thus contemplated). Nevertheless,

647

TIS2017_Book.indb 647 2/8/2017 8:08:20 PM

 ex ante analyses of users’ needs should result in compulsory provide accessible infrastructures,
“snapshots” of the bus stop requirements, to steer vehicles, and communication systems to all passen-
design of facilities towards a performance spe- gers and control possible conflicts. This requires a
cifically targeted to increase accessibility. Dealing strong commitment from the transit companies
with enlarged communities of passengers (includ- (KD d) to comply. Users must become aware that
ing those affected by less “common” impairments) accessibility is a right and informed of the avail-
includes not only the usual customer care process. ability of most suitable travel options. Marketing
Long-term observation studies on the fruition of of the innovative solutions (KD p) should be con-
the upgraded facilities by the different groups of sidered to disseminate the information to the dif-
users are also required to assess whether the whole ferent groups of passengers, according to different
built environment has gradually become less “disa- communication tools. Marketing can help increase
bling”. Again, funding and political commitment awareness and lessons learned from environmen-
are essential to ensure appropriate methodologies tal care can be applied to accessibility: mutatis
for the analyses of the users, and especially for mutandis, transit components, when compliant
those with less well-known disabilities. It is also with standards and operating within accessibil-
essential that the panel of stakeholders involved ity plans, may be, in the near future, labelled or
in the analyses and in the assessment of resources “tagged” as “accessibility-mindful”, thus increas-
includes representatives of all categories of disa- ing the attractiveness of public transport.
bled users and, further, that cooperation is ensured
(KD m). In this way the risk of conflicts may be
mitigated and the interests of all equally consid- 7 CONCLUDING REMARKS
ered (KD l). A strong supporting organizational
culture is obviously required (KD e). Alexander described the epitome of bus stops as:
“easy to recognize, and pleasant, with enough
activity around them to make people comfortable”
6.3 Integrated proactive design for all
(1977:453). To achieve this, there are still many ave-
To improve bus stop accessibility, practice has so nues to explore, and accessibility is the main one,
far relied on the provision either of ex novo facili- especially in political, regulatory and operational
ties, with the limits earlier mentioned (inclusion of terms. Issues such the EAA, the inclusion of less
requirements from “conventional” disabilities only, well known disabilities, the compulsory adoption
capacity as a priority criterion, conflict of inter- of common Universal Design criteria when plan-
ests unsolved, etc.) or of existing facilities, with ning and designing transit facilities, the mandate
“tacked-on” adjustments to make them more bar- to enforce accessibility plans for transit companies,
rier-free, more or less with the same limits and the and the possibility to award fully-compliant prod-
additional restriction that in some case the built ucts and components as “accessibility-mindful”
environment (for example historic centers) could are therefore necessary steps to make urban travels
thwart such rehabilitation attempts due to lack of really for all.
proper space.
These situations are far from acceptable as, in
both cases, the design approach is reactive: com- REFERENCES
mercial components (shelters, seats, displays, etc.),
designed for “able-bodied” users, are adjusted to Alexander, C. 1977. A pattern language. Oxford Univer-
meet the requirements of the considered majority sity Press, New York.
of impaired users. Subsequent additions of ramps Banister, D. 2005. Unsustainable Transport. City trans-
port in the new century. Routledge, London.
and tactile tiles represent “adjustment for all” vs Beirao, G. & Sarsfield Cabral, J.A. 2007. Understanding
“design for all”. To have the latter winning the attitudes towards public transport and private car:
action is two-pronged. On the one hand, the design A qualitative study. Transport Policy, Vol. 14, Issue 4,
approach must become proactive, thus it must: July, pp. 478–489.
i) include requirements from less well-known dis- Carmona, M., et al. 2003. Public places Urban Spaces.
abilities to develop more accurate design criteria, Architectural Press, Oxford.
and ii) rely on standardized universal design crite- CEN 2007. Business Plan—CEN/TC 293. Assistive prod-
ria for transit facilities, resulting in the production ucts for persons with disability. http://standards.cen.
of innovative systems and components enforced eu/BP/6274.pdf
CEN 2010. Business Plan—CEN/TC 251. Health infor-
by sound and durable regulatory supports and by matics. http://standards.cen.eu/BP/6232.pdf
the compulsory adoption of accessibility plans for CEN 2012. Business Plan—CEN/TC 224. Personal
transit. On the other hand, once again the politi- identification, electronic signature and cards and their
cal support must become unrestricted and com- related systems and operations. http://standards.cen.
pelling. Transit companies must be required to eu/BP/6205.pdf

648

TIS2017_Book.indb 648 2/8/2017 8:08:21 PM

 Church R. & Marston, J. 2003. Measuring Accessibility Lauria, A. 1994. Pedonalità urbana. Maggioli, Rimini.
for People with a Disability. Geographical Analysis, Littman, T. 2014. Accessibility, in: TDM Encyclopedia.
Vol. 35, n. 1, pp. 83–96. www.vtpi.org/tdm/tdm84.htm
CoE—Council of Europe (2006). Disability Action Plan Karlsson, M.A., (ed.). 2014. 3iBS—Del. 1.1.2. Guidelines
2006 2015. www.preventionweb.net/english/profes- for the introduction of) accessibility and safety concepts
sional/policies/v.php?id = 35144 in future design of bus systems (restricted document).
CSES 2009. Mid-term Evaluation of the European Lynch, K. 1995. City Sense and City Design: MIT Press,
Action Plan 2003–2010 on Equal Opportunities for Cambridge.
People with Disabilities. CSES, Sevenoaks. Lynch, K. 1981 A theory of good city form. MIT Press,
DfT—Department for Transport 2012. High level output Cambridge.
specification 2012 – Railway Act 2005 Statement. www. NRBTI 2009. Quantifying the Importance of Image and
gov.uk/government/publications/high-level-output- Perception to Bus Rapid Transit. www.nbrti.org
specification-2012 O’Neill, Y. & O’Mahony, M. 2005. Travel Behavior and
Dobbie F. et al. 2010. Understanding Why Some People Transportation Needs of People with Disabilities.
Do Not Use Buses. www.scotland.gov.uk/Resource/ Transportation Research Record, No. 1924, pp. 1–8.
Doc/310263/0097941.pdf Rapoport, A. 1982. The meaning of the built environment.
Engwicht, D. 1992. Reclaiming our cities and town, New The University of Arizona Press, Tucson.
Society Publishers, Philadelphia. Sinha, K.C. & Labi, S. 2007. Transportation Decision
European Commission 2010. European Disability Strategy Making. Wiley and Sons, Hoboken, NJ.
2010–2020: A Renewed Commitment to a Barrier-Free Stradling, S. et al. 2007. Passenger perceptions and the
Europe. COM/2010/0636 final. http://eur-lex.europa. ideal urban bus journey experience. Transport Policy,
eu/legal-content/en/NOT/?uri=CELEX:52010DC0636 Vol. 14, n. 4, pp. 283–292.
Geurs, K.T. & van Eck, J.R., 2001. Accessibility meas- Thomas, R., et al. 2007. L’accessibilite des reseaux de
ures: review and applications. RIVM report 408505 transport en commun en Europe. ÉNS d’Architecture,
006, National Institute of Public Health and the Envi- Grenoble.
ronment, Bilthoven. RFI 2012. Linee guida per la progettazione—Accessibilità
GMUAM—Gerencia Municipal de Urbanismo, nelle stazioni. http://www.webstrade.it/pescara/peba/
Ayuntamiento de Madrid 2001. Instruccion de Via RFI_2011_LG_accessibilita-stazioni.pdf
Publica. Nilo, Madrid. Rickert, T. 2007. Bus rapid transit accessibility guidelines,
Grava, S. 2003. Urban Transportation Systems. Choices World Bank, Washington, D.C.
for Communities. McGraw-Hill, New York. Vogt, S. 2013. Region baut Haltestellen behindertengerecht
Handy, S.L. & Niemeier, D.A., 1997. Measuring acces- um. Hannoversche Allgemeine, 14.2.2013.
sibility: an exploration of issues and alternatives. Envi- Vuchic, V.R. 1999. Transportation for livable cities.
ronment and Planning A, Vol. 29, n. 7, pp. 1175–1194. CUPR, New Brunswick, NJ.
Hanson, W. 1959. How accessibility shapes land use. Vuchic, V.R. 2005. Urban transit. Operations, planning
Journal of the American Institute of Planners, n. 15, and economics. Wiley and Sons, Hoboken, NJ.
pp. 73–76. Winnipeg Transit 2006. Designing for Sustainable Trans-
Lamont, D. 2010. The Effects of Dyslexia upon Personal portation and Transit in Winnipeg. www. winnipegtran-
Travel: Empirical Results of a Qualitative Study. Pro- sit.com/inside-transit/transitreports
ceedings of the 12th WCTR, Lisbon.

649

TIS2017_Book.indb 649 2/8/2017 8:08:21 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Shaping the future of buses in Europe: Outcomes and directions

from the EBSF and EBSF_2 projects

M.V. Corazza & A. Musso

Sapienza University of Rome, Rome, Italy

U. Guida & M. Tozzi

UITP, Brussels, Belgium

ABSTRACT: EBSF—European Bus System of the Future and its follow-up EBSF_2 are two research
projects funded by the European Union and led by UITP with the aim of developing a new generation
of buses across Europe. Within EBSF the accomplished goals were to increase the bus attractiveness by
testing innovative vehicles. Coherently, EBSF_2 is now committed to raise the image of the bus through
solutions for increased efficiency of the system. The environmental concern and the need to save energy
are also behind the majority of both projects’ innovations. These are tested through demonstrators, i.e.
on vehicles operating in several European cities, and performance are independently evaluated. The paper
describes the methodology and the test process adopted in both projects, and reports the results so far
achieved and the new challenges ahead, with the research objective to provide advanced knowledge for
further applications beyond the European projects field.

1 THE CONTEMPORARY SITUATION generations of buses are a very appropriate mode

to meet sustainability constraints in terms of
Passenger cars are still the dominant mode in the energy efficiency, emissions and space occupancy,
everyday mobility of many European urban areas, as well as operational effectiveness, since they can
even though the recent economic developments be more easily adapted to different requirements
have started to orient mobility demand towards of passengers and do not require heavy (re)design
transit, and more specifically toward increasing bus of infrastructures. Nevertheless, the modal switch
ridership. Car-based lifestyles are still fostered by from passenger cars to buses can be challenging
many behavioral patterns and recurring practice, in urban areas where inadequate performance
among these: the assumed personal convenience such as poor regularity, punctuality, comfort,
of this mode, continuously growing urbanization design and obsolete technologies contribute to
and increased tolerance to distance and congestion the general modest attractiveness of this mode,
(the average delay in minutes for one-hour journey if compared to other transit options, especially in
driven in peak periods is 29 min, as monitored by Central, Eastern and Southern Europe as stated
TomTom 2013). Air and noise pollution, space by the European Commission (2013), and gener-
consumption, climate change and environmental ally reported in scientific literature (Stradling et al
concerns are typically associated to the phenom- 2007, Beirao & Sarsfield Cabral 2007, NRTBI
ena above described, and integrated strategies to 2009, Dobbie et al. 2010).
fight congestion in urban areas have become more The European Union has a major role in advo-
and more mainstream issues on local, national and cating for a momentous shift in transport plan-
supranational political agendas in Europe. Among ning and operating, especially by promoting more
the many proposed countermeasures, the effective- sustainable urban mobility policies and the several
ness of (even cleaner) bus systems has been long research projects it funded since the 1990s repre-
acknowledged worldwide. sent continuous opportunities to test, in real urban
This is not surprising since buses remain the scenarios, innovative approaches. Great emphasis
most widespread public transport mode with is placed on promoting studies and implementa-
around half of all public transport passengers tion of solutions to meet more sustainable require-
(30 billion per year) in the EU, reaching 100% ments in local transport systems, and among the
in smaller towns and medium-sized cities (UITP research projects in this field, a series of them is
2011). Moreover, within transit, the most recent specifically dedicated to buses, such as EBSF—

651

TIS2017_Book.indb 651 2/8/2017 8:08:23 PM

 European Bus System of the Future (2008–2013),
3iBS—the Intelligent, Innovative Integrated Bus
System (2012–2015), ZeEUS—Zero Emission bUs
System (2013–2017), along with the more recent
EBSF_2 (2015–2018) and ELIPTIC—ElectrIfica-
tion of Public Transport In Cities (2015–2018). More
specifically, EBSF and EBSF_2 have a common,
two-pronged task: developing innovative solutions
to operate more environmentally-friendly and effi-
cient vehicles and, at the same time, increasing the
attractiveness of this mode of transport.
The paper first describes the measures tested in
EBSF and those to be tested within EBSF_2, then
it reports the results achieved in the former and
outlines those expected to accomplish in the latter
Figure 1. EBSF_2 Consortium partners.
and, eventually, critically analyze the relevance of
both in the field of the current practice for energy
management across Europe. consortia involve four leader European bus man-
ufacturers, along with many others as associated
members. Together they represent more than the
70% of the European bus market and are therefore
2 ACHIEVED RESULTS AND THE able to boost the introduction of key innovations
CHALLENGES AHEAD FOR on a large scale in Europe and beyond. In terms of
THE BUS OF THE FUTURE Public Transport Authorities and Operators, this
corresponds to operations by a fleet of more than
2.1 Innovation and demonstrations 45,000 vehicles and more than 6 billion passengers
Both EBSF and EBSF_2 rely on the synergy per year. They are complemented by suppliers
between research and demonstration activi- and research organizations as well as national and
ties under real operational conditions. Research international associations that will ensure the high-
involves six main key working areas, also called est dissemination and networking of the project
Priority Topics (PTs) within EBSF_2; these are: results.
i) energy strategy and auxiliaries, ii) green driver Central in both projects are the demonstration
assistance systems, iii) IT standards introduction activities. The innovative measures tested in such
in existing fleet, iv) vehicle design (in terms of real demonstrators are selected according to their
capacity, accessibility, and modularity), v) intelli- technological maturity, in order to ensure a short
gent garage and predictive maintenance, vi) inter- step for commercialization after the end of the
face between the bus and urban infrastructure. The project. The use of simulators and prototypes is
theoretical vision behind the PTs is that innovation conceived as a preliminary step for the validation
for buses is a multi-comprehensive concept, syner- of the innovations in real operational scenarios, or
gically involving the study of requirements and per- as a necessary task to prove the potential of more
formance of vehicles, infrastructures, operations futuristic solutions currently implemented at early
and the relevant actors (i.e., passengers, operators, stage of development (e.g. adaptation of bus seg-
drivers, staff, manufacturers, suppliers, decision ments to actual passenger demand through cou-
makers, etc., all equally considered). pling systems).
Both projects reflect this comprehensiveness as Several urban areas across Europe provide
they have been conceived as an open platform for test fields for the demonstrators (within EBSF:
dialogue on bus system efficiency that includes key Bremerhaven, Brunoy, Budapest, Rome, Rouen,
representatives of all stakeholders involved in urban Gothenburg, Paris city and Madrid; the latter
bus services. As already for EBSF, within EBSF_2, three also involved in EBSF_2, along with Lon-
a multiplicity of partners, in this case over 40 bod- don, Paris area, Barcelona, San Sebastian, Dres-
ies of international relevance, are involved in the den, Ravenna, Lyons, Helsinki and Stuttgart).
projects (Figure 1), covering all areas of expertise Even though the measures are very different,
(from manufacturing, to operations and research) environmental concern and energy management are
and geographical regions in Europe. Their mutual behind the majority of them. This also explains, for
collaboration, not limited by commercial relations, some PTs, the major emphasis placed on the vehi-
enables the development of solutions that repre- cles’ more challenging components, when the goal
sent the best technological answer to real opera- is to achieve a more balanced energy management:
tional requirements. To be noted that the projects auxiliaries, with a special attention to Heating,

652

TIS2017_Book.indb 652 2/8/2017 8:08:23 PM

 Ventilation and Air Conditioning (HVAC) and 2.2 Assessing the innovation: the methodological
driver assistance systems. Currently auxiliaries approach
account for 15% to 25% of the total energy budget
According to the dedicated evaluation methodol-
of an Internal Combustion Engine (ICE) bus,
ogy adopted in both projects, the assessment of
but they can rise up to 50% in the case of electric
each measure is a classical “before-vs-during-the-
propulsion with no excess heat energy to exploit.
implementation” comparison of results, with Key
Thus, requirements for energy management for
Performance Indicators (KPIs) measuring the
different propulsion technologies can vary, and so
performance variations at each single case study
do the optimal solutions. A combination of meas-
and cross-case. However, to have a more accurate
ures can increase the efficiency of air conditioning
assessment within EBSF_2, quantitative Perform-
units as well as reduce heat loss via the interior and
ance Targets have been introduced to assess (in
exterior design of buses. Effective circulation of air
percentage), at the end of the project, whether the
could also help maintain a comfortable tempera-
achieved results met the expectations. Demonstra-
ture inside the bus with less energy input. Fresh
tors can select one or more Performance Targets,
air is required to control moisture/condensation
each of them includes one or more KPIs. Among
on window surfaces which is imperative especially
the 65 Performance Targets thus far selected, those
during snowy conditions. A large number of driver
specifically targeted to assess a more balanced
assistance systems, including eco-driving systems,
energy management are: i) Improving the overall
are already available for passenger cars and trucks,
energy efficiency of fleets; ii) Reducing the con-
but not for public transport operations where the
sumption of conventional fuels or electric energy;
driver needs to keep to a set schedule and start/stop
iii) Increasing the uptake of fully electric and hybrid
frequently. “After-market” eco-driving systems
options, iv) Improvement of auxiliaries manage-
designed generically for vehicle fleets provide lim-
ment; v) Decrease harsh decelerations/accelera-
ited access to data other than vehicle accelerations
tions; vi) Reduction in HVAC power requirement;
and braking. Thus, they neither take into account
vii) Reducing noise and air emissions; viii) Reduce
several factors which influence energy consump-
the vehicle’s energy consumption; ix) Expected
tion, nor differences between the driver’s need for
increase in retrofitting programs, x) Energy savings
feedback in e.g. diesel compared to electric buses.
in auxiliary systems. Expectations are high for all
Moreover, in eco-driving, like in other driver assist-
of them, with improvement targets varying from 5
ance systems, the efficiency of the feedback relies
up to 30%, according to the different demonstra-
on the information transmitted to the driver but
tors (Musso & Corazza 2015b).
also on the design of the Human Machine Inter-
The performance comparative analyses are also
face (HMI). Present interfaces typically include
fostered by additional activities, such as modelling
displays showing the driving performance in terms
and surveys. Among these, a study on the theoreti-
of basic feedback which may suffice for the experi-
cal possibility to transfer the good results achieved
enced driver but not for the less experienced ones.
elsewhere in Europe (the so-called Transferability
At the same time excessive information will be dif-
Exercise) and a survey on the energy management
ficult to perceive and may even distract the driver
of the bus fleet (the 2013 EBSF “Survey on Energy
(Tozzi et al. 2016).
Efficiency”, specifically aimed at identifying criti-
Innovations which seem to be specifically
cal issues in public transport companies concern-
designed to improve operations or attractiveness
ing energy efficiency, submitted to 34 European
have also positive consequences in terms of energy
companies, all accounting for a total annual pro-
savings. The design of bus doors may serve as case
duction of 2.3 billion vkm) stressed also how the
in point: within EBSF, prototypes with larger door
perception of innovation may be affected by eco-
blades or with a fifth door were respectively tested
nomic concerns, as reported in the next section.
in Gothenburg and Budapest with the primary aim
to speed up boarding and alighting operations. But
the possibility of reducing overall fuel consump- 2.3 Lessons learnt from EBSF
tion due to the consequent shortened dwell time at
Although EBSF_2 is in progress and tests for the
stops may be not negligible. In light of the EBSF
majority of the measures just started, the EBSF
results (synthesized in Musso & Corazza 2015a),
ones evidenced some interesting results, namely:
energy efficiency proved to be a major issue and
therefore became a specific area of investigation • Ecodriving which led to a reduction of fuel con-
within EBSF_2 as well, where shortened dwell sumption in the two UCs where it was tested
time can be associated to controlled door open- (respectively by –17% and –8%);
ings for less heat exchange or to a totally indoors • New internal layouts, which sped up boarding
bus stop, to provide a controlled, single thermal and alighting operations, thus reducing dwell
environment. times by –13% and –6% in the two test sites where

653

TIS2017_Book.indb 653 2/8/2017 8:08:23 PM

 this performance was surveyed, increased capac- bus (up to 16%) according to Musso & Corazza
ity (respectively +3% and +8%), and improved (2015a).
punctuality up to 52% in one out of the three At the same time, the lesson from the Transfer-
test sites where this parameter was monitored, ability Exercise (in which a total of 35 European
whereas in the other two +3% and +6% were stakeholders took part) was clear: the EBSF meas-
recorded, respectively; ures, although successful and innovative, might be
• Predictive, remote maintenance systems, which transferred if rewarded by reduced fuel consump-
strongly reduced the amount of days in the tion or improved service efficiency, but in any case
workshop of the tested vehicles (respectively they do not have to increase operating costs. Energy
–60% at one test site and –75% at another), and consumption or improved driving styles affecting
contributed to decrease the ratio of non-work- fuel consumption are therefore considered just for
ing vehicles by 25% at one test site. their saving potentials, and not in terms of the pro-
spective environmental benefits. But this result can
In other test cases, some parameters for instance be also interpreted in light of the lower percent-
like fuel consumption, passenger demand, operat- age usually represented by fuel costs in the over-
ing costs, commercial speed and service efficiency all expenditure for operations: being this a minor
provided negligible variations (+1% if compared to item, it is important that remains unaffected by the
the baseline performance prior to EBSF) or did not introduction of innovations.
vary at all. Such results, shortly defined as of “no The Transferability Exercise highlighted one
change” were considered positive, as well, given the more issue: demonstrators more appreciated by
short testing periods, the limited amount of testing the respondents, thus more likely to be theoreti-
vehicles or extent of operations. cally transferable are those where a mix of meas-
An example of the quality of the “no-change” ures were contemporarily tested (for instance, cases
situation is provided by the test scenario for a pre- studies including new HMI driving devices along
dictive, remote maintenance system. In this test with energy management and ITS-based support
site (a small size city), the “no-change” situation systems), rather than a very specifically-focused
had to be ascribed to the relatively short duration trial of a single measure.
of the “during” phase and, partly, to the gist of The EBSF “Survey on Energy Efficiency” pro-
remote maintenance system itself. As a matter of vided, in turn, a snapshot of companies willing to
fact, the primary goal of this test was to develop a switch to more environmentally friendly propul-
reliable system to detect failures and prevent break- sion systems, but still anchored to conventionally-
downs whose efficacy in terms of saving resources fuelled fleets, mostly focused on reducing costs,
(both fuel and money) or improving performance and moderately interested in the use of less pollut-
should require implementation periods far beyond ing fuels, also because the majority of the surveyed
the 8-month duration of the test. Moreover, as the fleets were old and doomed to obsolescence within
demonstrator was applied on a small scale (just 10 a very short term (Carbone & Proia 2013).
vehicles on a single line, out of the local fleet of
152 vehicles, operating 22 lines), it was not sur-
2.4 The contribution of 3iBS
prising that it resulted uninfluential, especially for
what concerns the operating costs on their whole. These results can be analyzed in synergy with those
But, at the same time, the “no-change” situation from a complementary, but independent, sur-
proved that this remote maintenance system, even vey carried out within 3iBS (the 2013 “European
on such a short-range basis, worked properly: had Bus Systems in Europe: current fleets and future
it not worked so and problems strictly linked to trends” survey), involving a panel of more than
the installation and fine-tuning of the system had 70 stakeholders, from 63 cities of different size,
arisen in a reiterated way or/and with a high level corresponding to a total fleet of around 70,000
of severity, the test operations would have been vehicles (buses and trolleybuses) serving a popula-
inevitably affected and extra resources to manage tion of over 100,000,000 inhabitants in 24 Euro-
the service required, since the not-working vehi- pean countries. Some core questions addressed
cles involved in the test should have been replaced, the contemporary and future situations of pro-
anyhow, with consequent, unavoidable negative pulsion systems. As in the previous EBSF survey,
changes, especially in the service performance diesel-fuelled was the most common option (79%),
levels. followed by biodiesel (9.9%), CNG (7%), and elec-
Finally, passengers’ appreciation of the innova- tricity (1.2%); other fuels (hydrogen, bioethanol,
tions, according to the surveys run during the test liquefied petroleum gas, etc.) accounted only for
activities, was high especially for that which con- 3%. Looking at the future plans of the stakehold-
cerned improved quality of service (up to 13%), ers considering different propulsion systems, the
on-board comfort (up to 25%) and image of the majority of respondents are willing to change the

654

TIS2017_Book.indb 654 2/8/2017 8:08:23 PM

 current situation towards less “traditional” propul- passenger comfort and hence the attractiveness
sion systems. of the bus system. The assisted driving will also
In fact, electric traction seems to become the help the ICE-powered buses in complying with
real option of the future as more than 40% of the the real driving emissions limits set by the Euro
respondents want to move towards more electric VI emission regulations, because of smoother
vehicles. More specifically, respondents, who could driving and less transient duty-cycle of the vehi-
select more options, seemed to favor fully electric cle, and resulting into a more efficient way to
buses with batteries (45.5%) or hybrids (69.7%). manage the crucial interoperability of the engine
Among the “no-electric” options, the increased and the emission control system.
preference for CNG (28.3%), biogas (13.2%) and • The test of driver assistance system and HMI
biodiesel (18.9%) was one more proof of the will- technologies compliance with IT standards will
ingness to change, but about one respondent out contribute to move from vertical solutions to
of three still relied on diesel (as it totaled up to fully interoperable ones, both at on-board and
34%) as reported by Tozzi, Guida & Knote (2014). back-office level, for an easier integration of
This can be interpreted as the respondents’ aware- such systems in the existing fleets.
ness that diesel-based propulsion cannot be totally • Electric buses (thanks to clean, high tech appear-
avoided and is coherent with the EBSF survey ance, low noise, zero emissions) can provide a
result of the 70% of companies stating their aim challenging and appealing opportunity to design
to purchase (or having already purchased) Euro VI concepts and new functions for bus stops as
vehicles. well as address interactions between passengers,
vehicles and urban infrastructures. The highest
innovation potential identified in EBSF_2 is
2.5 More goals ahead, the EBSF_2 tasks the indoor stop for electric buses implemented
in Gothenburg which will be analysed in terms
The EBSF accomplishments opened the way for
of users’ perception and interplay between
further developments within EBSF_2.
vehicle and infrastructure, such as the passage-
More specifically, the EBSF_2 vision and
way between outdoors and indoors; this will be
expected outcomes can be considered able to pro-
also an opportunity to measure how to ensure
vide a paradigm shift if compared to the more con-
an ‘indoor’ climate and atmosphere, and study
servative positions of the Transferability Exercise
accessibility issues.
and the two surveys respondents. At the same time,
the trend to change towards more eco-friendly pro-
pulsion systems observed in the two surveys cannot 2.6 Selected case studies within EBSF_2
be ignored if the goal is to develop advanced energy
The activities of the demonstrators to pursue all of
management systems. This is the reason why great-
the above are complex, but a few examples are use-
est importance is attached to the role played by
ful to describe the EBSF_2 testing scenarios.
eco-driving, saving strategies and the enhancement
of Zero Emission Vehicle (ZEV) operations, which
2.6.1 Barcelona, boosting the efficiency
became central in a series of case studies. Therefore
of full-electric buses
it is expected that:
The Barcelona demonstrator is one of the bus
• The effective strategies for energy and thermal networks selected in EBSF_2 to test and validate
management of buses, in particular auxiliaries technological solutions to boost the efficiency of
may contribute significantly to an overall reduc- full electric buses. Two 12 meters electric buses are
tion in energy consumption for both ICE and already operating on routes crossing the city cen-
battery-powered buses. The proposed demon- tre as part of the ZeEUS Project (Figure 2). The
strators on HVAC will implement advanced results already achieved within ZeEUS paved the
technologies for the various components and way for more efficient operations currently about
will combine them individually. Improved HVAC to be tested in EBSF_2, based on extended drive
systems that can use about 20% less energy are ranges and reduced overall energy consumption
tested in different climate conditions. thanks to a multitasking demonstrator which
• Specific driver assistance technologies, encour- includes two technological innovations: an energy
aging an energy saving driving style, may help predictive system and an eco-driving system.
drivers contribute substantially to the overall The former is a system able to achieve up to 20%
goals for energy efficiency. The EBSF_2 solu- energy-savings for the auxiliaries, by implement-
tions aim to reduce the fuel/energy consumption ing a “self-learning” approach (i.e., by learning
between 5 and 8% (depending on the demon- and predicting energy demands in order to opti-
strator activities). These technologies not only mize the management of energy flows between
aim to save fuel/energy, but also to increase the auxiliary systems and the energy storage unit).

655

TIS2017_Book.indb 655 2/8/2017 8:08:24 PM

 Figure 2. The electric bus under test in Barcelona.

Figure 3. The hybrid bus under test in Lyons.

The consideration behind the test is that driving
cycles of urban buses on specific routes are usu-
ally similar, therefore they might successfully put
into operation an intelligent and adaptive system
to optimize usual energy requirements. This is sup-
ported by the awareness that frequent loading and
discharge of energy to and from the storage system
increases the consumption and the energy demand
for cooling system. Moreover, auxiliaries on full
electric buses are powered and run independently
from the driving styles.
Thanks to the predictive system, several online
vehicle-parameters (including the position and
the driving styles) will be collected, sent to the Figure 4. The testing environment in Lyons.
self-learning algorithm which in turn will provide
revised parameters to steer the vehicles towards
more efficient functioning of, among the others, The test to exploit a 2 km ZEV mode (corre-
the steering pump and the air compressor. This sponding to the bus ride across the Croix Rousse
will also provide the opportunity to improve the tunnel opened in 2013 for buses, pedestrians and
efficiency of more components (sensors, compres- cyclists, as in Figure 4), manages the batteries State
sor, touch panels, etc.). During the test, a driver of Charge (SOC) efficiently and assess the impact
assistance system will assess in real time the driv- on the durability of the batteries started in the sec-
ing style so as to have the driver always informed ond quarter of 2016.
about his/her driving efficiency. This means that More specifically, the test relies on a hybrid
quantitative data will be automatically collected vehicle (Euro 6 Urbanway Hybrid) currently
by a control unit and transferred to the back- operating in the so-called “Arrive & Go” mode,
office. A small group of drivers will be selected which enables operations for 50 meter in ZEV
and efficiency improvements will be measured for mode. Modifications on the SOC management
each of them, due to the high potential impact of are required to force batteries charging operations
driver behavior on energy consumption (Musso & ahead of the ZEV strip and perform the no emis-
Corazza 2015b). sion ride accordingly. Bus stops along the route are
simulated. The testing activities also require the
2.6.2 Lyons, exploiting ZEV-mode options preliminary training of the drivers and at least 5
The demonstrator in Lyons includes two main test running tests to have an average performance. Data
fields still based on the binomial “energy man- are collected daily, in real time, to assess the most
agement and ecodriving”: the former relies on relevant parameters (for example, distance, speed,
an energy management system on a hybrid bus acceleration, amps, SOC and voltage of batteries,
(Figure 3) modified to operate a 2 km ZEV-mode ICE parameters, vehicle weight, profile, position,
extension on the route, and to mitigate the conse- weather conditions). One might argue that a ZEV
quent detriment in the batteries lifetime. 2 km strip would appear a rather modest perform-
The latter relies on a new approach to eco- ance if compared, for instance, to the virtually tre-
driving, based on the use of a mathematical model bled ZEV mileage currently tested in Cagliari by
which calculates in real-time a dynamic “fuel econ- a trolleybus line, within ZeEUS (Gasparini 2015).
omy performance” index, taking into account the As a matter of fact, the focus of the test is not
vehicle operating conditions (Blain 2015). on the distance to cover but on the possibility to

656

TIS2017_Book.indb 656 2/8/2017 8:08:24 PM

 investigate how extended ZEV mode operations will be made between different solutions to assess
on hybrid vehicles can make them competitive the possibilities for storing and distributing heat,
with the full electric option, or eligible as a reliable for combining heat and cold storage as well as
alternative for operations in restricted areas of the testing solutions to achieve a distribution of tem-
city. One more field of investigation will include perature which is perceived as comfortable by pas-
the assessment of the environmental potential: fuel sengers and drivers. A new heating system driven
consumption and air and noise emissions of the by electricity and biofuels (instead of diesel) will
tested vehicle will be compared with those recorded be tested (on one of the buses) in the field trial. It
by a regular hybrid bus. will consist of this innovative heat pump and inte-
Unlike Barcelona where the demonstrator grated air conditioner on the roof of the vehicle.
is tested on the same vehicles and routes, or the In addition the coolant liquid is to be heated with
local ZEV mode trial which involve just one bus, a heater combining 16 kW heating capacity from
for the ecodriving system in Lyons a much larger bio-fuel with 7 kW electrical heating while driving
fleet is deployed, due to the specific test features and 9.2 kW electrical heating in depot. A key to the
of this technological innovation. This ecodriving improved performance is that the bus is pre-heated
system is based on a real time algorithm which while standing in depot, either by hot coolant water
assesses fuel consumptions, and identifies driving or by electric power. Such multifold test plan will
styles which may cause excessive consumption. certainly enable a comprehensive measurement of
This requires three different test categories: a trial the energy use for auxiliaries.
in a controlled environment (in which performance But in the case of Gothenburg, innovative find-
from a hybrid bus “as-it-is” will be compared to ings are also expected to come from the research
those coming from a test vehicle equipped with the questions on how design of vehicles and facilities
ecodriving system); a test in real operational condi- might affect the energy management. As said (and
tions (where the performance of a fleet of 40 diesel also already proved within EBSF) larger bus doors
buses equipped with this technological innovation contribute to reduce dwell times, but in this case the
will be compared with those of a fleet of 40 vehicles interplay between bus design and bus stop is also
operating without such equipment); and lastly, the expected to provide more accurate suggestions on
validation of this technological concept, by com- how to speed up boarding and alighting operations
paring performance with and without the design (thus reducing idle times and preventing unneces-
of the selected IT architecture compliant with the sary periods with opened doors which affect the
EN13149 standard and EBSF project results which overall thermal comfort levels on-board, both
are now integrated into the IT×PT (Information causing excessive energy consumption) according
Technology for Public Transport) platform (Tozzi, to the direct observation of the passengers’ behav-
Guida & Knote 2014, Blain 2015). iors. These issues will be assessed both by collecting
data based on a number of participants boarding
2.6.3 Gothenburg, improving energy management and alighting at bus stops and on-board, and by
For the Gothenburg case, the newly-opened line 55 questionnaires (or short structured interviews) to
will be the arena to test performance due to three drivers. Along with that, different bus stops with
technological innovations, more specifically: a new new layouts will be tested at different locations,
air-to-liquid heat pump and integrated air condi- including indoor areas (Figure 5).
tioner on the roof, driven by electricity and bio- The tests for these facilities will rely on real
fuels that will improve the efficiency of heating by trials and field studies. Quantitative data (for
30%; new buses (fully electric) with internal layout instance the number of passengers accessing/
features such as large double doors, low-floor, and
flexible/folding seats to increase accessibility, as
well as new and light interior design and on-board
WiFi; and an advanced design concept for bus
stops including traditional outdoor facilities and
an indoor one, previously mentioned.
Regarding energy strategies and auxiliaries, the
demonstration will include simulation, labora-
tory tests, and a field trial. The simulations and
laboratory tests will deal with investigations of
the impact of vehicle designs on energy use by
taking into consideration the number and type of
windows, insulation, heated surfaces rather than
heated air in compartment, as well as the effects of
the number of passengers on-board. Comparisons Figure 5. The indoor bus stop in Gothenburg.

657

TIS2017_Book.indb 657 2/8/2017 8:08:26 PM

 egressing each stop; dwell time calculations and Table 1. A selection of Performance Targets and KPIs.
average speed) will be collected through techni-
cal counting systems. Observations of travelers’ Performance
behavior at bus stops (activities, positions, cluster- Target KPI definition Unit
ing, etc) will complete the collection of data and Improve energy Average vehicle kWh/km
information. The questionnaires will also include efficiency energy
queries aimed at assessing the passengers’ appre- of fleets consumption
ciation due to improved on-board travel comfort, per km
appearance, and increased quality of service and Energy efficiency MJ/vkm
satisfaction (Tozzi et al. 2016, Musso & Corazza (energy used per
2015b). vehicle)
Reduce energy Average energy €/(vehicle*
costs per costs per vehicle 10,000 km)
3 LEARNING FROM THE RESULTS vehicle per 10,000 km
Improve Total average energy kWh/km
auxiliaries demand by
If results from the surveys and the Transferabil- management auxiliaries per km
ity Exercise are compared, the emphasis placed by Mitigate air CO2 emissions g/vkm
the respondents on the economic side of opera- emissions
tions seems to be a stronger and stronger criterion Assess passengers’ Noise perception Scale 1–10
for the renewal of the fleets in Europe, and an perception of by passengers (survey)
observed poor attention to environmental issues in on-board noise
the EBSF Transferability Exercise may corrobo- Decrease vehicle Average dwell time / %
rate this. dwell time total travel time
But the reiterated interest within EBSF_2 for during the test
measures such ecodriving or to manage energy
relies on the need to address the economic issue 4 CONCLUSIONS
(i.e. reduce the operating costs) not by simply sav-
ing, but by innovating. Innovation, therefore, may EBSF first and EBSF_2 now are continuously
be a driver affecting the overall cost structures building a comprehensive knowledge of the drivers
from many sides. The EBSF_2 measures within the and barriers to increase the attractiveness and the
Energy Management and Driver Assistance PTs efficiency of bus systems, by completing the devel-
are directly targeted to reduce the fuel consump- opment, test and validation of a set of promising
tion and control the energy balance by acting on technological solutions in real operational sce-
auxiliaries or operating ZEV modes. But improv- narios, and evaluating to which extent such inno-
ing the quality of maintenance operations, for vations can contribute to improve cost and energy
example by making parking operations speedier at savings. Both also contribute to the progress in the
depots, or progressing with the IT back-office and development of some key solutions that would
on-board interoperability strongly contribute to have a great impact on bus systems, but require
improve the overall efficiency of the system. Last more steps before wide acceptance on the market
but not least, improved ergonomics and design of (one example for all is modularity). The success
facilities and vehicles may speed up boarding and of EBSF is expected to be replicated by EBSF_2.
alighting operations, and more in general travel To conclude, a set of Guidelines and Tools for the
times with not negligible effects on the energy sav- introduction of innovations will be derived from
ing potentials. the experience of the demonstrators and it will
That these innovations may equally address pave the way to the introduction of the EBSF_2
both the economic and the energetic issues is also technologies and concepts beyond the test fields.
evidenced by the progress in the assessment proc-
ess. The evaluation process is on-going and a set
of comprehensive KPIs is available to detect the ACKNOWLEDGEMENT
performance variations. Evaluation categories
mostly are focused on operations and labour, costs, The authors wish to thank all the participants in
energy efficiency, passengers’ and drivers’ percep- the EBSF and EBSF_2 projects who actively con-
tion. Demonstrators are currently selecting the tributed to the results above described.
most suitable ones, and it is not accidentally that
the trend seems to assess the majority of the meas- REFERENCES
ures under the energetic and the economic points
of view. A selection of some recurring KPIs, thus Beirao, G. & Sarsfield Cabral, J.A. 2007. Understanding
far selected, is reported in Table 1. attitudes towards public transport and private car:

658

TIS2017_Book.indb 658 2/8/2017 8:08:27 PM

 A qualitative study. Transport Policy, Vol. 14, Issue 4, NRBTI 2009. Quantifying the Importance of Image and
July, pp. 478–489. Perception to Bus Rapid Transit. www.nbrti.org
Blain, L. 2015. EBSF_2 Deliverable 5.1. Demo description Stradling, S. et al. 2007. Passenger perceptions and the
and implementation plan. Restricted document. ideal urban bus journey experience. Transport Policy,
Carbone, D. & Proia, E. (ed). 2013. EBSF Delivera- Vol. 14, n. 4, pp. 283–292.
ble—Report on the Energy Efficiency of Bus System. TomTom International 2013. TomTom Conges-
Restricted document. tion Index 2013. http://www.tomtom.com/lib/doc/
Dobbie F. et al. 2010. Understanding Why Some People congestionindex/2013–0322-TomTom-CongestionIn-
Do Not Use Buses. www.scotland.gov.uk/Resource/ dex-2012-Annual-EUR-mi.pdf
Doc/310263/0097941.pdf Tozzi, M., et al. 2016. A European initiative for more
European Commission, 2013. Quality of life of cities. Pub- efficient and attractive bus systems: the EBSF_2
lications Office of the European Union, Luxembourg. Project, Transportation Research Procedia, Vol. 14,
http://ec.europa.eu/regional_policy/sources/doc pp. 2640–2648.
gener/studies/pdf/urban/survey2013_en.pdf. Tozzi, M., Guida, U. & Knote, T., 2014. 3iBS—the Intel-
Gasparini, P. 2015. ZeEUS Deliverable 27.1—Cagliari: ligent, Innovative Integrated Bus Systems. Proceedings
Demo Description and Implementation Plan. Restricted of the Transport Research Arena 2014, Paris.
document. UITP—Union International des Transports Public 2011.
Musso, A. & Corazza, M.V., 2015a. Visioning the bus Position Paper – A comprehensive approach for bus sys-
system of the future: a stakeholders’ perspective, tems and CO2 emission reduction. http://www.uitp.org/
Transportation Research Record: Journal of the Trans- sites/default/files/cck-focus-papers-files/20111122_
portation Research Board, No. 2533, pp. 109–117. UITP%20EU%20position%20paper_comprehen-
Musso, A. & Corazza, M.V., (ed.), 2015b. EBSF_2 Deliv- sive%20approach%20for%20bus%20systems%20
erable 2.2. Report on test scenarios and validation and%20CO2%20emission%20reduction.pdf
objectives. Restricted document.

659

TIS2017_Book.indb 659 2/8/2017 8:08:27 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Waterfront and sustainable mobility. The case study of Genoa

P. Ugolini, F. Pirlone, I. Spadaro & S. Candia

DICCA, University of Genoa, Genoa, Italy

ABSTRACT: The paper analyses European port cities. Port cities are cities which grow up in close con-
nection with their ports. Over the years, ports influenced cities development becoming the main driver for
urban sprawl. Nowadays many port areas are no more exploited for port’s trade. Several cities used these
spaces designing modern waterfronts for leisure, culture and tourism activities. Waterfront revitalisation
is also fundamental for urban mobility. Port cities can use these areas to develop new transport infrastruc-
ture promoting sustainable mobility. In a densely built up area it is increasingly difficult to find space for
bicycle lanes or pedestrian zones, waterfront revitalization projects can be the perfect occasion to solve
this problem. This paper analyses the case of one important port city in the Mediterranean: Genoa. This
analysis is necessary to define new forms of mobility and transport promoting sustainability inside port
cities’ centres taking often advance of port’s abandoned areas.

1 PORT-CITIES ties carried out in these villages were related to port

trades. The opening of the Mediterranean trading
Port cities are fascinating. They are a combination area coincides with the first crusade which revived
of many different cultures, architectural styles, and trade with the Near East interrupting the Mus-
people. Port and city interact in many ways and lim predominance on the eastern Mediterranean.
one is essential for the other. The intrinsic com- Crossing the Persian Gulf and the Red Sea, Arab
plexity of the port city has drawn attention from a caravans brought luxury goods to Egypt and Israel
vast number of scholars belonging to a variety of and from the port of Alexandria, Acre and Joppa
scientific fields (Ducruet C., 2011). sailors from Venice, Genoa, and Pisa transported
Port cities are cities which grow up in close con- the goods to Italy on their way to the markets of
nection with their ports. These two entities have Europe. The easiest way to reach the north from
been developed together over the years. The pres- the Mediterranean was by Marseilles, getting back
ence of a natural harbor or bay attracts boats and on the top of the Rhone valley. Thank to this afflu-
sailormen, and merchandises and good attracts ence of goods the Italian Repubbliche Marinare—
merchants: in this way the first urban settlement is Maritime Republics, established their own “State”
established around the main market. in the Mediterranean shores. In each port city
Different from port cities, it is the case of a city the merchants organized themselves into guilds,
with a port established a long time after the city becoming richer and more influential in social
foundation. This port is the result of the necessity and political every day life. The extent of trade
of a city to trade and it is normally settled far from and urban life were really connected. At the same
the old town centre. On the contrary port cities are with the ending of Viking and Barbarian attacks a
all built around a natural harbor. north trading route was opened, from the British
The Mediterranean Basin has been the cradle Isles to the Baltic Sea. The center of this northern
of world civilization since the first settlements in maritime traffic was the county of Flanders.
Jericho (Israel West Bank) and port cities were fun- Medieval port cities were really compacts: the
damental for Europe development. In the Meddle center with the main market was always established
age port cities were centers of primary importance immediately behind the principal harbor and the
for trade of merchandises and knowledge. Some built-up area was surrounded by high walls to pro-
of the most flourishing and prosperous towns of tect its inhabitants against enemy raid. There were
the XIII century were port cities, such as Venice, three characteristics that marked ports: in the first
Istanbul, Pisa and Genoa. All around the main place, ports had harbors that were the center of the
harbor were established new small ports strictly movement of people and products; secondly, the
connected to the first trade center. Also the hinter- urban morphology of ports always had particular
lands, in the proximity of a big port city, reported buildings or spaces that dominated the city, such
a fast rate of development. The majority of activi- as dockyards, warehouses, customs houses, open

661

TIS2017_Book.indb 661 2/8/2017 8:08:29 PM

 markets, inns and pubs; finally, ports could also be city and the port were no more a single entity but
identified by the particular socio-economic groups two different realities developing one close to the
that they sheltered such as merchants, bankers, other with few points of contact. Many barriers—
bookkeepers, shopkeepers, shipbuilders and for- fences, rails, walls, port police, and customs serv-
eigners (Antunes C., 2010). ice … -were placed between the port and the city
The Mediterranean supremacy started to crack dividing even physically port cities. The port and
with the discovery of America. The main trade the city were facing away from each other develop-
moved on the Atlantic cost facilitating the estab- ing their infrastructure separately. Huge highway
lishment of new port cities in Portugal and in and rail crossroads were also built to get easier the
Great Britain. The Mediterranean Sea became a container loading and unloading and to facilitate
peripheral basin of commerce, because the main the access to port for wagons and trucks. All these
maritime routes where no more between Europe elements contributed to the widespread of the idea
and the close Middle-East, but between Spain, that ports are modern walled city that make the
Portugal, France and England and the new colo- access to the sea impossible for port cities’ citizens.
nies settled in America, Africa and Asia. European This physical separation compounded the negative
colonial powers would be the first to establish a effects of isolation and of insecurity also linked
true global maritime trade network from the 16th to the development of criminality. -the common
century: most of the maritime shipping activity belief considers ports as danger zone were crimi-
focused around the Mediterranean, the northern nality is not sufficiently controlled. This felling of
Indian Ocean, Pacific Asia and the North Atlan- insecurity is another element which contributed to
tic, including the Caribbean (J.P. Rodrigue, 2013). port and city split.
With the opening of the Suez Canal the second The situation has completely changed at the end
half of the 19th century saw an intensification of of the 20th century: different old ports were reno-
maritime trade to and across the Pacific and inside vated and revitalized and many efforts have been
the Mediterranean Sea but here boats normally made to re-establish the relationship between the
were directed to England crossing the basin. port and the city.
Port cities have played a pivotal role in each In the wide and complex panorama of urban
Industrial Revolutions; ports have thrived because transformation, waterfront revitalization is one
all Industrial Revolutions up until now relied on of the most interesting phenomena of urban
the trade of raw materials: coal and iron ore since renewal of the last decades, bringing ‘cities on
the 18th century, whereas oil and gas were added water’ around the world to a new leadership (Gio-
to the commodity mix since from the late 19th vannazzi O., Moretti M., 2010). Many big cities
century (Jansen, M. 2016). After the second indus- worked towards water to offer a better quality of
trial revolution, many port cities renovated their life regenerating people’s space and respecting the
structure to host new factories handling coal, ore, environment. Waterfront and riverfront after New
oil, oil products, and chemicals. As for the Mid- York and San Francisco’s experiences are now of
dle Age, ports were the driver for urban develop- central importance for urban planning in many
ment. Many people moved to port cities from the international cities such as Buenos Aires, Shang-
countryside to work in the factories. Port cities hai, London, Barcelona and Porto. These port
significantly incremented their population and cities are good examples, best practices of urban
it was necessary to establish new settlements for renewal increasing the quality of life starting from
workers and their families outside the city wall. degraded port areas. This was possible paying par-
Since the 80 s of the XXI century the spread of ticular attention to the design of new public spaces,
container transport obliged port cities to build new transports, residential area respecting the environ-
and modern docks for container ships and cargo ment. The cruise industry had more benefits than
containers. These port infrastructures require other economic sectors, from port regeneration
much more space than the previous docks to stor- offering new activities for tourists near the sea. The
age both loaded and empty containers. Normally food chain (local products and tradition) also had
loaded containers are stored for short periods then its place in the renewal process. Many historical
they are sent to northern Europe by train, trucks markets closed to old ports were renovated thanks
and barges, on the contrary empty containers can to waterfront revitalization projects and now are
remain on the docks for longer periods awaiting the points of interest for tourists and citizens.
next use. The container terminals were often set up As maritime trade flows within and between
far away from the original/historic port with good ports change, the recognition of new values of
rail or highway connection. Because of containers the waterfront, including through tourism, have
trade many old ports were abandoned or under- spurred projects to attract attention, investment,
exploited and their infrastructures (merchandise and people: Opportunities to improve quality of
stores, cranes, quays…) began to degrade. But the life and environment and enhance economic vitality

662

TIS2017_Book.indb 662 2/8/2017 8:08:30 PM

 have led to the revitalization of degraded and proposing new solutions for sustainable mobility
underutilized port regions, often through strate- inside the waterfront area and the old town centre.
gies that incorporate their rich historical and cul-
tural heritage to create unique new public spaces
(Ivinis C., 2013). 2 THE PORT CITY OF GENOA
Waterfront revitalization is also fundamental
for urban mobility. Port cities can use old ports’ Genoa is one of the most famous examples of
areas to develop new transport infrastructures port-city. The port of Genoa is beautifully located
promoting sustainable mobility. In a densely built between the Ligurian Sea and the Apennine Moun-
up area it is increasingly difficult for urban plan- tains. It is the major Italian seaport with a trade
ner to find space for bicycle lanes or pedestrian volume of 51.6 million tons of cargo.
zones. At many locations, it might appear compli- Genoa is one of the main port cities of the
cated to provide for on-road cycling in a way that Mediterranean; in truth, the northern most port
satisfies the needs of bicycle riders and other road of the Mediterranean Sea. This position can pro-
users and waterfront revitalization projects can be vide remarkable advantages for the city, which
the perfect occasion to solve this problem. The can aspire to become a juncture between Europe,
European White paper of transport—Roadmap to Africa and Asia Minor (Gabrielli B., 2010).
a single European transport area, towards a com- Genoa has strongly been connected with its
petitive and resource-efficient transport system port since the beginning of its history. For years
(2011) – claimed the necessity of a competitive and the city and the port lived together in a state of
sustainable transport system across EU Countries interdependence and the local community based
reaching the 60% emission reduction target. Water- their identity on the relationship with the sea and
front renewal can contribute to reach this goal. One sailing, in a prosperous union.
of the first port cities which decided to renovate its The extraordinary shape of seaside cities—that
port in favor of sustainable transport was Portland we can better appreciate from the sea—is the result
in Oregon. In 1974 Portland tore down a portion of of the synthesis between urban culture and mari-
the Route 99 W freeway to reclaim the waterfront time culture; this harmonious union of maritime
for a pedestrian/cycling seawall and park. Recently and urban cultures offers a different point of view
the city of Boston even defined the “South Bos- that is full of charm and semantic values (Clem-
ton Waterfront Sustainable Transportation Plan” ente M. 2013).
(2015). The Sustainable Transportation Plan lays Genoa is a typically port city. The first settlement,
out a blueprint for transportation and public realm dated 6th and 5th centuries BC, was established on
improvements for the South Boston Waterfront a hill (called Castello) controlling a natural har-
over the next two decades (Boston Redevelopment bor. The port was created according to the nature
Authority, 2015). The main goals of the plan are: of the site and the city developed around the bay.
Improve Access and Mobility for All (improve Thanks to the prosperous commerce with the Mid-
multimodal access to/from/through and mobility dle East, in the XIst century Genoa emerged as an
within the South Boston Waterfront for residents, independent city-state. In the Middle Age, Genoa
workers, maritime-related commerce, and visitors); held one of the largest and most powerful navies in
Support Economic Growth and Vitality (deliver the Mediterranean financed with trade, shipbuild-
the transportation infrastructure needed to sup- ing, and banking. The commercial routes along
port a world-class economy); Reinforce Sustain- the Mediterranean coast enriched enormously the
able Policies and Programs (align programs and town which became one of the most influent port
policies to support more sustainable transportation cities in the Mediterranean with Venice, Pisa, and
choices and demand management to and within Amalfi, the so-called Maritime Republics. In the
the South Boston Waterfront); Enhance the Pub- XIIsd century was built the first port according to
lic Realm (contribute to enhancing the attractive- an established project innovating the pre-existing
ness and quality of the urban character through settlement. Legal and administrative means guar-
ongoing transportation investment); Contribute anteed the maintenance and the development of
Environmental and Health Benefits (Realize the port’s infrastructures. The old town centre was
positive environmental effects and health benefits characterized by the presence of copious markets
that result from a more sustainable transportation and there were also entire streets or arcades (for
plan. Invest Smartly for the Future); Advance stra- example the present Sottoripa—ex Ripa Maris)
tegic investment to ensure the long-term financial consecrated to the trading of goods. For this rea-
and operational sustainability. The plan for Bos- son, Genoa was called both a port and a market
ton waterfront is good example, a best practice, city. The main markets of S. Giorgio, Banchi and
which could be transferred to other port cities in Soziglia were placed on the principal streets which
Europe. This paper analyses the port city of Genoa connected the city with the port. The markets were

663

TIS2017_Book.indb 663 2/8/2017 8:08:30 PM

 the real point of contact between the city and the through the rediscovered use of collective and rep-
port, the physical place where merchandises, peo- resentative functions at a metropolitan level. In
ple, ideas and culture melted together. S. Giorgio addition, the overall image of the city is seen from
and Banchi are located perpendicularly to Sot- the outside not only as more linked to the harbor
toripa on the same axis; Soziglia is rearward in and the industry (often seen in a static or decadent
comparison to the first two in the middle of an way), but also as a new element linked to art cul-
artisanal district. ture, architectural beauty, and new form of accept-
Every day were sold textile products coming ance (in a dynamic and more propulsive vision)
from the north of Europe, spices from the Middle (Gastaldi F., 2014).
East and the refined silk coming from china. Dur- Other international events (the G8 meeting in
ing the Middle Age Genoa could count on its colo- 2001 and the European City of Culture in 2004)
nies settled in Africa (Tabarca), in Crimea (Caffa, provided resources for regenerating other portions
Cherson, Cembalo…), in the Caucasus region of the old port and of the old town centre. Nowa-
(Copa, Mapa, Casto…) and in Anatolia (Amasra, days the working port of Genoa is almost all set in
Carpi…) to trade metals, stones and other precious the districts of Sampierdarena and Voltri, far away
goods. The urban fabric reflected port configura- from the city centre. It covers an area of about 700
tion, indeed the roads were built in correspond- hectares of land and 500 hectares on water with
ence to the main quay to get easy the loading and 30 km of operative quays. It is incredible wide cov-
unloading of merchandises. ering around 22 kilometers of coastline. The cruise
Since the 1970s the predominance of the iron terminal instead is located immediately westward of
and steel industry and of the containers traffic sep- the Old Port. This is a privileged position because
arated the port of Genoa from the city. Huge high- cruise tourism has a direct access to waterfront lei-
way crossroads and fences were built separating sure activities. Further interventions restored other
port activities from Genoa daily life. In particular, port abandoned areas and buildings: on the site of
it was impossible for Genoa citizens to have any the medieval Arsenals were built the Sea Museum
visual contact with water. Maritime traffic affects and the new Faculty of Economics. Actually the
only container terminals set up in the west part of urban waterfront covers an area of 117.000 sq
the city far from the old port that was progressively m—of which 86.000 sq m are open spaces—and it
abandoned. In 1985 the remaining activities finally is made up of: 23% Culture, permanent exhibition
moved and the Municipality of Genoa formed a and museums; 20% Congress Centre; 17% Leisure,
Commission, in cooperation with the Region Lig- education; 15% Facilities and parking lots; 12%
uria and the Port Authority, to study all the pos- Offices; 7% Restaurants and bars; 6% Retail.
sible solutions to connect again the old port with In the last months, the Municipality of Genoa
the old town centre. In 1992 Genoa has a unique has decided to open an international ideas compe-
opportunity to renovate its waterfront thanks to tition to renovate the last portion (Figure 1) of the
an International Exhibit hosted inside the old port port inside the city center starting from a project
as part of the Columbus Celebration (for the 500 donated to Genoa by the Renzo Piano Building
years since the discovery of America). State invest- workshop.
ments helped the city to afford the restoration of The overall project—called “Bluprint for
the majority of the old port area. Waterfront revi- Genoa”—outlines the profile and the future func-
talization included the recovery of almost all the tions of the part of the city located near the east
historic docks, the rearrange of the open space
in a big square and promenade along the sea and
the realization of a big aquarium, a panoramic
lift and an ice rink. The old port was permanently
converted to urban uses, in particular for leisure,
culture and tourism activities. The waterfront
regeneration processes have also concerned the
relationship between the city and the port.
The restoration of the waterfront and the main-
tenance operations in the historic centre have led
residents to a change in their perception of their
own city; the re-ownership of the waterfront and
the notable quantity of historical places recovered,
but also the reconversion of large disused indus-
trial sites through the introduction of new uses,
have been translated in terms of growing sense of Figure 1. The waterfront of Genoa. The port, the city
belonging and have reinforced elements of identity and the historical markets.

664

TIS2017_Book.indb 664 2/8/2017 8:08:30 PM

 port entrance: the challenge is to combine tourism Anatoliki Makedonia—Thraki); Municipality of
with the industrial activities situated in the neigh- Limassol (Cyprus); University of Genoa—Research
borhoods, where companies of excellence work on Centre in Town Planning and Ecological Engineer-
ship repairing industry and refitting and where are ing (Italy, Liguria); Municipal Institute of Markets
foreseen other important projects. of Barcelona (Spain, Cataluña); PLURAL—Eu-
The part of the waterfront designed by the Blue- ropean Study Centre (Italy, Liguria); Municipality
print shows some strong features: a central posi- of Favara (Italy, Sicilia); National Research Centre
tion; the presence of two docks with more than (Egypt); Souk El Tayeb (Lebanon).
400 moorings; a big pavilion for events designed by The project collected and exchanged best prac-
Jean Nouvel; the historical “Palasport”—a struc- tices to enhance competencies and capacities of
ture with a great potential—and a close promenade, small entrepreneurs operating in city markets. The
“Corso Italia”, for leisure activities and bathing. partnership promoted the integration of productive
The requalification of areas of the former Trade chains strategies at Mediterranean level with focus
Fair gives to Genoa the possibility of realizing an on high quality agro-food and handicraft products.
essential tile to complete the waterfront restyling Local authorities and private stakeholders were both
started in 1992 by Renzo Piano with the renovation involved in the revitalization process to guarantee a
of Porto Antico, today one of the main touristic sustainable management of historic city markets.
attraction and livable places in Genoa. The guid- The participative approach proposed by the project
ing principle of the project is defined by Piano as firstly involved the local dimension in order to ensure
a “mending” of the urban pattern, the win back of the effective participation and territorial relevance
the sea by the city and the reopening to the water of of the initiatives carried out. Marakanda involved
part of land reclaimed from the sea (the canals). essentially local communities by working with them
on common aspects of the day to day life and ordi-
nary running of the economy and markets.
3 SUSTAINABLE MOBILITY FOR THE Marakanda established also common quality
WATERFRONT OF GENOA standards for the efficient management of historic
city markets. Particular attention was given to the
In this paragraph are reported different researches connection between markets, ports and old town
on the waterfront of Genoa. In particular, innova- centers. Sustainable mobility was preferred and the
tive solutions are here analyzed for a sustainable project supported pedestrian walkways and cycle
mobility inside the old port area. A better urban routes. A better connection was of fundamental
mobility Plan is of paramount importance to over- importance to revitalize city markets and to rees-
take the remaining barriers between the port and tablish the historical relationship between the city
the city. Port cities can use these areas to develop with its markets and port. These elements lived
new transport infrastructure promoting sustain- together in a state of interdependence and mar-
able mobility. It is increasingly difficult to plan kets renewal was necessary to restore a prosper-
bicycle lanes, new bus lines or pedestrian zones in a ous union. A good connection is also important
densely built up area. For this reason, waterfronts for tourism. Markets can be a veritable tourist
are an extraordinary opportunity for urban plan- attraction promoting local foods and traditions.
ners to have new areas to dedicate to sustainable Pedestrian walkways which connect old town cen-
mobility. tres, waterfront and markets are for this reason of
paramount interest.
In figure 2 is reported a study which analyses
3.1 Genoa as port-market city
every possible sustainable connection between the
Genoa is not only a port city but can be considered historic city markets, the waterfront and the old
as a port-market city. This is a result of the fact town centre of Genoa.
that the urban pattern is marked by the presence of New pedestrian walkways are identified to get
many markets and a central port. The authors have easy for citizens and tourists to move around the
developed a detailed study on market cities within city. Cruise tourism could also benefit from these
the European project “Marakanda Historical mar- interventions: from cruise terminal tourist can fol-
kets” (2012–2015) funded with the European Neigh- low a safe and pleasant itinerary to discover Genoa
bourhood and Partnership Instrument—ENPI historical buildings.
CBCMED (scientific responsible: P. Ugolini and
F. Pirlone). Marakanda brought together 10 direct
3.2 The old town centre and the port of Genoa:
partners from six different Mediterranean Coun-
identification of new itineraries
tries: Municipality of Florence (Italy, Toscana);
Municipality of Genoa (Italy, Liguria); Local The aim of this research is to enhance the rela-
Authorities Union of Xanthi District (Greece, tionship between the port and the city of Genoa

665

TIS2017_Book.indb 665 2/8/2017 8:08:30 PM

 markets were a part of a single productive system.
Now these elements could interact together again
preferring the realization of public spaces for lei-
sure, culture and tourism activities and directional
centers.
The authors have done a detailed study on
urban mobility inside the new waterfront and the
old town centre of Genoa (from the cruise termi-
nal—Piazza Fanti d’Italia—to Mura della Marina)
in favour of sustainable mobility. In particular, it
has been analysed the public transport network.
In this area there are 3 bus lines, a trolleybus, dif-
ferent public elevators, one underground line and
the main railway station (Figure 3).
On the whole the public transport system is suf-
ficient and it covers the entire area. Different prob-
lems are caused instead by car traffic. In this case
the introduction of interchange station parking
could push drivers to use public transportation.
There are different users that walk inside the study
area, mainly students, tourists and workers. There
are many pedestrian walkways both inside the old
town centre and the waterfront. These paths are
Figure 2. Identification of the main roads and ways generally safe inside the waterfront, but the same
between the historical markets of Genoa: Dinegro, Stat-
uto, Carmine and Sarzano.

Figure 4. The outline of the possible connections

between the port and the city.

Figure 3. Analysis on public transport inside the water-

front area.

supporting sustainable mobility. Waterfront revi-

talization is analyzed to a larger scale including the
old town centre and the historic markets inside a
more complete reclamation project. This is because Figure 5. Touristic routes (pedestrian and cycling areas)
throughout history, port, old town centre and city inside Genoa waterfront and old town centre.

666

TIS2017_Book.indb 666 2/8/2017 8:08:30 PM

 cannot be said for many pedestrian streets inside of the city, a driver for new interventions and for
the old town centre that, especially at night, can future restoration projects. This is because many
be dangerous. town planners consider old ports as a driver for
For this reason of primary importance is to economic development capable of attracting the
create new safe vertical connections between the interest of international investors. Particular atten-
waterfront area and the historical centre (Figure 4) tion is given to public spaces and sustainability. The
to get stronger the relationship port—city. aim is to crate open and permeable places, lived in
The most critical area is the old district of Prè. and felt to be a part of the city. New waterfronts
Here the waterfront is completely separated from are more enjoyable, fostering relations, displays,
the city by a physical barrier, the “sopraelevata” an development and enhancement in the fields of lei-
elevated road build in the ‘60 s. This isolation con- sure, sport, mobility and culture. Port cities look to
demned the main road of the district—via Prè—to the future transforming post-industrial areas and
become a degraded and unsafe place. This road, if buildings into new production areas.
better connected to the rest of the old town and the Genoa is an evident example of this revitaliza-
port, can become a market street to promote local tion process. Its renovated waterfront—the old
products with a touristic interest. In Figure 5 are port—interacts with the city and it is a magnet for
reported different sustainable ways that could con- resources and fluxes capable of exploiting oppor-
nect the study area preferring pedestrian walkways tunities to generate new economies. But there is
and cycle tracks. still a lot of work to do: this paper suggests new
All the old town centre’s districts can attract dif- itineraries and mobility solutions to carry on
ferent users according to their vocations: night life, the waterfront revitalization enhancing the cur-
culture, wine and food shops and markets, … all rent relationship between the city, the port and
moving in a sustainable way. its historical markets and supporting sustainable
development.
3.3 Conclusions
In conclusion, waterfront revitalisation is of para- REFERENCES
mount importance to relocate in the international Artuso, N. & Salvetti, L. 2013. Verso il completamento
context port cities and to improve the urban qual- del waterfront di Genova, dagli anni ’90 ai programmi
ity of life. It is preferable a larger-scale reasoning futuri, PORTUS: the online magazine of RETE, n. 26,
which includes inside the project area appropriate November 2013, Year XIII, Venice.
connections between the port, the city and its his- Boston Redevelopment Authority 2015, South Boston
torical markets in favour of sustainable mobility. Waterfront Sustainable Transportation Plan, VHB,
The concept of sustainability has been applied USA: Boston.
to port services and infrastructures revitalizing Busi, R. & Pezzagno, M. 2006. Mobilità dolce e turismo
sostenibile: un approccio interdisciplinare, Gangemi
historic port areas. Several port cities such as Editore, Rome.
Barcelona, Genoa and Marseille are reclaiming, Cátia Early, A. 2010, Modern Ports: 1500-1750, Insti-
or reclaimed, their waterfront from freeways and tute of European History (IEG), ISSN 2192–7405,
industry. These cities are using ports’ abandoned Germany: Mainz.
areas designing modern waterfronts for leisure, Carnevali, G. & Delbene, G. & Patteeuw, V. 2003. Devel-
culture and tourism activities because they have oping and rebooting a waterfront city, NAi Publishers,
realized that public access to water—for the sim- Nederland: Rotterdam.
ple pleasures of walking, cycling, beaches, boating, Clemente, M. 2013. Sea and the city: maritime identity for
etc. –is a necessary asset for any city that calls itself urban sustainable regeneration, TRIA International
Journal on city planning culture11 (2/2013) 19–34/print
livable. ISSN 1974-6849, e-ISSN 2281–4574.
Urban planners and policymakers, pursuing European Commission 2011. The European White paper
waterfront projects, conceive innovating approaches of transport—Roadmap to a single European transport
for interlinking socio-economic and environmental area, towards a competitive and resource-efficient trans-
dimensions. Public and private engagement is a port system. Commission services, Belgium: Bruxels.
precondition for project success, so it is preferred a European Commission 1992. Green Paper: Towards a
joint cooperation since the planning phase. Water- new culture for urban mobility, Commission services,
front renewal has in many cases a catalytic effect Belgium: Bruxels.
becoming the starting point for the revitalization of Fera S., Minella M. 1999. Genova porto e città, Sagep,
Genoa.
the entire city. Ports are a perfect place for urban Gabrielli B. 2010. The Renaissance of Cities: the Case
experiments because they are located in strategic of Genoa, PORTUS: the online magazine of RETE,
areas, close to city centers and transports routes. n.16, June 2010, Year XIII, Venice.
Sometimes this restoration process transforms Gabrielli B., Ghiara H. 2013., Genova, dal waterfront
an unused port in one of the most dynamic parts alla città portuale del XXI secolo, PORTUS: the

667

TIS2017_Book.indb 667 2/8/2017 8:08:33 PM

 online magazine of RETE, n. 25, June 2013, Year XIII, magazine of RETE, n. 27, May 2014, Year XIV,
Venice. Venice.
Gastaldi F. 2014. Waterfront redevelopment and gentrifi- Rodrigue J.P. 2013. The Geography of transport system,
cation in the inner city of Genoa, PORTUS: the online ISBN 978-0-415-82254-1, Routledge Edition, USA:
magazine of RETE, n. 30 Year XIV, Venice. New York.
Genoa Port Authority 1999. Piano, porto, città. Schubert D. 2013. L’ultima frontiera del waterfront
L’esperienza di Genova. Skyra Editore, Milan. rigenerazione urbana: Nord Europa, PORTUS the
Genoa Urban Lab 2008. Quaderno 1, ECIG Edizioni online magazine of RETE, n. 25, giugno 2013, Anno
Culturali Internazionali, Genoa. XIII, Venice.
Genoa Urban Lab 2011. Quaderno 2, ECIG Edizioni Troin J.F. 1997. Le metropoli del Mediterraneo. Città di
Culturali Internazionali, Genoa. frontiera, città cerniera, Jaca Book, Milan.
Ivins C. 2013. Waterfront Revitalization in Port Cities:
Risks and Opportunities for Emerging Economies,
Brics Policy Centre, Policy brief publication, Brazil: Notes
Rio de Janeiro. Selena Candia is responsible for the researches on water-
Jansen M. 2016. The Fourth Industrial Revolution and fronts and on cycle mobility as preferred sustainable
the future of ports, World Economic Forum, Switzer- mean of transport inside the port city of Genoa.
land: Geneva. Francesca Pirlone is responsible for port cities general
Municipality of Genoa 1995. White book on Genoa mobil- description and she carried out different researches on
ity, Mobility and Transport department, Genoa. sustainable mobility for the city of Genoa.
Municipality of Genoa 2012. Urban mobility plan for the Ilenia Spadaro is responsible for a detailed study on
city of Genoa, Mobility and Transport department, the historical evolution of port of Genoa.
January 2010 updated with C.C. n. 28/2012, Genoa. Pietro Ugolini is responsible for the researches on
Ribalaygua B., Latorre E.M. 2014. Port and City. An Mediterranean port cities and on Mediterranean histori-
approach to the study of new urban port role from cal markets.
its perception as node and place, PORTUS: the online

668

TIS2017_Book.indb 668 2/8/2017 8:08:34 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

New forms of mobility for an alternative territorial fruition:

The rediscovery of tourist footpaths

R. Papa & R.A. La Rocca

Department of Civil, Architectural and Environmental Engineering (DICEA), University Federico II of Naples,
Naples, Italy

ABSTRACT: The search for modalities of movement inspired by the sort of “soft mobility” that can be
an alternative to the use of a car represents an opportunity to promote sustainability in territorial use and
its fruition. In this sense, the definition of methods and tools aimed at aiding decision-makers in identify-
ing policies and technical interventions for both the promotion of tourism and the safeguarding of the
existing system of resources is one of the main targets of town planning. The enhancement of existing
cultural paths, as well as the realization of new foot routes, constitutes an opportunity for tourism devel-
opment and territorial sustainability. In the context of these premises, this study represents a proposal to
characterize a network of “cultural tourist paths,” also considering the application of GIS technologies in
supporting both the planning and use of such tourist routes.

1 CULTURAL TOURIST PATHS AS The Directive, then, also wants to be a tool

“COMMON HERITAGE” for supporting and fostering the less well-known
regions by encouraging tourist typologies different
The use of soft mobility is currently one of the from the mass tourism model. For these reasons,
primary agents of territorial fruition, especially for “cultural tourism” (WTO 1985; ICOMOS 2002,
those who put particular emphasis on environmen- Richards 2003) has been identified as the form of
tal issues and sustainability. The focus on a devel- tourism most adequate for the use and promotion
opmental model of conscious tourism that favors of this kind of territorial resource.
a “slow” model has concentrated attention on the Even though there is not yet a uniform defini-
current national and European policies which aim tion of cultural tourism, it is considered to be the
for the revaluation of cultural paths, as well as a form of tourism that favors territorial cultural
renewed interest in historical, environmental, cul- aspects, in its rediscovery of the primitive educa-
tural and territorial identity. tional function of the journey, intended as an expe-
In Italy, the recent directive from the Ministry rience, as it was for the nineteenth-century model
of Heritage, Culture and Tourism (2016) in estab- of the Grand Tour.
lishing the year of the routes, underlines the role of If the identification of cultural tourism as a
historical-cultural paths both for the enhancement sustainable typology of tourism is a recent acquisi-
of territorial resources (even economic ones) and for tion, the European project to activate a cooperative
the promotion of sustainable tourist typologies. network by the revaluation of cultural routes dates
The Directive, in particular, mainly emphasizes back to the Sixties, when the European Council
two aspects. On the one hand, it points out the certificated the first two historical routes: the Way
need for in-depth examination of the state of the of Saint James and the Rural Habitat.
existing routes capable of being adapted for the The first example is perhaps the best known
soft mobility manner of moving, in order to cre- among European routes; it has been the object of
ate an Atlas of Routes. On the other hand, it high- literary and cinematic interest, as well as historical,
lights the exigency of the different bodies (Region, cultural and religious. Set up in the ninth century,
Universities, Research Institutes, Associations, as a route for pilgrims to reach the grave of Saint
etc.) engaged in territorial development operating James, at present, the Way of Saint James repre-
in a collaborative vision in order to promote “slow sents an icon for cultural tourism. Data on fluxes
tourism” as a user typology compatible with the in tourism, in fact, show how cultural motivation
valorization of historical and cultural identities of has increased in recent years (Table 1), despite reli-
the territories through which the paths cross. gious motivation remaining the main impetus for

669

TIS2017_Book.indb 669 2/8/2017 8:08:35 PM

 Table 1. Types of motivation for traveling the Saint
James Way 2012–2015.

Motivations 2012 2013 2014 2015

Cultural and 101.171 117.715 120.412 141.969

religious
Only religious 79.490 86.291 101.013 99.680
Only cultural 11.827 11.804 16.461 20.809
Total 192.488 215.810 237.886 262.458

Figure 2. Percentage of types of motivation for walking

the Via Francigena in Italy (elaboration from TCI 2015).

existing cultural routes and supports new projects

for enhancing territorial resources. At present,
there are thirty-two certified routes representing
different, interesting geographical regions and
referring to different themes inspired by European
history and culture.
In particular, the classified routes can be clus-
Figure 1. Number of pilgrims in the period 2004–2015 tered into nine macro-categories:
(elaboration from www.prgrinossantiago.es consulted on
08/08/2016). • pilgrimages (the Way of Saint James since 1987;
the Via Francigena, the Saint Olav);
• European celebrity (European Mozart Ways; the
the flow of tourists, with numbers growing signifi-
Schickhardt Route; the Saint Martin of Tours
cantly on occasions of holy events (Fig. 1).
Route; the Huguenot and Waldensian trail; the
A recent survey carried out by the Italian Tour-
Via Habsburg; in the Footsteps of Robert Louis
ing Club in 2015, using a sample of 400 tourists,
Stevenson; Destination Napoleon; European
gives the main reasons that users indicate for hav-
Routes of Emperor Charles V);
ing decided to retrace the medieval Via Francigena
• monasteries (European Route of Cistercian
(Fig. 2).
abbeys; the Cluniac sites in Europe);
In particular, it can be noted that religious
• historical and architectural heritage (Routes
motivation inspired only 10% of the sample, who
of El legado andalusí, Transromanica; Euro-
were mainly driven by a cultural reason (22%). On
pean Cemeteries Route; ATRIUM; Réseau
the other side, the Habitat Rural paths refer to a
Art Nouveau Network; the European Route of
system of routes characterized by the presence of
Megalithic Culture);
particular, predominantly rural ecosystems identi-
• rural landscapes (Routes of the Olive Tree; Iter
fiable on the basis of specific local features (special
Vitis)
craft production and/or gastronomic interest, spe-
• trade lines (Phoenicians’ Route; the Viking
cificity of the natural landscape, etc.).
Routes; La Hanse);
The Program of the Cultural Routes of the
• industrial and craft lines (The Pyrenean Iron
Council of Europe is based on the idea that the
Route; The European Route of Ceramic);
valorization of leisure through culture can activate
• prehistoric art (Prehistoric Rock Art);
cooperation among different countries. The wander-
• specialized towns (European Route of Historic
ing journey, then, can represent the means to redis-
Thermal Towns).
cover heterogeneous heritages still interconnected
within common historical and cultural roots. The Cultural Routes of the Council of Europe
In this sense, the path can be assumed to be a express the will of a shared and cooperative project
shared “common cultural heritage,” able to link involving all political and administrative levels
different places through the physical and immate- (from local to national to transnational), as well as
rial resources of the lands across which it runs. the local population, in development of the pro-
Since 1997, the European Institute of Cultural gram. They also represent a project of cooperation
Routes (EICR) has been the technical body of the between cultural, educational, and tourism inter-
Council of Europe. The Institute evaluates already ests aimed at promoting European values.

670

TIS2017_Book.indb 670 2/8/2017 8:08:35 PM

 Among the European countries, Italy is one ent aspects involved. In particular, as regards the
of the most engaged in the Program of routes public component such actions should include:
(Fig. 3). In fact, the relevance of the historical
• the coordination between the target of the devel-
routes for tourism in Italy has also been recognized
opment plan and those of the town planning;
at the institutional level, particularly in the current
• the dimensioning of the infrastructure (physical
government’s program focused on the enhance-
and technological) in order to support the tour-
ment of the Via Francigena and the Route of the
ism fruition of the territorial resources;
Phoenicians.
• the implementation of facilitations (tax incen-
It should be noted that the Program of Cultural
tives and so on) in order to improve the involve-
Routes, rather than focusing on economic objec-
ment of the private sector and to increase the
tives, aims to promote sustainable and non-invasive
supply services for the better use of the routes
forms of tourism by proposing alternative modes
(accommodation, catering, management, etc.).
of visitation and transportation. The promotion
of sustainable tourism, in fact, is one of the cri- As regards the private component, actions
teria necessary to obtain European certification. should include:
The need to promote culture in order to propose
• the collaboration among the different interests
different models for utilizing territorial resources
involved in the development project;
has also been adopted in the cultural itineraries
• the promotion of entrepreneurial initiatives
defined by the Italian Ministry that focus on “soft
aimed at enhancing the local identities and ter-
mobility” as a way of travelling along the routes
ritorial resources;
with opportunity for territorial knowledge.
• the compliance with minimum standards of
This framework, on one side, highlights the
quality that have to be defined and monitored;
awareness (at the national and pan-European lev-
• the sharing of information to avoid
els) that cultural routes represent an important
redundancy;
means for promoting lesser-known destinations and
• the coordination among actions of promotions
contribute to the sustainability of tourism develop-
in order to ensure the brand awareness.
ment, while on the other side, it underlines the need
to make these resources available through appropri- The synergy between these two components (pub-
ate actions of territorial planning and design. Such lic and private) strongly involved is a prerequisite for
actions should provide for urban and territorial the realization of a shared, balanced and sustain-
policies targeted to the coordination of the differ- able project of territorial development. The support
for the realization of this project can came also from
the correct use of technology (Papa 2013).

2 CULTURAL PATHS AND TOURIST

DEMAND

Over the past few years, there has been a radical

change in tourist behavior and habits: personal use
of tablets and smartphones has, in fact, exceeded
the purposes exclusively connected to business
needs, with that use spreading to recreation and
entertainment, too. In this change, tourism can
have an active role in promoting sustainable life
style in this contributing to the opening of new
perspectives of the recent concept of urban smart-
ness (Papa & Fistola 2016).
In truth, our entire interpersonal system of com-
munication has changed quite rapidly. Referring to
tourism, in particular, the whole information sup-
ply system is modifying itself and a multiple “vir-
tual dimension” (Fistola & La Rocca, 2001) is being
established beside the traditional supply of tourists.
Some scholars (Orlando 2013; Garau 2014)
Figure 3. Country involved in the European Cultural identify at least three types of use for new technol-
Route Program (elaboration from www.camminideu- ogies in the fruition of cultural heritage as applied
ropa.eu, consulted on 08/08/2016). to tourism.

671

TIS2017_Book.indb 671 2/8/2017 8:08:36 PM

 The first is a communicative type in which the use • to promote the territorial potentialities accord-
of technologies refers to the diffusion of informa- ing to sustainable modalities;
tion to a wide audience (audio-guides, QR codes, • to enhance tourist typologies that generate low
smart plaques, etc.) using narration. environmental impacts (eco-tourism).
The second is an educational type in which the
use of technologies refers to the transmission of The realization of these objectives assumes the
complex information to a specialized audience definition of tools and rules mainly aimed at:
(students, scholars, researchers, teachers, etc.) • verifying the feasibility of the valorization
using augmented reality, visual ambience, inter- project or program;
active reconstructions of virtual environments, • defining criteria for the award of the paths at
immersive virtual environments, etc. local, regional and national levels;
The third type can be defined as customized, as • outlining the use and walkability of the paths;
it refers to the use of mobile devices (smartphones, • delineating the rules for the management, main-
tablets, cameras, etc.) capable of combining geo- tenance and usability of the paths;
graphical data with cultural information to be used • supervising the relationship between the private
on site (via an app, for example). and public sectors.
The development of these technologies is widely
The last point, in particular, underlines the
used in the cultural tourism sector and is increas-
need for cooperation among all the subjects that
ingly linked to the composition of the supply-side,
are interested in the development of the project of
oriented to this type of tourist demand.
valorization. Moreover, cooperation is an essential
The most frequent applications include integra-
condition if we consider that the paths run across
tion with geolocation tools present in the device;
different counties, beyond regional and adminis-
these provide, for example, for proposing a ter-
trative borders. On the other side, tourism devel-
ritorial area as a special route, characterized by a
opment involves different sectors and different
strong central theme (food, history, legend, wine,
stakeholders, and partnership among these sub-
etc.) and allows the users to personalize the path.
jects can foster economic development through
These applications, generally speaking, merge the
tourism (Briedenhann & Wickens 2004).
use of GPS (Global Position System) and Web GIS
In this sense, the planning of paths can be assumed
(Geographic Information System): in other words,
to be an occasion to define networks (physical and
through the GPS, users can visualize their localiza-
immaterial) among all the different subjects and
tion on a geo-referenced map and choose their own
levels involved (Meyer 2004; Mitchell & Hall 2005).
path according their own exigencies (disposition of
The importance of the involvement of all the levels
time, difficulty, interests, curiosity, etc.).
(administrative, economic sector, local population)
The segmentation of the information is par-
and the positive effect derived from cooperation can
ticularly significant in allowing users to customize
be indicated in five main points:
their path, as the afore-mentioned exigencies can
vary deeply by gender, age, composition, cultural • the establishment of a network among poles that
level, etc. In this sense, it is important that informa- might not hold the same attraction individually;
tion is organized in order to satisfy all the different • the enhancement of lesser-known destinations;
typologies of user. Nevertheless, it is also necessary • the increasing of employment and the training
to introduce information regarding the basic serv- of workers in new professional skills;
ice supply system (accommodation, food services, • the diffusion of such new tourist typologies as
leisure activities, etc.) that might be present along slow tourism;
the path. In fact, the density, variety, quality and • the decreasing of negative impacts on natural
organization of this supply system can deeply environments.
influence choices. Furthermore, this information These conditions should improve the distribu-
allows the users to know the path and, thus, to plan tion of benefits derived from the development of
their travel according to their needs (time available, tourism, including for those work categories not
difficulty of the path, personal objectives and moti- directly engaged in the tourism sector.
vations, etc.).
For these reasons, the planning of cultural tour-
ist routes has to be based on the presence of an effi- 3 PRINCIPLES FOR THE PLANNING OF
cient supply system able to support the choices of CULTURAL TOURIST PATHS
the users; the supply system has to be oriented along
at least three orders of objective, in particular: The planning of a cultural tourist path is primarily
a matter of landscape and territory, a project that
• to indicate safety standards for the fruition of the has to consider both the physical and immaterial
paths, both those existing and those to be planned; components of a territorial heritage (Berti 2013).

672

TIS2017_Book.indb 672 2/8/2017 8:08:36 PM

 The complexity of this heritage needs a holistic The Via Francigena represents another famous
and systemic approach in order to achieve the val- case of a linear path that connects England
orization of its potentialities and, at the same time, (Canterbury) to Italy (Rome). The Via Francigena,
to assure the safeguarding of its local identities. in particular, is an example of the effects that a
Such an approach, furthermore, allows us to con- path can trigger on territorial resources that would
sider not only the physical links among the differ- otherwise have not been considered.
ent poles that a path can connect inside a region, The “in stages” type of path is a subtype of the
but also the immaterial relations that the paths can first one, in the sense that it is composed of a net-
establish among their components. work of leg-poles that defines its geometry. These
According to this systemic vision, attention has to poles can be either attractions or facilities strategi-
focus on the relations generated between the differ- cally located along the path (Figs. 4–5) and they
ent poles that the path connects and the territories must respect specific requirements, such as:
it runs across. In this vision, three main dimensions
• locations at a fixed distance from one another
can characterize the project of the path:
(no more than 20 km);
• immaterial: the path can create an immate- • the presence of auxiliary services and facilities
rial link among attractions located in different along the path;
regions, connected by a unique theme; • the availability of these services all year long;
• material: the path creates physical links connect- • the provision of a network of low-cost or free
ing attractions along defined directions; accommodations (hostels, privates rooms, mon-
• uniqueness: the path itself represents an attrac- asteries, etc.);
tion on the basis of the elements that make it • the outlining of procedures aimed at labelling
unique. the accommodation services;
• the definition of agreements to identify users of
In the first case, attractions are not physically
the path and collect data on their characteristics
connected to each other; this allows users to create
and preferences.
their own route.
In the second case, the path constitutes a physi- In the case of a “closed track” path, the origin-
cal link between attractions and has a strong rela- pole and the destination-pole coincide and the path
tion to the territory. itself can constitute an attraction for its structure
In the third case, attractions are part of the path, alone. With this type of path, even more so than
and relations are generated inside the elements that with others, it is important to install an efficient
constitute it. system of signage in order to guarantee safety and
With reference to these relations, we can pro- beneficial conditions of fruition for users. The
pose a first classification of paths as: question of signage, in fact, represents a difficulty
in the planning of paths, especially in regard to the
• linear paths;
necessity of defining standards for homologating
• paths “in stages”;
information (Polci 2009; Zabbini 2012).
• “closed track” paths;
The “multipolar network” path can be consid-
• multipolar network paths.
ered to be a combination of the previous types and
The first type draws a unique and recogniz- is probably the most suitable model for the study
able path through the territory, generally retracing of existing routes and the design of new routes. In
ancient historical ways, proposing a main theme this type of path, the poles are connected to each
(natural, historical, religious, etc.). In this case, the
relationship between the path and the territory is
the result of temporal phases that have character-
ized the territorial history. Normally, the path leads
towards a node-pole that physically represents the
final destination of the route, and from the point
of view of perception, the main goal for users walk-
ing the path. In the most common cases, the path
connects two different poles of attraction (a source
pole and a destination pole).
The Way of Saint James is perhaps the best
known path in the world, and its particular-
ity stems from the fact that it connects different
origin-poles that all lead towards the same final Figure 4. Percentage of preference for the accom-
destination point, located in the city of Santiago modation structures along the Via Francigena.
de Compostela. (Elaboration from TCI 2015).

673

TIS2017_Book.indb 673 2/8/2017 8:08:36 PM

 It will need to be equipped with adequate accom-
modation services in order to host and manage the
incoming fluxes of users.

4 BIG DATA, GIS AND TOURIST PATHS

In recent years, new technologies for collecting,

sharing and managing Big Data represent a power-
ful backing for the identification and the design, in
real time, of routes aimed at different uses and not
Figure 5. Tourist Representation of a path “in stages” least at the use related to leisure activities.
in Molise where some leg-poles coincide with attractions.
(Source: Regione Molise Progetto Leader 2007–2013).
In order to manage and to analyze these data
GIS technology (Geographical information Sys-
other according to complementary relations, even tem) permit to geolocalize the data.
among poles that are hierarchically different. In order to be able to manage and to analyze
In other words, all the elements composing the these databases, there is the need to use the GIS
network are indispensable and complete each other. (Geographical Information System technology,
The hierarchy among poles depends on the role that that compared to the traditional software of alpha-
each pole plays within the network, rather than on numeric data management, offers the possibility of
their physical dimensions. It depends on the ability reading the alphanumeric data and to localize them
that each pole has in triggering cultural, environ- in the space (geolocalize), through the assignment
mental and economic exchanges inside the network. of spatial coordinates. Being able to analyze the
Within the network, we can individualize four data and to localize it position, offers the benefit of
main levels among the different poles: improving the knowledge of the phenomenon both
• the first level corresponds to the origin pole in quality and in speediness (Maguire et al. 2005).
(base pole); The GIS approach and the growing availability
• the second level corresponds to the set of attrac- of sophisticated software allows us to develop a
tive intermediate-poles; Decision Support System able to identify and to
• the third level corresponds to scattered-poles; select the most suitable and realizable solution to
• the fourth level corresponds to the destination satisfy the users and to preserve the original envi-
pole (magnet-pole). ronmental and cultural characteristics.
The GIS application to the fruition of routes
The origin pole is the physical place where the allows the users to choose the path that will be the
path begins and it plays a strategic role, especially more adapt to theirs needs and aspirations.
for all the activities regarding the knowledge and This can be achieved by integrating the informa-
planning of the path. This pole has to be character- tion developed in GIS with the “Mobile App” that
ized by good accessibility and it has to be specialized can be easily downloaded on the personal mobile
in order to provide dedicated services and facilities, devices as additional multimedia data.
for instance: meeting point; presentation and adver- Data and information that these applications
tisement of the path; accommodation services; offer refer to leisure as well as to the augmented
information services; user monitoring facilities; reality and so on. They are largely used for tourism
enrollment of users; specialized equipment, etc. as they allows users to integrate information and
The “intermediate-poles” identify the physi- to localize the attractions within a territorial area.
cal nodes that are equipped to support the use of Data and information, that these applications
the path. These may coincide with smaller urban offer, refer to leisure as well as to the augmented
centers, whose attractiveness stems from their reality. They are largely used for tourism as they
very well-equipped supply system of facilities and allow users to integrate information and localize the
adjunctive services or in a particular monument attractions within a territory (Angelaccio et al.).
localized within the node. These technologies have radically changed the
The “scattered poles” indicate the network of current way of communicating and living the tour-
possible stages in order to split the length and dif- ist experience; they certainly integrate cultural infor-
ficulties of the path. These nodes will be properly mation but especially they play a strategic role in the
marked and equipped so as to ensure their func- improvement of safety and reliability of the paths.
tion within the network. In the context of these considerations, the plan-
The “magnet pole” corresponds to the end of ning of paths can be developed by the adoption of
the path (destination) and it plays a special and a GIS methodology based on different geometries.
symbolic role within the network, as it represents On one side, the previous typologies of poles can
the purpose of the journey. be structured according to a punctual geometry

674

TIS2017_Book.indb 674 2/8/2017 8:08:37 PM

 linked to descriptive information about amenities different countries that the paths run across, gener-
and facilities located at each pole. On the other side, ating new balance and positive effects.
a linear geometry can be generated by connecting The technology platform, then, could support
the data about the characteristics of the track (slope, the physical design of the tourist footpaths, serv-
track fund, travel time, etc.) to each arc of the path. ing the choices of different tourism typologies, on
In a successive phase, the GIS can be imple- one hand, and the implementation of the supply of
mented through additional information that can goods and services connected to the fruition of the
be both graphical and alphanumeric, made avail- paths, on the other hand.
able on an appropriate web site. This allows users In a starting phase, the technological platform
to select information through their smartphone or could allow to both the public actors and private
tablet according their own needs. investors the optimal management of the foot-
As we have already stated, routes represent paths. At the same time, it could allow the final
complex and dynamic systems generating relations users to appreciate the whole system of the present
among and through different territorial elements resources (territorial, social, cultural, historical).
and levels. Even though they cannot be considered Technologically, the e-services can be located on
as exhaustive needing further in-depth research a specialized Web Portal aimed at integrating differ-
development, for the planning of routes compati- ent elements of the present supply system of tour-
bles both with territorial characteristics (physicals ist services. The architecture of the technological
as well as socio-economical) and the exigencies of platform can be articulated into four macro-areas
users some phases can be indicated: (Fig. 6) that, also by using specialized apps, can inte-
grate information for the fruition of the footpaths
• to determine the typology of the route in order network. In particular, the macro-areas refer to:
to understand the potential and critical issues
• territorial promotion,
related to its geometry;
• territorial safety,
• to define the extend of the route;
• institutional cooperation,
• to delimit the intervention considering all the
• local participation.
elements (physical, social, cultural, economic,
etc.) engaged in the planning of the routes; The macro-area of the promotion guarantees
• to consider the possibility of involvement of interactivity between the supply and the demand
local populations and stakeholders; components by using functions aimed at optimizing
• to individualize actors and applicable forms of the territorial fruition. In the framework of this con-
funding (cooperation, project financing, etc.). text this macro-area should contain telematics func-
tions able to ensure interactivity between the supply
operators and the different typologies of tourist user
5 POTENTIAL DEVELOPMENTS AND
(family, senior, adventure, young, etc.). In particular,
RESEARCH APPLICATIONS
in this macro-area, services should be oriented at sup-
porting the users’ choices with additional information
5.1 A technological platform as tool for shared
that can be displayed directly on the users’ mobile
solutions
devices.
Designing tourist routes aimed at promoting the
territorial knowledge through sustainable forms of
mobility necessarily involves different aspects that
are inherently included in the relationship between
tourism and land use.
Tourism, in fact, is characterized by being a cross
activity which involves multiple sectors (public and
private). As such, tourism is an ideal field for testing
models of territorial development based on the inte-
gration between the different interests involved.
By the use of innovative techniques and tools
it is possible to envisage solutions able to drive
tourism and territorial development towards some
shared conditions and to reduce environmental,
social and economic impacts. In this sense, the con-
struction of a technology platform can represent
an operational tool to deal with the “dimension”
of a network or an integrated cooperative system.
In particular, the platform can overcome the Figure 6. Diagram of the four macro-areas composing
lack of coordination and cooperation between the the platform.

675

TIS2017_Book.indb 675 2/8/2017 8:08:37 PM

 The safety macro-area should include a moni- “intorno” alle città intelligenti. Smart City and Digital
toring system for the tracking of users along the Tourism. Roma: Universitalia.
path. In this way, users can be assisted in case of Berti, E. 2013. Itinerari culturali del consiglio d’Europa:
need and can be supported in the selection of the nuovi paradigmi per il progetto territoriale e per il
paesaggio. Alma Tourism, 2013 (7): 1–12, www.alma-
difficulties that characterize the paths. tourism.unibo.it.
The institutional cooperation macro-area is tar- Briedenhann, J. & Wickens, E. 2004. Tourism Routes as
geted to the coordination among public and private a Tool for the Economic Development of Rural Areas
actors involved into the tourist development project. – Vibrant Hope or Impossible Dream? Tourism Man-
The area of local participation should be tar- agement, I, 25: 71–79.
geted to provide information and services to Fistola R. & La Rocca R.A. 2001. The virtualization of
improve the involvement of the resident and of all urban functions. In H. Bakis & W Huh (eds), Geo-
the social levels concerned. cyberspace: Building territories on the geographical
The architecture of the platform would require space of the 21th century, NETCOM, 15 (1–2): 39–48,
Montpellier, France.
further details that could constitute possible Garau C. 2014. From Territory to Smartphone: Smart
research developments. In this context, it seemed Fruition of Cultural Heritage for Dynamic Tourism
appropriate to underline that the platform could Development. Planning Practice & Research, 29 (3):
represent a technical tool to define a “dialog space” 238–255. DOI: 10.1080/02697459.2014.929837.
where cooperation between different interests can International Council on Monument and Sites 1999.
be realized also to improve local identities. International Cultural Tourism Charter Managing
Tourism at Places of Heritage Significance. Adopted
by ICOMOS at the 12th General Assembly in Mexico,
6 CONCLUSION October 1999. www.icomos.org.
La Rocca, R.A. 2010. Soft Mobility and Urban Trans-
formation. TeMA. Journal of Land Use, Mobility and
The demands of tourism are evolving along with Environment 3 (SP): 85–90. Naples: University of
the rise in demand for more conscious forms of use Naples Federico II. DOI:10.6092/1970–9870/125.
of territorial and environmental resources through Maguire, D.J., Batty, M. & Goodchild, MF. 2005. GIS,
the modification of behaviors and lifestyles. spatial analysis, and modeling. Esri Press.
This change is also linked to the social structure Meyer, D. 2004 Routes and Gateways: Key issues for the
composing the present demand within tourism development of tourism routes and gateways and their
that prefers the cultural aspect of the journey. On potential for Pro-Poor Tourism, background paper,
the basis of this change, the rediscovery of cultural output of the ODI project: Pro-poor Tourism Pilots in
Southern Africa, http://www.odi.org.uk.
paths can be an occasion to promote alternative Mitchell, M. & Hall, D. 2005. Rural tourism as sustain-
forms of fruition of territorial resources through able business: key themes and issues. In D. Hall, I.
the use of soft mobility (La Rocca 2010). Kirkpatrick & M. Mitchell, Rural Tourism and Sus-
This paper has tried to underline the strategic role tainable Business, Clevedon: Channel View.
that Italy and, in particular, the regions of southern Orlando, M. 2013. Didattica e turismo 2.0—Nuove tec-
Italy can play, considering both the variety and sig- nologie per la divulgazione del patrimonio culturale.
nificance of its heritage and the recent interest of Storia e Futuro 32 June 2013 http://storiaefuturo.eu
the political and administrative sectors in promoting Papa R. & Fistola R. 2016 (eds). Smart Energy in the
cultural routes. Nevertheless, cooperation between Smart City. Urban Planning for a Sustainable future.
Switzerland: Springer International Publishing.
the private and public sectors has been individual- Papa R. 2013. Editorial Preface Smart Cities: Researches,
ized as a structural condition and a factor of success Projects and Good Practices for the City. TeMA
in the planning of territorial path networks. Journal of Land Use, Mobility and Environment 1
With reference to a holistic approach, the plan- (2013): 3–4 DOI 10.6092/1970–9870/1544
ning of a network of cultural paths has to consider Polci S. 2010. La valorizzazione della Via Francigena. I per-
the role that they can play in the integration of dif- corsi, l’accoglienza, l’offerta culturale. CIVITA, Roma.
ferent territories. In this sense, paths are both an Richards, G. 2003. What is Cultural Tourism? In A. van
immaterial tool of sharing projects and a physical Maaren, (ed.), Erfgoed voor Toerisme. Nationaal Con-
tool of territorial development according to sus- tact Monumenten.
Touring Club Italiano 2015. Il turismo sulla Via Fran-
tainability and resilience models. cigena. Touring Club Italiano, Roma. “Survey sul
turismo lungo la via Francigena” Roma: Centro Studi
TCI, ottobre 2015.
REFERENCES World Tourism Organisation 1985. The State’s Role
in Protecting and Promoting Culture as a Factor of
Akerkar, R. 2012. Big Data & Tourism. In TMRF Report Tourism Development. Madrid: WTO.
edited by Technomathematics Research Foundation. Zabbini, E. 2012. Itinerari culturali e patrimonio intangi-
Angelaccio, M., Buttarazzi, B. & Onorati, M. 2014. bile. In Proceeding of the 32nd National Conference of
Modelli di Turismo e tecnologie Smart Cities per la The Italian Association of Regional Studies (AISRE),
valorizzazione di borghi, giardini, itinerari e territori 1–12, www.aisre.it.

676

TIS2017_Book.indb 676 2/8/2017 8:08:38 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Terminal operator liability

Francesca d’Orsi
Studio Legale d’Orsi—Neri, Roma, Italy

ABSTRACT: The figure of the terminal operator, which can best be defined as the main player in the
complex of port operations within the port area, is still one of the most controversial figures in the context
of maritime law, and more specifically as regards the juridical qualification to be given to the contracts
stipulated with them for the execution of the aforementioned operations and the consequent regime of
liability in terms of the fulfilment of their obligations. Given that the law does not give any definition of
a terminal operator, it does not discipline the contractual circumstances in which the juridical relations
between them and the user who is benefiting from their activities should be included either.

1 INTRODUCTION 28 January 1994 has introduced it into our legal

system, at least at the level of statutory regulations,
The figure of the terminal operator, which can best clearly distinguishing between, on one side, the
be defined as the main player in the complex of legal system of port operations (Article 16), includ-
port operations within the port area, is still one ing all operations regarding goods carried out in
of the most controversial figures in the context of a port, such as the loading, unloading, tranship-
maritime law, and more specifically as regards the ment, deposit and handling of goods in general,
juridical qualification to be given to the contracts which can be performed by specialised companies
stipulated with them for the execution of the afore- if authorised by the Port Authority or, if there is
mentioned operations and the consequent regime no Port Authority, by the Maritime Authority; and
of liability in terms of the fulfilment of their on the other by the successive Article 18, which is
obligations. a public law ruling governing the case of the con-
cession, always granted by the Port Authority or,
otherwise, by the Maritime Authority, of areas
2 DEFINITIONS and wharves (terminals) for the execution of port
operations on the concessionaire’s own behalf
Neither the Navigation Code nor Law no. 84 of 28 or for third parties, in which case the concession
January 1994, reforming the port system under our itself refers precisely to the specific figure of the
legal system, give a definition, even brief, of the so-called “terminal operator”.
terminal operator. The holding of an authorisation for the per-
A contribution in this sense could be that of formance of port operations can, in some cases, be
the 1991 UNCITRAL Convention (the United accompanied by the right to dispose exclusively of
Nations Convention on the liability of Operators port spaces and areas. And in practice, the holding
of Transport Terminals in International Trade) of a concession for the use of state land in a port
which, however, did not enter into force for lack means a port terminal. The company which holds
of the necessary ratifications for that purpose, and both the authorisation and a concession is thus
it is hardly likely to enter into force in the near defined as a terminal company or operator, and is,
future. The definition given by the UNCITRAL in practice, the subject on which all the other possi-
under Article 1 “Operator of a transport terminal” ble authorised companies hinge, in a system which
means a person who, in the course of his busi- can be defined as reticular.
ness, undertakes to take in charge goods involved Therefore, a qualifying element of terminal
in international carriage in order to perform or operators, under Law no. 84 of 28 January 1994,
to procure the performance of transport-related is the fact that they perform port operations within
services with respect to the goods in an area under state-owned areas and at state-owned wharves the
his control or in respect of which he has a right of use of which is granted to them exclusively. Nor-
access or use”. mally, the exclusive use of the areas is connected
As mentioned, the Navigation Code does not with the need to use fixed wharf equipment, the uti-
even mention this figure, whereas Law no. 84 of lisation of which requires the exclusive availability

677

TIS2017_Book.indb 677 2/8/2017 8:08:40 PM

 of port areas, but with increasing frequency port of the monopoly of the technical-nautical services
terminals are assigned to companies whose activity was a hot topic as regards the letter of the law and
does not involve the use of this type of equipment, case law, based mainly on the distinction between
but which nevertheless requires the exclusive use of port operations and technical-nautical services.
areas for storage or for the parking of vehicles. The law, intervening on the problem of the exe-
Precisely because of this exclusivity, Article 16, cution of technical-nautical services, has probably
paragraph 4, letter a), in fact requires the authority brought this knotty question to an end, finding a
appointed for the purpose to assess every authori- definitive solution with the introduction of Article
sation application, first and foremost on the basis 1 of Law no. 186 of 30.06.2000, which has substan-
of the requisites of a personal and technical- tially amended Article 14, 1 bis, of Law no. 84/1994.
organisational nature, and the financial capacity In fact, the law now specifies that: the technical-
and professional standing of the applicant opera- nautical services of pilotage, towage, docking and
tors and companies, which must be adequate for inshore services are services of general interest for
the activity to be performed, and they must also the purpose of guaranteeing the safety of naviga-
present an operational programme and the deter- tion and berthing in the ports where they have been
mination of a staff of workers directly employed instituted. With the introduction of this provision
by the same. Which, in other words, implies an into the legal fabric of Law no. 84/1994, therefore,
assessment aimed at ensuring that only the com- also the legislator has pointed out, once and for all,
panies with greater capacities are authorised to use bringing to an end the knotty question of the dis-
the port spaces, also taking into account the fact tinction between port operations contemplated by
that the port is, in any case, a limited area in which Law no. 84/1994 and the technical-nautical serv-
only a limited number of companies can reason- ices (the so-called navigation ancillary services),
ably operate. Consequently, it becomes essential that port operations are in fact different from the
for the port authority or the maritime authority technical-nautical services (pilotage, inshore serv-
to adopt selective criteria to assess all the possible ices, towage and docking), emphasising that only
applicant companies, and to guarantee access to the execution of these latter, because of the close
the services market and to allow for the execution connection between their performance and the
of port operations only on the part of those with fulfilment of specific public service obligations,
the best credentials, i.e. those with special charac- serves to guarantee the safety of the ports and the
teristics and entrepreneurial capacities, as well as a prevention of damaging events in the ports, thus
development and investment programme suitable justifying, albeit implicitly, a control of the prices
for guaranteeing the return of economic benefits of the same. Furthermore, in order to determine
for the port and for the ancillary businesses that the binding criteria for regulating and disciplining
depend on the port for their survival. port services (not to be confused by the afore-men-
Therefore, in view of the above, the law attributes tioned technical-nautical services) by the port and
to the port or maritime authority also the power maritime authorities, pursuant to the amended
to establish a maximum number of companies Article 16 of Law no. 84/1994, Ministerial Decree
which can be authorised to operate at a port (Arti- no. 132/2001 has been issued (see Council of State
cle 16, paragraph 7) as well as the duration of the opinion of 29 January 2001).
relative authorisations, which will depend on the Having fully considered questions of a general
operational programme proposed by the company, nature, this paper will continue with a discussion
always without prejudice to the port or maritime of the terminal operator, as a legal person, and its
authority’s power to periodically check on respect liability.
for the programmes on the part of the authorised
companies, naturally as well as the power to revoke
the authorisation also in the case of failure to 3 LIABILITIES AND RESPONSABILITIES
respect the programme (Articles 16, paragraph 4,
letter b) and 6). First and foremost, the terminal operator cannot
The terminal operators are, in fact, character- undoubtedly be considered either as merely a dis-
ised by the fact that the subjects which operate patcher or simply a mere depository for the goods
in the port are simultaneously concessionaires of delivered, but it should rather be considered as the
state-owned port spaces and companies author- subject around which the main activities within
ised, pursuant to Article 16, paragraph 3, of Law the port are carried out, the real protagonist of the
no. 84/94, to carry out port operations. port operations (Carbone & Funari 1994; Carbone
Over time, this has led to some perplexity 1995; Midoro 1995).
also regarding the possibility of the creation of Since the law does not give any definition of a
monopolies, and in fact, before this situation was terminal operator, it does not discipline the contrac-
remedied by the law (Law no. 186/200), the subject tual circumstances in which the juridical relations

678

TIS2017_Book.indb 678 2/8/2017 8:08:40 PM

 between the same and the user who is benefiting stances for the terminal operator when it is dem-
from its activities should be included either. onstrated that the subject, its employees and/or
Doctrine has highlighted the mixed nature of representatives or other subjects of which it takes
contracts between the owner of the goods and avail in the execution of its obligations, have taken
the terminal operator, which includes within its all measures deemed reasonably necessary in order
scope both the circumstance of depositing goods to avoid the event or its damaging consequences
(deposit in favour of third parties) and the con- (art 5. Uncitral).
tracting of services, consisting of the performance With regard to delays, the terminal operator/
of port operations on behalf of third parties by the company bears liability if it does not make the
terminal operator (Vincenzini 1989). goods available to the assignee within the term
It has also been maintained that the terminal expressly indicated in the contract, or within a rea-
operator could be included within the sphere of sonable time in respect of the assignee’s legitimate
multi-mode transport, in which the obliged party request for the redelivery.
supplies not only the transport by various means, Substantially, the convention adopts the stand-
but also the activities ancillary to the same and ard principles of international law; this tendency
strictly inherent to the main performance, namely can also be found in the drafting of the terminal
that of the transport (Riccomagno 1998). contracts, and in fact an examination of the con-
Independently of the exact juridical qualification tracts most commonly used in this specific sector
to be given to contracts stipulated by the terminal allows for listing the main obligations of a ter-
operator, as regards the relations between the lat- minal operator/company, which are represented
ter and the third party recipients of the goods, the by activities that are typical of various contracts:
use of so-called Himalaya Clause in the contracts deposit contracts; mandate contracts; services and
in use in international maritime traffic implies that transport contracts, as well as a series of additional
the specific regulations on transport contracts are activities linked to the main activities, such as the
applicable, with specific regard to exoneration from supply of cranes and structures, and the rental of
liability and compensation limits, to the terminal spaces (Carbone 1995).
operator which has supplied its services on behalf Thus, there are several performances to which
of the dispatcher, almost as if it was acting as its a specific discipline is applied, also as regards lia-
auxiliary/delegate (Piras 2008, Potenza 2002). bility. However, the purpose of the terminal con-
In fact, the relationship between the terminal tract is certainly unique since, in view of the fact
operator/terminal company and the maritime ship- that it is extremely common, it can be defined as
ping company should be examined in depth and a socially typical contract, in consideration of the
therefore it should be verified whether it is an agent fact that there is a certain uniformity of content
of the maritime shipping company or a contrac- between the various terminal operators/companies
tor independent of the maritime shipping com- and that it is based on a (typical) services contract,
pany; this distinction, far from being a question of the discipline of which, however, derives from the
mere nomenclature, is important on the legal level, application of the law to the single performances,
regarding the applicability of the waivers and of but within the framework of a complex works con-
the limitations of liability, contemplated in favour tract (Antonini 2008, Riguzzi 2006, Lefebvre et al.
of international maritime shipping companies, of 2008 Gaggia A., 2008).
certain provisions of international law (the Hague One doubt of a practical nature is that which
Visby Rule on bills of lading), even if, in the case of regards the identification of the moment at which
the above-mentioned Himalaya Clause, this prob- the prejudice is caused in order to render applica-
lem is practically annulled by the application of the ble the discipline of one or another type of con-
clause (Dani, 1986). tract of which the actual contract is composed,
Coming to the terminal operator’s liability, also from the viewpoint of the evidence. Accord-
an important contribution has been given by the ing to reason, adhering to the general principle of
Uncitral Convention, whereby the legal discipline the so-called “proximity of the evidence”, it must
of the terminal operator’s liability is strictly con- be maintained that the subject bearing the burden
nected to the definition of the terminal operator/ of proof is the terminal operator/company, also
company given by the Convention. because this latter is certainly the party most inter-
With regard to the liability for loss or damages, ested in proving, for example, the applicability of
liability begins when the goods are taken in charge, a specific discipline rather than others (e.g. the law
therefore it is strictly connected to the effective on transportation, for the compensation limits and
availability, and therefore involving full control, of the term of lapse); however, if it is not actually pos-
the goods on the part of the operator; it is a pre- sible to identify the precise moment in time when
sumed liability similar to that of the carrier (road the prejudice occurred, the general discipline of
or maritime), except for the exonerating circum- Article 1218 of the Civil Code must be applied.

679

TIS2017_Book.indb 679 2/8/2017 8:08:40 PM

 Always from analysing the general nature of the direct damage to containers or other equipment
contracts, one can maintain that it is possible to for transport, ships or other means of transport
establish compensation limits contractually, valid owned by third parties, caused by the insured party
within the limits contemplated by Article 1229 of while the goods are in practice under its custody
the Civil Code, but only in the case of slight neg- and responsibility;
ligence. In the same way, the exoneration for so- indirect damages deriving from or due to mate-
called indirect damages, such as loss of profit, is rial loss or damage caused to the goods or by the
established. equipment used to transport the goods; without
prejudice to the activities and/or risks excluded by
agreement.
4 CONCLUSION The guarantees provide can be extended to errors
and omissions, to the activity of the custodian and/
In conclusion, therefore, from the above analysis, or the warehouse, to damage to the insured party’s
one can maintain that the terminal operator’s lia- own equipment, and third party civil liability for
bility is subjective for presumed negligence, disci- damages to people and property.
plined by Articles 1218 and 1693 of the Civil Code.
Consequently the exonerating circumstances fol-
low the same discipline, without prejudice to the REFERENCES
fact that the terminal operator bears the burden
of proving the specific moment in time at which Antonini A., Benelli G. Carretta M., Gaggia A., Ingratoci
the damaging fact occurred. Similarly, compensa- C., Lobietti C., Morandi F., Pellegrino F., Pruneddu
tion limits and terms of lapse, if contemplated, are G., Rizzo M.P., Sia A.L., Tassinari G. 2008. Trattato
those laid down by law for the single cases. breve di diritto marittimo, Milano.
With regard to the position of the third party Carbone S.M. & Funari F., 1994. Gli effetti del dirtitto
which receives the goods, however, a party which is comunitaio sulla riforma portuale, in Dir. Mar.
extraneous to the contract with the maritime ship- Carbone S.M., 1995. Contratto di trasporto marittimo di
cose, Editore Giuffrè Editore.
ping company, in view of the Himalaya Clause this Dani A. 1986. Persone che possono beneficiare del limite
party benefits from the typical system of the mari- di debito, Diritto Marittimo 1986.
time contract, with the consequent applicability of Gaggia A., 2008 L’operatore terminalista, in Trattato
the relative discipline, whether it bases its action on breve di diritto marittimo (a cura di Alfonso Antonini)
contractual or non-contractual liability, qualifying Trattato breve di diritto marittimo, Milano.
the terminal contract as a contract in favour of the Lefebvre D’Ovidio A. – Pescatore G. Tullio L. 2008. Man-
third party. uale di diritto della Navigazione, giuffrè Milano).
From the insurance viewpoint, considering the Midoro R. 1995. Riflessioni sull’impresa terminalistica
very large variety of obligations imposed on termi- portuale, in Economia e diritto del terziario.
Piras M., 2008 La clausola Himalaya, in Trattato breve di
nal operators/companies, the insurance market has diritto marittimo, II, Milano, 281 ss.
not drawn up standard policies; however, PORT Potenza M., 2002. La clausola Himalaya al vaglio della
OPERATORS’ AND TERMINAL OPERA- giurisprudenza italiana: verso l’estensione a favore del
TORS’ LIABILITY POLICIES can be found on terminal operator del regime di polizza di carico? Dir-
the market, specifically designed to cover the civil itto dei Trasporti 2002, 245 Trib. Genova 20 novembre
liability of the insured party and holding the same 2000.
harmless for what it may be held to pay on such Riccomagno M. 1998 La risoluzione dlle controver-
grounds, in relation to and for damages occurring sie delle controversie nel Trasporto multimodale,
within the sphere of its professional activity, for: Trasporti.
Riguzzi M. 2006. Il contratto di Trasporto Giappichelli
material loss or damage caused to the goods while Torino.
they are under the care, custody and responsibil- Vincenzini E. 1989. Ina realtà portualee una figura giu-
ity or within the sphere of control of the insured ridica nascente: il terminal operator in Dir. Trasp.
party; 1989 II.

680

TIS2017_Book.indb 680 2/8/2017 8:08:40 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Information provision in public transport: Indicators and

benchmarking across Europe

M. Pirra & M. Diana

DIATI, Corso Duca degli Abruzzi, Politecnico di Torino, Torino, Italy

A. Castro
Epidemiology, Biostatistics and Prevention Institute (EBPI),University of Zurich, Zurich, Switzerland
(affiliation during METPEX project: ZHAW—Zurich University of Applied Sciences—Institute of Sustainable
Development, Winterthur, Switzerland)

ABSTRACT: The measurement of the quality of transport services has been the object of an intensive
research activity in the last decades. The EU project METPEX aims at advancing the state of the art in
this crucial research area through a targeted survey proposed in eight European cities. The dataset is then
analyzed to identify latent constructs that can measure the quality of the traveler experience. Several dif-
ferent indicators are proposed to represent detailed aspects related to the perceived quality of transport
services. In this paper, we concentrate on those indicators capturing the quality of information provision
services in different travel stages: before starting the journey, when passing through stations or stops and
while travelling. The results show the added value of separately considering the quality of information
provision at different stages of the journey experience, for which different entities could be responsible
(e.g. infrastructure managers versus service operators).

1 INTRODUCTION AND LITERATURE The present work reports some preliminary anal-
REVIEW yses done to build and test quality indicators accord-
ing to the above research perspective (Diana et al.
Previous scientific work has attempted to measure 2016), especially focusing on those that measure the
the quality of transport services. The EU project quality of information provision processes in public
METPEX (A Measurement Tool to determine the transport systems (METPEX 2015). These analyses
quality of the Passenger Experience) sets the goal have been carried out during the first phase of the
of increasing our knowledge on this topic by focus- research, where the METPEX team tried to deepen
ing on current research gaps and less explored the understanding of the different facets that con-
issues. Namely, METPEX studies how quality is cur in forming the quality perception of a transport
perceived for the whole traveler experience, not just service without, at this step, taking into account the
focusing on the journey made with a specific serv- travelers’ specific profiles and features. Therefore,
ice but taking into account all phases, from the pre- the following analyses are targeted at defining some
trip information acquisition process to the final leg latent variables related to the perceived quality of
to get to destination. Within such framework, the information provision, rather than considering any
ambition is to assess how the quality of individual objectively measurable quantity.
trip legs is combined to form an overall quality The definition of both the measurement tool
assessment of multimodal journeys. Rather than and of the subsequent synthetic indicators, includ-
an industrial and managerial viewpoint, which is ing those presented in this paper, was also based on
usually focusing on single services (or on services past studies in this field. Many articles are found
provided by a given operator) and is adopting in literature which analyze several different fac-
methodologies adapted from marketing research, tors influencing perceived service quality, as also
a transport policy and planning perspective is con- information provision. In the following, we shortly
sidered, where a more holistic view of the system review those more relevant according to this and the
has to be taken. Beyond the much investigated above introduced METPEX perspective. For exam-
quality issues in public transport, the project con- ple, the relation between the overall satisfaction for
siders active transport means (walk, bike) and the a travel trip and the satisfaction for specific aspects
specific viewpoints of special user groups, includ- composing the journey itself has been analyzed by
ing commuters, women and physically challenged de Oña et al. (2013), while their relative importance
individuals. has been explored by Stradling et al. (2007).

681

TIS2017_Book.indb 681 2/8/2017 8:08:41 PM

 In many works the perceived quality of a wide required by passengers should be provided during
range of components is analyzed through different the trip, at stops, and beforehand to plan the journey
methods and techniques (de Oña et al. 2014, Eboli properly (pre-trip information) (Eboli & Mazzulla
& Mazzulla 2011, Tyrinopoulos & Aifadopoulou 2012). Consistently with such framework, the indi-
2008). In these works, the contribution of both sub- cators defined in the following sections will be based
jective and objective aspects is considered. Among on all these aspects too.
the different aspects that users have to evaluate, Some examples of quantitative indicators that
information provision is certainly an aspect affect- are based on the evaluation of the information
ing transit service quality. In fact, “passengers need provision can be found in literature. In Nathanail
to know how to use transit service, where the access (2008), an indicator of the passenger information
is located, where to get off in the proximity of their during the trip is defined. It is graded by a trained
destination, whether any transfers are required, and checker and it is based on the type and quality of
when transit services are scheduled to depart and information provided at five, arbitrarily selected,
arrive” (Cirillo et al. 2011). The importance of the stations along the train itinerary. Then, the itiner-
information provision is found in the interesting ary is evaluated according to the average grade of
qualitative study proposed by Beirão & Sarsfield all five stations. In Eboli & Mazzulla (2011), a more
Cabral (2007) to understand the travelers’ attitudes objective indicator of the attribute “availability
towards transport. In fact, several respondents think of schedule/maps on bus, and announcements” is
that the bus system, in this case that of the metro- proposed. It is defined as the ratio of the number
politan area of Porto, is difficult to use and infor- of vehicles with functioning information device on
mation is difficult to obtain. In specific, bus users board to the total number of vehicles sampled in a
notice that problems occur when the bus company certain time period. The functioning of these infor-
changes timetables or routes and no information is mation devices on different days is monitored by a
provided to travelers. In recent years, many transit trained checker. The attribute “availability of sched-
agencies have realized that that service information ule/maps at bus stops” is evaluated too. A score in a
is important for travelers and can be effectively used scale [0–10] is assigned to each stop of a line, where
to increase ridership potentially attracting new rid- 0 goes to a stop without any kind of information
ers to the transit system (Eboli & Mazzulla 2012). device on the place and 10 to one with schedule and
In an ideal perspective, the user should be able to maps. The indicator is then the average value of the
access information at every stage of its journey. This scores assigned to all the line stops.
is fundamental in the case of multimodal travel and However, beyond such examples, most of works
it is a key point affecting customers’ modal choice on literature concentrates on the subjective evalu-
(Grotenhuis et al. 2007). Through pre-trip informa- ation given by travellers on some specific quality
tion the user could plan routes and connections, issues, which is also a specific focus of METPEX.
knowing the location of the nearest bus stop, fares, In de Oña et al. (2014) a three step approach, which
routes to the desired destination, possible transfer aims at monitoring the transit service quality of
locations, time of departure and approximate jour- the metropolitan area of Granada over the years,
ney duration. Schmitt et al. (2015) observe that this is developed. Passengers are asked to evaluate 14
aspect has a strong influence on the customer’ mode rather general service features, such as information,
choice, mainly during unfamiliar transit travels. frequency and punctuality, cleanliness of the vehi-
However, the need of an efficient information pro- cle, or speed of the trip and a Service Quality Index
vision is also relevant for habitual trips. During the is calculated from the passengers’ perceptions.
displacement, the traveler should be able to know the A further key point in understanding the users’
correct vehicle to board and where to transfer and perception according to the METPEX perspective
proper information should be given while onboard is the consideration of the whole journey experi-
in different ways (i.e., by audio announcements, ence, from the origin to the destination of the dis-
by monitors, by personnel, on mobile devices…). placement. This means taking into account elements
These good practices are important for passengers that could provide more satisfaction to the passen-
since they “decrease uncertainty and develop psy- ger, such as a good provision of information before
chological strategies to reduce the adverse effects of the departure, beyond the information both during
uncertainty and to raise feelings of security” (Cheng the trip and at public transport stations or stops,
& Chen 2015). Moreover, return trip information the latter being called “land-side” to contrast it with
should be available at the destination (i.e., departure the former one. Many researchers have addressed
times and changes in route numbers). The availabil- these aspects (Tyrinopoulos & Antoniou 2014, Car-
ity of a wide range of mobile apps equipped with real reira et al. 2013, 2014, Lu et al. 2009) and related
time transit information is triggering many new chal- findings were useful to define the structure and
lenging issues, concerning for example the influence the contents of the survey that is presented in the
they could have on travelers’ behavior (Brakewood following section. Section 3 will then describe the
et al. 2014, 2015). On the whole, the information analyses that lead us to the definition of an initial

682

TIS2017_Book.indb 682 2/8/2017 8:08:41 PM

 set of indicators capturing the underlying dimen- Table 1. Tier-1 categories description.
sions of the quality experience, and will particularly
focus on three indicators related to information pro- Question
ID T1 category name
vision in different phases of the journey. Section 4
will present some benchmarking exercises based on T1_1 Design of stations was adequate for my needs
such indicators, while the concluding section sum- T1_2 Design of transport interchanges (main
marizes the results and offers some perspectives on terminals) was efficient
the next steps of this work in progress. T1_3 Design of transport stops was adequate for
my needs
T1_4 The city supported my mobility needs
2 EXPERIMENTAL FRAMEWORK T1_5 The different modes of transport I used
worked well together
Different group of travelers have different needs
T1_6 My passenger rights (e.g. able to access all
and viewpoints, thus different variables describing transport services) were respected
service aspects that are relevant for different type of T1_7 The overall accessibility of my journey was
users and travel modes have been formulated. This adequate for travellers with additional
involved the definition of more than 500 satisfac- needs
tion questions for the measurement tool related to T1_8 Provision of information on arrivals and
a wide range of aspects of the journey experience. departures was adequate for my needs
Only a selection of these questions was presented T1_9 Public Transport Staff were receptive to my
needs
to each user, namely around 60 of them, according
T1_10 The quality of travel information available
to both the characteristics of the traveler and of the during journey was good
kind of journey under investigation. Unlike most of T1_11 The quality of pre-trip information before
previously reviewed studies, the measurement tool I started my journey was good
is in fact referred to a specific journey rather than to T1_12 The quality of transport infrastructure
a more or less generic opinion about a service. (e.g. whole transport service) during my
Specifically, a multi-level questionnaire was set up, journey was good
where the evaluation and the satisfaction with differ- T1_13 The quality of my ride was good
T1_14 My safety and security while travelling was
ent aspects is checked. The structure of the question-
good
naire can be found in Diana et al. (2016, p. 1166). T1_15 Support for intermodal (e.g. different forms
Firstly, some baseline characteristics of the user are of transport during same journey) travel
analyzed, as socio-demographic ones, together with was provided
some attributes related to the last journey done (ori- T1_16 Recognition of the needs of motorised
gin, destination, travel time, travel means etc.) that is vehicle users
the focus of the remainder questionnaire. Then, the T1_17 Ticket purchasing process was easy to follow
attention goes to the overall journey itself and to the T1_18 Time the journey took was as promised
satisfaction with 21 more general quality components. T1_19 Transport availability was adequate for my
needs
The corresponding 21 questions are named “Tier 1”
T1_20 Vehicle design was suitable for my needs
(T1_x in the following, where x ranges between 1 to T1_21 Value for money of services was good
21) and take the form of a satisfaction rating exercise
using a 5 points ordinal scale, ranging from 1, not
satisfied, to 5, really satisfied. Such T1 questions are travelers in their journeys, has not been analyzed so
presented in the first two columns of Table 1. Some much until now. Among T1 questions, instead, it is
of these indicators are usually found in the litera- considered through T1_2 but also in T1_1, in T1_3,
ture, as highlighted in the previous section. However, in T1_20 and, in a certain way, in T1_12 (Table 1).
some innovative aspects are presented and investi- The passenger experience, center of the METPEX
gated. For example, the evaluation of intermodality project, comes out mainly when it is asked to evalu-
topics, that is the use of different forms of transport ate the meeting of mobility needs (T1_4) and the
during same journey, is introduced thanks to the def- respect of travelers’ rights (T1_6). Niche users opin-
inition of questions such as T1_5 and T1_15. This ions are investigated too, mainly thanks to T1_7 and
is a cross-cutting concept in contemporary transport T1_16 which focus on travelers with additional needs
research and policy, since it is related to a variety of and motorized vehicles users.
strategic issues, ranging from the minimization of Finally, and most importantly for the purpose of
transport-related impacts through the use of an opti- the present paper, information provision aspects are
mal “modal mix”, till the possibility of successfully considered in three distinct questions, namely T1_8,
launching new transport services that are enabled T1_10 and T1_11, respectively related to land-side,
by new technologies whenever potential customers on-trip and pre-trip phases. Information to travelers
do not necessarily stick to more traditional means is in fact supplied in different phases of the journey
(Diana 2010). Also the “design” concept, related experience, and it is important to distinguish per-
to the physical infrastructures encountered by the ceived quality issues at different stages, for which

683

TIS2017_Book.indb 683 2/8/2017 8:08:42 PM

 different entities could be responsible. For example, Table 2. Cronbach’s Alpha of the Tier-2 questions within
due to the vertical separation between infrastructure the 21 Tier-1 categories.
managers and public transport operators, informa-
Question N of Items Sample Cronbach’s
tion systems at stops and on-board the vehicles might ID (T2 questions) size Alpha
be designed and provided by different entities.
In the following sections of the questionnaire, it T1_1 12 67 0.950
is asked to the user to concentrate on the longest T1_2 13 67 0.949
trip leg among those composing the journey under T1_3 13 67 0.958
investigation. In this part, the interviewee must eval- T1_4 14 139 0.965
uate some aspects of his/her displacement according T1_5 14 96 0.951
to the same 5 points scale to have uniformity with T1_6 14 73 0.967
the previous section. The set of asked questions T1_7 14 45 0.968
was personalized according to both the mode used T1_8 10 64 0.963
T1_9 11 48 0.946
in this leg and on the kind of user, on the basis of
T1_10 17 110 0.980
the answers given in the baseline part. Overall, 305 T1_11 14 119 0.965
questions were thus organized in 10 different mode- T1_12 13 107 0.957
specific sets and 11 different user group-specific sets, T1_13 7 192 0.918
each set comprising around 15 questions. T1_14 12 94 0.933
The last part of the questionnaire is named “Tier T1_15 13 100 0.968
2” (T2 in the following) and its questions are organ- T1_16 12 109 0.940
ized in 21 sets, each set representing a deeper assess- T1_17 14 45 0.965
ment of one of the above introduced “T1” quality T1_18 12 96 0.965
aspects. The interviewee had to answer to only one T1_19 13 126 0.961
set, being selected depending on the previous answer T1_20 13 45 0.913
to a control question. This structure is intended to T1_21 13 71 0.947
enable the analyst to acquire both general infor-
mation about the journey, as well as very detailed T1 and T2 questions and we analyze the relation-
information about specific aspects whilst maintain- ships and the latent factors that lie underneath the
ing a practical limit on the number of questions a answers given by the users. This is a preliminary
participant is asked, thereby responding to the con- work to the definition of proper indicators that
cern about developing an overlong survey tool. The could help in understanding the real feeling of
number of T2 questions asked for each T1 compo- users about the whole journey experience.
nents is shown in Table 2 (second column).
The surveying period began on September 15th, 3 ASSESSMENT OF THE INDICATORS
2014 and ran until October 29th, 2014, in eight dif- THROUGH CRONBACH ALPHA AND
ferent European cities: Bucharest (Romania), Cov- PRINCIPAL COMPONENT ANALYSIS
entry (UK), Dublin (Ireland), Grevena (Greece),
Rome (Italy), Valencia (Spain), Vilnius (Lithuania) In a first explorative test, the consistency of the
and Stockholm (Sweden). Additionally, the ques- hypothetical summated rating scales defined by
tionnaire was distributed by FIA (Fédération Inter- jointly considering all variables within each of the 21
nationale de l’Automobile) through their national T2 groups has been tested by using the Cronbach’s
networks of motorists to reach a wider target of Alpha analysis (Cronbach 1951). This analysis helps
both transit and private means users. Overall, to check if a certain group of variables can be con-
6,360 responses have been collected through a vari- sidered as a unique indicator of an underlying uni-
ety of surveying protocols. After the data had been dimensional construct, or if it is representative of
cleaned, the total number of valid observations more than one latent variable. In order to have a suf-
was 6,276: 1,472 responses from paper-and-pencil ficient number of observations, the whole dataset
method, 3,805 responses from on-line web survey comprising data collected in all demonstration sites
method, 289 responses from the navigator app, 420 was jointly considered for this analysis. Data groups
responses from game app and 290 responses from with Cronbach’s Alpha values above 0.7 are consid-
focus group method. More information concern- ered acceptably consistent (Cureton & D’Agostino
ing the data acquisition can be found in Susilo 1983). As the last column of Table 2 shows, all
et al. (2016), while the questionnaire contents and T2 groups are reliable and the variables are consist-
the implementation of the survey are presented in ent within the groups since the Cronbach’s Alpha
METPEX (2013) and in Diana et al. (2016, 2017). value is around 0.9 in all groups. Therefore, accord-
For the sake of briefness, we do not present ing to this test, no item can be removed in order to
here descriptive statistics on the sample; those are increase the consistency of the data.
available in METPEX (2014). In the following, As further step, in order to reduce the number
we concentrate on the data gathered through the of T2 questions representing the T1 categories, a

684

TIS2017_Book.indb 684 2/8/2017 8:08:42 PM

 Principal Component Analysis (PCA) is performed. (Table 2). By looking at the overall values in Table 3,
The PCA is a variable reduction technique that aims we can therefore conclude that PCA can provide val-
to minimize the number of representative variables uable insights if applied to our dataset.
(components) that show the highest possible variance From this preliminary analysis we can conclude
of the observations (Bishop 2006). An Explorative that there could possibly be more than one underly-
Factor Analysis (EFA), an alternative variable reduc- ing dimensions for six out of the above 21 groups,
tion technique, might also provide interesting infor- namely T1_5, T1_6, T1_7, T1_12, T1_17 and
mation since it identifies “underlying constructs that T1_20. On the other hand, each of the three sets of
cannot be measured directly” (Suhr 2003). However, questions pertaining to the quality of information
EFA has been rejected for this reduction of variables provision can be represented by a unique latent vari-
because it requires additional statistical assumptions able. We present those three sets in Tables 4–6. It can
concerning the correlation of factors and error terms be noticed that indicator T1_10 is more heterogene-
that might not be fulfilled in the METPEX database, ous than the others, since Table 5 contains items per-
due to the small number of observations that are taining to both on-board the vehicle and at stations
available for some of the variables. and stops. Yet a unique component can represent all
As a side note, the METPEX data of T1 and questions also in this case, according to both Cron-
T2 questions are ordinal, since they are expressed bach alpha and Principal Component Analysis.
through rating scales. Although PCA is originally
designed for interval data, this variable reduc-
tion technique can also be applied to ordinal data 4 CORRELATION AMONG
(Kolenikov & Angeles 2004). Furthermore, since INFORMATION QUALITY INDICATORS
no multicollinearity was found among the T2 ques- AND BENCHMARK ACROSS
tions, a varimax rotation was selected for the PCA, DIFFERENT CITIES
in order to ease the interpretation of the factors.
When performing the PCA, the first coefficient In T1 categories whose corresponding T2 repre-
checked is the Kaiser-Meyer-Olkin measure, which sentative questions are represented by only one
varies from 0 to 1. Values close to 0 indicate that factor according to Table 3, summated rating scales
the sample is inappropriate for the factor analy- are created by summing up the satisfaction scores
sis due to diffusion in correlation patterns, values of the T2 questions pertaining to the same T1
around 0.5 can be considered acceptable sample category (Spector 2006). This process can therefore
conditions and values close to 1 indicate suitability be applied also to the three indicators reported in
(Cureton & D’Agostino 1983). Table 3 summarizes
the results for the Kaiser-Meyer-Olkin measure in Table 3. Main parameters of the factor analysis within
21 groups of Tier-2 questions.
the 21 T2 groups. Thanks to the fact that values
goes from 0.848 to 0.946, the sample is said to be Sample Kaiser- Components
suitable for factor analysis. size Meyer- with
Furthermore, eigenvalues of an item analyzed in T1 of T2 Olkin Average eigenvalue
a factor analysis represent the variance associated to Category questions measure communality above 1
this item, while communalities indicate the common
T1_1 89–112 0.931 0.610 1
variance. The so-called Kaiser’s criterion states that
T1_2 88–110 0.925 0.658 1
the number of items that actually represent the whole
T1_3 84–103 0.920 0.639 1
group after a reduction corresponds to the number of T1_4 240–311 0.946 0.660 1
factors with eigenvalues above 1 if three conditions T1_5 150–180 0.924 0.695 2
are met: 1) there are less than 30 variables, 2) aver- T1_6 89–142 0.866 0.696 2
age communality is higher than 0.7 and 3) the sam- T1_7 59–81 0.848 0.694 2
ple size is higher than 250 (Cureton & D’Agostino T1_8 83–103 0.910 0.699 1
1983). According to this rule, as the last column of T1_9 60–95 0.932 0.665 1
Table 3 shows, six groups could be reduced to 2 items T1_10 159–201 0.937 0.677 1
while the other fifteen could be reduced to just 1 T1_11 146–231 0.935 0.650 1
item. However, only four groups report average com- T1_12 166–196 0.934 0.711 2
munality higher than 0.7, ranging in general from T1_13 209–242 0.911 0.684 1
0.562 to 0.778. Results of eigenvalues in groups with T1_14 149–281 0.930 0.605 1
average communality between 0.6 and 0.7 could be T1_15 125–149 0.940 0.707 1
considered if sample size is higher than 300 (Cureton T1_16 152–191 0.909 0.562 1
T1_17 72–105 0.906 0.670 2
& D’Agostino 1983), which is the case in only one
T1_18 124–200 0.936 0.713 1
group (T1_4). Thus, sample size is relatively small in
T1_19 172–218 0.938 0.646 1
many T1 categories to give a clear-cutting answer, T1_20 89–113 0.871 0.591 2
while requirements of Kaiser-Olkin measures and T1_21 110–180 0.933 0.778 1
average communality are met partially or totally

685

TIS2017_Book.indb 685 2/8/2017 8:08:42 PM

 Table 4. List of rating items included in indicator T1_8 Table 6. List of rating items included in indicator T1_11
(land-side information provision). (pre-trip information provision).

T1_8: Provision of information on arrivals and departures T1_11: The quality of pre-trip information before I started
was adequate for my needs my journey was good

Design of timetable information Accessibility of pre-trip information

Knowledge of staff about your transport needs Accuracy of fare information on the web/apps
Lighting of written information Accuracy of route information on the web/apps
Public address announcements for arrivals Accuracy of timetable information on the web/apps
Public address announcements for departures Comprehensiveness of information provided
Quality of information on mobile devices Ease with which seat reservations can be made
Reliability of real-time information on arrivals at transport Ease with which you can speak to transport staff in advance
stops of your journey
Reliability of real-time information on departures at trans- Level of information enabling you to get to the station or
port stops transport stop
Reliability of real-time information on station arrivals Level of information enabling you to make connections
Reliability of real-time information on station departures Understandability of information provided
Provision of information in different languages
Table 5. List of rating items included in indicator T1_10 Provision of service information in multiple formats
(on-trip information provision). Quality of service provided by call centre staff
Advanced Information on planned service disruption
T1_10: The quality of travel information available during
journey was good

Adequacy of information for onward journey planning rho. Spearman correlation rather than Pearson
Clarity and ease of use of audio information correlation is selected for the calculations because
Clarity and ease of use of information displays although factor scores are metric variables, T1 ques-
Clarity and ease of use of map information tions are ordinal. As explained previously, T2 ques-
Clarity of directional information tions have been designed to deepen the knowledge of
Clarity of warnings and hazards during journey the corresponding T1 quality component, therefore
Clearness of traffic information announcements in stations/ we would expect high correlation values. Regarding
transport stops
the three indicators of interest here, correlations are
En route information on mobile devices
Information updates which give sufficient time to change
higher for T1_8 (0.638) than for T1_10 and T1_11
journey (e.g move to new platform) (0.387 in both cases) and they are all statistically sig-
Reliability of information provided by transport staff nificant at the 0.01 level. Concerning T1_10, lower
The accuracy of fare information at stations correlations could be an indication of the higher
The accuracy of fare information at transport stops heterogeneity of questions composing the scale that
The accuracy of timetable information at stations was already previously noted. The finding related to
The accuracy of timetable information at transport stops T1_11, instead, points at the inherent complexity of
Clarity of travel information aspects pertaining to pre-trip information, that can
Provision of information in different languages be provided through a variety of channels and for
Provision of service information in multiple formats different purposes, ranging from fares to schedules,
from directions to reach service points to the man-
Tables 3–5 to come up with three different scales agement of unexpected events.
representing the perceived quality of information It is finally interesting to conduct a benchmark-
before starting the trip, when arriving at stations or ing exercise to check the mean values taken by these
public transport stops and while travelling. We are three indicators in some of the aforementioned cities.
aware of the potential shortcomings of such data We preliminarily notice that on one hand the three
treatment, due to the fact that ordinal rather than indicators have a different number of items, on the
metric variables are considered here; however we other some ratings are missing due to non-response.
notice that scales are commonly developed through In order to make the values of different indicators
such process in social sciences (Spector 2006). comparable and to consider also observations where
Additionally, the process is robust when using scales only part of the responses are available, the values
with at least five points (Bollen & Barb 1981). More of the scales has been normalized on the 0–1 range,
complex aggregation methods, tailored on ordinal where 0 indicates that the respondent has given the
data, have also been proposed and already applied minimum score to all questions s/he answered and
in the transport sector (Diana et al. 2009). 1 the maximum. Additionally, we do not consider
A correlation analysis is performed between these observations where less than three items from those
summated scales and the values of the correspond- composing an indicator have been answered.
ing T1 question. Since the variables are ordinal, It is also important to stress that, due to the pre-
Spearman’s rho is calculated instead of Pearson’s viously introduced survey design, no respondent

686

TIS2017_Book.indb 686 2/8/2017 8:08:42 PM

 5 CONCLUSIONS
The main target of the European project METPEX
is the study on how quality is perceived for the whole
traveler experience. The main goal is to build proper
indicators that could provide useful information on
users’ perception and that could help policy makers
in providing inclusive, passenger-oriented integrated
transport systems that are accessible by all citizens.
More than 6300 interviews have been collected in
Figure 1. Normalised mean scores for three informa- late 2014 in eight European cities to gain a rich data-
tion quality indicators in different cities. set that could help in creating and understanding the
latent factors underlying the quality components.
answered to more than one T2 questions set, there- In this paper, our attention was on the 21 qual-
fore ratings for the three different indicators are ity issues asked in the survey (Tier-1 part) and on
coming from different individuals. Additionally, the questions representing a focus on a particular
given the high number of T2 questions sets, sample sub-set of travel satisfaction aspects (Tier-2 part).
sizes in each one are relatively small, as reported in We searched for latent factors hidden in them that
the second column of Table 3. Therefore, in the fol- could more thoroughly represent the correspond-
lowing we are presenting mean values and stand- ing T1 categories. Cronbach alpha followed by
ard deviations only for those five cities, among the Principal Component Analysis allowed us to find
above listed eight, where at least 10 observations the most important components, which are two for
for each indicator are available. With such thresh- 6 cases over the 21 quality issues presented in T1.
old, only two out of three indicators could be com- Among those components, we focus then on three
puted for one of those five cities. indicators that can measure the perceived satisfac-
The results are showed in Figure 1. The first tion of information provision services in different
five groups of bars report the standardized mean phases of the journey experience, namely before
values of the three indicators for such five cities, starting the journey, when going through public
whereas the sixth group reports the mean values transport stations or stops and while travelling.
over all eight cities where the survey was deployed. The correlation between the three indicators and
The last two groups of columns report the mean the corresponding synthetic “T1” questions is high
scores for FIA motorists and the overall values, but far from perfect: this shows us that such indica-
averaged both over the eight cities and the FIA tors are giving us richer information that cannot be
observations. It can be noticed that FIA motor- gathered with a simple question, even if some of
ists were relatively more satisfied than those that the underlying rating items are very detailed and
answered in cities, so that the sixth group of bars maybe overlapping. Finally, a benchmarking exer-
would probably represent a correct term of refer- cise has shown the advantages of considering three
ence to evaluate the performances in individual separate rather than a unique indicator. The provi-
cities. sion of public transport services in most countries
It is insightful to compare indicator mean val- is by now the outcome of a complex process where
ues pertaining to the same city, but also to make several different entities interact; similarly, inform-
comparisons among cities concerning the differ- ing both prospective and actual passengers might
ent satisfaction patterns. In particular, cities “B” be a task falling under the responsibility of several
and “E” score relatively higher concerning the different actors in the system. An analysis of the
quality of land side information and lower con- patterns of the values of the three indicators across
cerning pre-trip information quality, while city some of the eight METPEX cities has shown the
“C” is showing the opposite trend. Too detailed added value for policy makers of the proposed sets
comments on such aspects would infringe the of indicators. Some cities perform relatively better
anonymity of this benchmarking exercise, how- in the provision of information before starting the
ever we can comment that this is clearly reflect- journey, whereas others receive higher satisfaction
ing the different organization of public transport ratings in the travelling phase.
in such cities, particularly concerning the amount Future research work will more systematically
of staff that is present on-board the vehicles and explore the complete set of indicators stemming
the efforts put in setting up a modern information from the Tier-2 questions, beyond information
provision system through the web. In summary, by provision issues, and will study how such issues
using such indicators, policy makers can therefore are viewed by individuals with different socio-
realize how they should priorities their efforts in demographic background. Transport quality pro-
order to improve customer satisfaction concerning files are also being developed to more effectively
information provision. communicate the results of the analysis to stake-

687

TIS2017_Book.indb 687 2/8/2017 8:08:42 PM

 holders (Diana et al. 2017). A measurement model Diana, M., Duarte, A. & Pirra, M. 2017. Transport quality
that is linking each indicator with the overall sat- profiles of European cities based on a multidimensional
isfaction rating for the whole journey experience is set of satisfaction ratings indicators. In Transportation
Research Board 96th Annual Meeting, Washington DC,
also foreseen through the use of Structural Equa- USA, January 2017.
tion Models (SEM). This would allow understand- Diana, M., Pirra, M., Castro, A., Duarte, A., Brangeon,
ing the relative importance of different aspects V., Di Majo, C., Herrero, D., Hrin. G.R. & Woodcock,
studied during the project to shape the overall sat- A. 2016. Development of an integrated set of indicators
isfaction for a journey experience. to measure the quality of the whole traveller experience.
Transportation Research Procedia 14: 1164–1173.
Diana, M., Song, T. & Wittkowski, K.M. 2009. Study-
ACKNOWLEDGMENTS ing travel-related individual assessments and desires by
combining hierarchically structured ordinal variables.
This study is part of the METPEX (a MEasure- Transportation 36(2): 187–206.
ment Tool to determine the quality of the Passen- Eboli, L. & Mazzulla, G. 2011. A methodology for evalu-
ger EXperience) project (www.metpex.eu), which ating transit service quality based on subjective and
has received funding from the European Union’s objective measures from the passenger’s point of view.
Transport Policy 18(1): 172–181.
Seventh Framework Programme for research, tech- Eboli, L. & Mazzulla, G. 2012 Performance indicators for
nological development and demonstration under an objective measure of public transport service quality.
grant agreement no 314354. European Transport 51: 1–21.
Grotenhuis, J.-W., Wiegmans, B.W. & Rietveld, P. 2007. The
REFERENCES desired quality of integrated multimodal travel informa-
tion in public transport: Customer needs for time and
Beirão, G. & Sarsfield Cabral, J.A. 2007. Understanding effort savings. Transport Policy 14: 27–38.
attitudes towards public transport and private car: A Kolenikov, S. & Angeles, G. 2004. The Use of Discrete
qualitative study. Transport Policy 14(6): 478–489. Data in PCA: Theory, Simulations, and Applications
Bishop, C.M. 2006. Pattern Recognition and Machine to Socioeconomic Indices. In Proceedings of 2004 Joint
Learning, Information Science and Statistics. New York: Statistical Meeting, Toronto, Canada, August 2004.
Springer-Verlag. Lu, A., Aievoli, S., Ackroyd, J., Carlin, C. & Reddy, A.
Bollen, K.A. & Barb, K.H. 1981. Pearson’s R and coarsely 2009. Passenger environment survey: representing the
categorized measures. American Sociological Review customer perspective in quality control. Transportation
46(2): 232–239. Research Record 2112: 93–103.
Brakewood, C., Barbeau, S. & Watkins, K. 2014. An exper- METPEX. 2013. Development of standard format for the
iment evaluating the impacts of real-time transit infor- measurement instruments. Project Deliverable D3.1.
mation on bus riders in Tampa, Florida. Transportation METPEX. 2014. Report on survey results and behavioural
Research Part A 69: 409–422. analyses from each location. Project Deliverable D4.2.
Brakewood, C., Macfarlane, G.S. & Watkins, K. 2015. The METPEX. 2015. A comprehensive set of quality and
impact of real-time information on bus ridership in New accessibility indicators for transport services. Project
York City. Transportation Research Part C 53: 59–75. Deliverable D5.2.
Carreira, R., Patrício, L., Jorge, R.N. & Magee, C. 2014. Nathanail, E. 2008. Measuring the quality of service for
Understanding the travel experience and its impact on passengers on the Hellenic railways. Transportation
attitudes, emotions and loyalty towards the transporta- Research Part A 42: 48–66.
tion provider - A quantitative study with mid-distance Schmitt, L., Currie, G. & Delbosc, A. 2015. Lost in transit?
bus trips. Transport Policy 31: 35–46. Unfamiliar public transport travel explored using a jour-
Carreira, R., Patrício, L., Jorge, R.N., Magee, C. & Eikema ney planner web survey. Transportation 42:101–122.
Hommes, Q.V. 2013. Towards a holistic approach to the Spector, P.E. 2006. Summated rating scale. In Jupp, V. (ed.),
travel experience: A qualitative study of bus transporta- The SAGE Dictionary of Social Research Methods: 295–
tion. Transport Policy 25: 233–243. 297. London: SAGE Publications.
Cheng, Y. H. & Chen, S.Y. 2015 Perceived accessibility, Stradling, S., Anable, J. & Carreno, M. 2007. Performance,
mobility, and connectivity of public transportation sys- importance and user disgruntlement: a six-step method
tems. Transportation Research Part A 77: 386–403. for measuring satisfaction with travel modes. Transpor-
Cirillo C., Eboli, L. & Mazzulla, G. 2011. On the Asymmetric tation Research Part A 41(1): 98–106.
User Perception of Transit Service Quality. International Suhr, D.D. 2003. Principal Component Analysis vs. Explor-
Journal of Sustainable Transportation 5(4): 216–232. atory Factor Analysis. SUGI30, Statistic and Data Anal-
Cronbach, L. 1951. Coefficient alpha and the internal ysis: 203–230.
structure of tests. Psychometrika 16: 297–334. Susilo, Y.O., Woodcock, A., Liotopoulos, F., Duarte, A.,
Cureton, E.E. & D’Agostino, R.B. 1983. Factor Analysis: Osmond, J., Abenoza, R., Anghel, L.E., Herrero, D.,
An Applied Approach. Lawrence Erlbaum Associates. Fornari, F., Tolio, V., O’Connell, E., Markucevičiūtė, I.,
de Oña, J., de Oña, R., Eboli, L. & Mazzulla, G. 2013. Per- Kritharioti, C. & Pirra, M. 2016. Deploying traditional
ceived service quality in bus transit service: A structural and smartphone app survey methods in measuring
equation approach. Transport Policy 29: 219–226. door-to-door travel satisfaction in eight European cities.
de Oña, R., Eboli, L. & Mazzulla, G. 2014. Monitoring WCTR 2016. Shanghai, China, July 2016.
Changes in Transit Service Quality over Time. Procedia Tyrinopoulos, Y. & Aifadopoulou G. 2008. A complete meth-
- Social and Behavioral Sciences 111: 974–983. odology for the quality control of passenger services in the
Diana, M. 2010. From mode choice to modal diversion: a public transport business. European Transport 38: 1–16.
new behavioural paradigm and an application to the study Tyrinopoulos, Y. & Antoniou, C. 2014. Public transit user
of the demand for innovative transport services. Techno- satisfaction: Variability and policy implications. Trans-
logical Forecasting & Social Change 77(3): 429–441. port Policy 15: 260–272.

688

TIS2017_Book.indb 688 2/8/2017 8:08:42 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Public transport resilience during emergency: A simulated

case in Torino

F. Deflorio, H.D. Gonzalez Zapata & M. Diana

Department of DIATI, Politecnico di Torino, Torino, Italy

ABSTRACT: Nowadays detailed Public Transport (PT) data are available thanks to open standards
such as the General Transit Feed Specification (GTFS), introduced to enable information services usually
provided via web. In this paper this dataset has been used, after some conversion and filtering operations,
for accessibility analysis, based on some indicators to measure how the zones of a city are well connected,
in terms of capacity, number of transit options and their efficiency.
The study focuses mainly on the assessment of the consequences of transport resources disruption,
considering a large flooding event as risk scenario, which could generate also the closure of bridges and
then limit the PT coverage. The selected case study is in Torino (Italy), to measure the impact on the
various zones of the city, in terms of variation in their transit service accessibility. Relevant capacity
reductions (more than 90%) are observed at the north-east of the city.

1 INTRODUCTION 2 MEASURING NETWORK

ACCESSIBILITY AND RESILIENCE
In recent years a number of analyses focused on
understanding the behavior of a transport net- Accessibility is a concept difficult to define, even if it
work either public or private, under extraordinary is widely used in different contexts. One of the most
working conditions. The goal is generally to meas- common definitions in the literature was proposed
ure the degree of reactivity and the strategic role by Hansen (1959), according to whom accessibility
of certain connections within the network, thus is “the potential of opportunities of interactions”,
strengthening them in order to minimize service where the concept is associated with the spatial dis-
disruptions. tribution of activities, considering the preference
The type of failures that may occur over a net- of people or companies to overcome spatial sepa-
work due to external agents can be quite varied, ration. In that way, the accessibility is not a com-
ranging from natural disasters such as floods, pletely objective parameter, since it depends on the
forest fires and landslides, to causes of anthropic ability and desires of the users to cover longer or
origin, such as transit incidents or terrorist attacks shorter distances to access to certain service.
within the system. This work will focus on the total An accessibility measure may be analyzed from
or partial disruption of public transport lines, different perspectives depending on the application.
without considering the possibility of still operat- Different categorizations can be found in litera-
ing them through detours. ture, which provide an overview of components of
The main objective is to understand the per- accessibility. For example, Geurs & van Wee, 2004
formance and behavior of the transit transport distinguish four basic perspectives on accessibility:
network under normal and anomalous conditions.
1. Infrastructure-based measures, which analyze
According this aim, a base network model without
the performance of transport infrastructure.
considering disruptions on the system is built and
2. Location-based measures, based on indicators
analyzed, in order to create a general view of the
related to the spatial distribution of activities.
transport infrastructure. In the second phase, serv-
3. Person-based measures at the individual level,
ice disruptions are introduced and some perform-
considering the individual requirements and
ance indexes are proposed and monitored. The
limitations.
following section will lay out the theoretical foun-
4. A utility-based measure, which considers the
dations to define such indexes, based on accessibil-
benefits that people derive from level of access
ity and resilience concepts.
to the spatially distributed activities.

689

TIS2017_Book.indb 689 2/8/2017 8:08:44 PM

 Geurs & van Wee, 2004 also define four basic and parameters, such as network connectivity, reserve
mutually interacting accessibility components: capacity of not affected links or residual capacity
of the affected arcs, influence the network reactiv-
1. The land-use component, given by the spatial
ity and survivability. The system recovery speed
distribution of opportunities.
depends on the spread/absorption capacity of the
2. The transportation component, considering the
network and the complementary management
supply and demand of transport.
tools available for the system.
3. The temporal component, reflecting the avail-
ability of opportunities.
4. The individual component, reflecting the needs, 3 APPLIED METHODOLOGY
abilities and opportunities of individuals.
The analysis of public transit performance is
In general those components interact between
mainly based on the methodology proposed by
them, but the emphasis on some of them depends
Mamun et al. 2013, including some modifications
on the study purpose. For example transport studies
in the definition of the indexes that will be later
concentrate on the transport component of acces-
discussed.
sibility, using transport models based on factors,
such as travel time, reliability, comfort, but ignor-
ing the land-use and individual components of 3.1 Trip alternatives definitions
accessibility. On the contrary, urban planners typi-
The selection of the alternative transit paths for
cally focus on the land use component and less on
any Origin Destination pair is made for a specific
the transport component and differences between
time of the day, considering only the lines which
population segments (Geurs & Östh 2016).
present trips during the analysis period. In this
Beyond accessibility issues, studies related to
case, both direct connections and the alternatives
infrastructure and network resilience have been
which include one line transfer are included. These
popularized in recent years, in part due to the
alternatives are classified as follows:
increase of natural disasters either in the number
and intensity, as well as terrorist attacks in trans- − Direct connections, which include the trips made
port networks (Mcmanus et al. 2007; Henry & using only one transit line from the origin to the
Ramirez-Marquez 2012; Barker et al. 2013). The destination.
concept of resilience is related with the capacity − Direct Transfers, related to the trips made with
of a system to recover after an external disrup- one transfer; in this case the end stop of the first
tion; this capacity determines the time to return at track represents the initial stop of the second.
the initial state, a more resilient system will have a − Indirect Transfers, which refer to the trips with
shorter recovery time compared to one less resil- one line transfer, but in this case, the walking
ient (Henry & Ramirez-Marquez 2012). distance between the transfer stops is also con-
Barker et al., 2013 illustrates the system behav- sidered; the final stop of the first track and the
ior in function of time, where the states through initial stop in the second line are different.
which the entity passes before returning to the ini-
tial or desired state are the following:
3.2 Transit accessibility analysis
− Reliability, governing the system in the absence
Accessibility is first considered to evaluate the exist-
of disruptions; it contemplates the availability
ing transit service, and then its variation is assessed
of service (Zhang et al. 2011).
to analyze the impact generated by disruptions on
− Vulnerability, representing the adverse effect on
the public transport network. Traditionally, acces-
the network performance caused by any interfer-
sibility was calculated as the combination of spa-
ence (Nagurney & Qiang 2007).
tial coverage (related to the proximity of people to
− Survivability, i.e. a mitigation approach to the
stop) and the service characteristics (frequency and
vulnerability pursuing a minimization of the
capacity); however, the transit connectivity (trip
original impact (Westmark 2004).
coverage) is included in order to take account of the
− Recoverability, considering the speed at which
journey comfort (Mamun et al. 2013). Therefore,
the system returns to the original or desired
transit accessibility (Aijl) is measured as the com-
state.
bination of the spatial coverage score at the origin
In terms of transport, network resilience requires zone and the total hourly service capacity (tempo-
robustness to minimize the impact due to disrup- ral coverage) between the origin and destination.
tions as well as a rapid recovery back to normal
operations and performance (Barker et al. 2013). 3.2.1 Spatial coverage
The impact depends on safety margins established Spatial coverage (Ril) is the proportion of the
during the service planning and on system operation area in the origin i served by the transit line l. It is

690

TIS2017_Book.indb 690 2/8/2017 8:08:44 PM

 computed using the ratio between the spatial cov- In real situations, the level of service is conditioned
erage area of a transit line (Bi,l,buffer) and the total by the network congestion. Therefore, it might be
area (Bi,total) (Mamun et al. 2013). useful to consider a demand parameter linked to the
use of the transport service. Since this information
Bi ,l ,buffer is not available in all cases, the original methodology
Rij = (1) proposed by Mamun et al. 2013 uses the population
Biii ,total
of zones. However, this requires the hypothesis that
the population is uniformly distributed and that all
The spatial coverage area is here considered as zones are populated, presenting problems in areas
the area covered by a particular route within a with a large number of transient population (indus-
walking distance of 300 meters from the stops. This trial zones, business districts or even parks). Since the
calculation differs from the original methodology, focus is on the loss of travel opportunities in emer-
where the buffer is calculated around the line and gency events, it was considered more appropriate not
not around the stop. The area covered by each stop to use any parameter related to service demand, in
is calculated in a first step, then, the coverage area order to avoid distortion of results.
of the line in the origin is determined as the sum of
the areas covered by each stop belonging to the line
found in the analyzed zone. 3.3 Travel time estimation

Bi ,l ,buffer ∑B k
k
i ,l (2) The total door-to-door travel time between the ori-
gin i and the destination j is estimated as the aggre-
gation of different components, where components
A limitation of this approach may be the over- change depending on the alternative analyzed. In all
estimation of the spatial coverage of those lines cases the access and egress time are considered and
where stops are very close between them. taken equal to 5 min each one, while in-vehicle time
is calculated through the service data available.
3.2.2 Temporal coverage Additionally, it is included a waiting time at the
The temporal score refers to the available seats per stops (entry and transfer stops) assumed as one-
hour from the origin i to the destination j through half of the scheduled headway. If the scheduled
line l; this parameter is calculated as the product headway of an entry stop is greater than 10 min
of the hourly public transport runs of l from i to j the waiting time is assumed as 10 min. This default
(service frequency) and the capacity of the vehicles value is taken from the Mamun et al. 2013 meth-
on the line. odology and it represents the average waiting
time from the National Household Travel Survey
Sijl VijlU (3) (NHTS) data in the United States.
In that way, the travel time (Tijl) calculated for
a direct alternative considers: access time, waiting
The capacity of the line is a function of the total
time, in-vehicle time and egress time:
number of seats by vehicle (U) and the frequency
of service (Vijl); this information is generally pro-
Tijl accesss + Twait
Tacces Tiinn −vehicle + Tegress (6)
vided by the transportation agencies.

3.2.3 Transport accessibility The travel time for transferred trips includes an
The spatial and temporal coverage are combined additional in-vehicle time for the second line and a
to obtain the transport accessibility (Aijl) for each waiting time at the transfer stop equals to one-half of
origin/destination pair by transit line l connected the schedule headway of the second route, even if the
to the origin: waiting time turns out to be more than 10 minutes.
An additional walking time from the ending
Aijl Ril Sijl (4) stop on the first track to the initial stop on the sec-
ond one, is calculated considering the air distance
between the involved stops and a walking speed of
Then the total accessibility for each i-j pair is
0,7 m/s.
calculated by considering the aggregation of the
Finally, the concept of connectivity decay factor
possible trip alternatives between the zones. It is
(fijl) is introduced in order to take into account the
important to note that in this case the selected alter-
perception of people when selecting the route to
natives are those resulting after the grouping done
travel from one place to another (Eq. (7)).
in function of the line connected to the origin:
L
Aij ∑Al
ijl (5) fijl =
1 + α e − βT
(7)

691

TIS2017_Book.indb 691 2/8/2017 8:08:44 PM

 where α and β are coefficients available in Mamun as well as the transfer alternatives between each
et al. 2013, estimated using information related pair of lines. In that way, each direct alternative
with the transport demand, and L is equals to 1. represents just one connection between the zones,
while in the case of options which includes a line
change, either directly or indirectly, each possible
3.4 Definition of performance indicators
combination of routes is taken as a link alternative,
The public transport network behavior is assessed and in the cases of two partially overlapping routes,
under three different components: capacity, net- which share more than one transfer stop, each
work flexibility and connection efficiency. All of one of them would represent an additional alter-
them related with the zones accessibility and the native linkage. According to this, it is assumed as
performance of network under unusual conditions. more flexible a transfer option which allows diverse
Each parameter has an associated indicator: TOI, change alternatives between the routes associated
CONX, EFF respectively. with a specific trip choice.
This study does not consider the origin/destina-
tion demand matrix. Therefore, in order to evalu- 3.4.3 Efficiency Estimation (EFF)
ate the performance of the network, each zone is The transit efficiency is function of the total travel
considered both as an origin and as a destination time between two zones and the distance between
Thus, a square matrix with the results of each index their centroids. It is used to measure how conven-
is generated, whose dimension correspond to the ient or expensive is a journey in terms of time, since
number of zones studied. Additionally, the analy- travel time is perhaps one of the most influential
sis is made hourly, thereby measuring the variation parameter in the selection of routes within an
of parameters during the day. urban context, even more important than the dis-
tance between places (Balijepalli & Oppong 2014).
3.4.1 Transit Opportunity Index (TOI) In the case of public transport, the time employed
The Transit Opportunity index measures the pos- is not directly proportional to the physic distance
sibility that people have to move from one zone to between the places as in the case of private trans-
another in terms of the total seats per hour avail- port (considering equal speed), where in general
able at a specific time of day (service capacity), the shortest or the fastest path is selected.
modified by the perception of journey alternatives Public transport lines are generally designed
in function of the travel time. According to the to pass through the areas of greatest interest for
definition of the decay function, trips with a travel the population, but in certain cases, some regions
time grater that 60 minutes contribute very little in which are rather close among them require a rela-
the index result (Eq. (8)). tively higher travel time. Therefore, the proposed
index considers a relative coefficient, which is func-
TOII ij ∑A
l
f
ijl ijl (8) tion of the minimum physical distance between the
interest points (centroid zones), determined using
the Dijkstra’s method on the road network graph,
The TOI index is calculated by the combination and the minimum travel time between them meas-
of spatial and temporal factors, summarized as the ured on the public transport network.
multiplication of the Transit access and Transit
connectivity. First, the index for each trip alterna-
tive is independently computed, then, the total TOI 4 SIMULATION OF SERVICE
for each Origin-Destination pair is calculated by DISRUPTIONS
adding all single alternatives between the zones. The
Transit Opportunity Index is also useful to measure The main objective of the transport analysis pre-
the level of opportunity of any origin or destina- sented in this work is to understand the performance
tion into the city; this value can be interpreted as and behavior of any public transport network under
the possibility of entering or leaving a specific zone special work conditions. First, the simulation in nor-
in a specific moment of the day (e.g., hospitals, mal conditions without considering disruptions on
schools, universities, railways stations, etc.). the system is performed to create a general view of
the transport infrastructure. Then, the calculation
3.4.2 Connectivity Index (CONX) process considering the disruptions into the network
The connectivity index measures the flexibility level provides further results for comparison. This proce-
of the public transport network in terms of the pos- dure has been implemented in Python, using QGIS
sible trip combinations between each Origin-Desti- to manage and represent geographical information.
nation pair. The index represents the total number All the transit data used to perform the analysis
of trip options considering both the possible are generally available through the General Transit
combinations of routes (direct or indirect changes) Feed Specification (GTFS), an open standard

692

TIS2017_Book.indb 692 2/8/2017 8:08:45 PM

 widely adopted to publish PT data on internet. The specific approach to simulate disruptions
This information includes the stops and routes, fre- can be feasible if only a network failure point is
quency of service, vehicle classification, etc. (More analyzed, or if the study focuses on few lines of the
information can be found on: https://developers. system, because in this case the input file is manu-
google. com/transit/gtfs/) ally set, then, any input data verification can be
managed by the analyst.
On the other hand, the general analysis considers
4.1 Modeling of network failure
a wider disruption area represented by a shapefile.
The model considers the network as a graph, In this case, the algorithm considers all the public
described by a sequence of nodes (stops) and arcs, transport stops and lines within the affected area,
linked between them through different data struc- filters them and generates the input disruption
tures, which connect the nodes with transport routes, file automatically following the format in Table 1.
and these in turn with the origin and destination Unlike the specific analysis, the classification of
zones. This organization, where there is not a unique failure type (0/1) is made following some param-
data structure which includes the graph information, eters, in order to generate a more realistic analysis:
makes difficult the removal of an element within the
− If the affected percentage of the line is greater
data set, therefore, an “ignore it” approach is chosen
than 80%, the line is assumed as completely dis-
during the calculation, without removing it from the
rupted (Type 0).
graph.
− If the line is not completely disrupted, it is ini-
In this order of ideas, the introduction of sys-
tially assumed as partially damaged and the
tem disruptions to the model is through as an
residual segments are verified. If the residual
established format shown in Table 1.
segments have less than 4 stops (3 arcs), the
− The field “ID” just represents an identification of segment is not considered useful (very short)
the disruption (it is not employed in the model) and it is removed; after that, the percentage of
− The “Type” column specifies the disruption clas- damage is verified again according the previous
sification, which indicates how to consider the condition.
line. There are two interruption types: 0 when
The model makes three validations before con-
the line is completely disabled; 1 if the line has a
sidering a line as potentially available after the dis-
partial failure (in this case the two segments of
ruption; the sequence is shown in Figure 1.
lines are assumed as independent lines not con-
nected between them).
− In the “Shape” field the line codes are intro- 4.2 Measurement of disruption effects
duced; one route is composed at least by two
The effects on the public transport network are esti-
shapes, one for each line direction. In some
mated through the relative variation in the indices
cases, there are more than two shapes, some of
defined (TOI, CONX, EFF and EFFAV, the latter to
them available only in specific hours.
be later introduced). These differences allow visual-
− In the column “From” the last enabled stop before
izing the effects for the various zones, in terms of loss
the onset of the failure must be indicated. Depend-
of capacity, variation on travel time and reduction
ing on the route direction analyzed, each direction
of network flexibility. The variation is calculated as a
will have a different “beginning” stop. This field is
percentage respect to the base scenario (Eq. (9)).
required only if the disruption type is 1.
− Finally, the column “To” represents the first stop
enabled after the disruption. In a similar way,
it depends on the route direction and is only
required when disruption type is 1.
In the network simulation two different types of
failure analysis are contemplated (namely, specific or
general analysis), depending on the disruption exten-
sion or the study aims. In both cases, the above intro-
duced classification of failure types and the standard
structure presented in Table 1 are considered.

Table 1. Established format for disruption data.

ID Type Shape From To

1, 2, …, n 0/1 Shape code Stop code Stop code Figure 1. Decision making structure for the alternatives
chosen (disruption analysis).

693

TIS2017_Book.indb 693 2/8/2017 8:08:45 PM

 indexbase i xddisrupted
inde 5 APPLICATION TO THE TORINO
isrupted
indexvar iation [%] * 100 (9) CASE STUDY
iindex
d base
As a case study, the city of Torino was chosen, due
4.2.1 Transit Opportunity Index Variation to its morphological features that were considered
The Transit Opportunity Index (TOI) variations relevant for the proposed analysis, such as the con-
are related to the transport capacity modification fluence of four rivers within the urban area of the
and the changes on the total travel time. This varia- city, which interact at different levels with the pub-
tion can take both positive and negative values due lic transport infrastructure, representing a poten-
to the way in which its calculation is structured. tial risk factor of the system.
The negative variations in the TOI are associ- Considering the position of these rivers, the most
ated for example, with the loss of a transit line with interesting to analyze is perhaps the Dora river,
a high hourly capacity. This negative variation which crosses the city near to the center dividing it in
helps to determine the most critical links in the net- two parts. The river is mostly channelized along its
work. By contrast, the positive changes are more route in the city and it is crossed by a large number
related with trip time variations. Indeed, if the of bridges, many of them built centuries ago, with
sample connection is selected based on the hourly relatively low clearances, therefore, more suscepti-
capacity, there is the possibility that the alternative ble if floods occur. This river is also connected to
chosen is not the best, considering the travel time, the Mont Cenis dam, for which inundation maps
and the result for TOI in the base analysis is penal- related to its failure are available (Fig. 2).
ized by the time decay function. For that reason,
if this connection is removed, it may penalize the 5.1 Simulation data
accessibility from the capacity point of view, but
improve it in terms of travel time. If the benefit For this analysis, the inundation area shown in
due to the travel time improvement is greater than Figure 2 is considered as the affected zone, accord-
the variation due to the loss of capacity, the TOI is ing to the emergency plan available. The simula-
larger compared to the base scenario result. tion was performed in normal conditions and in
the emergency scenario after the dam failure.
4.2.2 Connectivity index variation In both cases, the same city zoning employed by
This index is directly related to the number of urban demographic studies was used, thus consid-
travel alternatives and it is mainly sensitive to ering 94 zones.
those including any line change. The importance The needed information concerning the Turin
of a public transport line for a specific city zone public transport system is provided on the web by
is measured by the number of line connections the “Gruppo Torinese Trasporti (GTT)” and “5T
lost when the transit route is disturbed. In gen- S.r.l” transportation agencies, through the above
eral, higher hourly capacity lines (e.g. subway lines introduced GTFS standard format. The complete
or trams) have a larger impact compared to low transport network is composed by 1264 shapes
capacity lines, since they have a large number of (grouped in 212 transit lines) and 6965 stops.
line changes along its route.
5.2 Base scenario results
4.2.3 Efficiency indices variation
The variation of the efficiency indices (both the The model uses the travel times included within the
efficiency of the fastest route and the average effi- GTFS database which is not considering real time
ciency) is directly related to any change in travel traffic dynamics, but it rather takes into account
times due to the elimination of one or more con- the experienced vehicle travel times used to set the
nections between the studied areas, and it can be timetable. The indices are calculated on an hourly
associated with the loss of a less efficient or a more
efficient route.
A first index (EFF) estimates the efficiency of
the fastest route; a variation in this parameter
means that the affected public transport line is part
of the shortest path (in terms of time) between the
two analyzed zones. On the other hand, the index
of average efficiency (EFFAV) measures the global
variations in the travel time between zones. This
coefficient might take either positive or negative
percentage values, because it is based on an average
travel time, calculated considering all possible alter- Figure 2. Inundation map in Turin for the scenario of
natives between a specific origin-destination pair. Mont Cenis dam failure.

694

TIS2017_Book.indb 694 2/8/2017 8:08:45 PM

 basis for a whole week, in order to observe both the city, with a greater impact in the direct affected
their hourly and daily variations. However, Tues- zone, but also in the norther part of the city, where
day values on the morning peak hour (7:00 am) are the variation rises more than 90%, whereas in the
only presented here for briefness. south, the reduction is little (less than 20%).
The algorithm calculates the four indices for In the case of efficiency the trend is the same,
each origin-destination pair and, to visualize them but the difference between the north and the south
on the map, it aggregates the results for each zone of the city is much more visible.
as origin or destination. In the case of the Transit According to the information presented on the
Opportunity Index (TOI), the results is measured maps, the impact on the public transport network
in terms of transit capacity and it is shown in terms generates a complete disconnection of the zones
of relative values (Fig. 3), calculated respect to the at the north of the city, affecting both production
maximum absolute value of TOI reached in the and residential zones (Figs. 6–7).
week, equals to 59517seats per hour per zone. In A particular situation is presented in the zones
Figure 3, it is apparent that the transport supply is located on the hills (on the right side of the maps).
concentrated around the city center. Those zones are not directly affected by the inun-
According to the map shown in Figure 4, the dation, but the estimated level of impact suggests
travel efficiency is much higher on the peripheries that the related public transport lines pass through
of the city, having an opposite behavior pattern the north of the city to reach the city center. This
respect to that shown by the transit opportunity can be clearly confirmed by analyzing the connec-
analysis. tion loss shown in Figure 7, which reflects a high
This greater efficiency in the outskirts of the vulnerability of those zones.
city is associated with a decrease of travel time per
kilometer, usually due to a longer distance between
stops as well as more direct routes and possibly to
both less congestion and to the presence of road
infrastructures with higher free flow speed, all ele-
ments enabling higher travel speeds.

5.3 Variations due to PT disruption

The disruption analysis tries to estimate the effects
on the public transport network due to a failure
in the Mont Cenis dam, using as input data the
inundation area shown in Figure 2. The analysis is
made in general mode (see section 4.1) and the dis-
ruption lines are selected automatically. During the
simulation, 335 affected shapes were found (174
partially affected and 161 completely damaged).
In Figure 5 the variation of TOI is presented after
Figure 4. EFF by origin—base scenario analysis (Km/h).
the disruption: there is an evident general impact in

Figure 3. TOI by origin—base scenario analysis (%). Figure 5. Variation of TOI by origin (%).

695

TIS2017_Book.indb 695 2/8/2017 8:08:47 PM

 of the affected lines. More remote affected areas
are nevertheless identified, mainly in the eastern
part of the city. This means that among all travel
alternatives available, at least one of them crosses
the disrupted area. The most relevant values were
found in the north-eastern quadrant, where the
reduction is of more than 90% in terms of trans-
port capacity (TOI). The other indices indicate
a total disruption in the north of the city and in
some eastern zones, representing, in some cases, a
risk of total disconnection, even if the zone is not
included into the inundation area.
Finally, the model was built based on the public
transport system of Torino, but the way in which the
different algorithm’s modules were built makes the
model easily transferable to other cities, provided that
Figure 6. Variation of EFF by origin (%).
the input data are available in the GTFS format.

REFERENCES

Balijepalli, C. & Oppong, O., 2014. Measuring vulner-

ability of road network considering the extent of serv-
iceability of critical road links in urban areas. Journal
of Transport Geography, 39, pp.145–155. Available at:
http://dx.doi.org/10.1016/j.jtrangeo. 2014.06.025.
Barker, K., Ramirez-Marquez, J.E. & Rocco, C.M., 2013.
Resilience-based network component importance
measures. Reliability Engineering and System Safety.
Geurs, K.T. & Östh, J., 2016. Advances in the Measure-
ment of Transport Impedance in Accessibility Model-
ling. EJTIR Issue, 16(2), pp.294–299.
Geurs, K.T. & van Wee, B., 2004. Accessibility evalua-
tion of land-use and transport strategies: Review and
research directions. Journal of Transport Geography.
Figure 7. Variation of CONX by origin (%). Hansen, W.G., 1959. How Accessibility Shapes Land Use.
Journal of the American Institute of Planners, 25(2),
6 CONCLUSIONS pp.73–76. Available at: http://www.tandfonline.com/
action/journalInformation? journalCode = rjpa19.
Henry, D. & Ramirez-Marquez, J.E., 2012. Generic met-
The methodology proposed and applied in this work rics and quantitative approaches for system resilience
offers an alternative to traditional computing acces- as a function of time. Reliability Engineering and Sys-
sibility models for transportation systems based on tem Safety, 99, pp.114–122. Available at: http://dx.doi.
both supply and demand data. Considering that not org/10.1016/j.ress.2011.09.002.
always detailed and reliable demand data are avail- Mamun, S.A. et al., 2013. A method to define public
able, this model only uses information related to transit opportunity space. Journal of Transport Geog-
public transport service characteristics, as well as a raphy, 28, pp.144–154. Available at: http://dx.doi.
city zoning to perform the analysis. The study does org/10.1016/j.jtrangeo. 2012.12.007.
Mcmanus, S. et al., 2007. Resilience Management A
not consider the analysis of the passengers’ distribu- Framework for Assessing and Improving the Resilience
tion within the system, but it determines the weight of Organisations Executive Summary, New Zealand.
of the alternatives in terms of their travel time, which Nagurney, A. & Qiang, Q., 2007. A Transportation Net-
penalizes options that have higher values. This can work Efficiency Measure that Captures Flows, Behavior,
be directly associated to the users’ perception, since and Costs with Applications to Network Component
passengers tend to prefer faster alternatives. The Importance Identification and Vulnerability. In POMS
model quantifies three main parameters: the access 18th Annual Conference. Dallas, Texas, U.S.A., pp. 1–22.
to public transport and its capacity (TOI), the degree Westmark, V.R., 2004. A Definition for Information
of network connectivity (CONX) and the efficiency System Survivability. In Proceedings of the 37th
Hawaii International Conference on System Sciences.
of its connections (EFF). Honolulu, Hawaii.
Concerning the scenario with public transport Zhang, C., Ramirez-marquez, J.E. & Rocco Sanseverino,
network interruptions, as expected, the obtained C.M., 2011. A holistic method for reliability perform-
variations of the three indices show that the impact ance assessment and critical components detection in
is mainly concentrated in the direct influence area complex networks. IIE Transactions, 43, pp. 661–675.

696

TIS2017_Book.indb 696 2/8/2017 8:08:48 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The “form” of railways: Between construction and regeneration

of contemporary landscape

F. Viola
DICEA, Università degli Studi di Napoli “Federico II”, Napoli, Italy

C. Barbieri
DiARC, Università degli Studi di Napoli “Federico II”, Napoli, Italy

ABSTRACT: Transport infrastructures—in particular railways—have a “special beauty”, punctual and

systemic at the same time. On the one hand, infrastructural architecture can be exemplary works of tech-
nical merit which express adherence to practical functionality and the wise use of materials according
to their nature and costs. On the other, each infrastructural work has an “added value”: it belongs to a
complex landscape system of greater scale through which it affects the visual, physical and cultural qual-
ity of the inhabited landscape. The recent closure of many obsolete railway lines makes it necessary to
redefine the landscape character using a new design strategy more attentive to the preservation of identity
characters, to avoiding the homologation of the contemporary landscape. This kind of design strategy
proposes the enhancement of railway architectural characters, thus avoiding a nostalgic reconstruction of
their original form. The goal is to implement a new design paradigm which works with surviving elements
and signs, produced by time, and looking for to ensure successful integration between old and new. When
the engineering works are abandoned they gain a unique character—becoming aesthetically interesting.
Residual elements are also used in many fields of Contemporary Art to give them new life and new social
functions through post-production processes and formal contaminations. These artistic techniques can be
applied to reuse abandoned railways too: hybridization, repetition and serial fittings can generate surpris-
ing and expressive new projects.

1 INTRODUCTION offer to the redevelopment of urban and suburban

areas—originally marginal compared with regional
Over the past decades the railways network went dynamics, but now in a strategic position -, but
through a huge phase of restructuring in all indus- also for the intrinsic qualities of infrastructural
trialized countries due to the changing mobility places characterized by an unexpected and amaz-
needs of goods and people, to the reallocation of ing mix of archaeological finds and wild nature
productive activities in the territory—such as raw (VIOLA 2004). Industrial buildings, domestic and
materials extraction or shipbuilding industry—to rural artifacts, slender pylons, hardy steel portals
which the railroad was bound from birth. suspended on the tracks and massive brick sheds,
The disposal of a large part of the railways are one beside the other in these places with no
affected equally both the major metropolitan areas apparent relationship. The connections between
and the peripheral areas where there is a strong the repertoire of architectures and tracks, on the
competition with the road transport system— one hand, and the set of formal, perceptive and
cheaper and widely disseminated—forcing the clo- symbolic relations, on the other, are the leitmotif
sure of many kilometers of “deadwood”, lines with of our study and the qualifying aspect according
a limited number of users and unsustainable man- to which we intend to suggest some inputs for the
agement costs. A precious legacy of areas and archi- project of transformation of these places.
tecture made of tracks of rails, depots, workshops On the issue of brownfield infrastructure gener-
for the repair of trains, a loto of small stations and ally two different positions have been taken. On the
surveillance tolls, has suddenly been abandoned one hand there were those who only seize the oppor-
and, in many cases, is still waiting to know their tunity of building on abandoned areas. On the other,
fate. It is a set of artifacts with an extraordinary there were those who, by agreeing to compete with-
economic, urban and cultural value, not only for out bias with pre-existing conditions, tried to redeem
the great opportunities that the conversion could the qualities of the places through the redevelopment

697

TIS2017_Book.indb 697 2/8/2017 8:08:51 PM

 project. But along with an extraordinary heritage of
architecture that have marked the history of modern
transportation, the disposal of the railways ques-
tioned also the ways of crossing the space and the
perception of the landscape from these places, devel-
oped over time. A deep transformation of the areas
and buildings often followed the closure of railways:
within a few years these areas and buildings have
been homologated to many anonymous other that
make up the contemporary landscape.
All the traces of a previous infrastructural func-
tion have been deleted and where previously ran the
trains were opened new roads, built other houses,
shopping centers, industrial warehouses.
In Italy the railway network restructuring process
started later than in other European nations—even
for the decreased availability of resources—and the
conversion of railway areas can still be seized as
an opportunity to put in place interventions more
careful to identity of places and useful to trigger
virtuous recovery actions of the many abandoned
sites around the sediments of tracks. The recovery
of abandoned railroads can be an opportunity to
introduce a new architectural and environmental
quality in those marginal places among the city, the
infrastructure and the countryside, where signs of
degradation are more evident. Just this filed, so far
Figure 1. Intersection of tacks.
without significant experiments in planning, is the
topic of this paper.

Figure 2. Elements and legacy of railways.

698

TIS2017_Book.indb 698 2/8/2017 8:08:51 PM

 The hypothesis from which we started is that
the reuse of abandoned railways is an important
opportunity to implement regeneration proc-
esses of degraded or marginal areas compared to
regional dynamics, provided that they respect their
own specific identity, an identity made of recur-
rences and singularity.

2 BETWEEN ENGENEERING
AND ARCHITECTURE

The railways have traditionally been a specific

competence of the engineers and the only occa-
sions when architects had a role in the infrastruc- Figure 3. Railway Parma-La Spezia, 1879–1894.
tural design process were those in which it was
necessary to give a new form to values expressed
by the infrastructure themselves: i.e. the emblem- and give priority to the theoretical and virtual sim-
atic case of stations, that was a favored theme in ulations instead of field testing (BRANZI 2006).
the last century because of the experimentation of In recent years, among architects and urban
new typologies related to the experience of the trip. planners it has spread the belief that the city and
It was established the belief that the construction the territory of the future will be characterized by
of the railways has been exclusively determined by increasingly weak urbanization systems and intan-
technical and economic reasons, indifferent to the gible and enzymatic architectures, without defined
framework conditions in order to make the trans- formal features and material sediment. Further-
port as fast and economical as possible (MUM- more, engineers who deal with infrastructures
FORD 1961). are increasingly attracted by resources and traffic
After years of unconditional faith in the domain flows management rather than—as is their tradi-
of reason over reality, there is today more awareness tion-, the creation of artefacts able to leave tangi-
that vagueness and contradictions are intimately ble traces in the territory.
part of the architect’s work, creating opportunities
for his progress. In fact, the condition of perma-
nent instability is a urge to explore new paths, to 3 THE RE-USE OF RAILWAYS: TECHNICAL
experience more and more advanced expressive FORMS AND LANDSCAPE
approaches, up to reach those extreme limit points
beyond which the initial borrowings principles The topic of the relationship between railway infra-
are contradicted and replaced by new ones. Com- structure and territory, in its most recent meaning
pared to many studies of the recent years about of the re-use and environmental restoration, is
railways, in this paper we will focus especially on today in a complex scenario, in which the architec-
architectural built form that links the railway to ture skills appear wider and less defined than in the
the territory and on the role that it had—and may past: a dilation of the field of work involving both
still have—in landscape construction. From our the issues addressed and the methods and tools
point of view, the attention to the relationship that used in the project. Overcoming the traditional
architecture establishes with the reality of places opposition between the consolidated city and sub-
is important when “reality” is not interpreted as a urbs, between the built and natural space—that
condition to simply join or undergo, but as some- have long dominated the theory and practice in the
thing to positively transform, giving the project an last century -, the opportunities of transformation
active role and not simply a decorative one. have become much more frequent and important
On the contrary, many recent architectural and than the new constructions for architects. The con-
engineering studies about transport infrastructures temporary territory is no longer considered as a
seem to have overlooked the complex constructive neutral space, a void to fill with new objects, but
and type-morphological relationships which tradi- it is conceived as a complex place in which traces
tionally have bind the artifacts to places, focusing of human presence, heterogeneous in consist-
on increasingly evanescent and impalpable aspects. ence and identity, are stratified layer upon layer.
In Architecture, these studies transcend the pur- And in this emphatic fragmentation scenario—in
pose of projects’ realization to make them only which the individuality appears prevalent than the
producer of non-figurative quality; in Engineering, overall design -, the railways are the only elements
they split the skills in increasingly specialized fields that characterize the sites, according to the clear

699

TIS2017_Book.indb 699 2/8/2017 8:08:54 PM

 identity of their architectures, the autonomy of the inhabited space that has many points of contact
tracks compared to the surroundings, and the per- with the «soul-searching of places» elaborated by
ceptual opportunities of the crossed space. Norberg-Schulz in Genius Loci.
The strategy of disused railways recovery, wish- When faced with the trend currently prevail-
ing to protect the quality of places and infrastruc- ing in making indistinct places, the reuse of aban-
tural architectures, is part of the vast opinion doned railways is an opportunity to assert a new
movement that opposed, in the recent years, to and more advanced idea of environmental restora-
environmental desertification policies according tion, able to align the transformation of the terri-
to which humanity seems to have no more place tory with the enhancement of its identity.
to live. The need for protection of places iden- On the contrary, the prevailing attitude over
tity has become increasingly strong since the cri- landscape, especially in our country, systemati-
sis of Reason, in its various forms ranging from cally rejects any suggestion of change, demanding
ecological thinking - that in architecture fueled to match the reality of places with an ideal and
researches for environmental sustainability—to abstract image of them.
the regionalist claims in the political field, up to On the other hand, it’s necessary to define pre-
the formulation of a Geophilosophy in contempo- cisely what constitutes the aesthetic identity of a
rary thought. place to go beyond generic statements of principle,
The Geophilosophical reflection, as written by and it’s necessary to identify the shared characters,
Luisa Bonesio, in front of «a landscape ravaged though not exactly “objectives”, in order to base
by a careless use of technology, where identities on them the hypothesis of transformation.
are deleted from the effects of a senseless logic of Especially for railways, the landscape’s identity
production and consumption», interrogates the is made up of recurrences and singularity. Appar-
territory in view of «a new responsibility and a ently similar tracks and architectures, with similar
full awareness of thinking, of acting and living» typological, formal and constructive features in
(BONESIO 1996). The Geophilosophy is also a different and distant places, but all the more dif-
Physiognomy because it recognizes in every place ferent from each other the stronger the bond that
a particular aspect, an individual expression of unites them to the context and to local memory.
its unique character. It is an appreciation of the The intertwining of repertoires and individual-
character inherently spiritual and symbolic of the ity, on the one hand, and formal, functional and

Figure 4. Güterbahnhof Pankow, Berlin.

700

TIS2017_Book.indb 700 2/8/2017 8:08:55 PM

 symbolic relations between railways and territo- The serial logic used to build railways architec-
ries, on the other, may therefore be the main thread tures often prevents us from recognizing their aes-
of the same strategies of transformation. thetic value, traditionally attributed to individual
The issue of reuse of abandoned railways is often pieces rather than the products appearing as the
addressed in Industrial Archeology, a recently born result of cultural environments or extensive techni-
discipline and full of inner contradictory aspects cal expertise. In industrial production, objects can be
because of the short temporal distance between replicated in thousands of specimens without losing
the artifacts of the first industrial revolution and the original quality of the prototype, while for men
our time, unlike other protected historical arche- it is not possible to reproduce the same objects equal
ologies. Railways, built mainly during the last cen- to each other: human life is made of experiences and
tury, have not yet acquired that status because of the repetition is, in a sense, the end of the experience.
they belong to an age culturally too close to ours to The artisan object shows the secret of the trade and
let us make critic judgments. its tradition through its forms: it is the emblem of
The term “Archeology” is undoubtedly effective the experience gained over a period by human exist-
to describe the nature of railways’ finds: as many ence. For this we are spontaneously attracted by craft
old artifacts, they are objects bearing clear traces objects with a sentimental attitude and it is difficult
of physical consumption due to the use, to the time to love in the same way industrial products because
and to the weather, that are very important fea- they are the effect of a mechanical action. The aban-
tures especially in a design perspective. donment and the degradation by time, however,
As opposed to what happens for other kinds of often transform the serial elements present along the
architecture that draw from Beauty the reasons for tracks (machines, electrical equipment, sleepers and
their survival over time, the abandoned railway rails, metal trellises, lights, etc.), assigning them a
buildings gain this privilege thanks to a renewed new individual character and making them less per-
functional vitality. fect, unique and aesthetically interesting.
Considered in longer timescale, the current phase
4 WORKING WITH THE RESTS of railways infrastructures’ disposal can be seen as
the latest expression of the troubled relationship that
Forms determined by the technical do not have a architecture has established with the industry over
secondary role in transport architectures. Gener- the past two centuries (JÜNGER 1991). What we
ally Engineering is not pleased with its own beauty are experiencing in recent years is the phase in which
and it never worried about having to persuade the the abandoned industrial artifact is judged regardless
public with appearance. However, even engineers of the use for which it was designed, the choice of
know how to grasp the difference between a beau- materials, its solidity and resistance. The industrial
tiful bridge and one less successful. The attribution product has turned into an object to contemplate or
of a different value to build works is not as impor- to rethink in its appearance through post-production
tant for Engineering as a discipline in itself, as it is and assembly operations—as in the works of Kurt
for our culture and for how we assign a different Schwitters. After him, many others—like Fernan-
value to the artifacts that surround us, distinguish- dez, Arman, Rosenthal, Barry, Chamberlain—used
ing between objects that characterize the environ- the scraps, giving them a new life and a social func-
ment of our lives (MANTEROLA 2011). tion through art. The new assembly techniques used,
The principle that railways architectures and man- from “serial assemblies” to “accumulations”, offer
ufacts must be incorruptible, unalterable, emblemati- an extensive repertoire of compositive solutions that
cally reflected in oversizing the constructive elements, can be models for the design of railways recovery.
with generous use of solid materials—such as steel, Beyond the proposed objectives and the personal
stone and brick -, often ostentatious to enhance the artistic career, the use of anonymous industrial
solid character of the buildings in solids details of materials by modern architects has always been
the stations’ furnitures and railway coaches. Some functional to create unique and original works, in
kinds of architectures have been invented specifi- a time when architectural production—from com-
cally for railways, such as tunnels, stations, locomo- puterized procedures used in the design to indus-
tives remittances, coal and water tanks, workshops trialized building components—seems to refer to a
for maintenance of rolling stock, cabins for traffic mechanical assembly operation.
surveillance suspended on the rails. Working “in” and “with” the remains, however,
These are the elements that have contributed the become in recent years an ordinary condition in
most to build the image of a unified system both the architect’s work and, perhaps, the only possible
for their strong formal characterization both fot strategy to offer an adequate response the complex-
the ability to give life to a complex network able to ity of the contemporary. Many argue that our age
communicate—through the serial repetition of ele- is not destined, as in the past, to produce contents
ments along the lines—with landscape in its broad- but post-productions, collecting and regenerating
est extension. all that is already available. This is particularly

701

TIS2017_Book.indb 701 2/8/2017 8:08:58 PM

 true for the transformation of abandoned areas in The use of fragments as materials for the reuse
which the role of the objects is prevalent than the project has, in some ways, a role similar to the one
faded figures of the background. of quotations used by the architects in their works.
The contemporaneity space appears structur- In both cases, it is to affirm the idea of architec-
ally transient, a construction site animated by con- tural continuity over time against the belief that
tinuous processing activities in which the forms of history is made of fractures and that the past can-
architecture lose stability. not be in any way related to the contemporary.
As written by Ernst Jünger, in the present land- The relationships between objects, thier
scape nothing is made to last forever and with that sequences and rhythms, are more valuable than the
character of permanence that we appreciate in objects themselves and them spaces they define. In
ancient buildings, nor in the spirit in which art tries these conditions, an effective design strategy can
to establish a formal language: everything, how- only be implemented through a set of discontinu-
ever, is provisional and made for short-term use. ous elements, creating a system made up of geo-
According to this situation, our territory looks metric references between new and existing space,
like a landscape in transition: in it there is no sta- of triangulation based on benchmarks and emerg-
bility of shapes, continuously created by a dynamic ing discrete points in the territory rather than rely-
restlessness. ing on the continuity of the parties.
The unpredictability of the initial conditions It comes to design interventions of inter-sca-
characterizing the nature of abandoned territories lar transformation that do not develop, as nor-
is a key factor in the development of the project, mally happens, according to a concatenation of
introducing the opportunities of subversion of tra- choices—from the general to the particular-, but
ditional approaches as surrealist artists did: looking according to systems of elements, even partial and
for an art that could overcome the logic and shared discontinuous, able to create meaningful places
conventions, they artfully created unforeseen situ- through perceivable syntactic orders.
ations to go beyond reality immediately perceived. In this case, the additive process—which con-
As in the contemporary projects of transforma- sists in arranging different parts in a horizontal
tion, the surrealists tried to create an artistic space sequence—seems the most suitable because the
in continuous change, including simultaneously individual elements already have their own config-
the past, the present and the future, contradictory uration and their order is already dictated by the
conditions and conflicts. linear track of rails.
But designing with wastes requires a complex
approach to reality. On the one hand, it is neces-
sary to look at the pre-existing conditions with 5 GENERALIZATION OF THE RESEARCH
no preconceived ideas that may lead to exclude
what instead could be useful to the project. Look A condition to safeguard in the re-use project of
without preconceptions—also paying attention to abandoned railways is the public use of the tracks.
seemingly worthless materials—means being able On the contrary, often it happened that, once
to look tendentiously at reality, being interested in closed the lines, the property was split and sold to
the transformation of the existing rather than in its settle debts of the railway companies.
preservation, with a “creative eye” and not with a In Italy, in 1991, it was created a company called
“contemplative” one. In this sense, the repertoire of Metropolis with the aim to “enhance” the aban-
references for the reuse project of railways is very doned manufacts of ‘Ferrovie dello Stato’ by selling
broad and it also includes the trivial forms with no the buildings and the relative areas. In other Euro-
author, the engineering works and even the objects pean countries, more careful to the preservation of
of Nature without any aesthetic feature, like those historical heritage, such as France, Spain and Bel-
“objects with poetic reaction” that Le Corbusier gium, the divestment was followed by the establish-
loved to collect and draw in his notebooks to find ment of companies with public capital that reused
their “secret form”. These are objects that can also the closed lines for tourism and cultural purposes.
be found “by chance” in the territory, confused The case of the Spanish Vias Verdes—Fundación
among many other worthless materials, abandoned de los Ferrocarriles Españoles (FFE), is exemplary
wrecks and figures worn by time. In the railway con- from this point of view: in Spain, today there are
text, the designer defines some criteria of selection more than 1,700 kilometers of rail infrastructure
of the objects left along the tracks: formal affinities, in disuse that have been converted into cycle routes
analogies mnemonic, emotional suggestions, etc. and green public walks.
These criteria make the design process less acciden- Equally important it is to be able to preserve the
tal, as opposed to what happens in “automatic pro- public use of the path so that the knowledge experi-
cedures” that, by relying on mere chance, submit ence of the territory by train can survive, although
the materials to various treatments, removing the through other forms of transport (SCHWARZER
artist’s control of the creative process. 2004). It would be desirable to use new locomotion

702

TIS2017_Book.indb 702 2/8/2017 8:08:58 PM

 Figure 6. High Line, New York, 2009.

sites, where the task of excavations is not to bring

to light separate pieces of a buried reality but—
through a careful reconstruction of the various lev-
els—to restore the meaning of layered sediments, in
the same way the primary goal of the reuse project
Figure 5. Barry Rosenthal, Grid, 2014. for abandoned railways should be to integrate the
complexity and contradictions of current condition
by giving the lines a new life (VIOLA 2013).
systems with low environmental impact on the old To define the physical dimensions of the inter-
tracks, as well as in many European countries, to ventions, their articulation, their hierarchies, their
replace the old technically and economically not components and the different times of their muta-
sustainable train. tions, is the primary task of the recovery project
The reuse project has to face also with unsched- of tracks and railway architectures, linking the
uled and spontaneous aspects, using them as past and a possible future of a particular territory.
a new resource: the natural vegetation wedged Therefore, a project designed for recovering a lin-
between the rails, creating long green lawns and ear complex system and its different elements, “re-
large patches of trees and wild shrubs, graffiti and building” the contemporary landscape through an
inscriptions that covered the walls, gaps and breaks intervention that is a new “work of art” itself.
created from the ruins of buildings.
The green is a constant presence in infrastruc-
ture places, closely linked to the real identity of the REFERENCES
railway. The relationship is obvious in the case of
tracks that develop within the natural territory, less Bonesio, L. 1996. Appartenenza e località. L’uomo e il ter-
obvious in urban lines. Although initially the lines ritorio. Atti degli incontri di geofilosofia, Milano: SEB.
were outside the cities, the constructions have grad- Branzi, A. 2006. Modernità debole e diffusa: il mondo del
ually surrounded the tracks because of the suburbs progetto all’inizio del 21° secolo. Milano: Skira.
growth, changing the character of places: inside and Budoni, A. 2014. Catturare il valore del suolo per svilup-
pare reti di trasporto locale su ferro. Ingegneria Ferrovi-
above the city, in the interstices left free from build-
aria (n.5): p. 431.
ings, deep green corridors raised along the train Jünger, E. 1991. L’operaio. Dominio e forma, Parma:
paths right into the city. These are elements of con- Guanda.
tinuity with the countryside that introduced forms Manterola, J. 2011. Ingegneria come opera d’Arte. Milano:
of biodiversity in the city (RITTER 1994). Once the Jaca Book.
railways were closed, this unforeseen nature grew on Maternini, G., Riccardi, S., Cadei, M. 2014. Trasformazi-
the edge of the tracks is one of the most significant one a tramvia di un sistema ferroviario. Il caso studio
legacy to be valued in the reuse project, as hap- tram-treno nell’area metropolitana di Brescia. Ingegne-
pened emblematically in the 90’s transformation of ria Ferroviaria (n.3): p. 225.
Mumford, L. 1961. The city in history: Its origins, its
the Görlitzer line in Berlin or in the recent recovery
transformations, and its prospects. New York: Brace &
project for the High Line in New York. World.
Berlin and New York cases teach that the reuse Ritter, J. 1994. Paesaggio. Uomo e natura nell’età mod-
of abandoned railways requires a flexible strategy erna. Milano: Angelo Guerini e Associati.
of interventions, able to adapt to the large-scale, Schwarzer, M. 2004. Zoomscape. Architecture in Motion
with typical and repeatable solutions, but also able and Media. New York: Princepton Architectural Press.
to identify specific solutions for each place. Viola, F. 2004. Ferrovie in città. Luoghi e architetture nel
Once lost the original unity of the infrastructure, progetto urbano. Roma: Officina.
the project has to ensure a permanent condition of Viola, F. 2013. Pietra su Pietra. La storia come materiale di
progetto. Salerno: Cues
transformation, favoring the spontaneous evolution
of the parts and the physiognomic change over time
according to their use. As well as in archaeological

703

TIS2017_Book.indb 703 2/8/2017 8:08:58 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Modelling ambulance and traffic behaviour using microsimulation:

The LIFE project application

F. Galatioto, V. Parisi, E. McCormick & C. Goves

Transport Systems Catapult, Milton Keynes, UK

ABSTRACT: This paper presents the preliminary results of the application of a microsimulation model
(VISSIM, PTV Group) to replicate the behaviour of ambulances in urban area and how different reac-
tions of general traffic can impact on the travel time of an ambulance. The work is part of an Innovate
UK collaborative funded project, namely LIfe First Emergency Traffic Control (LIFE) with the aim to
develop an innovative application for an intelligent transport system that operates in real-time to ena-
ble ambulances to reach life threatening emergency cases quicker by integrating ambulance route finder
applications with traffic management systems.
The microsimulation model setup within the project has been developed to understand and evaluate the
impacts and the best scenarios to improve ambulance response time and gains in cost-saving, whilst on
balance mitigating adverse impacts such as residual congestion.

1 INTRODUCTION a higher speed without causing any harm to any-

body even more difficult for the emergency vehicle
1.1 Context drivers itself.
Currently, the issues of ambulance services
Given the aging demographics and rapid urbanisa-
to meet the Government’s target of responding
tion, cities need to be equipped to respond to emer-
to 75% of life threatening calls within 8 minutes
gency (eg. 999 calls) more quickly.
(Guardian, 2015), is becoming more complex year
By 2050, over 25% of the UK’s population will
on year. Although critical emergency calls are time-
be over 65. This has implications on the overall
sensitive (i.e. heart attacks, strokes etc.), the BBC
health services as well as the NHS Trust to cope
in 2015 reported a 116% rise in ambulance delays
with anticipated rise in ambulance call outs amidst
over the past year alone, proving the situation is
worsening urban congestion. Presently, the Gov-
becoming increasingly critical.
ernment has set out targets for ambulance services
An international best practice review suggested
to reach 75% of emergency calls within 8 minutes.
that the response time should be approx. 5 min. to
For this reason, there is a growing need to develop
increase the survival rate by 12% (NHS, 2014).
new and innovative applications for an even more
Aside from operational issues, ambulance serv-
intelligent use of the existing transport system that
ices currently face (e.g. 10% of 999 calls are genu-
will support in real-time emergency vehicles to
inely life-threatening conditions, while categorising
reach life threatening emergency cases quicker.
40% of calls, resulting in dispatched vehicles before
Hence, this paper will discuss the methodol-
having determined the exact nature of the prob-
ogy and the preliminary results of the modelling
lem), LIFE will approach the poor ambulance
implementation of a “LIfe First Emergency Traf-
response performance from a transport and city
fic Control” or “LIFE” system seeking to identify
level. This is further exacerbated that ambulances
the best solution to reduce the time to respond to
are not legally allowed to travel 10 mph above
emergency calls, whilst operating a resilient service
the speed limit, as this could potentially cause
with a cost and fuel efficient fleet.
more accidents. On average, four ambulances are
involved in crashes each day resulting in a total
1.2 Background
cost for compensation and repair bills of £300 K
A recurrent problem that affect emergency vehicles every month (EveningStandard, 2008).
is reaching their destination on time. This is caused In Plano, Texas, the Emergency Vehicle Pre-
by increasing traffic and congestion in modern cit- emption (EVP) system has dramatically reduced
ies. Especially in urban areas, the mix of road users the number of emergency vehicle crashes from an
and the presence of traffic lights, makes driving at average of 2.3 intersection crashes per year to less

705

TIS2017_Book.indb 705 2/8/2017 8:09:03 PM

 than one intersection crash every five years (US do not experience slowdowns at the intersection and
DOT, 2006), whilst in Fairfax, Virginia, the EVP additionally crashes are prevented (Viriyasitavat &
system has been shown to save anywhere from a Tonguz, 2012). Not many are the experiments car-
few seconds to a few minutes. Ambulance drivers ried out so far using traffic microsimulation software.
cited savings of 30–45 seconds at a single intersec- An example though is represented by (Wang, et al.,
tion (US DOT, 2006). 2013) where a Vissim model has been used to validate
England has 10 Ambulance Service NHS Trusts a mathematical model that estimates travel times for
all of which have similar needs and same pressure vehicles under pre-emption control. The approach
to meet the Government’s 75% target (NHS, 2015). used here sees the deployment of an emergency lane
Of 1.1 m emergency calls last year, approx. 68% that is activated at random times, from where vehicles
require emergency transport (NWAS, 2015). move to the outside lane to let the EV through. Other
Across the UK, the Diagnostic and Ambulance studies focused on the emergency vehicles routing
Services Market (D&AS) has an annual growth of decisions, where assignment models that take into
2.7% (2010–2015) with a revenue of £6bn, driven account the demand and travel costs variability on
by UK’s ageing population. Total demand of the the network have been used to develop a travel time
£800 m a year UK ambulance service industry is function able to forecast emergency vehicles routing.
expected to grow significantly in the next 10 years (Musolino, et al., 2013).
demonstrated by the double digit growth in spend-
ing on private ambulances by the NHS over the
1.3 Research gaps
past two years (Plimsoll, 2013).
Crashes are a significant problem for ambu- There is a clear need for the traffic systems of
lances and all emergency vehicles, in the US it was the city to support faster response times through
found that 25% of all accidents involving emer- prioritisation. In the US Emergency Vehicle
gency vehicles occurred at signalised intersections Pre-Emption (EVP) at signalled crossings has been
especially where line of sight was blocked by build- taken up widely across the country with improved
ings or vegetation (Viriyasitavat & Tonguz, 2012). ambulance journey times.
In the UK cost for compensation and repair bills This is not the first time a concept like this takes
of ambulance crashes totals £300 K every month place: other systems, such as the European Com-
(EveningStandard, 2008). pass 4D, looked at prioritising vehicles at traffic
Also the bus priority applications that can be lights through V2I/I2V communication, as well
seen in many cities worldwide use pre-emption sys- as warning drivers of possible road hazards (V2V
tems to save time especially when they are beyond communication). However, the system that has
schedule. (CORDIS, 2016) seen only a trial version of it, is not deployed yet
An application that is a precursor of the LiFE and even though it has been proved to be efficient
concept is the EViEWS system, deployed in Texas in terms of reducing emissions at intersections and
since 2014. Through GPS signals transmissions journey times. For these reasons a test with emer-
and a wireless communication, speed and position gency vehicles should be run and is the focus of
data of emergency vehicles are sent to traffic lights, this paper.
that are then turned green to provide prioritisation. Moreover, current and traditional traffic micro-
(Lloyd, 2014) simulation modelling tools, have been used using
The latter however is not the first emergency default functions which are definitely not designed
vehicles pre-emption technology used so far. Glo- to model an ambulance and the behaviour of traf-
bal Traffic Technologies (GTT) introduced the fic when an ambulance is going through it.
Opticom system is 1968, which again makes use Unlike EVP, the LIFE project idea goes beyond
of GPS signals and wireless communication, in traffic lights responding to an ambulance’s
combination with the siren activation, to trigger approach but integrates route finder applications
the green phase according to the ETA to the signal with traffic management systems. The proposed
(ITS Int, 2015). application will be developed so it can be inte-
The pre-emption technique has proved through grated with existing systems in ambulance vehicles
the different on site applications to actually reduce and will therefore compliment current solutions.
journey times, however also traffic modelling stud- Also, it is going to be designed using ethnographic
ies demonstrate that it easies the movement of emer- research into ambulance driver behaviour to ensure
gency vehicles in urban areas. (Sharma, et al., 2013) LIFE is fit-for-purpose for the end-user with feed-
Algorithm based approaches have been carried out to back loop mechanisms throughout the project.
determine whether enabling priority at intersections As the project has begun 10 months ago,
would actually reduce travel times for emergency this paper will focus mainly on the initial results
vehicles and it has been found out that since the green of the implementation of the microsimulation
wave allows the emergency vehicles to go faster, they modelling methodology and in the steps to model

706

TIS2017_Book.indb 706 2/8/2017 8:09:03 PM

 the improvements of the response time of ambu- The reason why it has been chosen among the
lance services through a predefined corridor. other commercial software available is the more
realistic behaviour of the vehicles, with the possi-
bility for them to overtake both in the same lane
1.4 Main aims
and in the opposite lane. Vehicles that have enough
In order to overcome the above identified issues, lateral space (defined by the user) and a desired
the LiFE project aims at reducing the travel time speed higher than those in front, may overtake in
of emergency vehicles in a safely manner, namely the same lane, as well as in the opposite lane if not
prioritising them at traffic lights or rerouting them upcoming traffic is detected.
where the combination of traffic and road condi- Since the main purpose of the application of
tions would allow a shorter drive. a microsimulation traffic models to analyse trips
Key and unique aspect of the work undertaken and travel time of emergency vehicles (EVs) was
in this paper is the process that has allowed the also to reproduce and monitor the reactions of
microsimulation tool to be adjusted in order to standard drivers in urban areas that traditionally
enable an existing microsimulation model to repli- in presence of ambulance react in different ways
cate traffic conditions when an ambulance on blue depending on the road layout and speed, therefore
light (highest emergency code) is going throughout the feature required was from one hand the ability
a representative urban corridor. for the EV to overtake standard traffic, but also for
the standard traffic to implement manoeuvres that
are not by default implemented in a microsimula-
1.5 Objectives
tion software.
The real-world datasets used to calibrate and vali- In fact, by default the overtaking feature in Vis-
date the microsimulation tool have been made sim allows all the classes of vehicles defined by the
available by the North West Ambulance Service user to be overtaken by vehicles with higher desired
(NWAS) based in Liverpool, UK. speed, but vehicles of same defined classes may
Within this paper several objectives will be overtake as well depending on the speed they desire
achieved: to travel at, as mentioned earlier. In other words, it
was not possible for EVs only to overtake all the
− For the first time a traffic microsimulation
other vehicles on the road network by default.
model will be used to reproduce realistic traffic
Hence, in order to also achieve a realistic behav-
behaviour in presence of ambulance;
iour of the traffic in presence of an EV, an external
− the modelling framework and API module set-
driver model able to control vehicles’ behaviour on
tings used within the microsimulation tool will
the Vissim network has been identified as the only
be presented;
solution to model realistically ambulances and
− a real study area will be reproduced using real-
traffic affected by EVs.
world ambulance data, and
− preliminary results of selected traffic manage-
ment scenarios will be presented alongside the 2.2 The external driver model
benefit of implementing the new modelling
By designing and implementing an external driver
software.
model, the intention is to override most of the
The reduction of the frequency and severity of vehicles’ behaviour. In this way, Vissim vehicles do
collisions with ambulances significantly decreases not follow anymore the behaviour embedded in
the cost to the fleet as well as public liability associ- the software, but instead the instructions the exter-
ated with fatalities, injury and damage to property. nal model send to them in certain circumstances
3) Managing vehicle intelligently could have huge (approaching traffic lights, EVs behind or in front,
environmental impacts for cities with benefits such etc.).
as reduction of CO2 emissions, improved air quality Emergency services are mostly operating in criti-
as well as cost savings resulting from less fuel con- cal conditions; thus an efficient journey planning is
sumption and wear-and-tear on emergency vehicles. vital in improving time-response to a distress call.
Modelling and analysing travel time in different
traffic conditions and road types could prove very
helpful in decision making and better allocation of
2 THE PROPOSED MODEL
existing resources.
Vissim is an established software package in
2.1 The microsimulation model
microscopic traffic simulation, capable of provid-
The traffic microsimulation software used to ana- ing a realistic model of road users while allowing
lyse the LiFE system is the German Vissim 8.12 the flexibility to inject external models for vehicle
commercially available through PTV Group. behaviour. VISSIM has an External Driver Model

707

TIS2017_Book.indb 707 2/8/2017 8:09:03 PM

 for as DLL interface which provides the option dispersion. In congested areas, the awareness falls
to replace the default driving model with a user to 10 meters.
defined one as long as the choice is activated for When an ambulance is approaching, every vehi-
that particular vehicle type. cle checks if the distance between the two is within
The possibility of incorporating the external the awareness threshold, if it is not, the DLL col-
driver model into the software has made it feasi- lects the values suggested by VISSIM and passes
ble to reproduce realistic conditions in the traffic them back, without any change. If the distance is
simulator, where standard vehicles are made aware in the awareness range and the vehicle has a veloc-
of EVs approaching and react accordingly. ity greater than 5 m/s, it changes lanes in order to
allow the emergency vehicle to pass. If the velocity
is small, the vehicle is caught in congestion and the
2.3 General Implementation Guidelines
ambulance is behind, the vehicle is moving out-
The DLL is implemented in C++ and has to follow side the road while decelerating with an exponen-
VISSIM structural guidelines. tial function of lane lateral position, allowing the
The DriverModel.h is a pre-defined header file ambulance to overtake.
containing all the definitions necessary during the Without coordinated commands, vehicles might
simulation, the DriverModel.cpp represents the decide to take inconsistent decisions such as, some
main source file which must implement three func- move left while some move to the opposite direc-
tions: DriverModelSetValue, DriverModelGet- tion, thus creating a situation where unrealistic sce-
Value and DriverModelExecuteCommand. narios could arise. In order to avoid this, the vehicle
During the simulation run, VISSIM calls the which leads the decision making is the one directly
DLL every time step for each vehicle type actively ahead of the ambulance. Any other driver affected
controlled by the external model in this order: SET, by the ambulance firstly checks the specified car
GET and EXECUTE COMMAND. action and then applies the same behaviour if the
The Set method passes the current vehicle conditions are favourable.
values referring to time, vehicle id, lane and link The first limitation experienced was the fact
information, velocity, acceleration, width, height, that at any time a vehicle could interrogate the ID
maximum acceleration, turning indicator, x and of the neighbours but not their types, thus a car
y coordinates, information regarding lane change would be unware of who is the ambulance in the
and lane target, road information, coupled with proximity.
data about adjacent vehicles. At any time, the The problem was solved by defining a ‘Vehi-
known neighbouring vehicles in the current link are cle’ structure holding all the necessary parameters
two ahead, two behind, two left and two right. If updated during ‘Set’ calls and a matrix with the
the signal DRIVER_DATA_VEH_CURRENT_ neighbours’ unique identification numbers as seen
LINK is active, the model also gets information in the Figure 1.
about adjacent links. Every Set, ExecuteCommand and Get calls are
The ExecuteCommand method handles the data dealing with data embedded within the current
received about a particular vehicle in Set and makes vehicle object.
the necessary adjustments to change the default In parallel, there is a hash map which holds an
behaviour. If no correction is needed, the data is array of ‘Vehicle’ objects and uses as a key the
passed back to VISSIM without alterations. vehicle unique identification number. The map
In the Get method, VISSIM retrieves the value contains all the vehicles active in the simulation.
associated with the vehicle id specified in Set and If a vehicle knows its neighbour’s id it can check
applies the new behaviour. The currently avail- on the map the type of that particular id alongside
able commands are driver init, create driver, move other vehicle specific parameters. Furthermore,
driver and kill driver. in order to improve efficiency there could be two
A minority of parameters are updated per vehi- structures, one holding regular vehicles and one
cle type or once per DLL initialization and are not holding the emergency vehicles.
called every tick.

2.4 Specifications
In order to implement the behaviour of ambu-
lances and the affected vehicles, a certain degree of
traffic disturbance was implemented.
Every vehicle has a different degree of ambu-
lance awareness distance which is a Gaussian
distribution of 70 meters mean and 20 metre Figure 1. Matrix for the neighbours vehicle’ ids.

708

TIS2017_Book.indb 708 2/8/2017 8:09:03 PM

 The latter category contains less elements and a
query takes considerably less resources.
The ambulance has the ability to identify gaps
in the upcoming traffic, to evaluate if the vehicle
ahead is offset lane by checking the lane lateral
position and accelerate past and to ignore the red
traffic warnings.
The scenario when the non-emergency behav-
iour vehicle has no ambulance behind in the cur-
rent lane is described in the Figure 3.
The scenario when the ambulance is two cars
behind is described in Figure 4.
The scenario when the ambulance is more than
two cars behind in the current lane is described in
Figure 4. Ambulance two cars behind.
the Figure 5.
The scenario when the ambulance is one car
behind is described in Figure 6.

Figure 5. The ambulance is more than two cars behind

in the current lane.
Figure 2. Vehicle Hash map.

Figure 3. No ambulance behind in the non-emergency’s vehicle current lane.

709

TIS2017_Book.indb 709 2/8/2017 8:09:03 PM

 Figure 6. The ambulance is one car behind in the cur-
rent lane. Figure 7. Traffic network layout and ambulance trip
(red).

3 RESULTS

3.1 The study area

For the application and calibration of the newly
developed external driver model an urban network
extended on an area of 3.5 km2 in the vicinity of
the Transport Systems Catapult premises have been
used. The network (Figure 7) was previously cali-
brated and validated for a different project, hence
the testing could rely on an existing proven tool.
Since red calls should arrive at destination within Figure 8. Screenshots of three different modelling
8 minutes, the ambulance journey (in red) was cho- scenarios of ambulance (long yellow vehicle), using the
sen of a length of 5 km which in normal traffic or novel external driver model.
overnight takes up to 6 minutes, but during peak
hours can easily take over 10 minutes.
The time modelled for this exercise was between
7:30 am and 10:00 am, with 30 minutes warm up
period. This is because the network is particularly con-
gested at that time. However, to stress test the external
model, higher level of flows were also considered.

3.2 The modelling results

As a result of the novel external driver model devel-
oped, several scenarios of ambulance going through
traffic both on links and junctions have been repli- Figure 9. Travel time of Ambulances (EVs) through the
cated and realistic behaviour have been possible to simulation period with and without implementation of
the external model (DLL).
replicate in the microsimulation model (Figure 8).
For each scenario result, 10 different unique
simulations (using different seeds) were carried through red light at signalised junctions). Simi-
out. larly, for highly congested period EVs cannot take
In Figure 9 are reported the results of running advantage of either the external model, since sur-
the simulation with 3 to 5 ambulances inserted in the rounding vehicles are not able to move easily, and
network every 10 minutes, this was done to limit mul- the ability to go through red light.
tiple ambulances to proceed on the same corridor at This, as highlighted in Figure 10, which refers
the same time, even though that is a real possibility. to results from the same network, but implement-
It can be observed that the external model pro- ing reduced flow levels, provides a useful perspec-
duces a real benefit of around 4 minutes for traf- tive on when both temporally and geographically
fic level which can be classified as busy, while for a LIFE system application will make a real differ-
free flow, as expected, the time saving is minimal ence in enabling the ambulance to reduce the travel
(mainly associated with the ability of EVs to go time to meet the Travel Time (TT) target.

710

TIS2017_Book.indb 710 2/8/2017 8:09:06 PM

 Figure 11. Preliminary results of Emergency Vehicle
Figure 10. Traffic network layout and ambulance trip speed/flow curves (orange) compared with standard traf-
(red). fic (red).

replicate ambulance trip accurately and also to

4 DISCUSSION AND CONCLUSIONS model the behaviour of the traffic when an emer-
gency vehicle is going through it, the next step the
This paper has presented the methodology to LIFE modelling group is currently developing is
develop an external driver model to enable a com- the extension of the modelling functionalities to
mercial microsimulation model to accurately model model new scenarios where intelligent priority sce-
ambulance (or generically Emergency Vehicle) narios can be designed and tested on any traffic
trips across a transport network and the prelimi- network.
nary results of the application of proposed driver Future steps are, a) to develop a traffic control
model. strategy that by detecting the ambulance and its
The external driver model has been designed route in advance will automatically develop a new
and developed to work independently from any traffic light planning, dependent on traffic condi-
microsimulation network configuration and geom- tion and congestion on the network; b) an algo-
etry, this is because it overrides the vehicle behav- rithm to calculate in real-time the ambulance TT.
iour according to the dynamics on the surrounding For the latter, specific speed flow curves for ambu-
environment (eg. speed limits, speed of the sur- lance will be derived using the new modelling tool
rounding vehicles, position of the EV in respect to developed and presented in this paper. This is cur-
each individual vehicle). rently under development and full results may be
The results obtained show that the applica- presented at the conference.
tion of the new driver model developed provides In anticipation, Figure 11 presents the result
the expected benefit that in real-world condition of an initial test performed using the external
an ambulance going through an urban area and driver model algorithm in a non-congested urban
for different traffic levels is expected to achieve, transport network. The different points have been
however the model results also highlight that with obtained by changing the percentage of total flow
congested traffic levels, ambulance travel time can- (standard vehicles) hence, the travel time of the
not be currently reduced beyond a certain limit, ambulances injected in the network at 10 minutes
which unfortunately is not enough to achieve the intervals have been averaged and reported in the
set target of 8 minutes. This reinforce the need to graph.
implement a new system which is currently under It is still an early stage result and needs calibra-
development as part of the LIFE project. tion and tuning, as well as a validation with the
A simple link based pre-emption strategy has AVL data available, however the preliminary results
been tested using the new external driver model show encouraging pattern, which is definitely dif-
and benefit up to a certain congestion level have ferent from the general traffic as expected.
been observed, however a more structured and This will help in developing the pre-emption
coordinated strategy seems to be required and this strategies as well as to accurately predict the
will be part of the future steps of the project. Ambulance travel time in the future.

5 FUTURE WORK REFERENCES

Since this paper presents the preliminary results BBC, 2015. http://www.bbc.co.uk/news/health-30742817
of the application of an external driver model to CORDIS, 2016. http://cordis.europa.eu/result/rcn/4102
enable a commercial microsimulation model to 9_en.html.

711

TIS2017_Book.indb 711 2/8/2017 8:09:07 PM

 EveningStandard, 2008. http://www.standard.co.uk/news/ NHS, 2014. Ambulance Services, England 2013–14.
ambu-lances-in-crashes-four-times-every-day− Workforce and Facilities Team, Health and Social
6653861.html. Care Information Centre. Version 1.0. 31 July 2014
Guardian, 2015 https://www.theguardian.com/society/ NWAS, 2015. North West Ambulance Service NHS Trust
2015/jan/16/englands-ambulance-services-trial-new- Annual Report 2014/2015.
system-cut-response-times. Plimsoll, 2013, Ambulance Services Market Report
ITS Int., 2015. Priority management saves time, money US DOT, 2006. Traffic Signal Pre-emption for Emer-
and lives. ITS International. September October 2015. gency Vehicles. A CROSS-CUTTING STUDY.U.S.
http://www.gtt.com/1521–2/. Department of Transportation. Federal Highway
Lloyd, R., 2014. Next-Generation Preemption Sys- Administration. January 2006.
tem Clears the Way for Harris County, Texas, First Viriyasitavat, W., Tonguz, O.K., 2012. Priority Manage-
Responders http://www.fireapparatusmagazine.com/ ment of Emergency Vehicles at Intersections Using
articles/print/volume-19/issue-3/features/next-gener- Self-organized Traffic Control. Carnegie Mellon Uni-
ation-preemption-system-clears-the-way-for-harris- versity. 9–2012.
county-texas-first-responders.html. Wang, J., L. Yu, and F. Qiao. 2013. Micro Traffic Simu-
Musolino, G., Polimeni, A., Rindone, C., Vitetta, A. lation Approach to the Evaluation of Vehicle Emis-
2012. Travel time forecasting and dynamic routes sions on One-way vs. Two-way Streets: A Case Study
design for emergency vehicles. Procedia—Social and in Houston, Downtown. Proc., 92nd Annual Meeting
Behavioural Sciences 87 (2013) 193–202. SIDT Scien- of the Transportation Research Board, Washington,
tific Seminar 2012. DC.

712

TIS2017_Book.indb 712 2/8/2017 8:09:08 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Development and application of national transport model of Croatia

Gregor Pretnar
PNZ Svetovanje Projektiranje d.o.o., Ljubljana, Slovenia

Uwe Reiter
PTV Transport Consult GmbH, Karlsruhe, Germany

Igor Majstorović
Građevinski Fakultet Sveučilišta u Zagrebu, Zagreb, Croatia

Ana Olmeda Clemares

Ineco, Madrid, Spain

ABSTRACT: The Croatian Ministry of Maritime Affairs, Transport and Infrastructure commissioned
the development of the National Transport Model, including collection of all available data, carrying out
necessary surveys, developing networks models and demand models for freight and passenger demand for
the base year, calibrating and validating the models, and developing forecast models. Model has 985 inter-
nal and 267 external transport zones and over 360.000 km of modelled links. The purpose is to identify
shortcomings, bottlenecks and issues in the current and the planned future transport systems. The model
is used to identify specific measures and projects for the different transport modes and their integra-
tion supporting the selected strategies. The model produces quantitative results allowing to determine
impacts of the strategy alternatives and of the measures on traffic conditions, on social and environmental
impacts. The availability of a National Model guarantees that similar approaches are used at the regional
level, improving the general transport planning approaches all over the country.

1 ABOUT THE PROJECT commissioning the model development was to

obtain a quantitative tool that could support the
The National traffic model for the Republic of development of the National Transport Strategy,
Croatia was co-financed by the EU from the Euro- help to analyse current conditions and forecast
pean Regional Development Fund under Trans- future conditions, provide the basis to identify nec-
port Operational Programme 2007–2013 within the essary strategies and measures and finally being
project “Support for the preparation of the Repub- able to calculate the impacts of strategies and
lic of Croatia´s Transport Development Strategy measures on the future transport system and the
and designing of the national Traffic Model for influencing processes, like social, economic and
the Republic of Croatia—National Traffic Model environmental processes. A National Transport
for the Republic of Croatia”. Beneficiary of the Model is the necessary tool to plan the sustainable
project and contact information: development of the transport system.
Ministry of Maritime Affairs, Transport and Infra-
structure; Directorate of EU Funds 3 MODEL DEVELOPMENT
Krležin Gvozd 1a, 10000 Zagreb, Croatia www.
mppi.hr; www.promet-eufondovi.hr
The model was developed using the software
suite PTV VISION. It follows the classical 4-step
The contents of this publication are the sole
approach. National Model of Croatia is a syn-
responsibility of Consortium led by PTV Trans-
thetic model using network data, socio-economic
port Consult GmbH and do not necessarily reflect
data and behavioural data as its foundation. Only a
the opinion of the European Union.
synthetic model, calibrated and validated to actual
empirical data, is capable of scientifically and
2 INTRODUCTION correctly forecasting future developments and of
calculating impacts of changes in influencing con-
The intention of the Croatian Ministry of Mari- ditions (exogenous factors like economic and social
time Affairs, Transport and Infrastructure in development) and of changes within the transport
713

TIS2017_Book.indb 713 2/8/2017 8:09:10 PM

 system itself, e.g. implementation of strategies and – Road network with all relevant levels (motor-
measures (endogenous). ways; state, county, local and urban roads) with
For passenger transport, a trip generation model design standard and condition characteristics;
was developed for the resident population and for – Railway lines of importance to international,
the tourists, information of the actual destinations regional and local transport with design stand-
of trips of residents and visitors was used for trip ard, traction, gauge, interoperability, condi-
distribution. Similarly, for freight transport, data tions—restrictions and reliability characteristics,
on import/export, production, processing and con- strategic national and international connections,
sumption of numerous commodity types was used capacity utilization, rolling stock, operators and
to develop the freight generation and distribution. safety;
For both models, mode choice and assignment – Public transport routes for rail, boat (ferry), tram,
were based on costs including actual travel times. city and intercity bus lines with timetables;
The socio-economic data was collected from dif- – Seaports and inland ports with their characteris-
ferent sources at the level of model zones mainly tics, purpose and capacities;
from National Statistics. The basis for the behav- – Airports for which the land-side traffic was
ioural data was a household survey with more than modelled (by car, public transport and freight
3,000 interviews and a survey of freight operators, traffic);
both carried out within the project. Empirical data – Intermodal facilities for passenger and freight
was complemented by traffic counts and public transport.
transport passenger counts. The empirical data
Republic Croatia was divided into 985 traffic
from external sources and from the own surveys
zones. Basic level for zoning were cities and munic-
was used to calibrate and to validate the model. To
ipalities. Traffic zones with population greater than
represent differences between the summer season
8.000 persons were broken down to lower level of
with high numbers of tourists visiting the country
territorial unit, i.e. city and municipality to settle-
and the rest of the year, two different models were
ment, settlement to statistical circle.
produced, an off-season model and a seasonal
model.
Model development consisted in the develop-
ment of a base year model and of forecast models
for the horizon years 2020, 2030 and 2040. Fore-
cast was based on the available and accepted offi-
cial data of future socio-economic development
of Croatia and the surrounding countries. For all
forecast horizon years, a so-called do-minimum
scenario was developed that is used as a refer-
ence scenario, including only those projects and
measures that are already under development or
that are planned and financed. The do-something
or strategy scenarios include additional projects
and measures being part of the national transport
strategy. Figure 1. Road and rail network.

3.1 Network and zoning

The network system was composed of road and rail
systems, airports, maritime and inland ports. The
national and international connections were mainly
established by primary road network (motorways,
trunk roads and state roads), railway network and
ports. In addition, roads in major urban areas have
relevant functions for the national network and a
significant impact on connectivity and accessibil-
ity. These network elements also serve as public
transport routes and as alternative routes. There-
fore, urban main roads were also included in road
network.
As result the following network elements were
included in the model: Figure 2. Zoning.

714

TIS2017_Book.indb 714 2/8/2017 8:09:10 PM

 3.2 Passenger demand model work-other and other-work, the concerned per-
son is an employee, or in school trips, the person
Passenger demand model consists of first three
concerned is a pupil, and for secondary school
steps in traditional four-step model:
or university a student. Based on surveys across
– generation (production and attraction), households in the Republic of Croatia certain
– distribution and mobility rates were determined for all origin-des-
– modal split. tination groups.
Development of the seasonal traffic was based
Production and attraction primarily depend
on the same basis as the average workday model.
on the gross domestic product, motorization rate
Thirteen origin-destination groups were taken into
and spatial socio-economic structure as well as of
account to calculate the generation of passenger
behaviour patterns, whilst attraction also partly of
traffic, but the destination groups home-school
the traffic supply. Distribution and mode choice
and school-home were replaced by the destination
are calculated simultaneously and consider both
groups home-vacation and vacation-home. Modi-
attraction (e.g. number of shopping area) and
fied destination groups include trips from home
transport supply (accessibility). Distribution
to holiday and private facilities intended for vaca-
approximately equally depends on the spatial socio-
tion use (home-vacation) and trips in the opposite
economic structure and behaviour patterns on one
direction (vacation-home); there are no school
hand and on the traffic supply on the other. Modal
trips during the holidays. Therefore, the purpose
split particularly depends on the traffic supply and
of (daily) leisure was extended for a special holiday
to a considerable extent also on the land-use.
subcategory of leisure (vacation).
Last step in the development of passenger
demand model is assignment of demand on the
3.2.2 Trip distribution and mode choice
multimodal network. Static stochastic learning
Distribution and modal split were calculated simul-
procedure (Lohse) was used. Public transport
taneously. That is to say, at the same time the des-
was assigned by the intermodal method based on
tination and the transport mode by which the trip
timetables. Method takes into account journey,
is done were chosen. The calculation was carried
transfer and waiting times, based on actual time-
out on the basis of the EVA probability function
tables. Both, internal and external transport, were
(developed by prof. Lohse from TU Dresden and
assigned simultaneously.
included in PTV Vision software), for the average
This approach enables that the forecast calcula-
workday traffic and for the traffic during the tour-
tion considers changes in spatial structure, gross
ist season.
domestic product, motorization rate, residents, jobs,
Input data for the distribution and modal split
etc., as well as transport supply while calibrated
sub-model were:
parameters of the model remain unchanged.
– productions and attractions,
– generalized prices or generalized times for the
3.2.1 Trip generation
road motorized and public transport,
Production and attraction were calculated by the
– EVA model parameters.
method of origin-destination groups (13 origin-
destination groups were considered for passenger Basic parameters of the model were set based
transport). These 13 origin-destination groups on the stated researches, recommendations of the
actually represent 5 trip purposes (work, school, software manufacturer and previous experiences
shopping, leisure and vacation, other) in conjunc- in modelling of the national and regional models.
tion with the home-bound trips and combinations Based on these data and the EVA functions, within
between them. The trip purpose of business was the multiple iterations, the trip matrices for pas-
specifically modelled as a separate demand model. senger car transport, public transport, cycling and
From the household survey (performed within walking were calculated.
the project itself) it was found that different parts
of Croatia have different travel behaviour patterns. 3.2.3 Assignment
Therefore, the model of passenger transport pro- Road assignment was carried out in several itera-
duction and attraction on the average workday was tions. Based on the information obtained in the
specially developed for Continental and Adriatic previous iteration, users find a new optimal route
regions of the country. in the next iteration. Therefore, in the iterative
For each of the thirteen origin-destination process search for optimal route runs until the net-
groups a mobility rate (number of trips per day work equilibrium and the appropriate impedance
per person concerned) was determined. In most matrix convergence were reached.
of these groups, the person concerned is a resi- Network assignment was based on the function
dent. However, in groups home-work, work-home, of generalized price or generalized time. Toll was

715

TIS2017_Book.indb 715 2/8/2017 8:09:11 PM

 incorporated through the function of generalized transhipment infrastructure for intermodal han-
time, where the monetary values were converted dling of goods.
into the equivalent of time. The modal freight trip matrices were calculated
In seeking the optimal routes, also the effects with a commodity based multi-modal model using
of traffic congestions and jams were taken into an enhanced 4-step approach. As a big advantage
account, i.e., the effects of driving speed reduc- of this synthetic multi-modal approach, the pro-
tions. BPR function was used for this, the most posed methodology guarantees:
common volume-delay function, which reflects
– the adequate consideration of commodity-
the travel time, depending on the volume and road
specific affinities regarding different transport
capacity. It was useful both for modelling of non-
modes,
urban and urban roads.
– the ability to reflect multi-modal transport and
Intermodal method for public transport assign-
inter-modal transport chains,
ment allowed the entire public transport network
– a realistic calculation of future freight demand
to operate as a unified system that includes rail, bus
based on socio-economic changes and/or net-
and maritime lines of various levels. The method,
work modifications (e.g. new links or tranship-
based on timetables, requires precise arrival and
ment hubs),
departure times of vehicles or trains at stations and
– the consideration of all possible transportation
stops to be set for all public transport lines. The
modes for route choice.
network was therefore modelled in that way.
For each origin-destination pair of zones a The different commodity types, which are con-
favourable connection was found or calculated. It sidered, range from agricultural goods (e.g. cere-
was assumed that passengers are aware of the time- als, fruits, vegetables), raw materials (e.g. coal, raw
table, and would take the first available line of pub- wood, ores), oil products, industrial products (e.g.
lic transport offering a favourable route. Among steel and metal products, chemical products) to
various combinations of routes, more favourable construction materials and consumer goods.
routes were chosen. The most favourable routes
were determined on the basis of the whole chain of 3.3.1 Freight generation
route segments, including ticket price, which was In practice, the reasons for the transport of goods
included in the function of generalized time. are the different locations of production and con-
sumption of a certain good and the resulting need
3.2.4 Input data of exchange.
The main input data of the passenger model is Hence, as a first step of the demand calculation,
listed below: the generated volumes per traffic zone are deter-
mined for each commodity. This is done both for
– Transport network data for all modes (road, the production side (also referred to as origin side)
public transport), and the consumption side (also referred to as des-
– Socio-economic data, i.e. population and tination side).
employment data, disaggregated to traffic zone In general, the determination of origin and des-
level, tination vectors is conducted in 2 steps:
– Behavioural data from household survey
(n = 3.000), 1. Determine production and consumption vol-
– Transport cost parameters (distance related, umes on national level for Croatia and the coun-
time related, toll costs) per transport mode. tries of the region
2. Break down of these national volumes to traffic
zone level
3.3 Freight demand model
On national level, there is the constraint that the
Freight transport as a whole is a very complex and total generated volume of all origin zones must equal
heterogeneous process. The multi-modal freight the total generated volume of all destination zones.
model follows a highly disaggregated approach to While the origin volumes consist of local production
calculate the freight volumes based on origins and volumes and import volumes, destination volumes
destinations of homogenous commodity types. are the sum of local consumption and export.
This includes both, domestic freight flows and The production and consumption volumes on
external freight flows (import/export/transit). The national level are broken down to traffic zone level
freight model considers all transport modes rel- by the distribution of the decisive land uses for each
evant from a national perspective, i.e. road (HGV commodity. Depending on the type of commodity,
and LGV), rail and vessel. The same network as decisive land uses can be for example population,
for the passenger model is used, enhanced with employees by economic sector as well output/pro-
additional mode-specific freight parameters and duction capacities of production facilities.

716

TIS2017_Book.indb 716 2/8/2017 8:09:11 PM

 Thus, calculating the local production and con- – Distance costs (mode specific distance related
sumption per zone and adding the import/export transport costs),
volumes, for each commodity two vectors are gen- – Handling costs (costs for loading/unloading and
erated. One includes the generated origin volumes transhipment).
per zone, the other vector the attracted destination
volumes. The modal transport costs were allocated to all
network links. Different sets of transport costs
3.3.2 Freight distribution consider:
Like traffic generation, distribution calculation is – The link attributes (permitted transport mode;
applied successively and separately for each com- transfer link) and
modity. Using a gravity model, the generated origin – The logistic system, the commodity is allocated
and destination volumes per zone are distributed, to.
resulting in yearly ton flows between the traffic
zones. The trip distribution calculation is carried As noted before, the ton flow matrices of each
out in 2 steps: commodity were assigned to the multi-modal net-
work with an iterative equilibrium assigning proce-
1. Calculation of evaluation matrix based on a dure, resulting in the most cost-efficient route and
skim matrix including the impedances between mode choice for each OD relation. From the multi-
traffic zones modal ton flow assignment, ton flow matrices by
2. Calculation of trip matrix (yearly ton flows) mode can be derived.
based on the evaluation matrix and the origin The modal ton flow matrices were converted
and destination volumes. to vehicle trip matrices by average loading factors
A monetary impedance matrix calculated from (depending on distance classes) and a factor for the
the VISUM network model is used as skim matrix. unloaded drives. Finally, the generated HGV and
The matrix values in [€] are calculated with the fol- LGV trip matrices were assigned to the network
lowing impedance function. together with the trip matrices from the passenger
demand model.
wij ∑C hhandlin
dli g
+ Length ckkmm + ∑ (Time ⋅ ch ) (1)
3.3.4 Input data
The main input data of the freight demand model
where wij = impedance between traffic zone i and are listed below:
traffic zone j [€]; Chandling = handling costs [€];
ckm = distance cost rate [€/km]; and ch = time cost – Socio-economic data, i.e. population and
rate [€/h]. employment data, disaggregated to traffic zone
For all commodities, different impedances level,
(costs) result from different cost rates for required – Data on national production (production vol-
logistic system (e.g. bulk, container) and different umes, location of major production facilities)
requirements for transport speeds due to varying for each commodity,
urgencies (loss of value). – Import/export data from trade statistics, aggre-
gated to about 50 freight model commodities,
3.3.3 Mode choice – Transit data from UN COMTRADE statistics in,
The route and mode choice were conducted simul- aggregated to the freight model commodities,
taneously in an interim assignment step for the ton – Transport cost parameters (distance related; time
flows. The ton flow matrices of each commodity related; transhipment costs) per transport mode
were assigned to the multi-modal network with an for the route and mode choice calculation,
equilibrium assignment procedure.
The mode choice decision for a certain commod-
ity and origin-destination relation was based on the
transport costs. It is always the most cost efficient
route and transport mode that is chosen. This can
be either a direct transport with one transport sys-
tem or a multimodal chain with a combination of
transport modes and transhipments in between.
The total transport costs, which are the deter-
mining factors for the route and mode choice dur-
ing the assignment, consist of:
– Time costs (mode specific time costs + commod-
ity specific loss in value), Figure 3. Mode and route choice for freight transport.

717

TIS2017_Book.indb 717 2/8/2017 8:09:11 PM

 – Operational parameters (e.g. average loading First step was checking the input data for demand
factors and share of empty trips by commodity that are socioeconomic and behavioural data.
and vehicle type), Socioeconomic data (number of population, work-
– Growth rates and macro-economic parameters places, school places…) were taken from the official
are required for forecasting internal and external databases where they have already been submitted
freight flows. to various checks and should be reliable. There were
some issues that required additional analyses (e.g.
number of workplaces assigned to company’s seat,
4 MODEL CALIBRATION no data available on shopping areas…).
AND VALIDATION Results of household survey were compared with
existing data from previous studies and practice. It
In modern society the transport model represents was established that most important indicators lie
one of the key bases for decision-making on trans- within expected benchmark values (e.g. between
port and spatial policy, on the investments in the 2.5–3 daily trips per person, cca. 40% of work and
infrastructure demanding time and funds, on the school trips, characteristic morning and afternoon
form and dimensions of roads and railways, their peak hours, average trip length and duration…).
impacts, etc. It is therefore important that the model First validation was done for trip duration dis-
results are reliable. Reliability and credibility are the tribution. Although such distribution is also input
key characteristics of good and useful transport data, it serves also for checking results of the model.
models. The necessary precision of the model is Modelled distribution is not only direct result from
achieved by the calibration, whilst the reliability and the survey, but also considers impedance (travel
credibility required are proved by the validation. time, length…) between all pairs of zones.
The growing role of the transport policy has ini-
tiated that transport models are becoming increas- 4.2 Transport flow validation
ingly complex. Validation became an obligatory
part of a model and the only way to justify its qual- Most traditional type of validation is validation of
ity. Increasing complexity of models also led to transport flows.
greater complexity of the validation procedures. Considering previously mentioned guidelines
Validation procedure for the verification of the and recommendations we suggested following cri-
adequacy of the National traffic model for the teria to be accepted by the client for the National
Republic of Croatia (NTMC) was based on fol- Transport model for the Republic of Croatia:
lowing documents: – R2 > 0.9
– JASPERS Appraisal Guidance (Transport): The – 65% of GEH <5
Use of Transport Models in Transport Planning – 85% of GEH <10
and Project Appraisal, August 2014 – difference in transport work < 3%.
– Design Manual for Roads and Bridges, Volume In the Table 1 goodness of fit is presented.
12, 1997. Only two indicators do not match expected
– Variable Demand Modelling—Convergence criteria, but this is for season traffic, where input
Realism and Sensitivity, TAG Unit 3.10.4, 2010. data were less reliable (no household survey, data
– Critical opinions on these documents and exam- about overnight stays, count data). With the use
ples of good practice have also been taken into of matrix estimation for internal (Croatian) traffic
account. numerical results would be much better and would
For the freight model, the calibration and vali- fit all criteria. We do not encourage it, as this would
dation process comprises three levels: decrease quality of forecasting models.
– total national freight volumes by mode [tonnes],
– port and border crossing volumes [tonnes],
– freight vehicle volumes at 480 count locations.

4.1 Passenger demand validation

Validation criteria for demand is less standardized
than for validation of the assignment. Mostly due
to lack of independent statistical data (as it is case
with traffic counts and assignment). Nevertheless,
few quantitative and qualitative tests were done to
prove the model. Figure 4. Validation of travel times distributions.

718

TIS2017_Book.indb 718 2/8/2017 8:09:12 PM

 Table 1. Goodness of fit. Table 2. Recommended elasticities of demand to a change
in supply.
Difference
Correlation GEH < 5 GEH < 10 [veh*km] Result
Travel Result [change in
Offseason 0.94 67% 88% < 1% time Recommended [change public
weekday Mode change elasticity in car] transport]
Offseason 0.91 65% 91% < 1%
peak hour car ± 20% <−2.0 −0.29 0.58–0.66
Season 0.96 70% 82% < 3% to –0.38
weekend public ± 20% 0.28–0.35 −1.42
Season 0.92 61% 84% < 1% transport to −1.66
peak hour

For public transport the sensitivity is also tested

In freight demand model for all validation levels, to a change in travel time. Sensitivity to the change
a satisfying model quality was achieved. The corre- in travel time is greater than for passenger cars and
lation factor R2 is 0.92 and 95 percent of the count is 1.42 to 1.66, but still well within benchmark.
locations have a GEH < 5. Based on the sensitivity analysis we concluded
that the model responds realistically on systemic
4.3 Realism test and developmental changes and is therefore suitable
to offer realistic traffic predictions. This test demon-
If a model adequately reproduces the existing situ- strates an ability for a real change in modal split.
ation, it does not yet mean that it is appropriate for
traffic forecasts. The demand model should also
behave realistically. A change in the traffic supply 5 APPLICATION AND USE OF THE
should lead to a realistic change in the demand. A NATIONAL TRANSPORT MODEL
change in the demand should be consistent with
general experiences. After development, calibration and validation of
A change in travel time or trip price (cost in one the transport model, it can be used to assess current
word) particularly affects the mode choice and and future conditions, to identify potential strate-
trip distribution. This impact must be in realistic gies and measures to improve conditions and to
limits. assess the impacts of these strategies and measures
Acceptability of the model response is deter- on the transport system itself and on influenced
mined by the elasticity of demand. The equation processes, like environmental impacts, accessibility
of arc elasticity is as follows: and social inclusion, social impacts and impacts on
economic development.
e = (log (T1) − log (To)) /(log(C1)
/ (log(C1) − log (Co (2)
5.1 Analysis of current conditions
where e = elasticity, T1 = demand after change,
To = demand before change, C1 = changed travel The model was used to analyse current conditions
cost and Co = original travel cost on the Croatian road network and on the National
Recommended elasticities for 20% changes in public transport system. Examples are described
costs are presented in Table 2. below.
According to the WebTAG recommendations, Of course the highest traffic flows can be
the elasticity on travel time must be significantly observed on the motorway and trunk road net-
greater than on monetary costs, but not greater works from Zagreb to all parts of the country and
than −2.0. This means that a change in travel time around the larger metropolitan areas in Croatia,
has a significantly greater impact on the demand mainly Zagreb, Zadar, Rijeka, Split, Varaždin,
than a change in monetary cost. However, also Osijek and Dubrovnik. The traffic flow analysis is
the later influences to some extent. But with the the basis for identifying the major OD relations in
increasing value of time, travel time gains impor- the country and to determine external impacts like
tance, whilst the impact of direct monetary costs environmental impacts emission of pollutants and
decreases. noise.
Sensitivity of the model must be within the For understanding the internal impacts and to
recommended limits. For passenger cars the identify bottlenecks in the network and potential
sensitivity is tested to a 20-percent change in travel shortcomings, it is necessary to relate the actual
time. Sensitivity for cars is between 0.29 and 0.38. This flows to the provided capacities. The Figure 5 shows
is consistent with the WebTAG recommendations. the volume/capacity ratio, again for the off-season

719

TIS2017_Book.indb 719 2/8/2017 8:09:13 PM

 conditions and the conditions within the summer
season.
Differences are clear; while in the off-season
situation, high volume/capacity ratios can be
observed mainly around the major cities, and here
more in the continental cities of Croatia, in the
summer season high volume/capacity conditions
occur also on the motorway network towards the
Adriatic coast and in and around the Adriatic cit-
ies. Levels above 75% volume/capacity ratio are
critical, potentially bearing the risk of congestion
and traffic breakdowns at peak hours.
Detailed analyses can be carried out for public
transport. To display supply/service quality, the Figure 7. Accessibility of city centres by public
Figure 6 to the right shows the accessibility of the transport.
city centres of the major cities from the whole ter-
ritory of the country in 30 minute steps. This could
also be compared to similar accessibility plots for Model results can also be used for a more com-
car traffic. plex analysis, for example for defining so-called
The accessibility analysis shows that there are “Functional Regions”. Functional regions are
service gaps between the cities, mainly between areas with a high frequency of internal regional
Zagreb and Zadar and Zagreb and Osijek, with interaction. The concept of functional regions is
very long access times or even no public transport used worldwide to understand and define func-
at all. However, these are remote areas with low tionally connected areas that need to manage the
population densities. transport system across administrative borders.
More graphical and quantitative analyses are The most commonly approach to define functional
possible and have been carried out for the current regions is using data on population commuting to
conditions represented in the base year model. work and school because the pattern of daily com-
muting rule is a good approximation for staging
other types of interaction.
The National Transport Model has been used
to determine commuter trips into and out of the
major cities as a basis to define functional regions
for Croatia (see Figure 7).

5.2 Analysis of future conditions

Since the National Transport Model of Croatia
not only represents current conditions but was
designed also for the forecast horizon years
2020, 2030 and 2040, the analyses can be carried
out also for the future transport conditions. To
identify future bottlenecks and shortcomings, a
so-called do-minimum scenario is used. In the do-
Figure 5. Validation of link volumes. minimum scenario, future demand is calculated
based in forecasts of socio-economic and behav-
ioural development, whereas the transport net-
works represent current conditions only enhanced
with those measures and projects that are already
under construction now or are planned and fully
financed.

5.3 Identification of strategies and measures

Analyses of current and forecasted future traffic
conditions, accessibility and transport impacts are
Figure 6. Road traffic—volume/capacity ratio for an important input to develop strategies and meas-
annual average (ADT) and for seasonal traffic (ASDT). ures to alleviate the identified shortcomings and to

720

TIS2017_Book.indb 720 2/8/2017 8:09:13 PM

 exercises. The National Transport Model forms the
basis for the development of these models repre-
senting smaller areas.
Technically, the National Model can be used to
extract so-called sub-network models, then form-
ing the basis for the implementation of further
details, like more detailed zoning structure in the
respective area of interest, more detailed allocation
of socio-economic data based on the refined trans-
port zones, adding details to the road network and
adding local public transport lines and services.

Figure 8. Display of commuter flows for the definition 7 SUMMARY

of functional regions.
The Ministry of Maritime Affairs, Transport and
transform the transport system in accordance with Infrastructure has commissioned the development
more general transport visions and objectives. of the National Transport Model in preparation of
The development of strategies and measures is the development of the National Strategy. The con-
based on the overall Vision and the defined objec- sultant team have finalised the work on develop-
tives of the National Transport System. These ing this powerful tool in form of a synthetic 4-step
objectives include: model, by developing the base year model with net-
work model, passenger and freight demand mod-
– Ensure economic, social and environmental
els, have calibrated this model with survey data and
sustainability,
have validated it against empirical data. The result is
– Provide accessibility and social inclusion,
a robust model capable of forecasting future trans-
– Increase traffic safety,
port conditions and calculating impacts of exog-
– Increase transport efficiency and quality of
enous changes (like political and socio-economic
services,
conditions) and endogenous changes in form of
– Reduce negative impacts on the natural, man-
transport strategies and measures.
made and social environment.

REFERENCES
6 DEVELOPMENT OF REGIONAL
AND URBAN TRANSPORT MODELS Dieter Lohse: Ermittlung von Verkehrsströmen mit
n-linearen Gleichungssystemen unter Beachtung von
Numerous regional and urban transport and Nebenbedingungen einschließlich Parameterschät-
mobility master plans are now under development zung (Verkehrsnachfragemodellierung: Erzeugung,
throughout Croatia. All regions and metropoli- Verteilung, Aufteilung), Dresden, 1997.
tan areas are carrying out these master plans. One European Commission—Directorates General for Energy,
important component of these regional and urban for Climate Action and for Mobility and Transport:
master plans is the quantitative analysis of current EU Energy, Transport and GHG Emissions—Trends
and forecasted conditions and the impact assess- to 2050, ISBN 978–92–79–33728–4, 2013.
NTM Croatia Consortium: NTM Croatia Model Devel-
ment of proposed measures and changes. Regional opment Report, 2016.
and urban transports model are needed for these

721

TIS2017_Book.indb 721 2/8/2017 8:09:14 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The effects of the different uses of mobile phones on drivers:

Experimental tests on an extra-urban road

S. Zampino
AIIT Puglia e Basilicata, Italy

ABSTRACT: It is well known that the alteration of the cognitive load due to the use of mobile phones
influences the driving behavior.
The development of smartphones, which allow mobile phone calls, driving navigation, text messag-
ing, etc., introduces more variables in terms of effects on the driver behavior and, consequently, on road
safety: the misuse of these phone devices is one of the most serious causes of accidents.
In order to develop an easy method to evaluate the relationships between cause and effect, a direct
experimentation on the road was performed, using a GPS navigation system enabled to monitor the driv-
ing behavior on a four lanes rural road.
Three groups of drivers were selected by age: their driving performances were evaluated while making
a phone call, writing a text message, setting a navigation route.
During the driving tests, different parameters were analyzed (i.e. acceleration/deceleration, lateral posi-
tion of the vehicle, use of the braking system).
The results of the driving tests highlighted dangerous behaviors during the most complex tasks, with
particular reference to the lane-keeping.

1 INTRODUCTION the “smart” devices can have on the vehicle speed

and trajectories is not yet well known.
1.1 Driving behavior and mobile phones
In normal conditions, driver behavior is not eas- 1.2 The multi-tasking interference
ily predictable and when motorists are engaged in It is widely acknowledged that the effects on driv-
multiple tasks, brain concentration can be danger- ing due to the interference of multiple tasks, affect
ously diverted from the main driving tasks. As a the brain performance, as confirmed by functional
consequence, the driving parameters (i.e. time of Magnetic Resonance Imaging (fMRI) experiments
reaction, speed, acceleration, lateral position) are conducted by the Carnegie Mellon University
significantly altered. (Pittsburgh, Pennsylvania). The experiments show
These effects are properly explained by the the- that the effects of cell phone speaking reduce the
ory of the “single channel bottleneck” in human activity of the parietal lobe related with driving
information processing, with specific reference to (Just et al. 2008).
high speed tasks (Craik 1948, Welford 1967).
From this perspective, the impairment caused
by the use of mobile phones while driving, has to
be considered a serious health problem. In many
countries, the legislation makes it illegal the use of
handheld cell phones while driving, although the
legislation does not always impose specific sanctions
related to the increasing complexity and to the multi-
ple possible uses of the most modern phone devices.
Although people overestimate their ability to
multitasking, the high level of impairment caused
by modern smartphones is perceived by drivers
(Wogalter & Mayhorn 2005) and road agencies (Fig
1), but to which extent the use of smartphones is Figure 1. Example of ITS message on a National Ital-
dangerous and which kind of effects the misuse of ian road.

723

TIS2017_Book.indb 723 2/8/2017 8:09:16 PM

 In order to comprehend the mental processes of or, in a different perspective, the effects of an ordi-
the human brain, many studies about the effects nary phone call are limited to the external loop of
of the use of mobiles have been conducted with the human brain by reducing the effects of the eye
specific reference to phone calls. movements (Cooper et al. 2013).
These studies prove that the consequences due In other words, the reasons why the lane-keeping
to the use of mobile phones are more complex than performance appears to be unaffected during ordi-
it was initially supposed. The driving behavior has nary phone calls can be due to the way in which the
been deeply investigated, in terms of sensitive cogni- hierarchical control operates: driving, while making
tion and cognitive workload, analyzing the different or receiving a low-concentration phone call, leaves
resources and the way in which the hierarchical con- less residual attention to other tasks directly related
trol, which characterizes the human mind works. to eye movements which affect the lateral positon, but
In more detail, the effects of the mental workload does not impair the driving automatic operations.
on visual search and decision making, studied by When the auditory tasks became more complex
visual detection and discrimination tests (performed or are combined with more complex visual tasks,
while driving) cause different levels of impairment their effects on the automated operations become
to spatial gaze concentration, visual detection and progressively more significant.
discrimination (Recarte & Nunes 2003). The different ways in which the auditory tasks
The four dimensions multiple resources model affect the driving behavior are demonstrated by
(Wickens 2002) states that there are four dimen- further experiments about the difference between
sions, which have influence on time-sharing per- the use of handheld and hands free cell phones,
formances. According to this theory, two tasks which, contrarily to common expectations, consid-
demanding the same performance (i.e. visual tasks) erably influence the not automated driving opera-
interfere with each other more than tasks demand- tions, in case of prolonged use (National Safety
ing separate levels of perceptual modalities (i.e. Council 2012).
visual and auditory tasks).
More recently, the hierarchical model proposed
by Logan and Crump (2009) states that there are 1.3 The advent of the smartphones technology
two separate control loops: the outer loop and If the effects of the use of mobile phones while
the inner one. The inner loop is automated and driving have been deeply investigated with regard
operates without knowing what the outer loop is to phone conversations, the most recent wide-
doing. spread use of smartphones has introduced a new
According to this theory, the ordinary driving perspective.
tasks, with practice, need a progressively reduced As a matter of fact, modern smartphones have
level of control: with extensive practice, the driv- progressively changed the way of using the tra-
ing tasks are gradually encapsulated in automatic ditional phone devices and the diffusion of the
operations (Medeiros-Ward et. al. 2014). social networks combined with the introduction of
When performing different tasks, such as driving innovative applications (i.e. Facebook, Instagram,
and speaking, the human brain switches between Whattsapp, Google maps) has progressively stimu-
primary (or automated) tasks and secondary tasks: lated a very different approach to the phone use.
consequently, the automated operations are par- The use of the most modern phone devices is no
tially impaired. While the effects on the “drained more simply based on auditory tasks, but it implies
attention” are not perceived during ordinary driv- more demanding brain tasks and the touch-screen
ing tasks, the switching process reduces the time of technology involves the direct observation of the
reaction whenever the driver has to deal with exter- device screen not only for text messaging (Reed &
nal not ordinary events or more demanding extra Robbins 2008)
tasks. Other unperceived effects are related to the The effects of such a different use are much more
visual information, which can be altered, affecting dangerous than the traditional use of mobiles for
the so-called “intentional blindness”. the following reasons:
The extreme unpredictability of the driver
behavior and the different impacts related to the − tasks which imply visual activities or text messag-
different kinds of simultaneous extra-tasks are ing are much more complex than simply speak-
demonstrated by experiments, which prove that ing, and can affect the information processing
the increase of external workload due to the most stages, more than simple auditory tasks;
ordinary use of handheld phones does not neces- − performances are less predictable and look-
sarily affect vehicles lateral position. The reason of ing at the phone screen inevitably affects visual
such a behavior can be differently explained, on channels;
the basis of both the theories reported: low-impact − focal vision is diverted from the road and
auditory tasks have no influence on visual tasks, from the car instruments, while unnatural eye

724

TIS2017_Book.indb 724 2/8/2017 8:09:17 PM

 movements, or the abnormal use of the periph- the peculiarities of the general driving attitudes of
eral vision alter the ordinary space perception each motorist (i.e. standard lateral deviation, use
(Maples et al. 2008, Robbins & Jenkins 2015). of the braking system).
In order to monitor the driver behavior during
The problem is emphasized by the constant
the driving tests, it was made use of three devices:
increase of the misuse of the “smart” devices.
a WI-FI Action-camera (VTIN—mod EyPro
According to a recently published AT&T research
VOD001B), a standard HD camcorder and a com-
(http://www.digitaltrends.com/mobile, May 2015),
mon phone device (Apple I-Phone 5 s) abled to
70% of drivers declare to use their smartphones
record the driver behavior with reference to speed,
while operating vehicles. While texting and sending
acceleration, lateral position (detected with the
e-mails are the most common activities, an unex-
GPS app—Sygic Navigation system—All regions
pected 27% of the interviewed drivers declared of
2016) (Fig. 2).
checking Facebook, 14% of Tweeting, 14% admit-
The choice of using ordinary devices and a
ted of using Instagram, 11% of using Snapchat, not
common application to record the data was made
to mention the use of smartphones to take “selfies”
in the prospect of reproducing the tests in other
while driving.
contexts.

2 GENERAL METHOD 2.3 Overall design

The tests were performed over the course of short
2.1 Objective of the study
repetitive drives.
The aim of this paper is to give a contribution The car used for the tests was a common model
to the comprehension of the actual effects of the of Peugeot (mod. 206-1.1, 60 HP), equipped
increasingly widespread and diversified use of with the system of cameras and GPS previously
mobile smartphones while driving, through a direct described (Fig. 3).
investigation, under normal but strictly controlled In particular, the phone Dash Camera was
and pre-defined driving conditions. positioned in correspondence to the central axis
In particular, the goal of the study was accom- of the vehicle, while the Action-camera was posi-
plished in a series of direct driving experiments in tioned on the left of the driver, in proximity to the
which a group drivers, selected by age and experi- windshield.
ence, performed different tasks involving different The driving experiments were conducted on a
uses of their smartphones while driving one a pre- dual carriageway (4 lanes) rural road (Fig. 4), on
defined road section. a pre-selected straight road section (S.P. n. 364 –
Lecce-S. Cataldo, Eastbound – km 1,200–km
2,640–length 1440 m–lane width 3.50 m–GPS
2.2 Description
coordinates of the initial point 40°22’0.48”N –
In order to collect the data, rather uniform driving 18°13’10.56”E).
conditions were pre-defined: the tests were con- Three groups of participants (16 males, 5
ducted on the same road section, driving the same females, 21 drivers in total), were selected accord-
car in the same hours of the day, under similar ing to the following age and Driving Experience
weather conditions (Temperature 27–32°C, Wind (DE) criteria:
Speed 2–5 m/s) and under uniform conditions of
light traffic.
The different tasks to perform during the tests
were standardized, while the drivers had to use
their personal devices, supposing a higher level
of familiarization and a better knowledge of the
Operating System.
Participants were selected on condition of
declaring to be regular users of smartphone apps,
to avoid dangerous consequences during the visual
tasks.
Before starting each experiment, participants
were carefully instructed about the tasks and the
security warnings.
Their standard behavior (without using any
device) was tested on the road and recorded, to
allow a preliminary evaluation and to highlight Figure 2. Devices and SW used during the tests.

725

TIS2017_Book.indb 725 2/8/2017 8:09:17 PM

 The driving tests were conducted at a reference
speed between 55–75 km/h, according to the trial
procedure reported below.
10’ Welcome and general instructions
5’ Familiarization with the car
6’ First Drive (baseline drive—full control) return
to the starting point
6’ Second Drive (Phone call) return to the starting
point
6’ Third Drive (Text messaging) return to the
starting point
Figure 3. Cameras and GPS devices positioned inside 6’ Fourth drive (Google maps route) end of the test
the car. For each participant the instructions to interact
with the smartphones were repeated at the same
point of the route by an instructor, according to a
general standardized procedure
The continuous presence of the instructor dur-
ing the tests was necessary in order to prevent the
risk of accidents.
No overtaking was allowed during the tests.
During the experiments, the basic data and the
driving parameters were simultaneously detected and
saved in the phone device (enabled to measure the
speed and acceleration parameters) in a cloud direc-
tory and in the memory card of the action camera
(abled to measure the lateral displacements with the
Figure 4. The straight section of the road where the use of optical viewfinder positioned upon the car
driving tests were performed. windshield). The standard HD camcorder was used
to record the general data of each test, inside the car.
Group 1 – young drivers aged between 18 and 27 The devices and the apps used to collect the data
years (YD—DE < 10 years – 7 drivers); were activated by the instructor seated on the front
Group 2 – expert drivers aged between 28 and 54 passenger seat of the car.
years (MD—11 < DE < 25 years - 7 drivers); With regard to the SW app used to record the
Group 3 – elder drivers aged between 55 and 65 test data, the Sygic application is not only a GPS
years (ED—DE > 26 years – 7 drivers). navigation software, but it also allows memorizing
Participants, who previously reported having a digital movie of each route by a Dash camera
normal or corrected-to-normal vision, were asked SW connecting the digital camera of the electronic
to interact with their phone, performing three sep- device and recording the main kinematic data of
arate tasks, according to the following standard- each route.
ized sequence:
A. Firstly, the drivers were asked to make a hand- 2.4 Definition of the tasks
held phone call (with a pre-defined content);
The tasks assigned to each motorist were pre-
B. Secondly, the selected drivers were asked to
defined, in terms of sequence and complexity, in
write and send a pre-defined text message with
order to analyze and compare different behaviors
the Whattsapp application;
under different circumstances.
C. Finally, each driver was asked to select a route
The smartphone exercises that participants were
to a predetermined destination, using the Google
required to perform while driving are showed in
Maps application.
the following Table 1.
After the first drive without performing any extra The phone call message was defined in order to
task (Full control), the pre-defined “smartphone evaluate the effects of a brief phone call, although
tasks” were performed in three different and con- the visual task of observing the speedometer to
secutive drives along the selected road section, each check the speed added a slightly more complex
task starting after returning at the same point (km activity contemporary to the auditory task.
1,400) of the road section and never exceeding the The following (visual) task was much more com-
final point (km 2,400) of the same road section. A plex and demanding than the previous one and
digital chronometer was used to record the duration was defined in relation to the length of the road
of every single task. section (1000 m).

726

TIS2017_Book.indb 726 2/8/2017 8:09:18 PM

 Table 1. Pre-defined messages/address for each task.

Task 1 Phone call Ciao, come stai? Io mi

trovo sulla strada Lecce
San Cataldo, e sto
andando a Sancataldo.
(After looking at the
speedometer) Attual-
mente viaggio alla
velocità di … chilometri
all’ora.
Task 2 Wapp Message Mi trovo a Lecce, Figure 5. Frames of the videos captured with the Dash
arriverò tra 10 minuti. Camera.
Task 3 Gmaps Route Piazza G. Mazzini,
10, Lecce.
transverse displacements relative to the central
lane axis.
In order to analyze and compare the driving
perfor-mances, the main data of each drive were
The final task was different from the second, extrapolated and reported in a digital worksheet.
shorter and less demanding in terms of brain
effort, but more demanding in terms of visual
application. 3 RESULTS AND DISCUSSION

2.5 Data acquisition and processing 3.1 The analysis of the driving performances
The movies recorded with the Dash camera appli- The analysis of each driver behavior was proc-
cation during the diving tests, contain the main essed through the extraction of the driving data
data of the route and, precisely, the GPS coordi- collected from the GPS app and from the video
nates and the instantaneous speed. Contemporary, camcorders.
the SW application recorded a set of kinematic In particular, for each participant and for each
parameters resulting in a series of diagrams (speed test, the following key measures were collected:
vs distance, speed vs time, acceleration vs distance, − Mean/maximum speed
acceleration vs time, etc.) collected in a “digital − Speed variation
travel book” instantaneously saved in a specific − Maximum deceleration/acceleration
external cloud directory. − Left side displacements
The videos of each test, simultaneously recorded − Right-side displacements
with the three cameras, allowed a double check of − Maximum left side displacements
the vehicle trajectories and were used to schedule − Maximum right-side displacements
each drive.
While, according to similar experiences, the The mean speed was evaluated with reference
speed data collected by the phone GPS antenna to the four drives, while the speed deviation was
and by the device internal accelerometer can be evaluated with reference to the second, third and
assumed to be precise with a margin of error of fourth drive.
almost 1% (Fazeen et al. 2012), the GPS app does Likewise, the standard lane-keeping behavior of
not allow precise lateral position measurements. the motorists was evaluated during the first drive.
For this reason, a simple method to calculate the The mean and maximum lateral displacements
vehicle lateral position relative to the axis of the recorded during the following drives were compared
road was implemented. with the data of the first drive. The driving param-
In particular, two optical viewfinders were posi- eters were analyzed for each group of drivers with
tioned upon the car windshield (exactly in front of particular reference to speed and lateral position.
phone camera and in front of the Action-camera). Further data were evaluated with regard both
The videos collected during the driving tests with to the longitudinal trajectories of the drivers and
the phone Dash camera SW and with the Action- to the relationship between speed and lateral posi-
camera were post-processed with DVDVideosoft tion during the tasks performed while using the
Free Studio software (Fig. 5). smartphones.
Each frame extracted from the videos (with an
average rate of 0.5 fps) was analyzed with C Thing 3.2 Speed
Meazure software, which (assuming as a refer-
ence the horizontal road markings and the marks The first aspect to be evaluated was the speed vari-
on the viewfinder) allowed the calculation of the ation during the phone tests.

727

TIS2017_Book.indb 727 2/8/2017 8:09:19 PM

 Assuming as a reference the baseline drive, all 0.10 m during third drive, 64% of drivers increased
the drivers decreased their speed while performing their mean displacement to the left by the same
the following three driving tests. measure, while during the second drive (phone
With regard to the mean speed, the data show call) the same increase of the mean fluctuation was
that the speed reduction was related to the com- limited to 18% of participants (Fig. 7).
plexity of the task. The mean of the maximum driver displacements
In particular, while the frequency of speed to the left were much higher during the third and
reduction was not significant during the second fourth drives, as shown in Figure 8.
drive, 53% of the motorists reduced their speed by During the third and fourth drives the mean
10 km/h during the third and fourth drive, when the of the maximun driver displacements to the left
tasks regarded the visual use of the smartphones. increased to more than 0.6 m.
The effects of the visual tasks were more signifi- While, during the second drive, for 52% of driv-
cant if related to the mean speed of each motorist ers the maximum displacement to the left was
during the tests (Fig. 6). more than 0.30 m, during the following drives, the
In particular, during the 3rd and 4th drive, more percentage of drivers whose maximun displacment
than 90% of drivers reduced their speed by more to the left was more than 0.30 m grew up to 82%
than 10% than the average speed. (more than 0.50 m, for 44% of drivers).
The average speed reduction during the second The displacments to the right during the sec-
drive was 12.70%, but it increased to 15.30% and ond, third and fourth drives were markedly lower
to 17.26% during the third and fourth drive. (Fig. 9).

3.3 Lateral position

The most significant results of the tests regarded
the lane-keeping performances.
For each driver, the lateral displacements
recorded during the driving runs were compared
to the standard behavior recorded during the base-
line drive.
In particular, assuming as a reference the cen-
tral axis of the lane, the mean fluctuation range Figure 7. Mean displacement to the left and to the
between the left and the right vehicle positions, right, during the different tasks (driver behavior).
measured during the baseline drive, was 0.28 m
(mean displacement to the left 0.17 m, mean dis-
placement to the right 0.11 m).
During the following drives, the mean fluctua-
tion range between the left and the right positions
increased to higher values (second drive: 0.34 m,
third drive: 0.44 m, fourth drive: 0.48 m).
The most significant result was that while the
mean range increased during the second drive
(phone call) by 0.06 m, it increased during the third
and fourth drive (Whattsapp and Google maps
exercises) by 0.16 and 0.20 m, respectively.
In more detail, 60% of participants increased Figure 8. Comparison of the mean displacements to
their mean displacement to the left by more than the left.

Figure 6. Mean speed: comparison with the baseline Figure 9. Comparison of the mean displacments to the
drive. right.

728

TIS2017_Book.indb 728 2/8/2017 8:09:19 PM

 In particular, for 52% of drivers the maxi- The following diagrams (Figs. 11, 12. 13) show
mum deviation to the right was more than 0.30 m the maximum lateral displacements recorded for
during the third and fourth drives (for 33%, more each motorist.
than 0.50 m). In more detail, while the maximum transverse
No significant difference was observed between fluctuations to the left and to the right were mostly
female and male drivers, except during the phone comparable during the second drive (phone call),
call exercise (second drive): in that case, the devia- the maximum fluctuation range increased during
tion to the left was slightly lower for female drivers. the the text messaging test and during the Google
The more significant differences were detected maps exercise, more significantly (Figs. 12, 13)
between the first group of drivers (Young Drivers) Table 2 reports the data of the statistical (nor-
and the third group of drivers (Elder Drivers). mal) distribution of the maximum lateral displace-
The mean of the maximum displacements to the ments, assuming as a reference the left edge of the
left was sharply higher during third and fourth drives lane (0.00 m left edge, 1.75 m central axis, 3.50 m
for elder drivers (0.45 – m—YD; 0.91 m—ED). right edge).
In comparison, the difference between the mean The prevailing tendency to occupy the left side
of the maximum displacements to the right was more of the lane during the third and fourth drives is
reduced (0.15 m—YD; 0.32 m—ED—IV drive). confirmed by the percentile curves (Fig. 14).
The mean of the maximum left-right fluctua- As reported above, during the third and fourth
tion recorded during the second, third and fourth drives a more noteworthy speed reduction was
drives show how the effects of the use of the smart- registerd.
devices during the three pre-defined tasks were lower
for young drivers than for elder drivers (Fig. 10).
All the data collected about the lane-keeping
performances, showed oscillating trajectories with a
prevailing trend to occupy the left side of the lane.
This trend was not exclusively related to the dif-
ficulty of the task (writing 3rd test more compli-
cated the 4th exercise), but it was more probably
influenced by the negative effect of the diversion
of the focal vision from the road. Figure 11. Maximum displacements during the second
The interruption to focal information and the drive.
consequent inadequacy of the sole peripheral
vision to define the correct spatial perception and
the vehicle trajectory (Land & Horwood, 1995)
induced the motorists to move away from the right
edge of the carriageway (perceived as more danger-
ous). For elder drivers, whose visual field is much
more reduced than young drivers, the tendency to
occupy the adjacent lane was more significant.
Figure 12. Maximum displacements during the third
drive.
3.4 Longitudinal trajectories and kinematic data
The tests conducted were used to study the longitu-
dinal trajectories of the vehicles while driving and
simultaneously performing smartphones extra tasks.

Figure 13. Maximum displacements during the fourth

drive.

Table 2. Statistical distribution of the maximum

deviation.

Mean St dev. N. Dist. 1.75

2nd drive 168,11 36,45 0,57

3rd drive 158,23 51,59 0,63
Figure 10. Mean of the maximum fluctuations for 4th drive 157,95 61,78 0,61
category.

729

TIS2017_Book.indb 729 2/8/2017 8:09:20 PM

 the lane (unconsciously perceived as safer than the
right side), increasing the risk of accident in case of
multiple vehicular interactions (i.e. frontal impacts
in case of single carriageway roads).
To prevent such a risky behavior, a suggestion
for smartphone manufacturers could consist in
implementing a GPS based application to inhibit
the activation of the most visual demanding smart-
phone apps, whenever the detected speed and
acceleration exceed a default maximum value.
Figure 14. Maximum displacements percentile curves
(axis 1.75 m).
REFERENCES

Cooper J.M., Medeiros Ward N., Strayer D.L. 2013. The

impact of eye movements and cognitive workload on
lateral position variability in driving. In Human Fac-
tors and Ergonomic Society (ed). Human factors Vol.
55 No. 5: 1001–1014.
Craik K.J.W. 1948. Theory of the human operator in
Figure 15. Example of diagram of the trajectories control systems II. Man as an element in a control
(driver n. 16). (2nd M 0.02, SD 0.04; 3rd M 0.07, SD system, British Journal of Psychology, 38, 142–148.
0.14; 4th M 0.03, SD 0.21). Fazeen M., Gozick B., Dantu R., Bhukhiya M., Gonzalez
M.C., 2012. Safe Driving Using Mobile Phones. In
However, the maximum deceleration (1.3–1.8 m/s2) IEEE Transactions on Intelligent Transport Systems.
demonstrates a limited use of the braking system Just M.A., Keller T.A., Cynkar J.A., 2008 - A decrease in
and for each test it was observed in correpondence brain activation associated with driving when listen-
ing to someone speak. In CMU Research Showcase.
to the most significant lateral displacments. Brain Research: 70–80.
For each drive, the longitudinal trajectories of Land M., & Horwood, J., 1995. Which parts of the road
the drivers were examined by diagrams deducted guide steering? Nature, 377 (6547), 339–340.
by the video processing (Figure 15). Logan G.D., & Crump, M.J.C., 2009. The left hand
For the vast majority of drivers, the analysis of doesn’t know what the right hand is doing. Psycho-
the trajectories showed a recurrent first displace- logical Science, 20, 1296–1300.
ment to the left followed by sudden trajectory cor- Maples W. C, De Rosier W., Hoenes R. Moore S., 2008.
rections to the centerline and to the right edge of the The effects of cell phone use on peripheral vision.
lane. Single fluctuations of variable entity, around In Optometry Journal of the American Optometric
Association, 79.
the central axis of the lane, followed (displacements Medeiros-Ward N., Cooper J.M., Strayer D.L 2014.
to the left of more than 1.5–2 m were observed). Hierarchical Control and Driving. In American Psy-
The trajectory Standard Deviation (SD) of each chological association (eds). Journal of experimental
driver increased during the 3rd an 4th drives (Mean psychology: General: Vol 143 No 3: 953–958.
SD 2nd drive, 0.07 m, Mean SD 4th drive 0.27 m—ref- National Safety Council, 2012. Understanding the dis-
erence central axis) and was higher for elder drivers. tract brain—Why driving while using hands-free cell
phone is risky behavior–White paper.
Recarte M.A., Nunes L.M. 2003. Mental Workload while
4 CONCLUSIONS driving: effects on visual search discrimination and
decision making. In American Psychological associa-
The results of the tests confirm the very dangerous tion (eds). Journal of experimental psychology applied,
Vol 9 No 2: 119–137.
effect of the visual use of smartphones while driving. Reed N., Robbins R., 2008. The Effect of Text Messag-
In particular, the data collected show how the ing On Driver behavior—A simulator Study. In TRL –
impairment due to the use of modern phone devices Published Project Report—PPR 367.
is not exclusively evaluable in terms of mental work- Robbins R. & Jenkins D., 2015. Eyes on the road, a
load but it is strongly influenced by the effects of the review of literature and an in-car study of driving
visual tasks. The concentration of the focal vision on whilst navigating–RAC Foundation—London.
the smartphone, diverts the look from the road and Welford A.T. 1967. Single channel operation in the brain,
produces the consequent negative effect of an incor- Acta Psychologica, 27, 5–2l.
rect reliance on the peripheral vision, with the further Wickens C.D. 2002. Multiple resources and performance
prediction. In Taylor and Francis Ltd (ed). Theoretical
drawback of the loss of the correct space perception. issues in Ergonomic Science, Vol 3, No 2: 159–177.
Drivers who rely on the sole guidance of the Wogalter M.S., Mayhorn C.B. 2005. Perceptions of driver
peripheral vision, perceiving of not checking exactly distraction by cellular phone users and nonusers. In
the space in front of the car, assume fluctuating tra- Human Factors and Ergonomic Society (ed). Human
jectories. They mostly tend towards the left side of Factors Vol.47 No. 2: 455–467.
730

TIS2017_Book.indb 730 2/8/2017 8:09:20 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Back-propagation neural networks and generalized linear mixed models

to investigate vehicular flow and weather data relationships with crash
severity in urban road segments

L. Mussone
Politecnico di Milano, Milano, Italy

M. Bassani & P. Masci

Politecnico di Torino, Torino, Italy

ABSTRACT: The paper deals with the identification of variables and models that can explain why a
certain Severity Level (SL) may be expected in the event of a certain type of crash at a specific point of an
urban road network. Two official crash records, a weather database, a traffic data source, and information
on the characteristics of the investigated urban road segments of Turin (Italy) for the seven years from 2006
to 2012 were used. Examination of the full database of 47,592 crash events, including property damage
only crashes, reveals 9,785 injury crashes occurring along road segments only. Of these, 1,621 were found
to be associated with a dataset of traffic flows aggregated in 5 minutes for the 35 minutes across each crash
event, and to weather data recorded by the official weather station of Turin. Two different approaches, a
back-propagation neural network model and a generalized linear mixed model were used. Results show the
impact of flow and other variables on the SL that may characterize a crash; differences in the significant
variables and performance of the two modelling approaches are also commented on in the manuscript.

1 INTRODUCTION of the research studies have been focused on the

likelihood of a crash without considering the crash
Road crashes are events that depend on a variety of outcome severity.
factors characterising human behaviour, weather, One of the main obstacles to investigations of
road pavement, vehicle stability and performance. this type is the limited availability of comprehen-
Crash events show different magnitudes when sive crash databases, and associated robust weather
evaluated with respect to the effects on road users and traffic databases. Nowadays, however, a con-
(crash severity), and the knowledge of the contrib- tinuous flow of environmental and traffic data
uting factors that affect the severity should be used is collected by local road agencies with sensors
to improve road safety through the action of trans- of increasing quality and performance (Chong &
port policy makers, designers and road agencies. Kumar 2003, Nekovee 2005). Contrary to what
Traffic volume, weather conditions, and road happened in the past, data are now frequently
characteristics affect crash severity in a multifac- collected in intervals of shorter duration. Hence,
eted way (Wang et al. 2013). Specifically, Theo- available databases can be associated and merged
filatos & Yannis (2014) pointed out that the few with others containing data collected over several
papers available mainly deal with roads operating years of observations, thus supporting new robust
under uninterrupted flow conditions, and recur inferences (El Faouzi et al. 2011).
mainly to logit modelling (Al-Ghamdi 2002, Golob To obtain reliable models and convincing results,
et al. 2008, Christoforou et al. 2010, Jung et al. the availability of high quality data representing
2010, Xu et al. 2013, Yu & Abdel-Aty 2013). Ear- the characteristics of drivers, together with traffic,
lier, Shankar et al. (1996) stated that crash severity weather, and pavement conditions is fundamental.
investigations had been historically limited to the Unfortunately, data describing every significant
localization of fatalities, even though the estimation factor affecting crashes needs work to associate
of the other severity levels (i.e., property damage the available contrasting information that could be
only—PDO –, possible injury, non-incapacitating long, arduous, and sometimes unproductive.
injury) could help in understanding the benefits The paper aims to bridge these gaps by providing
of safety-improvement projects. Seventeen years knowledge on factors contributing to crash severity
later, Xu et al. (2013) again underlined that most in an urban road network, considering only those

731

TIS2017_Book.indb 731 2/8/2017 8:09:22 PM

 influencing crashes along road segments. Data on Table 1 shows the number of crashes per year
crashes, traffic and the weather database of Turin’s and SL, and evidences the decrease in all the SL
road network (Italy) were collected and used to classes between 2006 and 2012. Assuming the year
calibrate and validate predictive models of crash 2006 as a reference point, the following years wit-
severity. The Back Propagation Neural Networks nessed a decrease in crash occurrence across all
(BPNN), a robust tool used to investigate complex severity classes.
phenomena without assuming any preliminary
hypotheses on the model, was used. But BPNN can-
2.2 Traffic data
not give an analytical formulation of the mathemat-
ical functions linking the variables that significantly Traffic data were provided by the 5T Company
affect a certain phenomenon, thus only a sensitivity (Telematics, Technologies for Traffic and Transport
analysis of the model can be performed. A General- in Turin), which monitors and controls over 300
ized Linear Mixed Model (GLMM) was also used urban traffic lights in Turin, and collects traffic data.
(with its analytical formulation) to compare and 5T uses induction-loop traffic sensors located along
assess results with those obtained by BPNN. the exiting lanes of the monitored intersections to
collect vehicle flow data at 5 minute intervals. It is
worth noting that from 2006 to 2012, the number of
traffic sensors available varied from 662 to 1051 due
2 DATABASE FORMATION
to the installation of new ones and the elimination of
some of the damaged ones. Figure 1a shows the por-
2.1 Crash classification
tion of the road network monitored by 5T in 2006.
Crash data were provided by the Istituto Nazionale di
Statistica (National Statistics Institute, ISTAT). The
2.3 Weather data
ISTAT database contains details on crash dynam-
ics and location, on the vehicles, and on gender and The Environmental Protection Agency of the Pied-
age of people involved, in accordance with current mont Region (ARPA Piedmont) provided data on
Italian legislation, specifically articles number 582, weather conditions. The Turin weather station
583 and 590 of the Italian Penal Code 2015 (Repub- considered in the paper is located in the city cen-
blica Italiana 2015). The Italian law considers a road tre (238 m a.s.l., 1.5 m off the ground, latitude
accident to be a crash when it results in at least one 45°.066667, longitude 7°.683333), and collects
injury, and crash consequences are classified into temperature, atmospheric pressure, wind speed
the following five Severity Levels (SL): and direction, solar radiation, and rainfall intensity
data on an hourly basis. The maximum distance
– Very Slight Injuries (VSI), when the most seri-
between the weather station and the farthest crash
ously injured person has a prognosis of less than
location included in the database was 9.6 km. Each
20 days;
crash record was associated with the weather data
– Slight Injuries (SLI), when the prognosis is
recorded at the time of the crash.
between 21 and 40 days;
– Severe Injuries (SEI), if the event causes an ill-
ness that endangers the life of the injured party, 2.4 Database formation
and/or the event results in permanent damage to
Only crashes that occurred along segments pro-
the brain or any body organ;
vided with valid and reliable traffic data were
– Guarded Prognosis (GPR), if the doctor can-
extracted from the main database and used. The
not determine the disability, and issues a report
database adopted for modelling is a subset (and
of “guarded prognosis” (pending resolution of
prognosis, the road crash must be considered
and treated as a determining factor); and Table 1. Injury crash records along segments of the
urban road network of Turin (Italy, 2006–2012).
– fatalities (FAT), including injured persons who
die within 30 days of the crash. Year VSI SLI SEI GPR FAT Total
The dearth of information in the ISTAT data-
2006 1317 173 36 33 22 1581
base was overcome by including crash data col-
2007 1303 205 69 42 29 1648
lected by Turin’s Municipal Police (TMP). All the
2008 1128 152 41 33 12 1366
records of the ISTAT database were matched up to
2009 1061 160 34 27 21 1303
the TMP database and implemented with the fol-
2010 1198 169 46 31 19 1463
lowing information: (a) historical data (time, near-
2011 1037 171 51 18 18 1295
est minute, day, month and year of the crash event); 2012 917 141 44 13 14 1129
(b) locality data (street name, house number); and 2006-‘12 7961 1171 321 197 135 9785
(c) generic information concerning crash SL.

732

TIS2017_Book.indb 732 2/8/2017 8:09:22 PM

 – pavement conditions (C2), which have been dis-
tinguished with a numerical variable indicating
the presence of water, snow or ice (0 = unknown,
1 = dry, 2 = wet, 3 = slippery, 4 = icy/frozen,
5 = snowy); and
– the road signage (C3), which indicates if it was
absent (0), if it was composed of vertical signs
only (1), horizontal markings only (2), if both
were present (3), or if a temporary construction
signage was present (4).
The variables that reflect the characteristics of
vehicles A and B involved in the crash are:
– vehicle type (C4 and C6), ranging from 0 (pas-
senger cars) to 20 (quad), also including the case
of vehicles that fled the crash scene (19); and
– vehicle category (C5 and C7), from 0 to 8, in
which 1 represents cars, 2 buses, 3 trams, 4 heavy
vehicles, 5 industrial vehicles, 6 bikes, 7 motor-
cycles, 8 vehicles that fled the crash scene and 0
unclassified vehicles.
In modelling, variables describing roads and
vehicles were assumed as categorical. The vari-
ables describing drivers involved in the crashes
were assumed as numerical. They are:
– age (C8 and C11), which ranges from a minimum
of 10 (driver B) to a maximum of 89 (driver A);
this variable also assumed the null value in cases
of unknown/unrecorded age;
– age class (C9 and C12), which groups the ages
into 6 intervals ranging from 0 to 5: 0 in the case
Figure 1. (a) Turin’s traffic monitoring network operated of unknown/unrecorded data, 1 for very young
by 5T in 2006 (highlighted in black); (b) spatial distribu- drivers (15–19 years old), 2 for young drivers (20
tion of road crashes that occurred in 2006; and (c) crashes 24 years old), 3 for adults (25–64 years old), 4 for
which were associable to 5 min traffic flow (487 in total). elderly drivers (from 65 to 79), and finally 5 for
very old drivers (over 80); and
– sex of drivers (C10 and C13), which assumes the
a random sample) of the total number of crashes
value 0 in cases of unknown/unrecorded data, 1
that occurred and were recorded in the official
for males, and 2 for females.
database. Figure 1b shows all the crash records in
2006, while Figure 1c shows only crashes associ- In Table 2, the lowest values for ‘age of driver
ated with traffic flow data. A’ refer to scooter drivers, while those for driver
B refer to pedestrians or cyclists. Air temperature
(C14), wind speed (C15), solar radiation (C16),
3 DATA ANALYSIS AND TREATMENT and rainfall precipitation (C18) were assumed
as numerical with values that correspond to the
3.1 Variables measured values. The lighting condition (C17) was
assumed as a Boolean variable (0 = dark, 1 = light).
Table 2 lists the independent variables, their num-
The Traffic Flow (TF) variables (C20 ÷ C25) are
bering and labels, the type of variable, the unit of
numerical and represent the volume of vehicles per
measurement, and the range. The variables refer-
hour (veh/h) measured every 5 min across the crash
ring to the road are:
event, according to the time scale reported in Fig-
– road type (C1), which indicates the organization ure 2. Finally, the standard deviation (C26) for the
of the carriageways and the directions served seven flow values was added to the list to take into
(0 = unknown; 1 = one carriageway, one way; account flow fluctuations for the 35 min period
2 = one carriageway, two ways; 3 = two carriage- before and after the crash. Finally, the output vari-
ways, two ways; 4 = more than two carriageways, able indicating the SL (C27) was assumed numeri-
two ways); cal and ranging from 2 (VSI) to 6 (FAT).

733

TIS2017_Book.indb 733 2/8/2017 8:09:22 PM

 Table 2. Number and labels of variables. variance explained by the first eight components.
They account for about 81% of the total variance
Range for both databases, while the first two components
account for about 61%. The variables most linked
# Code Description Type u.m. min max
to the first component are road type, road signage,
1 C1 Road type C _ 0 4 and age of driver A, whereas those linked to the
2 C2 PC C _ 0 5 second component are light/dark, light radiation,
3 C3 Road signage C _ 0 4 and air temperature. Finally, traffic flows (TF1-
4 C4 Veh. A type C _ 0 20 TF7) are mainly linked to the third component.
5 C5 Veh. A cat. C _ 0 8 This means that the set of variables relating to road
6 C6 Veh. B type C _ 0 20 and driver can explain about 45% of the variance;
7 C7 Veh. B cat. C _ 0 8 those related to meteorological conditions about
8 C8 Dr. veh. A age N _ 16 89 16%, and those related to flow about 8%.
9 C9 Dr. veh. A cl. age N _ 0 5
10 C10 Dr. veh. A sex N _ 0 2 3.3 Data treatment
11 C11 Dr. veh. B age N _ 10 86
12 C12 Dr. veh. B cl. age N _ 0 5 According to Table 1, the five SLs contained in the
13 C13 Dr. veh. B sex N _ 0 2 database were not equally represented and the dataset
14 C14 Air temp. N °C −7.5 +35.3 resulted imbalanced. This is not a problem for mod-
15 C15 Wind speed N m/s 0 9.95 elling approaches such as logistic regression, but it is
16 C16 Light radiation N W/m2 0 996 for machine learning tools, and especially Artificial
17 C17 Light/dark B _ 0 1 Neural Networks (ANN). With imbalanced datasets,
18 C18 Rainfall N mm/h 0 12.8 ANN could not find the correct relationships between
19 C19 TF1 (*) N veh/h 0 730 input and output for all categories present in the data-
20 C20 TF2 (*) N veh/h 0 750 set. Over-sampling (with data duplication) or under-
21 C21 TF3 (*) N veh/h 0 765 sampling (with data cancellation) techniques present
22 C22 TF4 (*) N veh/h 0 775 advantages and disadvantages: under-sampling can
23 C23 TF5 (*) N veh/h 0 570 remove important data and over-sampling can lead to
24 C24 TF6 (*) N veh/h 0 565 over-fitting problems. Studies on imbalanced datasets
25 C25 TF7 (*) N veh/h 0 494 have shown over-sampling to be more advantageous
26 C26 Flow st. dev. N veh/h 0 193 and useful than under-sampling (Chawla 2010).
27 C27 SL N – 2 6 The “focused re-sampling” method proposed by
Japkowicz (2000), which consisted of an oversam-
Notes: PC = pavement conditions, TF = traffic flow, pling of those examples that occurred in the minor-
Dr. = driver, veh. = vehicle, cl. = class, N = numerical,
B = Boolean, C = categorical.
ity classes (specifically FAT, GPI, and SEI), was
used. This approach implies duplication of the entire
subset of data until their count is of the same mag-
nitude as the most populated class. This approach
avoids other possible biases in re-sampling data.
Another task performed was data normaliza-
tion. Feature scaling, also called unity based nor-
Figure 2. Time scale used to aggregate Traffic Flows malization, was used for its simplicity. Let Xmin
(TF) across the crash event. and Xmax be the two extreme values (minimum and
A criticism may be made of the use of all seven maximum) of a variable X, the normalized vari-
flow values (TF1-TF7). In fact, some of these values able X’ (according to feature scaling) is:
(in particular TF5-TF7) refer to intervals after the
crash and therefore were caused by the crash itself. Table 3. Percentage of variance explained by the first
Nevertheless, the reason for using them is that they eight components in PCA.
belong to the same time series and the interrupted
nature of flow in urban roads makes each interval Component Simple value Cumulative value
(though not as long as 5 minutes) a story apart, even 1 44.72 44.72
when there is no crash. In addition, results will show 2 16.34 61.06
that they play a different role in the models. 3 8.32 69.38
4 6.72 76.10
3.2 Database information content 5 4.91 81.01
6 4.38 85.38
The Principal Component Analysis (PCA) (Lebart 7 3.49 88.88
et al. 1977) was used to investigate the informa- 8 2.28 91.16
tion content of the database. Table 3 reports the

734

TIS2017_Book.indb 734 2/8/2017 8:09:23 PM

 X X min pendent variables listed in Table 2, the hidden layer,
X' = (1)
and the output layer corresponding to the SL, tested
X max X min
with one neuron. All categorical variables are coded
in binary format to reduce connection between their
All the input variables were then normalized values. Finally, the best model was found to be made
according to eq. 1, hence with values falling within up of 25 neurons in the hidden layer for the model.
the range [0,1]. The output variable (the output) is It has a MSE lower than 0.08, which means that
numerical, ranging from 2 to 6. there are only 8 errors to each 100 classifications.

4 DATA MODELLING AND RESULTS 4.2 Generalized Linear Mixed Model (GLMM)
For the analysis of multilevel data, random clus-
4.1 Back-Propagation Neural Networks (BPNN) ters and/or subject effects should be included in the
The BPNN used in this work is an example of an regression model to account for the correlation of
Artificial Neural Network (ANN) model, ANN data. The resulting model is a mixed model includ-
models have a classical multilayer topology with ing fixed and random effects. Mixed models for
feed-forward connections. Cybenko (1989), and continuous normal outcomes have been proposed
Hornik (1991), described the capability of ANN in for non-normal data and are generically classified
approximating any function belonging to the Leb- as Generalized Linear Mixed Models (GLMMs).
esgue two space (L2 space) with minimum error. The extension of the methods from dichotomous
Applications regarding transport, planning, con- responses to ordinal response data was actively
trol fields, and crash analysis are numerous start- pursued in the reviews of Agresti & Natarajan
ing from the 90’s (Dougherty 1995, Mussone 1999, (2001).
Mussone et al. 1999). Other contributions have The GLMM model is a regression model of a
faced the problem of crash prediction or sever- response variable that contains both fixed and ran-
ity (Abdelwahab & Abdel-Aty 2001, Chong et. al. dom effects and comprises data, a model descrip-
2004, Delen et al. 2006, Baluni & Raiwani 2014). tion, fitted coefficients, co-variance parameters,
The downside in using the BPNN approach is design matrices, residuals, residual plots, and other
that the relationships between variables are in a diagnostic information. Fixed-effects terms usu-
black box (the hidden layer of Figure 3), and no ally refer to the conventional linear regression part
analytical formulation between input and out- of the model. Random effects terms are associated
put can be directly obtained. The effects of inde- with individual experimental units taken at ran-
pendent (input) variables can be interpreted only dom from a population, and account for variations
through a sensitivity analysis of the model. between groups that might affect the response. The
The BPNN models were calibrated and validated random effects have prior distributions, whereas the
with the Levenberg-Marquardt training algorithm. fixed effects do not.
Performances were evaluated according to Mean The GLMM model structure is:
Squared Errors (MSE) through the three phases
of train, test, and validation. The model was con-
structed with an input layer including the 26 inde- ⎛ σ2 ⎞ (2)
y i b ≈ Distr ⎜ μ i , ⎟
⎝ wi ⎠
g (μ ) X bZ δ (3)

where, yi = the i-th element (dependent variable) of

the y response vector, b = the random-effects vector
(complement to the fixed β), Distr = a specified con-
ditional distribution of y given b, μ = the conditional
mean of y given b, and μi is its i-th element, σ2 = the
variance or dispersion parameter, w = the effec-
tive observation weight vector (wi is the weight for
observation i), g(μ) = link function that defines the
relationship between the mean response μ and the
linear combination of the predictors, X = fixed-
effects design matrix (of independent variables),
β = fixed-effects vector, Z = random-effects design
matrix (of independent variables), and δ = model
Figure 3. Back-propagation Neural Network structure offset vector (residuals). The model for the mean
for SL modelling adopted in this investigation. response μ is:

735

TIS2017_Book.indb 735 2/8/2017 8:09:24 PM

 μ = g −1 (η
ˆ) 5 DISCUSSION
(4)
5.1 Sensitivity analysis of BPNN
where g–1 = inverse of the link function g(μ), and
η̂ = linear predictor of the fixed and random effects In the case of the BPNN model, a sensitivity analy-
of the generalized linear mixed-effects model: sis was carried out to assess how output changes
by varying input normalized variable values in the
η β δ (5) range [0,1] one by one. With this aim in mind, a
first set of scenarios, referring to a particular set of
According to the Wilkinson notation, the input variables, was prepared. In addition to basic
GLMM model has the following structure: scenarios where variables are all zero or 1, another
six scenarios were considered to study particular
y ∼ fixed + (random1|group1) + ... + (random N| combinations of variable values (in Figure 4, from
group N) (6) 4a to 4f). These scenarios aim to consider some
possible and typical crash situations involving
where, the terms “fixed” and “random” are associ- male or female drivers, during daytime or at night-
ated with independent variables and contain fixed time, with rainy weather or dry road surface, or
and random effects, N = number of grouping vari- with elder drivers.
ables in the model. Grouping variables are utility The effect of flow varies a lot for scenario and
variables used to group, or categorize, observa- flow itself. When TF4 is high (Figure 4a), the SL
tions, and are useful for summarizing or visualiz- is generally high for most of the scenarios, except
ing data by group. for elderly drivers. On the other hand, the effect of
For the SL output, a log link function and the TF7 (Figure 4b) depends very much on the par-
Probability Mass Function (PMF) for the Pois- ticular scenario, though, generally, a higher SL is
son distribution was used. The fit method was related to a higher flow. Low light radiation (night-
the ‘Laplace’ one. Finally, the best performance time to dawn) has a strong effect on young male
was calculated through the minimization of the drivers with dry pavement, while middle radia-
log-likelihood index; other indexes (i.e., Akaike’s tion has a strong effect on young female drivers in
information criterion—AIC, Bayesian information rainy conditions. Generally, a low radiation is more
criterions—BIC, and the Deviance parameter) were related to high SL than a high radiation.
also estimated to control the minimization process.
According to eq. 6 notation, the GLMM model 5.2 Model output comparison
that gives the best performance was:
According to Powers (2011), the two model outputs
C27 ∼ 1 + C8 + C9 + C11 + C14 + C16 + C18 were evaluated by confusion matrixes representing,
+ C21 + C24 + C25 + (1 | C1) + (1 | C2) for each output, the number of predicted cases (aij)
+ (1 | C3) + (1 | C4) (7) on the reference databases. In this case, the output
coincides to the SL, and aij are calculated on the
where Cx identifies the x-th variable (Table 2). In resampled databases obtained according to what
Table 4 the fixed effect coefficients are drawn with reported in Section 3.3.
a 95% confidence interval. The p values are lower There is s also interest in the measurement of its
than 0.001, with the exception of C9 (driver vehicle precision (the percentage of correct data predicted
A age) and C18 (rainfall intensity) which are lower in respect of the total predicted) and its recall (the
than 0.05. The Standard Error of estimates (SE) is percentage of corrected data predicted in respect of
generally much lower than the estimates, and lower the total to be predicted) capability. The main goal
and upper bonds of CI never include zero. Table 4 of learning is to improve the recall measurement
also reports the estimates for random parameters. without hurting the precision one. Tables 5 and 6
The effect of flows (C21, C24, C25) on SL has include the percentage of the predicted crashes to
a different sign, positive for C21 (TF4), which the total predicted for each SL, the “a priori” rate
anticipates the crash event, and for C24 (TF6), (PR), which expresses the complement of the recall
and negative for C25 (TF7). When flow after the rate, and the “a Posteriori” rate (PO) which is the
crash (C25) increases, it is more likely that the SL complement of the precision rate, according to the
decreases. The grouping variables are road type following equations (n is the matrix dimension):
(C1), pavement condition (C2), road signage con-
dition (C3) and vehicle A type (C4). The driver PRi = 1 - aii/(ai1 + … + ain) (8)
vehicle A age (C9) as well as the class age of vehi- POi = 1 - aii/(a1i + ... + ani) (9)
cle B driver (C11) are negatively related to the SL.
Furthermore, also light radiation (C16) and rain- Furthermore, comments on results are sup-
fall intensity (C18) are inversely related to the SL. ported by the estimation of their accuracy (A):

736

TIS2017_Book.indb 736 2/8/2017 8:09:24 PM

 Table 4. Fixed effects coefficients estimates and Random effects covariance parameters at 95% CIs) for the GLMM
segment model.

Variable Estimate SE p-value Lower Upper

Intercept 1.24210 0.2019 <10–3 0.84629 1.63790

C8 0.00594 <10–3 <10–3 0.00428 0.00759
C9 −0.06953 0.018 <10–3 −0.10540 −0.03367
C11 −0.00620 <10–3 <10–3 −0.00707 −0.00534
C14 0.00497 <10–3 <10–3 0.00318 0.00675
C16 −0.00016 <10–4 <10–3 −0.00023 −0.00010
C18 −0.07152 0.0203 <10–3 −0.11142 −0.03163
C21 0.00056 <2⋅10–4 0.002 0.00020 0.00092
C24 0.00111 <3⋅10–4 <10–3 0.00062 0.00161
C25 −0.00242 <3⋅10–4 <10–3 −0.00284 −0.00199
Group variable Estimate
C1 (Intercept) 0.13646
C2 (Intercept) 0.38224
C3 (Intercept) 0.19893
C4 (Intercept) 0.11375
Indexes
LogLikelihood −11465
AIC 22959
BIC 23053
Deviance 22931
R2 adjusted 0.3107

A = (a11 + a22 + … + ann)/Σaij (10)

Table 5 reports the confusion matrices for the

model calibrated through the BPNN. SL values
lower than 2 (corresponding to the PDO crash type)
and greater than 6 (which are unrealistic values) have
also been included in the tables considering that the
model output can fall outside the range of numeri-
cal values associated with each SL. The accuracy of
90% is certainly very high for BPNN model. PR and
PO rates are low with the exception of SL 2 and 3.
SL 2 is the more difficult to predict while SL 3 has
the largest number of wrong cases assigned to it.
Table 6 contains the confusion matrices for the
GLMM model. In this case, the capacity of SL pre-
diction is significantly lower than the one for BPNN
as indicated by the accuracy of 33%. GLMM has a
superior capacity to provide results within the SL
limits of 2 and 6, as confirmed by the absence of
values that fall outside of the two limits. PR and PO
rates are lower than the corresponding values for
BPNN, showing a greater difficulty in predicting SL
than the neural network modelling approach.
Comparisons with GLMM show a marked
superiority of BPNN modelling as regards per-
formance measured through confusion matrices.
Figure 4. Effect of TF4 (a) and TF7 (b) (variables 22 GLMMs, on the other hand, clearly show what
and 26) on SL (BPNN model) for different flow values variables are significant and their effect (sign and
(0.1, 0.5, 0.9). value of coefficients) though this is limited to the

737

TIS2017_Book.indb 737 2/8/2017 8:09:24 PM

 Table 5. BPNN model confusion matrix for segments (row percentage values in brackets), and “a Priori” (PR) and
“a Posteriori” (PO) rates.

Predicted SL
PR
Real SL <2 2 3 4 5 6 >6 CD (%)

2 21 859 257 71 39 60 7 1314 35%

2% 65% 20% 5% 3% 5% 0%
3 18 150 1056 6 0 0 0 1230 14%
1% 12% 86% 1%
4 0 0 0 1250 0 0 0 1250 0%
94%
5 0 0 0 0 1287 0 0 1287 0%
100%
6 0 0 0 0 0 1311 0 1311 0%
100%
PO – 15% 20% 6% 3% 4% – – –

Notes: CD = crash data in the resampled database.

Table 6. GLMM model confusion matrix (row percentage values in brackets), and “a Priori” (PR) and “a Posteriori”
(PO) rates.

Predicted PSL
PR
Real RSL <2 2 3 4 5 6 >6 CD (%)

2 0 79 534 593 100 8 0 1314 94%

52% 41% 45% 8% 1%
3 0 72 444 588 102 24 0 1230 64%
48% 36% 48% 8% 2%
4 0 0 300 850 100 0 0 1250 32%
24% 68% 8%
5 0 0 393% 702 507 39 0 1287 61%
55% 39% 3%
6 0 0 0 483 621 207 0 1311 84%
37% 47% 16%
PO – 48% 66% 74% 65% 26% – – –

linear effect of variables without considering their in the prediction of the SL when compared to the
possible reciprocal interaction. GLMMs. This is attributable to their greater capabil-
ity of accurately approximating any continuous and
non-linear function. On the other hand, GLMMs
6 CONCLUSIONS (like any analytical model) allow a readier inter-
pretation of model results. Other pros and cons in
The paper aims to achieve two goals: the eval- their use derive from the intrinsic characteristics of
uation of the crash Severity Level (SL) on statistical and neural network methods, as clearly
urban road segments using environmental vari- underlined by Karlaftis and Vlahogianni (2011).
ables (some of which, like short-term flow, are The authors suspect that the most significant limit
innovative for this type of research), and the of GLMMs for these applications is related to the
comparison of two different techniques for cal- constrained linearity of their functions. In addi-
culating SL, the Back-Propagation Neural Net- tion, missing data may have contributed to the
work model (BPNN) and the Generalized Linear fact that the BPNNs, which are known to be capa-
Mixed Model (GLMM). ble of overcoming this problem, achieved better
The results presented here provide new insights results.
into urban roads and fill a gap in the knowledge However, both approaches (BPNN and
acquired from the number of studies on rural free- GLMM), though with significant differences, indi-
ways and expressways reported in literature. cate that flows have a relevant role in predicting
From the use of the confusion matrixes tech- severity: this role is not limited to the flow when the
nique, BPNN models evidenced their superiority crash occurred (TF4), but also involves other flow

738

TIS2017_Book.indb 738 2/8/2017 8:09:25 PM

 data recorded before (TF1-TF3) and after (TF5- accidents using a series of artificial neural networks.
TF7) the crash. GLMM model shows the relevance Accid. Anal. Prev. 38, 434–444.
of TF3, TF6, and TF7 only, but the BPNN model Dougherty, M. 1995. A review of neural networks applied
evinces more complex relationships for all seven to transport. Transp. Res. Part C: Emerg. Tech. 3(4),
247–260.
variables. Weather variables also (i.e., rainy condi- El Faouzi, N-E., Leung, H. & Kurian, A. 2011. Data
tion and light radiation) show a strong relation in fusion in intelligent transportation systems: progress
some scenarios. and challenges—a survey. Inform. Fus. 12(1), 4–10.
In future research, generalized non-linear mod- Golob, T.F., Recker, W.W. & Pavlis, Y. 2008. Probabilis-
els will be used to consider the higher order effects tic models of freeway safety performance using traffic
and the interaction between variables. Moreover, flow data as predictors. Saf. Sci. 46, 1306–1333.
a mixed approach using both short-term flow and Japkowicz, N. 2000. The class imbalance problem: signif-
AADT values will be investigated to derive models icance and strategies. Proc. of the 2000 Intern. Conf. on
over a mid-to-long term period and investigate the Art. Intel. (IC-AI’2000), Las Vegas, Nevada.
Jung, S., Qin, X. & Noyce, D.A. 2010. Rainfall effect on
relationship between them. single-vehicle crash severities using polychotomous
response models. Accid. Anal. Prev. 42, 213–224.
Karlaftis, M.G. & Vlahogianni, E.I. 2011. Statistical
ACKNOWLEDGEMENTS methods versus neural networks in transportation
research: differences, similarities and some insights.
The authors thank Polizia Municipale di Torino, Transp. Res. Part C: Em. Tech., 19(3), 387–399.
and the Città Metropolitana di Torino for having Lebart, L., Morineau, A. & Tabard, N. 1977. Techniques
provided crash data. Thanks are also due to Con- de la description statistique: méthodes et logiciels pour
sorzio 5T s.r.l. for providing short-term flow data, l’analyse des grands tableaux, Dunod, Paris.
Mussone, L. 1999. A review of feedforward neural net-
and to the Environmental Protection Agency of works in transportation research, e&i Elektrotechnik
the Regione Piemonte (ARPA Piemonte) for pro- und Informationstechnik 116(6), 360–365.
viding weather data. Mussone, L., Ferrari, A. & Oneta, M. 1999. An analy-
sis of urban collisions using an artificial intelligence
model. Accid. Anal. Prev. 31, 705–718.
REFERENCES Nekovee, M. 2005. Sensor networks on the road: the
promises and challenges of vehicular ad hoc networks
Abdelwahab, H.T. & Abdel-Aty, M.A. 2001. Develop- and vehicular grids. Proc. Work. on Ubiq. Comp.
ment of artificial neural network models to predict e-Res., Edinburgh, Scotland, UK.
driver injury severity in traffic accidents at signalized Powers, D.M. 2011. Evaluation: from precision, recall
intersections. Transp. Res. Rec. 1746, 6–13. and F-measure to ROC, informedness, markedness
Agresti, A. & Natarajan, R. 2001. Modelling clustered and correlation. J. Mach. Learn. Technol. 2, 37–63.
ordered categorical data: a survey. Int. Stat. Rev. 69, Repubblica Italiana, 2015. Codice Penale (in italian).
345–371. Testo coordinato del Regio Decreto 19 ottobre 1930,
Al-Ghamdi, A.S. 2002. Using logistic regression to esti- n. 1398, aggiornato con le modifiche apportate dalla
mate the influence of accident factors on accident L. 28 aprile 2015, n. 58, dalla L. 22 maggio 2015, n. 68
severity. Accid. Anal. Prev. 34, 729–741. e dalla L. 27 maggio 2015, n. 69.
Baluni, P. & Raiwani, Y.P. 2014. Vehicular accident anal- Shankar, V., Mannering, F. & Barfield, W. 1996. Statis-
ysis using neural networks. Int. J. of Emerg. Tech. and tical analysis of accident severity on rural freeways.
Adv. Engin. 4 (9), 161–164. Accid. Anal. Prev. 28 (3), 391–401.
Chawla, N.V. 2010. Data mining for imbalanced datasets: Theofilatos, A. & Yannis, G. 2014. A review of the effect
an overview. Data Mining and Knowledge Discovery of traffic and weather characteristics on road safety.
Handbook, Springer US, 875–886. Accid. Anal. Prev. 72, 244–256.
Chong C-Y. & Kumar, S.P. 2003. Sensor networks: evolu- Wang, C., Quddus, M.A. & Ison, S.G. 2013. The effect
tion, opportunities, and challenges. Proceedings of the of traffic and road characteristics on road safety: a
IEEE 91(8), 1247–1256. review and future research direction. Saf. Sci. 57,
Chong, M.M., Abraham, A. & Paprzycki, M. 2004. Traf- 264–275.
fic Accident analysis using decision trees and neural net- Xu, C., Tarko, A.P., Wang, W. & Liu, P. 2013. Predict-
work. arXiv preprint cs/0405050. ing crash likelihood and severity on freeways with
Christoforou, Z., Cohen, S. & Karlaftis, M. 2010. Vehi- real-time loop detector data. Accid. Anal. Prev. 57,
cle occupant injury severity on highways: an empirical 30–39.
investigation. Accid. Anal. Prev. 42, 1606–1620. Yu, R. & Abdel-Aty, M. 2013. Using hierarchical Baye-
Cybenko, G. 1989. Approximation by superpositions of sian binary probit models to analyze crash injury
sigmoidal functions. Math. Control Signals Syst. 2(4), severity on high speed facilities with real-time traffic
303–314. data. Accid. Anal. Prev. 62, 161–167.
Delen, D., Sharda, R. & Bessonov, M. 2006. Identify-
ing significant predictors of injury severity in traffic

739

TIS2017_Book.indb 739 2/8/2017 8:09:25 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The influence of pedestrian crossings features on driving

behavior and road safety

A. Bichicchi, F. Mazzotta, C. Lantieri, V. Vignali, A. Simone & G. Dondi

Department of Civil, Chemical, Environmental and Material Engineering, University of Bologna, Italy

M. Costa
Department of Psychology, University of Bologna, Italy

ABSTRACT: Traffic safety depends upon the integrated and complex relationship between various
components: driver psychology, traffic, vehicles and road infrastructures. According to statistics, the
aspect that seems to be the most important, as it is responsible for most accidents, is the behavioral
component and therefore the psychology of the driver of the vehicle. When considering the interaction
between drivers and road infrastructure, pedestrian crossings play a significant role because theirs con-
figuration, in terms of traffic signs and road signage, implicates different behavior responses. For this rea-
son, it is important to identify which situations would be safer for vulnerable users. In the present study,
a sample of 24 drivers performed a trial route of 59.4 km to investigate the totality of non-signalized
pedestrian crossings located along the roadways. GPS kinematic parameters and mobile eye tracking data
were registered to evaluate their approaching behavior. Pedestrian crossings were classified according to
the configuration of road elements present in the each intersection. For each single crossing were evalu-
ated the approach speed, the distance of pedestrian crossings perception and the fixation time to specific
elements. The frame-by-frame video analysis shows that at least one element is visualized by 84% of driv-
ers and the presence of certain elements and their configuration determine significant differences in sight
distances and speed adaptation. Overall, driver behavior is significantly influenced by crossings complex-
ity, determining remarkable effects on the incidence of fatal pedestrian accidents.

1 INTRODUCTION Even premised that the risk of injury caused by

running over is especially high for the elderly popu-
1.1 Vulnerable road users lation, accident causes are multiple, ranging from
infrastructures critical issues to improper driver.
The pedestrian is one of the most vulnerable user
More specifically, at crosswalks pedestrians believe
involved in road accidents. In Italy, statistics con-
to be safe and they reduce their level of attention,
firm that the number of deaths among pedestrians
compromising the overall safety of the intersection.
is showing an overall decrease but it is likely that
Besides, it is clear that safety at the intersections
this is only due to the overall decline of fatal motor
depends on many factors and one most important
vehicle crash (−9.8%). Most of the crashes involv-
is represented by the behavior related to drivers.
ing pedestrians occur on rural roads (4.63 deaths
per 100 accidents), with a reduction on motorways
and urban roads (3.46 and 1.04, respectively). As
1.2 Drivers’ behavior
regards the roadway carriage, the highest mortal-
ity rate is recorded in single carriageway two-lane While approaching pedestrian crossings, drivers
roads (5.03 deaths per 100 accidents) (ISTAT 2014, visually incorporate multiple information that may
2015). The most interesting data is, however, that affect the perception of vulnerable users present
about ¼ of fatal pedestrian accidents occurs at within the area of operation. It is hence important
intersections with crosswalks. identifying how boundary conditions, in terms of
Research on this topic is therefore more than road signs, road markings, geometrical elements
actual. Recent interest on vulnerable users’ pro- of the road, can capture the drivers’ attention and
tection is also encouraged by the European Com- influence their driving behavior (Mazzotta et al.
mission that proposes to maintain the target of 2014). A recent naturalistic driving study by Habi-
halving the overall number of road deaths in the bovic focused on pedestrian accidents captured by
EU decade 2011–2020. onboard cameras and found that, at intersections,

741

TIS2017_Book.indb 741 2/8/2017 8:09:27 PM

 drivers failed to recognize the presence of the accidents and, in addition, a key role in those acci-
pedestrian conflict due to visual obstructions and/ dents that could be completely avoided (Anderson
or because their attention was allocated towards et al. 1997).
something other than the pedestrian (Habibovic For these reasons, roadway capability to attract
et al. 2013). For this reason, road design must take or distract drivers may be thus considered as the
into account not only traffic factors, but also the direct cause of speed variations, a sort of an action-
gaze behavior of road drivers. From one side, the reaction process that should be deeply investigated
environment at the intersection should be designed to find measures for improve the safety level of
avoiding any source of potential distraction; on the intersections.
other hand, road elements and their configuration This study aimed to quantify visual (with eye-
have the important role to alert drivers about the tracking system), and speed behavior of drivers as
presence of the pedestrian crossing, raising their a function of constituent elements of pedestrian
level of attention. The mechanism in which the crossing with different design.
user recognize and decide heavily affects the risk
of accident and, for this reason, influencing the
perception of the road users can induce precise 2 METHODOLOGY
behaviors and make driving role safer (Bucchi et al.
2012). When action needs to be taken, the driver Field tests have been carried out on C-type roads
seems to degrade the allocation of visual attention according to the Italian Road Code (1992), which
when it is overly general, but optimizes it when it is are constituted by a two-lane carriageway with a var-
specific to a manoeuvre (Eyraud et al. 2014). iable section width. Routes were chosen according
The recent interest of road engineers concerns on their high rate of car-pedestrian collisions. The
the study of which elements of the road environ- first was on a segment of the roadway SP610, known
ment actively attract the attention of users. as the “Selice-Montanara”, which connects Emilia
The users’ attention during the driving role and Toscana regions in the North of Italy. Similarly,
focuses mainly on some specific areas of the inter- the second segment was located on roadway SP26,
section, ignoring other elements. In particular, atten- known as the “Valle del Lavino”, which develops
tion is increased in more attention to more complex from Bologna city area to Monte San Pietro. Overall,
intersections, compared to those with less elements the experimental route was 59.4 km long, and inter-
(Werneke & Vollrath 2012). More specifically, hori- sected 15 small urban centers, in hilly terrain, with
zontal markings are more looked at than the verti- speed limit of 50 km/h. Because of the conspicuous
cal signals (Costa et al. 2014): the latter, in fact, have presence of urban areas, a high number of pedestrian
an important role, because of its advantage to be crossings was encountered (89) along the roadway.
placed at a certain height but force drivers to turn The sample consists of twenty-four drivers,
their gaze at the right or left side with exception of 15 men and 9 women (mean age: 30 ± 10 years)
central overhead signs. Horizontal markings are with normal vision, and without glasses or con-
thus more visible because located in the direction tact lenses that interfered with the eye-movements
of the user gaze, allowing drivers to keep the gaze recording system. Their mean driving experience
straight ahead. Nevertheless, literature lacks of tan- was 12 ± 10 years and the mean kilometers driven
gible evidence of horizontal markings effectiveness: per year were 9858 ± 10212. None of the drivers had
a research by Herms that regarded pedestrian acci- previous experience of the route and they were not
dents at 400 intersection, demonstrates that approxi- informed about the study true aims. Participants
mately twice as many pedestrian accidents occur were Psychology and Engineering undergraduate
in marked crosswalks as in unmarked crosswalks students and nobody of them was paid.
(Herms 1972).
A similar research assessed the design features
2.1 Devices
that affected signalized pedestrians crossings. The
developed methodology suggested design modifi- Participants drove a BMW 1 Series car equipped
cations and operational features to improve both with a Video VBOX PRO data logger. Two cam-
the safety level and the level of service of the inter- eras and a GPS antenna, positioned on the top of
section (Steinman & Hines 2004). the car, and connected to the Video VBOX system,
Concerning the travel speed in approach to inter- recorded the scene seen by the driver and their
sections, other researchers found that users usually video outputs were synchronized with data about
approach intersections with excess of speeding acceleration, speed and GPS coordinates.
and an aggressive driving style (Papaiannou 2007). Visual fixations were assessed using a mobile eye
Anderson also showed that small changes in the tracker, a device that has been extensively used in
travelling speeds of vehicles could have a remark- road safety research (Land & Tatler 2009; Under-
able effect on the incidence of fatal pedestrian wood 1998; Van Gompel et al. 2007). The system

742

TIS2017_Book.indb 742 2/8/2017 8:09:27 PM

 used in this study was an ASL Mobile Eye-XG
equipment constituted by two digital high-reso-
lution cameras attached to lightweight eyeglasses.
One camera recorded the scene image and the
other the participant’s right eye, creating a video
for each participant, where a red cross showed the
eye-fixations. This cursor, superimposed to the
video of the scene, allowed researchers to detect
what specific point of the scene the participant was
looking at. The ASL Mobile Eye-XG equipment
has a declared precision of 0.5°–1°, and a record-
ing sample rate of 30 Hz. Its reliability in target-
ing looking behavior has been recently tested at a
distance up to 150 meters from target point (Costa
et al. 2014), by correlating the theoretical angle and
the Interfix degree parameter obtained from the
ASL Results software (r = 0.993 p <0.001).
The eye tracking equipment along with a com-
puter and the VBOX PRO equipment were kept
on the back seat and were monitored by one of the
experimenters, who was instructed not to interact
to the driver except for instructions or request for
assistance.
Before the test, a calibration procedure was car-
ried out for each participant. This procedure took Figures 1a, 1b, 1c. Example of elements: gateway and
place in a parking lot with the car being stationary zebra markings (a), vertical sign and zebra markings (b),
and involved participants to look at least 15 visual central island and zebra markings (c).
points spread across the whole scene. Each driver
was given a two kilometers trial run to get used to
the car before starting along the experimental route. 3 RESULTS

A frame-by-frame analysis allowed noticing the

2.2 Pedestrian crossings
visual fixations to the above-mentioned elements,
As shown in Table 1, pedestrian crossings have been which were considered “looked at” when the red
divided into 4 classes, depending on the configura- cursor was superimposed on the specific element
tion of four different elements: “zebra” marking, for at least one frame (33.33 ms). In addition, in
vertical signs, overhead gateway and central island each case the approaching speed and the distance
(Figures 1a, 1b, 1c). For the reliability of this study, at which the elements were looked at were con-
only pedestrian crossings within urban areas, and sidered and entered in the analyses. A pedestrian
visible at more than 100 m were considered. crossing was considered “perceived” when at least
Considering the whole sample, 742 unobstructed one of the mentioned elements was seen.
intersections (cases with the presence of pedestri- Based on this criteria, in 114 cases (15% of the
ans were excluded) were considered. According to total investigated) drivers did not looked at any
the above-mentioned classification, their distribu- element and did not consider the presence of the
tion was: 29% of cases occurred on C1 configura- pedestrian crossing. The correlation between vis-
tion (214 on the total), 27% on C2 (203), 22% on ual behavior and drivers characteristics denoted
C3 (163), and 22% on C4 (162). that pedestrian crossing perception was influenced
by age (Pearson correlation r = 0.70) but not by
Table 1. Typological configurations of the pedestrian gender (r = 0.021, p = 0,572).
crossings considered in the study.

N° of Zebra Vertical Central 3.1 Visual behaviour

crossings markings sign Gateway Island A pedestrian crossing can be reasonably consid-
ered “well designed” when an approaching driver
C1 24 X X
detect it sufficiently in advance, in order to assess
C2 24 X X
the presence of pedestrians and eventually decrease
C3 19 X X X
C4 22 X X X
speed in safe conditions. It is also assumed that
drivers “safely” perceive a pedestrian crossing

743

TIS2017_Book.indb 743 2/8/2017 8:09:27 PM

 when the gaze at the first element occurs at an markings were the first element that was looked at
equal or greater distance than 50 m. This distance in C1, C2 and C4. When present, in fact, horizon-
is required to stop a vehicle travelling at an aver- tal signs are seen before other elements (vertical sign
age speed of 50 km/h with an average longitudinal and gateway); when central island was present, zebra
slope of 4% (Ministerial Decree n. 6792, 2001). markings decrease their possibility to be seen first.
A detailed data analysis permitted to determine For each configuration it was computed the visu-
the percentage of perceptions by each configura- alization of specific elements and then it has been
tion and, in particular, if it occurs more of less related to the approaching distances. Gaze duration
than 50 m from the pedestrian crossing (Table 2). was computed multiplying the number of frames
Thus, the percentage of not perceived configura- for the fixations to specific elements to the duration
tions (none element gazed) was determined. of each frame (33.33 ms). C4 resulted the configu-
Central islands and overhead signs were the ration whose elements were gazed for longer time
most looked elements. C4 was the pedestrian cross- (M = 724 ms), followed by C1 and C3 (M = 296.48 ms
ing configuration more looked at (in more than and M = 296.54 ms, respectively). C2 was the config-
90% of crossings) whereas the C1 configuration, uration less looked at (M = 281.91 ms). Average of
composed by zebra markings and vertical signs total fixation times to the four pedestrian crossing
only, was the less perceived (almost one crossing configurations are shown in Figure 3.
on four was ignored). The time spent gazing elements increased con-
Regarding the property of pedestrian crossings to siderably when the driver was closer to the intersec-
be perceived at a greater distance, it resulted that the tion, with the exception of C3 configuration where
C2 and C3 configurations had the highest percent- the peak of time average was reached between 100
ages, due to the presence of the overhead sign. These and 50 m in advance (see Table 3). Is therefore pos-
configurations were respectively perceived more sible to suppose that the simultaneous presence of
than 50 m in advance in 22% and 27% of the totality gateway and vertical sign anticipates the caption of
of cases. It is also possible to suppose that the central the attention level. Similarly, considering each spe-
island, even if ensures the crossing an high probabil- cific element, it resulted that zebra marking was the
ity to be perceived, in more than the 80% of cases is element most looked at (Figure 4) in every configu-
seen very close to the pedestrian crossing (<50 m). ration, followed by the central island when present.
At the same time, considerations emerge from Considering the relationship between specific
the investigation of first element seen for each con- elements and sight distance:
figuration. From Figure 2 it is evident that zebra
– the quantity of frames focusing on zebra mark-
ings, which occur in all configurations, signifi-
Table 2. Perception of the 4 pedestrian crossing con-
cantly increased as the driver approached the
figurations as a function of distance.
intersection, reaching a peak in the last 50 m
Pedestrian crossing Not before the pedestrian crossing (4.96 frames cor-
configuration >50 m <50 m perceived responding to 165.31 ms);
– the other elements (vertical sign, overhead sign
C1 17.29% 58.88% 23.83% and central island) got the attention of drivers
C2 21.67% 62.07% 16.26% before (from 150 m to 50 m before the pedes-
C3 26.99% 63.19% 9.82% trian crossing) and, consequently, the number of
C4 11.11% 80.25% 8.64% frames slightly decrease in the last 50 meters;

Figure 2. First element that was looked at in the differ- Figure 3. Mean total fixation duration for the four
ent pedestrian crossing configurations. pedestrian crossing configurations.

744

TIS2017_Book.indb 744 2/8/2017 8:09:27 PM

 Table 3. Average number of fixation frames for the different element of the four pedestrian crossing configurations.

Zebra Markings Gateway Vertical Sign Central Island

150–100 100–50 50–0 150–100 100–50 50–0 150–100 100–50 50–0 150–100 100–50 50–0

C1 4.8 67.9 165.2 17.4 27.5 13.5

C2 8.5 68.7 142.5 23.1 27.5 11.3
C3 11.4 77.5 92.8 18.6 14.7 7.3 17.5 39.0 17.3
C4 5.5 114.8 242.7 38.0 38.8 9.4 17.4 134.9 122.4

Figure 4. Average total fixation durations to the different elements of the four pedestrian crossing configurations.

(Bella & Silvestri 2015). Speed data were recorded

with the Video V-Box equipment. In particular,
were considered the speeds detected in correspond-
ence of 150, 100 and 50 m before the pedestrian
crossing. Mean speeds are shown in Figure 5.
The overall peak of average speed occurs in the
C3 configuration 50 m before the pedestrian cross-
ing (56 km/h) while at a distance of 50 m the aver-
age speed is always higher than 50 km/h. Instead,
V150 and V0 (average speed at 150 m and in cor-
Figure 5. Average speed at 150 m, 100 m, and 50 m respondence of the crossing) are very similar for
before the pedestrian crossing. every configuration; interesting data are the signifi-
cant speed decreasing in C2 and C4 (−1.26 km/h
– overhead signs are visualized before the other ele- and −2.23 km/h).
ments, confirming their key-role in the improve-
ment of pedestrian crossings in terms of safety
(Ariën et al. 2013; Lantieri et al. 2014); 4 DISCUSSION AND CONCLUSIONS
– central islands, were mainly fixated in the last
100 m. The described experimental field permitted to quan-
tify how the configurations of road furniture influ-
ence drivers’ behaviour. Horizontal, vertical signs
3.2 Driving behaviour
and geometric elements as central islands influence,
Previous studies in literature suggest to analyze indeed, the pedestrian crossing sight distance, the
the drivers’ speed approaching to the intersection, observation time, and speed modulation. In par-
starting from 150 m before the zebra markings ticular, assuming that pedestrian safety and, more

745

TIS2017_Book.indb 745 2/8/2017 8:09:28 PM

 in general road safety, are assured when a crossing Bella, F., & Silvestri, M. (2015). Effects of safety
is perceived largely in advance, it has been dem- measures on driver’s speed behavior at pedestrian
onstrated that the presence of gateways is a safety crossings. Accident Analysis and Prevention, 83,
indicator, independently of boundary conditions. 111–124.
Bucchi, A., Sangiorgi, C., & Vignali, V. (2012). Traffic
Similarly, overhead signs resulted to be the element psychology and driver behaviour, in Procedia—Social
visualized first with respect to other elements. and Behavioral Sciences, 53, 973–980.
In addition, the central island resulted the ele- Costa, M., Simone, A., Vignali, V., Lantieri, C., Bucchi, A.
ment that induces the highest speed reduction, & Dondi, G. (2014). Looking behavior for vertical road
probably a cause of the reduction in lane width. signs. Transportation Research Part F, 23, 147–155.
The increasing observation time resulted slightly Decreto Legislativo 30 aprile 1992, n. 285 Nuovo codice
related to speed reductions. For this reason, con- della strada (in Italian).
figurations that are more complex do not distract Decreto Ministeriale n. 6792 del 5 novembre 2001: Norme
drivers but induce them to increase their attention funzionali e geometriche per la costruzione delle strade.
G.U. n. 3, 04/01/2002 (in Italian).
and to be more careful (decrease of average speed) EU Commission (2010). Road Safety Programme
confirming previous studies (Werneke & Vollrath 2011–2020.
2012). On the contrary, almost 25% of the totally Eyraud, R., Zibetti E. & Baccino, T. (2014). Alloca-
ignored pedestrian crossings, had a simple config- tion of visual attention while driving with simulated.
uration (zebra markings and vertical sign only). Transportation Research Part F, 32, 46–55.
The aim of this study have been suitably reached, Habibovic A., Tivesten E., Nobuyuki U., Bärgman J. &
in particular: Aust M.L. (2013). Driver behavior in car-to-pedestrian
incidents: An application of the Driving Reliability
1. Have been provided useful insights for a better and Error Analysis Method (DREAM). Accident
comprehension of the elements attracting driv- Analysis and Prevention, 50, 554–565.
ers’ attention while approaching the crossings; Herms, B.F. (1972). Pedestrian crosswalk study: acci-
2. Regarding drivers’ speed behaviour approach- dents in painted and unpainted crosswalks. Highway
ing crossings, a comparative evaluation of con- Research Record, 406, 1–13.
figurations’ effectiveness have been performed. Land, M.F. & Tatler, B.W. (2009). Looking and acting:
Vision and eye movements in natural behaviour. Oxford;
The findings of this research offer interest- Oxford University Press.
ing application for the design and adaptation of Lantieri, C., Lamperti, R., Simone, A., Costa, M., Vig-
pedestrian crossings, in order to increase the level nali, V., Sangiorgi, C. & Dondi, G. (2015). Gateway
of safety offered to users. High-rate accidents design assessment in the transition from high to low
speed areas. Transportation Research Part F: Traffic
roadways, once upgraded and adjusted following Psychology and Behaviour, 34, 41–53.
those criteria, should be studied in order to realize Mazzotta, F., Irali, F. & Simone, A. (2014). La valutazi-
adequate before-after analyses. one della sicurezza delle utenze deboli negli attraver-
Continuing this research, the sample for experi- samenti pedonali non semaforizzati tramite Mobile
mental trials could be incremented and clustered into Eye Detector. In: L’utente Debole Nelle Intersezioni
different users’ categories; moreover, the evaluation Stradali, Parma, 27–28 Marzo 2014, 181–190, EGAF
of further configurations e.g. provided with signal- EDIZIONI.
ing lamps and back-colored markings could be con- Mazzotta, F., Vignali, V. & Irali, F. (2014) Evaluation of
sidered for future trials. Future research could also the readability of road signs and roadside elements
using Mobile Eye tracking device. InBo, 7, 253–260.
focus on the pedestrian crossings accident analysis, Papaioannou, P. (2007). Driver behaviour, dilemma
in order to search for a direct correlation between zone and safety effects at urban signalised intersec-
elements or configurations and pedestrian injuries. tions in Greece. Accident Analysis and Prevention, 39,
147–158.
Steinman, N. & Hines, D.K. (2004). Methodology to
REFERENCES assess design features for pedestrian and bicyclist
crossings at signalized intersections. Transportation
ACI-ISTAT, 2015. Rapporto ACI—ISTAT sugli inci- Research Record, 1878, 42–50.
denti stradali, anno 2014 (in Italian). Underwood, G. (Ed.) (1998). Eye guidance in reading and
Anderson, R.W.G., McLean, A.J., Farmer, M.J., Lee, B.H. scene perception. Amsterdam: Elsevier.
& Brooks, C.G., (1997) Vehicle travel speeds and the Van Gompel, R.P.G., Fischer, M.H., Murray, W.S., &
incidence of fatal pedestrian crashes, Accident Analy- Hill, R.L. (Eds.) (2007). Eye movements: A window on
sis and Prevention, 29, 5, 667–674. mind and brain. Amsterdam: Elsevier.
Ariën, C., Jongen, E.M.M., Brijs, K., Brijs, T., Daniels, Werneke, J. & Vollrath, M. (2012). What does the driver
S., & Wets, G. (2013). A simulator study on the impact look at? The influence of intersection characteristics
of traffic calming measures in urban areas on driving on attention allocation and driving behavior. Accident
behavior and workload. Accident Analysis and Preven- Analysis and Prevention, 45, 610–619.
tion, 61, 43–53.

746

TIS2017_Book.indb 746 2/8/2017 8:09:28 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The relationship between driving risk levels and drivers’ personality

traits, physical and emotional conditions

L. Eboli, G. Mazzulla & G. Pungillo

University of Calabria, Rende, Italy

ABSTRACT: This paper presents a study aimed to investigate the relationship between driver person-
ality traits and the level of driving risk taken by driver during a trip, and also the relationship between
driving risk levels and driver’s physical and emotional conditions while driving. To this end, a survey
was conducted to gain an objective measure of the risk level, which gives an indication of the driving
conditions (safe or unsafe), and subjective judgements of the drivers about their personal characteristics.
Objective measures were obtained from kinematic parameters (instantaneous speed, longitudinal and
lateral accelerations) recorded along each trajectory followed by the driver. For this aim, smartphones
equipped with GPS and accelerometer were used. On the other hand, drivers were asked to complete a
questionnaire about their personality traits, and physical and emotional conditions while driving, which
can influence their driving style. Each driver covers the same path several times in different days, in order
to capture different physical and emotional conditions and driving styles. The sample of drivers belongs to
a population of car drivers between 25 and 50 years old who every day drive by their private car for reach-
ing their destination of work or study. Overall, drivers made about 170 tests covering about 1,100 km. We
proposed a correlation analysis between the drivers’ characteristics and the risk level obtained by using
acceleration and speed data. As an example, we found that characteristics such as patience and meticu-
lousness positively influence the level of safety.

1 INTRODUCTION precision that we can obtain at low cost. In fact,

this kind of measures were used also to study the
The phenomenon concerning road accidents is bus comfort on board (Eboli et al., 2017, in press)
ever more relevant; every year about five mil- and to study the design consistency of a road seg-
lions of people in the world dies a violent death ments by drawing the speed profiles (Eboli et al.,
and a quarter of these are traffic accident victims. 2015). On the other hand, we conducted a survey
Driver’s behavior is the primary cause of traffic for collecting the judgments of drivers about their
accidents. In fact, road safety problems are more personal characteristics. Each participant involved
frequently related to certain drivers’ attitudes and in the survey was asked to complete, for each tra-
their physical and emotional conditions when driv- jectory, a questionnaire including some questions
ing; for this reason, it is necessary to study carefully according to which he/she has to express a rating
the human factor in road system. The aim of this about some personality traits, and physical and
paper is to investigate the relationship between the emotional conditions while driving.
level of driving risk taken by driver during a trip The database is interesting because all the driv-
and his/her personality traits. We also analyze the ers involved in the survey registered the same path
relationship between driving risk levels and driver’s run in different days, and complete the question-
physical and emotional conditions while driving. In naire for each path. This permits to capture the
order to collect this kind of information, a survey different conditions of the drivers and observe the
was conducted to gain an objective measure of the possible changes of their driving style as a function
driving style, and drivers’ opinions about their per- of their physical and emotional states.
sonal characteristics. The objective measures are In this paper we propose a correlation analysis
represented by the kinematic parameters (instan- between the risk level calculated based on the kine-
taneous speed, longitudinal and lateral accelera- matic parameters registered through smartphones,
tions) recorded along the trajectories followed by and the judgements expressed by the users through
the drivers through smartphones equipped with the questionnaire, about some personality traits,
GPS and accelerometer. The acquisition of this physical and emotional conditions which can influ-
type of data is increasingly spreading thanks to the ence the driving style.

747

TIS2017_Book.indb 747 2/8/2017 8:09:30 PM

 In the following we report a brief literature anxiety per se and its effects on driving has not as
review of the studies dealing with this kind of yet garnered much attention in the literature. The
issues. Then, we propose the methodology, by study proposed by Dula et al. (2010) found that
describing the sample, the objective measure of higher levels of general anxiety were related to a
the risk level, and the collection of the data con- wide variety of dangerous driving behaviours.
cerning drivers’ characteristics. After, we propose The research of Lu et al. (2013) explores how
the results of the correlation analysis and the main and why anger and fear influence driving risk
conclusions about the work. perception.
There are also studies where kinematic param-
eters were registered in addition to the information
2 LITERATURE REVIEW collected through the questionnaire addressed to
the drivers. As an example, Harder et al. (2008)
The major part of the literature studies dealing conducted a study to better understand the psy-
with the relationship between psychophysical fac- chological and roadway correlates of aggressive
tors and driving style comes from psychologists, driving. Survey data was used to investigate the
pedagogues, and so on. They have generally ana- relationship between personality, emotional, and
lysed personal traits, driving behaviour, and habit. behavioural variables and self-reported driving
Most studies on the causes of road crashes focus behaviour. Participants were asked to give infor-
on the relationship between specific traits and abili- mation about their personality characteristics
ties of drivers, and crash risk. Petridou and Mous- including hostility, anger expression, physical
taki (2000) proposed a review where they tried to and verbal aggressiveness, competitiveness, and
delineate behavioural factors that collectively rep- empathy; their emotional state including anger-
resent the principal cause of three out of five RTCs provoking experiences, anger states, and negative
(Road Traffic Crashes) and contribute to the cau- thought patterns; and their behavioural tendencies
sation of most of the remaining. The personality with regard to driving. In addition, a part of the
factors most often selected for that purpose include participants drove on a section of a simulated four-
the Big Five personality dimensions: openness, con- lane freeway.
scientiousness, extraversion, agreeableness, neurot- In a simulator study proposed by Roidl et al.
icism. The studies proposed by White and Dahlen (2014), some participants took part in certain traf-
(2001), Dahlen et al. (2005), and Dahlen and White fic situations which each elicited a different emo-
(2006) investigated the utility of combining the Big tion. Each situation had critical elements (e.g. slow
Five personality factors, which have been studied car, obstacle on the street) based on combinations
independently, in the prediction of self-reported of the appraisal factors. Driving parameters such
driving anger expression and the frequency of as velocity, acceleration, and speeding, together
aggressive and risky driving behaviours. with the experienced emotions, were recorded.
Much attention has been paid to older driv- Results indicate that anger leads to stronger accel-
ers in research on driving behaviour and safety. eration and higher speeds. Anxiety and contempt
The study by Adrian et al. (2011), in fact, inves- yielded similar but weaker effects.
tigated how executive functions and personality From this brief literature review it emerges that
traits are related with driving performance among the major part of the studies analysed personality
older drivers. Significant correlations were found traits that do not vary over the time. On the con-
between poor driving performances and low scores trary, our analysis focuses on personal conditions
on tests assessing shifting and updating functions. changing from day to day, that is the physical and
In addition, extraversion had a negative relation emotional conditions of the drivers at the time of
with driving performance and made the only con- the driving.
tribution, among the psychological factors, to the
prediction of driving performance.
Arnau-Sabatés et al. (2012) proposed a study to
3 METHODOLOGY
analyze the relationship between emotional abili-
ties, such as adaptability, stress management or
3.1 Sample
affectivity, and the influence of this relationship
on self reported drivers’ risky attitudes. The risky The drivers supporting this research belongs to a
driving attitudes and emotional abilities of driving population of car drivers who every day drive by
instructors were measured. The results demon- their private car for reaching their destination of
strate that risky attitudes correlate negatively with work or study, by running always the same path.
emotional abilities. This definition of the population is due to our
Driving anxiety that has developed following desire to investigate on the influence of psycho-
crashes has been studied relatively frequently, but physical factors on driving style; so, it is important

748

TIS2017_Book.indb 748 2/8/2017 8:09:30 PM

 to collect data concerning the same path run in dif- from smartphones provide accurate real-time traf-
ferent days in order to capture different conditions fic data for safety analysis.
of the drivers, and observe the possible changes of
their driving style as a function of their psycho-
3.2 Objective measure of driving risk level
physical state.
The drivers involved in the survey are five The level of risk is identified on the basis of the
licensed drivers who are between 25 and 50 years percentage of external points at borderline of a
old. The drivers can be grouped into two categories. safety domain. As a safety domain was chosen
The first group is composed of three experts in the the one identified by Eboli et al. (2016) by means
sector of transportation engineering, two females of a methodology that describes the relationship
and one male, who know very well the objectives of between lateral, longitudinal accelerations and
the research and the proposed methodology. How- speeds, and represents a tool for classifying driv-
ever, they become simple drivers who travel with ing behaviour of the car drivers. Each point, in the
their own cars from home to their place of work. reported diagram, represents the kinematic condi-
The choice to be components of the sample arises tions of the vehicle in a certain instant, and it is
from the desire to test the survey and personally representative of a safe or an unsafe driving behav-
experience the planned work. Two experts in driv- iour. Specifically, if the point is inside the safety
ing compose the second group of drivers; they are domain the driver is in safety conditions, else the
two police officers working for a long time. driving behaviour is unsafe.
Each driver registered through a smartphone As an example, Figure 1 shows a survey car-
the information of thirty paths approximatively. ried out by a driver having an unsafe behaviour.
Participants drove under good weather daytime It shows that there is a high percentage of external
conditions with light traffic. The differences in points at borderline of the safety domain; the per-
driver age, gender and driving experience were centage is 9.97.
treated as random effects, and were not considered
in this research.
3.3 Collection of data concerning drivers’
The survey interested some parts of the Italian
personal and psychophysical characteristics
National road n.107 (S.S. 107) which is a rural two-
lane road 138 km long connecting the Tyrrhenian Each participant involved in the survey was asked
with the Ionian Sea coast of the Calabria region, in to complete a questionnaire aiming to collect judg-
the Southern Italy. The high flow, combined with ments concerning personal traits and psychophysi-
heavy rains and frequent snowfalls, has gradually cal conditions that can influence driving style and
increased the danger of the road, especially in risk level (Eboli et al., 2016, in press). The final ver-
some tortuous and difficult parts. In fact, the road sion of the questionnaire has been obtained after
was in the list of the 10 most dangerous roads of a deep and wide research on character aspects that
Italy in 2007, occupying the 7th place (ACI, 2007). could influence people driving style. We designed
Some measurements were also carried out on one and realized a preliminary survey with the aim to
part of the Italian National road n.18 (S.S.18) that investigate on various aspects and to verify the
is an important arterial road connecting Campania suitability for our objectives (Salandria, 2013). The
with Calabria on the Tyrrhenian Sea coast. analysis of the results of the preliminary survey
The analysed road segments have a length of helped us in the selection of the definitive list of
about 6 km. Each segment has a roadway with aspects for the final survey, and in the observa-
two lanes, one for each direction, having a width tion of different categories of aspects. In fact, by
between 3.25 meters and 3.5 meters. observing some results of the preliminary survey
Overall, drivers involved in the survey make we succeeded in identifying three different cat-
about 170 tests covering about 1,100 km. egories: personality traits, physical conditions, and
With the aim of record the kinematic param- emotional conditions.
eters along each trajectory followed by the driver, Specifically, drivers made a self-evaluation by
drivers ran repeatedly the analysed road segments. expressing a level according to a numerical rat-
Smartphones equipped with GPS and accelerome- ing scale about five pairs of adjectives represent-
ter were used; by means of a free app, the kinematic ing general peculiarities of their character. We
values were recorded with a frequency of 1 hertz, adopted a scale on five levels, from −2 to +2. The
together with the instantaneous vehicle position “−2” and “+2” levels are labelled “extremely”,
(latitude and longitude). Today, these devices are while “−1” and “+1” are labelled “quite”. After a
widely used because they allow the investigation deep research and analysis of the attitudes, behav-
on kinematic parameters with a low cost and at iours, characters, emotions of people, we selected
good levels of reliability. As reported in Eboli et al. five pairs of personality traits that we retain possi-
(2016), it is proved that probe estimates obtained ble determinants of the driving style of car drivers.

749

TIS2017_Book.indb 749 2/8/2017 8:09:30 PM

 Figure 1. Graphical representation in the (V, ā) plane of the points inside and outside the safety domain, referred to
a generic path.

The five pairs of adjectives according to the layout We decided to analyse also five emotional con-
reported in the questionnaire are the following: ditions by considering the following five pairs of
adjectives:
− lazy/dynamic
− irresolute/resolute − gloomy/happy
− impatient/patient − worried/carefree
− impulsive/reflective − nervous/calm
− superficial/meticulous. − bored/interested
− angry/serene.
The core of the questionnaire is the part aiming to
collect information about the psychophysical condi- Also in this case the emotional conditions refer
tions of the driver before starting the analysed path. to the time before starting to drive the car.
Specifically, the conditions were divided in two
categories: physical and emotional. Also in this
case drivers express a level from −2 to +2 about a
4 CORRELATION ANALYSIS BETWEEN
series of adjectives representing physical and emo-
OBJECTIVE RISK LEVEL
tional conditions.
AND DRIVERS’ CHARACTERISTICS
This part differs from the previous one because in
this case drivers provide for information referred to
We propose a correlation analysis between the
their state in a specific moment, while in the first part
personal and psychophysical characteristics of
they express opinions about general personality traits
the drivers and the risk level, in terms of safe or
that however we retain as affectingtheir driving style.
unsafe condition, obtained by using acceleration
We selected three physical conditions that we con-
and speed data.
sider possible determinants of drivers’ driving style:
The aim of this kind of analysis is to understand
− tired/fresh which are the personal characteristics of the drivers
− sleepy/vigilant mostly affecting drivers’ risk level. In other words,
− sick/healthy we want to determine the personality traits, and the

750

TIS2017_Book.indb 750 2/8/2017 8:09:30 PM

 psychophysical conditions that influence the driv- is the resoluteness and the reflectiveness; the cor-
ing style, or the safe or unsafe drivers’ behavior. relation analysis shows that an irresolute person
We have to make some preliminary remarks has a safer driving, as well as an impulsive person.
before starting the comments on the correlation It seems that to be a resolute and reflective per-
analysis. The major part of the correlation coeffi- son does not help in having a safe behavior, or in
cients show values lower than 0.3, indicating a low decreasing the driving risk level.
correlation. Only for two coefficients, belonging Table 2 shows the correlation coefficients of the
to the group of the personality traits, we found a risk level with the drivers’ physical conditions. We
medium correlation, being the value of the coef- observe lower values than the personality traits,
ficient around 0.5. However, we retain that the due to the large variability of these conditions as
obtained results can be considered for making a regards the personality traits which are fixed over
preliminary analysis, with the intent to expand the the time; however, we can define the physical condi-
survey and improve the work. tions that mostly affect driving risk level and behav-
Table 1 shows the correlation of the risk level ior. More specifically, we can see that there are two
(objective measure) with the drivers’ personality coefficients with a negative value. We can state that
traits. We can observe that the pair of adjectives if the driver is fresh, the risk level decreases, as well
mostly correlated with the objective measure of the as if the driver is vigilant. On the contrary, if the
risk level is “impatient/patient”, while the personality driver is healthy the risk level increases.
trait having the lowest influence on the risk level con- In Table 3 the correlation coefficients of the risk
cerns the laziness/dynamicity of the person. There level with the drivers’ emotional conditions are
are some coefficients showing a negative signs, and reported. As for the physical conditions, because
others with a positive sign of correlation. The posi- of the variability of the data, also in this case the
tive sign indicates that when the value of the subjec- coefficients are quite low. The conditions mostly
tive judgement goes to the positive judgement (and affecting driving risk level are being “gloomy-
therefore towards the adjective on the right), the risk happy” and “nervous-calm”. More specifically, if
level increases, or the behavior becomes more unsafe; driver is gloomy or nervous, he has a safer driving,
vice versa, the negative sign means that if the value of as well as he is bored. It seems that being gloomy or
the subjective judgement goes to the positive judge- nervous is not a problem during the driving, but it
ment driver behavior is safer. More specifically, if the can be an element that induces the driver to main-
driver is patient, the risk level decreases, as well as tain a safer behavior, maybe because the driver is
if the driver is dynamic and also meticulous. On the conscious of his condition and for this reason, he
contrary, the risk level of people who are more reso- tries to drive with more cautiousness.
lute and more reflective increases. The value of the coefficients with a negative sign
These results can be interpreted by saying that are very low, so we cannot provide a very reliable
laziness, superficiality, and impatience are person- interpretation of the results; however, we can say
ality traits that negatively influence the risk level that it seems that being carefree and serene helps
and the safeness of the driving. In other words, to drive with a lower risk level.
an impatient person has a more unsafe driving,
as well as a superficial person. We registered the
Table 2. Correlation of the risk level with the drivers’
highest values of the correlation coefficients just physical conditions.
for these two personal characteristics. We can say
also that a lazy person has an unsafe behavior, Physical condition Correlation coefficient
although the correlation coefficient is lower than
the others. tired/fresh −0.101
On the other hand, two personality traits have sleepy/vigilant −0.085
a very different influence on driving risk level, that sick/healthy 0.061

Table 1. Correlation of the risk level with the drivers’ Table 3. Correlation of the risk level with the drivers’
personality traits. emotional conditions.

Personality trait Correlation coefficient Emotional condition Correlation coefficient

lazy/dynamic −0.095 gloomy/happy 0.140

irresolute/resolute 0.273 worried/carefree −0.060
impatient/patient −0.477 nervous/calm 0.108
impulsive/reflective 0.112 bored/interested 0.046
superficial/meticulous −0.445 angry/serene −0.019

751

TIS2017_Book.indb 751 2/8/2017 8:09:31 PM

 5 CONCLUSIONS driving behavior among a cohort of middle aged driv-
ers. Accident Analysis and Prevention 45: 818–825.
The aim of this paper was to analyze the relation- Dahlen, E.R., &White, R.P. (2006). The Big Five factors,
ship between drivers’ personality traits, physical sensation seeking, and driving anger in the prediction
of unsafe driving. Personality and Individual Differ-
and emotional conditions and the risk level to ences 41: 903–915.
be involved in a traffic accident calculated based Dahlen, E.R., Martin, R.C., Ragan, K., & Kuhlman,
on the kinematic parameters registered through M. (2005). Driving anger, sensation seeking, impul-
smartphones during the driving. Specifically, we siveness, and boredom proneness in the prediction of
proposed a correlation analysis of the drivers’ unsafe driving. Accident Analysis and Prevention 37:
judgements about their personal characteristics 341–348.
(collected through a questionnaire) and an objec- Dula, C.S., Adams, C.L., Miesner, M.T., & Leonard,
tive measure of the risk level obtained from kin- R.L. (2010). Examining relationships between anxiety
ematic parameters. This kind of analysis allows and dangerous driving. Accident Analysis and Preven-
tion 42: 2050–2056.
to determine the drivers’ personal characteristics Eboli, L., Guido, G., Mazzulla, G., & Pungillo, G. (2015,
which more affect driving risk level. published on line). Experimental relationship between
The main and more interesting results suggest operating speeds of successive speeds of successive
that characteristics such as patience and meticu- design elements in two-lane rural highways. Transport.
lousness positively influence the level of safety, (http://dx.doi.org/10.3846/16484142.2015.1110831)
because the level of risk decreases with the increase Eboli, L., Mazzulla, G., & Pungillo, G. (2016). Combin-
of these two human virtues. A finding apparently ing speed and acceleration to define car users’ safe or
unexpected is the influence of the conditions unsafe driving behaviour. Transportation Research
“gloomy-happy” and “nervous-calm”. Specifically, Part C 68: 113–125.
Eboli, L., Mazzulla, G., & Pungillo, G. (2017, in press).
we found that being gloomy or nervous has a posi- Measuring Bus Comfort Levels by using Accelera-
tive effect on the level of risk. A reasonable inter- tion Instantaneous Values. Transportation Research
pretation could be that gloom and nervousness Procedia.
induces the driver to maintain a safer behavior Eboli, L., Mazzulla, G., & Pungillo, G. (2016, in press).
just because the driver is conscious of his condi- The influence of physical and emotional factors on
tion and he tries to drive with more cautiousness in driving style of car drivers: a survey design. Journal of
order to avoid to occur in a traffic accident. Traffic and Transportation Engineering.
As above specified, a future development of the Harder, K.A., Kinney, T.A., & Bloomfield, J.R. (2008).
work is surely the increase of the sample dimen- Psychological and Roadway Correlates of Aggres-
sive Driving. Report CTS 08–30. Center for Human
sion, in order to obtain more solid and reliable Factors Systems Research and Design, University of
results from the analysis of the data. We retain that Minnesota.
the results reached in this paper are however inter- Lu, J., Xie, X., & Zhang, R. (2013). Focusing on apprais-
esting given that the main aim of the work was to als: How and why anger and fear influence driving risk
test a new methodology and survey to combine perception. Journal of Safety Research 45: 65–73.
subjective and objective data concerning driving Petridou, E., & Moustaki, M. (2000). Human factors in
style. the causation of road traffic crashes. European Jour-
nal of Epidemiology, 16, 819–826.
Roidl, E., Frehse, B., & Höger, R. (2014). Emotional
states of drivers and the impact on speed, accelera-
REFERENCES tion and traffic violations-A simulator study. Accident
Analysis and Prevention 70: 282–292.
ACI (2007). Incidenti stradali in Italia, 2007. Italian Salandria, A. (2013). L’influenza dei fattori psico-fisici
Bureau of Statistics. sui comportamenti di guida dei conducenti delle
Adrian, J., Postal, V., Moessingerc, M., Rascleb, N., & autovetture (Thesis). University of Calabria, Italy.
Charlesb, A. (2011). Personality traits and executive White, R. P., & Dahlen, E. R. (2001). The role of per-
functions related to on-road driving performance sonality and emotional factors in the prediction of
among older drivers. Accident Analysis and Prevention crash-related conditions, aggressive driving, and risky
43: 1652–1659. driving behavior. Paper presented at the Mississippi
Arnau-Sabatés, L., Sala-Rocab, J., & Jariot-Garcia, M. Psychological Association Convention, Gulfport,
(2012). Emotional abilities as predictors of risky MS.

752

TIS2017_Book.indb 752 2/8/2017 8:09:31 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The impact of a bus rapid transit line on spatial accessibility

and transport equity: The case of Catania

N. Giuffrida, G. Inturri, S. Caprì, S. Spica & M. Ignaccolo

Department of Civil Engineering and Architecture, University of Catania, Catania, Italy

ABSTRACT: Accessibility is a key issue to address spatial equity when planning for sustainable mobility.
Accessibility indicators can be used to measure the performance of public transport as basic strategy to
cause modal shift from private transport and reduce car dependence and urban sprawl. The purpose of this
paper is to verify if the realization of a set of Bus Rapid Transit (BRT) lines with high level of service can
provide an equitable access of residents to workplaces, when compared with a light improvement of the
commercial speed of conventional bus lines with low level of service but high spatial coverage. To this aim
we use a relative accessibility measure between private and public transport, weighted by socio-economic
data of population. A high spatial resolution spatial analysis is used to capture the relevance of different
stop density for walking access impedance, through a GIS transport modeling software. The methodology
is tested for the city of Catania (Italy).

1 INTRODUCTION its context, the application of the methodology

and the results obtained. Finally, conclusions are
In the last several years, in the field of planning presented.
and management of transport systems, a proc-
ess that can be described as the transition from a
1.1 Accessibility definitions and measures
mobility oriented to an accessibility oriented plan-
ning is taking place. While planning for mobility The concept of accessibility, especially in transport
uses performance indicators mainly based on travel planning, plays an increasingly important role in
time and level of congestion, planning for accessi- the process of policy decisions and it is a useful tool
bility takes into account a broader set of elements, to analyze the degree of access to an area above the
in order to integrate the transport strategies with surrounding territory. Accessibility measures may
urban and territorial ones. In other words, there has be valid social indicators, as they show the level of
been a gradual shift towards theories, approaches difficulty with which various categories of indi-
and support instruments to an integrated plan- viduals can reach the economic opportunities or
ning of land use and transport system that gives social interaction throughout the area.
increasing importance to accessibility. However, drawing up a strict and unambiguous
This work has multiple aims: on one hand it definition of accessibility is a complex task. One of
defines a new method for modeling public trans- the first scholars which considered its importance
port based on a high resolution spatial analysis in the context of spatial planning was Hansen, who
with the use of GIS; on the other hand, it identi- defined accessibility as “the potential of interac-
fies a suitable accessibility index that allows a com- tion opportunities” (Hansen, 1959).
parison with private transport and that includes A recent definition that highlights the mutual
social issues, in order to support an equitable tran- interaction between land use and transport sys-
sit planning. These goals are pursued by modeling tems has been provided by Geurs and van Wee
the impact of two different transport scenarios on (2004). According to the authors, the accessibil-
urban accessibility and testing the model in the ity can be considered as the measure with respect
multimodal transport network of Catania (Italy). to which the use of the territory and of transport
The paper provides at first a literature review systems allow groups of individuals to reach activ-
regarding accessibility and its relation with social ities or locations by a combination of modes of
inclusion; the successive section shows the meth- transport.
odology by focusing on the network model, the From these and other definitions in the litera-
GIS approach and accessibility measures used. ture, four major accessibility components can be
Afterwards the case study is explored considering identified: land use, the transport system, the time

753

TIS2017_Book.indb 753 2/8/2017 8:09:32 PM

 factor and the individual dimension (Geurs & Van The generic measure of cumulative opportuni-
Wee, 2004). ties can be considered a special case where f(Cij)
A classification of accessibility measures depend- is equal to 1 if Cij is less than the predetermined
ing on land use can be done considering the place threshold; it is equal to 0 otherwise.
in question as the origin or destination of the travel. This type of indicators offers the advantage of
We can therefore distinguish the origin accessibility requiring a relatively small amount of data (ease of
or active accessibility, and the destination accessi- processing and calculation), allowing to differenti-
bility or passive accessibility (Cascetta, 2009): ate the areas of study and to derive the accessibil-
ity indices for each of them. They are particularly
− Active accessibility refers to the need to carry
useful for assessing the potential of suburban resi-
out the activities located throughout the area by
dential areas in allowing access to activities such as
a user that is in a particular place and it meas-
shops, schools, workplaces, health care and other
ures the ease with which the user can reach vari-
services.
ous destinations from an origin. It is useful in
locating settlement decisions.
− Passive accessibility refers to the need for the var- 1.2 Accessibility and social inclusion
ious opportunities that are located in a certain
Social exclusion is a condition of poverty com-
area of the territory, to be achieved by the vari-
bined with social marginalization, which causes
ous users scattered throughout the study area.
the separation of certain groups or individuals
In other words, it measures the ease with which
from institutions and community, preventing full
individuals, business and the services of a target
participation in joint activities of the society in
area of the displacements can be reached by the
which they live.
users concerned. It is useful in the location deci-
With regard to the transport field, social exclu-
sions of public services and economic activities.
sion concept is closely linked to the accessibility
Most of the formulations in the literature refer to activities, goods and services. In this case, the
to an urban active accessibility, whose indicator, transport system plays an important role, since
in analytical terms, is generally a function of the limited accessibility to transportation, by public or
number of spatial opportunities and of the general- private transport, prevents a person to reach job
ized transport cost to reach them. In particular, the places, health centers and entertainment venues.
accessibility indices based on gravitational models Public transport plays a pivotal role in worsen-
provide a measure of the continuous type which ing or alleviating social exclusion of vulnerable
weighs the value of the opportunities with respect groups, since it affects their access to basic services
to a spatial impedance function. The impedance as well as employment positions and social rela-
function reflects the effect of decreasing accessibil- tions. It’s also likely that the negative impact of the
ity due to the increase of distance, travel time, or in transport system on the environment, safety and
general of the generalized cost of shipping. public health would fall disproportionately on the
The first application of the gravity model to most disadvantaged groups.
accessibility measures is attributed to Hansen To develop an inclusive transport system is
(1959), which suggested that accessibility across necessary for accessibility, safety and comfort of
regions was directly proportional to the attractive- modes of transport to become the first priorities
ness factors (jobs, shops, sports centers, etc.) and of transport policy. According to Litman (2009)
inversely proportional to the travel time between this is strictly related to the improvement of:
the zones, which represents the cost of moving.
− all phases of the path, including the environ-
The Hansen’s index has the following form:
ment related to the pedestrian access, in a way
n
that also disabled people could reach and use
Ai ∑O
j
j ⋅ f Cij ) (1) public transport services;
− planning of transport infrastructure, consider-
ing the specific needs of vulnerable groups;
where Oj is the number of opportunities in the − safety of public transport, a key issue that affects
zone j and f(Cij) the impedance function among mainly women and elderly people;
zones i and j. A negative exponential impedance is − frequency and punctuality of services;
often used, such as: − dissemination of maps and information guides.
Public transport may be able to reduce this
f Cij ) = e − βCij (2) mobility gap, by the promotion of social inclusion.
In fact, when it’s not accessible by the weakest
With Cij generalized cost of travel among i population groups and it’s unable to break down
and j zone and β is a parameter related to the the barriers that do not allow the participation to
cost, estimated by choosing a destination model. social activities, public transport fails its primary

754

TIS2017_Book.indb 754 2/8/2017 8:09:33 PM

 goal: to give access to employment or educational 2.3 Relative accessibility measure
opportunities, medical care services and entertain-
We define the Relative Accessibility Index as the
ment venues. In summary, public transport should
relationship between the accessibility guaranteed
offer everyone the ability to move and therefore it’s
by public transport and accessibility guaranteed by
a determining factor for social inclusion policies.
private transport. This concept can be expressed
by the equation (1):
2 METHODOLOGY
AiTPb
AR = (5)
2.1 TransCAD model AiT Pr
In the present study it used a software for the
management and analysis of transport data that where AR is the relative accessibility, AiTPb is the
is called TransCAD, produced by Caliper Cor- accessibility provided by public transport and AiTPr
poration (USA). TransCAD combines a GIS is the accessibility by private transport.
(Geographic Information System) with analytical According to the aim of this work and in order
models and transport planning in one integrated to include social issues, we define the Relative
environment. Weighted Accessibility (RWA) index as follows:

Ai
TPb
b( t )
⎛ Pop o iCar ⎞
2.2 Hansen’s index computation W =
RWA ⋅ ⎜1 − (6)
A
T Pr( t )
i ⎝ o iTot ⎟⎠
Pop
In order to evaluate Hansen Index an impedance
matrix among zones has been calculated.
where PopiCar is the population that has access to
The impedance matrix tells us how difficult is to
private transport (specifically people over 18 years
travel from an origin to a destination and considers
old and under 70 years old) and PopiTot is the total
the generalized cost of transport. It also translates
population at the origin. People that do not have
into monetary cost factors not monetized ones, as
access to private transport are assumed to be highly
the travel time (that the user tends to minimize),
correlated with the number of disadvantaged peo-
physical effort, comfort, safety and user stress.
ple, (such as children and elderly are) i.e. people
The matrix has been realized considering imped-
who have lower car availability and therefore will
ance as the generalized cost of transport and tak-
benefit more largely of higher transit accessibility.
ing into account parameters such as travel time,
In this way, the RWA index is decreased for
the cost of travel time, the number of transfers and
those zones where a high ratio of disadvantaged
travel fare.
people is present.
Hansen function used by TransCAD to perform
the calculations is:
3 CASE STUDY
∑
n
O j e − λCij
A = i=0
(3) 3.1 Territorial framework and transport system
∑
i n
i=0
Oj
Catania is a city of about 300.000 inhabitants and
it is located in the eastern part of Sicily; it has an
where Ai is accessibility of zone i, Oj are the
area of about 183 km2 and a population density of
opportunities in the zone j, Cij is transportation
1.754,54 inhabitants / km2 (Istat, 2015b). It’s part
cost between zones i and j and λ is the deterrence
of a greater Metropolitan Area (750.000 inhabit-
parameter. λ value should be calibrated fin each
ants), which includes the main municipality and
context according to the situation, but in this case
26 surrounding urban centers, some of which
it is used the value provided by default for each
constitute a whole urban fabric with Catania. The
transport mode and trip purpose.
main city contains most of the working activities,
Results will be then normalized with the nor-
mixed with residential areas. With reference to the
malizing function:
urban area, the transport service is provided by
51 bus lines, a Shuttle line (ALIBUS) connecting
Ai Amin
ANi = (4) the city center with the airport and a second fast
Amax Amin bus (called BRT1) connecting the parking Due
Obelischi with Stesicoro Square. BRT1 is the first
where ANi is the normalized value of accessibility of three lines provided by the City of Catania with
for zone i, Ai is the generic value of accessibility equipped lanes protected by curbs on the major-
for zone i, and Amin and Amax are respectively the ity of their path and was promoted commercially
minimum and the maximum value of accessibility as Bus Rapid Transit. In Catania it is also present
among all the case study’s zones. an urban subway line that currently connects the

755

TIS2017_Book.indb 755 2/8/2017 8:09:33 PM

 station “Porto” with the station “Borgo” from Relative Weighted Accessibility has been then
which continues as a surface long-distance line. evaluated for the base scenario by taking into account
By 2016 it is expected the undergrounding of Equation 6 and results have been mapped (Fig. 3).
the line until the station “Nesima” and it’s also Weighting the accessibility index as indicated before
planned the opening of a branch linking the sta- generally produces lower values for zones with a low
tion “Galatea” to Piazza Stesicoro. number of resident with access to private transport.
So it can be noticed that the weighted index seems
more adequate to support transit planner choices
3.2 Transport model and scenarios
when the spatial distribution of disadvantaged social
Transport model has been implemented in Tran- groups has to be taken into account properly.
CAD software, a software which combines Geo- Through the normalized accessibility values it
graphic Information System and analytical models is possible to derive a histogram showing the fre-
and transport planning in one integrated environ- quency distribution and the cumulative curve for
ment. The zonation used for the city is the one traffic areas defined.
given by ISTAT, which divides the study area in On the x-axis accessibility classes are shown,
2480 CENSUS sections (Fig. 1). while on the y-axis we can see the frequencies. The
Methodology has been applied with reference to cumulative establishes the percentage of areas with
the trips made by the citizens of Catania for work a certain accessibility value. The histogram can be
purpose. seen in Figure 4.
Three different scenarios have been considered:
3.2.2 Project scenario 1
− Base scenario: 51 urban public transport bus
In this case we tried to understand how accessibil-
lines; it is assumed for these bus a commercial
ity to workplaces varies when three lines of rapid
speed of 15 km/h.
buses, L-EX, BRT1 and BRT2 are introduced into
− Project scenario 1: insertion of three rapid tran-
the network. It is assumed for these buses a com-
sit lines (L-EX, BRT1 e BRT2), with a commer-
mercial speed of 20 Km / h, superior to that of all
cial speed of 20 km/h.
other buses (15 Km / h).
− Project scenario 2: increase of commercial speed
As you can see in Figure 5 the three buses cover
of all bus lines to 18 km/h.
the whole territory of Catania and in particular: the
BRT1 moves from north to south; the BRT2 from west
3.2.1 Base scenario to east and L-EX follows a route in the southeast.
Hansen accessibility has been evaluated for the first The calculation of RWA for the Project
scenario, through the use of TransCAD; results are Scenario 1 shows little improvements of accessibil-
showed in the thematic map in Figure 2. The map ity for all the zones in the study area; a thematic
reveals how the high-resolution zoning can repre- map of the improvements can be seen in Figure 6.
sent a significant difference in accessibility index
even for adjacent cells, which would have been
‘flattened’ by a lower resolution approach.

Figure 1. Transport model. Figure 2. Accessibility map for base scenario.

756

TIS2017_Book.indb 756 2/8/2017 8:09:34 PM

 Figure 5. BRT lines.

Figure 3. RWA map for base scenario.

Figure 6. RWA difference map for project scenario 1.

Figure 4. Histogram of base scenario’s RWA.
Once again we present the results in a histogram
Once again we present the results in a histogram (Figure 9) that shows the frequency distribution
(Figure 7) that shows the frequency distribution and the cumulative curve for the zones defined.
and the cumulative curve for the zones defined.
3.2.4 Results
3.2.3 Project scenario 2 Comparing the three scenarios we see how in all
In this second scenario we tried to understand the cases we have a low accessibility in the indus-
how accessibility to workplaces varies if we raise trial area (located in the south) and this is due
the commercial speed of all city buses already to the geographical spread of the central areas
presented in the scenario base. This speed was served by buses and the industrial zone, and the
increased from 15 km/h to 18 Km/h. low capillarity of the public transport network
The calculation of RWA for the Project in that area. It can be noticed, however, an incre-
Scenario 2 shows big improvements of accessibility ment in accessibility as we get closer to the center
for all the zones in the study area; a thematic map or we move toward the northeast, because they are
of the improvements can be seen in Figure 8. areas served by public transport lines. In any case

757

TIS2017_Book.indb 757 2/8/2017 8:09:35 PM

 it is clear that between the baseline scenario and
project scenarios we have a total increase of acces-
sibility to workplaces.

4 CONCLUSION

In this paper new transit accessibility measure and

indexes has been presented and their use has been
experimented to evaluate the impact of introduc-
ing BRT lines in the transit network of the city
of Catania. They have proved a good sensitivity
to: the walking trip segment along the pedestrian
network to access the transit system, the land use
opportunities at destinations and the social groups
using the transit system.
The relative weighted accessibility measure
Figure 7. Histogram of RWA for scenario 1.
can be considered a useful tool to address a new
approach in transport planning, when the main
goal is to provide people with a high potential to
reach the goods and services necessary for daily life,
regardless their social status and car availability.
A high transit accessibility compared to car acces-
sibility is a precondition for economic develop-
ment as it enables and empowers the opportunities
for people interaction (workers and consumers),
a performance indicator of the availability of less
energy-intensive and lower impact modes of trans-
port, a measure of equity in terms of distribution
of transport options among citizens with different
social and economic power.
In addition, the relative weighted accessibility
can be a useful transport planning indicator to
address changes in the transit network design and
operation, in order to avoid the risk of uncritical
increase in transit accessibility of zones where peo-
ple already have a high car ownership and land use
accessibility, thus increasing the social exclusion
of people who have less access to private car and
Figure 8. RWA difference map for project scenario 2. transit services.

REFERENCES

Cascetta E. (2009), Transportation System Analysis

Models and Applications. Springer.
Geurs K. T. and Van Wee B., 2004. Accessibility evalua-
tion of land-use and transport strategies: review and
research directions. Journal of Transport Geography,
Delft.
Hansen W. G., 1959. How Accessibility Shapes Land
Use, Journal of the American Institute of Planners.
Vol. 25, 73–76, USA.
Litman, 2009. Evaluating Transportation Land-use
Impacts, Victoria Transport Policy Institute, Canada.

Figure 9. Histogram of RWA for scenario 2.

758

TIS2017_Book.indb 758 2/8/2017 8:09:36 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Shared infrastructures: Technique and method for an inclusive

social valuation

Saverio Miccoli
Department of Civil, Building and Environmental Engineering, La Sapienza University of Rome, Rome, Italy

Fabrizio Finucci
Department of Architecture, Roma Tre University, Rome, Italy

Rocco Murro
Department of Civil, Building and Environmental Engineering, La Sapienza University of Rome, Rome, Italy

ABSTRACT: In Italy, the implementation of many infrastructures is stuck because of ongoing conflicts
between political and administrative decision-makers, operators and users. Aware of the current situation,
a recent Legislative Decree introduced the obligation of the Public Debate between the stakeholders in
order to reduce the likelihood of infrastructure projects not being implemented due to the lack of neces-
sary social consensus. Consensus-building activities require the use of valuation approaches that actively
involve the community. To this end, the paper proposes an inclusive valuation technique aimed at the
direct and conscious deliberation of individuals involved in an infrastructure project combining delibera-
tive methods, based on community debates, with a social multi-dimensional evaluation.

1 INTRODUCTION ing. To ensure funds were used efficiently, in the

Community Programming 2014–2020, Member
In the Italian regulatory framework, the two main States were asked to guarantee a series of ex ante
investment-planning tools are a transportation conditionalities of a regulatory, administrative or
plan and a programme for strategically important organizational nature. Within the framework of
infrastructure projects. In 1984 with Law n. 245, Community programming, the concept of condi-
the General Transport Plan (GTP) of Italy was tionality are prerequisites Member States have to
introduced with the aim of ensuring the existence fulfil in advance to guarantee they have the regu-
of a single national transport policy. The Strategic latory and organisational resources necessary to
Infrastructures Programme (SIP) was introduced ensure full adhesion to EU strategies and the effec-
in Italy’s national legislation more recently, in com- tive feasibility of the investments supported with
pliance with Law 443/2001 (the so-called Objective allocated funds.
Law). This measure provided for the identifica- The updating of national regulations and the
tion of the public infrastructures and production adjustments introduced to fulfil the prerequisites
plants of particular national interest, in full respect of the ex ante conditionalities, resulted in the
for the regions’ autonomy, through a programme Infrastructure Appendix (IA) to the Economic
established by the Ministry of Transport and and Financial Document (EFD), the single docu-
Infrastructure (MTI), in agreement with the com- ment that provides the framework for the plan-
petent ministers, after consultation with the Italian ning and investments of national interest in the
Inter-ministerial Committee for Economic Pro- transport sector. The IA of 2015 was the first to
gramming (ICEP). Furthermore, it was decided recognise the infrastructure planning reform. The
to include this programme in the Economic and first part of the document provides a general over-
Financial Planning Document (EFPD) specifying view of transport infrastructure planning based
the relative funding. on an analysis of transportation demand (goods
The need to coordinate the infrastructural inter- and passengers) and a SWOT analysis of the entire
ventions linked to national requirements with the national infrastructure sector. In the second part
main transport policies at Community level also of 2015 IA, strategies to address the critical issues
implied the thorough revision of Italy’s legisla- identified by extending and improving the entire
tion regarding both project planning and prioritis- infrastructural network (railways, roads, ports

759

TIS2017_Book.indb 759 2/8/2017 8:09:39 PM

 and airports) are proposed. To make this objective those with Binding Legal Obligations (BLO) and
operational a number of implementing tools are those without; c) the new projects proposed by local
used: a) the Strategic Infrastructures Programme authorities or institutions. The 25 strategic projects
(SIP); b) the ANAS Programme contract related along with those included in the SIP without BLO
to the restructuring of the motorway sector; c) the will be included in the Project Review after which
RFI Programme contract to extend and improve they will be inserted in the first MPD 2017–2019.
the national railway network; d) the National Plan The SIP projects without BLO and any new projects
for Ports and Logistics; e) The Airport Plan. proposed by local authorities or institutions will be
The implementing tools are linked to specific subject to an evaluation related to infrastructural
financial tools: a) the Connecting Europe Facil- needs. This evaluation, of a quantitative nature,
ity (CEF) and the European Fund for Strategic will consist in identifying the shortage of elements
Investments (EFSI); b) the European Structural in the infrastructure system necessary to satisfy the
and Investment Funds; c) the Development and demand. Once identified, feasibility projects will
Cohesion Fund 2014–2020, for the achievement be carried out to find the most suitable solutions to
of strategic objectives related to national thematic remedy the shortage. Subsequently, the feasibility
areas. The document also specifies 25 top priority projects will be subjected to an ex ante evaluation,
projects to include in the SIP; the selection is based and only after they have undergone this procedure
on their integration with European networks, the can they be included in the MPD.
progress of ongoing works, and the possibility The infrastructure planning and programming
of their being funded mainly with private capi- process is innovative also thanks to several proce-
tal. As regards the other projects included in the dural changes introduced with Legislative Decree
SIP (called Other Interventions), the document 50/2016, or the New Procurement Code (NPC).
suggests proceeding following a preliminary con- The most important innovations include: a) the use
sultation with the Regions. The other new develop- of the Project Review; b) the feasibility projects;
ments envisaged in the IA to the EFD 2015 include c) Public Debate in the decision-making process
the launching of a new IT system to monitor the for the creation of infrastructures.
progress of ongoing works (called Opencantieri) The Project Review is a tool provided for under
by the MTI. Art. 202 of the NPC to verify if a project selected
The IA to the EFD 2016 is defined by the MTI with previous planning processes has meantime
itself as a new Planning process for public works become oversized due to variations in the gen-
centred on the planning of interventions that are eral state of affairs or new market conditions.
useful from an intermodal and sustainable view- Depending on the outcome of the verification
point and that enhance of existing infrastructures. process, performed mainly with evaluation tools
It introduces two new tools: the first is the General and procedures, the functional characteristics of
Plan for Transportation and Logistics (GPTL) that the project in question can be redefined or fund-
will contain strategic guidelines for policies regard- ing may be revoked. The purpose of the feasibility
ing the mobility of people and goods, and for the project introduced and regulated in Arts. 23 and
infrastructural development of the nation. The 202 of the NPC is to improve the quality of infra-
second is the Multiannual Planning Document structural planning, programming and design.
(MPD) that will contain the interventions related The objective of design at this level is to verify the
to the transport and logistics sector, the feasibil- technical, economic, environmental and territo-
ity studies of which must have received a positive rial conditions for the creation of infrastructures
ex ante evaluation to show the projects deserve by selecting the solution with the best cost-benefit
funding. Furthermore, the scheme envisages the ratio for the Community. Lastly, the forms of
possibility for the MPD to be integrated with pro- participation are recognized in national regula-
posals from the Managing Bodies, the Regions and tions with the introduction of a structured process
the Local Authorities involved. As provided for in that involves communities in the decision mak-
Legislative Decree n. 228/2011, the MTI publishes ing regarding infrastructure projects, i.e. Public
a detailed annual report annexed to the EFD on Debate pursuant to Art. 22 of the NPC. Although
the state of implementation of the MPD. timing, modalities and techniques have still to be
After representing the framework of reference established through future implementing decrees,
and context of the EFD 2015 and summarising the the NPC clearly states that local administrations
ongoing activities, the IA to the DEF 2016 estab- have to make public the feasibility projects regard-
lishes a roadmap for the drawing up of the first ing infrastructural and architectural interventions
MPD. In this process, the public works are divided of social importance with a significant impact on
into three main groups: a) the 25 top priority stra- the environment, the city and the territory. The
tegic works (Annexed to EFD 2015); b) the works outcomes of the public consultations including
included in the SIP divided into two typologies— the minutes of meetings held and debates with the

760

TIS2017_Book.indb 760 2/8/2017 8:09:39 PM

 stakeholders also have to be published. Currently, it is possible to see that approximately one in every
open conferences with the participation of admin- ten is due to an infrastructural intervention.
istrations, stakeholders and citizen committees are There are ongoing conflicts linked to 5 of the
foreseen. This process will be mandatory for large 25 strategic projects: the Turin-Lyons high-speed
infrastructure projects and an ad hoc MTI decree train link, the Valico dei Giovi railway, the Pede-
will establish size thresholds and what kind of montana Lombarda motorway, the Tangenziale
project will undergo Public Debate. The debate will Est in Milan a the Experimental Electromechani-
focus on the feasibility project to give the propos- cal Module MOSE. The lack of social consensus is
ing institutions the possibility to listen and receive affecting all these public works in terms of delays
information and suggestions on every aspect of the and increased costs. In addition, the progress of
project, and to add new ideas and further clarifica- work on the 25 strategic projects does not mean that
tions if necessary. further protests will not arise. The overall situation
After outlining the regulatory framework, regarding the progress of work on the 25 strategic
the paper: a) deals with the topic of inclusion in projects is the following: a) under 15% of the work
infrastructure planning; b) searches for shared needed to complete seven works (28%) has been
decisions by proposing a deliberative multidi- carried out (four have still to be started); b) under
mensional evaluation procedure; c) ends with a 50% of the work needed to complete almost half
synthesis of the advantages of the proposed tech- (12 projects) has been carried out; c) under 2/3 of
nique and highlights the main obstacles to its full the work needed to complete sixteen projects has
implementation. been carried out. Italy’s infrastructural modernisa-
tion process is also linked to many public works
provided for in the SIP: a fragmented series of
2 EVALUATION AND INCLUSION interventions of different sizes regarding ports,
IN THE PROCESSES FOR THE airports, interchange junctions, and stretches of
CREATION OF INFRASTRUCTURES railways and road.
In Italy’s new infrastructural programme, the
Over the last two decades, the topic of conflict and evaluation and inclusion of citizens in the decision-
consensus when it comes to creating large infra- making process is considered crucial to the success
structures, and public choice in general, has influ- of the interventions. Evaluation plays a pivotal
enced Italy’s political agenda. Albeit later compared role in various phases: the drawing up of feasi-
to other countries, Italy has also put in place inclu- bility studies for new projects proposed by local
sive procedures to cope with the main decisional authorities or institutions and in the updating of
stumbling blocks linked to public works. Initially, the GPTL and MPD. During the implementation
a few regional laws were introduced, e.g. those of of the first MPD, evaluation is the main activity as
the Regions of Tuscany and Emilia Romagna, and far as decision making is concerned and includes
local council regulations such as those introduced selecting interventions, identifying priorities and
in Rome and Milan that regulate participatory Project Review. Each phase has to be supported
practises. The adjustment process for the adop- by evaluation procedures that are adequate for the
tion of these practises at national level was finally problems peculiar to the infrastructure project in
regulated in the recent NPC with the introduction question. These evaluation procedures must be
of Public Debate. The IA to the EFD 2016 also capable of dealing with the innovations related to
states that, “often, the projects handed down from scientific approaches and to the regulations that
on high, which do not take into account the micro- from a structural point of view guide and reflect
restrictions of the territory in question, give rise technological evolution. From an evaluation view-
to opposition while work is in progress and this in point, the MTI is preparing standardised guide-
turn gives rise to uncertain and approximate imple- lines for the evaluation of investments in public
mentation costs and deadlines”. In literature, this works with the aim of ensuring that all works are
kind of opposition is called the NIMBY (Not In evaluated homogeneously and selected with trans-
My Back Yard) syndrome. As the local community parency. As regards social inclusion, the Ministry
does not identify itself in the choices of the deci- itself is working on the choice of modalities for
sion makers and believes the consequent transfor- Public Debate as provided for in the NPC.
mation will prove detrimental to local interests, it As far as this paper is concerned, these develop-
begins to protest against the implementation of the ments are irrelevant as our intention is to propose
project in question (Miccoli et al. 2015a). Currently guidelines for an operational tool and the method
in Italy, according to the Nimby Forum, there to implement it capable of providing effective solu-
are 355 ongoing projects affected by the protests tions in situations where the objective is not to help
of local residents. By subdividing these conflicts administrators-politicians who are the legitimate
according to the category of project in question, representatives of citizens to find out about their

761

TIS2017_Book.indb 761 2/8/2017 8:09:39 PM

 opinions through Public Debate, but to allow the to take into consideration a number of alternative
citizens themselves to make informed, debated and projects, various criteria and their order of prior-
shared choices on the infrastructural problems in ity, and the multiple impacts of an environmental,
question. social and economic nature that the infrastructure
Therefore, preference regarding the two afore- in question would have on the territory. With a
mentioned decision-making pathways is not called view to foreseeing problems linked to distributive
into question; the same goes for their possible justice, the evaluation approach is also social, capa-
complementarity or coexistence, and the possibil- ble of taking into account the viewpoints of all the
ity to merge them into a single procedure. stakeholders—be they single, collective, private
In our opinion, in situations where it is neces- or public—who are directly or indirectly affected
sary to know what a community thinks about by a complex project. As the procedure is based
alternative infrastructural solutions, the formalisa- on dialogue and negotiation, as well as mediation
tion of an assessment pathway carried out directly between diverse interests and values, the evaluation
by the members of the community with techniques model envisages the systematic participation of the
and methods suited to dealing with the complex community in question. It goes beyond the more
factors at stake, can result in an efficient choice traditional social consultation models and tries to
because it is characterised by responsibility, trans- directly involve the community through techniques
parency, rationality, and full awareness of the pos- based on information, communication, opinions
sible distributive effects. Thereby the valuation and debate between stakeholders.
approach makes it possible to abandon the context In recent decades in many countries, the involve-
of opinions elicited during a public debate and ment of citizens in decision-making processes has
embrace that of certainties produced by scientific been consolidated through a variety of methods
approaches. Substantially, we are speaking about and tools that are indispensable for making pub-
a realistic choice that does not arise from theoreti- lic choice successful. The inclusive approach,
cal or reductionist scenarios, but from situations intended as a tool to manage and bring together all
associated with action and the effective adapta- the interests at stake in a given territory, combines
tion of the members of a community. We firmly the development and safeguard of the local com-
believe that the use of an evaluation technique of munity with a view to making sustainable decisions
a dialogic-deliberative nature, combined with mul- (Miccoli et al. 2015b). To this end, the action taken
ti-criteria approaches, makes it possible to devise a is based on transparent information, consultation,
more evolved and binding form of Public Debate, dialogue and negotiation, in a bid to ensure the
a tool otherwise destined to be used as a mere community’s participation in a decision-making
consultation mechanism with no totally inclusive process the aim of which is to make a shared
decision-making value. Taking note of the opin- choice based on consensus and full legitimacy.
ions stated during citizens’ assemblies undoubt- Participatory processes, therefore, imply a demo-
edly provides important information for public cratic value as they ensure that all the stakeholders
decision-making, but it is a far cry from knowing even when their interests differ have the right to
the measure of opinions regarding choices based express their opinion and the possibility to influ-
on analyses and on qualitative-quantitative data ence and determine a choice. Another aspect of
stated by those who for various reasons are inter- the utmost importance is a strategic value because
ested in an infrastructure project. In the evaluation public choice, which is founded on the democratic,
technique described below, the community itself, transparent and rational principles of inclusion,
with different modalities and roles, fixes criteria, is shared consensual and less conflicting, and has
establishes priorities, and states evaluations with more chance of being successful.
the aim of reaching a shared choice. An inclusive process can be developed through a
number of different techniques, which can be clas-
sified according to the share of decision-making
3 AN INCLUSIVE SOCIAL power that is transferred from the institutional
MULTIDIMENSIONAL EVALUATION decision maker to the participants. Some tech-
niques simply disseminate information among the
With the aim of providing an evaluation model that public, some involve the community in the consul-
not only takes into account the complexity, risk and tations, some seek the cooperation of all the stake-
uncertainty attached to infrastructure projects, but holders to define possible scenarios and identify
also supports public choice and actively involves solutions, while others leave the decision-making
the community in order to reach a shared choice, entirely to the citizens.
we propose a procedure based on an inclusive social In the most updated processes, the directly delib-
multidimensional evaluation. The purpose of a erative procedures based on assemblies, groups
multi-criteria based multidimensional evaluation is or citizen juries are becoming more common.

762

TIS2017_Book.indb 762 2/8/2017 8:09:39 PM

 Although born in the ’70 s, it was only in the ‘90 s − The second participatory evaluation step with
that these procedures began to be widely used first the stakeholders and the construction of multi-
in Germany and then in England, the USA and ple impact matrixes;
Australia. The various deliberative techniques − Defining partial rankings of the single stake-
developed over the years include the Plannung- holder categories—according to the impact
szelle (Dienel, 2002), Ned Crosby’s Citizen Jury matrixes (opinions stated by the stakeholders)
(Smith & Wales, 2000), the Opinion Deliberative and of the priority vectors (opinions stated by
Polls (Fishkin, 1991) and the Consensus Confer- the community)—and an overall multi-stake-
ence (Andersen & Jaeger 1999). The outcome of holder ranking.
all these techniques is a preference based on aware
and shared informed choices, obtained after all the In the evaluation procedure, the stakeholders
stakeholders have stated and discussed their points do not represent the entire community. The stake-
of view, and possibly changed their opinions after holders are subjects, at times of a collective nature,
discussing them (Miccoli et al. 2014b). Citizens are (current and future, real and potential) who are
no longer the receiving end of the decision-making directly or indirectly interested in an infrastructure
process but are full protagonists in the underlying project; they are asked to evaluate the alternatives
debate (Miccoli et al. 2014a). according to their personal interest in it. On the
other hand, the community is represented by all
the subjects that share a territory, but unlike the
3.1 Technique and method for inclusive social
stakeholders may have no interest whatsoever in
evaluation
the project, although they have the right to provide
The evaluation technique we propose is divided guidance to safeguard common interests. After
into the following phases (Fig. 1): consulting the community by means of a random
representative sample, the scenario of the evalua-
− Defining the key elements on which the evalua-
tion is obtained and it is the synthesis of the pos-
tion is to be based;
sible points of view present in a given territory.
− The first participatory evaluation step with the
During the initial phase of the procedure we
community;
propose, the course of actions and the elements
− Constructing the input matrix according to the
to prepare the ground for the actual evaluation are
criteria and sub-criteria identified through the
established through a) identifying all the alterna-
community’s evaluation;
tives to be evaluated; b) suggesting the set of criteria
and sub-criteria for the evaluation of the alterna-
tives; c) specifying the categories of stakeholders
interested in the infrastructure project, defined
according to the characteristics of the alternatives
to be evaluated and whether they impact at local,
urban or regional level.
In the first step of the participatory evaluation,
the community states its point of view through a
deliberative approach by selecting among the cri-
teria and sub-criteria submitted and adding oth-
ers (where necessary), specifying their priorities
and establishing the importance to attach to each
stakeholder category. Once the criteria and sub-
criteria to be used for the evaluation have been
defined, it is possible to move on to the construc-
tion of the input matrixes by establishing for each
sub-criterion the relative indicators (qualitative
and/or quantitative) and to measure them.
The initial phase of the second step of the par-
ticipatory evaluation consists in identifying the
stakeholder categories present on the territory
the members of which are called on to evaluate
the matrix inputs by stating whether, in their opin-
ion and according to their particular interests,
the alternative would be capable of satisfying the
sub-criterion/criterion taken into consideration.
Figure 1. Phases of the evaluation procedure. Once this phase has been completed, it is possible

763

TIS2017_Book.indb 763 2/8/2017 8:09:39 PM

 to define multiple impact matrixes for each stake- often-contrasting viewpoints of the various actors
holder category according to the points of view involved, suggest a common shared position. One
stated. Lastly, by taking into account the outcomes of the many techniques developed over the years
of the second step and the evaluation of the com- is Deliberative Opinion Polls, an idea pioneered
munity (priority vectors for the criteria, sub-crite- by James Fishkin. In this technique, the commu-
ria and stakeholder categories) different evaluation nity’s point of view regarding a particular problem
matrixes for each stakeholder category are formed. is stated only following a debate between a group
By applying a two-tier multi-criteria based eval- of citizens chosen by lot and a group of experts
uation method (Miccoli et al. 2013), it is possible to on the issue under examination. These new forms
presume multiple partial priority rankings of the of surveys differ greatly from the traditional ones
alternatives related to single stakeholder categories. as the latter do not gather opinions but something
From a purely operational standpoint, the evalua- else, because people often have no clear idea and
tion procedure is subdivided into two phases: respond without being informed. On the other
hand, the result of a Deliberative Opinion Poll is
− the first consists of a series of unidimensional
an aware and informed choice obtained “through
evaluations performed according to the sub-
debate, stating opinions and the possibility to
criteria, in accordance with the general criteria
change them after discussing and comparing them
(based on the impact matrixes and priority vec-
with those of others” (Fishkin et al. 2014).
tors among the sub-criteria), which makes it pos-
In the evaluation procedure we propose, the com-
sible to obtain a mono-dimensional ranking of
munity is asked to select and list the various factors
the alternatives for each stakeholder category;
by a) prioritising the general criteria proposed and,
− the second, provides the multidimensional evalu-
if necessary, single out new criteria; b) prioritising
ation of the general criteria for each stakeholder
the sub-criteria proposed divided by the general cri-
category, by taking into account the results
teria that can also be integrated; c) prioritising the
obtained from the unidimensional evaluations
categories of stakeholders involved in the evaluation
and from the priority vectors among the general
process. The outcomes will be used to evaluate the
criteria (indicated by the community), as well as
impact matrixes defined during the stakeholders’
defining the multidimensional rankings of the
evaluation. The preparatory phase of this partici-
alternatives for each stakeholder.
patory step consists in deciding on a work schedule
The rankings obtained in this way make it possi- for the duration of the procedure and forming a
ble to clarify the extent of conflict between the var- Steering Team whose task is to supervise the single
ious stakeholder categories and can be a starting discussion groups, ensure that everybody states his
point for further debate and possibly rethinking or her opinion, and provide the groups with unbi-
the alternatives under examination with the aim of ased information. In this phase, the individuals
minimising conflict on one hand and harmonising who interact with the community (experts and
the various interests at stake in the infrastructure stakeholders) are identified. The experts are people
project on the other. At a later stage, it is possible capable of clearing up any doubts the participants
to define a multi-stakeholder evaluation matrix, may have regarding the criteria and sub-criteria
based on the preferences stated by the single cat- under discussion; the selection of the stakeholders
egories (indicated by the partial rankings), and depends on their presence and activity on the
appropriately weighed by the community (priority territory, on the characteristics of the alterna-
vector of the stakeholder categories). By applying tives to evaluate and at which level they affect the
once again the two-tier evaluation method, an over- territory. Once the preparatory phase has been com-
all multi-stakeholder ranking of the alternatives is pleted, the participants who will form the sample
obtained. To simplify the community’s evaluation, are selected by means of the most suitable random
ordinal scales for the measurement of priorities sampling techniques. The techniques chosen must
have to be used. Therefore, during the final evalu- take into account the size and socio-demographic
ation, techniques that make it possible to employ characteristics of the target population, to ensure
the weights of the criteria expressed in the form of the sample is fully representative of the communi-
order of priority will be preferred. ties involved in the project.
Schedules for two kinds of meetings are drawn
up—group sessions during which the participants
3.2 The community’s evaluation
compare their ideas, and plenary sessions when
The first participatory step is an evaluation based the participants address the experts and stakehold-
on the approaches of Deliberative Democracy. ers. At the opening of the first session, the experts
The underlying principle is that of presenting a explain the purpose of the evaluation, the main
problem to a group of citizens who, after listen- regulations for each daily session and the project
ing to information provided by experts and the to be evaluated. At the end of the introduction,

764

TIS2017_Book.indb 764 2/8/2017 8:09:40 PM

 the participants are asked to fill in a questionnaire; by combining these matrixes with the priorities
their answers will be based on their knowledge expressed during the community’s evaluation the
of the project. The participants are also asked to final step of the proposed procedure can take place.
fill in a questionnaire at the end of the evaluation The choice of the method depends on the type of
process after which the outcomes are compared project, the specific characteristics of the territory,
and the participants’ preference shifts at the end to what extent the participation of the possible
of the deliberative process—i.e. after exchanging stakeholders is expected to be dynamic, and on the
information and opinions with each other—are relationship between the various individuals.
analysed. Individual Evaluation: With this method, the
The community’s evaluation is carried out in a stakeholders do not interact during the evaluation
number of sessions, one for each priority that has to process. They are contacted directly and provided
be defined. Each single session is based on a model with informative material, which they have to
session and comprises two group meetings and two complete and hand in with their evaluation. This
plenary meetings. A coordinator is chosen for each procedure requires a well-organized information
group with the task of guiding and moderating and communication campaign on the territory for
the debate to ensure participants total freedom of the entire duration of the evaluation process. The
expression and opinion, to encourage everyone to coordinators of this procedure set up an informa-
state their opinion and to prevent the more active tion point where the stakeholders can pick up the
participants from monopolizing the debate. Dur- material, ask for information and clear up any
ing the first sessions, the participants have to agree doubts, hand in the material once filled in, and
on and draw up a list of questions to put to the receive assistance regarding the evaluation itself.
experts and the issues to discuss with them; dur- The duration of the evaluation depends on whether
ing the subsequent plenary session the experts and the coordinators of the procedure or the Public
stakeholders will clear up any doubts and answer Administration involved have spread correct infor-
the questions that arose from the group meetings. mation to all the stakeholders, and on the reaction
A new group meeting during which the answers and response of the stakeholders themselves. The
provided are discussed, and a conclusive plenary outcome is an evaluation on the part of the stake-
session will bring the procedure to an end. At the holders who can state their opinion without having
end of each Model session, the participants fill in a to discuss their point of view or compare it with
questionnaire that enables them to formulate new those of others.
criteria and sub-criteria and to prioritise them. Compared Evaluation: Initially, this method
At the end of this first participatory evaluation foresees an individual evaluation as described
step, the opinions of the people called upon to above, followed by a meeting-assembly to which
represent the community are no longer based on all the stakeholders who have performed the evalu-
superficial notions but on aware informed knowl- ation are invited. During the first phase of this
edge stemming from the debates, exchange of public event, the outcomes of the first evaluation
information and a deliberative pathway. and the tendencies of each stakeholder category
are stated. Subsequently, everybody in turn has
the possibility to explain his/her intentions and
3.3 The stakeholders’ evaluation
the reasons behind his/her preferences. At the end
Before proceeding to the second participatory of the assembly (or assemblies depending on the
evaluation step, the Steering Team identifies the number of people involved), the stakeholders can
stakeholders. It is crucial for the territory and the rethink or modify their original evaluation—stated
project to be attentively examined before identify- in the first phase—and another session can be
ing the participants, and to do everything possible called if deemed necessary. This method gives the
to ensure that everybody involved has the oppor- stakeholders the possibility to compare, and possi-
tunity to state their opinion and that nobody is bly modify their original evaluation, after listening
excluded a priori (Abelson et al. 2003). The prepa- to the evaluation of others. This type of procedure
ration of informative material for the stakehold- can be repeated until the most widely shared posi-
ers involved is also indispensable. A wide range tion possible is achieved.
of existing techniques for public consultation and Co-valuation: Lastly, this method makes it pos-
participation can be used to carry out the evalua- sible to obtain a negotiated weighted evaluation
tion (questionnaires, interviews, market research, from the dialogic interaction between the vari-
mobile territorial observatories etc.). For our case, ous stakeholders. One, or if necessary a number
we propose the use of three different methods of meetings based on the fundamental steps of
a) Individual Evaluation; b) Compared Evaluation; participatory techniques are organized. During
c) Co-valuation. These methods produce impact these meetings, the participants are asked to evalu-
matrixes directly assessed by the stakeholders; ate the matter under examination and to discuss

765

TIS2017_Book.indb 765 2/8/2017 8:09:40 PM

 their evaluation with the other stakeholders. Once EFD, the IA is published annually. This demon-
all the possible stakeholders have been identified, strates the decisive role played by infrastructures in
they are asked to participate in a workshop. In this driving Italy’s economic growth, but also the urgent
phase, they acquire specific knowledge of the mat- need to realign the country’s infrastructural condi-
ter under examination, and they state their inter- tions with the more advanced European standards.
est and compare it with and relate it to that of the In fact, over the last few decades, the infrastructure
other participants. During the workshop, alternat- gap between Italy and the rest of Europe has grad-
ing sessions with the participation of homogene- ually widened with considerable repercussions on
ous stakeholder groups are held, as well as plenary the overall performance of the country’s system.
sessions during which the evaluations of each Bridging this gap by programmatically focusing
group are stated, debated and discussed. on priorities in infrastructure investments implies
The work pattern is similar to that used for a pressing need to correlate their creation with the
EASWs (European Awareness Scenario Work- principles of sustainable development considering
shops) (Andersen & Jaeger 1999) but in the the dimensions and effects of works of this kind.
procedure we propose the stakeholders work in Above all, however, citizens’ interpretation of the
homogenous (single category) and not hetero- impact these projects have on the environment, the
geneous groups; in this way the evaluations and economy and on the quality of social life needs to
opinions are always representative of a specific cat- be taken into consideration.
egory and its interests. After a brief introduction In order to ascertain if conditions of this kind
and illustration of the criteria and sub-criteria, are included in the implementation plan, a prelimi-
the group sessions during which the stakehold- nary evaluation of the feasibility of these projects
ers formulate their evaluation are held. Once this with regard to established prerequisites of sus-
phase is completed, two plenary sessions are held tainability must be carried out. This is necessary
during which each group briefly illustrates their because of both the multifaceted, heterogeneous
evaluation and the positions stated by the various and often conflicting nature of the interventions,
categories are discussed. At the end of the debate, and the expectations of those participating in the
the participants are once again divided into groups decision-making who for different reasons are
so they can re-formulate their evaluation based on interested in their implementation and who would
the positions stated by the other stakeholder cat- not hesitate to protest and state their opposition
egories. During the conclusive plenary session, the if they were not consulted. It was partly due to
participants decide if the outcomes are exhaustive this last consideration that a decision was taken to
evaluations shared by every single participant, or include the practice of Public Debate in the NPC.
if a new debate and a new evaluation are called for. Therefore, the evaluation proposed in this paper
This iterative process makes it possible to obtain is complex as it is necessarily multidimensional,
evaluations that are the result of negotiated and social and inclusive given the problems associated
shared interests. to the sphere in which it is used.
Within the framework of public evaluation,
those of a social and inclusive nature are radically
4 CONCLUSIONS different from ordinary institutional evaluations
because of the perspective in which they are con-
The relatively recent updating of regulations ducted. They are likely to become an important
regarding the infrastructure sector and contracts element of democratic life over the next few years
for public works prompted by the EU provides the as they could become the basis to forge collective
opportunity to introduce cutting-edge techniques choices and mould new, different, non-standard-
in the field of evaluation for public choice. These ized scenarios of public interest. The in-depth,
techniques are capable of indirectly generating multifaceted nature of this kind of evaluation,
innovation both in institutional administrative and the freedom with which it is conducted,
management and the enhancement of people’s definable as civic evaluation, extols all forms of
citizenship. The underlying consideration that sup- pluralistic democracy as it enables the members
ports the methodology we propose in this paper is of a community to give voice to their fears, needs,
oriented in this direction. We are aware that today and desires that otherwise would have remained
the infrastructure sector is deemed to be closely unknown. It enables them to compare differing
linked to the dynamics of national economy. This and dissonant convictions, opinions that were
belief is substantiated firstly by the fact that infra- latent or potential until revealed. It makes it possi-
structural projects considered strategic for the ble to achieve adhesion and consensus, and lastly,
development and modernisation of the nation have to reach a shared view of alternative solutions
been included in Italy’s national economic program- to collective problems. What becomes important
ming, and secondly by the fact that along with the from a political and cultural viewpoint is that a

766

TIS2017_Book.indb 766 2/8/2017 8:09:40 PM

 civic evaluation helps to bridge the increasingly Andersen, I.-E. & Jaeger, B., 1999. Scenario workshops
wider and more frequent gap between the prefer- and consensus conferences: towards more democratic
ences stated by citizens and the decisions taken decision-making. Science and Public Policy, 26(5),
by their institutional representatives. It becomes pp. 331–340.
Dienel, P.C., 2002. Die Planungszelle: der Bürger als
the tool to implement the social, cultural and eco- Chance, Wiesbaden: Westdeutscher Verlag.
nomic changes that are taking place, because if Fishkin, J.S. et al., 2014. Deliberating across Deep
the self-determination of citizens is disregarded, Divides. Political Studies, 62(1), pp. 116–135.
the very foundations of liberty and democracy Fishkin, J.S., 1991. Democracy and Deliberation: New
risk being distorted. Directions for Democratic Reform. Canadian Journal
Many factors in decision-making methods may of Political Science, 26(3), pp. 596–597.
hamper the general use of inclusive evaluation Miccoli, S., Finucci, F. & Murro, R., 2015a. A direct delib-
techniques. They include the pride and arrogance erative evaluation method to choose a project for Via
of certain politicians and bureaucrats who are Giulia, Rome. Pollack Periodica, 10(1), pp. 143–153.
Miccoli, S., Finucci, F. & Murro, R., 2014a. A Monetary
inclined to disparage the popular will expressed by Measure of Inclusive Goods: The Concept of Deliber-
majority votes; the fact that more time and money ative Appraisal in the Context of Urban Agriculture.
is required for deliberative evaluations; the reluc- Sustainability, 6(12), pp. 9007–9026.
tance to launch an experimentation program of Miccoli, S., Finucci, F. & Murro, R., 2013. La valutazi-
these techniques, a precondition for their efficient one partecipata per la valorizzazione del paesaggio.
and habitual application. However, most of these Evaluation-Sharing in Landscape Enhancement. In
obstacles could be removed simply by increas- M. Crescimanno, L. Casini, & A. Galati, eds. Evoluzi-
ing social culture on the most advanced forms of one dei valori fondiari e politiche agricole. Bologna:
democracy. MEDIMOND S R L, pp. 149–157.
Miccoli, S., Finucci, F. & Murro, R., 2015b. Measuring
Shared Social Appreciation of Community Goods:
An Experiment for the East Elevated Expressway of
REFERENCES Rome. Sustainability, 7(12), pp. 15194–15218.
Miccoli, S., Finucci, F. & Murro, R., 2014b. Social
Abelson, J. et al., 2003. Deliberations about deliberative Evaluation Approaches in Landscape Projects.
methods: issues in the design and evaluation of pub- Sustainability, 6(11), pp. 7906–7920.
lic participation processes. Social Science & Medicine, Smith, G. & Wales, C., 2000. Citizens’ Juries and Deli-
57(2), pp. 239–251. berative Democracy. Political Studies, 48, pp. 51–65.

767

TIS2017_Book.indb 767 2/8/2017 8:09:40 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Automatic calibration of microscopic simulation models

for the analysis of urban intersections

Sara Oliveira & Luís Vasconcelos

CITTA and Department of Civil Engineering, Polytechnic Institute of Viseu, Campus de Repeses, Viseu, Portugal

Ana Bastos Silva

CITTA and Department of Civil Engineering, University of Coimbra, Polo II, Coimbra, Portugal

ABSTRACT: Microscopic simulation models are being increasingly used in traffic engineering
applications, but various issues concerning the extent to which its outputs reproduce field data still need
to be addressed. In this perspective a proper calibration of the model parameters has to be performed so
as to obtain a close match between the simulated and the actual traffic measurements. This paper aims to
highlight the importance of calibration process, as the adjustment stage of the microsimulation models’
parameters, applied to the analysis of urban, at-level, intersections. After the selection of the case studies
(one roundabout and one intersection with traffic signal control), field observations were made, allowing
the creation of a database that supported both the models’ development and calibration. The next phase
focused on the application of an Aimsun microscopic simulation model to the selected intersections. This
involved the development of an optimization based calibration methodology, coupled with a sensitivity
analysis. The optimization framework was implemented in Matlab using the pre-defined genetic algo-
rithm in the optimization extension. With the models properly calibrated and validated, the performance
indicators were obtained and conclusions about their approximation to reality were drawn. The recom-
mended calibration methodology easily allows the replication of the observed conditions, revealing how-
ever poor adaptation to other intersections and a general lack of representativeness. The results obtained
by the genetic algorithm are very sensitive to small changes in the initial set of parameters. The calibration
methodology allows the replication of the observed conditions, revealing however poor adaptation to
other intersections and a general lack of representativeness.

Keywords: Microsimulation; Roundabout; Signalized Intersections; Performance indicators; Capacity;

Aimsun; Sensitivity analysis; Genetic algorithm

1 INTRODUCTION such as car-following, gap-acceptance and

lane-changing. In the literature several studies can
Microscopic simulation models emulate realistically be found addressing this topic but most of them
the traffic flow on a road network, which makes are focused in uninterrupted traffic, namely in
them appropriate tools for evaluating complex freeways (e.g. Punzo e Tripodi, 2007; Rakha et al.
traffic facilities and control strategies. Simulation 2007; Vasconcelos et al. 2014b).
models include components related to the infra-
structure, the demand, and associated behavioural
1.1 Objectives
models, which have complex data requirements
and numerous model parameters. Calibration, as The main objective of this research is to develop
the adjustment of model parameters such that the a sensitivity analysis to identify the role of several
model’s output closely represents field conditions, parameters in the simulation of urban intersec-
is vital for appropriate decision making process. tions, and then develop and apply an automated
A particular problem is the accurate model- calibration procedure to estimate optimal values
ling and calibration of unsignalized intersections, for the most relevant parameters. The Aimsun
including roundabouts or traffic light controlled microsimulation package (Casas et al. 2010) was
intersections. This is not a trivial task since sev- used in this application, however the methodology
eral behavioural microscopic models are involved, is transversal to similar simulation models.

769

TIS2017_Book.indb 769 2/8/2017 8:09:41 PM

 1.2 Procedure
This research followed the following analysis
procedure:
1. Selection of the urban at-level intersections and
data collection based on field observations;
2. Selection of the most relevant parameters
through a previous selection, followed by a sen-
sitivity analysis, involving a comprehension of
the Aimsun sub-models, such as car-following,
gap-acceptance and lane-changing;
3. Calibration of the selected intersections mod-
els using an optimization process by applying a
Genetic Algorithm (GA), selection of perform- Figure 2. Casa do Sal, signalized roundabout (east entry).
ance indicators and analysis of the results.
Table 1. Summary of the vehicle counts at the rounda-
bout (121 minutes).
2 DATA COLLECTION Entry lane Circulatory lane

Two intersections were studied: a roundabout, Left Right Inner Outer

located in Viseu, and a signalized intersection,
located in Coimbra, both in Portugal. The two- 683 1290 1274 534
lane west entry of the roundabout, “Nelas Round-
about” (Fig. 1) was selected for the study due to the
long queues that can be observed during the peak Table 2. Summary of the vehicle counts at the signal-
ized intersection (120 minutes, 28 cycles, green time: 33 s,
periods. In the off-peak, the traffic is fluid with yellow: 3 s).
very few stops. This roundabout has an inscribed
circle diameter of 57 m and two circulatory lanes. Approach lane
The traffic light approach selected for this study
is the east entry to the “Casa do Sal” junction, a Period Left Middle Right Total
roundabout that is now operating with traffic
lights (Fig. 2). The approach has three lanes and Green 404 400 388 1192
operates under heavy demand levels during most Yellow/Red 50 42 47 139
part of the morning and evening. Total 454 442 435 1331
The data collection was based on video record-
ings, and the data treatment was supported by the
vehicles (at approximately 450 m before the entry),
use of LUT|VP3 software (Vasconcelos et al. 2013).
at the entry (yield line) and in the circulatory ring,
The application log contains a list of records, each
in front of the entry. In the signalized intersection,
corresponding to a vehicle, with the corresponding
vehicles were recorded, cycle by cycle, also during
time stamp and a code that indicates the respective
120 minutes, as they passed the stop line during the
lane. In the roundabouts the vehicles were observed
green, yellow and red periods. A summary of vehi-
during 121 minutes in three sections: approaching
cle counts is presented in the tables.

3 SENSIVITY ANALYSIS

A sensitivity analysis was performed with three

main objectives: a) to provide a better insight
about the influence of the different parameters in
the simulation results; b) to facilitate the selection
of parameters to calibrate; c) to reduce the range
of values to use in the optimization procedure.
In a first approach, and based on previous stud-
ies (Vasconcelos et al. 2014a; Vasconcelos et al.
2014b), eight parameters were selected. Five of
these parameters are defined at the vehicle type
Figure 1. Nelas roundabout (and the selected west entry). level assuming a normal truncated distribution of

770

TIS2017_Book.indb 770 2/8/2017 8:09:41 PM

 values. This way, the user must provide, for each
parameter, the mean, the standard variation, the
minimum and maximum values. These parameters
are (Aimsun, 2012):
− Maximum acceleration, a (m/s2) – is the maxi-
mum acceleration a vehicle can achieve under
any circumstances. This parameter is required by
the Gipps car-following model but is also used
by the gap-acceptance and lane-change models.
A high value means that a vehicle can achieve
the desired speed faster and is able to change
lanes and cross traffic streams using short inter- Figure 3. Roundabout simplified model.
vals, thus resulting in a higher throughput;
− Normal deceleration, b (m/s2) – is the maximum
deceleration that a vehicle can use under normal
driving conditions;
− Minimum spacing between vehicles, s (m) – is
the distance that a vehicle keeps between itself
and the preceding vehicle when stopped. A vehi-
cle with a short spacing can seize short intervals
and contributes to reduce the queue lengths;
− Give way time, gwt (s) – this parameter is used
when a vehicle is in a give-way situation (e.g.: Figure 4. Traffic light approach simplified model.
lane changing or yield/stop sign in a junction).
When a vehicle has been at a standstill for more
than gwt (in seconds), it will reduce the accept- The main characteristics of each model are, con-
ance margins; cerning the roundabout: single lane approach to a
− Speed acceptance, SA – the speed acceptance roundabout with only a circulatory lane with pri-
parameter can be interpreted as the acceptance ority flow in the circulatory lane: 500 veh./h. For
level of maximum legal speed section (vmax), the traffic light approach the characteristics are:
i.e. the desired speed the driver will be vmax⋅SA, one single lane with a cycle of 90 s, 30 s of green
which is achieved if the geometric and operating and 3 s of yellow.
conditions permit. The next step consisted of evaluating the sen-
sitivity of all parameters for the two models. The
The remaining three parameters are usually
intersection capacities were measured for different
defined at the experiment level (same value for all
values of each parameter, maintaining the default
vehicles) and reflect the reaction time of drivers
value on the remaining ones. The results are pre-
under three different situations:
sented as the ratio between the current capac-
− Reaction time, rt (s) – is the time it takes a ity and the reference capacity (obtained when all
driver to react to speed changes in the preceding parameters take the default value) – Q/Qref.
vehicle;
− Reaction time at stop, rts (s) – is the time it takes
3.1 Roundabout
for a stopped vehicle to react to the acceleration
of the vehicle in front; With all the default values, the roundabout
− Reaction time at traffic light, rttl (s) – is the time approach has a capacity of 948 veh./h. The ratio
it takes for the first vehicle stopped after a traf- between the actual and the reference capacity for
fic light to react to the traffic light changing to different parameter values is shown in Figure 5.
green. The graphs also indicate the default value for the
cars (circle marker) and truck (box marker). The
The sensitivity analysis is based on two simple
following conclusions can be drawn:
models (Figs. 3–4) that reflect the base conditions
of the case studies: a roundabout and a traffic − The maximum acceleration, the minimum spac-
light junction. To get a better understanding of ing and the reaction time have a considerable
the way each parameter influences the simulation effect on the capacity in the whole range of
results, the driver-vehicle population was assumed tested values;
uniform, only with cars. In both cases a very high − The effect of the normal deceleration and give-
demand value at the approach lane was set (5000 way time is very low when its value is near
veh./h) to force a permanent queue at the entry. the Aimsun defaults (cars and truck) but is

771

TIS2017_Book.indb 771 2/8/2017 8:09:42 PM

 Figure 5. Sensitivity analysis for the traffic light approach and roundabout.

significant for values bellow 2 m/s2 (decelera- 3.2 Traffic light approach
tion) and 10 s (give-way time);
With all the default values, the traffic light
− The reaction time at stop only has a significant
approach has a capacity of 751 veh./h. The ratio
effect when its value exceeds the reaction time;
between the actual and the reference capacity for
− For the simulated conflicting flow (500 veh./h),
different parameter values is presented in Figure 5,
it was not possible to clarify the influence of
from where the following conclusions can be
parameter speed acceptance due to the exist-
drawn:
ence of continuous queue at the entrance that
prevents drivers from reaching its desired speed, − The maximum acceleration has a very signifi-
and there is therefore one significant variability cant effect on the capacity, particularly for val-
in the graph corresponds. ues bellow 4 m/s2;

772

TIS2017_Book.indb 772 2/8/2017 8:09:42 PM

 − Both the normal deceleration and the give-way observed during the whole data collection period, it
time do not have any effect on the capacity, simply indicates the difference between the observed
whatever the value tested; and simulated throughput (or entry capacity).
− Increasing the minimum distance between vehi- To measure the goodness of fit at the rounda-
cles tends to decrease the capacity; bout, as well as the signalized approach, the RMSE
− It was not possible to classify the influence of statistic was used. Its definition is:
parameter speed acceptance factor due to the
∑ (M )
existence of continuous queue at the entrance n 2
obs t M sim
sim t
that prevents drivers from reaching its desired RMSE = t =1
(2)
speed; n
− The capacity decreases for higher values of the
reaction time and reaction time at stop; The optimization framework was implemented
− The reaction time at traffic light has a very sub- in Matlab using the built-in genetic algorithm tool.
tle effect on the capacity. The algorithm starts by generating an initial popu-
lation (20 individuals, each of which corresponds
Considering the modest role of the normal to a set of 6 parameters in the roundabout and 4
deceleration and the reaction time at traffic light in parameters in the signalized junction).
the model, it was decided to exclude these param- For each individual, a Python script modifies
eters from the automated calibration process. the corresponding parameters in Aimsun, simu-
Likewise, the give way time parameter was used lates the model in console mode and compares the
only in the roundabout model. observation and simulation outputs to compute the
fitness value. The default Aimsun parameters that
4 CALIBRATION PROCEDURE were subject to optimization (for the roundabout
entry and traffic light approach) are presented on
The automated calibration procedure takes the form the following table.
of an optimization problem based on a Genetic When all individuals are evaluated, the GA gen-
Algorithm (GA). This is a widely used technique erates a new population: besides elite children, who
that employs ideas from natural evolution to find correspond to the individuals in the current gener-
good solutions for combinational parametric ation with the best fitness values, the algorithm cre-
optimization problems. ates crossover children by selecting vector entries
Building on the results of the sensitivity analysis, from a pair of individuals, and mutation children,
the parameter calibration problem was formulated by applying random changes to a single individual
in the following optimization framework: (MathWorks 2005).
After 73 generations and approximately 2 hours
min
a,s,GWT,RT,R
T TSF,SA
(
f M obs ,M sim ) of computing time, the algorithm reached a con-
vergence condition and returned the optimal solu-
s.a. 1 < a < 5 m s2 , tion for the roundabout analysis (see Fig. 6). The
0,5 < s < 5 , optimal parameters obtained are presented in the
0 < GWT
W < 30 s , (1) following table.
0,5 < RT <1,5 s , The effectiveness of the optimization method
1 < RTSF < 3, to fit the simulated time-series of densities to the
0,9 < SA <1,7 observation is presented on Figure 7, which high-
lights the proximity between the density obtained
where rtsf is the ratio between the reaction time at in the field observations with the density simulated
stop and the reaction time (rtsf = rts/rt) and f is with the Aimsun default parameters and with the
the goodness of fit function that measures the dis- optimal parameters obtained.
tance between the observed and simulated meas- It is noticeable that the difference between the
urements, Mobs and Msim. default and optimal parameters has some discrep-
This function varies according to the intersec- ancy, revealing that the optimization process does
tion: in the roundabout, we were particularly inter- not always returns values that can be considered
ested in understanding how Aimsun was able to as realistic; however the solution of all the optimal
predict the variation of the density (and indirectly
the queue lengths) during the simulation period and
so the vector M was defined as the vehicle density Table 3. Default Aimsun parameters.
between the upstream and downstream sections at
the approach (veh./km), at each one minute obser- a (m/s2) s (m) gwt (s) rt (s) rts SA
vation—simulation period i. At the signalized inter-
3 1 10 0,75 1,20 1,1
section, and given that the permanent queues were

773

TIS2017_Book.indb 773 2/8/2017 8:09:43 PM

 Figure 6. Convergence of the algorithm (roundabout entry analysis).

Figure 7. Time-series of densities (observations, simulation with default and optimal parameters) for the roundabout
entry.

Table 4. Optimal solution for the roundabout. Table 5. Optimal solution for the signalized approach.

a (m/s2) s (m) gwt (s) rt (s) rtsf SA RMSE a (m/s2) s (m) rt (s) rtsf RMSE

2,433 0,937 2,041 0,581 1,584 0,977 4,172 2,340 0,940 0,994 1,669 0,008

parameters obtained combined leads to a fitness min

a , s , RT , RTSF
(
f M obs , M sim
i )
value that approximates in a very reasonable form s.a. 1 < a < 5 m s 2,
the indicators in analysis (in the case of the round- 0, 5 < s < 5 m, (3)
about: density). 0, 5 < RT < 1, 5 s ,
In the signalized intersection, the objective was 1 S 3
to understand if Aimsun was able to predict the
intersection capacity. The optimization resulted in a RMSE of 0,008,
Using the same optimization procedure, with indicating that those parameters reproduced accu-
the following indicated optimization framework, rately the entry capacity. The optimal solution
the parameters were obtained (Table 5). obtained is presented in the following table.

774

TIS2017_Book.indb 774 2/8/2017 8:09:43 PM

 The value adjustment indicator shows a great fit (eventually using instrumented vehicles) and to
between the field observation data values and those generalize the results to a population of drivers.
obtained after calibration. This is an analysis “sim- It would also be relevant to apply the optimi-
pler” than the one performed for the roundabout zation procedure at intersections with different
entrance involving a smaller number of parameters, layouts, confirming the applicability of the param-
where the approximation of the approach capacity, eters for a particular type of intersection to other
as the number of vehicles that entry the intersec- intersections of the same type.
tion per cycle can be considered excellent, again
demonstrating the importance of the optimization
process applied and validating its results. The value REFERENCES
of the capacity obtained by optimization for the
approach was 47,54 veh./cycle, coinciding with the Aimsun, 2012. Aimsun Version 7 User’s manual, TSS -
resulting value of the field observations. Transport Simulation Systems, Barcelona, Spain.
Casas, J., Ferrer, J., Garcia, D., Perarnau, J., Torday, A.,
2010. Traffic Simulation with Aimsun, in: Barceló,
5 CONCLUSIONS J. (Ed.), Fundamentals of Traffic Simulation, 145.
Springer New York, pp. 173–232.
MathWorks, I., 2005. Genetic Algorithm and Direct
For the two cases studied, roundabout entry and Search Toolbox for Use with MATLAB®: User’s
traffic light approach, the calibration procedure Guide. MathWorks.
based on optimization is validated and adequate to Punzo, V., Tripodi, A., 2007. Steady-State Solutions and
study specific urban intersections. The proximity Multiclass Calibration of Gipps Microscopic Traffic
of the calibrated simulated models to the observed Flow Model. Transportation Research Record: Journal
data is significantly better than those without any of the Transportation Research Board, 1999, pp. 104–
parameter calibration, demonstrating the impor- 114, doi:10.3141/1999-12.
tance of the sensitivity analysis and posterior Rakha, H., Pecker, C., Cybis, H., 2007. Calibration Proce-
parameter calibration. However, the optimiza- dure for Gipps Car-Following Model. Transportation
Research Record: Journal of the Transportation Research
tion may lead to unphysical parameter values and Board, 1999, pp. 115–127, doi:10.3141/1999-13.
therefore it is recommended to define acceptable Vasconcelos, A.L.P., Seco, Á.J.M., Bastos Silva, A.,
variation ranges for each parameter. 2013. A comparison of procedures to estimate
Simulation models must be carefully applied to critical headways at roundabouts. PROMET—
the analysis of intersections, since its outputs are Traffic & Transportation, 25(1), pp. 43–53.
strongly related with the quality of inputs (e.g. traf- Vasconcelos, L., Neto, L., Santos, S., Bastos Silva, A.,
fic demand, the geometric design coding) and have Seco, Á., 2014a. Calibration of the Gipps
high sensitivity to the values of the behavioural car-following model using trajectory data. Transporta-
parameters adopted. tion Research Procedia, 3, pp. 952–961, doi:10.1016/j.
trpro.2014.10.075.
This way, future works should focus on the Vasconcelos, L., Seco, Á., Silva, A.B., 2014b. Hybrid Cal-
application of automated calibration procedures ibration of Microscopic Simulation Models, in: Sousa,
that can be applied to large networks and on the J.F., Rossi, R. (Eds.), Computer-based Modelling and
identification of simple methodologies that allow Optimization in Transportation, 262. Springer Interna-
to quantify some of the most important parameters tional Publishing, pp. 307–320.

775

TIS2017_Book.indb 775 2/8/2017 8:09:44 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

On-street parking management: Bridging the gap between

theory and practice

C. Piccioni, M. Valtorta & A. Musso

DICEA—Department of Civil, Building and Environmental Engineering, “Sapienza” University of Rome,
Rome, Italy

ABSTRACT: More and more cities in Europe are adopting parking schemes targeted at improving
urban accessibility, without increasing car dependency and the related negative externalities. However,
possible links between such parking systems and contexts within they work are generally not investigated.
This paper summarizes the findings of a study aimed at exploring, from a quantitative viewpoint, the gap
between theory and practice also showing that, at practical level too, there is not yet a common under-
standing in selecting specific parking measures. Starting from an in-depth analysis of parking policies
implemented by a sample of European cities, a set of indicators, specially defined for this analysis, were
calculated. The statistical variability of the related independent variables was also examined, in order to
find probable cause-effect links between local context features and selected measures as well as to provide
a first methodological approach to explore the effectiveness of the applied parking strategies.

1 INTRODUCTION More and more cities in Europe are adopting

parking schemes targeted at improving accessibil-
Parking in urban environments is increasingly ity to urban spaces, while limiting car dependency
becoming a crucial issue. Cities have to cope with and the related negative externalities. Neverthe-
high rates of parking demand but the on-street less, the maturity level of what experienced is quite
parking provision appears very often inadequate. uneven and measures are sometime applied at very
Such a supply likely results as a component of local level. Besides, possible links between parking
built environments designed many years ago, at a schemes and contexts within they work are gener-
time when ownership and use of cars had not yet ally not recognized at all. Finally, there is no objec-
reached present values and parking was not con- tive evidence of which measures are preferred with
sidered a land-consuming activity. respect to other ones.
Awareness that land is a precious and scarce To the Authors’ knowledge there are no appli-
resource, especially in high-density environments, cations explicitly interpreting, from a quantitative
has gradually grown over time; thus the main static viewpoint, parking management strategies applied
and dynamics activities featuring mobility patterns, at European level. To this end, such a paper aims
and their impacts on city users, are being revised at contributing to the current debate, by providing
in light of this concept (Musso & Piccioni 2012). a methodological approach through which inves-
On-street parking really provides direct accessibil- tigate and critically discuss whether and to what
ity to services/activities but shortage of accessible extent the strategies put in place differ from the
car spaces or, even more, their inefficient use (i.e. theoretical framework supporting parking man-
low turnover) can negatively affect local business agement policies.
also decreasing the livability of urban setting.
On-street parking affects street capacity and road
2 THEORY BEHIND PARKING
users’ safety; maneuvering vehicles as well as ille-
MANAGEMENT: ACCESSIBILITY
gally parked cars can create disruptions to traffic
AND ROLE OF INDICATORS
circulation, by increasing pollutant emissions. On-
street parking also reduce public transport supply
2.1 Framing parking management evolution
and performances; it actually can occupy portions
of carriageway that otherwise would be used by Many studies and researches have developed differ-
specific traffic components (e.g. reserved bus lanes, ent parking management approaches, over the last
cycling or walking paths). decades, in order to meet the changing dynamics

777

TIS2017_Book.indb 777 2/8/2017 8:09:46 PM

 affecting urban mobility and sustainability (Shoup of varying types of trips and activities and most
2005; Litman 2008, 2013; Bates & Leibling 2012, likely varies across people according to their tastes
Guo et al. 2012). Such analyses allowing marking and preferences (Dong et al. 2006). By focusing on
how the paradigm shift is progressing, by passing private individual transport is almost immediate
from a “predict and provide” traditional approach understanding the influence of parking on accessi-
(Owens 1995; Vigar 2001) up to a more contempo- bility in urban areas. The ease with which a driver
rary and comprehensive strategy focused on a set is able to reach a particular activity affects not only
of management levers (Litman 2013, 2016). the actual travel time but also the time required to
The former approach can be thought as bor- find an available car lot as well as the monetary cost
rowed by the transport planning theories explaining associated with its use (also including the payment
the concept of induced traffic demand (Næss et al., ways). Accordingly, the accessibility concept can be
2014). By transferring such a concept to parking declined both in physical and economic terms.
(i.e. induced parking demand), the predict and pro- Physical accessibility, as practical availability
vide approach sets wider minimum requirements of parking spaces, directly affects probability of
aimed at providing as much parking lots as pos- finding an available car spot as well as time spent
sible for free, or with the lowest price, in order to for parking. In congested areas such time is likely
offer parking options easily accessible at every des- related to the “cruising for parking” (Shoup, 1997)
tinations. According to the main reference literature phenomenon, affecting those drivers who, once
(Acutt & Dodgson 1997, Shoup & Manville 2005, realized that parking is at capacity in the destina-
Chatman 2008) the availability of parking at destina- tion zone, continue to drive at low speed all the time
tions seems to be a key issue influencing the users’ required to locate an available spot. In the most
choice by increasing car usage frequency and mile- critical cases, drivers leave the area without being
age. A generous parking provision also affects land able to park their cars and then, unable to carry
use planning in a long-term perspective by increasing out the activities for which they were moved.
risk to use valuable land improperly while encourag- Economic accessibility is directly linked to the
ing urban sprawl (Litman 2008). The latter focuses parking fee and, therefore, to the monetary trans-
on the improved efficiency of parking facilities, by port cost. Over the years, an increasing number of
applying a sort of “predict and prevent” approach Municipalities is implementing parking policies, by
(Owens 1995) without increasing minimum require- using just price as a management lever for achiev-
ments. This implies to rethink planning of parking ing a better equilibrium between parking supply
supply, with a special attention to the existing one, by and demand. A key target is to improve the supply
adopting pricing and/or regulations, creating a user’s efficiency, by increasing the related turnover index
information system as well as by improving overall (i.e. the number of cars served by a single spot dur-
walkable conditions in the parking lot catchment ing the reference period, generally calculated on a
areas (Litman 2016b). Some evidences stress how daily basis), limiting at the same time the related
synergy resulting by the application of a set of park- negative externalities.
ing measures allows to achieve increasing benefits,
because total impacts are greater than the sum of
2.3 Building accessibility indicators
their single ones (Litman 2016a). Such a paradigm
shift represents thus a sort of watershed: it changes According to the purposes of this study, the indi-
the way to conceive the “parking problem” and, so, cator is intended as a simplified measure, usually
the way to search for more proper parking solutions. expressed in quantitative form, composed by one
or more variables, able to describe in a plausible
and reproducible way the trend of a phenomenon
2.2 Evaluating accessibility for parking systems
or changes affecting the system to which this refers
Several definitions can be traced to the accessibil- (Musso & Piccioni 2010). Scientific literature sug-
ity concept. Among them, that proposed over time gests several indicators performing accessibility
by Dalvi (1976) as “the ease with which any land- measures, most of which can be traced to 2 macro-
use activity can be reached from a location using classes such as: (1) indicators based on the utility and
a particular transport system”, by Burns (1979) as (2) indicators based on opportunities. The former
“the freedom of individual to decide whether to type of accessibility indicators, generally pertaining
participate or not in different activities” as well as the gravity and random utility models. The latter
by the U.S. Department of Environment (1996) as type is used in isochronic measures, representing the
“the ease and convenience of access to spatially dis- cumulative count of opportunities reachable within
tributed opportunities with a choice of travel”, all a given travel time or distance (Cervero 2005) as well
well suited to the purpose of the present study. Nev- as in the opportunities perception measures (Cas-
ertheless, the quantitative assessment of such “ease cetta et al. 2013). Such measures, however, gener-
and convenience” is quite complex as it is a function ally return aggregated values not consistent with

778

TIS2017_Book.indb 778 2/8/2017 8:09:46 PM

 the purpose this research. Thus, specific ex novo socio-economic and transport aspects—increases
indicators suitable for the study have been defined, the descriptiveness of this analysis, which thus
in order to recognize which elements featuring the does not exhaust its usefulness to the only charac-
physical and/or economic accessibility can likely be terization of large cities.
related to parking management practices. An extensive technical review of databases and
statistical information (Urban Audit by Eurostat,
Knoema Dataset, Official websites of parking
3 METHODOLOGY operators and Municipalities) were carried out,
in order to reach relevant data on cities sample
3.1 Methodological approach according to the physical (e.g. size and spatial dis-
tribution of on-street, off-street and P&R) and
This research was structured according to a
economic (e.g. park ing rates, pricing approaches)
main pillar, consisting of a comparative analysis
parking supply dimensions. A preliminary outlook
involving a cities sample, aimed at characterizing
of collected data shows how on-street paid parking
accessibility from a quantitative viewpoint. More
supply is extremely variable. Moreover, on-street
properly, the methodological approach was
supply is always higher by at least an order of mag-
developed by performing a set of subsequent
nitude compared to the off-street parking.
macro-activities such as: (1) cities selection and
In central urban areas, parking price is presuma-
data collection; (2) accessibility indicators design;
bly higher than that applied in peripheral zones, so
and (3) statistical analysis of indicators.
special emphasis was given to the maximum hourly
Such steps paved the way for the further analy-
rate of on-street parking and its related indicators.
sis of correlation between indicators thus allowing
The minimum on-street parking fees is between
applying the Principal Components Analysis (PCA),
€ 0.24/h (Middlesbrough) and 2€/h (Vienna), the
through which to explore the gap between the theoret-
maximum hourly rate ranges from 1.20€/h (Rome)
ical approach and the practical one in implementing
to 5€/h (Amsterdam). Except for Middlesbrough
parking management strategies. Finally, a clustering
covering the last place in the ranking, Rome applies
process was performed in order to recognize cit-
the lower rate of the whole cities sample. Naples is
ies having an attitude towards one or more specific
the most expensive, the applied parking rate fol-
parking strategies as well as atypical contexts.
lows an incremental logic based both on zoning
and occupancy time: 1 hour costs 2€ but from
3.2 Data collection the second onwards, the unit price is increased by
25%. In France, e.g. in the central area of Lyon and
The study focused on 24 cities belonging to 13
Nantes, by increasing the occupancy time by 50%,
European countries (Fig. 1). It is worth mentioning
the parking fee rises by almost 150%. The rationale
that the heterogeneity of the selected urban envi-
of such pricing policy, widespread in many Euro-
ronments—in terms of spatial extent, urban form,
pean cities, is aimed at increasing the turnover rate,
thus improving the efficiency of the overall parking
system. The economic dimension of the P&R sup-
ply generally refers to the cost incurred by regular
users of Public Transport (PT). 61% of the sample
cites encourage P&R use, offering the service for
free; the maximum cost is recorded in Barcelona
(daily price is about 10€/8 hours).

3.3 Accessibility indicators

22 ad hoc indicators were defined and classified
according to dimension and typology of parking
supply (on-street, off-street, P&R). Concerning the
physical aspects, a set of descriptive indicators was
designed in order to investigate overall supply in
relation to the main features of the selected cities
(Table 1).
With reference to the economic accessibility,
the selected indicators allow featuring the parking
offer according to the disposable income (Gross
Domestic Products) and two cost items relating to
Figure 1. Framework of sample cities. the use of private car and PT (Table 2).

779

TIS2017_Book.indb 779 2/8/2017 8:09:46 PM

 Table 1. Physical parking supply indicators. Table 2. Economic parking supply indicators.

Descrip- Code Indicator* Descriptiveness

Code Indicator Unit tiveness
EION** 1 On-street max. hourly City disposal
PION* 1 Pay parking Lots/1000 Potential parking fee/GDP per income
lots/1000 inhabitants Parking capita
inhabitants demand EION2 On-street max. hourly PT cost
PION 2 Pay parking Lots/100 cars Parking parking fee/PT
lots/100 demand ticket
registered cars EION3 On-street max. hourly Car use cost
PION 3 Pay parking % Spatial parking fee/Fuel
area/Urban features price
area EION4 On-street max. hourly Pricing zoning
PIOF **1 Parking Lots/1000 Parking parking fee/On-
lots/1000 inhabitants demand street min. hourly
inhabitants parking fee
PIOF 2 Parking lots/100 Lots/100 cars Parking EIOF***1 Off-street average Parking
registered cars demand hourly parking fee/ demand
PIOF 3 Parking lots/Pay Lots/lots On-street GDP per capita
off-street parking EIOF2 Off-street average PT cost
parking lots hourly parking fee/
PIPR***1 P&R lots/1000 Lots/1000 Parking PT ticket
inhabitants inhabitants demand EIOF3 Off-street average Car use cost
PIPR 2 P&R lots/100 Lots/100 cars Parking hourly parking fee/
registered cars demand Fuel price
PIPR 3 P&R lots/Urban Lots/km Urban rail EIOF4 On-street max. hourly On-street
rail transit system parking fee/Off- parking
PIPR 4 P&R lots/Pay Lots/lots On-street street average hourly
on-street parking parking fee
parking lots EIPR****1 P&R average daily Context
parking fee/GDP prosperity
*PION Physical Indicator On-street, per capita
**PIOF Physical Indicator Off-street, EIPR2 P&R average daily PT cost
*** PIPR Physical Indicator Park & Ride. parking fee/PT
ticket
EIPR3 P&R average daily Car use cost
Existence of different parking fees in the same parking fee/Fuel
city has been also considered by calculating respec- price
tively EIOF4, EIPR4 as well as EION4. Such an EIPR4 P&R average daily On-street
indicator is also useful to detect the propensity fee/On-street max. parking
hourly parking fee
of Municipalities to manage parking demand by
using a sort of zoning approach consistent with *Such indicators are dimensionless as computed by
the attractive potential of the different city areas. the ratio of two items having the same measurement
Summarizing, a high number of indicators was unit (€),
intentionally chosen to do not miss interesting **EION Economic Indicator On-street,
information on the phenomenon for the purposes ***EIOF Economic Indicator Off-street,
of the next PCA. **** EIPR Economic Indicator Park & Ride.

3.4 Statistical analysis of indicators

for infrastructure, service and maintenance. Barce-
As far as physical accessibility indicators is con- lona, Brescia, Lisbon and Zurich recorded a ratio
cerned, PION2 returns an average value of 9 lots/100 higher than 1, showing a tendency to limit land
cars, which equals 11 cars/lot. The maximum value consumption caused by parking.
(28 lots/100 cars) is recorded by Copenhagen fol- In terms of economic accessibility, Naples, among
lowed by Bologna (18 lots/100 cars) and Pisa (12 all, is the city applying the most expensive fee com-
lots/car). The 2 Italian cities have higher PION1: pared to its disposable income. Rome and Zurich,
Bologna reports the value of 96 lots/1000 inhab- respectively, in the last and second to last position,
itants followed by Pisa (75 lots/1000 inhabitants). offer a very affordable rate compared to their real
PIOF3 highlights the attitude of Municipalities purchasing power. On the sample basis, the average
for the on-street parking type, thus shrinking costs parking hourly cost exceeds 1.6 times cost of a PT

780

TIS2017_Book.indb 780 2/8/2017 8:09:46 PM

 ticket. Birmingham, Leeds and Antwerp confirm Table 3. Physical accessibility: statistical variables.
to use the economic leverage to restrict car access
to the city center, e.g. in Leeds the hourly rate is 4.6 Sam- Std Max/
times higher than 1 bus ticket cost. In Italy, Naples Code ple Mean dev. CV Min Max Min
and Brescia offer the highest hourly rates, about PION1 18 38.6 21.3 0.6 11.9 96.2 8.1
2 times the ticket cost. A total of 19 cities record PION2 18 9.0 6.1 0.7 3.5 27.9 7.9
a value higher than 1 (i.e. 1 hour of parking costs PION3 18 1.8 × 73.2 × 0.7 33.1 × 29.9 × 9.0
more than 1 bus/metro ticket); Vienna offers both 10−5 10−5 10−5 10−5
services at the same cost; as opposed in Monaco PIOF1 15 27.9 27.5 0.98 2.2 87.5 40.4
(0.95), Rome (0.80) and Zurich (0.77) 1 hour is PIOF2 15 6.4 6.6 1.0 0.4 23.8 58.6
cheaper than a PT ticket. EION4 shows certain PIOF3 14 0.8 0.8 0.97 0.04 2.6 64.5
homogeneity between pricing policies implemented PIPR1 17 9.6 8.1 0.8 1.5 28.3 18.4
in British cities where there is a high rates difference PIPR2 17 2.0 1.6 0.8 0.4 5.4 12.5
(average value is close to 5). Besides, Copenhagen PIPR3 14 173.9 211.8 1.2 10.9 771.95 70.8
(9.7), Antwerp (7) and Amsterdam (5.5) show high PIPR4 13 0.3 0.2 0.7 0.1 0.83 7.9
rate variability. The off-street average rate records a
cost equal to the on-street maximum rate; Zurich,
Barcelona, Munich and Lyon report a value lower Table 4. Economic accessibility: statistical variables.
than 1. In cities where such data is available, off-
street parking is competitive when compared to on- Std Max/
Code Sample Mean dev. CV Min Max Min
street parking. With regard to P&R rates, two types
of cities were identified, depending on which they EION1 24 0.00 0.00 0.38 0.00 0.00 4.45
offer free or pay parking. 7 cities belong to the sec- EION2 23 1.58 0.57 0.36 0.77 2.73 3.56
ond group, among them Barcelona offers the most EION3 24 1.59 0.53 0.33 0.67 2.75 4.11
expensive fare (daily cost is about 5 times the PT EION4 24 3.39 2.02 0.60 1.00 9.73 9.73
ticket), even if a monthly subscription is available. EIOF1 15 0.00 0.00 0.41 0.00 0.00 3.97
EIOF2 15 1.22 0.60 0.50 0.53 3.14 5.97
3.5 Data scattering and heterogeneity of parking EIOF3 15 1.20 0.50 0.42 0.56 2.12 3.79
strategies EIOF4 15 1.48 0.61 0.41 0.68 2.72 4.00
EIPR1 18 0.55 1.02 1.87 0.00 3.98 *
In order to increase the descriptiveness of the
EIPR2 18 0.72 1.33 1.84 0.00 4.98 *
accessibility indicators, the main statistical vari-
EIPR3 18 0.95 1.90 1.99 0.00 7.03 *
ables were calculated. As known, mean describes
EIPR4 18 0.55 1.02 1.87 0.00 3.98 *
the sample through a single value, which is much
more representative as the sample has low data *Denominator = 0.
scattering. In the study context, a low data disper-
sion indicates a good homogeneity level of parking
vice versa a physical supply expansion, where
management approaches among different cities. In
allowed, requires a long-term perspective;
contrast, high dispersion ratios show lack of com-
− reliability and accuracy of available data. Data
mon practices or shared approaches; in such a case,
on physical parking supply are not always col-
the sample mean does not lend itself to represent
lected and/or updated by single Municipalities;
the sample. Thus, data scattering was estimated
− different approaches. In some cities, parking
by the Coefficient of Variation (CV) it allowing
management is performed by ignoring possible
comparing elements having different measurement
changes occurred in the functional mix/attrac-
units. According to the above accessibility indica-
tive potential of certain zones. In doing so, park-
tors, the main statistical variables are summarized
ing supply remains static without looking for a
in Table 3 and Table 4.
balance between on-street and off-street proxim-
The economic indicators (CV = 0.42) showed a
ity supply.
data dispersion minor than that calculated for the
physical ones (CV = 0.82). Among the economic indicators, those related
Such result shows how there was a greater con- to P&R recorded the highest data scattering; such
sistency in implementing similar pricing policies inhomogeneity can be likely attributed to:
(pricing lever is a recurrent element) if compared
− diversity of the proposed parking rates (e.g.
to approach managing physical supply. The higher
daily, monthly, zooning-related subscriptions,
heterogeneity of physical/spatial-related variables
etc.);
values can be likely traced to:
− different function assigned to P&R (data
− a greater ability in managing price-related varia- includes both on/off-street lots or only P&R
bles. Pricing plans can be changed in short term; supply for road-rail interchange nodes);

781

TIS2017_Book.indb 781 2/8/2017 8:09:47 PM

 − potential transport side effect. Really, P&R (PC) in case this is capable to explain a greater pro-
allows a reduction in the private cars use for portion of the total variance of that explained by
trips to the central areas; however, recent empiri- a single variable.
cal studies (Clayton et al. 2015, Noel 1988) stress Accordingly, 6 aggregate indicators were identi-
how a wide and free P&R supply incentives car fied, one for each group of indicators:
use for trips between the origin and the inter- − on-street (CPON), off-street (CPOF) and P&R
change node, by shifting congestion from the (CPPR) physical components;
city center to peripheral areas. A perception of − on-street (CEON), off-street (CEOF) and P&R
economic effects by Municipalities also plays (CEPR) economic components.
an important role, because they may fear losing
revenues from parking located in central areas Table 5 summarizes for each component, its
through introduction of P&R (Dijk & Montalvo eigenvalue, the accounted variance percentage
2011). and the equation expressing the PC based on
the original variables belonging to the respective
It is appropriate to note that for the other eco- group.
nomic indicators descriptive of on-street compo- CPON, CPOF, CPPR express the attitude (i.e.
nent, data are concentrated in relative small-scale supply-oriented approach) of Municipalities in
ranges, showing a certain spread of recurring facilitating physical accessibility to urban areas
practices. by private car. Vice versa CEON, CEOF, CEPR
quantify their propensity (demand-oriented
approach) in managing the potential demand by
4 CONTEXTUALIZING PARKING using pricing, thus reducing economic accessibil-
MANAGEMENT PRACTICES: THE ity. The accounted variance for all components
PROPOSED APPROACH exceeds 60% of the total variance, the minimum
value refers to the P&R physical component (61%)
4.1 Principal Component Analysis and the maximum one (99%) to the P&R economic
The proposed accessibility indicators provides a component. So, almost the totality of the data var-
good characterization of parking management iance is explained by a single component and the
strategies, according to the features of the single loading (equation coefficients) recorded the same
cities; however, just as they are do not allow a clear value (0.50) for each of the four (secondary) indi-
and concise picture of the main European trends cators, thus contributing with the same weight to
and attitudes. For this reason, also because of the the component structure. Such a result underline
correlation between some indicators, the Principal that CEPR does not depend on the links between
Components Analysis (PCA) was applied. This parking measures and the context variables, but
technique enables the explanation of new latent likely on the political choice to provide P&R car
factors as a linear combination of the original vari- spaces for free.
ables, limiting loss of information and maximizing
the variability explained through a limited number
of components. Table 5. PCA relevant outcomes.
The application of the PCA has requested 2
Eigen- Accounted
preparatory steps. The former involved the missing Code value variance Equation
data management, so for cities reporting only one
missing data by parking type, an unconditional CPON 1.92 63% CPON = 0,67PION1 +
mean imputation was adopted; cities with two or 0,52PION2 + 0,45PION3
more missing data were excluded from the sample CPOF 2.8 93% CPOF = 0,59PIOF1 +
for that specific parking category. The latter step 0,58PIOF2 + 0,56PIOF3
concerned the standardization of collected data, CPPR 2.37 61% CPPR = 0,59PIPR1 +
needed for managing variables having different 0,59PIPR2 + 0,31 PIPR3 +
unit of measurements; to this end the following 0,44 PIPR4
equation was used: Z = (X − μ)/σ, where Z = stand- CEON 2.57 64% CEON = 0,55EION1 +
ardized variable; X = original variable; μ = sample 0,42EION2 + 0,53EION3 +
mean, σ = standard deviation. 0,43EION4
PCA was applied to each of the 6 groups of indi- CEOF 2.5 62% CEOF = 0,48EIOF1 +
cators (PION, PIOF, PIPR, EION, EIOF, EIPR) 0,42EIOF2 + 0,56EIOF3 –
0,52EIOF4
and the components selection was made by using
CEPR 3.97 99% CEPR = 0,50RIPR1 +
the “Kaiser criterion”, then choosing only the 0,50EIP2 + 0,50EIPR3 +
components with eigenvalues greater than 1. This 0,50EIPR4
implies to consider only a Principal Component

782

TIS2017_Book.indb 782 2/8/2017 8:09:47 PM

 The estimated values for the 6 main components the sample shows a low or very low propensity
show the propensity of Bologna, Barcelona and in supporting off-street physical supply, although
Brescia—recording the higher values for CPON, compared to on-street type increases (+12%) cities
CPOF, CPPR—to ensure high physical accessibil- showing a medium or high propensity. Pricing poli-
ity to urban areas by private car through provi- cies supporting off-street parking are quite hetero-
sion of on-street (Bologna), off-street (Barcelona) geneous: 47% of cities prefer to keep low economic
and P&R (Brescia) car spaces. On the contrary, accessibility; other 40% provides service at a rela-
Birmingham, by recording the highest value for tively low price. With reference to the P&R physical
CEON, denotes a high propensity to shrink on- supply, 47% of cities show a very low propensity,
street parking demand. CEOF and CEPR refer the thus preferring to restrain availability of parking
higher values to Barcelona, where low economic spaces; vice versa 35% shows an attitude to ensure
accessibility (high fee) may reflect need to cover a high level of physical accessibility. The majority
high investment/operating costs. of the sample (61%) also provides service for free
In order to identify possible trends at Euro- by highlighting a policy encouraging use of P&R
pean scale, 5 levels (very low, low, medium, high system that is a very low propensity to disregard
and very high) related to Municipalities’ propen- P&R as part of the total urban parking supply.
sity in implementing parking policies were defined
(Fig. 2).
4.2 Cluster Analysis
78% of cities indicate a low or very low propen-
sity to improve on-street physical parking supply. A Cluster Analysis (CA) was carried out, aimed at
This figure manifests a clear policy to reduce cars identifying groups of cities applying similar park-
dependency by constraining physical parking acces- ing management practices and possible factors
sibility to private vehicles. The on-street economic affecting the decision-making choice on parking
component showed a low (34%) and medium (29%) measures. A hierarchical clustering aggregative
propensity to manage supply by using pricing for methodology was applied, by measuring clusters
controlling demand; besides a fairly high percent- similarity through the “complete-link proxim-
age (21%) amounted to very high values. 66% of ity” function: D(X;Y) = maxx∈X,y∈Y d(x, y), where
D(X;Y) = distance between cluster X and Y,
d (x, y) = distance between point x belonging to the
X set and point y belonging to the Y set. So, this
function calculates the Euclidean distance between
two specific clusters as the maximum distance
between two elements belonging to them.
Dendogram of Figure 3 shows outcomes of CA;
by choosing a similarity threshold value of 66.7,
suitable for achieving a representative and signifi-
cant amount of homogeneous groups, 6 clusters
were identified.
First cluster includes Naples, Antwerp and
Edinburgh that, although different in demo-
graphic, economic and transport terms, are all

Figure 2. Parking management attitude of the sample

cities. Figure 3. City clustering.

783

TIS2017_Book.indb 783 2/8/2017 8:09:47 PM

 characterized by a low propensity to meet demand Third cluster, composed by Barcelona, Berlin,
(low physical accessibility) and a high propensity to Pisa and Turin, record values close to the sample
manage it by pricing. In line with theoretical guide- mean both for physical and economic component.
lines, such a practice aims at reducing car use in This indicates a willingness to find a better balance
urban areas through control measures both supply between supply and demand by using pricing lev-
and demand-oriented, that is by increasing direct erage.4 Fourth cluster, including medium-size cit-
monetary costs for road users. Regardless of the ies (Brescia, Strasbourg and Nantes) shows a high
level of physical accessibility, Northern European propensity for managing and limiting on-street
cities (Antwerp, Edinburgh, Amsterdam, Copen- parking demand (very low value for CPON and
hagen) and even more the British ones (Leeds, Bir- medium for CEON). Brescia integrates on-street
mingham, Middlesbrough) show high propensity parking with a wide off-street and P&R supply,
in using pricing policies. The cost of parking is not both characterized by a high affordability (low
very convenient both when compared to the degree values for CEOF and CEPR).5 The French cities
of richness of the context whether it towards the address parking demand mainly towards P&R
cost of PT. This suggests that attitude to control facilities, by offering free parking. Finally, Copen-
demand by pricing tool does not depend on the hagen and Bologna belong to the last two clusters,
demographic/economic features, but mainly by both presenting a significant physical accessibility
cultural and social factors. (respectively high and very high values for CPON).
Second cluster includes Rome, Lyon, Munich, Both cities still turn away from the conventional
Zurich and Lisbon, all performing low physical paradigm by acting on price to manage on-street
accessibility and high economic accessibility that parking (CEON is medium for Bologna, very high
is few on-street parking lots offered at low price. for Copenhagen).
Such cities are all highly urbanized in central areas,
with consequent high functional mix also attract-
ing commuters and visitors from their surround- 5 CONCLUSIONS
ing metropolitan areas. The clear unfeasibility
to increase physical accessibility by building new The study provides an analysis of contemporary
on-street parking has been addressed in the sin- approaches to parking management implemented
gle realities through different practices. Lisbon at European level, by interpreting them accord-
and Zurich,1 e.g. compensate lack of on-street car ing to some main parameters (evident and latent)
spaces by offering extensive off-street supply at a affecting performances of the urban parking sys-
relatively low price (respectively a high and aver- tem. Findings confirm a decline of the pure con-
age value for CPOF was recorded). The resulting ventional paradigm albeit some cities have not
increase in public land availability has been used completely abandoned the typical “predict and
to encourage use of more sustainable transport provide” approach.
modes.2 Munich, given the high extension of its Pay parking is a common practice with varying
urban rail network (174 km, it is the 3rd high- prices depending on demo/socio-economic aspects
est sample value), increased physical P&R supply and on the vision pursued by single Municipali-
(very high value for CPPR) also offering affordable ties. Especially the Northern European cities show
service to PT users (low value for CEPR). Rome a clear willingness to implement a “car limiting”
and Lyon do not show particular strategies to policy, by emphasizing use of parking pricing, also
compensate their low physical accessibility. Rome ad hoc for specific valuable areas. There is also an
is characterized by a low propensity to control on- increasing propensity in encouraging use of off-
street parking demand (very low value for CEON) street parking. In cities dealing with a conventional
and the related high economic parking accessi- approach, the off-street typology is used to expand
bility (low price and a single fee rate in all zone), parking supply in built environments, where land
favors the motorized mobility, thus contributing to is a scarcer resource. As opposed, “greener” cit-
congestion.3 ies aim at redeveloping part of the urban space
by replacing on street parking lots with off-street
ones, keeping constant the urban parking capacity.
1
In Zurich this results from the “Historic parking com-
promise”, implemented since 1996 aimed at shifting car
4
parking from road to parking facilities. Barcelona integrates the on-street parking supply with
2
Zurich records highest cycle paths density (1.8 km/ a generous off-street provision (the highest value in the
km2). sample) to a higher but still competitive price if com-
3
Rome, with the highest motorization rate (734 cars/1000 pared to on-street.
5
people), is the 2nd congested city: on average a trip takes Brescia is the first, among the Italian cities, for density
38% of more time when compared with a free-flow of cycle paths (1.3 km/km2) and the fourth of sample as
condition. a whole.

784

TIS2017_Book.indb 784 2/8/2017 8:09:48 PM

 Besides, there is a common understanding in Dijk, M., & Montalvo, C. 2011. Policy frames of Park-
ensuring high affordability for P&R supply (less and-Ride in Europe. Journal of Transport Geography,
common in cities covered by wide urban rail net- 19:1106–1119.
work) although this may reduce PT’s ridership in Dong, X., Ben-Akiva, M. E., Bowman J. L. and Walker,
J. L. 2006. Moving from trip-based to activity-based
peripheral city areas. measures of accessibility Transportation Research Part
A lesson learned from this study suggests need A: Policy and Practice, 40 ( 2):163–180.
to identify common guidelines, also proposing Eurostat. Cities (Urban Audit). http://ec.europa.eu/euro-
“light” measures, as follows: stat/ web/cities. 25 February 2016.
Knoema. World Data Atlas. https://knoema.com/atlas. 18
− to adjust parking fees also according to PT serv- February 2016.
ice coverage and its real performances, mainly in Litman T. 2008. Recommendations for Improving LEED
car-oriented cities; Transportation and Parking Credits, Victoria Trans-
− to improve efficiency of urban parking systems port Policy Institute.
Litman, T. 2013. Parking Management: Strategies, Evalua-
by encouraging on-street parking turnover, also tion and Planning. Victoria Transport Policy Institute.
defining upper standards limits in central areas; Litman, T. 2016a. Parking Management—Comprehensive
− to increase P&R functionality, since it is designed Implementation Guide. Victoria Transport Policy
to improve urban intermodality without replac- Institute.
ing road–based PT. Litman, T. 2016b. Land Use Impacts on Transportation,
How Land Use Factors Affect Travel Behavior. Victoria
Transport Policy Institute.
REFERENCES Lois, D. & Lopes-Saez, M. 2009. The Relationship
Between Instrumental, Symbolic and Affective Factors
Acutt, M. & Dodgson, J. 1997. Controlling the Environ- as Predictors of Car Use: A Structural Equation Mod-
mental Impacts of Transport: Matching Instruments to elling Approach, Transportation Research Part A 43.
Objectives. Transportation Research Part D: Transport Musso, A. & Piccioni, C. 2010. Lezioni di Teoria dei
and Environment 2(1): 17–33. Sistemi di Trasporto. Edizioni Ingegneria 2000, Roma.
Barter, P. 2014. On-street parking Management, Musso, A. & Piccioni, C. 2012. Tourist Coach Mobility
Sustainable Urban Transport Technical Document. Plans: current practices and operational criteria, in
Federal Ministry for Economic Cooperation and Cappelli, A. Libardo, A & Nocera S. (edited by) Envi-
Development. ronment, land use and transportation systems, Franco
Burns, L. D. 1979. Transportation, Temporal and Angeli: pp.145–160.
Spatial components of accessibility. Lexington Book, Næss P., Andersen J, Nicolaisen M. S., Strand A. 2014.
Toronto. Transport modelling in the context of the ‘predict and
Cascetta, E., Cartenì, A., Montanino M. 2013. A New provide’ paradigm. European Journal of Transport and
Measure of Accessibility based on Perceived Oppor- Infrastructure Research, Issue 14(2): pp. 102–121.
tunities. Procedia—Social and Behavioral Sciences, 87: Noel, E. 1988. Park and Ride: Alive, Well, and Expanding
117–132. in the United States. Journal of Urban Planning and
Cervero, R., 2005. Accessible Cities and Regions: A Frame- Development, 114:1(2), 2–13.
work for Sustainable Transport and Urbanism in the Owens, S. 1995. From ‘predict and provide’ to ‘predict
21st Century, Working Paper UCB-ITS-VWP-2005-3, and prevent’?: pricing and planning in transport policy.
UC Berkeley Center for Future Urban Transport. Transport Policy, 2 (1): 43–49.
Clayton, W., Parkhurst, G., Ben-Elia, E. & Ricci, M. 2015. Shoup, D. 2005. The high cost of free parking. APA
Where to park? A behavioral comparison of bus Park Planners Press.
& Ride and city center car park usage in Bath, UK. in Shoup, D. & Manville, M. 2005. Parking, People and
17th EPA Congress, Berlin, Germany. Cities. Journal of Urban Planning and Development.
Chatman, D. 2008. Deconstructing Development Den- U.S. Department of Environment. 1996. Policy and
sity: Quality, Quantity and Price Effects on Household Procedure Guidelines, Planning policy guidance: Town
Non-work Travel, Transportation Research Part A 42. centres and retail developments, Revised PPG6.
Dalvi, M. & Martin, K. 1976. The measurement of acces- Vigar, G., 2001. Reappraising UK transport policy 1950–
sibility: some preliminary results. Transportation, 1999: the myth of ‘monomodality’ and the nature of
5:17–42. ‘paradigm shifts’. Planning Perspectives 16, 269–291.

785

TIS2017_Book.indb 785 2/8/2017 8:09:48 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Personality and driver behaviour questionnaire: Correlational

exploratory study

J.F. Dourado
CITTA, Faculty of Science and Technologies of the University of Coimbra, Coimbra, Portugal

A.T. Pereira & V. Nogueira

Department of Psychological Medicine, Faculty of Medicine of the University of Coimbra, Coimbra, Portugal

A.M.C. Bastos Silva & A.J.M. Seco

CITTA, Faculty of Science and Technologies of the University of Coimbra, Coimbra, Portugal

ABSTRACT: The aim of our research is to explore the association between personality traits and
dimensions of driver behaviour in a vast sample of the Portuguese population.
A community sample composed of 747 participants [417 (55.8%) women; mean age = 42.13 ± 12.349 years;
mean; mean driving license years = 21.30 ± 11.338; mean years of regular driving = 20.33 ± 11.328]
participated in an online survey. To evaluate personality traits the NEO-Five Factor Inventory-20 and
the Impulsive Sensation Seeking scale were used. The Driver Behaviour Questionnaire composed of 24
items, assessing Infractions and Aggressive Driving (IAD), Non-Intentional Errors (NIE) and Lapses,
was used.
IAD significantly and moderately correlated with age, gender (male), years of driving licence, Impulsivity
and Sensation seeking; NIE and Lapses with Impulsivity. Other personality traits and socio-demographic
variables presented significant but lower correlations with driving behaviour dimensions.
These results reinforce that personality traits should be accounted when studying driver behaviour for
road safety assessment.

1 INTRODUCTION drivers drive to analyse the type of actions they

take every moment. Additionally, Hennessy (2011)
Road trauma can be seen as one of the most signifi- argues that any discussion of factors that impact
cant diseases of the industrialized societies and it is driving outcomes should include personal factors.
an increasing public health and economic issue in Following this line of thought, traffic psychol-
developing countries. Organisation for Economic ogy researchers have been studding personal traits
Co-operation and Development reports that every that are believed to have positive or negative cor-
year approximately 1.24 million people die on the relation with aggressive and/or risky driver behav-
roads. In Portugal, the National Authority for iours. They include: sensation seeking, having a
Road Safety reported 30.604 road traffic accidents positive correlation with risky behaviours (Yang
with injuries in 2014, with a total of 482 deaths, a et al., 2013); anger with positive correlation to vio-
number that may increase in a near future. lations (Lajunen et al., 1998), to loss of concen-
Driving can be defined as a “complex process tration (Dahlen et al., 2005) and to poor vehicle
involving individual factors, expressed within a control (Deffenbacher et al., 2001); altruism with
social exchange among drivers, passengers, and negative correlation to aggressive driving (Dahlen
pedestrians, which is ultimately impacted by con- et al., 2006; Yang et al., 2013); and less consci-
textual and environmental stimuli found inside and entiousness that can be related to normlessness
outside the vehicle” (Hennessy, 2011). behaviour, which has been proven to be related
Oppenheim and Shina (2011) describe the driver to violations and risky behaviour (Iversen and
behaviour as what the drivers do given their limi- Rundm, 2002; Oltedal and Rundmo, 2006; Yang
tations, constraints, needs, motivation, level of et al., 2013).
alertness and personality. They also emphasize the Based on his taxonomy of human error Reason
importance of considering the context in which (1990) developed an Accident Causation Model,

787

TIS2017_Book.indb 787 2/8/2017 8:09:50 PM

 further developed by (Verschuur and Hurts, 2007), and an excellent adjustment by the confirmatory
as well as a Driver Behaviour Questionnaire (Reason factorial analysis.
et al., 1990) based on self-report questions to measure To assess Impulsivity and Sensation seeking,
behaviours, the Manchester Driver Behaviour Ques- which have long been associated with a tendency
tionnaire, to serve as tool for driver behaviour analy- to engage in behaviours without consideration
sis. Previous studies have demonstrated the validity of potential negative consequences (Caspi et al.,
of the DBQ (Reason et al., 1990; Parker et al., 2000; 1997), the principal researcher in collaboration
Lajunen et al., 2004), and it has been amply used by with a psychiatrist and a psychologist (both expe-
researchers, with Wahlberg et al. (2011) mentioning rienced in personality research and in the transla-
fifty-four studies that have included at least part of tion, adaptation and validation of psychological
the questionnaire is their studies. Yet, the relation- assessment instruments) translated the 19-item
ship between the results of the self-report scales and Impulsive Sensation Seeking Scale (Zuckerman,
objective driving data of the behaviours in question 1994). The ImpSS is part of the larger ZKPQ
(e.g. speeding, close following, abrupt lane-changing, (Zuckerman et al., 1993), showing an acceptable
etc.), need more validation (Helmand and Reed, internal consistency (alphas > 0.70) in heterogene-
2014), as for the relationship between the scales and ous samples (Mc Daniel & Mahan III, 2008; De
crashes (Wahlberg et al., 2011). Leo et al., 2010). This scale is composed of two
The aim of this study is to explore the associa- dimensions, Impulsivity (eight items) and Sen-
tion between personality traits and dimensions of sation seeking (11 items), in binary item’ score
driver behaviour in a vast sample of the Portu- “true” = 1 or “false” = 0.
guese population. This is part of a bigger research The driver behaviour evaluation was made using
project which aims to analyse these relationships the Portuguese translation of the Driver Behaviour
with real driving behaviours. Questionnaire (Reason et al., 1990; Reimer et al.,
2005; Correia, 2014) which consists of 24 items
describing a variety of errors and violations dur-
2 METHODS
ing driving. The questionnaire divides the items in
three dimensions—errors, lapses and violations—
2.1 Procedure
eight items each, on a 6-point scale, ranging from 0
A community sample was recruited and invited to to 5, where higher scores indicate more frequency
participate in an online survey on the relationship in engaging risky driving behaviours.
between personality and driver behaviour. Inclu-
sion criteria were: driving license and regular driv-
2.3 Sample characterization
ing for at least three years and age lower than 75
years old. The total sample comprised 747 participants [417
All the participants answered self-reported (55.8%) women; mean age = 42.13 ± 12.349 years;
questionnaires, containing socio-demographic mean driving license years = 21.30 ± 11.338; mean
questions, personal accident history, and validated years of regular driving = 20.33 ± 11.328].
instruments to evaluate personality traits and
driver behaviour.
2.4 Statistical analysis
To verify if these self-reported questionnaires were
2.2 Instruments
valid and reliable tools to evaluate the dimensions
To evaluate the personality traits the researchers in which the researchers were interested in, the
used the Big Five model (McCrae & John, 1992), in analysis started with the study of the psychometric
order to conciliate a global overview of the person- properties of the three instruments.
ality and the ability to differentiate specific traits. The total sample was randomly divided in two
To assess the Big Five personality traits (Neuroti- sub-samples, being sample A composed of 373 and
cism, Extraversion, Openness, Agreeableness and sample B of 374 participants, who did not signifi-
Conscientiousness), the Portuguese version of the cantly differ in relation to socio-demographic and
NEO-Five Factor Inventory-20-item (Bertoquini accident history variables (Qui-square tests and
& Ribeiro, 2006) was used. This scale divides the t-Student tests, with p-values > 0.05). Sample A
items in five dimensions, four items each, on a was used to Exploratory Factor Analysis and sam-
5-level Likert scale. It is a short version of the orig- ple B was used to Confirmatory Factor Analysis
inal NEO-PI-R (Costa & McCrae, 1992), validated (Figure 1).
for the Portuguese population by Bertoquini & SPSS version 23.0, StatsToDo and AMOS soft-
Pais Ribeiro (2006). The NEO-FFI-20 has shown a ware were used to Exploratory Factor Analysis,
clear factorial structure and an acceptable internal Parallel Analysis and Confirmatory Factor Analy-
consistency (alphas ≥ 0.7) for the five dimensions sis respectively.

788

TIS2017_Book.indb 788 2/8/2017 8:09:51 PM

 Following the exploratory analysis, the Con-
firmatory Factorial Analysis (CFA) was per-
formed (Kline, 2010). The fitness of the previously
obtained factorial dimensions from each scale to
the data from the sample B was assessed. A series
of indexes were calculated for the different possi-
ble dimensional structures for each scale, includ-
ing the original factorial structures adopted by
the authors of the scales. From the analysis and
comparison of the results of the CFA, the most
fitted models were chosen and then their internal
consistency was assessed using the total sample
(n = 747) and finally the structures for each scale
were established. Details from this analysis will not
be presented in this article, only the final factorial
dimensions. These factorial dimensions were con-
sidered as variables in the present study.
The dimensions’ internal consistency was meas-
ured by Cronbach’s coefficient (α). High values of
α indicate consistent and reliable measures (0.65
≤ α <; 0.70, acceptable; 0.70 ≤ α < 0.80, good; 0.80 ≤
α < 0.90, very good; DeVellis, 1991). To analyse the
ability of a single item to measure the attribute sup-
posed to be assessed by the scale/factor dimension,
Pearson correlations coefficients between each item
and the total score (excluding the item) were exam-
ined. In other words, this coefficient allows to inter-
pret the magnitude of the correlations between the
Figure 1. Statistical analysis methodology.
item and its dimension and/or total scale. The Val-
ues of Pearson’s Correlation Coefficient r ≥ 0.20
The Exploratory Factor Analysis was per- were considered to be acceptable, r ≥ 0.30 to be
formed using the principal components method, good and r ≥ 0.5 to be high (Cohen, 1992).
with varimax rotation (Pereira et al., 2013). The Also, the Cronbach’s alphas excluding each item
suitability of data for factor analysis was assessed were computed and compared with the coefficient
with the Kaiser-Meyer-Oklin (KMO) and the α of the total score. When this individual coeffi-
Barlett’s Tests of Sphericity (Kline, 1994). The cient α was lower than the coefficient α of the total
data was considered suitable for factorial analysis score, it meant that the item was contributing to
when KMO > 0.50 (Sharma, 1996) and the null the internal consistency of the scale, as its absence
hypothesis (H0) was rejected (p-value ≤ 0.05). To would decrease the total reliability of the scale.
help establishing the correct number of factors Finally, the Pearson correlation coefficients
to extract from the factorial analysis, the criteria between personality traits, the dimensions of the
used were: i) the Kaiser criteria, to retain fac- driver behaviour, and socio-demographic and per-
tors with eigenvalue >1 (Kaiser, 1958); ii) Cattel sonal accident history variables were analysed.
Scree Plot criteria, which implies the retention
of all components in the sharp descent part of
the plot before the eigenvalues start to level off, 3 RESULTS
where line changes slope (Cattel, 1966; Kline,
2000). The selection of the items for each factor 3.1 NEO-FFI-20
consisted of retaining items that showed strong
After performing the statistical analysis described
factor loadings. As other authors have done (Elal
in the chapter 2.4 and analysing in a qualitative way
et al., 2000), items with factor loadings >0.30 were
all the items, it was concluded that the best factorial
chosen.
structure for the analysed sample was the original
Using the Parallel Analysis, with the Monte
structure, excluding two items (14 and 16).
Carlo Simulation, the number of significant fac-
The factors were:
tors for each scale was confirmed (Ledesma &
Valero-Mora, 2007). Even though, different possi- Factor 1 - Neuroticism—tendency to experience
ble dimensional structures were obtained for each such feelings as anxiety, anger, envy, guilt, and
scale. depressed mood.

789

TIS2017_Book.indb 789 2/8/2017 8:09:51 PM

 Factor 2 - Extraversion—to enjoy human interac- The factors were:
tions and to be enthusiastic, talkative, assertive,
Factor 1 - Sensation Seeking—defined by the
and gregarious.
search for experiences and feelings that are
Factor 3 - Openness to experience—reflects the
novel and intense, and by the readiness to take
degree of intellectual curiosity, creativity and a
physical, social, legal, and financial risks for the
preference for novelty and variety.
sake of such experiences.
Factor 4 - Agreeableness—tendency to be compas-
Factor 2 - Impulsivity—a tendency to act on a
sionate and cooperative rather than suspicious
whim, displaying behaviour characterized by
and antagonistic towards others.
little or no forethought, reflection or considera-
Factor 5 - Conscientiousness—a tendency to
tion of the consequences.
be organized and trustworthy, show self-
discipline, act dutifully, aim for achievement, Both factors revealed to have good internal
and prefer planned rather than spontaneous consistency (0.70 ≤ α < 0.80) and the total score
behaviour. presented very good internal consistency with
a coefficient α = 0.82 (Table 2). The correlation
Each factor revealed to have an acceptable
strength between each item and the total score
and good internal consistency (0.60 ≤ α < 0.80)
of its respective dimension (excluding the item)
(Table 1). Pearson correlation coefficients between
showed to be good and very good, being the mini-
each item and the dimensional score of its respec-
mum the item 15, r = 0.329. Concerning correla-
tive dimension (excluding the item) ranged from
tion coefficients between each item and the total
0.359 (item 6—feeling helpless many times and
score (excluding the item), they ranged from 0.305
wishing someone to solve his/her problems) to 0.619
(item 2 – usually tending to think before taking an
(item 8—feeling great emotions when reading a
action) to r = 0.539 (item 14 – sometimes doing
poem and observing a work of art), with nine items
crazy things just for fun), having 13 items good
having very good correlations and eleven good
correlations and two items very good correlations.
correlations. Cronbach’s alphas excluding the
When analysing the total scale, Cronbach’s alphas
item, except for item 6, were all equal or smaller
excluding the item were all smaller than the alpha
than the alpha from the respective dimension. The
of the total score (0.82).
increase of the alpha when eliminating the item 6,
from the dimension Neuroticism, was very slight.
Additionally, eliminating this item would make the 3.3 Driver Behaviour Questionnaire (DBQ)
dimension too small, and consequently inducing
The best factorial structure for the DBQ was
the decrease of its internal consistency. So one
obtained as a result of the exploratory factor anal-
decided to maintain this item.
ysis, being composed of three factors, excluding
items 1 and 24.
3.2 Impulsive Sensation Seeking (ImpSS) The factors were:
Following the same methodological approach, the Factor 1 - Infractions and aggressive driving—
authors concluded that the best factorial structure deliberate deviations from practices believed
for the analysed sample was the two factors’ struc- necessary to maintain safe driving.
ture, resulting from the exploratory factor analysis, Factor 2 - Non intentional errors—to fail planned
excluding items 4, 6, 7 and 10. actions, which may result in dangerous
outcomes.
Factor - Lapses—failures related to distractibility
which can cause embarrassment but unlikely to
Table 1. Dimensions of NEO-FFI-20 and its internal have an impact on driving safely.
consistency (Cronbach’s α).
All factors revealed to have good internal con-
Internal sistency (0.70 < α ≤ 0.80) and the total scale a very
Number Consistency
Dimensions of items (α) Table 2. Dimensions of ImpSS and its internal consist-
ency (Cronbach’s α).
NEO- Neuroticism 3 (excluded 0.68
FFI- item 16) Internal
20 Extraversion 4 0.62 Number Consistency
Openness 4 0.74 Dimensions of items (α)
Agreeableness 3 (excluded 0.70
item 14) ImpSS Sensation seeking 10 0.79 0.82
Conscientiousness 4 0.74 Impulsivity 5 0.76

790

TIS2017_Book.indb 790 2/8/2017 8:09:51 PM

 Table 3. Dimensions of DBQ and its internal consist- Table 4. Pearson Correlation Coefficient (r) between social
ency (Cronbach’s α). demographic, personality traits, driver behaviour dimen-
sions and accidents per year of driving licence variable.
Number Internal
Dimensions of items Cons. (α) Accid/
driving
DBQ Impulsive and 7 0.77 0.84 IAD NIE Lapses DBQ lic year
aggressive driving
Non intentional errors 9 0.73 IAD r 0.214
Lapses 6 0.71 sig. <0.001
N 747
NIE r 0.431 0.221
good internal consistency (α > 0.80) (Table 3). sig. <0.001 <0.001
Pearson correlation coefficients between each N 747 747
item and the dimensional score of its respective Lapses r 0.326 0.540 0.176
dimension (excluding the item) ranged from 0.315 sig. <0.001 <0.001 <0.001
(item 2) to 0.639 (item 12). When analysing cor- N 747 747 747
relation coefficients between each item and the DBQ r 0.792 0.802 0.764 0.260
total score (excluding the item) they were all higher sig. <0.001 <0.001 <0.001 <0.001
than r = 0.311 (item 7—forgetting where the car is N 747 747 747 747
parked), The Cronbach’s alphas excluding the item Age r −0.239 −0.050 −0.092 −0.176 −0.211
were all lower than the alpha of the total scale sig. <0.001 0.173 0.012 <0.001 <0.001
(0.84). N 747 747 747 747 747
Gender r 0.254 0.014 −0.133 0.079 0.046
sig. <0.001 0.697 <0.001 0.030 0.743
3.4 Correlations N 747 747 747 747 747
Pearson correlation coefficients were estimated Education r 0.16 0.110 0.168 0.188 0.012
between socio-demographic variables, personal- sig. <0.001 0.003 <0.001 <0.001 0.743
ity traits and the DBQ dimensions—Infractions N 747 747 747 747 747
and aggressive driving, Non intentional errors and Years of r −0.174 −0.036 −0.072 −0.130
Lapses (Table 4). driving sig. <0.001 0.323 0.048 <0.001
license N
The strongest significant correlations (r > 0.2 747 747 747 747
and p-value < 0.05) with the dimension Infrac- Years of r −0.166 −0.044 −0.080 −0.132
tions and Aggressive Driving (IAD) were: age frequent sig. <0.001 0.225 0.029 <0.001
driving
(r = −0.239), gender (r = 0.254, male gender with N 747 747 747 747
higher scores on IAD), Impulsivity (r = 0.259), Neuro- r 0.092 0.188 0.166 0.182 0.058
Sensation seeking (r = 0.301) and the ImpSS total ticism sig. 0.012 <0.001 <0.001 <0.001 0.112
scale (r = 0.337). Years of driving license, Extra- N 747 747 747 747 747
version and Agreeableness showed also significant Extra- r 0.138 −0.038 −0.071 0.027 −0.001
but lower correlations with this driving behaviour version sig. <0.001 0.301 0.052 0.462 0.984
dimension. N 747 747 747 747 747
As for the dimension Non Intentional Errors Openness r −0.017 −0.015 0.113 0.031 −0.013
to experi- sig. 0.636 0.686 .002 0.400 0.731
(NIE), the variable with the strongest correlation
ence N 747 747 747 747 747
was Impulsivity (r = 0.212). Conscientiousness,
Neuroticism and the ImpSS total scale correlated Agreeable- r −0.142 −0.038 0.010 −0.082 −0.017
with lower strength, but also significantly. ness sig. <0.001 0.296 0.779 0.026 0.640
The Lapses variable presented stronger correla- N 747 747 747 747 747
tion with Impulsivity (r = 0.227, p-value < 0.001). Conscien- r −0.026 −0.188 −0.157 −0.146 −0.071
This dimension also presented correlations with ciousness sig. 0.474 <0.001 <0.001 <0.001 0.053
Neuroticism, Conscientiousness and the total N 747 747 747 747 747
ImpSS scale, but with lower strength. Impulsivity r 0.259 0.212 0.227 0.299 0.064
The total DBQ score correlated significantly sig. <0.001 <0.001 <0.001 <0.001 0.081
to Impulsivity (r = 0.299), Sensation Seeking N 747 747 747 747 747
(r = 0.246) and ImpSS (r = 0.312). Sensation r 0.301 0.115 0.131 0.246 0.047
The variable accidents per driving license
seeking sig. <0.001 0.002 <0.001 <0.001 0.202
years (in which the respondent had been identi- N 747 747 747 747 747
fied as responsible for the accident) correlated ImpSS r 0.337 0.177 0.195 0.312 0.062
significantly with IAD (r = 0.214), NIE (r = 0.221), sig. <0.001 <0.001 <0.001 <0.001 0.089
DQB (r = 0.260), and age (r = −0.211). N 747 747 747 747 747

791

TIS2017_Book.indb 791 2/8/2017 8:09:51 PM

 4 DISCUSSION AND CONCLUSION The results also showed that people with lower
scores in the Conscientiousness trait, that tend to
Personality traits are of paramount importance be less cautious and focused, scored higher in NIE
for traffic psychology, since they are believed to and Lapses.
have positive or negative correlation with aggres- For the behaviour Lapses younger people and
sive and/or risky driver behaviours. It is also of women got the higher scores. People with higher
great relevance to understand which factors and scores in Neuroticism, Openness to experience
how they influence the driver behaviour in order and Impulsivity and with lower scores in Consci-
to be taken into account when analysing the traffic entiousness had a significant relation to failures
safety conditions of a given infrastructure. related to distractibility, but that are unlikely to
With this research the authors aimed to explore have an impact on driving safety.
the association between personality traits and It was also interesting to found a gender pat-
dimensions of driver behaviour in a vast sample of tern in the correlations. Being female correlated
the Portuguese population (N = 747), contributing with Lapses and male with IAD. In future studies
for a clearer understanding of the role of personal- it will be useful to analyse if personality traits still
ity in this specific context. explain the DBQ levels after controlling for impor-
Exploratory and confirmatory factorial analy- tant demographic variables, such as age, gender
sis were performed to verify which factorial struc- and years of driving license.
ture best suited the sample. These psychometric Analysing the correlations related to accidents
analyses represent a strength of the study. For the per driving license year, higher number of acci-
NEO-FFI-20 and ImpSS scales the dimensional dents/year have significant and moderate correla-
structures obtained were similar to the original tion with higher scores in IAD and NIE dimensions
ones (Bertoquini & Ribeiro, 2002; Zuckerman, and with lower ages.
1994), with some adjustments, to better fit the The obtained results are in line with studies in
scales to the sample of this research. The excluded traffic psychology (Correia, 2014; Yang et al., 2013;
items were: 14 and 16 in the NEO-FFI-20 scale; 4, Dahlen et al., 2006) and seam to reinforce the idea
6, 7 and 10 in the ImpSS scale; and items 1 and that the personality traits should not be ignored
24 in the DBQ. The number of factors obtained in when studying the behaviour of drivers. Thus,
the DBQ was the same as in the original versions other personal variables like socio-demographic
(Reason et al., 1990; Reimer et al., 2005; Correia, variables as age, gender and driving licence years,
2014), yet the distribution of items in the factorial which also showed to be significant, should be con-
dimensions was slightly different in relation to the sidered when analysing the driver behaviour.
dimensions Non intentional errors and Lapses. The dimension of the sample (N = 747) is a
In each scale the internal consistency results were good strength of this work. On the other hand,
acceptable, good and in one case very good. having used the NEO-FFI-20 instead of the more
Regarding Infractions and aggressive driving, complete and detailed NEO-PI-R (240 items) or
younger and male individuals were the ones with the NEO-FFI (60 items) may be a limitation of
higher scores in this behaviour. this work. However, the use of shorter versions is
Our correlational analysis emphasizes the rel- increasingly recommended in observational survey
evance of the Impulsivity trait, which significantly research, as it decreases the probability of response
correlated with all the driving behaviour dimen- biases and increases the response rates.
sions. These results make sense, since people who are Analysing the results described in this article it
more impulsive, do not anticipate the consequences is possible to establish the hypothesis that people
of their behaviours. Additionally, people who like can be clustered by specific profiles that reflect
to search for new experiences and intense feelings, their personality traits and correspond to different
and at the same time are less compassionate and behaviour patterns on the road, and consequently
cooperative, may have more tendency to commit to distinct risk situations. It is our intention to test
infractions and to have an aggressive behaviour. this hypothesis in the near future. By using differ-
As for the Non intentional errors dimension, ent personality and driving profiles in an experi-
higher tendency to be impulsive and to be less self- mental design, we hope to increase the knowledge
disciplined and less organized is associated with the of human behaviour on the road to better calibrate
tendency to commit errors during the driving task, and validate tools (e.g. microsimulation models)
without intention, like fail to check the rear-view for road safety assessment.
mirror before pulling out and changing lanes, for
example. Furthermore, one can speculate that people
who score high in the Neuroticism trait, may be more ACKNOWLEDGMENTS
prone to Non intentional errors, due to their ten-
dency to worry and ruminate, which may consume This work was done with the support of the MIT-
their attention, particularly in stressful situations. Portugal Program and the Foundation for Science

792

TIS2017_Book.indb 792 2/8/2017 8:09:51 PM

 and Technology. The authors are deeply grateful Kline, R.B. 2010. (3rd ed) Principles and practice of struc-
to all the participants in this study and to all the tural equation modeling. New York: Guilford Press.
professionals involved. Lajunen, T., Corry, A., Summala, H., & Hartley, L. 1998.
Cross-cultural differences in drivers’ self-assessments
of their perceptual-motor and safety skills: Austral-
ians and Finns. Personality and Individual Differences
REFERENCES 24-539-550.
Lajunen, T., Parker, D., Summala, H., 2004. The Manches-
Bertoquini, V., & Pais-Ribeiro, J.L. 2006. Estudo de ter driver behaviour questionnaire: a cross-cultural study.
Formas Reduzidas do NEO-PI-R. Psicologia: Teoria, Accidents Analysis and Prevention 36 (2): 231–238.
Investigação e Prática 11(1): 85–101. Ledesma, R.D. & Valero-Mora, P. 2007. Determining the
Caspi, A., Begg, D., Dickson, N., Harrington, H.L., Number of Factors to Retain in EFA: an easy-to-use
Langley, J., Moffitt, T.E., & Silva, P.A. 1997. Personality computer program for carrying out Parallel Analysis.
differences predict health-risk behaviors in young adult- Practical Assessment, Research & Evaluation 12 (2).
hood: Evidence from a longitudinal study. Journal of McCrae, R.R. & John, O.P. 1992. An introduction to the
Personality and Social Psychology 73: 1052–1063. five-factor model and its applications. Journal of Person-
Cattell RB. 1966. The scree test for the number of factors. ality 60(2): 175–215.
Multivariate Behavioral Research 1(2): 245–76. McDaniel, S.R. & Mahan III, J.E. 2008. An examination of
Cohen J. 1992. A power primer. Psycholl Bulletin 112: the ImpSS scale as a valid and reliable alternative to the
155–159. SSS-V in optimum stimulation level research. Personal-
Correia, J.P. 2014. Traços de personalidade, estados emo- ity and Individual Differences 44: 1528–1538.
cionais e condução: um estudo comparativo entre con- Oltedal, S. & Rundmo, T. 2006. The effects of personal-
dutores de ambos os sexos. Doctoral Thesis, Faculty of ity and gender on risky driving behaviour and accident
Medicine, University of Lisbon, Portugal. involvement. Safety Science 44: 621–628.
Costa, P.T., & McCrae, R.R. 1992. Revised NEO Personal- Oppenheim, I. & Shina, D. Chapter 15, Human Factors and
ity Inventory (NEO-PI-R) and NEO Five-Factor Inven- Ergonomics. Handbook of Traffic Psychology. London:
tory (NEO-FFI) professional manual. Odessa, FL: B. Porter Ed, Academic Press in an imprint of Elsevier.
Psychological Assessment Resources. Parker, D., McDonald, L., Rabbitt, P., Sutcliffe, P. 2000.
Dahlen, E.R., & White, R.P. 2006. The Big Five factors, Elderly drivers and their accidents: the aging driver
sensation seeking, and driving anger in the prediction of questionnaire. Accidents Analysis and Prevention 32(6):
unsafe driving. Personality and Individual Differences 41: 751–759.
903–915. Pereira, A.T., Bos, S., Maia, B., Soares, M.J., Valente, J.,
Dahlen, E.R., Martin, R.C., Ragan, K. & Kuhlman, M.M. Nogueira, V., Pinto de Azevedo, M.H. & Macedo,
2005. Driving anger, sensation seeking, impulsiveness, A. 2013. Short forms of the Post-partum Depres-
and boredom proneness in the prediction of unsafe driv- sion Screening Scale: as accurate as the original form.
ing. Accident Analysis and Prevention 37: 341–348. Archives of Women’s Health 16 (1): 67–77.
De Leo, J.A., Van Dam, N.T., Hobkirk, A.L. & Earleywine, Reason, J., Manstead, A., Stradling, S., Baxter, J., & Camp-
m. 2010. Examining bias in the impulsive sensation seek- bell, K. 1990. Errors and violations on the roads: A real
ing (ImpSS) Scale using Differential Item Functioning distinction? Ergonomics 33: 1315−1332.
(DIF) – An item response analysis. Personality and Indi- Reimer, B., D’Ambrosio, L.A., Gilbert, J., Coughlin, J.F.,
vidual Differences 50. Biederman, J., Surman, C., Fried, R. & Aleardi, M.
Deffenbacher, J.L., Lynch, R.S., Oetting, E.R. & Yin- 2005. Accident Analysis and Preevention 37: 996–1004.
gling, D.A. 2001. Driving anger: Correlates and a test Sharma, S. 1996. Applied Multivariate Techniques. New
of state-trait theory. Personality and Individual Differ- York: John Wiley & Sons.
ences 31: 1321–1331. Verschuur, W.L.G. & Hurts, K. 2007. Modeling safe and
DeVellis FR 1991. Scale development - Theory and applica- unsafe driving behaviour. Accident Analysis and Preven-
tions. London: SAGE Publications. tion 40: 644–656.
Elal, G., Altug, A., Slade, P., & Teckcan, A. 2000. Fac- Wahlberg, A.E., Dorn, L. & Kline, T. 2011. The Manches-
torstructure of the eating Attitudes Test (EAT) in a Turk- ter Driver Behaviour Questionnaire as a Predictor of
ish university sample. Eating Weight Disorders 5: 46–60. Road Traffic Accidents. Theoretical Issues, Ergonomics
Hennessy, D. 2011. Chapter 12, Social, Personality and Science 12(1): 66–86.
Affective Constructs in Driving. Handbook of Traffic Yang, J., Du, F., Qu, W., Gong, Z. & Sun, X. 2013. Effects
Psychology. London: B. Porter Ed, Academic Press in of Personality on Risky Driving Behavior and Accident
an imprint of Elsevier. Involvement for Chinese Drivers. Traffic Injury Preven-
Iversen, H., & Rundmo, T. 2002. Personality, risky driving tion 14(6): 565–571.
and accident involvement among Norwegian drivers. Zuckerman, M., Kuhlman, D.M., Joireman, J., Teta, P.,
Personality and Individual Differences 33: 1251–1263. & Kraft, M. 1993. A comparison of three structural
Kaiser, H.F. 1958. The varimax criterion for analytic rota- models for personality: The big three, the big five, and
tion in factor analysis. Psychometrika 23:187–200. the alternative five. Journal of Personality and Social
Kline, P. 1994. An easy guide to factor analysis. London: Psychology 65: 757–768.
Routledge. Zuckerman. 1994. Behavioral expressions and biosocial
Kline, P. 2000. The handbook of psychological testing (2nd bases of sensation seeking. New York: Cambridge Uni-
ed.). London and New York: Routledge. versity Press.

793

TIS2017_Book.indb 793 2/8/2017 8:09:51 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Motorways of the sea: An outlook of technological, operational

and economic tools

R.G. Di Meglio & G. Mainardi

Livorno Port Authority, Livorno, Italy

S. Ferrini
Scuola Superiore Sant’Anna and Consorzio Nazionale Interuniversitario Per Le Telecomunicazioni, Pisa, Italy

I. Toni
Polo Sistemi Logistici—Università di Pisa, Livorno, Italy

ABSTRACT: The paper addresses recent trends in the Motorways of the Sea (MoS) through an
analysis of actions available to support and improve intermodal transport of people and goods among
European and Mediterranean ports. It highlights public policy tools to back and finance MoS services in
the European and Italian legal framework. It assesses how new technologies can be successfully deployed
within the MoS related logistic chains, analyzing RFID, e-seals, port monitoring platforms, in the light of
market needs for ITS. It discusses the design of port facilities in relation to MoS traffic. The paper deepens
operational issues both on port side as well as on port inland relations, in economical and in technical
terms. It deals with the Port of Livorno case, giving the chance to consider, in the real life of the leading
Italian Ro-Ro port, the implementation of technological, operational and economic tools for MoS.

1 MOS IN THE POLICY AND FINANCING investments in ports do not necessarily turn to
FRAMEWORK incentive modal shift, as technical barriers between
maritime and inland transport (like incompat-
1.1 Relevance of MOS for EU transport ibility for loading units, missing last mile links
policy framework etc) still exist (Baindur, 2011). Promoting MoS
EU policies in the field of MoS have targeted linkages, without solving inland connections, can
infrastructural, environmental and financial issues. thus result in greater road congestion in the port
Bearing in mind that the EU coastline is some hinterland and in the vicinity of port areas, and
70,000 km long and that waterborne transport ultimately lead to minor modal shift. Moreover, as
accounts for about 40% of intra EU trade (Suarez- far as MoS concept is concerned, there is lack of
Aleman, 2015). MoS linkages have been set to play clarity on what is MoS and is therefore fundable
a role both in terms of enhanced connectivity of with European schemes and what, on the contrary,
transport network, as well as of improved sustain- is pure (Short Sea Shipping) SSS service. In this
ability of shipments and economic viability of respect, Regulation EU 1315/2013 laying down the
services. Guidelines of Trans-European Network of Trans-
The European Commission White Paper on port (TEN-T), stated that MoS represent the mari-
Transport (COM 2001) acknowledged that Euro- time dimension of TEN-T, and thus recognized
pean funds, notably grants, were needed to ensure them as TEN-T corridor (Parliament, 2013). They
start up and make these services commercially must link two ports located in two different EU
attractive, as it was stated that “these lines will not countries, either core or comprehensive network
develop spontaneously” (Commission, 2001). ports, and possibly connect European network
In this respect, EU policy was intended to tackle with that of Third countries.
the growth of road transport, which, despite all If we look at figures on EU funding leverage, the
efforts made in the recent decades, is on the rise picture is quite positive, as 450 million EU grants
(Juan, 2016). for 45 MOS projects in the period 2007–2013
In fact, road transport proves to be still competi- yielded 2 billion euros investments. Leverage factor
tive, if we look at costs as well as at flexibility, while is then over 4, below the new target of 15–20 set for

795

TIS2017_Book.indb 795 2/8/2017 8:09:53 PM

 CEF funding but high when it comes to the rela- of road transport in the meantime. European
tively low funding rates, which hinder a real lev- ports. The reason behind these setbacks is rooted
erage effect (EU Parliament, 2016) (Expert Group in the instable shipping industry, in the challenging
5, 2010). Presently, CEF Funding grants 30% of integration of international supply chains and the
total eligible costs for works and 50% for studies, competitiveness of road transport, as European
percentages that are considered low to exert the freight is carried mostly on roads even if a mari-
potential of EU funding. As a consequence, only time solution exists. Another issue to be consid-
projects that would have carried out anyway can ered is the unbalanced trade flows for MOS which,
find attractive the application for European fund- in contrast with other branches of the shipping
ing, whereas riskier projects, or projects with a big- industry, have a greater need for balanced demand
ger need for initial ramp-up financial aid, are more of transport for starting the connection (Baindur,
likely to be disregarded by EU financing schemes. 2011). The present 16 MoS operating services (Juan
European funding has been increasingly ear- et al. 2015) face therefore relevant constraints as
marked to support sustainability of maritime for financial and economic issues, even after opera-
linkages, and the Connecting Europe Facility pro- tions have started. A public support scheme also
gramme (CEF) now in force has steadily focused available is State Aid. In the lack of EU funding to
on Liquefied Natural Gas (LNG) as more envi- back a MoS project, European Commission allows
romental-friendly fuel for transport. If we look Member States to support the project for a maxi-
at 2015 selected CEF proposals for the MoS mum duration of five years and up to 35% of total
priority, four out of twelve concerned LNG pro- operational costs (European Commission, 2008).
motion in shipping and installations in ports, other National financing schemes can be also envis-
three were linked to sustainability while only two aged. In this respect, the Italian Marebonus deserves
were more a “traditional” MoS project, financ- mention not only as example of national financing
ing a new maritime linkage and the related works scheme for backing modal shift through the imple-
along with IT application. The remaining three mentation of new MoS services, but also in the
proposals addressed navigation safety issues and light of a global transport policy laid down by the
data exchange among IT platforms. In 2014 MoS Government (Ministero Infrastrutture e Trasporti,
selected proposals were 27, of which 10 supported 2015). In fact, the 2015 unveiled National Strategic
the upgrade of Maritime services or initiatives in Plan for Logistics and Ports sets forth the improve-
ports or in the hinterland instrumental to carry-on ment of accessibility as well as maritime and inland
the proposed MoS, while five dealt with LNG connections. The financing scheme, totaling 138 mil-
distribution in ports and six targeted goals of sus- lion euros between 2016–2018, is structured as
tainability through enhanced MoS chains. This incentive for the shipping company operating a MoS
picture highlights how important are the environ- service, which has contributed to shift freight from
mental concerns in connection with MoS linkages road to sea transport; this incentive is then reversed
and how the focus shifted from the economic and to the haulier for each shipped unit, which has been
financial viability of project towards sustainability diverted from all road transport to MoS service. The
of MoS: within the TEN-T programme frame- Marebonus incentive is the follower of the Ecobo-
work, a total of 11 projects in the field of LNG nus, established with Law 265/2002, which managed
have been funded out of 45, reaching 103 million to divert 5% of total road traffic, not only by finan-
out of 369 million euros. cially support MoS services, but also by granting
As a consequence, the environmental side of aid to measures in the field of ICT and to bringing
European support to MoS is gaining room, while together small and medium haulers (Rete Autostrade
more commercially focused actions or ICT projects del Mare, 2014). It has been acknowledged that
somewhat lag behind. This trend can be partly intermodal solutions through national MoS services
due to the new goals set forth by the 2011 White have somewhat contributed to sustaining the coastal
Paper, in particular those goals referred to the use navigation demand, spreading the offer of transport
of more environmental-friendly fuels, which fore- solutions along the Italian peninsula. Marebonus is
see horizontal priorities of EU transport that can still waiting for EU Commission’s approval.
be found not only for MoS, but obviously for all Despite the attained levels of modal shift, financ-
transport modes. ing schemes such as Marebonus and Ecobonus
clash with traditional setbacks of MoS, which have
been partly explained above. First of all, although
1.2 Challenges and setbacks of financing
relevant players of the transport chains, hauliers
programmes
account for only 5% of global costs, while ports
The impact of previous relevant programmes on operations costs and dues play by far a bigger role
modal shift, what has been stated as major tar- (Isfort, 2014); furthermore inland transport costs
get for MoS, seems to be rather scarce or even are much more likely to raise as consequence of poor
negligible in comparison with the massive growth coordination among players. Different business

796

TIS2017_Book.indb 796 2/8/2017 8:09:53 PM

 models and unequal distribution of costs are to be In particular, the White Paper 2011 looks towards
blamed for bias in loading and unloading of freight a competitive and resource efficient transport sys-
time, delays and inefficient planning of capacity tem, breaking down barriers to the completion of
(Van De Horst, 2015), resulting in higher costs for the internal market of transport, by developing
the whole MoS chain. Only truly pan-European the concept of e-freight. E-freight consists in the
schemes would therefore reduce congestion in Core wide deployment of information technology tools,
European regions and foster intra-European trade, to simplify administrative procedures, provide for
for the benefit of more peripheral regions too. As a cargo tracking and tracing, and optimize schedules
consequence, to be more effective, MOS support- and traffic flows (European Commission, 2011). In
ing policies should look towards better inter-firm addition to this, the Maritime Transport Strategy
cooperation, global planning of capacities and 2018 highlights the importance to deploy e-mari-
investments in inland terminals and eventually time services at both European and Global level.
fine-tuning of operations of players through closer The e-maritime concept aims at promoting the
integration or agreements. In the following section, competitiveness of the European maritime trans-
we explore how ICT tools can effectively contribute port sector and a more efficient use of resources
to the viability of MoS service since the ramp-up through the widespread use of ICT tools (Euro-
phase, which has proven to be the most critical. pean Commission, 2009).
The e-freight concept, the e-maritime services
and the digitalization of all information related
2 HOW THE TECHNOLOGIES CAN BE
to cargo and passengers pave the way for the crea-
SUCCESSFULLY DEPLOYED WITHIN
tion of digitalized environments as Digital Single
THE MOS LOGISTICS CHAIN
Market (European Commission, 2015) and Digital
Administrations (MIT Transport, 2015).
2.1 ICT in MoS, introduction and regulations
Already in 2010, European Commission pro-
Seaports are considered as intermodal points in vided the basis for the creation of digitalized envi-
complex logistics chains, enabling to connect dif- ronments, within the concept of Maritime Single
ferent nodes of transport (i.e. other seaports, Window (MSW). Indeed, Directive 2010/65 aims to
inland nodes) in terms of both physical flows (i.e. establish MSW in the EU Member States, allow-
freight, cargo and passenger handling) and infor- ing the development of a platform to exchange
mation flows. In this context, the Motorways of data between EU Member States. MSW contrib-
the Sea are not only a matter of maritime transport utes to the development of a European informa-
between two ports, but are rather the establishment tion structure. Possibly, this would also provide the
of efficient and complete logistics chains among European maritime transport industry “a place to
origin and destination points, through upgraded stand on”. Directive 2010/65 supports the ONCE
transport nodes of a fully integrated network. The paradigm, based on data single electronic trans-
vision of a fully integrated network combines: mission and a single control declaration formality
the removal of existing barriers between nodes, for ship departing/arriving from/in member states’
the overcoming of obstacles for fast exchange of ports (European Commission, 2010).
information (for every transport node, independ-
ently of the concerned country), and the possibility
2.2 ICT in MoS, current practices
to track, trace, monitor and control cargo, pas-
sengers and events. To do that, the upgrading of To track, trace, monitor and control cargo, pas-
Information Communication Technologies (ICT) sengers and environments, and to facilitate the
tools is as important as the upgrading of handling communication and the exchange of informa-
equipment and physical facilities. tion among two (or more) nodes of transport
Regulation framework (at European and national involved in MoS, each of these nodes have to be
level) greatly emphasizes keywords as Intelligent upgraded and innovated. The upgrading process
Transport Systems (ITS) and Digitalization. Euro- consists in equipping each node of harmonized
pean Directive 2010/40 defines ITS as advanced info-structure, fostering the use of standards (in
applications which without embodying intelligence terms of data exchange and communication). On
as such aim to provide innovative services relating to the other hand, the innovation process consists in
different modes of transport and traffic management testing and setting-up advanced solutions, in order
and enable various users to be better informed and to both facilitate and accelerate the real-time data
make safer, more coordinated and ‘smarter’ use of collection, communication and dissemination of
transport networks (European Commission, 2010). information.
ITS are part of ICT and they integrate telecommu- The first step towards upgraded seaports is the
nications, electronics and information technologies design of a coherent ICT infrastructure, in order
with transport engineering, in order to plan, design, to create more interconnected Seaports within the
operate, maintain and manage transport systems. connection of the node both in itself (i.e. setting up

797

TIS2017_Book.indb 797 2/8/2017 8:09:53 PM

 a LAN of the communities that are based on the are read by readers placed in strategic points (i.e.
Port landside) and with other transport nodes along gates, berths.). Sensors Network is composed of
the MoS (i.e. setting up Internet wide-corridors of large number of sensor nodes that can be deployed
the member states communities) (Pagano et al, on the ground, in vehicles, inside buildings, on the
2016). The ICT infrastructure enables to gather berths, on the quays, on the gates, on port equip-
data from humans and machines (i.e. sensors and ment (i.e. cranes) (Jain et al, 2012). Each of the
vehicles), in order to create more interconnected distributed sensor node has the capability to col-
and smart environments and to facilitate the com- lect data and route them to a specific gateway,
munication inside and between seaports placed in in order to facilitate the data processing and the
different geographical areas, making way to new communication tasks with application layer. The
paradigms as Internet of Things and Big Data. WSN are usually used to detect and monitor the
The use of cutting-edge technologies in the environment. In MoS, integrating RFID with
transport field is the first step towards the Innova- WSN can provide both identity and location of an
tion process. Indeed, after creating an ICT infra- object (thanks to RFID technology) and informa-
structure ashore, seaports should consider which tion regarding the condition of the object carrying
are data to be gathered, and why and how to do the sensors (i.e. the status of the cargo should be
it they can do it; the more seaports can collect detected as quickly as possible, and alarms should
and manage data, the more MoS increase their be triggered when temperature gradients cross a
competitiveness. threshold). Sensors can “give expression” to con-
Data generated in the transport and maritime tainers, vehicles, freight, cranes and other objects,
field don’t come exclusively from the documents promoting the concept of Internet of Things or (as
accompanying cargo and passengers, but also from usually referred to in the field of transport) Inter-
transporting and handling activities and from the net of Goods.
port and logistics environment. In fact, users are Sensors, RFID, other smart devices, legacy sys-
interested in tracking, tracing, monitoring, and tems enable the digitalization of a huge amount of
controlling goods, people and events, knowing data, introducing the Big Data paradigm. Gartner
their position in real-time, for different purposes defines Big Data as high-volume, high-velocity and/
(i.e. administration bodies are interested in detect- or high-variety information assets that demand cost-
ing hazardous events in port areas and speeding-up effective, innovative forms of information processing
control activities, while importers are interested in that enable enhanced insight, decision making, and
knowing the position and the status of their cargo process automation (Gartner, 2016). Data could be
during the shipment). In so doing, seaports should investigated using semi-automatic and automatic
employ smart devices and technologies used to methodologies (business intelligence technologies),
enable customized ICT and ITS applications, in as OLAP, Data Mining (and its related techniques),
order to both monitor port areas and gather infor- for gathering, extracting, transforming, loading
mation related to efficiency level of MoS (in terms and reporting data from different data sources
of safety, security, speed, costs, times, etc.). Key aforementioned (Bhavin et al, 2004). The design of
enabling ICT technologies includes vehicle detec- distributed intelligent tools for MoS management
tion technologies (i.e. radar, laser, video image could be used for gaining competitive advantage
processing, infrared.), vehicle monitoring and through real-time collaboration with trading part-
tracking technologies (i.e. GPS, Radio frequency ners, and offer a new way to rapidly plan, organize,
identification technologies—RFID), distributed manage, measure and deliver new services.
sensors network, video cameras, connected vehicle
technologies (V2X technologies) and so on. The
2.3 Case study: ICT in the port of livorno
proper integration among these technologies ena-
bles to obtain heterogeneous and real time infor- The Port of Livorno is one of the largest multipur-
mation. Information, if rightly processed, allows pose Port in the Mediterranean Sea, and it is one
identifying value-added services, increasing the of the major Italian Port gateway belonging to core
competitiveness of seaports as gateways of MoS network of TEN-T (Scandinavian-Mediterranean
and, broadly speaking, of logistics chains. Corridor).
Two of the most interesting and widespread In order to speed-up the information flows inside
wireless technologies are RFID and Wireless the Port Areas and among the other transport
Sensors Network (WSN). RFID is the main low nodes, since 2000 Livorno Port is equipped with an
cost automatic identification technology used for optical fibre network. Thanks to this broadband
tracking and tracing cargo and vehicles, using sim- backbone in terms of connectivity, Livorno Port
ple tags attached or embedded into objects that are is pervasively monitored and centrally controlled.
to be identified or tracked (i.e. containers, trailers, The good connectivity inside the Port has estab-
pallets, track and so on). The information related lished the basis to create the smart environment,
to the objects are stored into the tags and they by making Livorno Port as interesting inter-modal

798

TIS2017_Book.indb 798 2/8/2017 8:09:53 PM

 point of interest, where to develop and validate 3 INFRASTRUCTURE PRIORITY FOR
Intelligent Transport Systems and innovative EFFICIENT MOS IN SEAPORTS
ICT technologies. The Port is working to set-up
a broadband backbone in the surrounding areas 3.1 Introduction
(with particular attention to the link with Guas-
Infrastructure development of the country, both
ticce Freight Village), to create controlled, moni-
on the maritime and earth side, by the Maastricht
tored and fast transfers.
Treaty onwards must comply with decisions taken
To support the design and implementation of
collectively and move in synergy with the European
the ICT Technical Agenda, in 2015 Livorno Port
mark. Thus, the decisions taken by the EU in 2010
Authority has created the “Joint Laboratory of
with the Regulation no. 913/2010 are influencing
Advanced Sensing Networks & Communication in
the development of the national port system-but
Sea Ports”, that provides a continuous and effec-
until 2020 since that regulation has established the
tive presence of CNIT (National Inter-University
criteria and rules for the construction of a Euro-
Consortium for Telecommunications) researchers
pean logistics network whose international access
at the Port.
points will be the main port systems continental.
The objective of Livorno Port and the mission
(Commission, 2010).
of the Joint Lab is to create smart and interop-
The way to relaunch SSS is to build real sea
erable Port environments, within which people,
motorways within the framework of the mas-
vehicles, objects, equipment, and cranes can com-
ter plan for the trans-European networks. This
municate with both each other and a central
requires better connections (accessibility) between
monitoring and control system. For this purpose,
ports and rail together with improvements in the
Livorno Port Authority has developed the moni-
quality of port services, port facilities, port com-
toring and control system of the Port of Livorno,
munity and port monitoring systems, shipping
MONI.C.A. (Monitoring and control application).
services (frequency, regularity, safety, security).
MONIC.A. collects and integrates all data coming
(Directorate-General for Internal Policies, 2014).
from distributed sensors network, information
Port infrastructures include not only facilities
systems (i.e Port Community System, Harbor’s
related to the delimitation of areas, but also port
systems, Operators’ systems, dangerous goods sys-
gates, road and rail connections to the hinterland
tem, etc.), tracking and tracing devices (i.e. RFID)
and dedicated facilities.
and other smart and intelligent devices. In addi-
The intermodal transport quality is a critical
tion to wireless sensors network, Livorno Port is
factor in the organization of the intermodal chain
testing the vehicular communication, to gather
based on maritime transport. In general, studies
information coming from vehicles passing through
on the intermodal quality, and in particular, those
the Port (i.e. destination, speed, hazardous events,
on maritime chains, are the pillars towards the
etc.), becoming, by so doing, the new test-bed for
optimal planning of intermodal transport and the
European Telecommunications Standards Insti-
removal of bottlenecks. (Isfort, 2015).
tute Plugtests 2016. To identify smart solutions for
MoS, Livorno Port has participated in different
3.2 Terminal layout
European Commission and National Initiatives as
Business to Motorways of the Sea (B2MoS 2012- Motorways of the Sea area is one of the MoSt
EU-21020-S), Monitoring and Operation Services complex to organize, implement and manage;
for Motorways of the Sea (MoS4MoS 2010-EU- must face the growing demand for mobility of
21102-S), Mediterranean Information Traffic Appli- passengers and cargo, matching the needs of busi-
cation (MED.I.T.A. Med Programme 2007–2013), nesses, ship owners, tourists and operators, aim-
towards the creation of intelligent smart MoS and ing at optimize the connections of the port with
smart logistics chains. RFID technology and the users and with the surrounding area. In this sce-
digitalization of transport documents have had nario, relevant issues are related to port security,
central role in these initiatives. customs management of goods and passengers,
In particular, RFID technology has been used: infomobility, the management of operational yard,
i) to improve customs procedures and port secu- the use of technologies and of advanced sensors.
rity in terms of cost and time, speeding-up cargo The main idea for the organization of area MoS
controls (active RFID), and ii) to track and trace is based on traffic flow management and the bal-
trailer and their cargo along both the MoS and ance of incoming and outgoing flows of vehicles,
the logistics chain (passive RFID). The creation making the distinction between heavy and light
of digital and smart ports is the real challenge to ones, already from the port access area. Even
increase the competitiveness level of the MoS, travelers without car or truck should benefit from
considering that nowadays, users are interested in a specialized connection through a mechanized
obtaining real time information about their cargo pedestrian path which allow them a direct link to
and the events occurred during the path. the boarding dock. MoS area also needs all the

799

TIS2017_Book.indb 799 2/8/2017 8:09:53 PM

 auxiliary services for workers, passengers and users to develop a functional railway system to handle the
such as: customs offices, control areas, local duty amount of traffic coming in and out. Catching flows
free, landing and boarding hall, shopping center, to be transferred from road to sea are those with a
bar. It is also possible, to make modern, environ- distance greater than 500 km. Such movements rep-
mentally friendly, sustainable and technologically resent a marginal share of heavy road transport. But
advanced terminal, to equip it with photovoltaic the environmental, social and economic impacts
panels. (Mare, 2014). of this traffic part are much more significant than
MoS must have a modern platform of infrastructure those generated by the short-range transport. The
and specific requirements: it must have an adequate movement of goods by road is mostly concentrated
surface for the movement of goods, including docks over distances below 100 km, however, from the
and maneuvering areas for loading and unloading, point of view of the negative and energy consume
customs stations, the long stock areas for trucks and impacts, one ton of goods carried over 600 km gen-
other means, buffer area for motor vehicle accumula- erates negative impacts and consumption energy six
tion, yards accumulation of heavy vehicles; the viabil- times higher than those required to transport the
ity input and output must be designed and managed same quantity of goods to a trip of only 100 km.
not only for the standard conditions of operation in Intermodal transport is hindered by the situa-
the winter months, when the passenger flow is con- tion of the national railway network which cannot
siderably lower, but especially consider summer influx guarantee the flow of any type of vehicle or goods
of conditions in which there is combination of heavy across the country. At European level, the Italian
and light vehicles. The MoS platform need for a com- railway network, from the infrastructural point of
mercial area for passengers and port operators, easily view, is the most backward: ports of central/south-
accessible from both categories of users. This area can ern Italy are not able to receive cargos (trailers,
be connected directly with the operational areas for trucks, swap bodies) because of rail loading gauge.
the best usability of services for heavy vehicles drivers. (Directorate-General for Internal Policies, 2006).
Buffer areas must be flexible and specialized by cat- A MoS railway terminal provides specific infra-
egory of traffic to facilitate the movement of vehicles structure and equipment as a freight railway sta-
and customs operations. The same concept should be tion, railway yard for handling wagons/goods,
applied to ship buffer area. The viability for light and lifting means and cranes.
heavy vehicles must be provided not only both for dif-
ferent types of means but also for different streams,
3.4 Extra Schengen area
load/unload. It is possible to provide the realization of
pre-boarding area where manage customs operations, The Extra-Schengen area is an area in which they
with dynamic and flexible accumulation lanes that are handled cargo and passengers, coming from
lead directly to the appropriated dock. The unloading countries outside the Schengen Agreement.
flow is disposed in advance with loading one without Within this area there must be space for the
overlapping. The management of the flow and of the movement of vehicles and passengers, spaces for
relative accumulation and outflow are the main aspects checks by police and agents of Customs, a viability
for a MoS terminal, because these aspects directly provides the unloading/loading without intersec-
related to passengers and port operators transit time. tion of the flows.
Another important feature of a functional MoS The design of a functional Extra-Schengen
terminal is the creation of a single access gate. The area provides the division of the different flow:
gate must be equipped with technological compo- entry and exit. There is the need to maintain sepa-
nents and sensors of the latest generation for moni- rate the viability not only as it regards the flow of
toring, data warehousing and terminal security. vehicles from internal countries of Schengen area
The core facilities should be located centrally within than those with external origin, but also of com-
the MoS terminal. The core of a MoS terminal can mercial vehicles by private ones, or users that are to
be identified in the boarding terminal, which should undergo inspections by customs. Separating vehicle
be conceived as a dynamic element. Inside are place- flows is essential for a complete functionality of the
ment offices, bar, shop, information desk, companies, area, allowing safety and reliability controls. An
customs, duty-free, waiting rooms, toilet facilities. It improper localization of different functions would
must in fact be easily reached both by external users imply traffic jams and queues at terminal gates.
and passengers. In addition to roads and yards, for the monitoring
of vehicles and passengers areas shall be equipped
for additional services to users and operators.
3.3 MoS and railway
Since MoS areas are created to move cargos from
3.5 Case study: Livorno port
the road in order to limit the negative impacts gener-
ated by the intensive use of road transport in terms The current MoS area in the port of Livorno is
of pollution, congestion and safety, terminals need located in the north of the city, within Porto Nuovo

800

TIS2017_Book.indb 800 2/8/2017 8:09:53 PM

 zone. Analyzing the area from the point of view of Within MoS terminal, temporary storage area
infrastructural layout, it has good connections to (TC) of goods are essential for current trades and
the road network but the railway sector still has the lack of these zones creates considerable prob-
expansion capabilities. lems for operators. In case of short-term changes of
The road network does not differentiate between the terminal, the creation of TC area is essential.
incoming and outgoing flows, creating promiscuity During 2015 it was approved the Port Master-
between commercial and private vehicles. plan for the Port of Livorno, which includes large
The rail network presents railways tracks out of infrastructure investment for the construction of an
service that should be reactivated, even in relation expansion to the sea insisting on the embankment
to recent demands of logistics operators, interested in part already realized and partly newly built.
in having dock directly linked to the railway. One of the main issues for the design of the
The access gate of the vehicles is the Varco Galvani, Darsena Europa was the need to offer not only docks,
which has two in/out lanes per way, differentiated for quays and yards but also an intermodal equipment
trucks and cars. The main problem is the placement and efficient logistics tools that will lay the founda-
of the gate, that does not have a buffer area sufficient tions to achieve the realization of the whole port-
for the vehicles exiting from the terminal. logistics platform as a single entity (Livorno, 2015).
Presently, all rail traffic in/out the port of In this sense it was necessary to characterize the
Livorno goes through the Livorno Calambrone port areas with specific logistics functions: ration-
freight station, which lies in the Genova-Roma alize and make efficient the different activities.
railway line and where all shunting operations are The project involves the port expansion in the
performed. The port railway network suffers from sea with the construction of new yards and new
poor connectivity, which currently hinders the rail docks. The functional areas are divided into Molo
operations as a result of slow shunting operations Nord, Molo Sud, Darsena Petroli and Darsena Flu-
and low quality accessibility to the shore. viale. These will be equipped with all the necessary
As a consequence, Livorno Calambrone rep- infrastructure for connecting the piers with the
resents a bottleneck of the Core rail network, as hinterland, areas for handling and movement of
time-consuming and expensive shunting opera- goods and direct connections to the internal mar-
tions are required to forward freight trains from kets (Livorno, 2015).
the port to the hinterland. Works are already ongo- Within this allocation, it also planned to build
ing to directly link the largest container terminal of a new area set to the Motorways of the Sea, in the
the port, ‘Darsena Toscana’ to the rail network. Molo Nord. This area has been designed according
It expects to achieve some interventions to to the expected traffic volumes and into considera-
develop the rail network so as to make more effec- tion that the Port of Livorno is the first Italian port
tive links between the port and the industrial for RO/RO traffic, as shown by the statistics for the
hinterland, reducing the problems created by the year 2015 (Autorità portuale di Livorno, 2015).
Calambrone station. An important intervention, The plan intends to develop the road link (with
that represents the first step towards the enhance- motorway features) and the main railway lines,
ment of better rail connectivity of the port to the such as the Tyrrhenian axis and connecting via
hinterland, is the realization of the rail overpass on Pisa-Florence to the Adriatic coast. Concerning
the Genova-Roma railway line to avoid the Calam- the organization of the new terminals of the port
brone Station and put the train on the network. of Livorno, it was identified a solution for railway
This infrastructure will be a source of development infrastructure considering the new configuration of
not only in a geographical perspective, but also the areas, Molo Sud and Molo Nord. The electrified
from the point of view of fully integrated logis- railway track, 750 m length, in accordance with the
tics systems. In fact, the rail overpass is needed to European interoperability standards, will split with
ensure fast and efficient transfer of goods between two new connections to two operative beams of
the port and the freight village of Guasticce. In different docks, one for the new container terminal
this sense, the expected benefits from the realiza- (Molo Sud) and the other for the new MoS terminal
tion of the rail overpass can be split into local and in RO/RO traffic service (the ferroutage terminal).
network ones, since the realization of the overpass Besides railway tracks, the new terminal will set
will enhance significant improvements both for the up by operational yard for the handling of goods
logistics within the Livorno Logistic Node and the and vehicles, lifting storage area, storage technical
rail transport along the corridor as well. warehouses and depots for terminal functions.
The current layout of roads and infrastructure
generates the need to make a short-term interven-
tion to increase the functionality of the entire area. 4 CONCLUSIONS
The implementation of buffer areas for commer-
cial vehicles would avoid traffic jams and the flows MoS often conjure up trucks loading and unload-
in/out would not suffer slowdowns. ing from ferries within terminals, featured as

801

TIS2017_Book.indb 801 2/8/2017 8:09:53 PM

 slightly more than parking slots. Actually MoS Commission European 2001. Transport White Paper.
are a rather complex world where onboard ICT Commission European 2008. Communication from the
tools communicate with infrastructures and smart Commission providing guidance on State aid comple-
devices placed in ports and within other logistic mentary to Community funding for the launching of
the motorways of the sea.
facilities. Such a framework is then subject to the Commission European 2009. The European Union’s
challenges of the shipping industry, that is volatile Maritime Transport Policy for 2018: Brussels.
freight rates, the availability of cargo and the sea- Commission European 2010. Directive 2010/40/EU.
sonality of traffic. Moreover, MoS face the com- Commission European 2010. Regulation (EU) No 913/2010
petition of all road cheap services, what is on the of the European Parliament and of the Council of 22
contrary hardly ever the case for deep sea connec- September 2010 concerning a European rail network for
tions. Viable MoS have to lay on financing schemes competitive freight.
that have been often provided through public fund- Commission European 2011. Transport White Paper,
ing, especially when it comes to the testing and COM(2011) 144 final.
Commission European 2015. Digital Single Market
implementation of ICT and the launch of new Priority.
maritime linkages. The paper shows how impor- Commission European Directive 2010. 2010/65/EU.
tant is to ensure the sustainability and affordabil- Directorate-General for Internal Policies European Parlia-
ity of infrastructural and infostructural solutions, ment 2014. Improving the concept of “Motorways of
through a greater involvement of private operators the Sea”.
and better integration of all these hardware and Directorate-General for Internal Policies European
software components. As a consequence, MoS can Parliament 2016. Motorways of the Sea—Modernising
have an impact on port communities and transport European short sea shipping links.
operators, broadening the logistic offer of ports European Parliament-Directorate General for Internal
Polcies 2016. Assessment of Europe Connecting Facility-
and port related logistic systems. MoS services call An in-depth analysis.
for global solutions, involving shipping industry’s Expert Group 5-TEN-T Policy Review 2010. Funding
stakeholders, public administrations such as Port Strategy and Financing Perspective.
Authorities and inland transport operators, fine- Gartner IT Glossary 2016. http://www.gartner.com/it-
tuning private investments and public policies. In glossary.
this respect, we pointed out how much important Giurini, A., La Tegola, O. & Miranda, L. 2012. La sicurezza
is the proper planning of port areas by competent sul lavoro nei porti I Working Papers di Olympus.—09.
public authorities for smooth and efficient work- Isfort Confindustria Italia 2014. Autostrade del Mare 2.0,
ing of MoS and MoS transport related services. Risultati, Criticità, Proposte per il rilancio.
Isfort Confindustria Italia 2015. Sviluppo
Administrations and public research bodies are dell’intermodalità—Autostrade del Mare 2.0 e combinato
also essential in the start-up phase of new tech- marittimo—Rapporto finale.
nological tools, as the industry may not be yet Jain, P.C. & Vijaygopalan, K.P. 2012. RFID and Wireless
capable of identifying which IT solutions are most Sensor Networks.
suitable for its needs. Public support, via European Juan C., Olomos, F. & Pérez E. 2016. Decision support
programmes, may still be needed in the future too, system to design feasible high-frequency Motorways of
especially in the light of developing innovative tools the Sea: a new perspective for public committment. The
and testing systems to trace, monitor and control Engineering Economics Vol. 61 (3): p. 163–189.
MoS related cargo. In so doing MoS chains and Ministero delle Infrastrutture e dei Trasporti 2015. Piano
Nazionale della Portualità e della Logistica.
ports can better compete with traditional all road MIT Transport Italian Ministry of Infrastructure and
transport through enhanced sustainability (both Transport 2015. Ministerial Decree n. 287.
economical as well as environmental) and security Pagano, P., Falcitelli, M., Ferrini, S., Papucci, P., De Bari,
of goods. Ports dedicated infrastructure can there- F. & Querci, A. 2016. Complex Infrastructures: the
fore become more efficient and better exploited, benefits of ITS services in seaports in Pagano P. (eds),
achieving greater return on investment and thus Intelligent Transport Systems: from good practices to
attracting the needed private funding. standards. CRC Press.
Parliament European 2013. Regulation of the Council and
the European Guidelines of the TEN-T.
Rete Autostrade del Mare, Sviluppo Italia 2014. Autostrade
REFERENCES del Mare: Il Master Plan nazionale-Infrastrutture.
Suarez-Aleman, A. 2015. Short Sea Shipping in today’s
Autorità Portuale di Livorno 2015. Relazione generale— Europe. A criticial review of maritime policy Maritime
Piano Regolatore del Porto di Livorno. Economics and Logistics. p. 1–21.
Baindur, D. & Viegas, J. 2011. Challenges to implement- Van De Horst, M. & De Langen, P. 2015. Coordination
ing motorways of the sea concept-lessons from the past in hinterland transport chains: a major challenge for
Maritime policy and Management. Vol. 1. - p. 673–690. seaports communities in Haralambides H. (eds), Port
Bhavin, B. & Vipul, J. 2004. Data Mining: Knowledge Dis- Management Palgrave Mac Millian.
covery in Databases International Journal of Research in
Computer Science and Management. Vol. 1(2): p. 74–76.

802

TIS2017_Book.indb 802 2/8/2017 8:09:53 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Assessing sustainability of road tolling technologies

D. Glavić, M. Milenković, A. Trpković & M. Vidas

The Faculty of Transport and Traffic Engineering, University of Belgrade, Belgrade, Serbia

M.N. Mladenović
Aalto University, Espoo, Finland

ABSTRACT: Facing a choice between a range of road tolling technologies, decision-making has to
account for the effects that specific road tolling technology causes. For example, road tolling technology
has direct impact on increase of vehicle operating costs, increase of greenhouse gas emissions, increase
in travel time costs, as well as decrease of road users’ level of service and safety. Thus, selecting the road
tolling technology becomes a question of managing the technology’s effects at an optimum level. To this
end, this paper evaluates both wide-spread as well as new emerging road tolling technologies, including
RFID, GNSS/CN, DSRC, smartphone, infrared, ANPR, smartcard, vignette, and ACM. The methodol-
ogy for analyzing these road tolling technologies focuses on technical, financial, efficiency, environmental,
and social aspects. As a result of this analysis, different technologies are comparatively ranked through a
multi-criteria analysis.

1 INTRODUCTION – Commission Decision 2009/750 / EC. This deci-

sion provides a description of the technical
An important element in the delicate task of road systems and interfaces necessary for the EETS
pricing is the selection of tolling technology. This (European Electronic Toll Service).
task is very complex and depends of many factors.
In accordance with Directive 2004/52/EC and
The conventional road tolling technology involves
the Decision 2009/75/EC European Union creates
toll plazas with manual toll collection in the form
the toll policy in Europe according to the principle
of cash, tokens, or other physically transferable
of “one OBU one Contract all EU”.
material. This system of toll collection is inher-
Besides the many and diverse interests at stake,
ently inefficient for many reasons. Consequently,
there is a multitude of TCT currently in use. Nowa-
transportation planners and road managers face
days, the TCT used worldwide include vignettes,
the challenge of selecting Toll Collection Technol-
Dedicated Short Range Communication (DSRC)
ogy (TCT) for upgrading the existing system or
with barriers, DSRC Multi-Lane Free Flow
when building new motorways.
(MLFF), barcode TCT, Radio-Frequency Iden-
A series of EU directives is trying to put some
tification (RFID) TCT, infrared TCT, smart card
order in the area of road tolling technology, both in
TCT, Automated Number Plate Recognition system
terms of technology and price systems as well as on
(ANPR) TCT, Global Navigation Satellite System/
other outstanding issues. The main objective of the
Cellular Network (GNSS/CN) TCT, tachograph—
EU is to achieve interoperability by policy called one
tolling, Automated Coin Machine (ACM) and
market, one contract, one OBU (On Board Unit).
smartphones tolling (Glavić & Milenković 2016).
Other objectives relate to the unification of prices
Taking into account the diversity and conflict of
and vehicle categories (Glavić 2016). EU based its
interests as well as a multitude of TCT, the decision-
policy on toll in following two documents:
making is further complicated having in mind that
– Directive 2004/52/EC. This Directive prescribes each TCT has certain advantages and drawbacks
the conditions necessary to ensure interoper- from the standpoint of road managers and users.
ability of road tolls in the EU. This applies to For example, different TCT respond differently
the electronic collection of all types of road fees, to the need to maximize the toll revenue, which is
on the entire road network, urban and intercity, often an important criterion for road managers.
highways, high-speed roads, and various infra- On the contrary, different TCT cater differently
structure and facilities, such as tunnels, bridges to the user needs for reducing delays and costs, or
and even ferries. improving safety. Moreover, TCT needs to achieve

803

TIS2017_Book.indb 803 2/8/2017 8:09:55 PM

 important sustainability goals from environmental, database and a particular amount of money is
social, and economic aspect. deducted from the user’s account. If the recorded
Similar decision-making issues have been plate number is not read out properly or not found
observed before in selecting transport technology, in the records, an enforcement violation alarm is
e.g., traffic signal controllers (Abbas et al. 2013, generated to alert the authorities.
Mladenovic et al. 2013, Mladenovic & Abbas, Barcode TCT is a subcategory of ETC. Within
2013). In order to deal with a range of decision fac- this system, the barcode is a sticker applied to the
tors and a multitude of TCT, this paper uses Multi- windshield of the vehicle which is read by a laser
Criteria Analysis (MCA). In order to cater for all scanner while the vehicle passes through the toll
the dimensions of sustainability, the decision cri- plaza. It is the simplest as well as the oldest tech-
teria include Level Of Service (LOS), environmen- nology (Sharma 2014).
tal, social and economic impact of TCT. The aim RFID TCT system contains the OBU installed
is to identify a group of TCT that satisfy sustain- on the front windshield of the vehicle. At toll
ability criteria, that have satisfying LOS for road plazas, this system is read by a RFID frequency
users, and can fulfill tolling manager’s objectives. reader or by an antenna. It can be either prepaid
The focus of the analysis is on the groups of TCT or postpaid, with or without a gate (Blythe 1999,
instead of individual TCT, aiming to test the dec- Sorensen & Taylor 2005).
ision-support framework and normalize the impact GNSS/CN technology includes a global navi-
that different groups of TCT might have on the gation satellite system incorporated with the com-
users, environment, and society. The grouping of munication mechanism. It functions using a global
TCT is additionally reasonable taking into account positioning sytem unit (GPS/Gallileo/Glonas)
that there are more than 13 TCT worldwide, con- mounted on the OBU which stores the coordinates
straining the collection of necessary input data. of the vehicle and sends the transaction informa-
tion to toll collection authorities via GSM/3G/4G
(global mobile communication system) (Blythe
2 TOLL COLLECTION TECHNOLOGY 1999, Catling 2000, Charpentier & Fremont 2003).
OVERVIEW Infrared TCT is similar to RFID and ETC
DSRC systems; the only difference is that it has an
This section provides an overview of all the TCT active infrared unit installed in the vehicle which
present worldwide at the moment of evaluation. contains all the information. (Shieh et al. 2005,
Furthermore, this section will focus in particular Staudinger & Mulka 2004, Tropartz et al. 1999).
on the dimensions of environmental sustainability, Tachograph TCT records the mileage driven by
system architecture, system efficiency, and user the user through an OBU connected electronically
friendliness. to the vehicle´s odometer (the instrument that meas-
ures the mileage). A tachograph system which is in
place in New Zealand requires manual, rather than
2.1 Description of TCT
electronic, data collection (Technology options for
Vignette is a sticker which, once bought, pays the the European Electronic Toll Service 2014).
toll for a specific time period. The vignette must be ACM is a machine with a slot for inserting coins
applied to the car windshield (Glavić, 2013). and paper money. Automated coin machine some-
DSRC with barriers is a type of non-contact times has a basket in which drivers insert coins
TCT where a vehicle does not have to stop at the in order to pay toll. This machine based on the
toll plaza. The vehicle only needs to reduce its dimensions as well as weight of the coin recognizes
speed in order to establish the contact and recogni- the value which should be paid.
tion through the On-Board Unit (OBU), in order Smart card TCT represents a memory card in
to receive the permission to pass through the gate which the details of a particular person and certain
(Glavić, 2013). amount of money are stored. The smartcard based
DSRC Multi-Lane Free Flow (MLFF) system, toll-gate automated system functions on the basis
have antennas above particular locations that of contact communication between the smart card
detect traffic flow and record the use on the OBU. and the reader. Whenever a smartcard is inserted
The technology used in MLFF system is designed into the smartcard reader, the reader will read the
so that the vehicles can maintain their speed and data stored in the card (Sridhar & Nagendra 2012).
change lanes while passing below the toll collection Smartphones TCT is still in its early stages of
portal (Glavić 2013). development. An example of mobile and Smart-
ANPR TCT system uses a stationary camera phone ETC integration is the m-Toll project.
for recording and identifying the registration plate The m-Toll project relies on the use of Smart-
numbers of vehicles passing through the toll plaza. phone WiFi connection to authenticate, validate
The identified registration plates are paired in the and charge road users without the need for any

804

TIS2017_Book.indb 804 2/8/2017 8:09:55 PM

 dedicated hardware for the end user. Servers passing an MTC lane and an ETC lane (Tseng
installed at the toll plaza can detect smartphones et al. 2014). According to the results of The Insti-
from a distance of 600 meters and deduct a par- tute of Transportation in Taiwan (2001), the aver-
ticular amount of money from the connected age toll transaction time for the ETC lane is 2.06
account through NFC (Technology options for the and for the MTC lanes it is 7.72 s.
European Electronic Toll Service 2014).
2.5 TCT aspects of user friendliness and mobility
2.2 TCT aspects of environmental sustainability
Users usually prefer technologies which are user
One of the aspects by which the toll technologies are friendly, that having high level of service and com-
different is the environmental aspect. Some technolo- fort during driving, which do not require stops or
gies, which require a speed change and stop the vehicle, speed reduction due to enforcement or fee reasons,
require more fuel consumption and therefore a higher and for which travel delays do not exist.
emissions of pollutants. For example, Coelho et al. The technologies which require minimum stop-
(2005) quantified emission impacts of toll facilities on ping or do not require stopping at all and where the
urban corridors in Lisbon, Portugal, and found that capacity of the toll lane is higher provide greater
61–80% CO2 reduction could be achieved by entirely mobility of the users especially in peak periods
switching MTC to ETC. Also, according to the data where queuing can be common on toll booths.
collected between 1999 and 2005 on the New Jersey
Turnpike in New Jersey, USA, Bartin et al. (2007) con-
ducted a before-and-after study to examine the benefits 3 METHODOLOGY
of ETC on reducing air pollution. Results indicated
that the overall emissions reduction of Carbon monox- Considering the semi-structured nature of the TCT
ide (CO), Hydrocarbon (HC), Nitrogen Oxide (NOx) selection problem, there is a need for a decision-
and particulate matter (PM10) was 10.1%, 11.2%, 3.4% support framework. MCA refers to the process of
and 13.0%, respectively, after the deployment of ETC. making decisions between a number of alterna-
Perez-Martinez et al. (2011) evaluated the energy con- tives, by defining the criteria and their weights. The
sumption and the associated CO2 emissions at three application of MCA thus results in the ranking of
toll systems in Spain and found that the type of toll alternatives, from the most to the least favorable,
collection is a major factor affecting energy efficiency. thus allowing comparison of alternatives.

2.3 TCT aspects of SMART system architecture 3.1 PROMETHEE decision-support framework
It is to be noted that sometimes apart from the main The MCA method selected for this research is Pref-
objective the same technology may also serve in sev- erence Ranking Organization Method for Enrich-
eral other manners. Utilization of ETC systems, which ment Evaluations (PROMETHEE) was used. The
represent smart technology, for navigation, theft pre- PROMETHEE method is one of the most recent
vention of automobiles and traffic surveillance are MCDA methods developed by Brans (1982) and
well examples of such secondary objectives. expanded by Brans et al. (1986). PROMETHEE is
primarily useful as the outranking method contrast
2.4 TCT aspects of system efficiency with other multi-criteria analysis methods (Barns
et al. 1986, Behzadian et al. 2010). One of the rea-
Different technologies have different toll collection sons for using PROMETHEE beside mathemati-
efficiency parameters. The efficiency of a particu- cal model is the fact that this method has excellent
lar toll can be measured by the various parameters, visualization of MCA process and results.
by the capacity of a toll lane, travel speed, number
of stops, the time spent waiting and others.
ETC lanes improve the speed and efficiency of 3.2 Definition of alternatives
traffic flow and save drivers time. Manual toll col- All toll technologies can be grouped into three
lection lanes handle only about 350 Vehicles Per groups (three alternatives), depending on whether
Hour (vph), and automated coin lanes handle they require stopping, speed reduction or free flow
about 500 vph. An ETC lane can process 1200 vph, driving of vehicles without any changes in vehicle
with ORT lanes allowing up to 1800 vehicles per speed. Alternative for MCA are:
hour (Tri-State Transportation Campaign 2004).
Since ETC allows toll transactions to be com- Alternative 1: Manual tolling, Automated coin
pleted while vehicle travels at a higher speed, they machine toll, vignette stickers,
may significantly reduce travel delay. Time saving smart card tolling, bar code toll-
is apparent from the difference between times for ing and tachograph tolling.

805

TIS2017_Book.indb 805 2/8/2017 8:09:55 PM

 Alternative 2: DSRC with barriers, INFRARED
with barriers, RFID with barriers,
ANPR with barriers
Alternative 3: DSRC MLFF, GNSS/CN, RFID
ORT, SMARTPHONES, ANPR.
The following figures describe speed time
dependences of each of 3 alternatives. The first
group, Alternative 1, require speed reduction from
about 120 km/h to 0 km/h and increase to around
120 km/h. This case is described by speed-time
curve (Figure 2). The second group, Alternative
2, require speed reduction from about 120 km/h Figure 4. Speed-time profile for the alternative 3.
to around 30–50 km/h and increase to around
in queue and stop and go conditions while queu-
120 km/h. In this case exists delays only because of
ing. This technology does not require enforcement
speed changes (Figure 3). In the case of Alterna-
for controlling toll users. Speed decrease to zero
tive 3 there was no delays because this technology
causes significant increase of VOC, TTC, GHG
do not require speed changes (Figure 4).
emission, lower LOS and decreased traffic safety.
The first group, Alternative 1, includes technol-
The second group, Alternative 2, consist of the
ogies which require stopping with possible waiting
technologies which require deceleration to lower
speed. This technology also do not require enforce-
ment for controlling toll users. The decrease of
speed to 30–50 km/h causes an increase of VOC,
TTC, GHG emission, lower LOS and traffic safety
compared to free flow speed.
The third group includes the technologies which do
not require changes in the vehicles speed. This tech-
nology does require possible stopping due to enforce-
Figure 1. Group of the alternatives. ment for controlling toll users. Speed decrease can
only be in the case of enforcement for controlling of
toll users. This can increase VOC, TTC, GHG emis-
sion and lower LOS for users that are controlled.

3.3 Selection of criteria

As Table 1 below shows, the list of criteria con-
sists of the following criteria: interoperability, effi-
ciency, Enforcement, level of service, traffic safety,
number of stops, an increase in travel time, visual
impact of defacing the ambient environment, air
pollution, noise, fair pricing for all users, the user
friendliness and aesthetic effect to the vehicle.
Figure 2. Speed-time profile for the alternative 1.
3.4 Criteria weights
Determination of the weights is an important step
in most MCA methods. Modified Digital Logic
(MDL) method of subjective evaluation of criteria
weights was used to determine the criteria weights.
Each expert individually evaluated weights of the
criteria and assigned points to the alternatives
in a chart. Then, mean values of the obtained
individual weights were used. For applications
in which the number of design criteria is fairly
large, assigning the importance weights among
multiple criteria simultaneously may be very dif-
ficult for the decision maker. The Modified Digital
Figure 3. Speed-time profile for the alternative 2. Logic (MDL) method is used to address this issue

806

TIS2017_Book.indb 806 2/8/2017 8:09:55 PM

 Table 1. The list and description of criteria. olds. Type of function and indifference and prefer-
ence thresholds are presented in Table 2.
Criteria Description

Interoperability Interoperability with other toll 3.6 Scoring of alternatives

collection systems (with other
countries). In the PROMETHEE the qualitative scale is
Efficiency This criterion includes the capac- defined by
ity of a lane, data rate and sys- • a number of ordered levels (from worst to best),
tem complexity.
• numerical values associated to these levels,
Enforcement Control of the vehicles.
• whether the numerical values should be mini-
Level of service The impact of the system on the
level of service of traffic flow. mized or maximized (scale orientation).
Number of stops The number of stops during the For purpose of this paper criteria are deter-
tolling. mined by min/max orientation, mostly scale 1–5
Traffic safety The impact of the system on the or 1–9 are used to evaluate alternatives by each
traffic safety. criterion. For each criteria expert from that field
Delays The increase in travel time that conducted scoring (Table 2).
each of the TCT causes.
Visual impact The effect of TCT on visually
ambient defacing the environmental
aesthetics.
Air pollution To what extent the systems con-
Table 2. PROMETHEE base matrix.
tributes to the pollution of the
environment.
Noise To what extent some systems
influence the increase of noise.
Fair pricing Whether the system is fair to all
users by ensuring payment for
what is used.
User friendliness The easiness of using the system.
Aesthetic effect to The effect on the aesthetics of the
the vehicle vehicle.

by suggesting pair-wise comparisons of criteria

(Dehghan-Menshadi et al. 2007). The decision
makers use digital scoring scheme of {1, 2 and 3}
to represent the less (1), equal (2), or more impor-
tant (3) criteria. After all pair-wise comparisons
are made, the MDL weights can be calculated as:

∑ C
n
jk
wj = k =1
, j and k { , , n} and j k
∑ ∑ C
n n
j =1 k =1 jk

If two criteria j and k are equally important,

then Cjk = Ckj = 2, otherwise Cjk = 3 and Ckj = 1 if
the criteria k is more important than the criteria j.
If the criteria k is less important than the criteria j,
then Cjk = 1and Ckj = 3.
The final values of the weights were obtained
by determining the mean value of individual MDL
weights for all 10 experts (Table 2).

3.5 Preference functions

Preference function are selected by analysing each
criterion with indifference and preference thresh-

807

TIS2017_Book.indb 807 2/8/2017 8:09:56 PM

 3.7 PROMETHEE base matrix Table 3. Ranking of the toll alternatives using PRO-
METHEE model.
All necessary input data regarding alternatives, cri-
teria’s, preference functions are shown in Table 2. Rank Action Phi

1 A3-Toll with free flow 0.266

4 RESULTS 2 A2-Toll with speed decrease 0.186
3 A1-Toll with stopping −0.452
4.1 PROMETHEE I & II ranking
4.1.1 PROMETHEE I Partial Ranking 4.1.3 GAIA visual analysis
The PROMETHEE I Partial Ranking is based The objective of GAIA is to describe the major
on the comparison of the leaving flow (Phi+) and features of the decision problems graphically:
the entering flow (Phi-). The PROMETHEE II • How much are actions different or similar to
Complete Ranking is based on the net flow Phi. each other?
Figure 5 shows PROMETHEE I Partial Ranking. • Which criteria are conflicting with each
The left column corresponds to the Phi+ scores other? Are there strong conflicts to solve? Are
and the right column to the Phi- scores. They are there groups of criteria expressing similar
oriented such that the best scores are upwards. The preferences?
alternatives are ranked: • What is the impact of the weighing of the crite-
– according to the Phi+ (A3, A2, A1); ria on the PROMETHEE rankings?
– according to the Phi- (A2, A3, A1). The GAIA plane is a descriptive complement
Interesting is to notice that according to Ph- to the PROMETHEE rankings. GAIA starts from
PROMETHEE I Partial Ranking (first place is a multidimensional representation of the decision
A2) and Ph- PROMETHEE I Partial Ranking problem with as many dimensions as the number
(first place is A3) we have different ranking. of criteria. A statistical method called the Prin-
cipal Components Analysis is used to reduce the
number of dimensions while minimizing the loss
4.1.2 PROMETHEE II Complete Ranking of information.
According to the PROMETHEE II Complete In Visual PROMETHEE three dimensions are
Ranking i.e. Ph net flows given in Table 3 it can computed:
be concluded that PROMETHEE ranking is A3 • U is the first principal component, it contains
(Phi = 0.226) followed by A2 (Phi = 0.186) and A1. the maximum possible quantity of information,
• V is the second principal component, providing
the maximum additional information orthogo-
nal to U,
• W is the third principal component, providing
the maximum additional information orthogo-
nal to both U and V.
Analyzing the GAIA plane given in Figure 6
one can conclude that tolling technologies with
free flow are closest to decision stick and in same
direction, closely followed by A2, while A1 is far-
thest and in the opposite direction from the deci-
sion stick.

4.2 Sensitivity analysis

Table 4 shows the final results of sensitivity analysis
through possible range of criteria weights needed
to keep the complete ranking unchanged and in
order to retain the first place on the list.
From the above sensitivity analysis, the main
conclusions are:
– Regarding sensitivity of the 1st place one can
conclude that 1st place is stable according to
Figure 5. PROMETHEE ranking. most of criteria

808

TIS2017_Book.indb 808 2/8/2017 8:09:57 PM

 Analyzing the results according to PRO-
METHEE II complete ranking, one can see that
the A3 (TCT which enable free flow of vehicles) are
considered as the optimal solution, especially hav-
ing advantage in ecological, social, level of service
and traffic safety criterion. The results also indi-
cate that the A3 is superior to A1 group of TCT.
The lack of analytically determined data in the
scoring procedure for each of the alternatives per
criterion is a limitation of this paper.
However, it is extremely difficult to analytically
determine all values of the alternatives per criteria.
In the situations where real data did not exist or
was not available, the problem was overcome by
including the experts from various fields who have
the knowledge and experience and are therefore
competent to subjectively evaluate the alternatives
per each of the criteria. Thus, the MCA objectivity
was realized through the expert evaluation using
the scales 1–5 or 1–9 (Table 2).
Figure 6. Ranking of the toll alternatives using GAIA The selection of the optimal toll collection sys-
visual analysis. tem represents an important decision considering
the fact that an inappropriate decision related to the
Table 4. Sensitivity analysis for weight of criteria selection of the optimal toll collection system can
regarding the complete ranking and the first place. lead to economic, environmental and social prob-
lems both for the present and future generations.
First place
remains Complete ranking
The selection of the optimal toll collection sys-
Criteria unchanged remains unchanged tem is significant for the road managers as well as
for the users and society as a whole.
Interoperability 11.5%–100% 39.6%–100% In addition, this paper provides a procedure for
Efficiency 0%–100% 0%–43.2% the proper selection one of the toll collection tech-
Enforcement 0%–100% 0%–100% nology from an optimal alternative. The overall
Level of service 0%–100% 0%–53.7% conclusion is that the Smartphone TCT, GSNN/
Number of stops 0%–100% 0%–43.2% CN TCT, DSRC MLFF, with RFID ORT are
Traffic safety 0%–100% 0%–100% greenest and most efficient TCT.
Delays 0%–83.3% 0%–69.2%
Visual impact 0%–100% 0%–100%
ambient 5.1 Future research
Air pollution 2.1%–100% 2.1%–100% There are several potential areas for further
Noise 1.1%–100% 1.1%–100% research. First, considering the fact that the smart-
Fair pricing 0%–100% 0%–100% phone technology is not yet a mature technology
User friendliness 1.1%–100% 1.1%–100% and it is in its development stage, there is a need
Aesthetic effect to 0%–100% 0%–100% for further testing and data collection on using
the vehicle
smartphone-based TCT.
Second, relation between self-driving vehicles
– Regarding sensitivity of the complete ranking and TCT need to be investigated, as emerging tech-
one can conclude that complete ranking is hav- nology. Currently there is no available data about
ing lower stability then 1st place ranking stability this issue.
but again according to most of criteria complete Third, when selecting the optimal toll collection
ranking is stable. technology, it is also extremely significant to con-
sider the attitudes of the highway and toll users. In
order to achieve this, a survey should be conducted
5 CONCLUSION to determine the attitudes of toll users.
Surveying toll users would provide the answers
The A2 is having advantage over A3 according to about the demands and expectations of the toll
PROMETHEE I ranking with negative flow Ph-. users. In other words, a series of opinions would
According to PROMETHEE I ranking with posi- be obtained which would significantly contribute
tive flow Ph+ A3 is having advantage over A2. to the ranking of toll collection systems from the

809

TIS2017_Book.indb 809 2/8/2017 8:09:57 PM

 users’ point of view. The authors of this paper Glavić, D. 2016. Analiza naplate putarine u Srbiji i
intend to make this the next step in their research uporedna analiza sa novim trendovima u EU i svetu.
on the topic. Second Serbian Road Congress, Belgrade, Serbia, 9–10.
June: 554–561.
Institute of Transportation. 2001. Highway Capac-
ity Manual in Taiwan, Institute of Transportation:
REFERENCES Taiwan.
M. Mladenovic, M. Abbas, S. Gopinath, & G. Faddoul.
Abbas, M., Mladenovic, M., Ganta, S., Kasaraneni, Y., 2013. Decision-Support System for Assessment of
& McGhee, C. 2013. Development and Use of Criti- Alternative Signal Controllers using Expert Knowl-
cal Functional Requirements for Controller Upgrade edge Acquisition, 92nd Annual Meeting of Transpor-
Decisions. Transportation Research Record: Journal of tation Research Board, Washington, D.C.
the Transportation Research Board, (2355), 83–92. M. Mladenovic, M., Abbas. 2013. Improved Decision-
Bartin, B., Mudigonda, S. & Ozbay, K. 2007. Impact of Support Making for Selecting Future Traffic Signal
electronic toll collection on air pollution levels-estima- Controllers using Expert-Knowledge Acquisition.
tion using microscopic simulation model of large-scale 16th International IEEE Annual Conference on Intel-
transportation network. Transp. Res. Rec. 2011: 68–77. ligent Transportation Systems, Delft, Netherlands.
Behzadian, M., Kazemzadeh, R.B. Albadvi, A. & Agh- Perez-Martinez, P.J., Ming, D., Dell’Asin, G. & Monzon,
dasi, M. 2010. PROMETHEE: A comprehensive A. 2011. Evaluation of the influence of toll systems on
literature review on methodologies and applica- energy consumption and CO2 emissions: a case study
tions. European Journal of Operational Research 200: of a Spanish highway. J. King Saud Univ. Sci. 23 (3):
198–215. 301–310.
Blythe, P. 1999. RFID for road tolling, road-use pricing Sharma, P. & Sharma, V. 2014. Electronic toll collection
and vehicle access control. Proceeding of the IEE Col- technologies: A state of art review. International Jour-
loquium on RFID Technology, London, UK, October nal of Advanced Research in Computer Science and
1999: 8/1–8/16. Software Engineering 4 (7): 621–625.
Brans, J.P. 1982. Lingenierie de la decision. Elaboration din- Shieh, W.Y., Lee, W.H., Tung, S.L. & Ho, C.D. 2005.
struments daide a la decision. Methode PROMETHEE. A novel architecture for multilane-free-flow electron-
In: Nadeau, R., Landry, M. (Eds.), Laide a la Decision: ic-toll-collection systems in the millimeter-wave range.
Nature, Instrument s et Perspectives Davenir. Presses de IEEE Transactions on Intelligent Transportation Sys-
Universite Laval, Quebec, Canada: 183–214. tems 6 (3): 294–301.
Brans, J.P., Vincke, Ph. & Mareschal, B. 1986. How to Sorensen, P.A. & Taylor, B.D. 2005. Review and synthe-
select and how to rank projects: The PROMETHEE sis of road-use metering and charging systems. Report
method. European Journal of Operational Research Commissioned by the Committee for the Study of the
24 (2): 228–238. Long-Term Viability of Fuel Taxes for Transportation
Catling, I. 2000. Road user charging using vehicle posi- Finance, UCLA Institute of Transportation Studies.
tioning systems. Proceedings of the 10th International Sridhar, V. & Nagendra, M. 2012. Smart card based
Conference on Road Transport Information and Con- toll gate automated system. International Journal of
trol, London, UK, 4–6 april 2000: 126–130. Advanced Research in Computer Engineering & Tech-
Charpentier, G. & Fremont, G. 2003. The ETC system nology 1 (5): 203–212.
for HGV on motorways in Germany: first lessons Staudinger, M. & Mulka, E. 2004. Electronic vehicle
after system opening. Proceedings of the European identification using active infrared light transmis-
Transport Conference (ETC), Strasbourg, France, sion. Proceedings of the ITS America 14th Annual
8–10 october 2003: 8. Meeting and Exposition, San Antonio, Texas, USA,
Coelho, M.C., Farias T.L., & Rouphail, N.M. 2005. 26–28 April: 357–376.
Measuring and modeling emission effects for toll Tri-State Transportation Campaign. 2004. The
facilities. Transp. Res. Rec. 1941: 134–144. Open Road. The Region’s Coming Toll Collection
Dehghan-Manshadia, D., Mahmudib, H., Abediana, Revolution.
A., Mahmudic, R. 2007. A novel method for materi- Tropartz, S., Horber, E. & Gruner, K. 1999. Experiences
als selection in mechanical design: Combination of and results from vehicle classification using infrared
non-linear normalization and a modified digital logic overhead laser sensors at toll plazas in New York City.
method. Materials & Design 28 (1): 8–15. Proceedings of the IEEE/IEEJ/JSAI International
Glavić, D. & Milenković, M. 2016. Comparative analy- Conference on Intelligent Transportation Systems,
sis of road tolling technologies. Second Serbian Road Tokyo, Japan, 5–8 October 1999: 686–691.
Congress, Belgrade, Serbia, 9–10. June: 562–568. Tseng, P.H., Lin, D.Y. & Chien, S. 2014. Investigating
Glavić, D. 2013. SWOT analysis of toll systems in Europe. the impact of highway electronic toll collection to the
Put i saobraćaj 59 (4): 21–30. external cost: A case study in Taiwan. Technological
Glavić, D. 2015. Analiza sistema naplate putarine u Srbiji Forecasting & Social Change 86: 265–272.
sa predlogom mera unapređenja. Put i saobraćaj 61(4):
56–62.

810

TIS2017_Book.indb 810 2/8/2017 8:09:57 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Traffic impact of automated truck platoons in highway exits

R. Mesa-Arango & A. Fabregas

Florida Institute of Technology, Melbourne FL, USA

ABSTRACT: This paper assesses the impacts on travel time and travel time reliability associated with
the operation of Autonomous Truck Platoons (ATPs) at freeway diverge areas. ATP technologies are
mature, and commercial deployments are expected in the next few years (earlier than autonomous passen-
ger cars). However, there is insufficient information about the impacts that ATPs will generate on the traf-
fic stream, especially around exit lanes. This paper proposes an interdisciplinary framework to integrate
ATPs into a microscopic traffic simulator. A combinatorial experiment is performed to test the impact of
four experimental variables on travel time and reliability, i.e., (i) traffic volume projections, (ii) ATP pen-
etration rates, (iii) ATP sizes, and (iv) ATP gaps. Two performance metrics are employed and statistically
tested to quantify the impact of these variables on (a) though- and (b) divergent-traffic. Numerical results
demonstrate the significance and impact of experimental variables on travel time and reliability.

1 INTRODUCTION 2016). ATP demonstrations, like the European

Truck Platooning Challenge (ETPC) (Alkim et al.
This research investigates the impact of Automated 2016), have been motivated by public agencies to
Truck Platoons (ATPs) on the traffic stream in the understand the interdisciplinary challenges associ-
vicinity of a freeway diverge area. The impacts on ated with this new technology. Given the imminent
average travel time and travel time reliability are deployment of ATPs for commercial purposes,
considered for freeway users that drive through and public agencies in the U.S. are energetically con-
turn right onto a right-side off-ramp. Truck pla- ducting projects to demonstrate their capabilities
toons are groups of multiple trucks traveling close in connected freight corridors (Transportation
behind each other to reduce the power required to Pooled Fund (TPF) Program 2016), and improve
overcome aerodynamic resistance, which improves their understanding on the system (Bevly et al.
gas consumption and, hence, reduces operational 2015, National Cooperative Highway Research
costs, and undesired emissions. Although closer Program (NCHRP) 2015).
inter-truck gaps represent higher savings, small Many researchers have developed studies to
gaps are unsafe and inconvenient for human driv- anticipate the impact of ATPs from different per-
ers. However, new automation technologies facili- spectives, i.e., operations research (Saeednia &
tate the implementation of ATPs with shorter gaps Menendez 2016, Liang et al. 2016, Van De Hoef
and higher financial benefits for the trucking com- et al. 2015, Larsson et al. 2015, Larson et al. 2015),
panies. Simulations and testbeds report the signifi- game theory (Farokhi & Johansson 2015), human
cant energy savings that can be achieved with ATP factors (Zheng et al. 2015, Skottke et al. 2014), and
systems, i.e., between 14% and 25% (Tsugawa et al. business/industry (Bevly et al. 2015, Shladover
2011, Vegendla et al. 2015, Bevly et al. 2015, Nowa- et al. 2015). However, few is known about the traf-
kowski et al. 2015). fic impacts associated with the insertion of ATPs
Within the last three decades, several ATP- into the traffic stream. Although initial works
related projects have been developed around the suggest that increased ATP penetration rates will
world, i.e., in Europe, Japan, and the United States improve traffic flow stability, highway capacity,
(U.S.), to test the feasibility of this technology and and critical/jam density (Bevly et al. 2015, Arem
measure its benefits (Tsugawa et al. 2011, Nowa- et al. 2006, Schakel et al. 2010, Arnaout & Bowl-
kowski et al. 2015, Bishop et al. 2015, Shladover ing 2014), there is uncertainty about these results
2012, Kunze et al. 2009, Kunze et al. 2011, Jans- because a clear microscopic traffic simulation
sen et al. 2015, Shladover 2010, Nowakowski et al. framework to model the interaction between traf-
2015). Furthermore, in the last two years several fic and ATPs under different conditions, e.g., ATP
truck manufacturers have demonstrated their capa- size, gap, penetration rate, among others, has not
bilities to develop and operate ATPs in real-world been formally presented. Developing a frame-
settings (Markoff 2016, Abt 2016, Alkim et al. work that facilitates the design of scenarios for

811

TIS2017_Book.indb 811 2/8/2017 8:09:59 PM

 different ATP conditions is a fundamental require- by Bishop et al. (2015), Nowakowski et al. (2015),
ment to confirm the traffic benefits claimed in pre- and Shladover (2012), present a detailed review of
vious studies and test scenarios related to the new previous ATP projects around the world, which are
challenges that ATPs will impose on decision mak- summarized next. ATP research efforts in Europe
ers, i.e., traffic operation managers, and transporta- were initiated and promoted by a number of ATP
tion planners. One of this scenarios correspond to projects, e.g., CHAUFFEUR, SARTRE, KON-
the conditions that traffic users will face at freeway VOI (Kunze et al. 2009 and 2011), and COMPAN-
diverge areas. This is important because when large ION. Likewise, the Netherlands Organization for
number of ATPs with multiple coupled trucks and Applied Scientific Research (TNO) (Janssen et al.
short inter-truck gap drive close to off-ramp exits, 2015) and the Eindhoven University of Technol-
the other users might experience reduced gaps that ogy (TU/e) have consistently promoted research on
will generate dangerous conditions or delays in Cooperative Adaptive Cruise Control (CACC) for
order to take the right exit at the right time. On the trucking operations in the Netherlands. In Japan,
other hand, large penetration rates might reduce the Energy ITS project (Tsugawa et al. 2011) was
the number of uncoupled trucks on the road and developed to demonstrate the benefits of ATPs—
hence, balance the negative vision stated above. among other Intelligent Transportation Systems
Given the complexity of this interaction, simula- (ITS)—for energy savings and global warming
tion experiments are required to assess the impacts prevention. Moreover, intensive research on ATPs
of ATPs on travel time and travel time reliability has been conducted by the California Partners for
for multiple users under these conditions. Advanced Transportation Technology (PATH)
This paper is one of the few addressing the traf- in the U.S. for about 2 decades (Shladover 2010),
fic impacts of ATPs and contributes to related lit- Nowakowski et al. 2015). These global efforts have
erature by (1) proposing a novel interdisciplinary demonstrated the benefits, feasibility, and techni-
framework to integrate the complex operation cal requirements of ATPs. Likewise, models and
of ATPs into a microscopic traffic simulator and real-world demonstrations have motivated recent
conduct an efficient combinatorial experiment to implementations of the system.
assess the validity of experimental variables, and Truck manufacturers and governments are
(2) providing meaningful insights about the opera- engaged in the implementation of autonomous
tional aspects of ATPs in the vicinity of freeway trucks and ATPs. In 2015, Daimler—one of the
diverge areas. larger truck manufacturers in the world (Mercedes-
The paper is organized as follow. The introduc- Benz, Freightliner Trucks, Western Star Trucks,
tory section illustrates and motivates the current among others brands)—introduced the first
work. Then, a comprehensive literature review on licensed autonomous commercial truck to operate
the state of research and practice related to ATPs is in the U.S. (Nevada), i.e., the Freightliner Inspi-
presented. Afterwards, the methodological frame- ration (Markoff 2016). Later on, In March 2016,
work to incorporate ATPs into a microscopic traf- Daimler demonstrated its ATP capabilities by run-
fic simulation is proposed. Next, a combinatorial ning a platoon of three heavy-duty Mercedes-Benz
experiment is performed to assess the impact of Actros trucks on a public stretch of the Autobahn
experimental variables on travel time and reliabil- highway system in Germany (Abt 2016). In the
ity. Finally, conclusions and future research direc- next month (April 2016), Rijkswaterstaat (traffic
tions are provided. agency in the Netherlands), organized the ETPC
to assess the current operations and requirements
related to ATPs. As part of the challenge, several
2 LITERATURE REVIEW truck manufacturers (i.e., DAF Trucks, Daimler,
Iveco, MAN Truck & Bus, Scania, and Volvo)
This section presents a comprehensive literature drove their own ATPs from Sweden, Germany and
review on ATPs, which includes previous and cur- Belgium to Rotterdam (Netherlands). The experi-
rent implementations and projects, benefits and ment collected and analyzed relevant information
challenges of ATPs, and interdisciplinary research about ATPs in the real world (Alkim et al. 2016).
initiatives. Subsequently, the few works studying Additionally, companies in the U.S. are actively
the traffic impacts of ATPs are summarized, which developing business based on ATP concepts. For
shows the gap on appropriate simulation frame- example, Otto is a start-up company led by former
works that integrate ATPs and regular traffic, and Google employees and engineers that develops
indicates the lack of research to quantify the traffic technologies to provide autonomous upgrades
impacts of ATPs on freeway diverge areas. to regular trucks. Likewise, Peloton offers ATP
The ATP concept has evolved from prelimi- services and technologies to the trucking indus-
nary tests—about two decades ago—to real world try. The U.S. government is also very interested in
implementations in the last few months. The works the future of ATPs. Five jurisdictions allow to test

812

TIS2017_Book.indb 812 2/8/2017 8:09:59 PM

 and operate ATPs (i.e., California, Nevada, Michi- platoon formation and find that a hybrid approach
gan, Florida, and the District of Columbia), and is the best strategy. Liang et al. (2016) propose
there are several projects studying their opportu- an iterative pairwise coordination method that
nities. For example, the California, Arizona, New sequentially joins the closest trucks and platoons
Mexico, and Texas DOTs are planning to demon- to improve fuel efficiencies on the fly. Van De
strate ATP capabilities for the I-10 Western Con- Hoef et al. (2015) propose optimization methods
nected Freight Corridor Project in the next few for platoon coordination based on pairwise clus-
years (TPF (Transportation Pooled Fund (TPF) tering. Larson et al. (2015) propose a distributed
Program 2016)). Similarly, the National Coopera- network of controllers to enhance platooning and
tive Highway Research Program (NCHRP) is con- maximize fuel savings. Larson et al. (2015) com-
ducting a project to understand the challenges of bine platooning and vehicle routing to minimize
connected/autonomous vehicles in truck freight the total fuel consumption of trucks dispatched
operations (NCHRP 2015). Additionally, Bevly between different origins and destinations in a net-
et al. (2015) recently released the first phase of a work. Similarly, Farokhi & Johansson (2015) use
comprehensive study on the commercial feasibil- game theoretical approaches to study the strategic
ity of driver assistive truck platooning, part of interaction of passenger cars and trucks under
the Federal Highway Administration (FHWA)’s truck platooning incentives to improve congestion.
Exploratory Advanced Research Program. The Most of these works focus on the optimization
substantial interest of researchers, governments, of individual firms and do not account for social
and companies on ATP systems is explained by the benefits/impacts. Although Farokhi & Johans-
prominent benefits that this technology offers to son (2015) approach this perspective, their games
business and society, which are summarized next. are broad and do not account for specific traffic
When trucks drive close next to each other, i.e., interactions. Other considerations include driver
with small inter-truck gaps, the power required to discomfort and unexpected behavior related to
overcome the aerodynamic resistance (drag) gen- small gaps (Zheng et al. 2015, Skottke et al. 2014),
erated by the air flow around the vehicles is less adoption willingness by trucking and logistics
than the power needed by trucks that are driven firms (Shladover et al. 2015, Bevly et al. 2015), and
in isolation, which generates fuel savings and emis- human factors important for successful implemen-
sion reductions that are more critical at higher tations (Nowakowski et al. 2015). However, scarce
speeds. Based on the results from the Energy ITS attention have been placed to the impact of ATPs
project in Japan, Tsugawa et al. (2011) indicate that on the surrounding traffic.
a platoon of three trucks driving at 80 km/h with Public agencies need to anticipate the traffic
a 10 m. gap has a 14% average fuel consumption impacts associated with ATPs. Janssen et al. (2015)
improvement. Likewise, an ATP penetration rate suggest that appropriate conditions need to be pro-
of 40% along an expressway is related to a 2.1% vided for the interaction between ATPs and other
reduction in CO2 emissions for 10 m. gaps, and drives. Nowakowski et al (2015) indicate that sev-
4.8% reduction for 4 m. gaps. Similarly, Vegendla eral variables of the ATP system will have impacts
et al. (2015) study the aerodynamic influence of on traffic, e.g., number of trucks allowed in an
five different platooning configurations and show ATP, determination of exclusive ATP lanes, ATP
with simulations that trucks obtain the highest coupling/decoupling procedures, among others.
team fuel savings (up to 23%) when they travel one Likewise, the ETPC demonstrates the requirements
behind another (savings are reduced as they drive to enhance the ATP-traffic interaction (Alkim
farther). Moreover, Bevly et al. (2015) demonstrate et al. 2016). Few works have studied the impact
that the closer the platoon distance, the better of ATPs, and related technologies, in the traffic
the fuel savings. According to Nowakowski et al. stream. Arem et al. (2006) use microscopic simula-
(2015), shorter following gaps at highway speeds tions of mixed traffic (passenger cars and trucks)
can generate energy savings as high as 20% to 25% with CACC. Although they conclude that CACC
for large trucks. Other ATP benefits include traf- improves traffic efficiency, the authors acknowl-
fic safety improvements (Kunze et al. 2011), traffic edge that extensive research is needed to antici-
flow benefits (Bevly et al. 2015), and congestion pate traffic impacts. Schakel et al (2010) highlight
mitigation (Farokhi & Johansson 2015). the benefits of CACC for traffic flow stability but
Furthermore, the interdisciplinary nature warn that fast shockwaves can be problematic for
of ATP systems have captured the attention of human drivers. Arnaut & Bowling (2014) develop a
researchers from multiple perspectives. For exam- traffic simulation to assess improvements related to
ple, interesting problems related to the operation mixed-traffic using CACC at different penetration
and optimization of ATPs have been approached rates, and conclude that CACC improves traffic at
by several authors. Saeednia & Menendez (2016) penetration rates above 40%. Most of the optimiza-
study acceleration/deceleration strategies for tion methods proposed in previous literature do not

813

TIS2017_Book.indb 813 2/8/2017 8:09:59 PM

 consider the interaction between traffic and ATPs to the Heavy-Duty Truck (HDT) driver and the
e.g., Saeednia & Menendez (2016), Larsson et al. HDT will behave as regular simulated vehicle. In
(2015). Only two studies investigate the impacts of the algorithm, this will yield vehicle control back
ATPs on the traffic volume. Bevly et al. (2015) use to VISSIM. ATP control was performed using a
CORSIM to show that two-truck platoons improve proportional controller taking into consideration
traffic at penetration rates above 60%. Secondly, acceleration, speed and gap between ATPs in the
Deng & Burghout (2016) compare the impacts of platoon. These data are passed from VISSIM to the
short gaps on the overall traffic flow performance. DLL at the selected time step which was chosen as
However, the specific characteristics of the ATPs 1/20 seconds (flow 3, Figure 1). The ATP control
considered in their derivations are unclear. action runs the gap control algorithm and returns
In order to develop accurate models and experi- the selected acceleration for execution in VISSIM
ments to determine traffic impacts, it is very (flow 2, Figure 1) When a vehicle reaches the end
important to understand the specific features that of the simulated segment is removed automatically
differentiate ATPs from the rest of the traffic. For by VISSIM.
example, the expected platoon size (number of
trucks clustered in an ATP) is uncertain. Typical
3.2 Traffic pattern generation
ATP sizes include two to three trucks (Tsugawa
et al. 2011, Bevly et al. 2015, Janssen et al. 2015, Traffic pattern for ATPs was created using collected
Shladover 2010, Zheng et al. 2014) but are envi- traffic counts from the analysis segment. The traffic
sioned to be as large as five to ten trucks (Abt pattern generation process is depicted in Figure 2-a.
2016, TPF Program 2016). The process starts with an ATP generation pattern
Similarly, the magnitude of the gap between base on inter-platoon times and platoon size on an
consecutive trucks is still unclear. Implementations exclusive modified lane. The arrival creation lane
and experiments have considered inter-truck gaps allowed the ATPs to quickly form a platoon and
that range from 3 m. 15 m. (Tsugawa et al. 2011, move onto the traffic stream using a parallel merge
Shladover 2010, Abt 2016, C Nowakowski et al. area, i.e., platoon formation stage in Figure 2-a.
2015, Zheng et al. 2014). Nowakowski et al. (2015) Such initial process guarantees that the platoon will
indicate that drivers feel uncomfortable 15 m. gaps form and stay formed as part of the initial traffic
and unsafe at less than 7 m. gaps. However, opti- pattern. The input traffic pattern is then formed
mal fuel savings are achieved around 6 to 8 m. and given a road segment to achieve stability. Once
The next section describes the methodology fol-
lowed to study the traffic impact of ATPs on travel
time and travel time for drivers in the vicinity of
highway exits based on the lack of previous studies
approaching this problem.

3 METHODOLOGY

The analysis framework used traffic microsimula-

tion with automated scenario generation to model
and analyze the segment near a freeway exit. This
section provides an overview of the technical
aspects involved in ATP simulation. The microsim- Figure 1. Control framework for ATPs.
ulation tool chosen in the study is VISSIM, which
gives the possibility to control individual vehicles
via Dynamic Link Library (DLL).

3.1 ATP control

The overview of the control process is described in
Figure 1. The process starts with an arrival genera-
tion (flow 1 in Figure 1) or a time step event from
VISSIM. Connected vehicle applications use the
dedicated short range (DSRC) frequency with a
communication range about 300 ft. It was assumed
that if the trucks get separated beyond the DSCR Figure 2. ATP traffic pattern generation process and
range connection is lost and control is returned analysis segment.

814

TIS2017_Book.indb 814 2/8/2017 8:09:59 PM

 the traffic contains the ATPs and it is in steady con- 4 RESULTS AND DISCUSSION
ditions, it enters the analysis segment. The analysis
segment was chosen as two miles upstream of the This section describes the data inputs required to
exit ramp. An illustration of a truck platoon near implement the traffic simulation experiment, pro-
the exit ramp is presented in Figure 2-b. vides a case study to test its implementation, and
discusses meaningful insights obtained from the
computations.
3.3 Computational experiment
A 2.5-mile 2-lane segment of State Road
A microscopic simulation experiment is proposed 528 (Martin Andersen Beachline Expressway)
based on several scenarios generated from a base- approaching Exit 20 is selected for the analysis.
line calibrated with real-time traffic data. Data from traffic counts reported by the Florida
The main variables used to test the impact of Department of Transportation (FDOT 2015) are
ATPs on travel time and travel time reliability are collected to develop the Directional Design Hourly
(i) traffic volume projections (Passenger Cars (PCs) Volume (DDHV) for the base scenario, where
and HDTs), (ii) ATP penetration rates, (iii) ATP DDHV = AADT × K × D, AADT is the average
distance gaps, and (iv) ATP sizes. Furthermore, annual daily traffic, K is the design hour factor, and
the following pseudocode describes the execution D is the directional design hourly factor. The cor-
of the simulation-based numerical experiment. responding truck DDHV (DDHVT) is computed
as DDHVT = DDHV × T, where T is the percent-
Step Description
age of trucks. FDOT (2015) indicates that for this
0-01 Define analysis segment;
intersection AADT = 20,000 vehicles per day, K =
0-02 Compile traffic data from upstream/down-
0.094, D = 0.632, and T = 0.08. Therefore, DDHV
stream sensors;
= 118 vehicles per hour, and DDHVT = 95 trucks
0-03 Calibrate baseline traffic simulation;
per hour. Additionally, FDOT (2015) reports that
0-04 Develop and integrate ATP behavior in
34.01% of the traffic takes the Exit 20 off-ramp.
traffic simulator;
The previous inputs are used to generate a base
0-05 For all traffic flow projections
scenario calibrated in VISSIM, where traffic vari-
0-06 For all ATP penetration rates
ables are fine-tuned to replicate realistic traffic
0-07 For all ATP distance gaps
behavior. Likewise, it is assumed that the ATPs only
0-08 For all ATP sizes
drive through, i.e., do not diverge on the off-ramp.
0-09 Run traffic simulation;
Four sets of experimental variables are selected
0-10 Compile travel times and delays;
to analyze the changes in travel time and travel time
0-11 End;
reliability (delays) as a function of ATP deploy-
0-12 End;
ments, i.e., (i) traffic volume projections Xi = {100,
0-13 End;
150, 200, 250}(%), (ii) ATP market penetration rates
0-14 End;
Xii = {20, 60, 100}(%), (iii) ATP sizes Xiii = {2, 3, 5,
0-15 Return travel time & delays.
10}(trucks), and (iv) ATP gaps Xiv = {5, 10, 15}(m).
Intuitively, a representative highway section with Thus, a total of M = 144 scenarios are considered
and off-ramp access is selected on the study region. in the experiment. The computational experiment is
Then traffic counts are compiled from downstream designed in Visual Basic, and it takes about 12 hours
and upstream sensors. Such data is used to calibrate to run all the cases. A total of 137,367 observations
the baseline in the traffic simulator with only trucks are collected. Each observation describes a simu-
and passenger cars. Furthermore, the ATP driving lated vehicle trip in a specific simulated scenario
behavior needs to be developed and implemented with the corresponding travel time and delay. Three
in the traffic simulator. Subsequently, a number of vehicle types are considered, i,e, PCs, HVTs (not in
scenarios are tested under multiple combinations ATP formation), and HDTs operating in ATPs.
of the experimental variables. Finally, outputs The dataset is used to estimate four Ordinary
associated with travel time and delays are collected Least Squares (OLS) models that test the impact
for subsequent analyzes. Delays are used to quan- of experimental variables on travel time and reli-
tify reliability, i.e., the higher the delays, the most ability (described by delays) for off-ramp users
unreliable the operation. (Table 1) and through traffic (Table 2).
As recognized by Bevly et al. (2015), develop- The numerical results in Table 1 show how differ-
ing the ATP driving behavior and implementing it ent ATP attributes impact travel time and delays for
into the traffic simulation is a major challenge for multiple off-ramp users. The average travel time for
ATP research. However, such challenge is properly off-ramp PCs in the experiment is 87.22 second, and
addressed in this paper and meaningful insight are in average divergent HDTs spend 8.14 more seconds
obtained from the numerical experiment described in the system, which is associated with the opera-
in the next section. tional characteristics of each user type, i.e., HDTs

815

TIS2017_Book.indb 815 2/8/2017 8:10:00 PM

 Table 1. OLS model for off-ramp travel time and Interestingly, the most critical ATP attribute
delays. impacting travel time is its market penetration. In
average a 1% increment in ATP market penetration
Travel Time Delay reduces off-ramp-related travel time by 2.06 seconds
Type Variable Coeff. t Stat Coeff. t Stat for PCs, and 4.06 seconds for HDTs. Such results
highlight the potential benefits that will be perceived
Constant 87.22 516.46 5.07 21.83 when platooned HDTs free additional highway space
PC ATP Size 0.04 4.05 0.04 2.92 that can be efficiently used by other vehicle classes.
PC ATP Gap 0.01 1.86 0.01 1.35 The benefit might be larger for single HDTs (not in
PC Traffic Intensity 2.83 45.32 2.91 33.98 ATP formation) because their speeds will be similar
PC ATP penetration −2.06 −20.50 −2.23 −16.12 to the speeds of the ATPs and time losses associated
HDT Constant 8.14 10.03 −6.34 −5.68 with maneuvers within ATPs are smaller. In general
HDT ATP Size 0.03 0.56 0.03 0.39 they might just follow leading ATPs until reaching
HDT ATP Gap 0.01 0.34 0.01 0.24 their exits. On the other hand, the higher speed of
HDT Traffic Intensity 1.53 5.13 2.23 5.44 PCs will generate more circumstances in which they
HDT ATP penetration −4.06 −4.90 −1.17 −1.03 will have to decelerate to change lanes, overcome
Observations 44487 44487
ATPs, and break through ATPs to reach highway
Multiple R 0.277 0.228
exits on time. Still the benefits obtained from the
Adjusted R Square 0.076 0.052 additional space gained after re-organizing HDTs
into ATPs are significant. Furthermore, a 1% incre-
ment in market penetration in average reduces delays
Table 2. OLS model for through-traffic travel time and
for PCs by 2.23 seconds and trucks by 1.17 seconds,
delays. and aligns with the insights presented for travel time.
The previous claims are also supported by the
Travel Time Delay results associated with ATP size and ATP gap. In
average each additional truck in ATP formation
Type Variable Coeff. t Stat Coeff. t Stat increases PC travel time and delays by 0.04 seconds.
Likewise, each meter of inter-truck gap distance
Constant 81.43 632.80 −1.07 −8.60 increases PC travel time and delays by 0.01 second.
PC ATP Size 0.02 2.77 0.02 2.85 As ATPs cluster more trucks, the interaction between
PC ATP Gap 0.02 2.58 0.02 2.61 the platoons and divergent PCs will generate more
PC Traffic Intensity 4.01 84.74 3.62 79.06 delays because it will be harder to break through the
PC ATP penetration −1.40 −18.18 −1.43 −19.13 ATP as more HDTs will have to adjust their speeds.
HDT Constant 13.27 21.34 −0.33 −0.55 Larger gaps provide more opportunities for PCs to
HDT ATP Size 0.02 0.64 0.02 0.66 change lanes through ATPs, which might promote
HDT ATP Gap 0.01 0.20 0.01 0.21 risky behaviors, and reduces overall travel times for
HDT Traffic Intensity 1.17 5.22 1.53 7.01
PCs and even ATPs. On the other hand, very small
HDT ATP penetration −1.71 −2.67 1.48 2.38 gaps (unacceptable for lane changing) might demo-
ATP Constant 5.04 10.66 4.64 10.14
tivate PCs to pass through the ATP and encourage
ATP ATP Size 0.16 6.10 0.07 2.55
drivers to anticipate their exit maneuvers taking
ATP ATP Gap 0.01 0.73 0.01 0.74
advantage of the additional longitudinal space gen-
ATP Traffic Intensity 2.48 16.14 2.34 15.68
erated by the ATPs. These results motivate further
ATP ATP penetration −0.63 −2.09 −1.07 −3.68
investigations to assess specific instances where the
Observations 92879 92879 PC-ATP interaction will generate delays, inefficien-
Multiple R 0.352 0.291 cies, and unsafe conditions. Although current results
Adjusted R Square 0.124 0.084 suggest similar conclusions for trucks, the low sig-
nificance of ATP size and gap on travel time and
delays (assessed by their low positive t-stats) indicate
are overall slower than PCs. Furthermore, the aver- that these variable might have a negligible effect on
age delay is 5.07 for PCs and 5.07–1.17 = 3.90 sec- travel times and delays for regular HDTs, which is
onds for trucks, which can be attributed to the lower expected as they are expected to follow ATPs rather
desired speeds associated with HDTs. than passing through them to take highway exits.
Traffic growth has differential impacts on travel Furthermore, Table 2 shows how different ATP
time for each user type, i.e., in average 2.83 seconds attributes impact travel time and delays for the
per 1% traffic increment for PCs and 1.53 seconds through traffic. The average travel time for PCs in the
for HDTs, which is expected from classic traffic experimental environment is 81.43 seconds. In general
theory. By looking at delays, a 1% increment in HDTs and ATPs are slower than PCs. While HDTs
traffic volume in average represents 2.91 seconds that drive through in average spend 13.27 seconds
of delay for PCs and 2.23 seconds for HDTs. more than PCs, ATPs in average spend just 5.04 more

816

TIS2017_Book.indb 816 2/8/2017 8:10:00 PM

 seconds. This results highlight the overall time-savings An interdisciplinary method is proposed to
achieved by HDTs engaged in ATPs. Reduced travel incorporate ATPs into a microscopic traffic simu-
times for HDTs are amplified because each vehicle lator and a combinatorial experiment assesses the
needs to account for specific behavioral characteris- impacts of (i) traffic volume projections, (ii) ATP
tics of the human driver. On the other hand, only one market penetration rates, (iii) ATP sizes, and (iv)
driver will lead an ATP, and, therefore, the travel time ATP gaps, with four OLS models for travel time
for the followers will be enhanced. However, from the and reliability (delays).
classic traffic-flow-theory perspective, ATPs will face Traffic volume growth deteriorate travel time
higher delays because each follower will not have and delays for all user/movement types. Increased
the possibility to improve their current speed as they usage of ATPs improves travel time for all user/
are forced to follow the leader. In the experiment, movement types. As the number of vehicles in the
each vehicle in an ATP experiences an average delay ATPs grows, travel times and delays are slightly
of 4.66–1.07 = 3.59 seconds. However, in average PCs deteriorated due to impedance on lateral move-
reduce delays by 1.07 seconds and HDTs obtain an ments and a resulting loss on automated efficiencies
additional 0.33-second reduction because they can of the ATP. Although inter-truck ATP gap has the
move freely, are more flexible, and have better oppor- lowest impact on travel time and delays, it is espe-
tunities to reduce delays based on their independent cially important for PCs in highway exits. Larger
decisions. Furthermore, traffic volume growth incre- gaps will represent a dilemma for users that might
ments travel times for all user types, i.e., a 1% growth riskily accept them for lane changing. On the other
in overall traffic increases travel time and delays by hand, smaller gaps might force traffic to plan ahead
4.01 and 3.62 seconds for PCs, 1.17 and 1.53 seconds and anticipate maneuvers taking advantage of the
for HDTs, and 2.48 and 2.34 seconds for ATPs. The additional longitudinal space freed by ATPs.
impact is more dramatic for PCs because they tend to The paper is one of the first efforts to study the
drive faster, so increased traffic intensity will generate impact of ATPs on freeway traffic. The main con-
more congestion with higher travel time and delays. tributions of this research include: (1) providing a
On the other, travel times and delays will be higher for novel interdisciplinary framework to integrate the
ATPs than for HDTs because they are forced to travel complex operation of ATPs into a microscopic
as a group and have low flexibility to avoid undesir- traffic simulator and conduct an efficient combi-
able circumstances generated by increased traffic. natorial experiment to assess the validity of experi-
From the market penetration perspective, the over- mental variables, and (2) providing meaningful
all through-traffic travel time will be reduced when insights about the operational aspects of ATPs in
larger fractions of HDTs operate as ATPs. The aver- the vicinity of freeway exits.
age travel time reductions per 1% increment in ATP Future research directions include running mul-
market penetrations are 1.40, 1.71, and 0.63 seconds tiple simulations associated with each scenario of
for PCs, HDTs, and ATPs respectively. In average, the combinatorial experiment, which is achievable
delays are mainly reduced for PCs and ATPs by 1.43 with multiple random seeds for each instance of the
and 1.07 seconds per 1% increment in ATP market experimental variables; generalizing the results with
penetration. PCs will have more longitudinal space additional experiments over different geometries,
to maneuver and flexibility to reduce their delays and traffic behaviors, and traffic volumes; enhance ATP
vehicles in ATPs will benefits by driving closer with- behavior to allow lane changing; incorporating
out human driver inefficiencies. Although HDTs will Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastruc-
perceive lower travel times, in average their delays will ture (V2I) communication and coordination for
be increased by 1.48 seconds per 1% increment in PCs, HDTs, and ATPs; and developing a systems-
ATP market penetration. This might happen because of-systems framework to understand the interac-
as PCs gain more mobility, they might slightly affect tion between technical, behavioral, institutional,
HDTs. In general, additional entities in an ATP and and physical components encompassing the ATPs.
larger inter-truck gaps will generate low increments
on the travel for all users.
These numerical experiment provides interesting REFERENCES
insight on future issues and opportunities related to
the operation of ATP in freeway diverge areas. A sum- Abt, N., 2016. Daimler’s Autonomous Platoon Debuts
mary of the work and conclusions are provided next. on Germany’s Autobahn. Transport Topics. Avail-
able at: http://www.ttnews.com/articles/basetemplate.
aspx?storyid = 41377.
Alkim, T. et al., 2016. European Truck Platooning Challenge
5 CONCLUSION 2016. Creating next generation mobility. Lessons Learnt,
Rijkswaterstaat, RDW, and the Ministry of Infrastruc-
This paper studies the impact of ATPs on travel ture and the Environment. Netherlands. Available at:
time and reliability (delays) on the vicinity of a https://www.eutruckplatooning.com/PageByID.aspx?se
freeway diverge area. ctionID = 131542&contentPageID = 529927.

817

TIS2017_Book.indb 817 2/8/2017 8:10:00 PM

 Arem, B. van, Driel, C.J.G. van & Visser, R., 2006. The Markoff, J., 2016. Want to Buy a Self-Driving Car? Big-Rig
Impact of Cooperative Adaptive Cruise Control on Trucks May Come First. The New York Times, p.B1.
Traffic-Flow Characteristics. IEEE Transactions on National Cooperative Highway Research Program
Intelligent Transportation Systems, 7(4), pp.429–436. NCHRP, 2015. Challenges to CV and AV Application
Arnaout, G.M. & Bowling, S., 2014. A Progressive in Truck Freight Operations. Transportation Research
Deployment Strategy for Cooperative Adaptive Board, (2015). Available at: http://apps.trb.org/cms-
Cruise Control to Improve Traffic Dynamics. Inter- feed/TRBNetProjectDisplay.asp?ProjectID = 3936
national Journal of Automation and Computing, 11(1), [Accessed July 29, 2016].
pp.10–18. Available at: http://dx.doi.org/10.1007/ Nowakowski, C. et al., 2015. Cooperative Adaptive Cruise
s11633–014–0760–2. Control (CACC) for Truck Platooning: Operational
Bevly, D. et al., 2015. Heavy Truck Cooperative Adaptive Concept Alternatives. eScholarship, University of Cali-
Cruise Control: Evaluation, Testing, and Stakeholder fornia, pp.1–37. Available at: http://www.escholarship.
Engagement for Near Term Deployment. Technical org/uc/item/7jf9n5wm.
Report - Federal Highway Administration, pp.1–135. Nowakowski, C., Shladover, S.E. & Tan, H.-S., 2015.
Bishop, R. et al., 2015. White Paper: Automated Driv- Heavy Vehicle Automation: Human Factors Lessons
ing & Platooning - Issues and Opportunities. ATA Learned. Procedia Manufacturing, 3, pp. 2945–2952.
Technology & Maintenance Council, Future Truck Saeednia, M. & Menendez, M., 2016. Analysis of Strat-
Program, Automated Driving and Platooning Task egies for Truck Platooning: A Hybrid Strategy. In
Force, (2015), pp.1–48. Available at: http://orfe.prin- Transportation Research Board 95th Annual Meeting.
ceton.edu/∼alaink/SmartDrivingCars/ITFVHA15/ pp. 16–5380.
ITFVHA15_USA_FutureTruck_ADP_TF_White- Schakel, W.J., Arem, B. van & Netten, B.D., 2010. Effects of
Paper_Draft_Final_TF_Approved_Sept_2015.pdf Cooperative Adaptive Cruise Control on traffic flow sta-
[Accessed July 29, 2016]. bility. Intelligent Transportation Systems (ITSC), 2010
Deng, Q. & Boughout, W., 2016. The impacts of heavy- 13th International IEEE Conference on, pp.759–764.
duty vehicle platoon spacing policy on traffic flow. In Shladover, S., 2010. Three-Truck Automated Platoon
Transportation Research Board 95th Annual Meeting. Testing. Intellimotion, 16(1), pp.7–9.
pp. 16–0403. Shladover, S.E. et al., 2015. Industry Needs and Oppor-
Farokhi, F. & Johansson, K.H., 2015. A Study of Truck tunities for Truck Platooning., (2015), pp.1–85. Avail-
Platooning Incentives Using a Congestion Game. able at: http://www.escholarship.org/uc/item/6723k932
IEEE Transactions on Intelligent Transportation Sys- [Accessed July 27, 2016].
tems, 16(2), pp. 581–595. Shladover, S.E., 2012. Recent International Activity in
Florida Department of Transportation (FDOT), 2015. Cooperative Vehicle--Highway Automation Systems.
Florida Traffic Online. Traffic Data Shapefiles - Truck Federal Highway Administration, pp.1–81.
Traffic Volume, (2015). Available at: http://www2.dot. Skottke, E.-M. et al., 2014. Carryover Effects of Highly
state.fl.us/floridatrafficonline/viewer.html [Accessed Automated Convoy Driving on Subsequent Manual
July 29, 2016]. Driving Performance. Human Factors: The Journal
Janssen, R. et al., 2015. Future of Transportation Truck of the Human Factors and Ergonomics Society, 56(7),
Platooning, (2015), pp.1–36. Available at: https://www. pp.1272–1283. Available at: http://hfs.sagepub.com/
tno.nl/en/about-tno/news/2015/3/truck-platooning- content/56/7/1272.abstract.
driving-the-future-of-transportation-tno-whitepaper/ Transportation Pooled Fund (TPF) Program, 2016.
[Accessed July 7, 2016]. I-10 Western Connected Freight Corridor Project.
Kunze, R. et al., 2009. Organization and operation of National Cooperative Highway Research Program
electronically coupled truck platoons on German (NCHRP), (2016). Available at: http://www.pooled-
Motorways. Second International Conference on Intel- fund.org/Details/Study/599 [Accessed July 13, 2016].
ligent Robotics and Applications, 5928, pp.135–146. Tsugawa, S., Kato, S. & Aoki, K., 2011. An automated
Kunze, R. et al., 2011. Automated Truck Platoons on truck platoon for energy saving. 2011 IEEE/RSJ Inter-
Motorways—A Contribution to the Safety on Roads. national Conference on Intelligent Robots and Systems,
In S. Jeschke, I. Isenhardt, & K. Henning, eds. Auto- pp.4109–4114.
mation, Communication and Cybernetics in Science and Van De Hoef, S., Johansson, K.H. & Dimarogonas, D.
Engineering 2009/2010. Berlin, Heidelberg: Springer V, 2015. Coordinating truck platooning by clustering
Berlin Heidelberg, pp. 415–426. Available at: http:// pairwise fuel-optimal plans. In 2015 IEEE 18th Inter-
dx.doi.org/10.1007/978-3-642-16208-4_38. national Conference on Intelligent Transportation Sys-
Larson, J., Liang, K.Y. & Johansson, K.H., 2015. A tems. pp. 408–415.
distributed framework for coordinated heavy-duty Vegendla, P. et al., 2015. Investigation of Aerodynamic
vehicle platooning. IEEE Transactions on Intelligent Influence on Truck Platooning. SAE Technical Paper.
Transportation Systems, 16(1), pp.419–429. Zheng, R. et al., 2014. Study on Emergency-Avoidance
Larsson, E., Sennton, G. & Larson, J., 2015. The vehicle Braking for the Automatic Platooning of Trucks.
platooning problem: Computational complexity and IEEE Transactions on Intelligent Transportation Sys-
heuristics. Transportation Research Part C: Emerging tems, 15(4), pp.1748–1757.
Technologies, 60, pp.258–277. Zheng, R. et al., 2015. Biosignal analysis to assess mental
Liang, K.-Y., Mårtensson, J. & Johansson, K.H., 2016. stress in automatic driving of trucks: palmar perspiration
Heavy-duty vehicle platoon formation for fuel effi- and masseter electromyography. Sensors (Basel, Switzer-
ciency. IEEE Transactions on Intelligent Transporta- land), 15(3), pp.5136–50. Available at: http://www.mdpi.
tion Systems, 17(4), pp.1051–1061. com/1424-8220/15/3/5136/htm [Accessed July 12, 2016].

818

TIS2017_Book.indb 818 2/8/2017 8:10:00 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The potential of energy-efficient driving profiles on railway

consumption: A parametric approach

Mariano Gallo, Fulvio Simonelli & Giuseppina De Luca

Dipartimento di Ingegneria, Università del Sannio, Benevento, Italy

ABSTRACT: This paper proposes a parametric approach to evaluate the effects of optimising driving
profiles on railway consumption. To this aim, a kinematic optimisation model is formulated and solved
for different case studies, also allowing for the energy recovered in the braking phase. The numerical results
were used to build tables and abacuses to provide a preliminary evaluation of effectiveness of energy-
saving strategies for different sections and train characteristics. The approach was tested on regional rail
services in the Italian region of Campania, showing that energy-efficient driving strategies could produce
significant reductions in energy consumption.

1 INTRODUCTION Other papers (De Martinis & Gallo, 2013;

Dominguez et al., 2012) dealt with the utilisation
Managing rail systems from different perspectives of energy recovery systems, while Gallo et al. (2015)
is an important part of transportation planning and Zhao et al. (2015) proposed a general optimisa-
and operation. As regards the real-time operation tion framework based on simulation approaches.
of rail systems, some recent issues based on simu- In the following, we consider a suburban railway
lation of rail convoys have included: managing rail line partitioned into sections. Each section links
disruptions (D’Acierno et al., 2016) and eco-driving two stations and is protected by a signalling system;
(Gallo et al., 2015). Further, in recent decades the departure of a train from a station is allowed
attention to energy efficiency has increased in all only if the section is entirely free. The scheduled
sectors both for economic and environmental rea- timetable is known and indicates the departure
sons. The transport sector consumes about 30% of time of the trains from the stations. Scheduled
energy in Italy (ENEA, 2015) and 27% worldwide timetables are designed so that the starting time
(IEA, 2014). Although energy consumption in the from a station is equal to the starting time from the
transport sector is due mainly to road transpor- previous station plus the minimum running time,
tation, energy efficiency in public transport and Tmr, between stations, plus the dwell time (neces-
a policy for increasing public transportation use sary for the boarding and alighting of passengers),
represent crucial objectives from the sustainability Td, plus the reserve time, Tr. The reserve time is
angle. In this context, the challenge to reduce rail provided in order to recover an eventual delay of a
energy consumption involves several aspects. Such train so that the starting time from the next station
aspects are both strictly technological (designing is respected (if the delay is lower than the provided
low-consumption electrical engines, designing reserve time).
energy-efficient braking recovery systems, etc.) If a train starts late, the driving style has to
and organisational (optimising rail service sched- ensure the minimum running time, i.e. maxi-
ules and driving styles so as to minimise con- mum acceleration, maximum cruising speed and
sumption on a railway line). The importance of maximum deceleration (compatible with comfort
energy-efficient driving strategies for reducing con- standards, speed limits and available power). This
sumption was highlighted by Strobel et al. (1974) driving style is also known as time-optimal (or all-
and relevant theoretical references for identifying out driving) and is the driving style with the maxi-
optimality conditions are provided in Howlett mum energy consumption. If the train is on time,
(1988). More recently, Liu & Golovitcher (2003) instead, the reserve time can be used for optimising
proposed a solution algorithm to optimise trajec- the driving style in order to minimise energy con-
tories using optimal control theory, Ke & Chen sumption. Indeed, reduction of the cruising speed
(2005) analysed optimal driving models under fixed and/or the introduction of a coasting phase (the
block and mobile block conditions, while energy- propulsion system is turned off and the train runs
saving methods for freight trains were studied by using kinetic energy, reducing its speed) make it
Lukaszewicz (2004) and Ke & Chen (2005). possible to reduce energy consumption.

819

TIS2017_Book.indb 819 2/8/2017 8:10:02 PM

 In this paper we use a model for optimising train 1 v2
speed profiles, with a view to minimising energy Eb M vb2 M g i b
consumption, based on kinematic variables pro- 2 2 | a′′ |
posed by Simonelli et al. (2015). Our aim was to M g ⎡ 1 vb2 1 v3 1 v4 ⎤
+ ⎢ r0 + r1 b + r2 b ⎥
ascertain the effectiveness of energy-efficient driv- 1000 ⎣ 2 | a′′ | 3 | a′′ | 4 | a′ | ⎦
ing and scheduling for different railway sections
scr vcr tcr sacc = vc2r a sb vb2 / 2 | a′′ |
and train characteristics.
The paper is organised as follows: section 2 2
⎛ vcr k ⎞
describes the optimisation model; section 3 intro- 1+
duces the parametric approach; section 4 focuses D ⎝ D ⎠
scs = ln 2
+
on the case study; finally, section 5 concludes the Z ⎛ vb k ⎞
paper. 1+
⎝ D ⎠
K ⎡ v k v k⎤
− ⎢ arctan cr − arctan
a b
2 OPTIMISATION MODEL Z ⎣ D D ⎥⎦

The overall train trajectory along a section encom- where the space and mechanical energy consumed
passes four motion phases: acceleration, cruising, in acceleration (Eacc, Sacc), in cruising (Ecr, Scr), in
coasting and braking. Under the assumption that coasting (Ecs, Scs) and in the braking phase (Eb,
acceleration and deceleration are at the maximum Sb) are calculated by assuming the resistances to
(comfort) feasible rate (a and a’), consistent with the motion, expressed in N over kN of weight,
the optimality condition found in Howlett (1988), as a polynomial function of train speed through
the problem of minimising mechanical energy can parameters r0, r1 and r2. Obviously, in the absence
be set as an optimisation problem in three vari- of braking energy recovery systems the term Eb
ables: cruising speed, vcr, cruising duration, tcr, and will be equal to zero.
initial braking speed, vb. Two constraints of the In the expression of Scs, parameters K, Z and D
problem are: (a) the constraint of compatibility are related to the resistance parameters and to the
with the train schedule, expressed by imposing that slope of section i. In particular:
the total space covered during the given travel time
T equals the distance L between the two stations; r1 ro 1000 i
(b) the inequality constraints related to the section K= D= − K2
2 r2 r2
speed limit and minimum initial braking speed. By
introducing these constraints the model can be for- r2 D g
mulated as follows (Simonelli et al., 2015): Z=
1000
[v ^
cr , vb^ ] Importantly, the previous expression of distance
= Arg
A gvcr ,vb i {E acc (vcr ) E cr (vcr , tcr ) + E b (vb )} in the coasting phase cannot be used for downhill
slopes where:
s.t.
4 r0 r2 − r12
i≥−
s (vcr ) + s [vcr ,t ] s (vcr ,vb ) + sb (vb ) = L
acc cr cr cs
4000 r2
0 ≤ vcr ≤ vmax
When slope i does not respect this inequality, the
vb, ≤ vb ≤ vcr motion equation during the coasting phase changes.
Nevertheless, the case of steeply descending slopes
tcr ≥ 0
will not be considered in the following analysis.
The optimisation model defined above, referring
with: to homogeneous sections, can be extended to mul-
tiple homogeneous subsections by considering the
1 M g ⎡ 1 vcr2 1 vcr3 1 vcr4 ⎤ same kinematic variables for each subsection and
E = M vcr2 + r0 + r1 + r2
2 1000 ⎢⎣ 2 a 3 a 4 a ⎥⎦ by introducing further variables, time and speed
related to the crossing points between subsections.
vcr2
+ M gi With regard to the parameters of resistance to
2a motion, although they should be evaluated experi-
Mg mentally for each train and truck, several formu-
E cr = [rro vccrr + r vccr2r + r vccr3r ] t M g i vcr tcr lations have been proposed in the literature and a
1000 sensitive analysis of mechanical equations to track

820

TIS2017_Book.indb 820 2/8/2017 8:10:02 PM

 and train parameters is provided in Boschetti &
Mariscotti (2012). Resistance to motion will be
calculated below through the Strahl formula for
traction units:

g
R(v ) = ( .3
.3 * Mt .03v 2 )
1000

where R(v) is expressed in kN, v is the train speed

(km/h) and Mt is the mass of traction unit in tons.
For passenger wagons, the Sauthoff formula
will be used:
Figure 1. Analysis of energy saving (% w.r.t. total
energy) with respect to train mass and section length with
Mw ⎛
ggM n + 2.7 2 ⎞ a reserve time equal to 5% of minimum travel time.
R(v ) = 1.9 + 0.0025v 0.0048 Fe w v ⎟
1000 ⎝ Mw ⎠
figure plots the isolines of the percentage of energy
where R(v) is expressed in kN, v is the train speed saving as a function of train mass and section
(km/h), Mw is the train mass in tons, Fe is a coef- length for a fixed percentage of reserve time. To
ficient related to the train front area characteristic, build the abacus plotted in Figure 1, steps of 5 tons
usually taking the value 1.45, and nw is the number were considered for train mass and 200 metres for
of wagons. section length. Hence the number of combinations
tested was 27,270.
The optimisation model for each combination
3 PARAMETRIC APPROACH was solved by a MatLab optimisation tool.
It is worth noting that the energy saving is always
By applying the model described above, the sen- significant, varying from a minimum of 10% for
sitivity analysis of the effectiveness of timetable long sections and low train mass to a maximum
scheduling and trajectory optimisation is carried of 28% for very short sections. Apart from where
out. In particular, our aim was to identify the kinds sections are shorter than 5 km, the trend is quite
of services suitable for the implementation of an regular and indicates that the percentage of energy
optimised travel time and subsequent optimised saving increases with train mass and decreases with
trajectory in order to obtain significant energy section length. Moreover, the effect of mass tends
savings. Indeed, long distances or very short sub- to decrease for increasing mass and decreasing
urban services do not appear so sensitive to the length (the slope of the isolines grows). Similarly,
optimisation of travel time and trajectory, due the distance between the isolines increases with sec-
to the preponderance of cruising or acceleration/ tion length, showing that the maximum variability
deceleration phase respectively. Below we therefore occurs for section lengths below 15 km, while for
focus on regional rail services with section lengths longer sections the sensitivity of the energy-saving
between two successive stations from 1 to 55 km percentage appears less significant.
and train mass from 45 to 600 tons. Train depar- In Figure 2 the same analysis was carried out
ture/arrival times can be scheduled as a function of considering a braking energy recovery system with
the minimum travel time, usually defined through an efficiency of 60%. Obviously, in the latter case
a trajectory—known also as all-out in the litera- energy saving is greater on all sections. However,
ture—encompassing an acceleration phase at the it is worth noting that the differences between
maximum feasible acceleration rate, amax, a cruising the optimisation of trajectories with and without
phase at the maximum allowed speed, vmax, and a braking energy recovery systems are more sig-
braking phase at the maximum feasible decelera- nificant in the case of shorter sections: when the
tion rate, |amin|. Nevertheless, scheduled timetables length is less than 10 km the percentage of energy
are usually designed by adding a reserve time to saved is almost always more than 30% and it tends
minimum travel time, fundamental for service reg- to become irrespective of train mass. This aspect is
ularity. Thus the analysis will be carried out for dif- due to the relative importance of the kinetic energy
ferent reserve times, Tr, expressed as percentages collected during the acceleration phase which, for
of the minimum travel time. short sections, represents a great percentage of
In Figure 1 the energy saving percentage is plot- total energy spent and can be partially recovered
ted against train mass and section length varying through the braking energy recovery system.
in the range specified above with reference to a In the previous figures the slope of the sections
reserve time of 5% of the minimum travel time. The was assumed equal to zero and speed limit equal

821

TIS2017_Book.indb 821 2/8/2017 8:10:04 PM

 Figure 2. Analysis of energy saving (% w.r.t. total
energy) with respect to train mass and section length with
a reserve time amounting to 5% of minimum travel time Figure 4. Effect of speed limit.
and a braking energy recovery system.

present services organised through round-trip runs,

such that not considering the effect of the slope
can lead, in the worst case, to underestimating the
potential energy saving. Moreover, the effect of the
slope is more sensitive to the train mass and sec-
tion length for short sections (less than 5 km) and
high train masses.
In Figure 4 the differences in energy saving are
reported for several speed limits (Vmax) compared
to the base with a speed limit of 100 km/h. Gen-
erally, energy saving increases as the speed limit
increases, although the effect of the speed limit is
not so significant. For differences in speed limits of
Figure 3. Percentage differences in energy saving for 10%, energy saving is nearly equal to the base case
ascending and descending slopes of sections compared in 90% of the combinations of mass and length
to the case of slope i = 0 with a reserve time equal to 5%
analysed.
of minimum travel time.

to 100 km/h, while the train mass was assumed 4 CASE STUDY
the same for a traction unit and passenger wag-
ons (50% each of total mass), and the number of The results obtained with the parametric analysis
wagons was assumed equal to 5. To assess the sen- can be useful to provide a preliminary evaluation
sitivity of energy saving to the section slope and of the effectiveness of energy-saving strategies on
speed limit, the same calculations were carried out real networks.
for several values of these parameters. Generally, In Figure 5 the regional railway network of
uphill slopes involve a reduction in energy saved Campania and the distribution of section lengths
while downhill ones involve an increase consistent are plotted. Looking at the railway network topol-
with the reduction in duration and effectiveness of ogy, independently of the actual services, over
the coasting phase. 97% of the station-station distances are less than
The results are reported in Figure 3, where the 10 km. Given the results reported in Figures 1
percentage differences in energy saving are repre- and 2, in Campania most of the regional station-
sented, through a contour plot, as a function of station sections are suitable for energy savings
train mass and section length, for a descending between 20% and 25% through optimisation of
slope (continuous isolines) and ascending slope timetables and trajectories. These percentages rise
(dashed isolines) of 5‰. Importantly, energy sav- to over 30% in the case of braking recovery sys-
ing is more sensitive to descending than ascending tems. Starting from this preliminary consideration
slopes. Indeed, the percentage increase in energy related to infrastructural and topological charac-
saving always outweighs the decrease due to the teristics, our analysis was enhanced by consider-
same, in absolute value, ascending slope. This ing the actual services defined from timetables and
allows preliminary analysis of potential energy sav- the actual train composition for each line. To this
ing by using parametric evaluation without a slope. aim 3,175 station-station runs were extracted from
Indeed, in real networks the timetables usually timetables and for each service the actual train

822

TIS2017_Book.indb 822 2/8/2017 8:10:05 PM

 Figure 5. Regional rail network in Campania. Figure 8. Cruising speed distribution of regional serv-
ices in Campania.

Table 1. Rolling stock in Campania.

Total Total
Number weight of weight
of traction Number of
traction units of trailers
Train units (tons) trailers (tons)

Ale 126(1R+2M)AL 2 48 1 31
Ale 126(2M)AL 2 96 0 0
Ale 126(1M+1R)AL 1 48 1 31
Figure 6. Distribution of the lengths from stop to stop Ale 724(2M+2R) 2 110 2 60
of actual rail regional services in Campania. Ale 125(1M+1R)AL 1 48 1 31
Ale 125(2M+1R)AL 2 96 1 31
Ale 501/502(2M+1R) 2 92 1 31
Ale 506/426(2M+2R) 2 125 2 88
Ale 663(1101-1204)(1M) 1 40 0 0
Ale 663(1101-1204)(3M) 3 120 0 0
Ale 663(1101-1204)(2M) 2 80 0 0
Aln 668(2M)AL 2 74 0 0
E464 1 72 6–15 200–410

Figure 7. Mass distribution of rolling stock used in

regional services in Campania.

composition was considered. The distribution of

section length, cruising speed and train weight are
reported in Figures 6, 7 and 8 while Table 1 reports
the trains with their mass characteristics. Figure 9. Energy saving (% w.r.t. total energy) with
The distribution of rail service characteristics respect to train mass and section length with a reserve time
show that 85% of daily routes cover less than of 5% of minimum travel time and scatter plot of train
mass and stop-stop length of services in Campania.
10 kms, while in terms of train mass two main
categories may be considered: medium (nearly
130 tons) and high weight (and 330 tons). The given the length and mass characteristics for each
cruising speed shows greater dispersion and route, most rail services may suitably achieve over
the average is 60 km/h, although, as described 20% energy saving. In detail, the evaluation of
above, this parameter does not significantly affect potential energy saving based on the rail services
the effectiveness of trajectory optimisation. In database and parametric analysis, neglecting the
Figure 9 the scatter plot of train mass and length slope, with the cruising speeds, provides an esti-
of services from stop to stop of the regional rail mate of the total saving of 18.5% with respect
services in Campania is positioned inside the con- to a reserve time of 5%. The same analysis can
tour plot of energy saving. The figure shows that, be carried out more accurately by optimising all

823

TIS2017_Book.indb 823 2/8/2017 8:10:07 PM

 trajectories, taking into account the character- De Martinis, V. & Gallo, M. 2013. Models and methods to
istics of rail services such as train mass, section optimise train speed profiles with and without energy
length, slope and cruising speed determined from recovery systems: a suburban test case. Procedia - Social
schedules. The total energy saving arising from and Behavioural Science 87: 222–233.
Dominguez, M., Fernandez-Cardador, A., Cucala, A.P. &
this further evaluation amounts to 20.5%, very Pecharroman, R.R. 2012. Energy savings in metro-
close to the estimate provided by the simplified politan railway substations through regenerative energy
parametric analysis. recovery and optimal design of ATO speed profiles.
IEEE Transactions on Automation Science and Engi-
neering 9: 496–504.
5 CONCLUSIONS ENEA 2015. Rapporto Annuale Efficienza Energetica
RAEE 2015. ENEA Agenzia nazionale per l’efficienza
In this paper the potential of railway energy sav- energetica.
ing strategies was studied, highlighting the ranges Gallo, M., Simonelli, F., De Luca, G. & De Martinis, V.
of train masses and railway section lengths that 2015. Estimating the effects of energy-efficient driving
are more promising in terms of consumption profiles on railway consumption. Proceedings of IEEE
EEEIC 2015–15th International Conference on Environ-
reduction. ment and Electrical Engineering: 813–818.
Using a parametric approach applied to the rail Howlett, P. 1988. Existence of an optimal strategy for the
network in the region of Campania (Italy) it was control of a train. School of Mathematics Report 3,
shown that up to 20% of total energy could be University of South Australia.
saved with proper optimisation of speed profiles, IEA 2014. Key World Energy Statistics. IEA.
and up to 35% by combining speed optimisation Ke, B.R. & Chen, N. 2005. Signalling block layout and
with a braking energy recovery system. In gen- strategy of train operation for saving energy in mass
eral, the speed limit slightly affects the optimisa- rapid transit systems. IEEE Proceedings: Electric Power
tion process, and the abacus in Figure 1, where the Applications 152: 129–140.
Ke, B.R. & Chen, N. 2005. Signalling block layout and
slope is also neglected, can be used for preliminary strategy of train operation for saving energy in mass
evaluation of potential energy saving for round trip rapid transit systems. IEEE Proceedings: Electric Power
rail services, leading at worst to an underestima- Applications 152: 129–140.
tion of energy saving. Liu, R. & Golovitcher, I.M. 2003. Energy-efficient opera-
tion of rail vehicles. Transportation Research Part A 37:
917–932.
ACKNOWLEDGEMENTS Lukaszewicz, P. 2004. Energy saving driving methods for
freight trains. Advances in Transport 15: 885–894.
Partially supported under research project PON— Placido, A. & D’Acierno L. 2015. A methodology for
assessing the feasibility of fleet compositions with
SFERE grant no. PON01_00595. dynamic demand, Transportation Research Procedia 10:
595–604.
Simonelli, F., Gallo, M. & Marzano, V. 2015. Kinematic
REFERENCES formulation of energy-efficient train speed profiles.
Proceedings of AEIT International Annual Conference
Boschetti, G. & Mariscotti, A. 2012. The parameters of 2015, Naples, Italy.
motion mechanical equations as a source of uncertainty Strobel, H., Horn, P. & Kosemund, M. 1974. Contribution
for traction systems simulation. XX IMEKO World to optimum computer-aided control of train operation.
Congress. 2nd IFAC/IFIP/IFORS Symposium on traffic control
D’Acierno, L., Placido, A., Botte, M. & Montella, B. 2016. and transportation systems: 377–387.
A methodological approach for managing rail disrup- Zhao, N., Roberts, C., Hillmansen, S. & Nicholson, G. 2015.
tions with different perspectives. International Journal A Multiple Train Trajectory Optimization to Minimize
of Mathematical Models and Methods in Applied Sci- Energy Consumption and Delay. IEEE Transactions on
ences 10: 80–86. Intelligent Transportation Systems 99: 1–10.

824

TIS2017_Book.indb 824 2/8/2017 8:10:08 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

The importance of railways for developing countries. The feasibility

study for the rehabilitation of the Mchinji to Nkaya rail line in Malawi

Antonio Placido
D’Appolonia S.p.A., Transport Engineering—Mobility and Logistics, Napoli, Italy

Marco Marchesini
D’Appolonia S.p.A., Technical Engineering Development IT—Railway Systems, Roma, Italy

Vincenzo Cerreta
D’Appolonia S.p.A., Transport Engineering—Mobility and Logistics, Napoli, Italy

ABSTRACT: Malawi is a landlocked country in South-Central Africa which ranks among the world’s
least developed countries. The economy is predominantly based on agriculture and depends on substantial
inflows of economic assistance from the International Monetary Fund (IMF), the World Bank, and
individual donor nations. Because of poor reliability of rail infrastructure, import and export goods
are mainly moved by road from/to neighbouring countries, with South Africa being the main partner.
In particular, more than 95% of current freight is transported by road resulting in high transport costs as
well as high negative impacts, such as pollution and car accidents. However, the Malawi Railway Network
has the potential to be a vital asset for the entire southern region of Africa, as it is part of an international
corridor connecting the Indian Ocean through the port of Nacala in Mozambique to the inner regions of
the Continent. Indeed, besides Malawi, other landlocked countries like Zambia and Democratic Republic
of Congo (DRC) are extremely interested in finding a competitive outlet to lower the cost of import
and export trades. Currently most exports from these countries are being transported by road over a
distance of more than 3,400 km to Durban port in South Africa, instead of less than 2,000 km by rail to
Nacala port passing through Malawi. This paper provides a brief description of the traffic forecast, the
infrastructure design, the operation and maintenance action plans, the environmental and social impact
assessment, the multicriteria analysis as well as the cost benefit analysis culminating in a feasibility study
report which identifies an optimal investment solution for the rehabilitation of the Mchinji to Nkaya rail
section. The study also highlights the strategic importance of the railway to support the fragile economy
of Malawi and that of the neighbouring countries.

1 INTRODUCTION

This paper summarizes the main activities carried

out by D’Appolonia SpA in joint venture with the
company TEAM Engineering SpA, as part of the
feasibility study for the rehabilitation of the rail-
way infrastructure from Mchinji in the Centre to
Nkaya in the South of Malawi (Fig. 1). The Feasi-
bility Study is expected to inform the Malawi gov-
ernment of the best possible investment solution
to meet the current and future traffic demands on
the line as well as a balanced maintenance plan,
on the basis of the declared national and specific
objectives. In doing so the railway line will play Figure 1. Proposed methodology for the Feasibility Study.
a vital role in national development of Malawi in
particular by reducing the time it takes to move provides evidence of the importance of transport
goods and people between locations in Malawi as infrastructure for developing countries and gives
well as from outside the country. Furthermore, it some useful indications about rail service costs

825

TIS2017_Book.indb 825 2/8/2017 8:10:10 PM

 (e.g. unitary cost of service, cost of rolling stock, 3 TRAFFIC FORECAST ANALYSIS
cost of signalling system) in that region of Africa,
which can be used as references for other academic The aim of the Traffic Forecast Analysis is to assess
and/or professional studies. the future traffic levels which will presumably affect
the Mchinji-Nkaya railway line in 15 years’ time
and, consequently, estimating the revenues associ-
2 BACKGROUND INFORMATION ated with the forecasted traffic scenarios.
As regards the first task, in order to simulate the
The Malawi Railway Network is single track, with future consistent traffic which will likely influence
narrow (1,067 mm – “Cape”) gauge. The rail- the circulation of goods within the entire South-
way was constructed between 1908 and 1980 and ern African region, an appropriate transporta-
extends for approximately 797 km, running from tion model is used to assess the performance of
Mchinji at the Zambian border in the west, via transportation infrastructures (both road and rail)
Lilongwe, to Blantyre and Makhanga in the South, within Malawi taking into account its relations
up to the border with Mozambique (Fig. 2). with neighbouring countries and ports (Fig. 3).
At Nkaya Junction, it links with the “Nacala In particular, the freight demand is estimated by
Corridor” line, running eastwards via Nayuchi to means of a gravitational model (Cascetta, 2009)
Mozambique’s deep water port at Nacala on the with mass variables corresponding to the employed
Indian Ocean. The link south from Makhanga to people of each origin and destination.
Mozambique’s Beira corridor has been closed since After calibrating and validating the models on
the Mozambique Civil War, with plans for recon- present traffic conditions (i.e. 2016), an in-depth
struction not yet implemented. There is no direct analysis of the future economic growth of Malawi
link with neighbouring Tanzania, as there is a dif- and Zambia is carried out so as to estimate the future
ference of gauge (1,067 mm/1,000 mm). An exten- freight demand. Indeed, both these landlocked coun-
sion from Mchinji to Chipata in Zambia opened in tries are extremely interested in the rehabilitation
2010, and there is a proposal to eventually link up of the Mchinji-Nkaya line, which would represent
from there with the TAZARA railway at Mpika or a cheaper and faster corridor to the Indian Ocean
Serenje (i.e. to the heart of the continent). Direct than those currently used. At the moment, most of
linkage is available with Mozambique, however, the products to Europe, the United States and China
which has the same gauge track. are exported via Durban port in South Africa.
Between 2000 and 2012, the Malawi Railways net- The economic projections are thus based on
work was concessioned to Central and East African estimated GDP trends1 in future years and four
Railways (CEAR), a private company, which made possible scenarios are considered:
losses during every year of the concesssion. The
very poor traffic performance and the recurrence
of derailments/accidents are further evidenced in
several technical/programmatic documents issued
by various Authorities (SADC, 2012) according
to which it is stated that CEAR’s derailment rate
results as being “one of the highest worldwide”.

Figure 3. Transportation model adopted for the simula-

tion of the rail and road freight traffic.

1
GDP trends for both Malawi and Zambia have been esti-
Figure 2. CEAR Railway Network railway and its con- mated according to data provided by The World Bank
nections with Mozambican Corridors. (www.theworldbank.com).

826

TIS2017_Book.indb 826 2/8/2017 8:10:10 PM

 Scenario 1: in this scenario, both Malawi and Table 2. Results of the macroscopic simulations.
Zambia are not expected to experience a massive
growth, since it is assumed that trading levels will Modal split Modal split
increase following a linear trend. This has to be at 2015 tons at 2031 tons
considered as the most pessimistic scenario; Total OD freight 3,021,198 15,111,572
Scenario 2: Malawian economy is considered to Total rail traffic 271,298 2,824,732
grow slowly, while Zambia is expected to increase % Rail 8.98% 18.69%
its trade exponentially;
Scenario 3: in this case, it is hypothesised that
the Malawian GDP and import/export trade will situation at 2015 and the most likely traffic forecast
develop considerably, in contrast to the Zambian at 2031 gives an idea of the benefits provided by the
economy, which is not expected to grow rapidly; rail line to all the Sub-Saharan area (Table 2).
Scenario 4: in this scenario, it is assumed that the As can be seen, almost 3 mtpa will travel on rail,
economy of both countries will run at full speed, corresponding to 18.69% of the total OD freight
growing according to an exponential trend. This has matrix. This is even more significant if it is consid-
to be considered as the most optimistic scenario. ered that railways in Malawi are not densely distrib-
Based on these assumptions, the adoption of uted, and some areas (e.g. the Northern region and
macroscopic simulations provide important indica- the Southern region) can be reached only by road
tions about the amount of goods affecting the rail vehicles. Therefore, the traffic captured by rail could
section between Mchinji and Nkaya in 15 years’ time be probably even higher if further investments in
(i.e. time horizon mentioned in the ToR), as well rail infrastructure on other corridors are made (e.g.
as the resulting financial revenue. These values are Sena corridor—from Blantyre to Beira Port).
required for the revenue estimation which depends In addition, it is necessary to consider that,
on the cost that customers have to pay for moving besides freight transport, railways can be extremely
goods. To this purpose, after carrying out a specific strategic for passengers. The Mchinji-Nkaya rail sec-
Sensitivity Analysis on fare structure, the rail trans- tion has proved to increase the accessibility of the
port cost considered is 0.089 USD/ton*km. Southern and the Central regions, within a country
Among the four different scenarios, the one where transport mobility at the moment is very dif-
which foresees an exponential increase of the Zam- ficult. The rail line is, therefore, a valid alternative to
bia trading and, at the same time, a more linear bus transport, enabling a faster and a more comfort-
growth of the Malawian economy (i.e. Scenario 2) able connection between Blantyre and Lilongwe, as
is identified, according to the actual situation of well as between all the other important cities on the
both countries, as being the most likely and, there- corridor. Therefore, even though the main profits
fore, is analysed in detail. for the rail concessionaire are generated by freight
In particular, results highlight that the amount traffic, the social benefits provided by the introduc-
of tons travelling per year on the Mchinji-Nkaya tion of a passenger rail service are noteworthy.
section is 2,706,290, which corresponds to
3,484,329 ton*km per day. This value should guar-
antee a total annual revenue of 86,841,949 USD 4 INFRASTRUCTURE ASSESSMENT
to the rail operator. The majority of this traffic
is for transit flows and concerns Zambia trades In order to correctly define the type and level of
through the Nacala corridor. However, the simula- interventions necessary for the rehabilitation and
tions demonstrate that the rehabilitation of the rail upgrading of the Mchinji-Nkaya line section, a
line is attractive even for Malawi. In fact, the main site inspection of the line enabled a detailed assess-
commercial hubs within the country (i.e. Nkaya ment of the current conditions of the line, which
and Kanengo station near Lilongwe, the capital resulted to be extremely bad.
of Malawi), consistent freight flows are attracted According to this analysis and to the techni-
from road transport, resulting in lower negative cal characteristics of the existing railway system,
externalities. A comparison between the current four different options for the rehabilitation and
upgrading of the railway infrastructure have been
Table 1. Results of the macroscopic simulations at considered:
2031. Option 1: Rehabilitation of infrastructure
aimed to return the railway to its original capacity.
Tons/year Tons*km/day $/year Under this option the line would have an estimated
15 years life for the railway trackage, but would
Scenario 1 1,143,635 1,584,398 $ 39,483,198
Scenario 2 2,706,290 3,484,829 $ 86,841,939
require an ordinary programmed maintenance.
Scenario 3 1,984,374 2,575,270 $ 74,135,739 The forecasted capacity is eight (8) trains per day
Scenario 4 3,168,679 4,075,692 $ 101,566,245 with 1,836,800 tons/year. The estimated cost of this
option is about 146,000,000M$ and the estimated

827

TIS2017_Book.indb 827 2/8/2017 8:10:12 PM

 duration of the infrastructure rehabilitation is costs and adequate revenues?” This criterion
18 months based on the assumption of four con- has been given a 50% weight.
struction sites, with an extension of approximately C. Environmental criterion: “which is the reha-
100 km each; bilitation option that is less impacting on the
Option 2: Rehabilitation and upgrading of the surrounding environment, in terms of air qual-
infrastructure to increase the capacity of the line by ity, noise, soil, water and biodiversity?” This
extending crossing loops at the main stations and criterion has been given a 10% weight.
constructing new triangles at Mchinji and Nkaya D. Social criterion: “which is the most profitable
stations. The estimated cost of this option is about option from the point of view of the local com-
164,000,000 M$ and the line would have an esti- munities?” This criterion has been given a 10%
mated 15 years life for the railway trackage, with an weight.
ordinary programmed maintenance. The forecasted
capacity is eight (8) trains per day amounting to For each of the above mentioned criteria, spe-
2,663,360 tons/year. The estimated duration of the cific sub-criteria have been used in the MCA.
infrastructure rehabilitation is 20 months based on More in detail:
the assumption of four construction sites;
A. Technical (transport) criterion: This technical
Option 3: Rehabilitation and upgrading of the
criterion has been represented by the following
infrastructure as for Option 2, plus increase of
sub-criteria:
the axle load up to 18t. The estimated cost of this
A.1. Railway Freight Capacity between
option is about 261,000,000 M$. Under this option
Mchinji and Nkaya, calculated as the
the line would have an estimated 20 years life for
maximum annual ton*km that can
the railway trackage, but would require an ordinary
be moved for each upgrading option,
programmed maintenance. The forecasted capac-
obtained by an optimal operational serv-
ity is eight (8) trains per day with 3,442,880 tons/
ice. The measurement of freight capacity
year. The estimated duration of the infrastructure
is given in number of tons per year.
rehabilitation is 24 months, based on the assump-
A.2. Transport Level of Service, expressed
tion of four construction sites of approximately
in terms of ratio between the expected
100 km extension.
traffic flow (tons/year) and the capacity
Option 4: Rehabilitation and upgrading of
offered by each option.
the infrastructure with a complete renewal of the
A.3. Railway Passenger Services, expressed in
permanent way and increase of the axle load up
terms of average speed (km/h) along the
to 21 tons. The estimated cost of this option is
line.
about 507,000,000 M$. Under this option the line
A.4. Improved Road Operations, expressed in
would have an estimated 30 years life for the rail-
terms of volume of freight traffic (tons)
way trackage, but would require an ordinary pro-
deviated from the road.
grammed maintenance. The forecasted capacity is
B. Financial criterion:
ten (10) trains per day with 5,278,000 tons/year.
B.1. Investment cost, Total financial invest-
The estimated duration of the infrastructure reha-
ment cost, represented in million USD
bilitation is 36 months based on the assumption of
B.2. Operational costs, expressed in USD/
four construction sites.
ton*km.
B.3. Traffic Revenues, expressed in total yearly
5 IDENTIFICATION OF OPTIMAL traffic revenues in USD.
INVESTMENT SOLUTION THROUGH C. Environmental criterion:
MULTI-CRITERIA ANALYSIS (MCA) C.1. Synthetic parameter, ranking each option
impact on air quality, noise & vibrations,
The elaboration of the MCA has been based on the soil & landscape, waste, water resources
best recognised international guidelines (Dogdson and biodiversity.
et al. 2009). The different MCA criteria adopted D. Social criterion: The social criterion will be
can be summarized as follows: represented by the following sub-criterion:
D.1. Synthetic parameter, ranking each option
A. Transport criterion: “how will the different
impact on employment, economic devel-
options respond to the transport demand,
opment, poverty reduction, community
improving the transport capacity, reliability
health & safety and land acquisition.
and safety in comparison with the present
transport infrastructure?” This criterion has The results of the MCA base case indicate that
been given a 30% weight. the preferred option is Option 3. In particular, the
B. Financial criterion: “which is the most efficient Transport criteria would favour Options 4, 3 & 2,
way to rehabilitate the railway, with the mini- which ensure higher values of railway capacity
mum investment cost, sustainable recurrent and level of service. Only Option 4, with higher

828

TIS2017_Book.indb 828 2/8/2017 8:10:12 PM

 Table 3. Values of each sub-criterion for the different is that Option 3 (restructuring and upgrading the
upgrading options. railway infrastructure, increasing the capacity of
the line and stations, increasing the axle load to
Opt. 1 Opt. 2 Opt. 3 Opt. 4 18t and installing welded rails) can be considered
Cr. A1 [M tons] 1.5 2.7 3.4 5.3 as the best rehabilitation option according to the
Cr. A2 [f/Cap] 1.84 1.02″ 0.79 0.52 selected criteria.
Cr. A3 [km/h] 33.48 33.48 33.48 44.66
Cr. A4 [%] 12.15 18.40 18.69 18.69
Cr. B1 [M $] 146 164 261 507 6 ENVIRONMENTAL SCREENING
Cr. B2 [$/ton*km] 0.054 0.054 0.047 0.047 AND POTENTIAL IMPACTS
Cr. B3 [M $] 47.2 85.5 86.9 86.9
Cr. C1 30 29 27 24 The Environmental Brief has been developed for
Cr. D1 42 44 45 44 Project Option 3 and provides a description of the
Project and of the selected Project option focussing
on the analysis of alternatives, national policy and
Table 4. Distribution of weights for the different criteria. legal framework, environmental and social setting
of the area and, finally, on the identification of the
Trans. Finan. Envir. Social main environmental and social impacts (positive
and negative) associated with the rehabilitation of
Weights case 1 30% 50% 10% 10% the railway line, both during the construction and
Weights case 2 35% 45% 10% 10% operation phases.
Weights case 3 40% 40% 10% 10% The main components likely to be affected are:
Weights case 4 25% 45% 15% 15%
Weights case 5 30% 40% 15% 15% 1. Natural environment: this includes air quality,
noise & vibrations, soil, subsoil, land use and
landscape, waste, water resources, biodiversity
Table 5. Results of the MCA for each distribution of and sensitive habitats;
weights adopted. 2. Socio-economic environment: this includes
labour, local and national economy, poverty
Opt. 1 Opt. 2 Opt. 3 Opt. 4 reduction, community health and safety, land
access, work related safety, transport.
Case 1 40.0 62.4 65.4 37.8
Case 2 37.0 60.2 62.9 37.1 The Environmental Project Brief concludes
Case 3 34.0 57.9 60.6 37.5 that the Project has very positive impacts. In par-
Case 4 42.0 64.6 67.1 37.3 ticular, it has the potential to contribute to the
Case 5 39.0 62.3 64.7 37.1 economy of the country through improved move-
ment of exports/import goods, job creation, busi-
ness promotion, enhancement of Community
speed would guarantee a substantial attraction of Health and Safety, and support to the agriculture
passengers from Lilongwe and Blantyre. Options industry, among others. From an environmental
2, 3 and 4 would all attract freight traffic from point of view, the Project will be responsible for
the parallel road, improving the road safety con- the improvement of air quality as a consequence
ditions. The Financial criteria, considering the of diverting passenger and freight transport from
construction cost (investment), operating cost road to rail, as well as reducing soil erosion. Lim-
and revenues, gives advantages to Options 3 & 2. ited negative impacts are expected to arise from the
Option 1 is the most advantageous for investment Project and are mostly limited to the construction
costs, while Options 3 & 4 for operation costs and phase (namely the interventions for the railway
revenues. The Environmental criterion (impact on rehabilitation), thus of a temporary nature. They
air, noise, soil, water and biodiversity) presents mainly refer to limited impacts on air quality due
the most advantageous synthetic parameter for to particulate emissions from construction engines,
Option 1, followed by Option 2. The Social criteria the increase of the noise level, extraction and use
(impact on jobs, health, poverty and land acquisi- of earth materials, harmless solid waste genera-
tion) shows the advantage of Option 3, with a dif- tion, potential work related accidents, limited loss
ference with the other options. of land and properties, risks of vandalism. It is
In order to analyse the stability of the results of important to consider that most of the railway
the MCA base case, a Sensitivity Analysis has been sections cross uninhabited areas, thus limiting the
conducted for a different distribution of weights impacts on human receptors. Moreover, through
(Table 4). The results of the Sensitivity Analysis the implementation of appropriate mitigation
in comparison with the base case (i.e. case 1) are measures, negative impacts will be minimised with
collected in Table 5. The conclusion of the MCA no unacceptable residual impacts.

829

TIS2017_Book.indb 829 2/8/2017 8:10:12 PM

 7 OPERATION MODEL AND
DIMENSIONING OF TRAIN FLEET

On the basis of the results of the macroscopic

simulations concerning the selected scenario (i.e.
Scenario 2), it has been possible to design plau-
sible Operation Plans to obtain indications about
the requirements in terms of infrastructure equip-
ment, the signalling system and rolling stock for
the rehabilitation of the Mchinji-Nkaya rail line.
Basically, at the moment, the rail track would
support a load of at most 15 tons per axle, and the
actual status of the infrastructure enables a limit of
about 0.25 mtpa. As far as the signalling system is
concerned, train operations are currently managed
by means of a telephonic paper order system. This
method can be used safely with low traffic levels. Figure 4. Hypothesized timetable of the Mchinji-
For denser timetables (as expected), systems that Nkaya rail line.
provide train movement detection, as well as fail-
safe interlocking, should be considered. Further- Table 6. Average unitary cost for purchasing and
more, single-track line capacity is limited by the revamping the rolling stock fleet.
need for trains to decelerate, stop, and accelerate
out of sidings to allow other trains to use the inter- Cost of new locomotive $ 3,500,000
mediate single-track sections. As a consequence, Cost of new freight wagons $ 57,500
the need of frequent crossings of trains running in Cost of new passenger wagons $ 700,000
opposite directions consumes a lot of capacity and Cost of revamping locomotives $ 1,750,000
can become a critical factor. In addition, meeting Cost of revamping freight wagons $ 14,375
at these sidings is the largest cause of train-inter- Cost of revamping pax wagons $ 280,000
ference delays on single tracks. Therefore, in order Cost of new shunting locomotives $ 1,000,000
to best increase the capacity of the Mchinji-Nkaya Cost of revamping shunting locomotive $ 500,000
line, for the Operation Plan no crossings between
trains travelling in opposite directions are consid-
(average fleet age is 49 years) and has to be refur-
ered. As a consequence, it is hypothesised that on
bished and or increased to meet the expected
odd days trains run from East to West (e.g. from
demand. In order to estimate the investment cost,
Mozambique to Zambia) and from West to East
the following assumptions have been taken into
on even days. This solution has the great advan-
account:
tage of reducing travel times, avoiding inefficient
stops at stations for crossings and, above all, mini- • for freight service, double-headed trains coupled
mising safety risks. Obviously, due to the alternate “in multiple” are considered, which means that
circulation, since trains travelling in one direction at least two locomotives are necessary to haul
are allowed every two days, the number of runs per each train;
day has to be doubled. • Besides the total locomotive fleet required to
Regarding the signalling system, Manual Block perform the service, an additional 30% has to be
Signalling (MBS) or Automatic Block Signalling dedicated as a reserve for ordinary and extraor-
(ABS) could be adopted. The cost for signalling dinary maintenance programmes or in case of
system has been estimated as 46,304,560 for the disturbances to the ordinary conditions (e.g.
whole line (i.e. almost 8,900 USD/km). spare locomotives);
By means of simple calculations, it has been pos- • For the same reasons, a surplus of 40% of the
sible to assess that, the upgrading Option 3 would total freight wagon fleet might also be sufficient
meet the expected freight demand (i.e. Scenario 2) to balance the empty wagons which have to be
considering 6 freight runs from Nkaya to Mchinji moved each day;
on odd days and 6 runs travelling in the opposite • The supplementary rolling stock (both locomo-
direction on even days (Fig. 4). In addition to the tives and wagons) could also be used to meet
freight service, a daily train connecting Lilongwe possible peak demand periods (i.e. more than 6
and Blantyre is included so as to provide also a runs per day needed);
passenger service. • for passenger service, each train is composed of
Regarding the dimensioning of the fleet, one locomotive and 8 passenger wagons for a
the actual rolling stock available is out of date total capacity of 600 passengers;

830

TIS2017_Book.indb 830 2/8/2017 8:10:12 PM

 • Additional rolling stock for maintenance and No significant differences in terms of ordinary
reserve purposes is expected to be 30% of the maintenance costs were considered between the
actual fleet. hypotheses “with” and “without project”. The
• The fleet belonging to other rail operators, espe- assumption is that there will be a rough compensa-
cially CDN from Mozambique will contribute tion between the differences that will occur among
to moving the expected freight, thus reducing the two situations: a) increase of maintenance
the resources needed. costs due to increased volumes of traffic for the
situation “with project”, and b) increased costs
According to these considerations, after a brief due to the deteriorated existing conditions in the
analysis of the average unitary costs (Table 6), “without project” case.
the estimated investment costs for rolling stock A periodic maintenance intervention is sup-
amount to 52.09 M USD. posed for both scenarios every ten years. The cost
of such maintenance was assumed by experience
to be 10% of required cost for the upgrade of the
8 COST BENEFIT ANALYSIS line, i.e. 23.84 M USD. Differences between the
two scenarios are only due the different phasing,
The CBA methodology is based on the EU Guide with the first extra maintenance starting at year
to Cost-Benefit Analysis of Investment Projects, 2017 and 2026 for the “without project” and at five
Economic Appraisal Tool for Cohesion Policy years later for the “with project” case.
2014–2020 (European Commission, 2014), which Financial inflows are represented by the pro-
combines financial, economic and risk analyses. ceeds from the charges applied to users for the sale
The Financial Analysis takes into consideration of transport services. The estimation of revenues
Investment costs and Operation and Maintenance has been based on the traffic volume forecasts, tak-
(O&M) costs. Investment costs have been evalu- ing the charge system (i.e. 0.089 USD/ton*km) as
ated through Bills of Quantities and Cost Esti- invariant.
mates regarding Option 3 and resulting from the The financial CBA, as expected for this kind of
infrastructure assessment (14-1397-H2, 2016a). investment, does not reach profitability, but the
The financial discount rate adopted for this study relevant indicators are still positive:
was 7.5% and the time extension of the cash flow
has been taken equal to 30 years after the finalisa- • the IRR is 6.45%, below the discount rate;
tion of the upgrade, i.e. until year 2050 included. • the NPV is −31.61 M USD;
The traffic flows that were considered are those • the B/C ratio is 0.94, just below 1.
estimated for the Scenario 2.
O&M costs have been estimated in the “with” In the Economic Analysis, the main direct bene-
and “without project” scenarios, grouped into the fits related to changes of the following measurable
following categories: have to be considered:

• infrastructure operations, e.g. repairs, current • The consumer surplus, defined as the excess
maintenance, materials, energy, train operations, of users’ willingness to pay over the prevailing
etc; generalised cost of transport for a specific trip.
• services operations, e.g. staff cost, traffic man- The generalised cost of transport expresses the
agement expenses, energy consumption, mate- overall inconvenience to the user, computed as
rials, consumables, rolling stock maintenance, the sum of monetary costs borne (railway tar-
insurance, etc.; iff) plus the value of the travel time, calculated in
• services management, e.g. services management equivalent monetary units. Any reduction of the
itself, fare/tolls collection, company overheads, generalised cost of transport for the movement
buildings, administration, etc. of goods and people determines an increase
in the consumer surplus. The main items to be
The unitary freight train operation cost was esti- considered for the estimation of the consumer
mated as 0.0534 USD per ton × km. surplus in the railway project are: reduced rail-
Maintenance costs by contrast, should cover way travel time and reduced road users Vehicle
(and are usually distinguished into): Operating Costs (VOC) & accident costs;
• routine maintenance: yearly work required to • The producer surplus, defined as the revenues
keep the infrastructure technically safe and accrued to the transport operators and owners
ready for day to day operation as well as to pre- (railway and road) minus the costs borne. The
vent deterioration of the infrastructure assets; change in the producer surplus is calculated as
• periodic maintenance: all activities intended the difference between the change in the pro-
to restore the original condition of the ducer revenue (e.g. rail ticket income increase)
infrastructure. and the change in the producer costs (e.g. train

831

TIS2017_Book.indb 831 2/8/2017 8:10:13 PM

 operating & maintenance costs, road mainte- The modal shift from the road to the rail mode
nance costs). The main items to be considered produces some additional socio-economic benefits
for the estimation of the producer surplus are that have a significant impact on the final result.
the variation of the railway producer operat- These benefits are mainly represented by:
ing & maintenance costs and the parallel road
• The environmental impact of different transport
maintenance costs.
modes;
The analysis is therefore carried out under the • The benefits due to the reduction of road
point of view of the general society, and this is the accidents;
reason why this approach is frequently referred • The reduction of road maintenance costs pass-
also as “socio-economic” approach. Excluding ing from the road mode to the rail mode (espe-
transfers of money between the parts involved in cially as regards the freight component);
the analyses means to take into account also taxes,
The environmental impact of the different
excise duties, levies that are collected by the State
transport modes was estimated on the basis of
and by local administrations and used in order to
international standards. This because the cost
deliver public services and, at the same time, means
of pollution cannot be differentiated on the basis
to exclude from the expenditures those implicit
of the characteristics of local economies, giving
benefits that are linked to the new job creation.
that pollution has a global impact and cannot be
These are the notions that are at the base of the
estimated at local level. The values adopted in this
so-called “shadow prices” that are used inside eco-
study are therefore:
nomic CBAs and that from the practical point of
view bring to the reduction of some costs consid- • Truck environmental cost: 0.0335 USD per
ered by the financial approach with the application ton × km;
of financial/economic conversion factors. In our • Bus environmental: 0.0215 USD per pass × km;
case a unique financial/economic conversion factor • Freight train environmental cost: 0.0122 USD
of 0.7 was adopted. A discount rate of 8% was also per ton × km;
considered, as indicated in a previous pre-feasibility • Passenger train environmental cost: 0.0335 USD
study (Gibbs & Parsons Brinckerhoff, 2014) on the per pass × km.
same rail line. As a consequence, for the economic
The above values are the result of a Consultant’s
analysis, the different cost items are based on the
adaptation of data from other relevant European
same values adopted for the financial analysis, but
studies. Environmental values for rail services are
they were reduced by transforming these costs into
relatively high because they are referred to trains
economic values. In addition, the “with project”
hauled by diesel locomotives only. On the other side,
scenario is bound to produce a modal shift from
values for road transport can be considered overes-
road to rail for both freight and passenger traffic.
timated and therefore quite prudential for an Afri-
Such modal shift will produce a reduction in terms
can scenario, given that the environmental quality
of operating costs, due to the better efficiency of
of the local vehicle fleet still does not match that one
the rail mode as a whole (and especially in terms of
of the European vehicle fleet at year 2011.
reduced energy consumptions). The calculation of
The basic cost for road accidents was estimated
the overall benefits was carried out by considering
at 0.01 USD per ton × km (Gibbs & Parsons
the tons × km and the passengers × km saved by
Brinckerhoff, 2014) and a comparable value was
moving traffic from the road mode to the rail mode
adopted for passenger road traffic (0.01 USD per
and by comparing the relevant reduction of costs.
pass per km). Finally, following again the indica-
The following unitary economic operating costs
tions of Gibbs & Parsons Brinckerhoff (2014), the
were adopted in order to estimate such aspect:
cost for road maintenance was also estimated at
• Truck operating cost: 0.0690 USD per ton × km; 0.01 USD per ton × km. The results of the eco-
• Bus operating cost: 0.0260 USD per pass × km; nomic analysis are:
• Freight train operating cost: 0.0374 USD per • the IRR is 13.63%, that is over the discount rate
ton × km; (8%);
• Passenger train operating cost: 0.0395 USD per • the NPV is 170.77 M USD, therefore positive;
pass × km. • the B/C ratio is 2.09, therefore much higher
than 1.
Values for trucks and buses have been assumed
on the basis of a comparison with similar studies Due to the difficulty in forecasting values in the
carried out by the Consultant, while values for rail future, besides CBA, it is good practice to assess
are the result coming from the economic/financial the project risks by carrying out a sensitivity analy-
conversion of the operating costs, as considered sis and a risk analysis. In the Risk Assessment, risks
for the financial CBA. are analysed through the probability distributions

832

TIS2017_Book.indb 832 2/8/2017 8:10:13 PM

 indicating the likelihood of the actual value of a the Mchinki-Nkaya rail line are briefly described.
variable falling within stated limits, while uncertain- Results have demonstrated that the project would
ties have been analysed through defined variations help the Malawian economy to grow up support-
of the main variables of the project (sensitivity ing the GDP and reducing costs of import/export
analysis). The Sensitivity analysis allows the deter- trades. In addition, the rail line would provide
mination of the ‘critical’ variables or parameters of several social benefits which make the investment
the model. Such variables are those whose varia- worthwhile especially from the economic point of
tions, positive or negative, have the greatest impact view.
on a project’s financial and/or economic perform-
ance. The analysis has been carried out by varying
the investment cost or the Benefits/Revenues val- REFERENCES
ues, one element at a time and also as a combined
Cascetta, E. 2009. Transportation systems analysis:
scenario, determining the effect of that change on Models and applications. Springer: New York (NY),
EIRR/FIRR and on NPV. The results of both the USA.
above-mentioned analyses support the convenience Dodgson, J.S., Spackman, M., Pearman, A. & Phillips, L.D.
of the investment. In fact, there is a 95% probabil- 2009. Multi-criteria analysis: a manual. Department for
ity that the upgrading Option 3 economic IRR will Communities and Local Government: London.
be greater than 11.35% and the minimum value European Commission 2014. EU Guide to Cost-Benefit
(11.90%) is reached if the investment cost is 20% Analysis of Investment Projects, Economic Appraisal
higher and simultaneously the benefits are 20% Tool for Cohesion Policy 2014–2020.
lower than estimated in the CBA. GIBB, & Parsons Brinckerhoff 2014. Scoping study for
the railway line in Malawi from Mchinji to Nkaya—
including the section between Mchinji and Chipata in
Zambia”. Final Report.
9 CONCLUSION SADC 2012. Southern African Development Commu-
nity. Regional Infrastructure Development Master
In this paper, the main activities carried out as part Plan.
of the feasibility study for the rehabilitation of

833

TIS2017_Book.indb 833 2/8/2017 8:10:13 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Identifying driving behaviour profiles by using multiple correspondence

analysis and cluster analysis

D.S. Usami, L. Persia, M. Picardi & M.R. Saporito

Sapienza University, Rome, Italy

I. Corazziari
ISTAT, Rome, Italy

ABSTRACT: Many studies investigated the relationship between accident risk factors related to driv-
ing behaviour (e.g. distraction, aggressive driving) and the involvement in a road accident. However less
attention has been given to aspects related to driver insecurity and the effects of the overall driving pat-
terns. The main target of this research were twofold: 1) The identification of driving behaviour profiles
taking into account also insecurity. 2) The analysis of the association between the identified profiles and
their accident involvement. A survey was undertaken among a sample of Italian drivers to assess driving
distraction, aggressiveness, indiscipline and insecurity. The items of the used questionnaire were mostly
derived from the literature (e.g. Driver Behaviour Questionnaire). Behavioural tendencies of drivers to
distraction, aggressiveness, indiscipline and insecurity were studied through Multiple Correspondence
Analysis. By using Cluster Analysis seven groups of drivers with similar behaviours were identified.
A significant association between the seven groups and road accident involvement was found. This statis-
tical approach allows the identification of driving behaviour profiles that could be used for driving train-
ing purposes. The answers to the questionnaire can for instance highlight an aggressive and/or insecure
driving behaviour thus tailoring the theoretical and practical driving exercises to the specific driver needs,
especially for novice drivers.

1 INTRODUCTION Extensive research investigated the relationship

between personality traits and driving behaviour,
1.1 Background referring to the ways drivers choose to drive or
habitually drive (e.g. speeding, driving aggressively,
The prevalence of human factors in the causation
complying to norms, etc.).
of road accidents has been demonstrated by sev-
Methods used within these studies to measure
eral studies (e.g. Usami et al. 2015). This led many
driving behaviour are mostly based on question-
researchers to focus on human related predictors
naires, data from simulators or number of traffic
of road accident involvement.
offences.
A review of the literature indicates that vari-
Several questionnaires have been proposed in
ous socio-demographic factors like gender and
the last decades to measure driver behaviour. Some
age have been found to be related to road accident
of them, like the Driver Behavior Questionnaire
involvement (Abdel-Aty et al., 1998). Also infor-
(DBQ, Reason et al., 1990) consider more aspects
mation processing (e.g. selective attention), cog-
of driving behaviour. For example DBQ examines
nitive ability and personality traits (e.g. sensation
errors made while driving, deliberate violations of
seeking, anger, respect of authority, locus of con-
normal safe driving practice, and harmless mistakes
trol) have been found to be related to involvement
that result from inattention. Other questionnaires
in road accidents (e.g. Arthur et al., 1991).
are more specific examining for instance aggressive
It is also recognize that different personality
driving (Deffenbacher et al., 1994).
traits lead to different driving styles. Many studies
Based on a review of several existing question-
examined the association of a potential predictor
naires, Taubman-Ben-Ari et al. (2003) initially pro-
alone. Dahlen et al. (2005) investigated the com-
posed four driving styles: (a) reckless and careless
bined effect of driving anger, sensation seeking,
driving style, (b) anxious driving style, (c) angry and
impulsiveness, and boredom proneness on driving
hostile driving style, (d) patient and careful driving
behaviour; also highlighting the importance of the
style. After a factor analysis they found eight main
use of multiple predictors in understanding unsafe
factors, each one representing a specific driving
driving behaviour.

835

TIS2017_Book.indb 835 2/8/2017 8:10:15 PM

 style: dissociative, anxious, risky, angry, high veloc- aggression, careless (distraction) and anxiety
ity, distress reduction, patient and careful driving (insecurity).
style. They found that these factors significantly The “driving experience” section is character-
predict self-reports of accident involvement and ized by a number of questions related to driving
traffic offenses. frequency, type, mode of travel etc. Questions were
However, a driver hardly would fall on a single derived from technical meetings held in collabora-
driving style category, his driving profile would tion with the EFA (European Federation of driving
probably fit more factors. This information is schools and driving instructors) and in EU studies
important especially from a driving school per- (e.g. NovEV Project of CIECA, the International
spective. Knowing in advance the profile of a Commission of driving examiners.
driver would help a driving instructor in defining a The “reckless” section has been drawn from the
customized course, aimed at addressing the issues Driver Behaviour Questionnaire (DBQ, Reason
of a person way of driving. et al., 1990) by selecting questions on deliberate
choices of the driver. Estimating the consequences
1.2 Objective of deliberate manoeuvres is difficult for a common
driver and has different perceived consequences.
Based on the existing knowledge, the present study For instance, parking a vehicle in second line is
combines four predictors of driving behaviour perceived differently from running a red traffic
with two specific objectives: 1) The identification light.
of driving behaviour profiles. 2) The analysis of The section of “aggressive behaviour” has been
the association between the identified profiles and addressed both by means of the Driver Anger
their accident involvement. Scale (Deffenbacher et al., 1994), and the Driver
The basic assumption is that a not balanced Behavior Questionnaire (DBQ, Reason et al.,
and patient driving behaviour is positively related 1990). Among the standards of human sciences
to accident involvements, and also to a lacking used by the American Association of Psychology it
knowledge of traffic rules and laws. includes the Driver Anger Scale, containing differ-
The driving behaviour is a complex and multidi- ent traffic situations attributable to the behaviour
mensional construct and it has been measured con- of others (disrespectful), presence of other types
sidering four main dimensions: reckless, aggressive, of users (co-existence among different) or external
careless and anxious driving behaviour. Each of the situations (speed radar) and the consequences of
four dimension is complex and multidimensional in these on your driving attitude.
turn, so it is described by a set of indicators, meas- “Distraction” section is once again mostly
ured by the corresponding questions (categorical derived from the DBQ. It addresses the difference
variables) in the developed questionnaire. between distraction and inattention, where the
To summarize and measure the four behavioural first may be attributable to the conduct of a series
dimensions, a Multiple Correspondence Analysis of activities in parallel at driving (using the naviga-
(MCA) has been performed to each of the corre- tor or mobile phone) and the second to a mental
sponding set of categorical variables. state of disconnection (i.e. absorbed in thoughts)
The drivers’ scores on the derived factors (prin- from the main task.
cipal components) have then been analysed by a The “insecurity” section makes use of two stud-
Cluster Analysis (CA) to identify profiles of driving ies (Ma et al., 2010; Ulleberg, 2001) in addition to
behaviors more related to accidents occurrences. the support received from the research of techni-
cal workshops in collaboration with the experience
of the training carried out within the team EFA.
2 MATERIAL AND METHODS The risk perception, the assessment of the situation,
the sharing of the traffic environment, the decision
2.1 Survey questionnaire making and past experiences are the factors that
By applying to the field of driver education the characterize this part of the questionnaire.
European principles of Lifelong Learning, a ques-
tionnaire was developed able to identify different
2.2 Sample description
types of driving behaviour, so that we can better
address initial and post-license training. Several drivers, clients of Italian driving schools
The aim is to get targeted training to obtaining from all the country, were contacted by e-mail and
a driving profile that would be safe, responsible, invited to take part to the survey.
beyond a preparation aimed at passing an exami- More than 200 drivers initially agreed to answer
nation to obtain a driver’s license. to the questionnaire. However, only a part of them
The questionnaire is divided into 5 sec- were correctly completed and were selected for the
tions: driving experience, reckless (indiscipline), analysis.

836

TIS2017_Book.indb 836 2/8/2017 8:10:15 PM

 The final sample was 104 participants (41% by other multivariate quantitative methods. In the
women) from all over Italy (mean age 40 years). present study an ascending hierarchical classifica-
The demographic characteristics and the per- tion analysis was performed (Everitt BS 1993), to
centage of involvement in a road accident in the determine the most plausible number of drivers
last two years are presented in Table 1. groups, homogeneous in term of their four predic-
tor profiles. The variable created by the clustering
process indicating which cluster every individual
2.3 Statistical methods
belongs to, represents an indirect measure of the
2.4 Multiple Correspondence Analysis and driving behaviour, to be related to accident occur-
cluster analysis rences. The cluster method was the Ward’s mini-
Multiple Correspondence Analysis (MCA) is a mum variance method applied to the whole set of
factorial analysis method allowing summarizing principal components obtained in the four MCA
a set of categorical variables by one or few quan- analysis. The Ward’s method is normally used to
titative and not correlated (orthogonal) ones. The aggregate cases when euclidean distances make
main aim of MCA is to describe the association sense, as it is the case with quantitative variables,
among the original variables, and to measure the and so in the case of the MCA scores. With this
underlying latent factor not directly observable, method, the sum squared distances of points from
represented by the set of observable variables. their cluster centroids are minimized. The descrip-
Separate MCA were performed for the four set tion of clusters, that means the definition of driver
of variables relating to the corresponding four pre- profiles, is made through the analysis of the propor-
dictors of driving behaviour. tions of units with a specific category in the cluster,
To perform MCA, every item of each cate- compared with the corresponding proportion of
gorical variable has been coded in a binary form, units reporting the same category in all the sample.
1 indicating the presence of the item, 0 otherwise, After having defined the clusters, new units can
only one item presence allowed for each categori- be classified ex post in one of the groups, by a dis-
cal variable. Performing an MCA on the resulting criminant analysis (Anderson TV 1984). The first
matrix (a complete disjunctive one) is equivalent to step is to calculate the new units scores, projecting
a principal component analysis on the same matrix as supplementary units the new ones on the fac-
(Jobson JD 1992). torial planes obtained in the MCA analysis. Then
The method provides a set of coordinate values a nonparametric discriminant analysis can be
which allow to analyse and possibly to graphically performed on these scores, to classify units in the
display on a Cartesian plane, the association among given clusters, minimising distances between the
the original variables (among their categories). new units and the clusters barycentres.
MCA also provide coordinate values (scores on
the principal components) for individuals (drivers)
allowing to evaluate drivers’ profiles in term of their
3 RESULTS
similarity referring to the considered predictor (their
proximity can be graphically displayed on the same
3.1 Factors contributing to driving reckless,
Cartesian plane where variables are displayed).
aggressive, anxious and careless behaviour
The drivers’ scores on the derived factors (no
more than two principal components for each of The first MCA performed on the reckless indicator
the four predictors of behaviour) can be analysed variables is summarized in Figure 1, displaying the
first two principal components: behaviours respect-
Table 1. Characteristics of the sample. ing the rules are identified by negative values on the
first (horizontal) principal components axis, with a
Male Female tolerant attitude towards the other drivers. Nega-
tive values also on the second axis show less respect
Age
when the car ahead drives too slow, so that the risk
<25 9,3% 11,5%
to overtake other vehicles, even on the right when
25–34 25,6% 11,5%
driving on the motorway, is not implausible. Such
35–44 48,8% 45,9%
profile is associated with drivers aged more than 55.
45–54 9,3% 19,7%
Positive values on the second axis identify drivers
55–64 4,7% 6,6%
that do not always respect the public buses right of
> 65 2,3% 4,9%
Accident involvement in the last 2 years
way. It is associated with younger female drivers.
0 crashes 88,4% 85,2%
Positive values on the first axis represent more
1 crash 9,3% 11,5% reckless behaviours. When associated to negative
2+ crashes 2,3% 3,3% values on the second axis, more extreme behaviours
are described in term of risks: driving after alcohol

837

TIS2017_Book.indb 837 2/8/2017 8:10:15 PM

 consumption, no speed limit respect, risky over- (with negative values on both the axis), mainly
taking, no tolerance towards the other drivers, car resident in the North-east of Italy, with a primary
races. Such reckless profile is associated with males. school educational level, and (with positive val-
To summarize the reckless predictor, only the ues on the second axis) the no aggressive profile
first principal component is taken for the follow- except in case of being forced to reduce its own
ing CA. driving speed due to ahead bikers or ahead cami-
The second MCA performed on the aggres- ons limiting the view (drivers aged 25–34). With
sive indicators provide two principal components positive values on the first axis the more aggressive
axis, identifying four profiles of drivers (Figure 2). profiles: from drivers who feel nervous if forced to
Negative values on the first axis describe the less reduce their speed due to speed controls, traffic,
aggressive profile, yet willing to show who is right bikers, camion limiting the views or loosing sends

Figure 1. Map of reckless behaviour associations (first and second principal components).

Figure 2. Map of aggressive behaviour associations (first and second principal components).

838

TIS2017_Book.indb 838 2/8/2017 8:10:15 PM

 Figure 3. Map of anxious behaviour associations (first and second principal components).

Figure 4. Map of reckless behaviour associations (first and second principal components).

(positive values on the second axis), to more first axis, while with positive values on the first one,
aggressive drivers feeling nervous in all the con- drivers are characterized by less worrying about acci-
sidered dimensions, but not interested in showing dent involvements, injuries and weather conditions.
their own reasons (aged more than 55, negative The fourth and last MCA has been applied to
values on the second axis). the careless indicators variables. The two orthogo-
The third MCA refers to the anxious style of nal axis are also conceptually orthogonal, as the
driving. Again two axes have been considered, the first axis identified with positive values less care
second one more clearly identifying anxious drivers drivers, in every all the considered dimensions,
(with positive values) with avoiding behaviours due while the second axis’ positive values identifies
to anxiety, against less anxious drivers with negative very care drivers.
values on the same axis. Drivers showing higher con- Negative values on both the dimensions identify
sciousness of actual risks have negative values on the more care drivers’ profiles.

839

TIS2017_Book.indb 839 2/8/2017 8:10:15 PM

 3.2 Driving behaviour profiles The individuals in Cluster 4 are characterised
by very high anxiety. They also reported high val-
To identify the overall driving behaviour in term
ues in aggressive driving. They respect the traffic
of balance and patience, the first reckless axis
rules and pay attention while driving.
discriminating between more respectful drivers
Reckless but mild behaviour mostly character-
and less ones, and the first two axis for the other
izes Cluster 5.
three predictors, has been analysed by a cluster
Drivers in Cluster 6 are anxious even if they
analysis. An ascending hierarchical classification
appear to pay a lot of attention when driving.
has been adopted, using the Ward’s algorithm.
All the other variables have scores close to the
Starting from each driver being a single cluster,
median.
the method aggregates more and more couples
Drivers in Cluster 7 appear to be less anxious
of units, or a cluster with a unit, or two clusters,
and conscious of driving risks. They are also
according to the minimization of the variance of
aggressive and careless drivers.
the new cluster formed. The process stops when
Drivers in Clusters 2 seem to be those closer to a
some criterion are met, considering the gain in
balanced average driving profile.
minimizing the within clusters variance, and
maximizing the between clusters one (Everitt BS
1993). 3.3 Driving behaviour profiles and road accident
Seven clusters were identified with this method. involvement
The study of the characteristics of the drivers
Finally, an attempt was made to assess the asso-
belonging to each cluster, allows identifying seven
ciation between the seven driving profiles and self-
driving profiles in terms of balanced and patience
declared accident involvement.
behaviours as described by the four predictors, and
A significant association (95%) between the
other demographic characteristics.
seven groups and road accident involvement was
In each driving profile does not exist a pure
found (see Table 3).
driving style. Driving profiles differ in the preva-
Figure 5 shows the percentage of drivers report-
lence of a driving style (reckless, aggressive, care-
ing to be involved in a crash in the last two years
less and anxious driving) on another. A driver
by cluster. Drivers belonging to cluster 4 (aggres-
could act for instance aggressively (honks easily,
sive and anxious, but observing rules), 6 (careful
discusses with other road users), disregarding
but anxious) and 7 (aggressive and careless) show
traffic rules but confident in what he is doing
the highest crash involvement. While drivers of
(not being afraid of the complexity of the driv-
clusters 1 (aggressive but observing rules), 2 (anx-
ing task).
ious but mild behaviour) and 5 (reckless but mild
The cluster medians and the median for each
behaviour) seem to be the safest.
factor are reported in Table 2.
Even if the sample is limited and general con-
Drivers in Cluster 1 appear to be very aggressive
clusions cannot be stated, it seems that the com-
but observing traffic rules.
bination of different risky driving styles leads to
In Cluster 2 there are drivers who show scores
a higher accident involvement. The exception
close to the median on most variables. However,
is Cluster 6 for which the level of anxiety seems
they seem to be anxious (e.g. worried about driving
to lead these drivers to be more aware avoiding
other vehicles, in bad weather conditions, etc).
distractions. However even the higher attention
Cluster 3 includes drivers with moderate aggres-
appears to not compensating the increased risk
siveness and recklessness.
due to anxiety.

Table 2. Cluster medians.

Reckless1 Aggress.1 Aggress.2 Careless1 Careless2 Anxious1 Anxious2

Cl. 1 −0,53 1,12 −0,37 −0,31 −0,22 0,05 −0,19

Cl. 2 −0,19 −0,45 −0,06 −0,44 −0,29 0,19 −0,01
Cl. 3 0,08 −0,09 0,85 −0,43 −0,24 −0,48 −0,04
Cl. 4 −0,73 1,47 −0,52 −0,43 1,01 −0,15 1,56
Cl. 5 1,13 −0,58 −0,61 −0,01 −0,23 0,15 −0,26
Cl. 6 −0,11 −0,44 –0,02 −0,19 0,78 0,36 −0,23
Cl. 7 0,11 −0,08 0,61 1,08 −0,05 −0,57 −0,21
Median −0,11 −0,18 −0,06 −0,29 −0,14 0,14 −0,15

840

TIS2017_Book.indb 840 2/8/2017 8:10:16 PM

 Table 3. Margin settings for A4 size paper and letter In these situations, the theoretical lessons should
size paper. provide information about the related traffic rules,
road signs, the safety distance, but also aspects
Statistic DF Value Prob aimed to provide greater driving confidence such
Chi-square 12 218.124 0.0397 as understanding of perception/reaction time at
Chi-square MH 1 77.900 0.0053 different speeds, the importance of visual search,
Phi coefficient 0.4580 the importance of the rear mirrors etc. The practi-
Contingency Coefficient 0.4164 cal exercises should then allow the driver to experi-
Cramer’s V 0.3238 ence the concepts learned on the road. The driver
will be able to tackle those situations that caused
anxiety and insecurity with a better understanding
and a higher confidence.

REFERENCES

Abdel-Aty M.A, Chen C.L, Schott J.R. 1998. An assessment

of the effect of driver age on traffic accident involvement
using log-linear models. Accident Analysis & Prevention,
30, Issue 6, November 1998, Pages 851–861.
Anderson, T.W. (1984). An Introduction to Multivariate
Statistical Analysis. 2nd ed. New York: John Wiley &
Sons.
Arthur Jr., W., Barrett, G.V., Alexander, R.A., 1991.
Prediction of vehicular accident involvement: A meta-
Figure 5. Percentage of drivers involved in a crash by analysis. Human Perform. 4, 89–105.
cluster. Dahlen, E.R., & White, R.P. 2006. The Big Five factors,
sensation seeking, and driving anger in the prediction
4 CONCLUSIONS of unsafe driving. Personality and Individual Differ-
ences, 41, 903–915.
The paper identified seven driving profiles by clus- Deffenbacher, J.L., Oetting, E.R., Lynch, R.S. 1994.
tering drivers according to four different driving Development of a driving anger scale, Psychological
styles: reckless, aggressive, anxious and careless Reports, 74: 83–91.
driving behaviours. Three profiles were mostly Everitt BS (eds), Cluster Analysis. London: Arnold, 1993.
Jobson JD. Principal components, factors and corre-
characterised by a single (risky) driving behaviour
spondence analysis: In: Fienberg S, Olkin I (eds),
(e.g. reckless, aggressive and anxious). While the Applied Multivariate Data Analysis. Volume II:
other four profiles were mostly a combination of Categorical and Multivariate Methods, Paris: Dunod,
two risky driving styles. Careless driving was usu- 1992: 343–482.
ally found in combination with other behaviours. Ma M., X. Yan, H. Huang, M. Abdel-Aty. 2010. Occupa-
The analysis of the identified profiles with acci- tional driver safety of public transportation: Risk percep-
dent involvement shows a significant association tion, attitudes, and driving behavior. Proceedings of the
(i.e. some clusters show a significantly higher level Transportation Research Board 89th Annual Meeting.
of accident involvement than others). Reason, J.T., Manstead, A., Stradling, S., Baxter, J.S.,
Campbell, K., 1990. Errors and violations on the
This statistical approach allows the identification
roads: a real distinction? Ergonomics 33, 1315–1332.
of driving behaviour profiles that could be used for Sanders, N. & Keskinen, E. (eds.) (2004). EU NovEV
driving training purposes. The answers to the ques- project; Evaluation of post-licence training schemes for
tionnaire can for instance highlight an aggressive novice drivers. Final Report. International Commis-
and/or insecure driving behaviour thus tailoring sion of Driver Testing Authorities CIECA, Rijswijk.
the theoretical and practical driving exercises to the Taubman-Ben-Ari O. Mikulincer M, Gillath O. 2003.
specific driver needs, especially for novice drivers. The multidimensional driving style inventory—scale
Answers to the questionnaire can be used to classify construct and validation. Accident Analysis and Pre-
new drivers in one of the above clusters. A non-para- vention 952: 1–10.
Ulleberg, P., 2001. Personality subtypes of young drivers.
metric discriminant analysis can be the statistical tool
Relationship to risk-taking preferences, accident involve-
to attribute drivers to one of the seven individuated ment, and response to a traffic safety campaign. Trans-
profiles, requiring different safe driving trainings. port. Res. Part F: Traffic Psychol. Behav. 4, 279–297.
The analysis has shown that among the traffic Usami, D.S., Giustiniani, G., Persia, L., Gigli, R. 2015.
situations that can create insecurity fall, for exam- Aggregated analysis of in-depth accident causation
ple, driving in congested traffic conditions or on data. International Journal of Injury Control and
roads with many trucks. Safety Promotion. Article in Press.

841

TIS2017_Book.indb 841 2/8/2017 8:10:16 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Solution to a capacity problem using non-conventional intersections

M. Vsetecka, M. Novak & T. Apeltauer

Brno University of Technology, Czech Republic

ABSTRACT: When designing the road network for the urban tender for new southern part of Brno
city center, the authors of this paper came across a problem of heavily loaded intersection, which due to
capacity reasons could not be designed as one-level—due to architectural issues and because of a nearby
small river it was also not possible to use a multi-level design. Focus was therefore put on designing a
non-conventional intersection, which diverts left turns to a neighboring intersection, not to nearby streets.
The nearby streets can therefore be designed as purely service roads. In contrast to a one-way design or a
design with only right turns permitted, this does not lead to lengthening the routes and is also important
for the environment. The traffic at the traffic light-controlled complex of intersections has been tested
using a microscopic traffic simulation, which proved the higher capacity of this solution. Sufficient capac-
ity margin is important not just for the vehicles, but also pedestrians, because otherwise there would be a
pressure to reduce the number of crossings. Several parameters of the results of the model were compared
with conventional one-level and multi-level intersections.

1 INTRODUCTION intersection includes a reverse semi-direct ramp

and turns across the middle separating lane.
Many intersections, which could be labeled as Another example is the multi-level interchange
non-conventional, have been described (Esawey A1 and D549 near Seclin, France (50°32'41’’N
2012). Some moves, especially left turns, are dealt 3°3'21’’E), a type of so-called diverging diamond
with using various non-standard ways in order to interchange (Maji 2013, Hu 2014, Leong 2015) –
increase capacity, by separating the moves into left side traffic is used at a short stretch, which cre-
more intersections. ates free left turns and thus increases the capacity
A good example where such solution has been and reduces delay time.
used is the intersection of the streets Aachenre Nevertheless, use of non-conventional intersec-
Straße and Innere Kanalstraße in Köln/Rhein, tions is rather rare and so the authors welcomed the
Germany (50°56’11.9”N 6°55’28.7”E). This opportunity to co-operate on an urban study, which
required use of such non-conventional intersection.
The goal was to design a non-conventional inter-
section, including traffic control and then compare
this layout with conventional intersection types.

2 SOURCE OF THE PROBLEM

Figure 1. Non-conventional left turns at the intersec- The city of Brno has in long term been expand-
tion in Köln. From the left: semi-direct ramp, turns ing quite unevenly—the southern part of the city
across separating lane and reverse ramp.
center has not been very interesting for investors
for several decades because of the unresolved issue
of the main railway junction placement and the
risk of floods from the nearby Svratka River. The
area therefore currently consists of mostly brown-
fields. One challenge of building a new residence
area here is designing an intersection that would
connect the main city ring with the major road to
Bratislava, which in 2014 had a traffic load of 73
Figure 2. Diverging diamond interchange near thousand vehicles/day. Based on traffic model that
Seclinu—left-side traffic on the bridge. takes into account the new housing development,

843

TIS2017_Book.indb 843 2/8/2017 8:10:18 PM

 the traffic load will increase to 89 thousand vehi- concentrating the traffic corridors in one place,
cles/day and after the planned roads on the city so that the negative effects of the heavy traffic
outskirts are finished, it should drop to 65 thou- impact as small area as possible (see Figure 4).
sand vehicles/day. During the worst period (assum- − No left turns, which would be replaced by traffic
ing the standard ratio of 8% between peak and in minor streets (see Figure 4). This solution is
daily traffic in Brno) the traffic load of the four- not suitable because it would place a load on the
way intersection will therefore be over 7 thousand nearby minor streets.
vehicles/hour, which is close to the upper limit of − Non-conventional intersection, which would
the capacity of a traffic-light controlled four-way thanks to its specific layout, allow handling such
intersection, especially if the intersection is also heavy traffic load required.
supposed to include pedestrian crossings. In addi-
tion, the urban design assumes a denser housing
development compared to the traffic model.
3 NON-CONVENTIONAL INTERSECTION
In general, the following solutions are possible:
− Multi-level interchange, which however is not The requirement was to concentrate all traffic to
suitable due to architectural and urbanistic rea- just one corridor and also not increasing the driv-
sons in an above-ground layout and very compli- ing distance. This limited the possible types of non-
cated to build in underground layout due to the conventional intersections to those, which partly
nearby river and high levels of groundwater. use (in continental Europe quite unusual) left-side
− Distributing the traffic load to nearby roads, traffic that allows left turns without having to give
for example by separating the traffic into one- right of way to oncoming traffic.
way system in parallel streets, which instead of Cross section ahead of the intersection is from
one heavily loaded two-way intersection, would the perspective of a driver entering the intersec-
result in four less loaded intersections and non- tion: On the left a lane for left turns, in the mid-
problematic left turns. This solution was not cho- dle driving lanes for oncoming traffic, on the right
sen because it did not satisfy the requirement of lanes for driving straight and turning right. Vehi-
cles turning left therefore do not have to give right
of way to the oncoming vehicles and thus do not
reduce the intersection capacity.
Crossover of the position of the left-turning vehi-
cles, i.e. the point where they cross the oncoming
traffic, is located ahead of the intersection. Instead

Figure 4. Parallel road for opposite-direction traffic (left),

no left turns, which are replaced by minor roads (in the
middle with just right turn, but with a diversion affecting
four blocks, i.e. approximately 700 m, on the right with left
turns at a different intersection and with no diversion).

Figure 5. Possible solutions to left turns: 1. for capac-

ity reasons, left turns are not possible at the transverse
intersection; 2. left turns are performed at a previous
Figure 3. Map of the location of interest in the context intersection using minor roads; 3. left turns at a previous
of the city traffic system. intersection is maintained, but realized using the main
Map layer—urban design by architect Jana Kaštánková. road and thus creating a non-conventional intersection.

844

TIS2017_Book.indb 844 2/8/2017 8:10:18 PM

 of one collision area in the intersection, there are accelerating), there would be a risk of complete
therefore additional, partial collision areas outside blockage of the non-conventional intersection if
the actual intersection. In dense housing develop- the traffic jams ahead of the individual traffic lights
ment with lots of intersections, it is possible to use extended to the previous intersection. Another risk
nearby intersections for the crossover. associated with non-functional coordination is the
Non-conventional intersection therefore consists so-called stop-and-go effect, which arises as a result
of several intersections—one, which will be called of gradual stopping and accelerating of vehicles,
the central, and others, which will be called satellite. during which congestion arises, which moves in the
Central intersection is the actual intersection of the opposite direction of traffic and reduces the traffic
two major roads. Satellite intersections are used for capacity of the road.
the crossover of vehicles to the opposite side and One of the factors affecting the design of coor-
also for connecting minor roads with the major dinated traffic control is the distance between the
one, although some turns can be limited, and also individual intersections. The size of the housing
for pedestrian and bicycle crossings. The number of blocks (i.e. the distance between the intersections)
satellite intersections corresponds to the number of in the area of concern is 170 to 180 m. The almost
left turns that have to be dealt with non-convention- perfectly symmetrical structure makes the traf-
ally to satisfy the capacity based on traffic load. fic control coordination significantly easier and
allows achieving much better results.
For maximal use of the capacity it is desirable that
3.1 Traffic control
at the central intersections, the opposite direction
In order to maintain smooth traffic it is necessary traffic drives at the same time, and in contrast, at the
to guarantee coordination of traffic control of the satellite intersections, in order to allow the crossover
whole complex of non-conventional intersection of direction, that the opposite direction traffic does
(green waves), in other words mutual coordinated not drive at the same time. One shortcoming of this
control of the central and satellite intersections. If solution is that transverse movements are not pos-
this was not the case, apart from increased delay sible, which is not ideal especially for pedestrians.
time (and the associated increased ecological Assuming the standard speed of 50 km/h and
footprint resulting from additional braking and the above mentioned distance of approximately
175 m, the cycle time necessary for coordination is
50 s. This is unusually short for such heavily loaded
intersection. It is obvious that doubling the cycle
time could be achieved by placing the crossover
point one block further away (assuming same block
size), which would create one additional atypical
intersection between the central and satellite one,
which would allow transverse movements.
Taking into account the surrounding traffic in
the city of Brno, where the cycle time is unified to
100 s, the solution with 50 s cycle time was aban-
doned. The alternative with placing the crossover
point to the intersection further away and creating
one additional intersection was also abandoned,
Figure 6. Non-conventional intersection complex: cen- because it is desirable that the stretch where the
tral (c) and satellite (s) intersections, stretches with partial vehicles are driving on the left side of the road is
left-side traffic highlighted in bold. as short as possible. It was decided that an interval
of 100 s will be used and the crossover performed

Figure 8. Coordination (green wave) of Control plans

of central (C), additional intermediate (I) and satellite
Figure 7. Driving scheme at one part of the (S) intersections for the variant with 100 s cycle time and
intersection. reduced capacity of the crossover.

845

TIS2017_Book.indb 845 2/8/2017 8:10:19 PM

 at the nearest intersection, which however meant a pedestrians (see Figure 12 and 13). Second pos-
reduction in capacity of the crossover. sibility is non-collisional—the green phase for
the crossing pedestrians and turning vehicles
is not concurrent (see Figure 10 and 11). If the
3.2 Pedestrians
Due to urbanistic and architectural reasons, it is
required that the pedestrians cross the intersec-
tion on the same level, even though designing the
pedestrian crossings using different levels would
increase the capacity and also since the pedestrians
would not have to wait for green light, reduce the
delay time.
One-level pedestrian crossing offers in general
two possibilities. One possibility is concurrent
(partially collisional) crossing and vehicle turn-
ing, where the drivers of turning vehicles must
make sure they do not threaten the crossing

Figure 12. Phases of the control plan of the central

intersection for the variant with partially collision pedes-
trian crossing.

Figure 9. Coordination (green wave) of Control plans

of central (C) and satellite (S) intersections for the vari-
ant with 100 s cycle time and reduced capacity of the Figure 13. Control plan of the central intersection for
crossover. the variant with partially collisional pedestrian crossing.

Figure 10. Phases of the control plan of the central Figure 14. Phases of the control plan of the satellite
intersection for the variant with non-collisional pedes- intersection for the variant with transverse movement
trian crossing. possible (see also Figure 9).

Figure 11. Control plan of the central intersection for Figure 15. Control of the satellite intersection for the
the variant with non-collisional pedestrian crossing. variant with transverse movement possible.

846

TIS2017_Book.indb 846 2/8/2017 8:10:19 PM

 ratio of turning vehicles is high, there is no point A possibility is then to create a green phase for
in shortening the green phase for turning and all pedestrian crossings at the same time and
leaving green phase for straight direction, espe- thus also allow diagonal crossings.
cially if there are several driving lanes available.
3.3 Safety and orientation of drivers and
pedestrians
Vehicles driving in the opposite direction obviously
raise the question of safety and orientation.
Ahead of the satellite intersection will be signs
that will clearly navigate the drivers to the correct
driving lane. Using horizontal and vertical traffic
signs and road markings, the driver will be guided
to perform the crossover.
Traffic on the left side of the road will be physi-
cally separated—depending on the final architec-
tural solution of the street either by verge posts or
dividing belt.
Figure 16. Phases of the control plan of the intermedi- Turns at the central intersection create a prob-
ate intersection (see also Figure 8). lem that stems from the (Czech) traffic regulations.
Drivers turning right are not obliged to drive in the
lane most on the right and instead can choose a
lane. Drivers on the left therefore cannot smoothly
turn to the left lane and instead have to give way
to the oncoming right-turning vehicles. Solution to
this problem is road markings, which would con-
verge the right-turning vehicles. Physical separa-
tion is not an option because it would block the
vehicles driving straight.
Pedestrians would cross the road with partial
Figure 17. Control plane of the intermediate left-side traffic only at traffic-light controlled cross-
intersection. ings, without collisions with vehicles coming from
the unusual direction. In addition, a situation where
vehicles drive in the opposite direction also arises in
one-way streets or intersections with traffic islands
and the authors did not encounter any problems
with respect to the behavior of pedestrians in such
situations.

4 ASSESSED VARIANTS OF THE

INTERSECTION

Several variants of intersection configuration were

selected for multi-criterial assessment:

Figure 18. Phases of the control plan of the satellite − Non-conventional, described in previous
intersection in variant that the intermediate intersection section
is used (see also Figure 8). − Conventional four-way intersection with left
turns possible
− Conventional four-way intersection with no left
turns possible
− Multi-level intersection
These variants are then further subdivided
depending on the number of driving lanes (and
thus the required width of the road) based on the
possible crossings. One-level crossing is required in
Figure 19. Control plan of the satellite intersection in the given location, but for general comparison also
variant that the intermediate intersection is used. variants with no one-level crossings were created.

847

TIS2017_Book.indb 847 2/8/2017 8:10:20 PM

 All variants use traffic light control, which is time and two getting lanes at the neighboring inter-
designed uniformly for all variants based on the section, the calculated capacity is the same, however
Czech technical standards TP 81 and TP 235. The given the short distance between the intersections,
cycle time used was 100 s (standard cycle time of there would be significant disruption of smooth
the nearby road network) and for non-conventional traffic and also reduced capacity due to the zip and
intersection variant also 50 s (see section 3.1). stop-and-go effect. An alternative would be to move
Multi-level intersection considered is that of the ramps for the underpass/overpass further away
the so-called “diamond” type, i.e. with one bridge/ from the intersection, in other words extending the
tunnel object for one of the pair of two-way traffic bridge or tunnel and using four-lane configuration
roadways. Taking into account the expected width (see Figure 20). The length of the ramp, assuming
of the road, only one driving lane for each direction 8% slope, 1000 m radius of concave gradient deflec-
is directed via underpass/overpass, which is fully tion, 700 m radius of convex gradient deflection
sufficient for the desired capacity, but could cause and 6 m vertical alignment difference, is 143 m.
problems for intersections nearby. For 50% green
5 ASSESSMENT CRITERIA

Traffic-engineering, architectural-urbanistic and

economic criteria were selected for the assessment
of the created variants. Architectural-urbanistic
criteria cannot be objectively expressed and can-
not be transformed directly into costs. Traffic-
engineering criteria can usually be objectively
described, but directly expressing them as costs is
very complicated and questionable due to estimat-
ing the costs of human resources. For these rea-
sons, objective multi-criterial assessment based on
one parameter (costs) was not used and instead an
approach was used where the individual criteria are
assessed proportionally and anyone can interpret
them based on their own personal preference.

5.1 Traffic engineering: Capacity

Figure 20. Configuration of multi-level intersection
with an underpass (alternatively overpass) for one of the Capacity calculation was done using the standard
pairs of two-way traffic—at the top 4 lanes in underpass, procedure and equation 1:
at the bottom 2 lanes in underpass.
Table 2. Capacity of the intersection variants.
Table 1. Assessed intersection variants.
left direct right sum
C Var. veh/h veh/h veh/h veh/h
Var. Type Lanes* Left turn Ped** s
A1 780 1760 780 13300
A1 non-conventional 1,2,1 yes no 50 A2 640 1760 640 12160
A2 non-conventional 1,2,1 yes yes 50 B1 780 880 780 9780
B1 non-conventional 1,1,1 yes no 50 B2 640 880 640 8640
B2 non-conventional 1,1,1 yes yes 50 C1 840 1880 840 14200
C1 non-conventional 1,2,1 yes no 100 C2 680 1880 680 13000
C2 non-conventional 1,2,1 yes yes 100 C3 590 1320 590 9970
C3 non-conventional 1,2,1 yes yes*** 100 D1 390 880 840 8430
D1 conventional 1,2,1 yes no 100 D2 320 880 680 7540
D2 conventional 1,2,1 yes yes 100 E1 390 440 390 6670
E1 conventional 1,1,1 yes no 100 E2 320 440 680 5780
E2 conventional 1,1,1 yes yes 100 F1 0* 1880 840 10860
F1 no left turn 0,2,1 no no 100 F2 0* 1880 680 10260
F2 no left turn 0,2,1 no yes 100 G1 390 880** 840 11560
G1 multi-level 1,-,1 yes no 100 G2 320 880** 680 10500
G2 multi-level 1,-,1 yes yes 100
Values rounded to the nearest ten, *Left turns realized
* Number of lanes for driving left, straight and right, using minor roads, **Multi-level traffic has capacity
**Pedestrians, ***All-green phase for pedestrians. equal to the saturated flow, i.e. 2000 veh/h.

848

TIS2017_Book.indb 848 2/8/2017 8:10:21 PM

 G the delay time for diagonal movement (delay at
K= SL (1) two crossings) was performed using Equation 3,
C which is derived from the assumption that if the
where K = capacity, G = green time, C = cycle time, imaginary green time on first crossing is 1 s (i.e.
S = saturated flow and L = number of getting lanes. the last second), then the delay time corresponds
The saturated flow is assumed to be 2000 vehicles/ to the sum of delay time based on Equation 2 and
hour for straight direction, for turns, depending on (fixed) green time between the end of green time
the radius (8 to 12 m) it is reduced to 1690 or 1780 at the first crossing and beginning of green time at
vehicles/hour respectively. If there is a collision the second crossing.
of turning vehicles and pedestrians, the saturated
flow is reduced in accordance with the Czech tech- R12
D= +I (3)
nical standards to 1460 vehicles/hour—in reality 2C
of course, this value can be significantly different
depending on the number of pedestrians and the where W = pedestrian delay time, R1 = imaginary
ratio between the green time for pedestrians and red time at first crossing (in reality the cycle time
green time for turning vehicles. The value used is minus one second) and I = interval between end of
based on the Czech standards and corresponds to green time at first crossing and beginning of green
300 pedestrians per hour for one pedestrian cross- time at second crossing.
ing. The actual pedestrian load is very hard to
predict—in this particular case however it is not 5.3 Architectural-urbanistic: Street area
so important, because all the variants assume the Based on discussions with architects and urbanists,
same number of pedestrians or the same coeffi- the following is considered to be undesirable:
cient is used so the results are comparable.
To make the capacity calculation a bit more
clear, Table 3 also includes a more detailed calcula- Table 4. Pedestrian delay time.
tion for one of the variants (A1).
C G R D1 D2
5.2 Traffic engineering: Delay time of pedestrians
Var. s
Calculating delay time for transverse movement
(delay at one crossing) was done using Equation 2, A2 50 12 38 14 37
which can be derived from weighted average delay B2 50 12 38 14 37
(0 s for green light, half of the red time for red light): C2 100 37 63 20 62
C3 100 14 86 37 43*
R D2 100 12 88 39 87
G* R* 2
D= 2 =R (2) E2 100 12 88 39 87
C 2C F2 100 37 63 20 62
G2 100 12 88 39 87
where D = pedestrian delay time, G = green time,
R = red time and C = cycle time. Calculation of C = cycle time, G = green time for pedestrians, R = red time for
pedestrians, D1 = delay time for transverse crossing (1 cross-
Table 3. Capacity of the individual movements for the road), D2 = delay time for diagonal crossing (2 crossroads).
selected intersection (variant A1). * Crossing via 1 diagonal crossing.

G R S K
Turn* L s m veh/h veh/h
VAB 1 22 12 1778 782
VAC 2 22 0 2000 1760
VAD 1 22 12 1778 782
VBA 1 22 12 1778 782
VBC 1 22 12 1778 782
VBD 2 22 0 2000 1760
VCA 2 22 0 2000 1760
VCB 1 22 12 1778 782
VCD 1 22 12 1778 782
VDA 1 22 12 1778 782
VDB 2 22 0 2000 1760
VDC 1 22 12 1778 782
total 13298 Figure 21. Current pedestrian overpass closed to solved
intersection. With this experience is clear, why overpass
*Labeling—see Figure X. is unacceptable.

849

TIS2017_Book.indb 849 2/8/2017 8:10:21 PM

 Table 5. Summary of intersection assessment. 7 CONCLUSION
Var. Capacity Delay Urbanism Costs Non-conventional intersection which uses direc-
tion crossover of left-turning vehicles at a neigh-
A2 88% 100% 100% 100%
boring intersection is a suitable design for the
B2 40% 100% 100% 100%
C2 100% 59% 100% 100%
location of interest from traffic-engineering,
C3 58% 61% 100% 100%
architectural- urbanistic and economic perspec-
D2 24% 0% 100% 100% tive. The capacity of the crossroad would be 50 to
E2 0% 0% 100% 100% 70% higher compared to a conventional four-way
F2 62% 59% 0% 100% intersection because crossing over left-turning
G2 65% 0% 0% 0% vehicles at a neighboring intersection allows bet-
ter and more efficient design of traffic control.
*Number of getting lanes for driving left, straight and right. The average delay time for pedestrians would
also shorten. In the meantime, the architectural-
− Distributing (partly) traffic load to minor streets, urbanistic requirement for one-level intersection
which are only to be used by residents and concentrating the traffic to main roads (not
− Multi-level intersection as a foreign element, affecting the nearby minor roads), would be satis-
which does not fit into the inner city area fied. In addition, the construction costs will be
− Multi-level pedestrian crossings—overpasses lower compared to the variant with multi-level
and underpasses. Pedestrians should be able to intersection.
cross the road on just one level. Non-conventional intersection can be a suitable
solution for other locations as well, if they satisfy
Based on the above it is quite obvious, which var-
the following criteria:
iants can be considered as suitable and which ones
not, from the architectural-urbanistic perspective. − Heavy traffic load (8000 to 12000 vehicles/h/
intersection)
5.4 Economic: Costs − Partially distributing the traffic also to nearby
minor roads is not possible or not desirable
Cost estimates depend on the extent of construction
− Multi-level intersection is not suitable because
considered. Non-conventional intersection does not
of the urbanistic situation or because the costs
include just the central, but also the satellite inter-
of building one are too high.
sections, which would however be built either way
in one form or the other, because the rebuilding of
the traffic skeleton in the location in question is not
limited to just this problematic intersection. ACKNOWLEDGEMENT
The variants assuming multi-level intersections are
quite different in terms of their costs. Similar object This paper has been worked out under the project
was already built in Brno previously for approxi- No. LO0108 “AdMaS UP—Advanced Materials,
mately 100 mil. CZK (approximately 3.7 mil. EUR). Structures and Technologies”, supported by Min-
istry of Education, Youth and Sports under the
“National Sustainability Programme I” and under
5.5 Assessment summary the project FAST-J-14-2360 supported by Brno
Table 5 shows a comparison of the four assessed cri- University of Technology.
teria; all expressed using the same unit (%) – 100%
being the best, 0% being the worst possible result.
REFERENCES
6 MICROSCOPIC SIMULATION Esawey, M. E., and Sayed, T. 2012. Analysis of unconven-
tional arterial intersection designs (UAIDs): State-of-
The purposed configuration of the traffic signals, the-art methodologies and future research directions.
including traffic control and calculation of non- Transportmetrica
conventional intersection capacity, was realized Hu, P., Tian, Z., Wu, X. & Xu, H. 2014. A Proposed Sig-
using microscopic simulation and the Aimsun nal Operation and the Effect of ITS Cycle Length on
8.1.3 software. The simulation confirmed that inter- Diverging Diamond Interchanges Institute of Trans-
portation Engineers. ITE Journal.
section traffic control has a major impact on the Leong, L. V., Mahdi M. B. & Chin K. K.. 2015. Micro-
intersection capacity. Very close attention must be scopic Simulation on the Design and Operational
paid to this in order to prevent stop-and-go effect, Performance of Diverging Diamond Interchange.
which would mean the theoretical capacity esti- Transportation Research Procedia.
mated based on green time would not be fully used Maji, A., Mishra, S., & Jha, M. 2013. Diverging Dia-
because the entrance to the intersection is not fully mond Interchange Analysis: Planning Tool. J. Transp.
saturated as a result of this stop-and-go effect. Eng.

850

TIS2017_Book.indb 850 2/8/2017 8:10:23 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

A cycling mobility study case: The European cycling challenge 2015

in Naples

C. Aveta
Reconstruction Special Office, L’Aquila, Italy

M. Moraca
Freelance Engineer, Naples, Italy

ABSTRACT: In May 2015, the City of Naples and forty others European cities have joined the fourth
edition of the European Cycling Challenge. This event has been created to promote the bicycle as a sus-
tainable transport mean in urban areas, and lasted for all the month of May.
The event organizer, SRM Reti e Mobilità, provided an app where all citizens/cyclists could enroll and
track their cycle journeys; the app was tracking, with a time interval of five seconds, the cyclist position
(using GPS) and some journey details (journey name, length, speed, besides day and schedule).
These data were registered by the app and saved in a database. Several months after the event, each
database has been sent to the participating cities.
Naples’ database has been utilized to evaluate, for the first time, the cycling mobility in the city. Indeed,
the City of Naples developed a cycle network longer than twenty kilometers in the recent years. Therefore,
this evaluation aimed to understand how, when and where the cyclists have been using these paths.
The evaluation required the development of a methodological framework to analyze the database, with
operations in Gis, Excel and Access environment.
The final product has been disaggregated in two categories, the territorial evaluation and the temporal
evaluation. The territorial evaluation contains O-D matrices, an analysis involving Naples’ districts, and a
flow count analysis for road. The temporal evaluation includes an analysis for each day of the month, for each
day of the week (Monday, Tuesday, and so on), and for two time slots (7,30–9,30 A.M. and 4,30–6,30 P.M.).
The overall results registered over 7961 kilometers covered and 1308 registered trips, with an average
journey length of 6,07 kilometers and a massive use of the Waterfront cycle path (Via Francesco Carac-
ciolo and Via Partenope).

1 EXECUTIVE SUMMARY was guaranteed. This analysis represents the first

cycling mobility survey carried out in the City of
In the last ten years Italian cities extended consider- Naples.
ably their cycle route networks. Indeed, the “Urban
Mobility Report” (ISTAT, 2013) reveals that the
1.1 The European Cycling Challenge—event
length of the urban cycle routes is increased by
description
40%, from a value of 13,5 km per every 100 km2
of urban surface in 2008 to a value of 18,9 km per The ECC (www.cyclingchallenge.eu)—whose first
every 100 km2 of urban surface in 2013. edition was held in 2012—is a challenge among
Then, the analysis of the cycle route use is impor- urban cyclists which starts every May 1st and ends
tant for cities, because of the need of quantifying the up in May 31th. The event foresees a virtuous com-
transiting flows and getting information useful for petition among the citizens of involved cities based
future planning: having useful tools is necessary. on the total mileage that the active cyclists/citizens
The aim of this paper is to analyze the results of each city will have traveled in that month. To
extracted from the 2015 European Cycling Chal- participate, is necessary to register all journeys by
lenge (ECC) edition in Naples—utilized as a bicycle through the free App “Cycling365”, or input
case study—defining a methodology for the data own journeys manually via the event website.
processing and showing the obtained results. All journeys are permitted and included in the
The methodology foresees progressive steps: database except journeys with an average speed
in this way, the repeatability to all future sur- greater than 30 km/h, a maximum speed higher
veys based on a similar georeferenced database than 40 km/h, or a length bigger than 30 km.

851

TIS2017_Book.indb 851 2/8/2017 8:10:24 PM

 At ECC can participate all citizens using the bike regional aggregator. These data have been disag-
within the boundaries of the involved city, includ- gregated for transport modes in Table 1.
ing also residents out of the city who move them- In 2011, the City Administration started the
selves into the city by bicycle. For the 2015 Edition, construction of the first cycle route—from Bag-
99 citizens of Naples (and from the neighboring noli to Castel dell’Ovo (Via Nazario Sauro), with
urban centers) have registered their journeys in the a length of 7,7 km—ended up in 2012. The City
month of May. Administration was well aware of the benefits of
the cycling in an urban contest highly congested by
traffic like Naples (Van Hout, 2008) and the con-
2 THE CYCLE ROUTE NETWORK IN struction of a new cycle route wanted to increase
NAPLES the modal split between private vehicles and the
bike in favour of the second.
The City of Naples has an area of 117,27 km2, The cycle route network in Naples was finally estab-
a road network of approximately 1200 km long, lished with the Major order n.1233 of the 9th Novem-
and it is divided in 10 Districts (Figure 1). Naples ber 2012. The cycle route network goes through 4
owns a population of 974.074 inhabitants— Districts, and has a length of about 19 km.
reference to 31/12/2015—and every District has The existing cycle routes in the city of Naples
a population grouped in a range from 80.000 to can be classified in 3 categories:
140.000 inhabitants.
− Type 1: separated cycle routes. The routes are sep-
According to the data provided by the
arated from the road, having a width of 1,25 m
“2011 Italian Census” (ISTAT, 2011), the City
for lane (total width 2,50 m) and defined by
of Naples is interested—on a typical working
delineator strips of width 12 cm. Plus, they have
day—by 574.916 journeys; approximately the 59%
a separator curb from the road—in concrete—
(342.109) of journeys are within the city, while the
with a height of 20 cm and a width of 50 cm,
remaining 41% are classified internal-external or
and are separated from the footpath by the verge
external-internal journeys. Concerning this 41%,
(variable width between 30 and 50 cm);
193.928 journeys are towards Naples, while 38.880
− Type 2: routes on footpaths. They are indi-
depart from Naples; these numbers confirm the
cated through delineator strips having a width
importance of the city of Naples as the principal
of 12 cm, the orange coloring the intermediate
zone and an overall width of 2 m (the width of
the single cycle lane is 1 m). Pedestrians have pri-
ority over cyclists;
− Type 3: routes in pedestrian areas or Restricted
Traffic Zones (“RTZs”). The cyclists follow the
rules of the other road users, and have indica-
tions through markings drawn on the road/
pedestrian area (the bicycle symbol).

2.1 The cycle route network in District 10

The cycle route falling into the District 10 is unique
and has a length of about 3,6 km. The cycle route

Figure 1. The City of Naples, its Districts and the

neighboring cities.

Table 1. Working daily journeys in Naples (for trans-

port modes).

To From
Internal Naples Naples Total

Railway/ 38.019 59.914 2.793 244.176

Underground
Bus 60.135 29.482 4.585 100.726
Car 115.081 98.660 30.435 135.813
Pedestrian/ 128.874 5.872 1.067 574.916
Bicycle
Figure 2. The cycle route network in Naples.

852

TIS2017_Book.indb 852 2/8/2017 8:10:24 PM

 starts from Via Nuova Agnano and ends up in Via and ends up in Piazza Dante, foreseeing ramifica-
Caio Duilio. However, the path is not interrupted tions in Via dei Tribunali and Via Benedetto Croce,
but continues towards Piazza Sannazzaro—through and has been developed in pedestrian area (Via
Galleria delle Quattro Giornate—in District 1. Toledo) and on footpaths (except for the section
The path falling in District 10 has Type 1 near Piazza Dante, RTZ). The intermodal points
stretches (Viale Augusto and Via Caio Duilio) located in the District are several, but the principal
for a length of over 1,2 km, Type 2 stretches (Via is Piazza Garibaldi—the main intermodal hub of
Nuova Agnano and Viale J.F. Kennedy) for almost the city—where converge: Underground Line 1,
1,8 km, and Type 3 stretches (Piazzale Tecchio and Underground Line 2, the railway called “Circum-
Largo Lala, pedestrian areas). There are several vesuviana” (which connects all the cities around
intermodal points that intercept the cycle route: Vesuvius, starting from Naples) and the bus hub,
Underground Line 2, Underground Line 6 (under starting point of the main city bus lines. In the area
construction), the railway line called “Cumana” it is also possible the connection with the ferry area
and the bus station located in Piazzale Tecchio. starting and arriving in the Port of Naples (Calata
Porta di Massa). From 2015, there are stations of
2.2 The cycle route network in District 1 bike-sharing located in Piazza Bovio, Via Brin, Via
Toledo corner Via Diaz and Piazza Dante.
The overall length of the cycle routes falling into
District 1 is about 7 km. In District 1 there is a
main cycle route and two secondary routes that are 2.4 The cycle route network in District 4
linked to the main. The main route carries on from The cycle routes located in District 4 transit both
Galleria delle Quattro Giornate and ends up in Via into the Historical Center and ends up in Piazza
Toledo, entering into Historical Centre. The sec- Garibaldi, following two parallel routes. Indeed,
ondary paths start from Piazza Vittoria, crossing the two cycle routes (exclusively Type 3) move
the historic area of Chiaia neighborhood (Piazza through the pedestrian area of Via dei Tribunali
dei Martiri, Via Morelli, Via Chiaia) and Via and Via San Biagio dei Librai. The cycle route net-
Toledo, through Via Santa Brigida and Via Paolo work in District 4 has a total length of 3 km, and
Emilio Imbriani, ending up in Piazza Matteotti. it has been developed in pedestrian areas and into
In this District there are Type 1 stretches RTZ of the Historical Center.
(Galleria delle Quattro Giornate and the “Water- The first cycle route, called “Decumano Mag-
front” from Via Francesco Caracciolo intersection giore cycle route” with the historical reference to
Via Sannazzaro to Via Nazario Sauro intersec- the road on which the route stands, starts from Via
tion Via Raffaele De Cesare), for a total length of Benedetto Croce, and ends up in Piazza Garibaldi.
3,5 km, while the rest is Type 3 in pedestrian areas The second cycle route, named “Decumano Supe-
(Via Toledo, Via Chiaia, Piazza Plebiscito) or in riore cycle route” (similarly to the previous),
promiscuous with vehicles (the “30 km Zones” in starts from Piazza Dante, and ends up in Piazza
Via Santa Lucia and Via Cesario Console). Again Garibaldi too. Unlike other Districts, intermodal
the intermodal points are numerous along the cycle connections are scarce because of the size of the
network: Underground Line 1, Underground Line streets and high population density: indeed, the
2, Underground Line 6, the Cumana railway, and only exception is the Duomo station of Under-
the Central and Chiaia funiculars. ground Line 1, next to activation. However, in this
From 2015, bike-sharing stations (currently sus- District were located since 2015, the bike-sharing
pended service, but in reactivation) are present near stations of Largo Donnaregina, Via Benedetto
Castel dell’Ovo, Piazza Vittoria and Via Partenope. Brin and Piazza Garibaldi (side District 4).

2.3 The cycle route network in District 2

The cycle routes crossing the District 2 are connected 3 THE METHODOLOGICAL
to the Historic Centre of Naples (UNESCO World FRAMEWORK
Heritage), have a total length of 5,2 km and the
characteristic of being only in a promiscuous way The App “Cycling 365” saved, for each journey
with vehicles and pedestrians (cycle routes Type 3). registered and with a frequency of 5 seconds, the
The main cycle routes are two: the first, defined cyclist position (using GPS) and some journey
as “Decumano del Mare”, originates from Piazza details (journey name, length, besides day and
Garibaldi, and ends up in Via Roma after crossing schedule) in a database from the start to the end of
Via Benedetto Croce. This cycle route is developed the journey. Several months after the event, each
both in pedestrian areas and on footpaths, 30 km database was sent to the participating cities.
Zones and the RTZ so-called “Mezzocannone”. The extraction of results from the European
The second cycle route starts from Via Toledo, Cycling Challenge database has required the devel-
continuing the cycle route coming from District 1, opment of a methodology to define all the stages of

853

TIS2017_Book.indb 853 2/8/2017 8:10:25 PM

 the elaboration process, from the survey objectives
to the specific type of investigation to be carried
out, in order to ensure the replicability and com-
parison of results, obtaining understandable results
by an unskilled audience. In detail, interested stake-
holders have been divided in two main categories:
− Municipal technical offices, having the task of
assessing the current status of the situation,
according to the objectives identified, and invest-
ing resources in the development of the urban
cycle network;
− Decision makers and carriers of public interests:
subjects that, for institutional reasons or role,
should guide the choices of the public adminis-
trators in the sector.
Figure 3. Methodology framework.
The methodological framework foresees the
following steps, and is based on the typical scheme
of transport surveys (Richardson & Ampt & the overall analysis of the month of May, from
Meyburg, 1995): 1st to 31th.
Classifications introduced allow to obtain a bet-
− Definition of the objectives; ter detail level for the analysis. Joining one single
− Definition of the key factors; layer belonging to each key factor allows to get a
− Phenomenon analysis according to the identi- 3 × 4 matrix, in which each layer of the Space key
fied key factors; factor is analyzed according to 3 different layers of
− Data processing; the Time key factor (Figure 3).
− Evaluation of the results. The join permits to make a progressive detailed
The survey objective is to provide an assessment analysis and assesses whether there are particular
of the satisfaction degree and the use of existing phenomena, such as home-work journeys, home-
cycle routes in the City of Naples, providing infor- school journeys, and how the movements are dis-
mation on the city areas for developing future cycle tributed over a time period and on the territory.
routes. These outputs require the definition of a set Data processing operations have been devel-
of “key factors”, which allow the analysis to achieve oped with the aid of different software tools (GIS,
the proposed objectives. The database provided spreadsheet, SQL database). Indeed, the cycling
from the European Cycling Challenge has a struc- flow analysis in a GIS can provide information to
ture in which each row show information about the evaluate where cyclists choose to move and to pre-
registered journey (journey's identity code, single dict where they could move in the future (Aultman-
point identification code belonging to the specific Hall & Hall & Baetz, 1997).
journey), time information (day of detection, hour, Operations have been performed starting from
minute and second of detection) and spatial infor- a first cleaning job of the database in GIS envi-
mation (latitude and longitude detected by GPS). ronment to eliminate some redundant information
For this reason, the key factors identified are 2: in the survey, such as the elimination of registered
“Space” and “Time”. The key factors have been dis- journeys outside Naples, or registered journeys
aggregated in different “layers”, to evaluate differ- on railway lines or motorways. The database thus
ent aspects of the phenomenon at different levels: obtained has been divided in accordance with pre-
dicted slots of the Time key factor, representing the
− Time key factor: mobility demand. The three sub-databases have
− 0 24 time slot layer; been combined with the GIS containing informa-
− 7,30–9,30 time slot layer (“Morning” time slot); tion about the City of Naples (which represents the
− 16,30–18,30 time slot layer (“Afternoon” time supply model of this procedure), using as joining
slot); factor the geographical coordinates of individual
− Space key factor: points (the database rows) presents in the sub-
− City layer; databases. The join of three sub-databases with
− District layer; the supply model, declined in the various detail
− Matrix Origin/Destination (O/D), layer obtained scales of Space key factor, has allowed to obtain
using the start and end points of the registered the 3 × 4 matrix displayed in Figure 3.
journeys; The evaluation of the results foresees the analy-
− Street layer. sis of the information obtained during the data
Each Time layer is referred to the entire processing in order to provide to all stakeholders
month: for example, the 0–24 time slot layer is the information necessary for its own needs.

854

TIS2017_Book.indb 854 2/8/2017 8:10:25 PM

 4 SURVEY RESULTS of registered journeys of Districts 3, 6, 7, 8, and 9 is
less than 10% of the registered overall journeys.
The ECC database registered more than half a The journeys shown in Table 3 are higher than
million points (equal to rows), which represent the those counted in Table 2: this is because a single jour-
“photographies” of all journeys made in May, with ney can go through in more than a single District.
a time interval of 5 seconds. The District with the highest mileage registered
The operations performed on the database, is the District 10 with about 2.080 km, while the
necessary to extract the data, have been predomi- District 3 has registered the lowest mileage with
nantly of “spatial join” type, and various query almost 64 km. The sum of the registered mileage in
types using the SQL environment that is present in District 1 and in District 10 is equal to the 50% of
the GIS software. Extracted data have been finally the overall mileage registered in the ECC, symptom
exported to a calculation sheet to create the tables that a separated cycle route is more attractive for
available in the successive paragraphs. cyclists. Results are similar for the other time slots,
The accuracy of the GPS data varies in some with prevalence of the District 10 in the morning
cases in a 5–10 meters range from the street. This time slot and District 1 in the afternoon time slot.
is a limit also evidenced in literature (Lindsey &
Hankey & Wang & Chen, 2013) because of it is Table 2. Overall journeys and mileage about the ECC
impossible to understand if cyclists used cycle route 2015.
or went in promiscuous with vehicles, nevertheless it
allows to understand in which streets they passed. Time slot Journeys Mileage (km) Km per journey
The results have analyzed the most significant
data for the survey: the registered journeys and the 0–24 1.293 7.960,94 6,16
covered mileage. Morning 213 1.616,78 7,59
Afternoon 251 1.409,01 5,59
4.1 Overall analysis
The total number of journeys registered in the Table 3. Registered journeys inside Districts.
month of May 2015—during the ECC—is 1.293,
while the total mileage was 7.960 km, for a 6,16 km Journeys per time slot
average mileage per journey.
District 0–24 Morning Afternoon
The day with the largest number of registered
journeys was on May 11th, with 84 journeys and 1 713 104 110
a total mileage of 426,39 km (5,08 km per jour- 2 257 41 35
ney). The day with the most mileage covered was 3 28 8 1
on May 19th with 448,37 km, 70 journeys and of 4 293 36 36
6,41 km average mileage per journey. 5 422 80 87
The analysis of the database by time slots shows 6 118 27 33
that the recorded journeys are distributed during 7 27 0 5
the day, without recording special spikes refer- 8 47 3 5
able to home-work journey or home-school jour- 9 38 8 2
ney. Indeed, only 16% of the total journeys were 10 542 110 95
recorded in the “Morning” time slot, while 19,4% Total 2485 417 409
of those registered in the “Afternoon” time slot.
Furthermore, the morning journeys are longer
Table 4. Mileage per registered journeys.
than the afternoon journeys (7,59 km per journey
in the morning time slot, compared to the 5,59 km Mileage per time slot (km)
on average registered in the afternoon time slot).
District 0–24 Morning Afternoon
4.2 District analysis
The second level of analysis provides a count of 1 1.765,50 192,04 327,36
the registered journeys inside the 10 Districts. 2 680,55 101,47 164,19
3 63,76 31,36 –
In Table 3 is shown that in the District 1 were
4 778,83 25,84 159,79
recorded 713 journeys, while in the District 10 were
5 1.414,50 441,20 189,51
recorded 542 journeys. Surprisingly, also the Dis-
6 478,91 64,90 195,76
trict 5 registers a high number of registered jour-
7 78,09 – 6,47
neys, considering is a hilly terrain. The same trend
8 149,04 9,67 26,38
was true also for the other time slots: however, in 9 108,28 34,61 –
the District 5 journeys in the afternoon time slot 10 2.080,01 667,94 267,25
are almost the same than in District 1, and higher Total 7.597,46 1.569,04 1.336,72
than those recorded in the District 10. Plus, the sum

855

TIS2017_Book.indb 855 2/8/2017 8:10:25 PM

 4.3 O/D Matrix analysis directrix (17 journeys and 93 km covered). Jour-
neys from neighboring cities to Naples (47) have
The third level of detail has analyzed the origin and
generated about 355 km (Table 9).
the end of journeys among the Districts (consid-
Journeys starting from Naples to neighboring
ered as internal centroids) and between the Naples
cities have generated about 343 km (Table 10). Also
Districts and the neighboring cities (considered as
in this case the Naples-Pozzuoli and the Napoli-
external centroids). Over 90% of the total journeys
Volla directrixes are the most traveled directions,
registered have been done within the city of Naples
the first with 143 km and the second with 80 km.
(with over 7200 km counted), with a small fraction
of journeys to and from Naples (internal-external
journeys and external-internal journeys). Table 9. Mileage per journeys starting from neighbor-
ing cities towards Naples.
The journeys registered in the morning and
afternoon time slot followed this trend (Table 6 Mileage per time slot (km)
and Table 7).
The preferential directions for external-internal Starting City 0–24 Morning Afternoon
journeys related to the 0–24 time slot (Table 8 and
Table 9) are the Pozzuoli-Naples directrix (15 jour- Arzano 25,93 21,85 –
neys and 113 km covered) and the Volla-Naples Casoria 27,68 – –
Cercola 11,81 – –
Table 5. Overall O/D Matrix, 0–24 time slot. Marano 5,05 – 5,05
Melito 23,20 – –
Internal External Mugnano 47,06 – –
Destination Destination Pozzuoli 113,68 25,89 9,65
Quarto 7,15 – –
Internal Origin 1193 51 Volla 93,71 – 50,97
External Origin 48 1 Total 355,27 47,75 65,66

Table 6. Overall O/D Matrix, Morning time slot. Table 10. Journeys starting from Naples towards neigh-
boring cities.
Internal External
Destination Destination Journeys per time slot
Internal Origin 204 5 Destination City 0–24 Morning Afternoon
External Origin 4 0
Arzano 2 1 0
Casoria 2 – –
Table 7. Overall O/D Matrix, Afternoon time slot.
Cercola 2 – –
Internal External Marano 1 – 1
Destination Destination Melito 1 – –
Mugnano 6 – –
Internal Origin 197 5 Pozzuoli 15 3 1
External Origin 10 1 Volla 1 – –
Total 17 – 8

Table 8. Journeys starting from neighboring cities

towards Naples. Table 11. Mileage per journeys starting from Naples
towards neighboring cities.
Journeys per time slot
Mileage per time slot (km)
Origin City 0–24 Morning Afternoon
Destination City 0–24 Morning Afternoon
Arzano 2 1 –
Casoria 2 – – Arzano 20,58 2,83 2,54
Cercola 2 – – Casoria 4,75 – –
Marano 1 – 1 Cercola 22,40 – –
Melito 1 – – Marano 9,90 – 9,90
Mugnano 6 – – Melito 21,25 – –
Pozzuoli 15 3 1 Mugnano 40,16 7,34 11,58
Quarto 1 – – Pozzuoli 143,73 32,18 14,10
Volla 17 – 8 Volla 80,92 – –
Total 47 4 10 Total 343,70 42,35 38,12

856

TIS2017_Book.indb 856 2/8/2017 8:10:25 PM

 From the internal-internal component of the Table 16 and Table 17 show the results about
matrix O/D, journeys with origin and destination journeys starting from District 10, with 144
falling inside the same District have been extrapo- journeys and 1433 km. The connection between
lated. The District where have been registered the District 1 and District 10 is the most evident
largest number of internal journeys is the District (449 km total), then the connection with District 2
1, followed by District 10 and District 5. The jour- (378 km), and District 5 (264 km).
neys shown in Table 12 cover the 25% of the total
mileage. 4.4 Street analysis
In Table 13 are presented data on total mileage
divided by District and by time slots. The District 10 The ECC database permitted an evaluation of the
and District 5 registered almost the same mileage. cycling flows transiting in the streets of the City of
Table 14. Journeys starting from District 1 towards the
4.3.1 District 1 and District 10 detail other Districts.
From the internal-internal component of the O/D
matrix the journeys starting from District 1 and Journeys per time slot
District 10 have been extrapolated, because of the
data registered the highest number of journeys in Arrival District 0–24 Morning Afternoon
that zones. For District 1 were recorded 167 jour-
10 53 11 5
neys per 1033,73 km.
2 35 5 7
The mileage counted for the District 1–District 3 6 – 1
4 connection was 250 km, 245 km was the mile- 4 38 1 3
age for the District 1–District 10 connection and 5 20 4 9
181 km for the District 1–District 2 connection. 6 5 2 –
7 1 – –
Table 12. Internal journeys per District. 9 9 – 1
Total 167 23 26
Journeys per time slot
Table 15. Mileage per journeys starting from District 1
District 0–24 Morning Afternoon towards the other Districts.
1 167 23 26 Mileage per time slot (km)
2 87 9 21
3 9 4 0 Arrival District h24 7.30−9.30 16.30–18.30
4 98 0 22
5 121 45 10 4 312,89 81,06 23,29
6 30 4 14 10 181,69 27,29 33,99
7 10 0 0 2 63,53 – 14,87
8 8 1 1 5 257,04 3,94 15,51
9 20 8 0 3 121,90 20,18 46,70
10 145 46 21 6 50,98 22,56 –
Total 695 140 115 9 4,07 – –
7 41,63 – 5,30
Total 1.033,73 155,03 139,67
Table 13. Mileage per internal District journeys.
Table 16. Journeys starting from District 10 towards
Mileage per time slot (km) the other Districts.

District 0–24 Morning Afternoon Journeys per time slot

1 1.033,73 155,03 139,67 Arrival District 0–24 Morning Afternoon

2 542,70 78,07 162,54
3 55,76 31,36 – 1 57 18 7
4 605,82 – 129,88 2 35 13 3
5 841,90 289,69 79,51 5 26 – 9
6 344,42 37,92 188,53 6 15 14 –
7 69,96 – – 8 4 – –
8 85,49 7,28 16,48 4 3 1 1
9 108,28 34,61 – 9 3 – 1
10 1.437,33 568,87 157,66 3 1 – –
Total 5.125,38 1.202,83 874,28 Total 144 46 21

857

TIS2017_Book.indb 857 2/8/2017 8:10:26 PM

 Table 17. Mileage per journeys starting from District 5 CONCLUSIONS
10 towards the other Districts.
The results showed that the cyclists preferred
Mileage per time slot (km) Districts where there are separated cycle routes
Arrival District h24 7.30–9.30 16.30–18.30 (District 1 and District 10), or in close proximity.
Flows drop progressively with the increasing dis-
1 449,52 164,95 62,63 tance from these areas, a sign that is necessary to
2 378,98 171,76 17,61 increase the trust of citizens in the outlying areas,
5 264,86 – 64,47 possibly creating new protected routes.
6 238,90 221,93 – The results prompted the City Administration
8 47,15 – – at approving the project of a cycle route in Corso
4 28,86 10,23 9,70 Umberto, one of the busiest streets used by cyclists
9 13,14 – 3,26 according to the data available. In addition, in 2013
3 12,04 – – the City Administration had already approved the
Total 1.433,45 568,87 157,66 construction of new cycle routes in own way in
the stretch adjacent to the Port (Via Marina, Via
Amerigo Vespucci and Via Alessandro Volta, Dis-
Table 18. The 15 busiest streets in Naples per time
trict 2), and in the Industrial Area (Via Emanuele
slots.
Gianturco, District 4), projects currently under
Journeys per time slot construction, for further another 5 km of cycle
routes in the east area.
Street 0–24 Morn. After. The success of the 2015 edition of the ECC
was confirmed by the results of the 2016 edition,
Via F. Caracciolo (1) 324 54 65 in which there were recorded 14.701 kilometers
Via Nazario Sauro (1) 235 40 46 (nearly double of the 2015 edition), and where
Via Caio Duilio (10) 185 44 26 from a first analysis of flows (available on the event
Via Partenope (1) 185 30 38 website), emerges that the favorite routes are the
Galleria Laziale (10–1) 182 41 22 same as the 2015 competition. In the future, travel
Via Nuova Agnano (10) 171 47 27 demand models could be developed using informa-
Via Toledo (1–2) 159 41 25 tion taken by this kind of survey.
Corso Umberto I (2) 156 22 23
Viale J. F. Kennedy (10) 141 37 22
Corso V. Emanuele (1) 137 56 27
REFERENCES
Via Bagnoli (10) 133 45 19
Via Salvator Rosa (2) 120 36 15 Aultman-Hall, L., Hall, F. & Baetz, B. 1997. Analysis of
Viale Augusto (10) 91 22 18 Bicycle Commuter Routes Using Geographic Infor-
Via Torquato Tasso (1) 68 31 7 mation Systems. Implications for Bicycle Planning.
Via Domenico Fontana (5) 63 11 18 Transportation Research Record 1578.
Istituto nazionale di statistica (ISTAT), 2011. 15° Censi-
mento della popolazione e delle abitazioni 2011.
Naples utilizing the single journey characteristics Istituto nazionale di statistica (ISTAT), 2013. Urban
(with geographic coordinates) and the map (in GIS Mobility Report.
environment) of the City of Naples, to get flows Lindsey, G., Hankey, S., Wang, X. & Chen, J. 2013. Fea-
as the sum of journeys falling in a single street. sibility of Using GPS to Track Bicycle Lane Position-
ing. University of Minnesota, Center for Transportation
Table 18 highlights the top 15 busiest streets, with Studies.
in parentheses the belonging District. Richardson, A., Ampt, E. & Meyburg, A. 1995. Survey
Seven streets on the list are Type 1 cycle route, Methods for Transport Planning. Eucalyptus Press.
and fall in District 1 and District 10. Except Via Van Hout, K. 2008. Literature search bicycle use and
Toledo—pedestrian area—the others are all main influencing factors in Europe. Instituut voor Mobiliteit
traffic streets. (IMOB).

858

TIS2017_Book.indb 858 2/8/2017 8:10:26 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Secondary or no longer used rail infrastructures: A new tool for the

regional railway and land planning

A. Cappelli
Department of Architecture and Arts, IUAV University of Venice, Venice, Italy

G. Malavasi & S. Sperati

Department of Civil, Building and Environmental Engineering, Sapienza University of Rome, Rome, Italy

ABSTRACT: To reach a sustainable development of the territory, it’s implied that the choice of trans-
portation methods must be related to the city form: compact or sprawl.
On opening or upgrading railway infrastructures or stations, have increased the real value (m2) of both
existing and new buildings located next to the station of the railway lines.
This research aims to demonstrate how the process of locationing and the development of the rail
transportation supply could re-organize the growth of the demand in exploited areas by providing an
innovative contribution on transportation planning using the Smart Regional Railway System.
Analyses were performed on two railway lines: Civitavecchia-Orte and Civitavecchia-Roma.
For the research purposes, it has been created an index, called “Cross Section Spatial Index”. The inno-
vation of this index is the way real estate value was considered, namely depending on the distance from an
efficient transportation system.

1 RESEARCH OVERVIEW specialization of the transport system, still strongly

tied to the car and, consequently, to the private
The research aims to provide an innovative contri- sphere (Brovarone E. 2010).
bution to the regional transport planning, through The reviewed studies have generally said and
a modeling approach defined as “Smart Regional confirmed that:
Railway” [SRR] calibrated on two case studies: the • Intermodal transport is the essential approach
lines Civitavecchia-Orte and Civitavecchia-Roma. to transport-related issues, as it allows to con-
The effectiveness of the existing regional trans- tain the external costs of transport and make it
port is part of a sustainable vision of collective more efficient (Cappelli A. 2006);
transport (Ricci S. 2013), where the environmental • The individual transport may be discouraged, in
consciousness and the need to modify the territory favor of public transport, through an innovative
have consolidated the importance of a connection territorial planning mindful ofcompact spatial
between land use and transportation planning distribution of the assets (Cappelli A. 2009);
(Maternini G. 2014). • The geographical dispersion does not create
The focus on the study of the existing rail net- competitiveness and convenience to the public
work rather than on expanding or building new transport operator, as it produces movements
infrastructure stems from the interpretation of the over long distances through private means of
current policies that are driving their intervention transport (Cappelli A. et al 2013);
on modernizing and upgrading the existing (Min- • A higher density (Fig. 1) creates more profitabil-
istry of Economy and Finance, 2016). ity for public transport as it collects more move-
The scientific literature confirms the link ments and increases the transportation demand
between land use and transport planning, moreo- (Laconte P. 1996);
ver it underline show correct urban and transpor- • A reduction in the use of private cars can only
tation planning decisions can affect the method of be sustained by a range of efficient and competi-
transport, the modal choice and the resulting qual- tive public transport (Cascetta E. et al 2013);
ity of life of the citizen. • The sustainability of the transport system
The urban planning decisions made in these descend from a balance between the economic,
years have defined the territorial form of today, environmental and social component (Cappelli
while future urban form is overwhelmed by the A. et al 2015).

859

TIS2017_Book.indb 859 2/8/2017 8:10:27 PM

 growth were the border areas, the ones between the
old town and the countryside.
These areas had the feature required for that era:
low density and greater comfort. In these areas,
created quickly and supported by urban planning
strategies, appropriate means of support were not
always present, for instance an adequate public
infrastructure network.
Therefore this implied that most of the move-
Figure 1. Difference between Low-High density. ments were with private vehicles (Laconte P.
1996).
Even nowadays, the regional public transport
The analyzed literature confirms, therefore, as presents different complications depending on the
the sitting choices of residences and businesses can territory: for example, in “third world countries”,
affect the development and success of public trans- the system is threatened by lack of funds, while
port and vice versa. in developed countries is declining because of the
A design aimed at the sustainable development predominant use of private cars.
of public transport, with particular reference to
rail, must be based not only on the performance of
the transportation system and on operating costs;
2 EVALUATION INDEX
even of different factors, such as: population size,
territorial density and the sustainable and orderly
2.1 Methodology
development of the territory.
In summary, it can be said that the choice of “The soil is an asset of common interest for the
mode of transport is related to the form of the city, Community, although it is mainly privately owned,
meant as an urban or suburban area, diffuse or and which, if not protected, will undermine the
compact. In compact areas, the organization and sustainability and long-term competitiveness of
management of public transport are more profit- Europe.”(European directive, 2014).
able. Therefore they represent a valid alternative to The soil, as a resource, produces public benefits
the private car (Abbadessa C. 1996). (transport, environmental and social terms).
Each area has its own characteristics, opportu- Nowadays, however, the accessibility to land use
nities and problems, which affect the user’s pref- information is not entirely guaranteed: the regions
erences: for example, in the case of a low density have taken the lead in preparing databases on soil
area, the choice generally falls on the private car, uses, but only few of them are now able to provide
probably in view of a lack of effective alternatives. accessible data in a simple and free way.
Studies on new or upgraded high-speed lines have This research deals with the study of the rela-
shown that, on opening or upgrading a rail infra- tionship between transport planning and land use
structure, or a station, the property value of new or planning, placing a constraint on a better use of
existing buildings located near stations or railway the existing rail infrastructure at a regional level
lines (Martinez F. et al 2013) is increased (m2). (Italy) and defining economically sustainable
The bibliography (Leonetti M. 2015) deals with development plans.
the reuse of brown field or abandoned rail infra- To understand the structure of the territory in
structure from different points of view: in the ex-ante and ex-post stage, indicators represent an
various examples, the plans are used for different ideal key-role tool to read urban phenomena.
purposes from the reactivation of the line (tourist The indicators, since aim different objectives,
paths, cycle paths and greenways). are fundamental. For this reason, it’s important to
The guideline of the European Union is clear choose them carefully to prevent them from being
(European Commission, 2011): the need to reduce ineffective and harmful in decision-making (Mead-
transport externalities, the land consumption and ows D. 1998) or analysis.
emissions into the atmosphere by 2050 is a priority. In this context, the choice of the indicators was
In 1996, at the European Conference Habitat important (Pileri P. 2011): simplicity of applica-
II, the debate was on demographic urban develop- tion, statistical significance and consolidated
ment. The discussion underlined how metropoli- application were used.
tan areas grew both in size both of inhabitants, The interpretation is another delicate and cru-
while rural areas were increasingly subject of cial phase of the research: the indicator use is not
abandonment. automatically true. In fact although it reduces
The growth of urban areas had been so fast that uncertainty, the indicator could be effected from
the services (including transport) couldn’t follow the context in which it applies (Greeuw S. et al
the urban development. The areas with the greatest 2000).

860

TIS2017_Book.indb 860 2/8/2017 8:10:28 PM

 To analyze the changes of land use compared The research, carried out on British study (Mar-
to rail infrastructure and to conduct comparative tinez et al 2013) cases which concern the trend of
analysis, it’s important to rely on indicators (or the real estate market at the time of the construc-
indices) which are normalized on reference areas tion and opening of a new AV line (using various
(eg. m2) or on demographic variables (eg. The parameters such as the proximity to the station and
number of inhabitants). This will make possible to the hedonic pricing), shows that the investments in
properly compare different territorial units. transport infrastructures have often played a central
role in the renovation of urban fabric. Two types of
2.2 Indicator choice and measures effects can be observed: one concerning the territory,
In order to perform the analysis of the territory mainly linked to the localization of residences and
affected by railway lines Civitavecchia-Orte and economic activities; the other one regarding econ-
Civitavecchia-Roma, paying particular attention omy and the price variation of properties in proxim-
to the relationship between spatial density (residen- ity to the station due to the increase in accessibility.
tial, production, etc.) and regional railway system The study shows that the opening of a new railway
performance (expected or possible), the following infrastructure contributes to the development of the
indicators were measured (OSDDT 2014). area and that people chose residences in the closest
proximity to it even if they have to pay more.
For example, in the case of the Stratford station
2.3 Territorial indicators
opened in 2010, during the first years of construc-
a. “Territorial population density” and “intensity tion (2004) there was a +8,1% rise in the local real
of land use” indicators [ab./ha or km2]: estate value, which then slightly decreased between
• Territorial population density: (permanent 2005 and 2008, when the build process started; it
residents)/(total land area). reached a negative peak of –13,5% with the advent
• Intensity of land use or territorial net popu- of the recession, but then grew to +4,56% in 2010
lation density: (permanent residents)/(urban when the station was first opened.
land area). Given the proper fiscal measures, the price vari-
b. Composition use index/land cover or coverage ation resulting from the improvement of the trans-
ratio composition (CLi): port services performance could support collective
Relationship between the surface of a certain transport by reaching the value (Budoni A. 2014)
class/land use and the entire area considered as the (once common procedures are defined) which
study-unit (eg. Common). Composition indicators however would not cover the investment entirely
are also used to measure the heterogeneity or, on and would still need public financing.
the contrary, homogenization of the landscape. The efficacy of indicators originates in its avail-
CLi = (LCi / LCtot) * 100 where: LC = land cover/ ability, but in the case of studies on real estate value
surface LCtot = total surface study-area and it’s rather complex to detect its variations over time.
i = type of land cover. Because of that, an indicator named “Cross Sec-
c. Used land per capita indicator: tion Spatial Index [CSSI]” was created for research
Relationship between the used land surface and purposes: this indicator is based on measuring a
the resident population in the area, expressed in real estate value in relation to its distance from an
m2/hab—ha/hab. efficient transport system offer.
Cab = CS/hab where: Cab: used land per inhab- However, even if the variation of real estate value
itant; CS: used land [m2; has]; hab: number of overtime is unknowable, we can still measure it
inhabitants [num]. through space and location; namely, we can observe
d. Building number and built surface average: how it changes in different territorial environments
These indicators are useful to understand the when it’s close or far from the local transport sys-
quality and the main features of the built area. tems. This way it is possible to evaluate a real estate
value in relation to the distance from an efficient
2.4 Transport indicators transport offer. The indicator was applied with a
1–2–3–5 km buffer from the station involved.
a. Coefficients and equipment index of the struc- The Spatial Cross Section substitutes for Time
tures/human infrastructure: Cross Section and allows us to create an economic
Infrastructures equipment or infrastructure evaluation model and improve our knowledge in a
density of the transport network (rail and road) field that is currently lacking contribution.
for the reference surface [m/km2].

2.5 Real estate value indicators 2.6 Calibration and application of indicators
in study cases
An additional indicator is the variation of the real
estate value compared to the efficiency of the rail- Once picked the most suitable indicators, territorial
way infrastructure. and infrastructural analysis have been carried out

861

TIS2017_Book.indb 861 2/8/2017 8:10:28 PM

 and will be carried out on three regional railway
lines: the unproductive line of Civitavecchia-Orte
(Fig. 2), the exploited line of Civitavecchia-Rome
(Fig. 3), focusing the efforts at locating the critical
points linked to territorial density and to the cor-
relation with the railway transport system.
These analysis were carried out using desktop
open source QGIS software with which each railway
line was studied, transposing every building (regular
or not) of the relevant district into planimetry and
tracing the existing road and railway infrastructures.
Furthermore, the green areas and the most fre-
quently reached destinations were depicted as well.
The area of study was divided into 1 km quad-
rants of territory crossed by the railway comparing
that to a pedestrian section which could be easily
traversed (Winkler B. 1990).
The choice of the routes is not random: the Civ-
itavecchia-Orte line is a dismissed regional railway
which shows how the presence of a railway infra-
structure can transform and guide territory devel-
opment; Civitavecchia-Rome is redeveloped line,
from an organizational point of view.
The purpose of the research program is to fully Figure 3. Civitavecchia-Roma line.
optimize the existing regional railway transport
system by renovating and enhancing it (namely innovative and inter-modal transport system, which
increasing both the quality and the quantity of the through a precise regional planning could allow to
offer), in areas where nowadays the demand would obtain an improvement in the transport quality,
not seem to be enough to justify such a work, but a change in the modality selection and finally an
which could be driving forces for the local develop- enhancement of life quality for the citizens.
ment in the near future. The research also aims to show the possibil-
Fundamental elements for the project are the ity of the opposite process as well; namely that
improvement of accessibility and the creation of an the improvement and enhancement of the public
urban railway transport system could bring about
the increase of demand in areas which are not been
taken advantage of for whatever reason (comfort,
accessibility, travel duration and frequency). Public
transport has to adapt to the environment in order
to increase its market share over the private means
of movement.
These studies will result in the creation of a
simulation model which could depict the situation
and simulate its changing, thus providing sup-
port to decisions concerning territory planning
with the purpose of improving collective public
transport.
The originality of this model consists in the intent
to increase the collective transport offer as a con-
tribute to a territory development that could justify
and represent a potential demand for mobility.

3 RESULTS AND FUTURE PROCESSING

To sum up, from the attached tables (Tables 1–2)

we can conclude how the application of the above-
mentioned indicators allows to define the main
Figure 2. Civitavecchia-Orte line. features of the two examined railway lines.

862

TIS2017_Book.indb 862 2/8/2017 8:10:28 PM

 Table 1. Civitavecchia-Orte indicators.

Figure 4. Comparing the average population density

(Italy, Civitavecchia-Orte line and Civitavecchia-Roma
line).

Table 2. Civitavecchia-Roma indicators.

The information refer to 2015 ISTAT sources

and allows to draw an up-to-date outline of the Figure 5. Hotspot.
situation.
Regarding the single track line of Civitavecchia-
Orte, how shown in the table (Table 1), the average To support the information brought about by
residential density in the involved districts is 139 the first indicator, the CAB indicator shows an
ab/km2, compared to the national average of 201 limited built up areas spread on a wide area.
ab/km2. (Fig. 4) Lastly, the dominant infrastructural outfit is
Peculiar situations are represented by the Civi- road network.
tavecchia district, in which the average residential The double track line of Civitavecchia-Roma
density is 719 ab/km2, and by Fabrica di Roma dis- (Table 2) has a population density of 719 ab/km2,
trict, in which it is 243 ab/km2; being the only two far above the national average.
districts above the national average. The CLI indicator spreads out evenly on the
The territory coverage index (CLI) shows a really territory, as we can deduce by observing the map
low percentage for the Civitavecchia-Orte line with (Fig. 3), in which the distribution of the buildings
values which go from 0% (regarding the buildings far along railway tracks is depicted.
from the station) to 30% (buildings near the station): In this line the special shape is dictated by the
these values show how the framework of buildings linear distribution of the buildings alongside the
is distributed in hotspots, meaning that tiny built up existing railway line, differently from the hotspot
areas are spread throughout the territory. pattern of the previous line (Fig. 5).

863

TIS2017_Book.indb 863 2/8/2017 8:10:28 PM

 Leaving out the case of Rome, the CAB varies The presented results are to be considered par-
between 175 m2 and in 75 m2 in all the districts, tial since new studies are being carried on in order
which displays similar land use per inhabitant. to reach further developments and more in-depth
Lastly, like in the previous case, the infrastructures analysis is with which to build the model and apply
shape stresses a predominance of roads network. it in the final phase.
Concerning the new CSSI indicator, the results The research turned out to be very difficult,
shown in tables 3 and 4 confirm how the station acts both for the collecting of the information to be
as an attracting agent in the territorial context. analyzed and for the related numerical and carto-
For the Civitavecchia-Roma line, it was more graphic processing.
problematic to acknowledge the same role to the Therefore more in depth studies and generaliza-
station due to land features that made it ambigu- tions are being carried out. However, the already
ous case: within the four analyzed buffers, the accomplished results allow us to confirm the esti-
value of real estate sales for an 85 m2 apartment mation method and to make some first interesting
are mostly similar in some cases a further distance considerations about the potential development of
from the station resulted in an increase of value, railway transport systems in not densely populated
due to the proximity to the sea. areas.
In this particular context the territorial structure
has two main attracting agents, the station and the
shoreline represents the two major poles of interest. 4 FINAL REMARKS
The Civitavecchia-Orte line instead, is the bet:
given the hotspot distribution of the line, some of With the methodology of the study defined and on
the stops display very limited built up areas nearby; the basis of the elaboration of data and analyses
furthermore, the wider the distance from the sta- undertaken, it is possible to express considera-
tion the lower the real estate value will be. tions which will allow for the continuation of the
research.
In the first instance, in the opinion of the
Table 3. Real estate sales Civitavecchia-Orte line. authors, the use of the parameter “Cross Section
Spatial” will allow for the evaluation of the effects
of the aperture or closure of one railway line on
the value of real estate and the consequent effects
on territorial development.
At present it is possible to affirm that the relative
effects are “strong”, in particular in the medium/
low earnings areas where the accessibility to trans-
port is strongly required and the cost of private
transport impacts significantly on the families’
budget.
The situation differs in the urban areas among
the medium/high earnings bracket where accessi-
bility to transport is far better and the presence of
the railway line (barrier effect and acoustic impact)
is more significantly felt.
The research, therefore, will carry out an in-
depth study of these differences highlighting
that the principal objectives of the study are the
regional systems and railway lines in disuse or in
various states of abandon and not densely popu-
Table 4. Real estate sales Civitavecchia-Roma line. lated, urban areas. It is in these areas that the prob-
lem appears more relevant and in need of further
study inasmuch as in the dense urban areas the
railway services, even if inadequately served, main-
tain their role as a valid public transport.
During the research it became apparent that
the information pertinent to real estate values was
lacking with regards to specific territorial settle-
ments and above all historic trends.
This lack of information does not allow for a
historic data bank that can establish a direct link

864

TIS2017_Book.indb 864 2/8/2017 8:10:29 PM

 between the evolution of the transport system and Il sistema di metropolitana regionale della Campania.
the presence or absence of railway connections. Railway Engineering.
For the reasons mentioned above the “Cross Sec- Coppola E., February 2012. Densificazione vs dispersione
tion Spatial” parameter was suggested as a means urbana. TeMa.
European Commission, 2011. White Paper on trans-
to investigate the property market trend in relation portation. Roadmap to a Single European Transport
to the distance of efficient transport (in particular Area-for a competitive transport system.
railway in areas of medium/low earnings). European directive n° 2004/35/CE.
Obviously, were historic data banks available Greeuw, S., M. van Asselt, J. Grosskurth, C. Storms,
over a sufficiently extended period of time it would N. Rijkens-Klomp, D. Rothman and J. Rotmans. 2000.
have proved simpler to study the various correla- Cloudy crystal balls. An assessment of recent European
tions of the role of railway transport. It is the and global scenario studies and models. Environmental
authors’ belief that such information be collected issues series No17. European Environmental Agency,
and stored and that the proposed indicators may Copenhagen.
Laconte, P. November 1996. Regional Rail in Low-Density
consent to a sufficient in-depth study as to warrant Areas. Japan Railway & Transport Review.
an evaluation of a strategic nature. Leonetti M. July 2015. Reti ferroviarie secondarie: analisi
e prospettive per uno sviluppo sostenibile dei territori
REFERENCES marginali. Thesis, La Sapienza.
Martinez F., Pagliara F., Tramontano A., 2013. Valore
Abbadessa C. September 1996. L’integrazione dei sistemi edonico dell’accessibilità relativo agli immobili ad uso
di trasporto nell’area metropolitana diffusa veneta. residenziale: processo di offerta casuale ed applicazi-
Railway Engineering. one ad un nuovo collegamento ferroviario. Railway
Brovarone, E. March 2010. Pianificazione urbana e com- Engineering.
portamenti di viaggio. TeMa. Maternini G. March 2014. Trasporti e Città: mobilità e
Budoni A. May 2014. Catturare il valore dl suolo per pianificazione urbana, EGAF Edizioni, Transportation
sviluppare reti di trasporto locale su ferro. Railway Engineering Editorial series.
Engineering. Meadows, D. September 1998. Indicators and information
Cappelli A., Libardo A., Nocera S., Sardena A. & systems for sustainable development. The Sustainability
Antognoli M. 2015. Accessibilità e qualità dei nodi di Institute.
interscambio per lo sviluppo dei servizi di trasporto fer- Ministry of Economy and Finance, Economic and
roviario regionale. Proceedings of the IV conference Financial Document 2016, attachment: Strategie per
Safety and Exercise Rail: solutions for the Rail Trans- le infrastrutture di trasporto e logistica.
port Development (SEF). OSDDT. 2014. Come calcolare gli indicatori.
Cappelli, A. & Libardo A. & Nocera S. November 2013. Pileri P. 2011. Misurare il cambiamento. Dalla percezione
Teorie, strategie ed azioni per uno sviluppo efficiente del alla misura delle variazioni d’uso del suolo. Land use
trasporto regionale. Railway Engineering. in Lombardia region in the last 50 years. Lombardia
Cappelli, A. 2006. Il nuovo corridoio ferroviario Torino- region.
Lione. Railway Engineering. Ricci S. December 2013. Ingegneria dei Sistemi Ferroviari:
Cappelli, A. September 2009. Il costo sociale del trasporto tecnologie, metodi ed applicazioni. EGAF Edizioni,
e della logistica in Italia. Il Mulino. Transportation Engineering Editorial series.
Cascetta E., Cartenì A. & Carbone A. March 2013. La Winkler B. Marzo-Aprile 1990. Piano della mobilità per la
progettazione quality-based nel trasporto pubblico locale. città di Bologna. Parameter.

865

TIS2017_Book.indb 865 2/8/2017 8:10:30 PM

TIS2017_Book.indb ii 2/8/2017 8:00:28 PM
 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Delay estimation on a railway-line with smart use of micro-simulation

F. Cerreto, S. Harrod & O.A. Nielsen

Department of Management Engineering, Technical University of Denmark, Kongens Lyngby, Denmark

ABSTRACT: This paper formulates a delay propagation model that estimates total railway line delay
as a polynomial function of a single primary delay. The estimate is derived from a finite series of delays
over a horizon that spans two dimensions: the length of the railway line and the number of trains in the
service plan. The paper shows that the total delay estimate is a cubic relation for small primary delays.
A probabilistic approach is presented to combine the total delay functions of primary delays given to dif-
ferent trains. The final estimate is the total delay on railway lines, after a random incident has occurred.
The model can be integrated in railway timetable analysis to reduce the number of necessary simulations,
and can be used when the computation speed is an issue, such as on-line rescheduling algorithms. The
model is demonstrated with an analysis of a Danish suburban railway.

1 INTRODUCTION disruptions on a railway line depends on the inter-

actions between the trains, and a different total
Operational stability and robustness are crucial for delay function is derived for each scheduled train.
railway transport. Not only are the passengers or The probabilistic approach presented in this paper
users of the service sensitive to these measures of allows to estimate the contribution of the individ-
quality, but railways are usually integrated systems ual trains to the general function of the total delay
or networks, and failures at one location of the sys- on a selected railway line.
tem affect other locations and services, sometimes Using microsimulation, the model can be shown
quite catastrophically. Railway network planners to be robust to deviations in assumptions, and the
are faced with many decisions about what quality results may be used to establish bounds of expected
of service to provide and what resources to allo- performance of simulation models, and thus reduce
cate to deliver this service. Much of the literature the use of simulation models in preliminary, explor-
demonstrates that there are frequently multiple atory studies. Railway microsimulation is known
feasible alternatives to allocate resources, and each for its heavy computational requirement, and the
alternative has a unique performance profile with models proposed in this paper introduce new esti-
characteristic statistics, especially with regards to mation of the total delay on railway lines with a
punctuality and robustness. The analysis of these very limited used of microsimulation, restricted to
alternatives frequently requires laborious and the initial calibration phase.
inconclusive modeling with simulation software.
This paper contributes to the literature with a
1.1 Literature survey
closed form function estimate of aggregate railway
line delay propagation in response to a primary Prior literature on operational stability and delay
delay. Many railway and transit services are of the propagation may be classified as proposing para-
form of a single terminating railway line, and this metric methods, providing analytical methods, or
function may supplement or replace the applica- demonstrating applications of simulation.
tion of simulation for exploration of alternatives. Parametric measures are functional relations
On many railway lines passenger traffic is distrib- fitted to empirical or experimental data. In these
uted over the line destinations, and aggregate delay measures, the cause and effect relation may not be
is an appropriate measure of system performance clearly understood, or it may be strongly limited
and customer service. to selected environments. Krueger (1999) presents
This formulation is closed form under a set of many capacity estimation functions proposed for
assumptions that is later shown to be robust to and validated on North American long distance
variance. The formulation is derived from a finite railways. International Union of Railways (2004)
series of deviations from the service plan (second- defines procedures using timetable compression
ary delays) caused by a singular initial disruption to estimate the capacity of European high density
(primary delay). The total delay generated by railway lines. Gorman (2009) fits linear multiple

867

TIS2017_Book.indb 867 2/8/2017 8:10:32 PM

 regression functions to large data sets of train oper- capacity consumption and the development of the
ations on a North American railway to estimate disruptions, but does not take into account the
delay as a function of train planning decisions. recovery of train delays according to the timetable
Analytical methods derive system performance allowance. Cerreto (2016) extends Landex’s delay
measures from known or presumed cause and propagation model to include the timetable allow-
effect relations in the railway service plan. Among ance. The total delay on a railway line is described
these, Hasegawa et al. (1981) applies a hydrody- as a composite polynomial function of the pri-
namic analogy to model railway traffic. The study mary delay generated at a station, which is cubic
models the delay propagation as a shock wave in for small primary delays. The model allows to cal-
a compressible fluid and finds the total delay as culate the total delay with a limited use of micro-
a cubic function of the primary delays by means simulation, but it returns a different total delay
of propagative velocity. Harker and Hong (1990) function depending on the first train delayed.
estimate the delay on a mixed single and double Simulation is widely used, experimentally and in
track railway where the train path is not defined in practice. Relevant publications include Lindfeldt
advance and is subject to a stochastic dispatching (2015), which extensively applies RailSys com-
decision en route. Higgins et al. (1995) formulate mercial railway simulation software to a variety
decision rules for operation on a single track rail- of capacity and delay propagation topics. In par-
way, and then calculate in closed form the expec- ticular, Lindfeldt simulates 336 timetable scenarios
tation of system delay given a traffic pattern and and then applies linear regression to determine the
defined probabilities of delay for trains, track seg- significance of many common heterogeneity meas-
ments, and terminals. ures in predicting aggregate secondary delay. Lind-
Railway delay models often lead to innovations feldt finds that the mean pass coefficient, a measure
in mathematics, such as Meester and Muns (2007) of the frequency of meets and overtakes, is the
application of phase-type distributions. Meester most significant indicator. Mattson (2007) uses
and Muns derive the net delay distribution on con- microsimulation to study the interferences between
nected railway network segments given the distri- trains under different capacity utilization values:
bution of primary delays on each segment. The Mattson finds this to be the most precise way to
derivation asserts that recursive calculation of the analyze secondary delays, but it is also demanding
solution may be attained with just three opera- for very detailed input and the process is very time-
tions: sum, nonnegative excess beyond a bound, consuming. Lastly, Cerreto (2015) applies Open-
and maximum. The paper states that a phase- Track commercial railway simulation software to
type distribution, a distribution of the absorption the analysis of a 21 km. line (nine stations) in the
time of a continuous time Markov chain, can be Netherlands with four configurations ranging from
contained in the three operations in closed form. double track to quadruple track. Cerreto investi-
However, the method depends on the assertion of gates methods to reduce the computation neces-
independence of the primary delays. The method sary for simulation based analyses and limits the
is demonstrated for a sample network of 24 direc- number of simulation runs required with a heuris-
tional line segments with seven transfer points. tic process called the skimming method. Instead of
Goverde (2010) presents an efficient delay prop- simulating all combinations of trains and delays, a
agation algorithm where timetables are modeled as composite profile of train delay is estimated from
timed event graphs (using max-plus algebra) and an initial simulation analysis, and this composite
initial delays are known. The algorithm is very delay function is used to calculate the aggregate
fast and in a few seconds can calculate the delay system delay. The results demonstrate that capacity
propagation over a large network consisting of utilization is not strongly correlated with aggregate
many interdependent services, such as the Dutch secondary delay, which contradicts the findings of
national railway timetable. However, the model some other literature.
offers no functional relationship, and results must
be calculated for each scenario separately. Kroon
et al. (2007) proposes a stochastic linear program 2 INCIDENT, PRIMARY DELAY
for the optimal allocation of supplement time PROBABILITY AND TOTAL DELAY
along the route of a train path and finds that in
a variety of realistic scenarios the supplement Delays are positive deviations between the realized
time should not be allocated uniformly along the times and scheduled times of activities. In the litera-
train path. Finally, and most closely related to ture, different classifications of delays are available.
this paper, Landex (2008) proposes a delay prop- Most of the classifications distinguish between
agation model computing the transfer of delay delays that are due directly to the variability of proc-
between trains through the scheduled buffer times. ess times and delays that are originated by the subse-
This model is used to study the relation between quent conflicts in the actual operation (Goverde &

868

TIS2017_Book.indb 868 2/8/2017 8:10:32 PM

 Hansen, 2013). The primary delays are unexpected
extensions of the planned times of the individual
processes scheduled. For instance, equipment fail-
ures and large passenger flows generate primary
delays. The secondary delays, on the other hand,
are delays generated by operation conflicts, which
are due to primary delays themselves. When a train
is delayed, it needs to use infrastructure elements at
different times than planned. A conflict arises when
two or more trains request to use the same element
at the same time: they will be queued by dispatching
decisions, since only one train at time can use one
element or track section. The delay that generates
from the queuing is called secondary delay.
The cumulative delay, or total delay, on a railway
line is the sum of all the total positive deviations
registered for all the trains at all the time measure-
ment points.
The delay generation process begins with a dis-
ruption or incident. A primary delay generates
when the failure intersects a scheduled event in the
timetable, and secondary delays evolve from the
interaction between different scheduled events in
the timetable.
The model presented in this paper translates the
probability density distributions of incidents on a
railway line into the probability densities of primary
delays and of secondary and cumulative delays.
The section below describes the probability of
generation of primary delays to a selected train,
given the characteristics of the incident and the
timetable.
Figure 2. Probability density functions of the start-
2.1 Probability of primary delay to one train ing time, the duration, and the ending time of an inci-
dent. The last graph shows the joint probability density
We consider those incidents that prevent trains domain of t ′ and t ′′ .
from moving. Such events can be, e.g. failures at
signal boxes, extended boarding times at stations, (2013) in Switzerland; finally Zilko et al (2016) pro-
failures at other ground or onboard systems. pose an on-line model to predict the duration of
In several cases, it is possible to describe an inci- a failure, based on the available knowledge at the
dent by the distributions of its starting time and beginning of the failure. The model uses the Cop-
duration. The distribution of the starting time could ula Bayesian Networks to estimate the contribution
be assumed uniform in preliminary studies, when of given influencing factors, based on historical
detail information is not available. The distribution data. In early studies the information available on
of the incident duration is still subject of studies in the incidents may be insufficient to estimate these
the railway field. Meng and Zhou (2011) propose distributions, so we take the relaxed assumption
the Normal distribution to model the disruption of uniform distribution also for the incident. The
duration on single track lines, while the Exponen- formulation is, thus, simplified, but could be inte-
tial distributions is used by Schranil & Weidmann grated with specific distributions fitted to the given
incidents. Our model translates the probability of
incidents into the probability of primary delays by
integration of the probability densities. The struc-
ture of the model would not be affected choosing
different distributions of the incident durations.
Figure 1. Delay generation process: Primary delays We define t ′ the starting time of an incident,
happen when incidents cross scheduled events. Second- t ′′ its ending time, and τ its duration, so that
ary delays generate from delayed scheduled event cross- t ′′ = t ′ + τ . Both t ′ and τ are assumed uniformly
ing other events in the timetable. distributed, on independent ranges:

869

TIS2017_Book.indb 869 2/8/2017 8:10:32 PM

 t ′ ∈U (tmin ′ )
′ ,tmax
τ ∈U (0,τ max )

Consequently, t ′′ follows a trapezoidal distribu-

tion in [t′min, t′max + τmax] (Figure 2).
The central segment of the distribution spans
from min{tmax ′ , max } to max{tmax ′ , max }, and its
2
constant value is q= .
t′ − t′ + τ
max min max
+ τ max
− t′
max

We define Li the event “Train i experiences a pri-

mary delay”. The departure time of train i from
the considered station is named θi, and the time
separation between the train i−1 and the train i is
the headway hi = θ i − θ i −1. The incident generates a
primary delay to the train i if it starts between the
departures of trains i−1 and i, and it ends after the
scheduled departure of the latter, θi.
Figure 4. Joint conditional probability P (t ′′ | t′ ∈ ( , θ ]).
(t′t ∈( )) ∩ (t′′
t > i)
i i −1 i

Li i i

The intersection probability is expressed by

⎧ hi
means of the conditional probability:
⎪1 − 2τ for hi < τ max
⎪
(
P ( Li ) = P t ′′ > θ i | t ′ (θi θ i ]) P (t ′ ∈(θ i 1 θ i ]) (1)
max

⎪ τ
P ( E1i ) = ⎨ max for hi > τ max
t″ depends on t″ through τ, and the conditional ⎪ 2 hi
probability is derived hereunder. ⎪ 1
⎪ for hi = τ max
The conditional probability density of ⎩ 2
( | t ′ ∈( i i ]) has a trapezoidal shape in the range
[θi–1, θi + τmax], with a central constant segment in P(E2i) is proportional to the headway of the
the range { i ,τ max + θ i },
} max{{θ i , max i 1}], train in the timetable cycle:
and height q = hi max 2
max τ max hi
. The joint condi-
tional probability corresponds to the striped area hi hi
P (E 2i ) = =
in Figure 4.
In the following formulation, equation (1)
∑h i i
c
is split into two factors for a simpler explana-
tion. We name the conditional delay probability where c is the timetable cycle, and is given by the
sum of all the headways. The probability of every
P ( E i ) = P (t i | t ′ ∈ ( i ,θ i ])
and the event
probability P ( E i ) = P (t (θ i −1,θ i ]) that corre- train in the cyclic timetable to experience a primary
sponds to the start of the incident between trains delay is given by the following:
i−1 and i. The probability of E1i depends on the
relation between τmax and hi and is described by the ⎧ hi ( max hi )
⎪ for hi < τ max
⎪ 2c ⋅ τ max
following:
⎪ τ max
P ( Li ) = ⎨ for hi > τ max (2)
⎪ 2c
⎪ hi
⎪ for hi = τ max
2c
⎩

Note that the probabilities of the individual trains

to receive primary delays do not sum up to 1. We
denote P(0) the probability that no train is delayed,
Figure 3. A train receives a primary delay if the inci- that is
dent begins (t′) in the previous headway (hi) and it ends
(t′′) after the scheduled departure time (θi). P (0) 1 ∑P ( L ). i (3)
i

870

TIS2017_Book.indb 870 2/8/2017 8:10:32 PM

 2.2 Combined total delay functions is a common operating plan in Europe and urban
North America. The time horizon of the model
A total delay function describes the relation
then begins with the departure of the first train at
between primary delays given to a train and their
the beginning of the line and ends with the arrival
cumulative effect on the railway line. Different train
of the last train at the end of the line.
paths in a timetable are characterized by different
The total delay d represents the unweighted util-
stopping patterns, running time supplements and
ity loss experienced by the railway service due to
headway buffer times towards the following trains.
a disruption. It is the sum of all individual delays
Therefore, at every train path scheduled corre-
at measurement points in the timetable over the
sponds a characteristic total delay function of the
analysis horizon, and is presented in (6).
primary delay.
We combine the characteristic total delay func-
tions of different trains in a general total delay d ∑( d
js
js dj s ≥ ) (6)
function that represents the effect of a primary
delay to any of the trains in the timetable. The gen-
eral function is a weighted average of the individ- with dj,s being the delay of train j registered at sta-
ual curves, where the weights are proportional to tion or timing point s (difference between real and
the individual probabilities of the trains to receive scheduled time).
a primary delay. The individual train delay dj,s is a combination of
The general total delay function is expressed by the hindrance from previous trains and the residual
delay from the previous station. The delay is trans-
ferred to following trains due to a lack of buffer
d ∑w ⋅ d ( i )
i
i (4) time, while a train keeps a residual delay from the
previous station due to a lack of running time sup-
with weights plement. Equation (7) expresses the delay propaga-
tion on the two dimensions of the model, under the
P ( Li ) relaxed assumption of equal running time supple-
wi = (5) ment a for all the trains between any pair of stations
∑ P (L )
j j and equal buffer time b between any pair of trains.

Equation (4) allows the estimation of the general dj s p − (s )a − ( j )b (7)

total delay given by an aleatory incident through
the combination of total delay functions generated Subject to non-negativity constraint: d j , s. j s
by selected trains. The estimation of individual p is the primary delay, which corresponds to the
total delay functions is relatively simple. In the first train’s delay at the first station d1,1 = p.
following section we describe a delay propagation The total delay is derived summing up the
model to calculate the total delay d(i) as a cubic individual train delays at all the stations. It results
function of the primary delay given to train i. in (8).
We reduce considerably the simulations neces-
sary to estimate the general total delay combining p p (s −1)a
the model described below and the probabilistic a b

approach. d ∑ dj s = ∑
j , s |d j s > 0 s =1
∑
j =1
p (s − 1)a ( j − 1)b
(8)
(a + 3ab )
2
a b 2 1 3
2.3 A finite series model of the total delay as a = p+ p + p
function of the primary delay 12ab ab 6ab

Previous literature demonstrates that the total The equation is valid for small values of primary
delay on a railway line can be described as a cubic delay that expire before the last train and before
function of the primary delays given to a train. the last station.
Cerreto (2016) models the total delay from the Cerreto validates the model using microsimula-
service timetable at all measurement points, as a tion on a Danish suburban railway line with het-
function of timetable supplement, timetable buffer, erogeneous timetable. The model is robust and
and a single initial delay to one train. The model is holds valid when the assumptions of equal running
summarized in this section. time supplement and buffer times are removed.
The total delay model has a two dimensional The total delay on the line can be regressed to a
analysis domain, namely the length of line and cubic polynomial function. The application to a
the number of trains included in the cumulative heterogeneous timetable, though, returns a differ-
delay statistic. Trains on a single line with a sin- ent cumulative delay function for each train that
gle direction of movement are considered, which receives a primary delay.

871

TIS2017_Book.indb 871 2/8/2017 8:10:35 PM

 We introduce the index i to identify the total The line stationing and the schedules are sum-
delay function d(i) resulting from a primary delay marized in Table 1.
given to train i. The defined set of {1,…,10} minutes of primary
The general total delay function is derived in delay was assigned separately to each train depart-
section 2.2 combining the individual functions ing from Hellerup. The individual total delay func-
through the probability of each train to receive a tions were regressed from the corresponding total
primary delay. delay measured in the simulation, independently for
line A and line E. The general total delay function
of the line is calculated by the weighted average of
3 CASE STUDY: THE NORDBANE IN the individual total delay functions of the trains.
COPENHAGEN For the model validation, we Monte Carlo sam-
pled n = 200 failures at the departure signal from
We simulated the operation of a suburban railway
Hellerup, starting at a random time independent
line in Denmark to validate the combination of
of the timetable. We identified the first impacted
different polynomial functions to describe the total
train and measured the primary delay generated
delay against the primary delay. The suburban
by the failures. We measured the related total delay
railway network in Copenhagen is a very densely
developed on the line and regressed it to individual
occupied network with 2 minutes headway in the
cubic functions for lines A and E. The starting time
busiest section. Six different lines operate on the
of the disruption was extracted from a uniform
network, five running on the same central section.
distribution between 0 and 80 minutes, spanning
The suburban line is operated by uniform rolling
over 8 consecutive timetable cycles. The duration
stock in cyclic timetable. The selected section of
of the failure was extracted from a uniform distri-
the suburban network is the line from Hellerup to
bution between 0 and 10 minutes.
Hillerød. Overtakes in this section are prevented.
Table 2 compares the cases of primary delay
Though it is theoretically possible at selected sta-
experienced by each train line and the calculated
tions, it hardly occurs in real operation, due to the
very high frequency of the train service.
The micro-simulation software OpenTrack by Table 1. Line stationing and scheduled.
OpenTrack Railway Technology Ltd. and the Swiss
Federal Institute of Technology (ETH Zurich) was Station Stationing Schedule*
used for the simulation. This micro-simulation uses
continuous computation of train motion equations Name Code km A E
and simulates the interaction between trains through
Hellerup Hl 7,8 05 07
discrete processing of signal box states (Nash &
Bernstorffsvej Btf 9,3 | 09
Huerlimann, 2004). Given user defined infrastruc-
Gentofte Gj 10,9 | 11
ture, rolling stock, and timetable databases, it is pos-
Jægersborg Jæt 12,6 | 14
sible to calibrate the train paths defining the running
Lyngby Ly 13,9 11 16
time supplements; moreover, different driving behav-
Sorgenfri Stf 15,9 | 19
iors can be modelled for on time trains and delayed Virum VG 17,7 | 21
ones. The strength of the micro-simulation models Holte Hot 19,0 16 23
is the higher accuracy than the analytical models, Birkerød BG 23,8 21
and their flexibility to represent different contexts. Allerød LG 29,3 26
Changes in the infrastructures and operating rules Hillerød HG 36,5 32
can easily be implemented and tested. The accuracy
comes, though, at the cost of much longer computa- *Departure minutes of the hour reported. Each train
tion time, as well as set-up time. Other micro-simula- path repeats every 10 minutes. | = pass-through.
tion software is available on the market, like RailSys
by Rail Management Consultants GmbH (RMCon). Table 2. Cases of primary delay registered in the simu-
Despite some differences in the approach, both the lation and probabilities modeled.
mentioned software suffer from long time needed to
compute such detailed models (Landex, 2008). Cases of Model
Two different train paths run every ten minutes recorded probability of
Course primary delay primary delay Weight
on the line between Hellerup and Hillerød with
two different stopping patterns: (i) # % P(Li) wi
− Line A: runs throughout the entire line, skipping 0 76 38.0% 34.0%
5 stops in the first stretch A 91 45.5% 48.0% 0.73
− Line E: only runs the first stretch, stopping at all E 33 16.5% 18.0% 0.27
the stations.

872

TIS2017_Book.indb 872 2/8/2017 8:10:36 PM

 In this case study, a series of 200 microsimulation
of a random failure at a signal box was well approx-
imated by a reduced series of 20 microsimulations
of primary delays to individual trains.
The modeled total delay function and the meas-
ured total delay hold tight up to 500 s of primary
delay. This is due to a higher number of trains
from line E that received smaller primary delay.
As opposite, trains from line A tended to be the
first delayed trains with higher values of primary
delay. For this reason, the joint regressed total
delay function is closer to line E for small primary
delays and closer to line A at higher values of pri-
mary delay. Instead, the modelled general total
delay function keeps the distance ratio between
the two individual total delay functions through-
out the entire primary delay range. A solution to
this issue could be to cluster the distribution of the
failure duration. In this way, different probabilities
to be delayed can be calculated for different ranges
of primary delays. The averaging weights would be
calculated for individual clusters and the general
total delay function would adapt to the probability
to be delayed of individual trains.

5 CONCLUSIONS

This paper derives the total delay on a railway line

as a closed function of the primary delay, under
the assumption of equal buffer time between trains
Figure 5. Total delay on the line as a function of the and equal running time supplement over the line.
primary delay given to Line A (dotted dark gray line) and We turn the estimation of the total delay given
Line E (dotted light grey line). Modelled (dashed black by an aleatory incident into the combination of the
line) and measured (dot-dashed line) general total delay. total delay functions of different trains. We deter-
mine each function’s contribution to the general
probability. The weights for the general total total delay with a probabilistic approach. The indi-
delay function are calculated from the modelled vidual total delay function of each train is regressed
probabilities. from microsimulation. The result is a combined total
The total delay general function of the railway delay function that does not depend on what train
line was regressed from the whole set of simula- receives the primary delay. It is now possible to esti-
tions. and compared to the combination of the mate the consequences of given incident, simulating
individual delay functions. independent primary delays on the individual trains
Figure 5 compares the modelled general total instead. This allows broader timetable analyses with-
delay on the line and the measured general total out increasing the number of simulations needed.
delay from the simulation. The model allows to calculate the total delay on a
railway line with high accuracy from the microsim-
ulation, reducing the amount of simulation runs
4 RESULTS AND DISCUSSION needed. Using this model, we needed only one tenth
of the microsimulations used to estimate the total
The total delay on a railway line can be regressed to delay from the incident distributions. The number
a cubic function of the primary delay. Every train of microsimulations needed for the analysis may
that receives the primary delay returns a different be further reduced, taking advantage of the good
function, due to different interaction with the fol- regressions of the individual total delay functions.
lowing trains, i.e. different buffer time. This is relevant for railway planners, because it
The weighted average total delay function allows timetable accurate analyses with a limited
reflects the total delay function given by the joint computational power, or on extended railway net-
simulation of failures independent of the timetable. works. At the same time, the accuracy of the model,

873

TIS2017_Book.indb 873 2/8/2017 8:10:36 PM

 together with the reduced computation needs, in Computer Science (Vol. 4359 LNCS, pp. 41–68).
allows new applications in real-time rescheduling Springer. http://doi.org/10.1007/978-3-540-74247-0_2
models, based on the total delay estimation. Krueger, H. (1999). Parametric modeling in rail capacity
The model accuracy could be further improved planning. In WSC’99. 1999 Winter Simulation Confer-
ence Proceedings. “Simulation - A Bridge to the Future”
in the future clustering the distribution of the (Cat. No.99CH37038) (Vol. 2, pp. 1194–1200). IEEE.
incident duration and introducing more complex http://doi.org/10.1109/WSC.1999.816840
distributions of the incident duration and starting Landex, A. (2008). Methods to estimate railway capacity
time. and passenger delays. Technical University of Den-
mark (DTU). Retrieved from http://findit.dtu.dk/en/
catalog/2185768953
REFERENCES Lindfeldt, A. (2015). Railway capacity analysis - Meth-
ods for simulation and evaluation of timetables, delays
Cerreto, F. (2015). Micro-simulation based analysis of and infrastructure. KTH Royal Institute of Technol-
railway lines robustness. In 6th International Confer- ogy. Retrieved from https://www.kth.se/polopoly_
ence on Railway Operations Modelling and Analysis fs/1.613049!/15_002PHD_report.pdf
(pp. 164-1–164-13). Tokyo: International Associa- Mattsson, L.G. (2007). Railway Capacity and Train Delay
tion of Railway Operations Research. Retrieved from Relationships. Critical Infrastructure: Advances in
http://orbit.dtu.dk/fedora/objects/orbit:140777/ Spaital Science. Berlin, Heidelberg: Springer Berlin Hei-
datastreams/file_112408001/content delberg. http://doi.org/10.1007/978-3-540-68056-7_7
Cerreto, F. (2016). A Cubic Function Model for Railway Meester, L. E., & Muns, S. (2007). Stochastic delay
Line Delay. In E. Spessa, M. Biscotto, & G. Smyrnakis propagation in railway networks and phase-type
(Eds.), TRAVISIONS 2016 Contest book (pp. 68–69). distributions. Transportation Research Part B: Meth-
Warsaw. Retrieved from http://orbit.dtu.dk/en/activi- odological, 41(2), 218–230. http://doi.org/10.1016/j.
ties/travisions-competition-2016(4d4a77a0-0541- trb.2006.02.007
4fb2-8548-e2ad7bd8269b).html Meng, L., & Zhou, X. (2011). Robust single-track train
Gorman, M. F. (2009). Statistical estimation of railroad dispatching model under a dynamic and stochastic
congestion delay. Transportation Research Part E: environment: A scenario-based rolling horizon solu-
Logistics and Transportation Review, 45(3), 446–456. tion approach. Transportation Research Part B: Meth-
http://doi.org/10.1016/j.tre.2008.08.004 odological, 45(7), 1080–1102. http://doi.org/10.1016/j.
Goverde, R. M. P. (2010). A delay propagation algorithm trb.2011.05.001
for large-scale railway traffic networks. Transportation Nash, A., & Huerlimann, D. (2004). Railroad simulation
Research Part C: Emerging Technologies, 18(3), 269– using OpenTrack. Computers in Railways IX, 45–54.
287. http://doi.org/10.1016/j.trc.2010.01.002 article.
Harker, P. T., & Hong, S. (1990). Two Moments Estima- Pellegrini, P., Marlière, G., & Rodriguez, J. (2014). Opti-
tion of the Delay on a Partially Double-Track Rail mal train routing and scheduling for managing traffic
Line with Scheduled Traffic. Journal of the Transpor- perturbations in complex junctions. Transportation
tation Research Forum, 31(1). Research Part B: Methodological, 59, 58–80. http://
Hasegawa, Y., Konya, H., & Shinohara, S. (1981). Mac- doi.org/10.1016/j.trb.2013.10.013
ro-Model on Propagation-Disappearance Process of Schranil, S., & Weidmann, U. (2013). Forecasting the
Train Delays. Railway Technical Research Institute, Duration of Rail Operation Disturbances. In TRB
Quarterly Reports, 22(2), 78–82. Retrieved from http:// 92nd Annual Meeting Compendium of Papers (pp.
trid.trb.org/view.aspx?id = 180725 1–20). Washington, D.C.: Transportation Research
Higgins, A., Kozan, E., & Ferreira, L. (1995). Modelling Board of the National Academies. Retrieved from
delay risks associated with train schedules. Transpor- https://trid.trb.org/view.aspx?id = 1241000
tation Planning and Technology, 19(2), 89–108. http:// UIC. (2004). Leaflet 406 - Capacity, (June).
doi.org/10.1080/03081069508717561 Zilko, A. A., Kurowicka, D., & Goverde, R. M. P. (2016).
Kroon, L. G., Dekker, R., & Vromans, M. J. C. M. (2007). Modeling railway disruption lengths with Copula
Cyclic railway timetabling: A stochastic optimization Bayesian Networks. Transportation Research Part
approach. In F. Geraets, L. Kroon, A. Schoebel, D. C: Emerging Technologies, 68, 350–368. http://doi.
Wagner, & C. D. Zaroliagis (Eds.), Lecture Notes org/10.1016/j.trc.2016.04.018

874

TIS2017_Book.indb 874 2/8/2017 8:10:37 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Smart cyber systems incorporating human-in-the-loop towards

ergonomic and sustainable transport systems

F. Galatioto
Transport Systems Catapult, Milton Keynes, UK

C. Salvadori
New Generation Sensors, Pisa, Italy

M. Petracca
Scuola Superiore Sant’Anna, Pisa, Italy

J. Garcia Herrero
University Carlos III de Madrid, Madrid, Spain

M.J. Santofimia
University Castilla—La Mancia, Ciudad Real, Spain

A. Pollini
BSD, Milano, Italy

ABSTRACT: This paper aims to introduce and explain a new framework that using standardised engi-
neering principles and methods for Human-in-the-Loop (HiL) when human interactions and feedback are
considered in smart Cyber-Physical Systems (CPS). This comes from an industrial need to better understand
and predict human behaviour when interacting with CPS, which so far has been mainly addressed for and
from ad-hoc solutions. The proposed framework objectives (fully described in the paper) will be achieved
integrating in an adaptive Internet of Things system existing and new low-cost technologies, both on street
and in-vehicle, with personal experience data coming from psychological, behavioural and physiological
processing. This new approach is expected to enormously enhance the effectiveness of the proposed plat-
form enabling a new focus on emotional states (anxiety, stress) and collective/shared feelings (eg. panic) that
go substantially beyond the state of the art in urban management delivering new capabilities to the ITS
sector.

1 INTRODUCTION communities. In particular, the issues and linkages

between traffic system status perception and deci-
1.1 Context sion making processes, at individual up to collective
level, has not been properly investigated within the
The 2025 European City transport scenario is
Intelligent Transport Systems (ITS) sector.
expected to be heavily affected by huge flows of per-
vasive information generated by people and vehi-
cles, at all scales from local to global. The multitude 1.2 Background
and heterogeneity of participants, each with their
own cultural models, needs and expectations must The so-called “Internet of Things” has been defined
be considered as a key component in the EU City as the network of physical objects or devices, inde-
transport innovation process to address the long- pendently of their nature (vehicle, entire buildings or
standing problem of poor air quality, congestion in other object), embedded with electronics, software,
urban areas, and achieve greenhouse gas emission sensors and network connectivity that enables these
target reductions. Deployment of smart-city sus- objects to collect and exchange data on the web.
tainable systems aspire to provide the right informa- The main idea behind the IoT concept is to have
tion to optimise “behaviours” in real towns. Most of worldwide interconnected objects (sensors), each
the effort so far has been devoted to technological one individually discovered and addressed as a
solutions, with less focus on their use by people and resource in the network.

875

TIS2017_Book.indb 875 2/8/2017 8:10:39 PM

 As sensors spread in the environment, and the In particular, the human today, through the use
capabilities of the so-called wireless sensor net- of smart and mobile technologies can be consid-
works (the considered building blocks of Cyber- ered as another sensor/actuator which brings new
Physical Systems, a.k.a. CPS) are enhanced, several opportunities, as a result of the new capability to
communication protocols must be considered. The gather real-time context and human behaviour
first standard protocol, specifying both the Physi- information. Human beings have therefore become
cal (PHY) and Medium Access Control (MAC) an inherent element of the CPS, in what has been
sub-layers of the ISO/OSI communication model, referred to as HiLCPS.
is the IEEE 802.15.4 (IEEE, 2003). The standard Future HiLCPS not only consider humans as
has been released in its first version in 2003 with the operators affecting the loop control, but more
aim of enabling energy-efficient communications importantly, as the source of the observation-
in Low-Rate Wireless Personal Area Networks analysis-adaption loop (Nunes, 2015).
(LR-WPANs). Kushalnagar et al. (2007) proposed
the adaptation of the IPv6 over Low-Power Wire-
1.3 Research gaps
less Personal Area Networks (6LoWPANs), thus
specifying a Network (NET) layer for Internet-like Transport innovation cannot happen without an
communication in IEEE 802.15.4-based networks. extensive sharing of information among all involved
The 6LoWPAN concept comes from the idea that stakeholders. Ensuring interoperable transport
“the Internet Protocol could and should be applied data and services, where possible based on existing
even to the smallest devices” (Mulligan, 2007), standards and technology, is one of the main com-
and that low-power devices with limited process- ponents of the proposed conceptual platform called
ing capabilities should be able to participate in the SMART H-LOOP. Currently, multi-modal infor-
envisioned IoT. mation services across Europe lack interoperabil-
The Routing Protocol for Low-power and Lossy ity and are fragmented in terms of what they offer
networks (RPL) (Winter, 2012) is the state-of-the- including modal and geographical coverage, real-
art routing algorithm developed by the network- time information and quality levels. The project will
ing community to enable routing capabilities in investigate how to reduce the costs and exploit the
6LoWPAN networks. benefits of sharing information in a context where
However, RPL is a “best-effort” routing protocol different players do not always have the same goals.
and does not satisfy the real-time requirements of Thus, SMART H-LOOP is expected to contrib-
several distributed applications for networks of smart ute towards EU-wide continuity and harmonised
cameras (e.g., tracking and action recognition). delivery of multimodal travel information services.
Recent routing protocols permit (He et al. 2003) This in turn is expected to encourage a positive
and (Bocchino, 2011) are characterised by highly modal shift to sustainable modes of transport and
efficient and scalable solutions able to tackle the therefore improve the efficiency of Europe’s trans-
needs and requirements of next generation sensor port network management.
networks in which cameras are going to be part of SMART H-LOOP will specifically aim to go
the IoT scenario. Other protocol solutions such beyond current “informational” MultiModal Travel
as the Constrained Application Protocol (CoAP) Information and Planning services (MMTIPs)
(Shelby, 2013), are advanced standards working at which allow travellers to plan their journey, com-
the Application (APP) layer to guarantee a seam- paring different travel options and combining
less integration of IoT smart objects in the Inter- different variations of transport modes. Even the
net world. In fact, CoAP is designed to enable the most modern MMTIPs includes a combination of
RESTful architecture in the constrained environ- several transport modes only (e.g. air, rail, water-
ment, thus providing an easy stateless mapping borne, coach, public transport, demand responsive
with HTTP in order to reach a full Machine-2- transport, walking and cycling), while SMART
Machine (M2M) interaction. H-LOOP will integrate also structural user profile
This paper presents an ambitious concept which data with information coming from their cultural
is the development of a new framework to embed models, emotional states, risk and social cohesion
more generic and standardised engineering princi- perception. A possible solution to achieve that will
ples and methods for Human-in-the-Loop (HiL) be through an ubiquitous and adaptive CPS based
when human interactions and feedback are con- on integrated ecosystem of innovative technologies
sidered in smart CPS, generating what has been (IoT, Common Sense Reasoning and Data Fusion,
defined as Human-in-the-Loop CPS (HiLCPS). Big Data analysis) capable of solving different prob-
This comes from an industrial need to better lems in different environments where the human-
understand and predict human behaviour when machine interaction is essential. This will be used
interacting with CPS, which so far has been mainly as sources of information for real-time transport
addressed for and from ad-hoc solutions, and the decision-making, even in contingency plans, as well
need for more robust and transferable approaches. as for medium to long-term trip planning.

876

TIS2017_Book.indb 876 2/8/2017 8:10:39 PM

 In this scenario, the SMART H-LOOP sustain- by receiving events from other cars, e.g., warn-
able city-user MMTIPs will allow the traveller to ing messages or simple position notifications, can
receive personalised routing results according not decide to notify other cars to slow down or change
only to their specific travel preferences or needs, direction. The messages collected from multiple
including the fastest or cheapest route, the fewest base stations can be used in a urban mobility con-
number of connections, the most environmentally trol centre, performing the T domain, for statistic
friendly or the most accessible option for persons related to safety and traffic flow, or traffic manage-
with reduced mobility, but also their decision will ment strategies in particular areas of the city. In
be supported taking into account emotional states this three layer-based architecture, a fast and local
(e.g. anxiety and stress) or feelings (vulnerability or reaction to events can be performed, e.g., inside the
exposure to risk). car, in a road covered by an ITS base station, thus
distributing the logic of the system from the root,
the urban mobility control centre, to the leaves,
1.4 Vision
smart sensors deployed in roads and vehicles.
The vision for the proposed SMART H-LOOP
platform is disruptive in the Internet paradigm.
In fact, it aims to completely distribute the intel- 1.5 Main aims
ligence at every level of the CPS chain, expanding Through the implementation of the SMART
the concept of smart everything and centred on the H-LOOP platform the global aims are to provide
human factor. In fact, the idea is to propose a sys- a holistic framework for Human-in-the-Loop
tem capable of understanding and reacting to dif- Cyber-Physical System (HiLCPS) that addresses
ferent type of events. Thus, the approach consists the systematization of the design, implementation,
of three different points of view: and quality assurance processes as well as the for-
1. The “dT point of view”, capable of understand- malization of model and programming interfaces.
ing isolated events occurred at a certain time This framework is aimed to be devised and tested
instant and the consequent immediate and con- through the implementation of use cases where
tingent reaction. Time is considered as a discrete smart dynamic reconfiguration of services and
variable. service composition are essential capabilities. For
2. The “ΔT point of view”, capable of understand- example, extreme event management in smart cities
ing both the cause and effects of events and their and safety in smart transportations may be two use
impact in undergoing situations. Situations are cases that could be develop and tested in the future.
considered to take place in a time interval. This The idea behind these scenarios is to prompt the sys-
perspective deals with managing situations in an tem to deal with unexpected situations and involve
accurate and specialised manner. human beings on the basis of the system’s available
3. The “T domain”, in which time is considered means (services and devices) and according to a sys-
in a broader way, capable of understanding temic view on real world including the community,
trends, attitudes, and behaviours. This per- the groups and the personal experience.
spective aims at maintaining the system in a The proposed approach will offer the possibility
continuous upgrade, capable of automatically to establish an engineering process for the construc-
reconfigure its models to the contingent reality tion of instructed knowledge-based system to be
and environments. demonstrated in hot areas such as smart cities and
smart transportation, where the interactions between
The above reported points of view can be human beings and machines are very strong.
mapped in a three-layer based system. In the low-
est layer, the dT point of view, decisions are taken
in a short time by using very few and local data. 1.6 Scientific and technological objectives
Systems in this layer have a local visibility of a
particular event. The layer above performs a ΔT The following five scientific and technological
point of view, and takes decisions considering objectives have been identified:
multiple events in a wide, but still local, area. In 1. To investigate and define a methodology to
the last layer, the T domain, a global view of the implement a systematic design, implementation,
system is maintained. For instance, by considering and quality assurance of HiLCPS. This meth-
the ITS use case in which vehicles are connected odology will be enhanced with a toolchain, pat-
to the infrastructure, the dT point of view can be tern designs, and reusable human behavioural
realized by smart sensors installed inside the car models to ensure a cost-effective and delimited-
taking local decisions, e.g., a pedestrian is crossing time-to-market construction process. The
the road outside the crosswalk, then the car stop methodology will be developed according to
and such event is sent towards an ITS base station an innovative and foundational interdiscipli-
covering a specific area. The same base station, nary approach trying to achieve an individual

877

TIS2017_Book.indb 877 2/8/2017 8:10:39 PM

 psychological comfort optimisation balanced the adaptation capability of the system is going to
with network optimisation, following a cog- depend on its understanding of what it is happen-
nitive and emotional ergonomics conceptual ing. Also, quality metrics will be used to validate
framework integrated with methods from oper- the output of the reasoning framework in terms
ations management and modelling. of correct alignment and interpretation of diverse
2. To contribute to standardisation efforts on sensor inputs and context events, such as detection
HiLCPS designs and protocols (networks and of humans, traffic elements, track continuity, etc.
Human Machine Interfaces).
3. To develop an advanced high level understanding
platform, here referred to as Distributed Rea- 2 THE PROPOSED PLATFORM
soning Platform (DRP). This platform enhances
with intelligence the server and gateway layer, 2.1 The three-layered architecture
by combining data fusion and common-sense The proposed SMART H-LOOP is a three-layered
reasoning techniques to model and semanti- architecture where key elements are individually
cally enrich IoT devices, their sensing values, addressed, as depicted in Figure 1.
and their programming interfaces. This is the At the top level, the sensing and actuators layer
core of the “ergonomic reaction”, since it has is characterized by its heterogeneity and dynamism.
the knowledge of both the real world event and Sensor and actuator devices need to be seamlessly
human models. integrated as knowledge producer or consumer.
4. To define a modular self-adaptive, self-recon- SMART H-LOOP will allow bringing reliable, effec-
figurable architecture compliant with already tive, and personalised information to the end users
established cooperation (e.g., FI-WARE). This through the implementation of web services that can
architecture, guided by the DRP, will articulate be easily accessed from personal mobile devices as
the process how humans devise clever uses of well as embedded actuators in the real context of
things they have at hand to overcome the short- use. Particularly, an extensive search will be carried
age of a certain item (e.g., switch your cell phone out to look for: (i) the most appropriate sensors to
during a blackout to illuminate). This architec- represent as real as possible the traffic environment,
ture along with the reasoning framework and the and (ii) the most effective actuators for showing a
big data processing platform comprise the core warming signal, from mobile phones, to street pan-
of the cognitive framework enabling a smart els, following the cognitive ergonomics guidelines.
context. Quantitative aspects such as easiness At the intermediate layer, based on an intelligent
to integrate new devices in the system and align gateway (re-named as “ser-way” for its improved
with other sources will be evaluated. processing and understanding capabilities), the
5. To build an open-source integrated set of indi- Distributed Reasoning Platform or DRP is capa-
vidual, social and collective human behaviour ble of understanding the evolution of monitored
models in the framework of the socio-technical events and immediately reacting to them (see sec-
systems. When studying complex real-life phe- tion 2.3). The DRP is intended to identify, model,
nomena, they need to be represented as net- and collect the domain specific knowledge involved
works of activity systems. Human actions will in the considered context, including that related to
be described following embodied cognition human behaviour, and combine it in real-time with
theory (Wilson, 2014) (Bagnara, 2015). The
achievement of this objective will be evaluated
over the generated models, their accuracy, com-
pleteness, consistency, clarity, and adaptability.

This platform will play a double role. Firstly,

it will accept general requests and feedback from
humans in an almost-natural language, such as
“illuminate a certain physical space”, which will
be translated into specific services activation. Sec-
ondly, context events and sensed magnitudes will
be modelled and combined with domain specific
and common-sense knowledge in a fully distrib-
uted architecture. This combination, along with
big data analysis (spatio-temporal pattern analysis
and activity recognition), will support the under-
standing of the ongoing situation(s), the elabora-
tion of behavioural models, data visualisation, etc. Figure 1. The proposed SMART H-LOOP framework
The target will be jointly considered with O2, since layers architecture.

878

TIS2017_Book.indb 878 2/8/2017 8:10:39 PM

 context events. Additionally, transversal knowledge, layers. This supports the distributed processing
also known as common-sense knowledge, will be and reaction implementation, thus holding the
provided to support the understanding of context “dT point of view”.
events. This knowledge is mainly devoted to describ- 2. The dynamic re-configurability and re-program-
ing and analysing the understanding tasks required mability of sensors and actuators, thus export-
as a primary need for both the decision making proc- ing parameters, constants, and processing code
ess and the direct ergonomic reaction exploiting the as CoAP resources (Alessandrelli, 2013).
CPS actuators.
In this scenario, the development and deploy-
The upper layer, Big Data Analysis and Model-
ment of smart and connected actuators is necessary
ling layer, (see section 2.4) will enable the analysis
to support the Situation Awareness (SA) (USDoT,
and modelling of mental states or propositional
2003) and the cognitive and emotional ergonomics
attitudes of the system (beliefs, intentions, desires,
guideline described in the following subsection.
hopes, etc.) exploiting the big data analysis on the
To better support SA applications, new class of
data gathered from the sensors. Eventually, the sys-
embedded visual sensors (e.g. smart cameras), can
tem behaviour should be driven by these mental
be used for advanced applications. Smart cam-
states since the functional requirements of HiLCPS
eras organized as IoT smart objects can be used to
cannot extensively captured, in advanced, as it hap-
execute advanced low-complexity computer vision
pens with traditional software systems. Thus, Big
algorithms able to extract mobility related data (Sal-
Data Analysis and Modelling layer is thought to
vadori 2012, 2014), to send images or video stream as
deeply understand behaviours, trends and attitudes
a function of a detected event (Salvadori, 2013; Mag-
typical of a contingent scenario and to feedback
giani, 2015), and more in general to extract advanced
and adapt the models residents in the ser-way.
information from the real world (Figure 2).
An example of how these layers will jointly
work will be the following. Given traffic light sen-
sors and induction loops, both geo-positioned sen- 2.3 The “Ser-Way” a Distributed Reasoning
sors, it could be possible to detect cars jumping Platform (DRP) layer
a traffic light (at the sensing and actuator layer)
This is the core of the proposed architecture and
by combining the knowledge of the traffic light in
what mainly distinguishes SMART H-LOOP
red and the induction loop detecting a car passing
from the other state-of-the-art proposals. In IoT
by. However, it might be the case in which one car
systems, gateways usually perform a basic packet
has jumped all traffic lights, in a row, of a certain
forwarding operation toward both, the nodes and/
avenue. To get to that conclusion and to achieve
or the servers.
to that understanding, the intermediate layer
This layer is mainly intended to support the
(exploiting the DRP installed inside the ser-way)
modelling, recognition, and understanding of on-
will model the jumping-light in a spatio-temporal
going activities from in the interpretation of sen-
manner. A causal explanation for those events can
sor information and then, dictate based on the
therefore be a car breaking the speed limit (dis-
inference of that understanding with the SA and
tance-time between two traffic lights) and breaking
cognitive ergonomics guidelines. In this sense, two
traffic rules (hypothesis). From the big data analy-
dimensions can be identified for the DRP: the
sis, these events match the behavioural pattern of a
understanding and acting. The understanding proc-
reckless high-speed driver. This hypothesis can be
ess is conceived here as a seek for causal explanation
used to update the ser-way resident models improv-
ing the HiLCPS toward the nearby pedestrian acci-
dent prevention by such dangerous behaviours.

2.2 Smart sensors and actuators layer

SMART H-LOOP will adopt and extend the Inter-
net of Things paradigm, by enabling the complete
seamless interoperability among different networks
(e.g., IEEE802.15.4/g, LoRa, etc.). Interoperabil-
ity support is provided through the realisation of
gateways (re-named as “ser-way” for its improved
processing and understanding capabilities) and
sensors nodes that embed one or more communi-
cation interfaces. Additionally, the exploitation of
the RESTful paradigm will enable:
1. Sharing a common semantic model, not only Figure 2. Image processing of a camera tracking
among devices at this layer but also with upper pedestrians.

879

TIS2017_Book.indb 879 2/8/2017 8:10:40 PM

 to sensed information. According to Woodaward and system versatility. This analytical capability will
(2003) a causal explanation is “any explanation that be driven by a planning scheme that given a desired
proceeds by showing how an outcome depends (where result (illuminate a room) it is going to look for a
the dependence in question is not logical or concep- reconfiguration or composition setup through the
tual) on other variables or factors counts as causal”. whole space of available resources. To enable such a
For example, an explanation for a big crash noise setup, a common semantic model should be shared
and a sudden stop of the traffic flow can be the among the different layers of the architecture.
occurrence of a car accident. Explanations should We plan to use Scone1 to model and hold explicit
be considered as preliminary hypothesis since there and implicit knowledge. Scone is also particularly
might be many different explanations catering for well suited to manage possible worlds, through its
the same sensorial information, and it cannot be implementation of the multiple-contexts mecha-
obviated that events are neither always clear nor pre- nism (Santofimia, 2011).
cise due to sensor malfunctioning or low precision. On the functionalities themselves, the “brain”
For that reason, a first stage will be intended to of the server, that is, the different algorithms, data
associate sensor measures (as captured effects of analytics methods and techniques to be imple-
events) to actions that might have caused them. mented in order to process the huge expected
amount of data collected and infer valuable infor-
2.4 Big data server layer mation regarding the identification and estimation
It is an important research activity to provide of the evolution and impact of event and non-
a high level understanding of both, the event recurring incidents, accident prediction models and
causes, evolution and impact (e.g., degree of the relations between the citizen behaviour and all
risk of the different sections of the road) and the possible influencing factors.
the human behavioural patterns that character- Specifically, long term data analysis will allow
ize the response of the citizens in such situations. planners and managers to locate or reconfigure
Here, we hypothesize the role of citizen sensing in sensors and/or actuators especially for accident-
extracting relevant event and situational informa- prone “hotspots” or risk level assignation of a
tion, therefore proposing innovative algorithms certain section of the road. On the other hand,
based on the semantically enriched data gathered our solution goes beyond current approaches, pro-
by DRP, historical registries, and user informa- viding methods as predictive analytics aimed at
tion (e.g., crowdsourced, smartphone, social discovering behavioural patterns and correlations
networks). and aggregation techniques, useful to evaluate the
SMART H-LOOP will integrate, in an advanced goodness of the DRP configuration and quantita-
computing environment, different massively parallel- tively guiding its parametrization.
processing technologies which can acquire, store
and elaborate a relevant and heterogeneous quan-
tity of mobility, events, behavioural and contextual 3 SPECIFICATIONS
data. These data will draw from asynchronous
user-generated information (e.g. mobility pat- 3.1 The FI-WARE compliance
terns), synchronous user-generated data (e.g. posi- FI-WARE constitutes an investment to promote
tion in real time, crowdsourced information such the innovation of ICT sector thanks to European
incidents or traffic status), and historic databases technologies; the architecture is comprised of a set
(e.g. weather reports and trends). of software pieces, all of them open source, known
as Enablers, tackling different aspects of ICT
2.5 Innovation components technologies such as big data, IoT, security, cloud
computing, etc. FI-WARE proposes a set of open
We consider that the innovation at the three levels
standards defined through APIs which offer differ-
identified resides in the following components.
ent technological capacities for developers. From a
Perception of important elements is fundamen-
practical point of view, it is possible to define it as a
tal for forming a correct picture of the situation.
set of open source software tools based on existing
Understanding is beyond simple perception; this
solutions and standards which have been designed
level promotes an accurate current understanding
to facilitate the creation of innovative ICT tools
of the environment through the integration of mul-
and services by developers.
tiple pieces of information and a determination of
Particularly, the functionalities defined in
their relevance to the person’s goals. Finally, pro-
SMART H-LOOP have the ambition to be raised
jection refers to the ability to foresee accurately
with the aim of creating a new model of smart city.
future situation events and their dynamics.
SMART H-LOOP will implement a semantic
composition approach based on the deep analysis
1
capabilities, held at this level, about device, service, http://www.cs.cmu.edu/sef/scone/

880

TIS2017_Book.indb 880 2/8/2017 8:10:40 PM

 Additionally, the key market application pro-
posed by SMART H-LOOP consists in a com-
prehensive framework for smart environments,
provided as a FI-WARE enabler, in which services
and devices can be automatically reconfigured and
combined, if required, under special circumstances
such as extreme event management, or device
unavailability. The project will integrate a full CPS
system from the sensor and actuators layer up to
the server level in to demonstrate the feasibility of
the approach and validate the methodologies and
tools developed during the project. One ambition is
to experiment the designed architecture at Technol-
Figure 3. Transport System Catapult Virtual Reality ogy Readiness Level (TRL) 5 through collaborative
Environment. opportunities and funding streams such as H2020.
SMART H-LOOP plans to advance state-of-the-
art software engineering by providing principles,
3.2 The Virtual Reality environment methods, and tools that support the software life-
SMART H-LOOP will offer a unique opportunity cycle for systems such as HiLCPS that are mainly
to study user reaction, interaction, and behaviour. characterized by working under open world assump-
It is intention of the authors to develop further tions. This implies that all the interactions cannot be
the concept using and testing scenarios initially by predefined nor prescribed, but the system is thought
implementation in a Virtual Reality (VR) experi- to be adaptive to different cultures and environ-
mental environment. Specifically, the use of VR will ments, exploiting the knowledge generated from Big
be across different use cases. However, before the Data analysis. Thus, the proposed methodology will
implementation in a VR environment it is expected aim at supporting the complex SMART H-LOOP
that an extensive use of simulations through which socio-technical systems integration and adaption in
dynamic and iterative shifts from stepwise real- a seamless way. The modelling and design process
world implementations and VR experiments will of such systems is co-evolutionary since hardware
be implemented. Such methodology will allow design (e.g., modularity, scalability), software design
SMART H-LOOP to achieve an individual psycho- (e.g., programmability, real-time computing) and
logical comfort optimisation balanced with HiLCPS Interaction design (e.g., products and services) are
optimisation following the cognitive and emotional carried out in parallel (divergence phase) and then
ergonomics conceptual framework integrated with integrated (convergence phase) in form of outputs.
methods from crisis management and smart trans- By developing a newly emerging approach
port operations management and modelling. of embodied cognition to HiLCPS, SMART
Moreover, the VR suite owned by the Transport H-LOOP recognises that there is no clear-cut sepa-
Systems Catapult (UK) has been recently enhanced ration between body and cognition; and between
by integrating the existing Omnideck platform with body in action into the environment, and between
a driving and a bike simulator. This will provide, the human being as a whole and the technologies
for the first time, the opportunity to test simultane- she is going to adopt to transform its cognition and
ous and synchronised simulation for the different action. Such an action-perception loop resembles
VR suite users that will be able to interact to each the humans-CPS circularity expanding further the
other like in the real-world environment. notion of integrated and complex socio-technical
VR users can be equipped with smart sensors, systems as a whole.
including eye tracking, body sensors, etc., repro- SMART H-LOOP approach to ergonomics
ducing the different scenarios identified as the use deals with this level of complexity, as one of its
cases for SMART H-LOOP. hallmarks is focusing on interactions among com-
ponents, and on avoiding considering elements in
4 DISCUSSION isolation. Such analytic and design approach is
systemic by nature; its focus being the study of sys-
The proposed platform originates from the needs of tems (as opposed to individual elements), context,
industrial stakeholders with common problems in and complex interactions (in order not to isolate
relation to the lack of methods and tools supporting elements), holism, and emergence (in order to cap-
the process of building HiLCPS, highly distributed ture the various levels of explanation of one phe-
and connected digital technologies. This is because nomena) (Wilson, 2014; Bagnara, 2015).
those companies seek common solutions through From the point of view of the physical environ-
the implementation of predictive engineering meth- ment, the proposed methodology plan to support the
ods and tools for reducing costs and time to market. topological infrastructure modelling, deployment,

881

TIS2017_Book.indb 881 2/8/2017 8:10:40 PM

 and test. The challenge here is address the process tions for Low-Rate Wireless Personal Area Networks
of quality assurance, when interactions cannot be (LR-WPAN). The Institute of Electrical and Electron-
predefined nor prescribed. Approaches like Test- ics Engineers.
Driven Development (Beck, 2002) are based on Kushalnagar N., Montenegro G., & Schumacher C., 2007.
IPv6 over Low-Power Wireless Personal Area Net-
writing small pieces of code specifically design to works (6LoWPANs): Overview, Assumptions, Prob-
pass a test. Tests are intended to cover all system’s lem Statement, and Goals, RFC 4919.
functionalities. In this sense, this methodology that LoRa. https://www.lora-alliance.org/What-Is-LoRa/
has gain credit over the last years, would fail to Technology
test the system reaction to unforeseen situations. Maggiani, C., Salvadori, C., Petracca, M., Madeo, S.,
Preparing a test for it means that the situation has Bocchino, S., Pagano, 2013 “Video Streaming Applica-
already been foreseen. The novelty here consists in tions in Wireless Camera Networks: a change detection
considering a mixed real-world and Virtual Reality based approach targeted to 6LoWPAN” in Journal of
(VR) experimental methodology that will consider Systems Architecture, vol. 59, no. 10, pp. 859–869,
May 2013.
the real SMART H-LOOP use context and the vir- Mulligan, G. 2007. “The 6lowpan architecture”. Proc. 4th
tual world as a continuum. IEEE Workshop on Embedded Networked Sensors
Finally, the ambition behind the proposed (EmNets), pp. 78–82, June.
SMART H-LOOP platform is to develop a method- Nunes, D.S., Zhang, P., & Sá Silva, J. 2015. “A Survey on
ology and tools enabling the design and the optimi- Human-in-the-Loop Applications Towards an Internet
zation of CPS in their integrity reinforcing Europe’s of All,” in IEEE Communications Surveys & Tutorials,
know-how on a diversity of connected subject. The vol. 17, no. 2, pp. 944–965.
CPS in smart cities is envisaged as one of the case Salvadori, C, Petracca, M., Ghibaudi, M., Pagano, P. 2012
studies. The whole architecture will be based on IoT “On-board Image Processing in Wireless Multimedia
Sensor Networks: a Parking Space Monitoring Solu-
for the sensors and actuators which lead to the use tion for Intelligent Transportation Systems” in book
of distributed intelligence and finally to analytics Intelligent Sensor Networks: Across Sensing, Sig-
(BIG DATA) to cope with the data to be analysed. nal Processing, and Machine Learning”, CRC Press,
This data will be elaborated throughout the differ- December 2012, pp. 245–266.
ent layers of the CPS architecture. The aim is to be Salvadori, C., Petracca, M., Bocchino, S., Pelliccia, R.,
able to understand past events, monitor the present Pagano, P. 2014 “A Low-cost Vehicle Counter for
situation and predict future behaviours and ways of Next Generation ITS”, in Journal of Real-Time Image
improving user lives. In order to involve individuals Processing, pp. 1–17, March 2014.
with these CPS, the SMART H-LOOP platform, Salvadori, C., Petracca, M., Madeo, S., Bocchino, S.,
Pagano, P. 2013 “Video Streaming Applications in
through the use cognitive ergonomics will put the Wireless Camera Networks: a change detection based
user at the centre of the CPS. approach targeted to 6LoWPAN”, in Journal of Systems
Architecture, vol. 59, no. 10, pp. 859–869, May 2013.
Santofimia, M.J., Fahlman, S.E., Moya, F., & Lopez, J.C.
REFERENCES 2011. Possible-world and multiplecontext semantics
for common-sense action planning. In Ambient Intel-
Alessandrelli, D., Petracca, M. Pagano, P. 2013. ‘T-RES: ligence and Smart Environments Ebook, Volume 14.
Enabling Reconfigurable in-network Processing in Shelby, Z., Hartke, K., Bormann, C., & Frank, B. 2013.
IoTbased WSNs’, Distributed Computing Sensor Sys- “Constrained Application Protocol (CoAP),” IETF
tems (DCOSS), 2013 IEE International Conference. Draft Version 18.
Bagnara, S, Pozzi, S. 2015. Embodied Cognition and USDoT, United States Department of Transportation.
Ergonomics. Journal of Ergonomics. 2003. Bicycle and pedestrian detection, Minnesota
Beck. 2002. Test Driven Development: By Example. Add- Department of Transportation-Office of Traffic Engi-
ison-Wesley Longman Publishing Co., Inc., Boston, neering/ITS Section, SRF Consulting Group, Inc.
MA, USA. (2003).
Bocchino, S. et al. 2011. “SPEED routing protocol Wilson, J.R. 2014. Applied Ergonomics, Volume 45, Issue
in 6 LoWPAN networks”. 16th IEEE Conference 1, January 2014, Pages 5–13.
on Emerging Technologies & Factory Automation Winter, T., Thubert, P., Brandt, A., Hui, J., Kelsey, R.,
(ETFA), pp.1–9, Sept. 2011. Levis, P., Pister, K., Struik, R., Vasseur, J.P. & Alex-
He, H. et al. 2003. “SPEED: A Stateless Protocol for Real- ander, R. 2012. RPL: IPv6 Routing Protocol for Low-
Time Communicationin Sensor Networks”. Proceed- Power and Lossy Networks. RFC 6550.
ings of the International Conference on Distributed Woodward, J. 2003. “Making Things Happen: A Theory
Computing Systems, pp. 56–55, May 2003. of Causal Explanation”. Oxford University Press
IEEE, 2003. Computer Society, Wireless Medium Access (2003).
Control (MAC) and Physical Layer (PHY) specifica-

882

TIS2017_Book.indb 882 2/8/2017 8:10:41 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Development, analysis and prediction of efficiency of bulk carriers

in the world market

S. Galić & T. Stanivuk

Faculty of Maritime Studies, Split, Croatia

A. Marušić
Herzegovina University, Mostar, Bosna and Herzegovina

ABSTRACT: Today, a significant part of the world trade is performed by sea, of which a large share
refers to bulk cargo. Bulk cargo shipping industry makes an essential part of the international shipping
with ocean-going vessels that are the most effective and sometimes the only way to transport large quanti-
ties of bulk commodities. Transportation of bulk cargo by sea can be defined as transport of homogene-
ous cargo with vessels following irregular schedules of sailing. This paper provides the statistical analysis
of bulk carrier vessels and further forecast of supply and demand of bulk cargo by type. The analysis also
includes the key factors that encourage or discourage the growth of bulk cargo industry. One of them is
the fulfilment of the new prescribed rigorous regulations on reduced discharge of greenhouse gases from
these types of vessels, because the shipping industry is currently the only industrial sector which is already
covered by a legally binding global agreement on the reduction of CO2 emissions through technical and
operational measures adopted by the International Maritime Organization (IMO).

1 THE FLEET OF THE SHIPS In the category of large and very large ships,
WORLDWIDE AND THE SEABORNE bulk carriers (30.7%), oil and chemical tankers
TRADE (29.5%) and container ships (22.9%) make about
85% of the fleet.
At present, the global number of merchant ships In terms of Gross Tonnage (GT), the large and
exceeds the figure of 80,000. Research results indi- very large size ships represent 80% of the fleet
cate that in the period of 12 months the world’s (Table 2), with oil and chemical tankers, bulk carri-
merchant fleet increased by 3.5% until 1 January ers and container ships dominating both categories
2015. That is the lowest annual growth rate in more at 86.1% (large) and 83.1% (very large).
than a decade (UNCTAD, 2015). At the beginning The average age of the world fleet increased slightly
of 2015, the world’s commercial fleet consisted of according to data from 2014. Taking into considera-
89,464 vessels with a total tonnage of 1.75 billion tion the delivery of several new buildings combined
DWT. The largest share in the global fleet is dry with reduced scrapping activity, higher tonnage can-
bulk that reached a share of 43.5% of the total not compensate for the natural cycle of aging fleets.
seaborne shipping capacity at the beginning of The world economy embarked on a slow-moving
2015. The result is an increase of 4.4% between recovery led by uneven growth in developed econo-
2014 and 2015 and even greater expansion from mies and a slowdown in developing countries and
2010 to 2013. Bulk carriers are making one third economies in transition. In 2014, the world’s Gross
of the world fleet, and are therefore the most com- Domestic Product (GDP) increased by 2.5% com-
mon type of ships (Table 1). According to research pared to 2.4% in 2013. Meanwhile, the world trade
results (Equasis, 2014) the total percentage of ves- increased by 2.3%, which was below the pre-crisis
sels by type and size are: levels. According to the analyses and statistics,
the global maritime shipments increased by 3.4%
• Small and medium sized ships: tugs (19.6%),
in 2014, which is the same growth rate as in 2013.
general cargo ships (19.1%), oil and chemical
This situation may be the result of several factors:
tankers (14.5%) and bulk carriers (12.9%) are
the most common ship types by number, making • a slowdown in development of major emerging
about two thirds of the global fleet. economies;
• Large and very large ships: bulk carriers (30.7%), • the price level of oil and the development of new
oil and chemical tankers (29.5%) and container refinery capacity; and
ships (22.9%) amount to about 85% of the fleet. • uneven recovery in developed economies.

883

TIS2017_Book.indb 883 2/8/2017 8:10:43 PM

 Table 1. Total number of ships in the world by size and type.

Small Medium Large Very Large

Ship type (GT < 500) (500GT < 25000) (25000GT < 60000) (GT ≥ 60000) Total

General Cargo Ships 4356 13.90% 11650 30.90% 212 1.90% 16,218 19.10%
Specialized Cargo Ships 8 0.00% 201 0.50% 56 0.50% 2 0.00% 267 0.30%
Container Ships 17 0.10% 2255 6.00% 1619 14.80% 1193 22.90% 5084 6.00%
Ro-Ro Cargo Ships 30 0.10% 653 1.70% 619 5.70% 180 3.50% 1482 1.70%
Bulk Carriers 320 1.00% 3700 9.80% 5374 49.20% 1602 30.70% 10,996 12.90%
Oil and Chemical Tankers 1815 5.80% 6597 17.50% 2414 22.10% 1537 29.50% 12,363 14.50%
Gas Tankers 39 0.10% 1070 2.80% 216 2.00% 378 7.30% 1703 2.00%
Other Tankers 315 1.00% 531 1.40% 5 0.00% 851 1.00%
Passenger Ships 3657 11.70% 2528 6.70% 271 2.50% 156 3.00% 6612 7.80%
Offshore vessels 2531 8.10% 5227 13.90% 115 1.10% 157 3.00% 8030 9.40%
Service Ships 2405 7.70% 2361 6.30% 23 0.20% 6 0.10% 4795 5.60%
Tugs 15,747 50.40% 946 2.50% 16,693 19.60%
TOTAL 31,240 100% 37,719 100% 10,924 100% 5211 100% 85,094 100%

*Source: Equasis statistics—world merchant fleet in 2014.

Table 2. Total number of ships in the world according to gross tonnage.

Small Medium Large Very Large

Ship type (GT < 500) (500GT < 25000) (25000GT < 60000) (GT ≥ 60000) Total

General Cargo Ships 1455 17.60% 49,812 22.80% 6942 1.70% 58,209 5.00%
Specialized Cargo Ships 2 0.00% 1596 0.70% 2107 0.50% 153 0.00% 3858 0.30%
Container Ships 7 0.10% 26,425 12.10% 62,925 15.20% 116,771 22.20% 206,128 17.70%
Ro-Ro Cargo Ships 11 0.10% 6306 2.90% 29320 7.10% 11696 2.20% 47,333 4.10%
Bulk Carriers 125 1.50% 54,518 25.00% 19,8021 47.90% 157,251 29.90% 409,915 35.10%
Oil and Chemical Tankers 586 7.10% 39,595 18.10% 88,677 21.50% 168,038 31.90% 296,896 25.50%
Gas Tankers 15 0.20% 6349 2.90% 9477 2.30% 40,813 7.80% 56,654 4.90%
Other Tankers 94 1.10% 1350 0.60% 162 0.00% 1606 0.10%
Passenger Ships 922 11.10% 10,579 4.80% 9649 2.30% 15,397 2.90% 3,6547 3.10%
Offshore vessels 719 8.70% 13,334 6.10% 5258 1.30% 15,374 2.90% 34,685 3.00%
Service Ships 595 7.20% 7510 3.40% 850 0.20% 992 0.20% 9947 0.90%
Tugs 3750 45.30% 931 0.40% 4681 0.40%
TOTAL 8281 100% 218,305 100% 413,388 100% 526,485 100% 1,166,459 100%

* Source: Equasis statistics—world merchant fleet in 2014.

Table 3 Ships world fleet (%)—according to the size of

the ships and gross tonnage.
Table 4. Seaborne trade worldwide in
Total number of ship’s world fleet by number 2014.

Very large 6% Crude oil 17%

Large 13% Container cargo 15%
Medium 44% Minor bulks 15%
Small 37% Iron ore 13%
Coal 12%
Total number of ship’s world fleet by gross tonnage
Other dry cargo 9%
Very large 45% Petroleum products 9%
Large 35% Gas and Chemicals 6%
Medium 19% Grain 4%
Small 1%
* Source: United Nations Conference on
* Source: Equasis statistics—world merchant fleet in Trade and Development—UNCTAD:
2014. Review of maritime transport 2015.

884

TIS2017_Book.indb 884 2/8/2017 8:10:43 PM

 It is expected that the growth of world GDP and According to the statistics from 2014, the world
seaborne trade will continue to grow at a moder- merchant fleet grew up to 42 million GT. This was
ate pace over the next few years. However, odds the result of the new construction of 64 million GT
remain uncertain and are subject to many risks, and 22 million GT ships that went for the scrap.
including the continuation of moderate growth in However more than 91% of GT delivered
global demand and trade, the fragile recovery in in 2014 was built in just three countries: China
Europe, geopolitical tensions and potential faster (35.9%), South Korea (34.4%) and Japan (21.0%),
slowdown of the developing economies. where China delivered dry bulk carriers followed
China achieved tremendous growth in the world by container ships and tankers, while Japan deliv-
seaborne trade in 2009. The challenge for shipping ered mostly bulk carriers (UNCTAD, 2015).
is to ensure that the dynamics of trade continues
to be generated through the expansion of China
and to replicate elsewhere. Therefore, the greater 3 BULK CARRIER FREIGHT RATES AND
integration of China is the key driver in global MARITIME TRANSPORT
production.
The Capesize market has a cyclical nature. Such
market circulates from depression into expansion,
2 DRY CARGO TRADE depending on a number of economic and political
variables. This kind of market is very complex and
The demand for imports in developing coun- reacts easily to various variables. It is not easy to
tries, particularly China and India, remains the predict a shipping cycle due to irregularities. Even
key driver of bulk carrier cargo growth in 2016. the slightest fluctuation in freight rates can have a
During 2014, the increase in world seaborne dry major implication on profits. The ability to accu-
bulk shipments was estimated at 5%, and that rately predict the Capesize market has potential
was lower than the previous four years (Dry Bulk advantage of allowing ship owners to better plan
Trade Outlook, 2015). The growth was driven by how to maximize profits or minimize losses or risks
strong expansion in iron ore trade (12.4%), which incurred on the shipping market. It is not possi-
accounted for about 30% of the entire bulk cargo ble to know how many years the market will be in
and reached 1.34 billion tons. In contrast, the expansion or recession. The cycles can last for sev-
transport of coal was estimated to be increased by eral years. However, every cycle obviously depends
2.8%, which was significantly slower than in 2012 on the law of supply and demand. If demand is
(UNCTAD, 2015). strong and supply is low, then the market will be in
Exports of dry bulk cargo such as bauxite, nickel expansion, but if the situation is reversed, then the
ore, iron ore and coal were constrained by, among market will be in depression.
other factors, restrictions on exports, weather pat- According to the data of Baltic and Interna-
terns, bans on mining activities, regulatory meas- tional Maritime Council—BIMCO from 2016, on
ures and policies seeking to promote national 10 February, the Baltic Dry Index (BDI) reached
producers and industries. 290. At that moment, one bulk carrier, regardless
of its size, age and fuel-efficient qualities earned a
time charter average between 2417 to 2776 USD
per day.
After that period, in the next few months the
earnings for Capesize vessels were very small, but
in August and September they achieved greater
profits. Despite the fact that the earnings of Cape-
size vessels doubled in the next two months, these
gains remained below the OPEX level for the larg-
est part of the fleet.
During 2015, China exported 112 million
tons of crude steel, which led to a decrease in
the price of scrap steel. In January and February
2016, the export of steel from China decreased by
1.6% (17.85 million tons). In March 2016 exports
Figure 1. World tonnage order for propelled seagoing
amounted to 10 million tons (BIMCO, 2016).
merchant vessels of 100 GT and above (from 2000 to 2015). According to data of Worldsteel Association,
* Source: United Nations Conference on Trade and global crude steel production for January and
Development UNCTAD based on data supplied by February 2016 was 5.6% lower than in the same
Clarckson Research. period of 2015, but at that time the production of

885

TIS2017_Book.indb 885 2/8/2017 8:10:43 PM

 Figure 3. The supply growth for 2016–2018 contains
Figure 2. Baltic Exchange time charter averages USD/ the existing orders only and is estimated under the
day (from 2015 to 2016). assumptions that the scheduled deliveries would falls
*Source: BIMCO, Baltic Exchange, Clarksons Research. short by 10% due to various reasons and that 50% of
the remaining ordered vessels are delayed/postponed.
* Source: BIMCO.

crude steel in China decreased by 6.5%. There are

three key main elements to watch out for in 2016, in the freight market. By mid-March, there were
including the amount of the dry bulk tonnage to only four new contracts signed, three in Japan and
be demolished. one in China. During March and April, the long-
Over the first two months in 2016, imports of awaited orders for 30 VLOCs with a capacity of
iron ore to China increased by 6.4% to 155.8 mil- 400,000 DWT were confirmed. In March 2016,
lion tons compared to the same period in 2015. On China COSCO Shipping Corporation and Vale
the other hand, imports of coal in China decreased signed a 27-year-long agreement that will allow the
by 10% to 28.8 million tons compared to the same Chinese corporation to carry 16 million tonnes of
period in 2015. iron ore per year for the Brazilian multinational
In March 2016, imports of Chinese iron ore corporation (BIMCO, 2016).
amounted to 85.8 million tons, while imports It is estimated that each VLOC can handle
of coal increased to 19.7 million tonnes. Strong 1.6 million tons of iron ore from Tubarão (Brazil)
imports of coal in March equalized the decline of to Baoshan (China) per year. This new series of
coal in January and February. VLOC carriers will open the market of 48 million
In the first quarter of 2016, import of coal tons of iron ore, and that is rather bad news for the
was smaller by 1.2% on an annual basis (BIMCO international owners and operators. The delivery
2016). of 30 new VLOC ships has been ordered for the
period 2018–2019 (BIMCO, 2016). In 2015, Brazil
exported 191.6 million tons of iron ore to China;
3.1 Efficiency and opportunities for bulk carrier
the existing and the new valemax ships (VLOC)
fleet on maritime market
will be able to carry more than half of Brazil’s cur-
Over the first three months of 2016, the number rent annual exports of iron ore.
of demolished dry bulk shipping capacity reached The export of grains and soybeans volume
a very high number, but regardless of that, the from Argentina and Brazil is expected to increase
overall total dry bulk fleet continued to increase. in 2016. According to research results (BIMCO
Specifically the 16.7 million DWT of new capac- 2016), there will be no massive increase in freight
ity entered the fleet, and 14 million DWT went to for handysize, supramax and panamax ships,
scrap. because these ships are already waiting around the
However, not all dry bulk sub-segments grew main loading areas. Another positive indicator for
evenly. For example, the Capesize fleet, which has China is the rise of domestic steel prices from mid-
doubled in the last 6.5 years, was reduced to 7 ves- February 2016. In mid-April 2016, iron ore rose
sels as well as the capacity which decreased by 0.2% from 41 USD per MT to 56 USD per MT on the
in the first quarter of 2016. international market, which is also positive indica-
According to research results (BIMCO 2016), tor. However, the concern for the sustainability of
a growth of the fleet by 1.1% or 10 million DWT freight rates in the years to come remains, because
is expected in 2016. Shareholders and investors there is a fear that the activity of destroying ships
work hard to postpone the delivery of new ships could slow down if the BDI starts to grow.

886

TIS2017_Book.indb 886 2/8/2017 8:10:44 PM

 If ship owners significantly slow down the Moreover, a similar rate of demolition and
destruction of vessels, the fleet of ships will inevi- delivery in the next year will further help to deflate
tably increase and this will lead to a fundamental the oversupply.
imbalance, because the market analyses predicted The growth of demand, which did not seem to
that the demand would grow slowly in the com- be likely to occur during the last year and early
ing years. In order to change the trends in fleet part of 2016, has helped the market dry bulk cargo
enlargement, a trend of negative growth in the fleet that is slowly recovering.
on multi-year basis should be developed. It is esti- It is expected that in 2016, the domestic produc-
mated that the current utilization rate of dry bulk tion of coal in India will continue to be high.
fleet approximates the state as it existed at the end If La Nina forms it would bring above average
of the 1970s (BIMCO, 2016). winter spring rainfall over central and Western
What can be changed regarding the import/ Australia which may affect the transport of cargo
export are the coal imports into India, because and disrupt rail traffic and port activities. Such an
India is planning to increase domestic production event could result in increased slowdown at ports
of coal and to completely stop importing coal in the in the country, which will be reflected in the avail-
next two to three years. India imported 176 million able supply of coal from other sources.
tons of thermal coal according to data from 2014. Freight rates for iron ore routes to Brazil-
In 2015, this country imported 171 million tons China and Australia-China for Capesize vessels
of thermal coal. According to the analyses (Simp- will strengthen in the next two quarters, but the
son Spence Young’s Dry Cargo brokers—SSY), it return of laid-up ships to transport cargo could
is expected that India will import about 170 mil- disrupt the expected progress according to analy-
lion thermal coal in 2016. It can also be expected ses (Drewry Dry Bulk Forecaster research) for
that the domestic production of coal in India will August 2016.
continue to be high (Drewry Maritime Research Demand for coal-carrying vessels will increase
2016). in the coming quarters because Japan has decided
It seems that the ship owners will seek to recover to increase its coal-fired power generation, while
their costs by reducing the proportion of the fleet China plans to cut domestic coal production which
so the market may return to a profitable situation will increase import demand. Also, China plans to
in 2017 (Drewry Dry Bulk Forecaster research). reduce domestic production of coal, which will
However, the recovery of market to such an extent, increase the demand. It is expected that the grain
where the ship owners will start earning profits, trade will remain strong as recent heavy rains have
shall remain elusive for another year. enhanced the prospects for grains and Europe, CIS
Key drivers for the shipping fleet will remain the and North America, especially for maize and bar-
steel industry and imports of iron ore and steam ley crops.
coal to China. In addition to iron ore, import of
steam coal to China will ensure future employment
of the dry bulk tonnage.
4 ENVIRONMENTAL CHALLENGES FOR
Increase in trade and contracting of supply will
BULK CARRIERS AND OTHER SHIPS
support the recovery of charter rates on major dry
bulk shipping routes. There is a possibility that
4.1 Implementation of the existing technologies
China will import more coal and iron ore (Drewry
Dry Bulk Forecaster research). Macroeconomic According to the 3rd International Maritime
policy decisions of the Chinese authorities had Organization—IMO greenhouse gas GHG study,
an impact on the overall dry bulk market increas- Maritime transport emits about 1,000 million tons
ing the demand for ships on major routes, which of CO2 per year, and is responsible for about 2.5%
resulted in the recovery of charter rates. China’s of global greenhouse gas emissions. The latter is
effort to revive its economy has helped to revive expected to reach around 5% by 2050, and this is
the trade of iron ore, so the further increase in ton/ not in accordance with the internationally agreed
mile demand could be expected where most of the objective of maintaining global temperature rise
cargo is coming from Brazil. below 2°C.
Stricter conditions of financing options and The international agreements require that global
restraint in new orders have helped the dry bulk emissions of gases into the atmosphere should be
market to keep supply under control so far. A high at least halved by 2050. Although seaborne ship-
number of slippages and cancellations will impede ping, in most cases, is more fuel-efficient than other
any substantial growth in deliveries over the next types of transports, greenhouse gas emissions from
few years. Demolition of old vessels is expected to ships grow more every day and, unless adequate
result in a slowdown in growth and the fleet size measures are taken, greenhouse gas emissions
in 2016. from ships will increase more than double by 2050.

887

TIS2017_Book.indb 887 2/8/2017 8:10:45 PM

 This will happen because of the expected growth type of the vessel. Studies show that there is poten-
of the world economy and the associated demand tial for reducing fuel consumption in the shipping
for maritime transport. sector up to 55%. This result is excellent compared
Nowadays in the seaborne shipping there is a to other forms of transport and economic sec-
growing pressure in reducing greenhouse gas emis- tors. Further reduction of emissions from ships
sions and the industry asks for bulk carrier manu- can be achieved by introducing new innovative
facturers to respond. With this in mind, statistics technologies.
from 2013 predict that the new building bulk car- Eco-ships have been considered as a solution
riers will reduce the average carbon dioxide emis- to the dry bulk market for some time. However, as
sions by 40% by 2040 (Fig. 4). the research results show, not all stakeholders in
According to 2nd IMO GHG Study, ship energy the maritime shipping industry share the view that
consumption and CO2 emissions could be reduced the eco-ships are worth paying attention to, and
up to 75% by applying operational measures and that they may affect the general situation. The ben-
implementing the existing technologies (Fig. 5). efits accruing from eco-efficient vessels on freight
There is a great potential in reducing emissions rates seem to be marginal with 43% of the respond-
from ships through fuel-saving techniques provided ents saying these ships have no significant impact
by modern technologies (Table 5). This could be at the moment. This can partly be attributed to the
achieved with relatively little cost. The implemen- current low bunker prices that limit the fuel savings
tation of such techniques and measures would also that modern tonnage gives (Platts, 2015).
significantly reduce ship running costs. However,
not all of these solutions can be applied for every

Figure 5. Existing technologies that can be imple-

Figure 4. Statistic of estimated CO2 emission reduction mented in the low emission concept—study for the 8,500
rate for new builds of bulk carriers between 2013 and 2050. TEU container carrier and the 35,000 DWT bulk carrier.
*Source: http://www.statista.com/statistics/216058/world *Source: http://www.greenship.org/lowemissionconcept
wide-bulk-carrier-co2-emission-reduction-rate/. study/lowemissionbulkcarrierstudy/technologiesused/.

Table 5. The most effective existing technical and operational measures to reduce CO2 emissions from ships.

RELATIVE CO2 SAVINGS/COSTS TAKE-UP

SOLUTION SAVINGS PER TONNE CO2 2007–2011

Speed reduction 17–34% −280€/t 0% 50%

Propeller and rudder upgrade 3–4% −150€/t 0% 0%
Hull coating 2–5% −280€/t 0% 50%
Waste heat recovery 2–6% +60€/t 0% 0%
Optimisation of trim and ballast 1–3% −200€/t 0% 50%
Propeller polishing 1–3% −280€/t 75% 75%
Hull cleaning 1–5% −200€/t 75% 75%
Main engine tuning 1–3% −250€/t 75% 75%
Autopilot upgrade 1–1.5% −280€/t 75% 75%
Weather routing 1–4% −280€/t 75% 75%

*Source: http://ec.europa.eu/clima/policies/transport/shipping/index_en.htm.

888

TIS2017_Book.indb 888 2/8/2017 8:10:45 PM

 4.2 Current situation
Significant efforts to reduce emissions from
ships have appeared through the IMO and the
United Nations Framework Convention on Cli-
mate Change (UNFCCC) over the recent years.
After years of international discussions, the
IMO’s Marine Environment Protection Commit-
tee set some important milestones including the
adoption of the Energy Efficiency Design Index
(EEDI) and the Ship Energy Efficiency Man-
agement Plan (SEEMP). Unfortunately, these
Figure 6. The statistic of current and estimated future
transport efficiency improvement associated with bulk measures alone cannot lead to an absolute reduc-
carriers between 2005 and 2050, by level of CO2 used per tion of harmful emissions from the ships. How-
tonne/mile. ever, international discussions have yet to bring
*Source: http://www.statista.com/statistics/216033/tran agreement on global market-based measures that
sport-efficiency-improvement-associated-with-bulk- would cut emissions from the sector as a whole,
carriers/. including the existing ships.

Table 6. SWOT analysis of supply and demand for bulk carriers and environmental challenges in bulk cargo shipping
industry.

STRENGTH WEAKNESSES POSSIBILITIES THREATS

The ability to accurately The Capesize market has The ability to accurately predict It is not easy to predict a
predict the market has a cyclical nature. Such the Capesize market has shipping cycle because
potential advantage of market circulates from potential advantage of it is irregular. Even the
allowing ship owners to depression into expansion, allowing ship owners to slightest fluctuation in
better plan how to maximize depending on a number better plan how to maximize freight rates can have a
profits or minimize losses or of economic and political profits or minimize losses or major implication on
risks incurred in the variables. risks incurred on the shipping profits.
shipping market. market.
It is expected that the world The recovery of the market Increase in trade and supply If ship owners significantly
GDP and seaborne trade to the point that ship will support the recovery of slow down the
will continue to grow at a owners will start earning charter rates on major dry destruction of ships,
moderate pace over the profits will remain elusive bulk shipping routes. the fleet will increase and
next few years. for some time. this may lead to a
fundamental imbalance.
Today, seaborne shipping According to the 3rd IMO According to 2nd IMO GHG Greenhouse gas emissions
is, in most cases, more GHG study, Maritime Study, ship energy consumption from ships are getting
fuel-efficient than other transport emits about and CO2 emissions could be larger and, unless
types of transport. 1,000 million tons of CO2 reduced by up to 75% through appropriate measures
per year, and is responsible the application of operational are taken, greenhouse gas
for about 2.5% of global measures and implementation emissions from ships may
greenhouse gas emissions. of the existing technologies. increase more than double
by 2050.
There is a great potential in Not all of fuel-saving Studies show that there is The benefits accruing from
reducing emissions from techniques can be potential for reducing fuel eco-efficient vessels on
ships through fuel-saving applied for every type consumption in the seaborne freight rates seem to be
techniques ensured by of the vessel. shipping sector up to 55%. marginal with 43% of
modern technologies. respondents who say
that these ships have no
significant impact at the
moment. This can partly
be attributed to the current
low bunker prices that
limit the fuel savings that
modern tonnage gives.
New ideas on technical and According to the 3rd IMO Reliable information on the Requirements imply that
operational measures that GHG study, maritime effectiveness of technologies global greenhouse gas
are based on a gradual transport emits about to improve energy efficiency emissions should be at
approach to monitoring, 1,000 million tons of CO2 reduces the financial risk to least halved by 2050.
reporting and verification per year, and is responsible investors.
of emissions from ships. for about 2.5% of global
greenhouse gas emissions.

* Source: author.

889

TIS2017_Book.indb 889 2/8/2017 8:10:45 PM

 New ideas on technical and operational meas- applying the saving fuel techniques through the
ures have been proposed by the European Com- existing technologies. However, such solutions
mission and the IMO recently. The measures are cannot be applied to all types of vessels, but stud-
based on a gradual approach to Monitoring, ies show that there is potential for reducing fuel
Reporting and Verification of Emissions (MRV) consumption in the maritime shipping sector up
as the first step of further measures of efficiency to 55%, while further reduction of emissions from
for the existing ships. ships can be achieved through implementation of
The MRV regulation was adopted in 2015. It new innovative technologies.
creates a EU-wide legal framework for monitor- The stricter environmental regulations as well
ing, reporting and verification of CO2 emissions as low freight and charter rates should encourage
from maritime transport. The regulation requires further demolition of older ships. Also, these eco-
all ships over 5000GT, calling at EU ports from 1 nomic and regulatory initiatives would contribute
January 2018, to collect and eventually publish veri- to environmental protection and help to reduce
fied annual data on CO2 emissions and other rele- overcapacity of the maritime shipping market.
vant information. Statistics show that the transport Reliable information on the effectiveness of
efficiency measures will reduce emissions by seven technologies to improve energy efficiency reduces
grams of CO2 per tonne/mile in 2030 (Figure 6). the financial risk to investors. However, shipping
companies are likely to take measures to improve
the energy efficiency of their vessels if they are
5 CONCLUSIONS based on accurate and reliable information, as such
information on the effectiveness of technologies to
Every year the number of merchant ships in the improve energy efficiency reduces the financial risk
world increases, and the largest share of the glo- for investors.
bal fleet belongs to dry bulk tonnage. Taking into
consideration the delivery of several new buildings REFERENCES
combined with reduced scrapping activity, higher
tonnage cannot compensate for the natural cycle Ančić, I. 2015—Influence of the required EEDI reduction
of aging fleets. It is expected that the growth of factor on the CO2 emission from bulk carriers, Energy
world GDP and seaborne trade will continue to Policy, article.
grow at a moderate pace over the next few years. Baltic and International Maritime Council—BIMCO
However, odds remain uncertain and are subject https://www.bimco.org/
to many risks, including the continuation of mod- Clarckson Research https://sin.clarksons.net/
erate growth of global demand and merchandise Drewry Dry Bulk Forecaster research 2016 http://
trade, the fragile recovery in Europe, geopolitical www.drewry.co.uk/publications/view_publication.
php?id=314
tensions and the potential deceleration of econo- Equasis and European Maritime Safety Agency EMSA
mies in developing countries. Statistics. 2014—The world merchant fleet in 2014,
Rebalancing of the Chinese economy can sig- study.
nificantly reshape maritime transport, changing European Commission http://ec.europa.eu/clima/polic
the overall image of shipping and maritime trade. ies/transport/shipping/documentation_en.htm
Import of steam coal to China will be the future Green Ship of the Future http://www.greenship.org
challenge for dry bulk transport. The reduction of International Maritime Organization IMO. 2009 — Sec-
coal import to India can also affect the dry bulk ond IMO GHG Study Executive Summary and Final
transport, because India is planning to increase Report, London.
International Maritime Organization IMO. 2014 —
domestic production of coal and entirely cease Third IMO GHG Study Executive Summary and Final
importing coal in the next two to three years. The Report, London.
key drivers for the bulk fleet will remain the steel International Maritime Organization IMO http://www.
industry and imports of iron ore and steam coal imo.org/en/MediaCentre/PressBriefings/Pages/01-2016
to China. -MTCC-.aspx
These days there is a growing pressure to reduce Kyong, H. 2013 — Forecasting the capesize freight market,
greenhouse gas emissions from ships and the sea- World Maritime University, Sweden, dissertation.
borne shipping industry is asking manufacturers Platts-McGraw Hill financial. 2015: Special Report Dry
of bulk carriers to respond to the situation. With Bulk Market Survey 2015, study.
Simpson Spence Young’s Dry Cargo brokers — SSY
that in mind, eco-ships have been considered as https://www.ssyonline.com/services/dry-cargo.
a solution to the dry bulk market for quite some Statista http://www.statista.com
time. Analyses predict that future new buildings United Nations Conference on Trade and Development
will be able to reduce carbon dioxide emissions by UNCTAD: Review of maritime transport 2015, United
40% by 2040. Today, there is a great potential in Nations publication, ISSN 0566-7682.
reducing carbon dioxide emissions from ships by Worldsteel Association https://www.worldsteel.org/

890

TIS2017_Book.indb 890 2/8/2017 8:10:46 PM

 Transport Infrastructure and Systems – Dell’Acqua & Wegman (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-03009-1

Enhanced financial analysis to evaluate mass transit proposals in terms

of contribution on resilience increasing of urban systems

A. Spinosa
Cityrailways Engineering, Rome, Italy

ABSTRACT: The prevailing model of development of urban areas is increasingly oriented to a territorial
isotropy, notwithstanding the evidences of a considerable literature. This is due to the absence of mathemati-
cal models able to translate this awareness into appropriate financial models. Nowadays a city no longer exists
as mere physical built place, but as a dematerialized “cloud” of flows: people, goods, information. The vicious
circle is all here: the car has allowed the urban sprawl; mass transit does not allow the widespread use of a city.
This is the time of urban sustainability. If each urban element (home-work-facilities) is framed dynamically
within the daily circadian rhythm that connects them, it will be possible to develop a sustainable solution
that is not partialized on a specific sector. A balanced city produces social wealth: that turns into a financial
one because it can be measured. The paper presents a methodology that combines the most modern epide-
miological knowledge and the energy aspects of mobility. The Enhanced Financial Analysis (EFA) approach
considers the Public authority as investor and manager of a planned infrastructure. Revenues come from fare
and savings on current health expenditure. The model explores the morbidity of some peculiar diseases of
which there is an identified cause-and-effect pollution (direct and indirect) and air quality. The avoided health
expenditure is measured by health services that will not be paid out in the medium and long term compared
with the do-nothing scenario. EFA allows introjecting some aspects that usually are relegated to the economic
analysis, but are part of the current spending of the funder. This realizes a comprehensive financial analysis
of the possible alternative technologies for a mass transit corridor, correcting the bias that may arise on the
electrical transport—even on the rail transport—compared with apparently less expensive choices.

1 INTRODUCTION productivity over time). Summarizing in a word, it

means increasing the resilience of a territory.
The objective of social and environmental sustain- Although this is the information age, it does not
ability requires to look at the costs of the processes mean that the displacements are decreasing. The cit-
that animate the city every day from a broader ies are being renovated but the per capita movement
perspective. If until now we have focused on the of people and goods will increase by 30% at 2050
sustainability of housing on one side and on the (The World Bank). This means that the mass trans-
reduction of transport emissions on the other, it port will be essential for the sustainability of cities.
is time to approach the theme of mobility from a
joint perspective.
It is the time of urban sustainability. A city is 2 MOTIVE OF THE PAPER
a set of buildings but also a complex network of
relationships in which homes, offices and trans- This is the time of urban sustainability. If each
port are only the most visible part: 60% of CO2 urban element (home-work-facilities) is framed
emissions is not in the car that produces it, but it is dynamically within the daily circadian rhythm
dematerialized in the choices that are upstream of that connects them, it will be possible to develop
the decision to take that car. a sustainable solution that is not partialized on
If there is a territorial sustainability there is also a specific sector. A balanced city produces social
an effective indicator of risk in case of unsustain- wealth: that turns into a financial one because it
ability? The answer is in the affirmative. Territo- can be measured. The paper presents a methodol-
rial sustainability is not only a virtuous exercise ogy that combines the most modern epidemiologi-
but an investment that a territory does in resources cal knowledge and the energy aspects of mobility
(human and environmental), productivity (how to support administrators and designers in the
these resources become or generate usable wells) choice of the most effective technology for a mass
and stability (preserving these resources and their transport project.

891

TIS2017_Book.indb 891 2/8/2017 8:10:48 PM

 3 METHODS with total and cause-specific mortality. This led to
outdoor air pollution standards for PM10 in the
For each technology, the operation and mainte- European Union for daily and yearly averaging
nance costs are derived from the figures reported by times. PM10 was seen as the most health relevant
122 European operators for the period 2004–2014. PM size fraction and is therefore measured on a
Unit prices are derived from the actual selling of regular basis in specified locations in all EU coun-
orders as of the value communicated by the con- tries. Evidence is increasing, mainly based on stud-
tracting authorities. ies from the US, that the smaller PM2.5 fraction is
The figures related to energy costs are homog- more consistently associated with health outcomes,
enized assuming as reference the Italian value and that this fraction is more specific for anthropo-
(=100). genic combustion-related emissions. EU registra-
Similarly, the figures related to labor and main- tion therefore decided to put, following regulation
tenance costs are homogenized taking as reference in the US, a PM2.5 standard value in force starting
the Italian value (=100). The reference indices for 2015 with an anticipated limit value of 25 μg/m3
each country are provided by Eurostat. (yearly average). Contrary to the large amount of
The health care cost is assessed in terms of US studies on health effects of PM2.5, European
health services provided (i.e. avoided) on the useful studies on PM2.5 are sparse and studies in the lit-
life of the project. erature on mortality effects of PM2.5 could only be
All evaluations refer to the maximum daily identified for 10 European cities (Anderson et al.,
demand i.e. the average weekday, identified as the 2001, Atkinson et al., 2010, Branis et al., 2010,
day in which an urban transit system set the maxi- Garrett and Casimiro, 2011, Halonen et al., 2009,
mum trip demand. Weekdays equivalent per year Mate et al., 2010, Mallone et al., 2011, Meister et al.,
(Istat) are derived from the ratio between the meas- 2012, Ostro et al., 2011 and Stolzel et al., 2007).
ured average displacements: In addition to effects of PM2.5, there is increas-
The daily average weekday is obtained by the ing evidence that coarse particles (PM2.5–10) may
annual demand dividing by 290 and vice versa. play a role in generating adverse health effects
To estimate the load in a particular day (e.g. (Brunekreef and Forsberg, 2005 and Sandstrom
summer, holiday or weekend) is necessary to apply et al., 2005). In a national, multi-city time-series
the reduction factors shown in the second column study of the acute effects of PM2.5 and PM2.5–10 in
of Table 1 (Istat*). the US, Zanobetti and Schwartz (2009) found that
both PM2.5 and PM2.5–10 were significantly asso-
ciated with increased mortality for all and specific
4 SHORT-TERM EXPOSURE TO AMBIENT causes. Less evidence of effects of PM2.5 and
FINE PARTICULATE MATTER PM2.5–10 is available from European studies. Sev-
eral European studies failed to demonstrate signifi-
A large number of epidemiological studies have cant effects of PM2.5 on all cause or cause specific
shown that short-term exposure to ambient Par- mortality (Branis et al., 2010, Garrett and Casimiro,
ticulate Matter (PM) in outdoor air is associated 2011, Halonen et al., 2009, Mallone et al., 2011 and
Stolzel et al., 2007). However, these studies gener-
Table 1. Number of equivalent weekdays (day in which ally comprised relatively small populations (≤1 mil-
an urban transit system set the maximum trip demand) lion) (Garrett and Casimiro, 2011, Halonen et al.,
per year. 2009 and Stolzel et al., 2007) or short study periods
(1 year) (Branis et al., 2010). In a recent study that
Weekday
included 6 years of data on particulate matter air
Type of day Value Number equivalent
pollution and mortality in London, UK, PM2.5
Weekday 100% 203 203 was not statistically significantly associated with
Saturday 62% 48 30 all cause or cardiovascular mortality, and the effect
Sunday 48% 48 23 on respiratory mortality was only statistically sig-
Holyday 32% 7 2 nificant for one of the seven lags studied (Atkinson
July weekday 68% 22 15 et al., 2010). Even less information is available for
July w-e 33% 8 3 PM2.5–10, especially for North-West Europe.
August weekday 56% 22 12 PM10 and PM2.5 levels were statistically signifi-
August w-e 27% 7 2 cantly (p < 0.05) associated with all cause and cause-
365 290 specific deaths. For example, a 10 μg/m3 increase in
previous day PM was associated with 0.8% (95%
CI 0.3–1.2) excess risk in all-cause mortality for
PM2.5 and a 0.6% (CI 0.2–1.0) excess risk for
* Italian National Statistics Center. PM10. No appreciable associations were observed

892

TIS2017_Book.indb 892 2/8/2017 8:10:48 PM

 for PM2.5–10. Effects of PM10, and PM2.5 were 5 SHORT-TERM EXPOSURE TO
insensitive to adjustment for PM2.5–10, and vice- AMBIENT NOISE
versa. PM10 and PM2.5 were too highly correlated
to disentangle their independent effects. Noise is pervasive in everyday life and can cause
Li MH et al. (2016) indicated that short-term both auditory and non-auditory health effects.
exposure to a 10 μg/m3 increment of ambient PM2.5 Noise-induced hearing loss remains highly preva-
is associated with increased COPD hospitalizations lent in occupational settings, and is increasingly
and mortality (N.A.H. Janssena et al., 2016). caused by social noise exposure (e.g. through per-
In this analysis, health expenditure is assessed sonal music players). Evidence of the non-auditory
in terms of morbidity and health care costs by effects of environmental noise exposure on public
an increase of the PM10 concentration (in μg/m3, health is growing. Observational and experimen-
yearly average). Vice versa a reduction of PM10 tal studies have shown that noise exposure leads
concentration measures the avoided health care to annoyance, disturbs sleep and causes daytime
expenditure. The health services are calculated on sleepiness, affects patient outcomes and staff per-
the national tariff of health services of the Italian formance in hospitals, increases the occurrence
Ministry of Health (SSN, 2016). of hypertension and cardiovascular disease, and
The assessment of a rapid transit line health impairs cognitive performance in schoolchildren.
effects must be related to local conditions of the In principle, the noise/stress hypothesis is well
project corridor. The proposed methodology understood: Noise activates the pituitaryadrenal-
requires the following characteristic data: cortical axis and the sympathetic-adrenal-medullary
axis. Changes in stress hormones including epine-
− Average daily PM10 concentration (γ in μg/m3)
phrine, norepinephrine and cortisol are frequently
− Traffic average PM10 emission rate (εR)
found in acute and chronic noise experiments. The
− Daily road trip (RT)
catecholamines and steroid hormones affect the
− Daily road mileage (RM)
organism’s metabolism. Cardiovascular disorders
− Public transport rate
are especially in focus for epidemiological stud-
− Larger urban area resident population
ies on adverse noise effects. However, not all bio-
− Population density (inh. /km2)
logically notifiable effects are of clinical relevance.
The Rapid Transit (RT) project corridor has a The relative importance and significance of health
length (km) and a weekday average demand i.e. a outcomes to be assessed in epidemiological noise
weekday average mileage (trip km). studies follow a hierarchical order, i.e. changes in
The new corridor subtracting a share of traffic physiological stress indicators, increase in biological
by the road: risk factors, increase of the prevalence or incidence
ΔRT = Road subtracted trip = RT demand ⋅ (1- of diseases, premature death (Basner et al., 2014).
Public transport rate) Common noise related problems are interfer-
On safe side: ence with communication and sleep disturbance
ΔRM = Road subtracted mileage (trip km) = Road (Griefahn et al., 2000). Decreased quality of sleep
subtracted trip ⋅ corridor length is considered to be a major health outcome of envi-
Finally: ronmental noise (Berglund and Lindvall, 1995).
Noise exposure can also cause other nonauditory
ΔPM10 = Average daily PM10 concentration red. effects such as annoyance, changes of behavior and
deterioration in performance. Long-term effects of
RM
= εR ⋅ ⋅γ road traffic noise on psychosocial health and wellbe-
ΔRM ing are also described (Ohrstrom et al., 1998). Stud-
ies have shown that people living near streets with
busy traffic or airports close their windows, spend
Table 2. Morbidity and health care costs by an increase
of the PM10 concentration (in μg/m3, yearly average). less time in their gardens, and have less visitors than
people living in more quiet areas (Griefahn, 2000).
Annual cost Primary physiological effects of noise expo-
Morbidity each of assistance sure are vegetative reactions such as increases in
Pathology 1.0 μg/m3 (EUR) blood pressure, heart rate and finger pulse ampli-
tude, cardiac arrhythmia and changes in respira-
Adenocarcinoma 6.4E-04 69,400 tion and body movements (Berglund and Lindvall,
Lynphoma 2.7E-05 54,800 1995). Secondary effects are reduction in perceived
Asthma 9.2E-03 810 sleep quality, increased fatigue, decreased mood
Chronic obstructive 8.8E-03 4,440 or wellbeing and deterioration in performance.
pulmonary
Cardiovascular effects of traffic noise such as
Disease (COPD)
hypertension and ischemic heart disease have also

893

TIS2017_Book.indb 893 2/8/2017 8:10:48 PM

 been described but epidemiological evidence is still With regard to electric traction technologies are
limited in this respect (Babisch, 2000). In a recent considered the emissions on the source (the power
study a relationship between aircraft noise expo- plant). In Italy, the national energy mix (GSE,
sure and hypertension has been reported (Rosen- 2014–2015) is 42.3% renewable source; 19.3% car-
lund et al., 2001). bon; 28.9% natural gas; 4.9% nuclear; 4.6%, petro-
Noise policy largely depends on considerations leum products and other sources. Power plants
about cost-effectiveness, which may vary between are located outside urban centers; for this it is
populations. Kim et al., (2012) has shown that the assumed a reference density for the health impact of
effect of traffic can cause annoyance in 11% of the 100 inh. / km2.
population living within 500 meters from a major Health impact is assessed by including both
road; sleep disorders in 2% of the residents. For the fine particulate that nitrogen oxides. Consid-
Bluhm G., Nordling E., Berglind N. (2004), fre- ering the specific effects on health (Medparticles,
quent annoyance was reported by 13% of subjects 2001–2011), the relation is:
exposed to Leq H24 > 50 dB(A) compared to 2%
PM10EQ = 25⋅NOX + PM10 [mg/kWh]
among those exposed to as lower of 50 dB(A),
resulting in a difference of 11% (95% Confidence Imagining a project RT corridor of 10 km in an
Interval 7%, 15%); sometimes or frequently occur- urban area of 100,000 inhabitants (density of at
ring sleep disturbance was reported by 23% at Leq least 1000 inh. /km2) is:
H24 > 50 dB(A) and by 13% at levels <50 dB(A), a In the same example, it is possible to calculate
difference of 11% (95% CI 4%, 18%). the overall cost per energy unit. Purchase cost are
In this analysis, health expenditure is assessed the current ones at 2016 in Italy.
in terms of morbidity and health care costs by an
increase of 1 dB (A) of the average noise level Leq Table 4. Noise emission [in dB(A)] per mode of trans-
H24 (or Lden, day-evening-night). Vice versa a port at normal travel conditions.
reduction Sound Noise Level measures the avoided
health care expenditure. The health services are cal- Average
Specific Average standard
culated on the national tariff of health services of emission speed emission
the Italian Ministry of Health (SSN, 2016). Technology per kW (km/h) per vehicle
The proposed methodology requires the Lden
average value along the RT project corridor: health Car 1.7E+00 30 83
effects are calculated only on the resident popula- Bus 5.8E-01 24 87
tion within a distance of 250 meters from each side Trolleybus 4.2E-01 24 63
from the corridor. Tram (streetcar) 2.6E-01 24 66
The reduction is proportional to the transit Light Rail Transit 2.4E-01 27 72
reduction of private vehicles and buses. Specific Metro 1.9E-01 35 77
noise emissivity is given by the following values Commuter railway 2.0E-01