Pedestrian

microsimulation
modelling

UK good practice guide

First edition 2023
Contents

Executive summary 04 5. Base model validation

5.1 Definition 22
1. Introduction 5.2 Criteria to assess model validation 22

1.1 Purpose of this guide 06

1.2 Project overview 06 6. Model outputs
1.3 Relevant guidelines 08
6.1 Level of service plots 24
1.4 Common issues 08
6.2 Space utilisation plots 26
1.5 Terms 09
6.3 Journey time outputs 27
6.4 Model screenshots 28
2. Model planning 6.5 Model videos
6.6 Generalised journey times and social cost
29
30
2.1 Modelling flowchart 10 6.7 Flow rate and density profiles 31
2.2 Scoping 11
2.3 Data collection 11
2.4 Inputs and assumptions 11 7. Model reporting
2.5 Base year model 11
7.1 Introduction 32
2.6 Scenario testing for rail stations 12
7.2 Inputs and assumptions note 32
2.7 Scenario testing for venues/stadia 14
7.3 Model development report 33
7.4 Modelling report 34
3. Model building requirements
3.1 Overlay 16 8. Model auditing
3.2 Physical characteristics and behaviour 16
8.1 Introduction 36
3.3 Demand data 16
8.2 Internal audits 37
3.4 Other data requirements 17
8.3 External audits 37
3.5 Model build reference documents 17
8.4 Audit report 37

4. Base model calibration

4.1 Definition 18
4.2 Criteria to assess the level of calibration 18
4.3 Address the issues 19
4.4 Recalibrating model 19
4.5 Automatic ticket gate example 19
4 | Pedestrian
Mott MacDonald
microsimulation
| Pedestrian
modelling
Microsimulation
UK goodModelling
practice guide
Guide

Executive summary

This pedestrian microsimulation

modelling UK good practice
guide has been written to assist
with pedestrian model building,
analysing, reporting, auditing,
interpretation and understanding
– focusing in particular on
railway station and venues/stadia
infrastructure design requirements.
Modellers producing increasingly consistent
and accurate deliverables will consequently help
enable project stakeholders to make sufficiently
informed decisions.

Published guideline document references are

provided, as are good practice examples of
what to do and why.

Authors
Stuart Fraser, Mott MacDonald
Tony Kivistik, Mott MacDonald

Peer reviewers
Samya Ghosh, AECOM
Andy Leeson, WSP
Brett Little, Arup
Emmet Ruxton, AECOM

Use by external organisations:

Should this guide be used by an organisation other than

Mott MacDonald, the company (Mott MacDonald) takes no
responsibility for the user’s interpretation and application of
the guidance.
6 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 7

Introduction

Figure 1.1: Microsimulation modelling project Gantt chart example

Task Week
Task 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Inception and
site visit

Determine available data

and further requirements

Scope and
commission survey

Undertake survey

Build basic network

Develop matrices

Assign trips

Calibrate model

1.1 Purpose of this guide 1.2 Project overview Validate model

The purpose of this guide is to provide a foundation Figure 1.1 is a Gantt chart representation of a typical
for pedestrian model building, analysing, reporting, microsimulation consultancy project. The chart shows
auditing, interpretation and understanding. It is hoped that some parallel activity is possible and, in the case of Audit model – internal
that the guide will help raise the quality of the modelling the model calibration, necessary. The timescales shown
process and result in the development of more robust are only indicative, as all models differ in timescale from
and accurate pedestrian microsimulation models. inception to completion. Task durations differ for various Scenario testing
reasons, including size, purpose, complexity, input data
This guide has been written with a particular focus on availability and the number of scenarios to test.
railway station and venues/stadia infrastructure design Reporting
requirements. However, many of the principles could be
applied to other environments.
Present outputs
The guide is not specific to a particular software
package; it is designed to help modellers produce
consistent and accurate deliverables, which will Audit model – external
consequently help enable clients and project
stakeholders to make sufficiently informed decisions.
8 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 9

1.3 Relevant guidelines The guidelines further state that there are numerous 1.5 Terms
In the UK there are a few key industry standard parameters involved in the development of a simulation Table 1.1 provides a brief description of the common terms that are typically referred to throughout the life cycle of
guidelines that are commonly referred and adhered model. It is impossible to assess the technical validity of a pedestrian microsimulation modelling project. The descriptions are similar to those more generically given in The
to when undertaking pedestrian microsimulation a model solely based on its output and visual animation Microsimulation Consultancy Good Practice Guide8 by SIAS Limited.
modelling. These include: of the results. Documentation of the model calibration
and validation process and the rationale for parameter Table 1.1: Descriptions of common terms
• Network Rail, Station Capacity Planning design adjustments are required so that the technical
manual1 robustness of the model can be assessed. Term Description
• Transport for London, G371A Station Planning Project The work undertaken for a client which, in this good practice guide, requires the use of a
Standards and Guidelines2 The Department for Transport in its Principles of pedestrian microsimulation model.
• Transport for London, S1371 Station Capacity Modelling and Forecasting7 guidance states on
Planning3 page 15 that: Study area The physical area to be represented by the model. The study area may be larger than the
• Transport for London, Modelling with Legion: TfL Best area in which the client is interested, to ensure that the needs of all the reporting authorities
Practice Guide4 “The risks of using disproportionate time are met.
• Sports Grounds Safety Authority, Guide to Safety at
Sports Grounds5 and resources can be minimised by Model build The physical representation of the infrastructure and associated features, but not including
assigned trips.
specifying the model scope correctly
Given the abundance of traffic microsimulation guidance from the outset. Models should be Origin – The representation of trips originating in one area and destinating in another area within the
publications available from public and private UK destination (O-D) network. Often referred to as demand matrices.
organisations, some of those considered as good sufficiently sophisticated to represent trip matrices
practice from a generic microsimulation perspective travel movements for the scheme
Assignment The process of applying the demand matrices to the modelled network.
(i.e. regardless of mode) are also specifically referred to effectively, whilst avoiding unnecessary
in this guide.
complexity.” Calibration The iterative process of defining appropriate parameters to ensure that the interactions of
pedestrians are realistic. Any adjustments to the model intended to reduce the differences
1.4 Common issues between the modelled and observed (surveyed data) should be regarded as calibration.
The National Highways Guidelines for the Use of It should be noted that a model is a representation of
Microsimulation Software6 states on page 6 that the what could happen and is not 100% accurate. As such, Validation A measure of the accuracy of the model suitability. Following validation, no further changes
following issues are often encountered when managing microsimulation modelling should be used as one of should be made to the base model.
microsimulation projects: many tools to inform infrastructure, planning and design. Validation report A report produced as a final validation check, after which no further changes are made to
the base model.
• Model extents and parameters not agreed prior to
model commencement.
• Insufficient data for calibrating/validating the model
• Model checks not extensive enough to identify errors.
10 | Pedestrian
Mott MacDonald
microsimulation
| Pedestrian
modelling
Microsimulation
UK goodModelling
practice guide
Guide Pedestrian microsimulation modelling UK good practice guide | 11

2. Model planning
2.1 Modelling flowchart 2.2 Scoping • Input/output/interfaces – consideration of the data
The flowchart shown in Figure 2.1 sets out the typical 2.2.1 Site visit available, evaluation output data produced by the
steps involved when undertaking a pedestrian A site visit is recommended in any modelling job. If software and any potentially required interfacing with
modelling project. possible, it should be attended by the entire project team. other software products.
• User training/support – availability of software
If a site visit was conducted during the scoping phase, manufacturers/experts to resolve potential modelling
review its findings and revise them if necessary. issues.
• Ongoing software enhancements – software updates
Figure 2.1: Modelling flowchart Undertake the site visit during the time period to be and patches to resolve bugs, as supported software
modelled, if possible. Separate site visits may be required can become unusable if improvements are made to
if the modelled periods are not continuous, e.g. AM and operating systems and hardware.
PM peak hours.
Scoping Data collection 2.3 Data collection
There may be specific issues relating to each time period, If available, existing data should be obtained from
therefore it is important to observe how the site operates the relevant authorities as soon as possible following
in each. the scoping stage. Review any existing data received
Inputs and Client sign-off and then ascertain whether it can be used directly in
assumptions note Always assume there will only be one site visit, and plan the modelling process (e.g. as calibration or validation
carefully to ensure that all possible observations are data) and also whether it can be used to determine
made during that visit. If it transpires later in the project the most appropriate time periods to conduct any
that further site visits are required, they should be supplementary surveys.
Base year model undertaken.
Further details regarding the data collection process,
2.2.2 Study area including the types of data, are provided in Section 3 of
It is good practice to get agreement from all interested this good practice guide.
Calibration and parties on the extent of the study area at the earliest
validation of base model opportunity. 2.4 Inputs and assumptions
The inputs and assumptions used to develop the
The study area forms the modelled area and should be base model should be documented and agreed with
determined in discussion with the client and the reporting the client and/or the reporting authorities prior to
Model development Internal audit
authorities. The authorities may have different views from the base model build and included within an inputs
report (MDR)
the client. and assumptions note. It is also good practice to
reference/document all data sources used as part of
Consider the whole sphere of influence that may occur the model development.
with your design scheme. The extent of the study
Reference case Design year scenarios Sensitivity testing area is particularly important to the economic Further details are provided in Section 7 of this guide
scenarios (Proposed options) assessment of schemes. including a list of typical inputs and assumptions that
(Problem definition) should be documented.
2.2.3 Software selection
The Use of Microsimulation Software National 2.5 Base year model
Highways guidelines6 states on pages 6-7 that the When developing a base model, the model should
Analysis of model selection of the most appropriate microsimulation aim to replicate current conditions and pedestrian
results software is a key part of study scoping and is tied to the behaviours as closely as is practically possible. This is
choice of approach. achieved through the calibration and validation
process (refer to Sections 3, 4 and 5 of this guide) by
Evaluation of model When selecting a modelling tool, it is good practice to which the model is adjusted and compared against
outputs consider the needs of the project and client along with observed data.
specific software strengths and limitations including:

• Technical capabilities of the software – the

Reporting External audit
appropriateness of the software to undertake the task
at hand. Specific attention should be paid to model
extents and modelling detail requirements.
12 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 13

An appropriately calibrated and validated model is 2.6 Scenario testing for rail stations Therefore, it is important to consider the impact of • During construction – In many circumstances a
essential when a project involves the assessment of Section 4.1 on page 69 of Network Rail’s Station abnormal conditions in terms of overall station operation station will remain operational during periods of
proposed options or future scenarios. All further models Capacity Planning design manual1 states that: and pedestrian safety. In addition to normal operations construction. Therefore, it is important to consider
should use the calibrated and validated model as a base. (i.e. what is typically expected to occur) it is also the impact of such works in order to minimise the
good practice to consider assessing the following:
“It is important to certify that stations disruption on station operations and maintain a
The calibration and validation results should be safe environment for pedestrians. It is particularly
documented within a model development report. The continue to function acceptably • Perturbation – This is defined as a significant delay or important to assess the impact of temporary
report should include all observations and information during planned and unplanned events cancellation of a train service which results in a build- construction works during perturbed and emergency
gathered from the initial site visit and survey. affecting train service, passenger up of waiting pedestrians on the station concourse or evacuation conditions.
platforms.
The base model should be capable of consistently demand or station layout.’’ • Special events – Many stations are located within The typical scenario tests and type of assessment,
producing replicable results in order to minimise the close proximity to venues that host major events. based on Network Rail’s Project Acceleration
level of variably between model runs. This guide Consequently, there may be periods of increased
provides examples of base model run comparisons
‘‘The temporary nature of these pedestrian flows, which require crowd management
in a Controlled Environment (PACE), for new or
improved stations are shown in Figure 2.2.
showing the average and upper and lower confidence scenarios increases the thresholds in to ensure pedestrians can safely access and egress
limits to the 95th percentile. Such determining of terms of acceptable density conditions the station. Previously, the Governance for Railway Investment
averaged multiple model run statistical significance is • Emergency evacuation – This can be defined as the
modelling good practice as it provides a range
and at the same time allows passenger evacuation of all or part of a station as a result of fire,
Projects (GRIP) tool was used to manage project delivery
in Network Rail. However, Network Rail has replaced
of likelihood. safety to be maintained.” security issues or overcrowding. Typically, there are GRIP with PACE, a new tool that allows project managers
two evacuation scenarios to consider: to adapt their approach, make decisions that best
― Train on fire; and meet the needs of the project and overlay stages.
― Station on fire.

Figure 2.2: Scenario tests at each design stage for rail stations

Project Acceleration in a Controlled Environment (PACE)

Stage 2​ Stage 3 PACE ES3​ Stage 4​ Stage 5

Static capacity analysis​ Dynamic microsimulation modelling

What is static analysis?​ What is dynamic microsimulation modelling?​

• Pedestrian flow calculations based on a • Detailed agent-based microsimulation modelling
quantitative spreadsheet assessment.​ using specialist software packages.​
• Used to determine infrastructure requirements • Build and analyse at scales ranging from
and identify potential ‘pinch-points’ early in the small individual stations to large multi-level
design stage.​ interchanges. ​
Which stage of design?​ Which stage of design?​
• Stage 2 – project feasibility • Stage 3 – single option identification
• Stage 3 – single option identification • Stage 4 – approval in principle
Typical scenarios of assessment:​ • Stage 5 – detailed design
• Normal operation​ Typical scenarios of assessment:​
• Emergency evacuation • Normal operation​
• Emergency evacuation​
• Perturbation​
• During construction
14 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 15

2.7 Scenario testing for venues/stadia “Such modelling can also help management ‘‘If considering the use of crowd simulation Suggested minimum scenario tests, at RIBA design
Section 6.7 on page 90 of the Guide to Safety at stages 2 to 49 (i.e. concept through to technical design),
and designers decide whether any modelling, management should ensure
Sports Grounds5 by the Sports Grounds Safety of a new or improved venue/stadium are shown in
Authority refers to crowd simulation modelling as a improvements need to be made to the that the relevant building control and Figure 2.3.
tool which offers an insight into how well existing or layout and/or management of the ground.’’ certifying authorities are involved in the
proposed sports ground infrastructure operates using process from an early stage, that any Normal operation is what is typically expected to occur
demand data determined by static capacity analysis at an event, and so scenario testing of ingress and
simulations are carried out by competent
calculations. The guide further states that: egress under this condition is good to undertake first.
modellers and that the resulting data is If the design doesn’t perform adequately under normal
subject to detailed analysis by competent operations, it is unlikely to perform under other more
persons.’’ arduous tests and, as history tragically tells us, often
events don’t operate as expected, envisaged or ideally
planned to.
‘‘For example, when measuring flow
In addition to planning for ‘normal operation’ scenarios,
rates during ingress or egress, crowd
we should plan and test for scenarios that are
simulation modelling can typically result in abnormal and somewhat worst case from a crowd flow
a differential of +/- 10%-20% compared with capacity perspective in order to either design out or
static calculations. In such situations, the significantly reduce crowd flow safety risk at proposed
new and improved venues/stadia. It is also therefore
response of management should always
good practice to, in addition to normal operation,
be to consider why these differences undertake perturbed operation (commonly referred to
have emerged, what lessons can be learnt as exceptional egress) and evacuation scenario testing
from the data and what, if any, mitigating using pedestrian crowd flow microsimulation software.
measures might be necessary.”

Figure 2.3: Scenario tests at each RIBA design stage

Royal Institute of British Architects (RIBA) Plan of Work

RIBA stage 2 RIBA stage 3 RIBA stage 4

Static capacity analysis Dynamic microsimulation modelling

Which stage
Which stage of
ofdesign?
design?​ Which stage
Which stage of
ofdesign?
design?​
• RIBA stage 2 ––concept
• RIBA stage 2 concept design
design • RIBA stage 3 ––spatial
• RIBA stage 3 spatial coordination
coordination
Typical scenarios of assessment:​ • RIBA stage 4 – technical design
Typical scenarios of assessment: • RIBA stage 4 – technical design
• Normal egress Typical scenarios of assessment:​
•• Normal ingress
Normal egress • RIBAscenarios
Typical stage 3: of assessment:
• Emergency egress
• Normal ingress
― Normal
• RIBA stage 3:
egress
― Normal ingress
• Emergency egress ‐ Normal
• Emergency egress
egress
• RIBA‐ stage 4:
Normal ingress
― Exceptional egress – directional
• Emergency egress
evacuation
• ― Exceptional
RIBA stage 4: egress – invacuation
‐ Exceptional egress – directional evacuation
‐ Exceptional egress – invacuation
16 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 17

3. Model building requirements Such surveys are not usually required for every facility
in all model time periods; however, observations of
3.4 Other data requirements
Other required and desirable data items may include:
particularly congested areas are recommended. • Items to aid model construction;
• Any proposed development drawings;
Local knowledge of queue locations and lengths can
• Any other project specific data; and
Microsimulation model data requirements typically fall • Gradients (where they are likely to be important to the be important. Discuss them with the client and/or local
Authority officers who know the area. • Historic pedestrian models in the study area.
into 3 categories, as stated in Figure 3.1. project, e.g. ramps);
• Any features that affects flow (e.g. street furniture,
Figure 3.1: Data requirements one-way systems); Identify if CCTV exists in the area, and if it could be used 3.5 Model build reference documents
• Bus or rolling stock stop locations, queuing patterns to identify and verify queues. In the UK there are key industry standard guidelines for
rail and metro stations that are referred and adhered
1
and behaviour before, during and after boarding and
Overlay alighting; Queue length and congestion observations, in to when developing a pedestrian microsimulation
• Pedestrian crossing locations and details of operation conjunction with classified count data, assist in modelling. The first type of reference document
(e.g. crossing demand and signal timings); and calibrating the base network; see Section 4. recommended for microsimulation modellers to refer
• Minimum, maximum and average waiting and to when model building is the user guide provided by

2 Physical characteristics
and behaviour
transaction times at facilities (e.g. ticket machines,
information screens, retail units/kiosks).
Without adequate queuing and congestion observations
it will be difficult to calibrate, and hence later validate,
the software provider. It is usually accessible as a pdf
document following download of the software onto
the model. If the model does not demonstrate queuing appropriate hardware.
3.3 Demand data and congestion in the expected areas, confidence in the

3 Demand data This is one of the most important aspects of the project.
There is usually only one opportunity to define the data
model outputs will be compromised. Further good practice guidelines for model building
are available from authorities such as Network Rail and
Transport for London. The Station Capacity Planning
used to construct the model, so consider it carefully. 3.3.3 Journey time surveys
Journey time surveys can help to pinpoint areas of the design manual1 by Network Rail and Modelling with
Any existing data should be obtained as soon as Determine the existence of current or historical data. model that pedestrians may be moving through too fast, Legion: TfL Best Practice Guide4 are, for example, both
possible after study inception from the relevant Ensure then that sufficient additional data is gathered to or too slowly, so that action can be taken to remedy the commonly adhered and referred to by modellers in the
authorities. enable the model to fulfil its purpose. situation. profession.

Review any existing data received and then ascertain 3.3.1 Classified counts They are nearly always required for study areas of more The executive summary on page 10 of the Station
whether it can be used directly in the modelling process The most fundamental data to be collected are than one route, but very short journey times can be Capacity Planning design manual states that its purpose
(e.g. as calibration or validation data) and also whether classified counts. difficult to validate. Check with the reporting authorities is to:
it can be used to determine the most appropriate time whether the validation will require them.
periods to conduct any supplementary surveys. ‘‘This involves counting the number of Observed journey times for key routes through the
“Provide guidance for undertaking
3.1 Overlay pedestrians at key locations within a network can be compared against model outputs and capacity assessments for stations
It is difficult to accurately build a microsimulation model model area/network during particular used to help calibrate and validate the model. across the network. It specifies the
network without an overlay. It is good practice, therefore, intervals and classifying these by type 3.3.4 Video surveys
thresholds for planning and design of
to obtain an overlay (2D or 3D) as soon as possible
after study commissioning to help enable planning and (e.g. person of restricted mobility, Conduct video surveys whenever possible. passenger areas with a consideration
construction of the network to commence. encumbered, non-encumbered).’’ for safety, ‘value for money’ and
Survey companies can provide video services, including
recording the video, quantifying data (classified counts,
passenger experience.”
When building a design year scenario, use the original The data should be used to construct the base matrices
base overlay as the starting point for the modified file. and may also be used in the model validation. queue lengths etc.), and provide the footage in a
This helps ensure that the design year modelled area downloadable format via a cloud network or company Section 1.1 on page 5 of the Modelling with Legion: TfL
matches the base year modelled area where applicable. Ideally all key locations in the site area should be file transfer portal. Best Practice Guide states that its purpose is to:
surveyed by classified counts, preferably on the same
3.3.5 Data checking
3.2 Physical characteristics and behaviour day. If this is not possible, group adjacent areas should
Check that there is consistency in the data and it “… develop robust and accurate models
Network data requirements are as follows, although not be surveyed on each day of survey.
all may be applicable: conforms with specification before use. Basic checks with consistency and extensibility for
• Detailed plans of infrastructure such as gatelines, 3.3.2 Queuing and congested area surveys
include cross-checking of flows entering and exiting TfL rail, underground and the urban
facilities, throughput at ticket gatelines and directional
furniture, retail units, ticket machines, areas It is typically necessary to conduct queuing and
flow screenlines. realm environment. Recommendations
inaccessible to pedestrians, locations of emergency congestion surveys or observations anywhere in of modelling approaches are provided
exits etc. the study area where major queuing already occurs.
• Rough sketches, photographs or videos should Determine this during the site visit.
Failing to eliminate errors at this stage can be time which modellers can follow to obtain
consuming and therefore costly to correct later.
be used if detailed plans are not available; optimal results.”
18 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 19

4. Base model calibration

4.1 Definition Many changes at once will make it difficult to assess

The simple definition of calibration stated by the Figure 4.1: GEH Statistic which actions had the most significant effect or may
Department for Transport in Section 3.3.1 on page 18 of even result in the model behaving unpredictably and
their Highway Assignment Modelling10 guidance is: diverging further from calibration.

“Any adjustments to the model intended 2(M-C) 2 Repeat the cycle of calibrating, assessing the
GEH= calibration, and then recalibrating, only until the model
to reduce the differences between the is deemed acceptable.
modelled and observed data should be M+C
Further iterations once the model already demonstrates
regarded as calibration.” acceptable calibration is time consuming, costly and
It is recommended that at least 85%
of turning count/link flow differences, unnecessary.
The calibration exercise involves running the model between modelled and observed,
simulation several times, observing pedestrian have a GEH of less than 5. Read the relevant guidelines to determine whether
behaviour throughout the study area and making small the calibration is acceptable, even if the guideline
adjustments until the simulated behaviour closely thresholds are not fully met. If the calibration
matches the observed behaviour. The key indicators assessment falls a little short of recommended
used for calibration are, typically, pedestrian behaviour, 4.2.2 Queuing and congested areas threshold criteria, it does not necessarily mean the
classified counts and queue lengths. On-site observations, photographs and video surveys model is unacceptable.
are the typical sources of queuing and congested area
determination used to assist in model calibration. Guidelines are typically flexible regarding calibration
criteria, for example queuing is regarded in the industry
Video surveys are preferred as they can be revisited as rather subjective and therefore best compared
at the convenience of the modellers and viewed faster visually and not reported on quantitatively.
than real time.
4.5 Automatic ticket gate example
Viewing observed and modelled of the same In accordance with Section 9.2 on page 108 of the TfL
areas side-by-side on screen at the same speed best practice guide4, a standard automatic ticket gate
is recommended for comparison and resultant (ATG) delay is set at 1.8 seconds.
identification of areas requiring further calibration.
It is often evident, however, from detailed reviewing
4.3 Address the issues of survey footage at stations that ATG delay and
Areas of the model that are evidently not calibrated corresponding flow rate can vary significantly.
require investigation. The recommended procedure is to:
4.2 Criteria to assess the level of calibration varying over a wide range. For example, 3,000 The following ‘Station A’ example has been devised
4.2.1 Classified counts pedestrians per hour may be exiting a railway station, • Check the model build for errors and to point out the importance of localised calibration of
The GEH (named after Geoffrey E Havers) Statistic is a while the interchange volume between two platforms is correct them; and infrastructure facilities such as gatelines. Please note
formula used in transport modelling to compare two sets 100 pedestrians per hour: in that situation it would not be • Check the demands and assignment through that example Station A demonstrates that a fixed gate
of counts or flows. It is an empirical formula often used to possible to select a single percentage of variation that is the study area and change and adjust as necessary. delay of 1.8 seconds may not always be appropriate
determine whether or not a microsimulation model has a acceptable for both volumes. The GEH Statistic reduces at every station and it should be reviewed on a case-
significant calibration problem. this problem as it is non-linear and, therefore, a single 4.4 Recalibrating the model by-case basis, if survey data is available. This may be
acceptance threshold can be applied effectively over a It is good practice to recalibrate (against criteria stated particularly useful if a station has a high volume of non-
Using the GEH Statistic avoids some pitfalls that occur wide range of flows. The GEH Statistic formula is shown in Section 4.2) after making a small number of changes commuter demand and passengers who are unfamiliar
when using simple percentages to compare two sets in Figure 4.1. where M is the hourly flow from the model and assess the impact those have had, rather than with the station e.g. tourists.
of flows. This is due to flows in transportation systems and C is the observed flow from site survey. making a large number of adjustments all at once.
 20 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 21

4.5.1 Station A second gate delay per pedestrian as stated in Section The difference in this calibration detail is further localised calibration applied. Queuing observed on
Figure 4.2 graphically shows the observed (surveyed) 9.2 on page 108 of the TfL Best Practice Guide.4 demonstrated, see Figure 4.4, by the visual difference approach to the gateline was not modelled effectively
gateline exit flow versus modelled flow based on the 1.8 in gateline congestion at Station A with and without until the 2.6 second delay was applied.

Figure 4.2: Modelled flow vs observed flow (1.8 second delay) Figure 4.4: Gateline queuing comparison

120

Without calibration (1.8 second delay) Time: 08:34:00

100
Gateline Flow Rate (Passengers per Minute)
Gateline flow rate (passengers per minute)

ed
30 80
41
34
26
37
42 60
32
30
27
36
53 40
42
13
39
46
4 20
42
59 Surveyed
44
19
46
Modelled (1.8s delay)
0
27
29
50
49
26 Time (hh:mm:ss)
33 Time (hh:mm:ss)
31 Surveyed Modelled (1.8s)
37
43

Figure 4.3 graphically shows observed (surveyed) It is clear from the graphs shown in Figure 4.2 and
gateline exit flow versus modelled flow with a 2.6 Figure 4.3 that when the 2.6 second calculated delay
second gate delay per pedestrian (as calculated from per pedestrian at the gateline is modelled, rather than a
survey video footage). standard 1.8 second delay, it evidently results in a much
closer fit between observed and modelled flows.

With calibration (2.6 second delay) Time: 08:34:00

Figure 4.3: Modelled flow vs observed flow (2.6 second delay)

120

100
Gateline Flow Rate (Passengers per Minute)
Gateline Flow Rate (Passengers per Minute)

ed Modelled
25 26
34 32
80
39 37
33 31
33 36
37 38
33 31
60
37 35
30 28
35 31
41 40
43 42
40
29 30
36 38
41 42
14 10
35 39
20
44 43
43 45 Surveyed
42 43
43
42
45
44
Modelled (2.6s delay)
0
32 36
41 42
44 42
40 42
36 36 Time (hh:mm:ss)
31 30
33 31 Time (hh:mm:ss)
Surveyed Modelled (3.0s)
41 36
22 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 23

5. Base model validation

5.1 Definition 5.2.1 Origin–destination (O-D) matrix validation Based on the guidance provided in Section 8.2.4 on
The definition stated by the Department for Transport In accordance with Appendix B on page 87 of the page 104 of the Transport for London Modelling with
in Section 3.3.1 on page 18 of its Highway Assignment Station Capacity Planning design manual1, the Origin- Legion: TfL Best Practice Guide, journey times are
Modelling10 guidance is: destination (O-D) matrix is considered validated if considered valid when the difference between the
the variation between the modelled output and the simulated journey time outputs and on-site observations
observations (survey data) is less than 10%. (or survey) is less than 10%.
“Validation simply involves comparing
modelled and observed data that is However, it is considered good practice to also 5.2.3 Visual validation
independent from that used in calibration.” undertake a more robust comparison by applying the In addition to O-D matrix and journey time validation,
GEH Statistic to the (O-D) matrix. At least 85% of the it is also good practice to undertake a visual validation
O-D differences between modelled and observed exercise. This involves comparing key infrastructure
Model validation is undertaken only once. It is done only
should have a GEH of less than 5. elements (ticket barriers, platforms, queuing areas
until the modeller is satisfied that the model has been
etc.) from the microsimulation model with on-site video
developed to a sufficient level of accuracy that further
5.2.2 Journey time outputs footage or photographs.
model calibration will yield no further useful benefit.
It is recommended that data is plotted on a graph
where the modelled mean journey time is plotted The simulated pedestrian movements should be
showing a 95% confidence range, and the observed checked against the real-life footage for any variations
mean journey time is shown against it — example in behaviour, levels of queuing or congestion and
provided in Figure 5.1. pedestrian volumes.

Time Surveyed Modelled Upper Limit Lower Limit 350

16:00:00 260 268 252 Figure 5.1: Journey time comparison – observed vs modelled example
16:10:00 240 255 263 247
16:20:00 270 275 265
16:30:00 290 285 293 277
16:40:00 280 287 273
16:50:00 285 275 283 267 300
17:00:00 270 275 265
17:10:00 270 265 270 260
17:20:00 245 254 236
17:30:00 255 250 258 242
17:40:00 255 262 248
17:50:00 265 270 278 262 250
18:00:00 260 269 251
18:10:00 260 255 260 250
18:20:00 250 255 245
18:30:00 245 240 248 232

Seconds
18:40:00 235 240 230
18:50:00 230 225 230 220 200

5.2 Criteria to assess model validation The validation checks should go into a validation report, 150
Model validation is a comparison of model output data either a separate document for large studies, or a Surveyed
with observed (surveyed) data to assess the accuracy of chapter in a more general report for smaller studies.
the calibrated network. There are guidelines specifying Modelled
100
the criteria that determine whether the calibrated model It is considered good practice to ascertain from the *Upper and lower
is considered to be a valid representation of reality or reporting authorities, at study scoping stage, whether or confidence limit
not. This chapter discusses how the guidelines can be not a separate validation report is required. In addition,
applied to a pedestrian microsimulation model. in some cases it is good practice to get authority 50
16:00:00 16:10:00 16:20:00 16:30:00 16:40:00 16:50:00 17:00:00 17:10:00 17:20:00 17:30:00 17:40:00 17:50:00 18:00:00 18:10:00 18:20:00 18:30:00 18:40:00 18:50:00
or stakeholder agreement on the type and level of 16:00 16:10 16:20 16:30 16:40 16:50 17:00 17:10 17:20 17:30
Time (hh:mm:ss) 17:40 17:50 18:00 18:10 18:20 18:30 18:40 18:50
The guidelines may be different in other parts of the validation required at the study outset. Time (hh:mm)
Surveyed Modelled

world. Check which guidelines are appropriate for the *Dashed lines indicate 95th percentile upper and lower confidence limits.
locale and what the most up-to-date versions are.
24 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 25

6. Model outputs

6.1 Level of service plots “The Fruin LoS framework is a concept A CMD plot provides a ‘mean’ LoS for areas that are 6.1.2 CHD
There are two main types of level of service (LoS) occupied during the simulation. Areas which are not CHD plots display how long various areas of a site have
plots which are commonly used within the industry, developed by John J. Fruin11 which occupied are not included as part of the density plot registered densities greater than a specified limit. The
cumulative mean density (CMD) and cumulative high provides a measure of crowd density calculation. For example, if there are periods of high range of colours represents time rather than density.
density (CHD). Typically, CMD plots are a required and movement.” level activity followed by periods of inactivity the CMD The plot is similar to a temperature map: areas that
output from pedestrian microsimulation models. CHD plot would only show the average LoS for the periods have experienced high levels of density for a long
plots are occasionally requested as an output by a client Although Fruin LoS is commonly adhered to in the rail of activity. time appear darker, while those that have experienced
or authority at the study scoping stage. industry, the performance criteria i.e. the LoS thresholds, shorter periods of high density appear lighter.
should be determined and agreed at the study scoping It is good practice to use an appropriate LoS based on
6.1.1 CMD stage with the relevant authorities. For example, for the area or element that is being assessed (walkways, A legend should be attached to each CHD plot to
CMD plots display the mean levels of density registered venue/stadia design more applicable LoS for crowd queuing, milling, standing etc.). illustrate the duration that each colour represents
in an area within a specified time period. They are densities are stated in the Green Guide5 Figure 11, above the specified density. The LoS threshold
generally used in combination with value ranges page 119. A legend should be attached to each CMD plot to applicable for a CHD plot should be in accordance
corresponding to LoS. illustrate the LoS density that each colour represents and with guidelines appropriate to the specific scenario
LoS is typically categorised into six bands. LoS A is the also the time period to which the plot corresponds. and facility type.
highest level of service, providing the most circulation An example of a typical CMD plot is shown as Figure 6.1.
freedom. LoS F is the most restrictive and is considered An example of a typical CHD plot is shown as Figure 6.2.
unsafe if predominant over a sustained period.
Figure 6.1: CMD plot example Figure 6.2: CHD plot example

D
sq

A
26 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 27

6.2 Space utilisation plots units. Colours represent how much time the space is 6.3 Journey time outputs Figure 6.4 and Figure 6.5 are typical examples of how
Space utilisation (SU) plots are typically used to highlight used during a specified time period e.g. peak 15 minutes: It is recommended that data is plotted on a graph where journey time comparisons should be reported.
popular movements (e.g. pedestrian desire lines) and blue being lightly used and red being heavily used. the modelled mean journey time is plotted showing a
are useful to identify over - or underutilised areas of the 95% confidence range, and the observed mean journey
modelled environment. A legend should be attached to each SU plot to illustrate time is shown against it.
the range of time each colour represents.
In addition, SU plots can often help determine the
optimum location to site facilities such as toilets and retail An example of a typical SU plot is shown as Figure 6.3. Figure 6.4: Journey time comparison – observed vs modelled example

Figure 6.3: SU plot example

Seconds
Surveyed

Modelled

*Upper and lower

confidence limit

Time (hh:mm:ss)

*Dashed lines indicate 95th percentile upper and lower confidence limits.
Time Scenario 1 Upper Limit 1 Lower Limit 2 Scenario 2 Upper Limit 2 Lower Limit 2 Scenario 3 Upper Limit 3 Lower Limit 3
400
06:30:00 140 142 138 145 147 143 141 149 133
06:40:00 145 147 143 147 148 146 145 153 137
06:50:00 140 143 137 145 147 143 145 153 137
07:00:00 144 145 143 150 153 147 143 151 135
07:10:00 143 145 141 153 155 151 145 153 137 350
07:20:00 145 146 144 155 157 153 147 155 139 Figure 6.5: Journey time comparison – scenario testing example
07:30:00 141 143 139 151 152 150 143 151 135
07:40:00 143 144 142 153 156 150 145 153 137
07:50:00 142 144 140 152 153 151 145 153 137 300
08:00:00 141 144 138 151 153 149 143 151 135
08:10:00 143 144 142 153 154 152 250 280 220
08:20:00 144 146 142 154 156 152 270 295 245
08:30:00 142 143 141 152 155 149 180 210 150
08:40:00 142 144 140 152 154 150 170 200 140
250
08:50:00 143 144 142 153 155 151 250 280 220
09:00:00 141 143 139 151 152 150 160 175 145

Seconds
09:10:00 142 143 141 152 154 150 145 150 140
09:20:00 142 145 139 152 153 151 155 160 150 200

150

Scenario 1
100
Scenario 2

Scenario 3
50
*Upper and lower
confidence limit
0
06:30:0006:40:0006:50:0007:00:00 07:10:00 07:20:00 07:30:00 07:40:00 07:50:0008:00:00 08:10:00 08:20:0008:30:0008:40:0008:50:0009:00:00 09:10:00 09:20:00
06:30 06:40 06:50 07:00 07:10 07:20 07:30 07:40 07:50 08:00 08:10 08:20 08:30 08:40 08:50 09:00 09:10 09:20
Time (hh:mm:ss)
Time (hh:mm)
Scenario 1 Scenario 2 Scenario 3
*Dashed lines indicate 95th percentile upper and lower confidence limits.
28 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 29

6.4 Model screenshots It is recommended to include labels to highlight key 6.5 Model videos type. Separately, videos can be produced which assign
Model screenshots are typically used to show the features of the modelled environment in order to assist Videos are an extremely effective output to show colours based on the LoS the pedestrian experiences at
position of entities, extent of queuing and congestion those less familiar with the layout. stakeholders how pedestrians move around the that point in time.
at certain points in time within the simulation. Model space and highlight different features of the modelled
screenshots can be used to compare the same location Examples of typical 2D and 3D model screenshots are environment. It is again recommended to include labels to highlight
and time period for different scenarios. shown as Figure 6.6 and Figure 6.7. key features of the modelled environment in order to
When recording a video, it is recommended to colour assist those less familiar with the layout.
code the pedestrians based on their final destination or

Figure 6.6: 2D model screenshot example

Figure 6.7: 3D model screenshot example

30 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 31

6.6 Generalised journey times and social cost For the GJT output, different weightings should be
Social cost outputs are the monetised generalised applied to the different activities e.g. walking, waiting, in
journey time (GJT) and congestion factor (CF) order to represent the desirability of each.
associated with all of the simulated pedestrians within For the social cost output, the assumed value of time
a model. It should be noted that not all pedestrian (VoT) and days per year should be reviewed per project
modelling packages are capable of producing such and agreed with the client and/or reporting authorities.
outputs. If this output is required then the most
appropriate software package should be selected It is good practice to undertake a minimum of five
at the beginning of the project. model runs of each scenario. The average result of the
five runs should then be used for reporting. Should
Social cost outputs are predominately used by the results of the five model runs not converge well,
Transport for London (TfL) as part of the business case additional runs are recommended to ensure the average
development process in order to help determine the is representative.
potential benefits or disbenefits of proposed scheme
implementation. The social cost and especially GJT outputs are useful
when comparing two or more layouts or scenarios.
The Business Case Development Manual (BCDM) In addition, highlighting the disbenefit of a layout or
parameters are stated in Section 13 on page 124 of the scenario. An example of the social cost output for one
TfL Best Practice Guide.4 scenario test is shown as Table 6.1.

6.7 Flow rate and density profiles This type of output is useful when determining how long
Flow rate and density profiles for a specific and how often a particular area or element exceeds
infrastructure facility or area (stairs, walkways, an agreed LoS criteria threshold. Additionally it can be
concourse etc.) can be extracted from the useful when compared against surveyed data as part of
microsimulation model. It is considered good practice the model validation process.
Table 6.1: Social cost output example to display this output on an LoS banded line chart.
An example of a density profile on a LoS banded line
chart for a staircase is shown as Figure 6.8.
By activity
Weighting JT GJT CF Cost Cost Total Annualised
journey congestion cost

Walking 2.000 50775.2 101550.4 239.6 £214.10 £0.51 £214.61 £53,651.78

+ 0.500 CF
Figure 6.8: Flow rate and density profiles example
Waiting 2.500 228.6 571.5 0.0 £1.20 £0.00 £1.20 £301.23
+ 1.000 CF 3.0

Queuing 3.400 7.2 24.5 0.0 £0.05 £0.00 £0.05 £12.90

2.5

Delayed 2.500 6266.4 15666.0 0.0 £33.03 £0.00 £33.03 £8,257.29

(pax/m2)
Density (pax/m 2)
2.0
On stairs up 4.000 1546.8 6187.2 0.0 £13.04 £0.00 £13.04 £3,261.17 ExperiencedDensity
LoS F
LoS E
1.5 LoS D
Experienced

On escalator 1.500 0.0 0.0 0.0 £0.00 £0.00 £0.00 £0.00 LoS C
up LoS B
LoS A
Average

1.0
On stairs 2.500 3435.6 8589.0 0.0 £18.11 £0.00 £18.11 £4,527.12
Average

Density
down
0.5
On escalator 1.500 0.0 0.0 0.0 £0.00 £0.00 £0.00 £0.00
down
0.0
08:15 08:20 08:25 08:30 08:35
Total: 62259.8 132588.6 239.6 £279.54 £0.51 £280.05 £70,011.49 Time
Time (hh:mm)
32 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 33

7. Model reporting
7.1 Introduction All data sources should be referenced within the note.
It is important to document as much as possible
throughout the entire project life cycle. The key items The note should contain the following information:
should include:
• Overview and scope
• Project objectives • Modelling approach
• Data requirements • Model scenarios and time periods
• Modelling approach • Model extent and drawing references
• Model inputs and assumptions • Survey data (if applicable)
• Performance assessment criteria • Pedestrian demand (matrices)
• Modelling results • Arrival and departure profiles
• Conclusions and recommendations • Pedestrian characteristics
• Pedestrian routing
Documentation and agreement are essential,
• Flow rates
as model inputs and assumptions can often be
challenged during the model audit/review period. In • Operational information
addition, any changes to the model could result in a • Outputs and assessment criteria
different set of results and outcomes. • Model limitations
7.3 Model development report (MDR) The report should contain the following information:
7.2 Inputs and assumptions note An example of a typical table of contents is shown as The modelling methodology and assumptions, including
A model inputs and assumptions note should be Table 7.1. how the model has been sufficiently calibrated and • Survey methodology
prepared and agreed with the client and/or the validated against the observed data.
reporting authorities prior to the model build stage. • Observed data
• Modelled data
It is good practice to undertake a survey of the
modelled area in order to collect the required inputs. • Model calibration results
The survey may include classified counts, journey • Model validation results
times and flow rates at key locations (staircases,
Table 7.1: Example of table of contents for model inputs and assumptions note passageways, escalators etc.). An example of a typical table of contents is shown as
Table 7.2.
1. Scope of works 1.1 Project objectives and scope
1.2 Modelling approach and methodology Table 7.2: Example of model development report table of contents

1.3 Model scenarios and extent 1. Survey methodology 1.1 Survey methodology
2. Pedestrian demand 2.1 Survey data (if applicable) 1.2 Data collection and time periods
2.2 Pedestrian demand matrices 2. Survey data 2.1 Classified counts
2.3 Demand profiling and distributions 2.2 Journey times
3. Model inputs and assumptions 3.1 Pedestrian characteristics 2.3 Flow rates
3.2 Routing 3. Model calibration 3.1 Screenline counts
3.3 Infrastructure flow rates and capacity 3.2 Arrival and departure profiles
4. Performance assessment criteria 4.1 Level of service framework 3.3 Infrastructure flow rates
4.2 Assessment criteria 4. Model validation 4.1 Origin-destination matrix
4.3 Limitations 4.2 Journey times
34 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 35

7.4 Modelling report The report should contain the following information: Table 7.3: Example of modelling report table of contents
The modelling report should clearly document the
results and findings of the study. It should demonstrate • Executive summary 1. Scope of works 1.1 Project objectives and scope
how the project objectives and scope have been • Project objectives and scope 1.2 Modelling approach and methodology
achieved, including any further recommendations.
• Summary of the key model inputs and assumptions 1.3 Report structure
The report should be written in a simple manner so • Reference to the inputs and assumptions note and 2. Model inputs and assumptions 2.1 Pedestrian demand scenarios
that it can be understood by technical and non-technical calibration and validation report (if applicable)
2.2 Key model inputs and assumptions
readers. The inputs and assumptions note, and the • Performance assessment criteria
3. Performance assessment criteria 3.1 Level of service framework
model development report should be included as • Description of the scenarios modelled (demand years,
appendices or referenced within the modelling report. layouts etc.) 3.2 Assessment criteria
• Modelling results for each scenario 4. Modelling results 4.1 Level of service plots
• Summary of results 4.2 Journey times
• Conclusions and recommendations 4.3 Screenshots
5. Summary and conclusions 5.1 Summary of results
An example of a typical table of contents is shown 5.2 Recommendations
as Table 7.3.
36 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 37

8. Model auditing

8.1 Introduction 8.2 Internal audits any errors and the possible benefits of any suggested
The purpose of a model audit is to ascertain whether It is considered good practice to carry out internal audits improvements.
the model is fit for its purpose. at key stages in the development of the model.
8.3.2 Authority or client
This should ensure that the model progresses smoothly Authorities and clients typically receive completed
The primary difference between an internal through its development phase. models from an external consultant and are more
and external audit is that internal audits interested in the implications of any model issues or
should be carried out several times during Noting, communicating and retaining any flaws identified rather than how to correct such flaws.
model development. External audits are recommendations is required but there is no need for
a detailed report – a simple email or technical note is It is not typically necessary to go into the details of how
generally only requested once, when the acceptable. to rectify a modelling issue; instead it is good practice
model is considered complete. to state the pragmatic implications that not rectifying
The auditor should ideally be someone in the the issue could have on the modelling results and,
organisation with a detailed knowledge of the particular consequently, wider project and stakeholders.
microsimulation software used, but with no direct (day-
to-day) involvement in the project. 8.4 Audit report
The audit report is typically a written document. It will
8.3 External audits probably contain recommendations and suggestions for
External audits fall into two categories, and the tone will any steps required for the model to be deemed
vary accordingly: fit for purpose.

• Requested by a consultant prior to submitting the Table 8.1 contains suggested base model audit report
project to a client or authority; or headings, and a brief description of what to look for
• Requested by an authority when the model has been under each. The headings can be altered to suit the
built and submitted by a consultant. audit. Different models will require different approaches,
and therefore additional headings should be added
Always observe the simulation running. A lot can be if appropriate.
learned very quickly by running a model and observing
the behaviour. Occasionally the client will request that output data from
the model should be checked against observed data
When suggesting changes, check that the change as part of the audit, e.g. as a validation check, and that
actually works by incorporating it in a copy of the audit would require its own section in the report.
model before making the recommendations.
When writing the audit report highlight any
8.3.1 Consultant recommended changes in bold text. This makes it
The consultant is seeking assurances that the model is fit easier to identify the parts of the model that the auditor
for the purposes of the project. The audit should take the suggests requires extra work.
form of a technical note pointing out where the model
does not conform to good practice.

The techical note should guide the consultant towards

ensuring that the model can be made fit for purpose
prior to the undertaking of analysis to be submitted to
the client. This means explaining the implications of
38 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 39

Table 8.1: Suggested base model audit report structure

Report heading Contents Report heading Contents

1. Introduction Concise introductory few paragraphs. 4. Model build The physical building of the model including: accessible space, object placement and
configuration.
1.1 General A description of the commission which should include a reference to the client
that commissioned the audit. Also, mention the consultant that built the models (if 4.1 Accessible space / Have drawing references been provided? Does the modelled layout accurately reflect
different). layout – base model the existing layout e.g. stair and walkway widths?

1.2 Purpose of audit A statement explaining the purpose. 4.2 Accessible space If several models are being audited, any changes between the base and option test
1.3 Format of audit A statement explaining the format. / layout – differences models (or AM and PM models) should be mentioned. If there are no differences in
from other models parameters between the various models, say so.
2. Parameters Explain all relevant global parameters used in the model.
4.3 Entries and exits Have all potential entry and exit points been included within the model? Are the entry
If several models are submitted for audit then the main text should refer to the base – base model and exits points appropriately located within the model?
model, and only the changes made between the base and the option test models
should be commented on. 4.4 Entries and exits If several models are being audited, any changes between the base and option test
– differences from models (or AM and PM models) should be mentioned. If there are no differences in
2.1 Description of model A description of the area covered by the model, supported by a graphic image. other models parameters between the various models, say so.
2.2 Behaviour and State the version of software used to build the models. Ensure that the version of
assignment – base software used to construct the model is the version used to conduct the audit! 4.5 Vertical circulation Have all potential vertical circulation elements been modelled correctly in terms
model Any changed defaults should be highlighted. objects – base model of useable width and flow rates? Has evidence been provided which compares
observed flow rates versus modelled?
2.3 Behaviour If several models are being audited, any changes between the base and option test
4.6 Vertical circulation If several models are being audited, any changes between the base and option test
and assignment – models (or AM and PM models) should be mentioned. If there are no differences in
objects – differences models (or AM and PM models) should be mentioned. If there are no differences in
differences from other parameters between the various models, say so.
from other models parameters between the various models, say so.
models
4.7 Route assignment Check that objects have been reasonably used to direct and/or modify the behaviour
3. Demand Comment on the way pedestrians have been added.
– base model of pedestrians within the simulation. Has justification been provided for the use of any
3.1 Pedestrian types Check the pedestrian types; have they changed from the defaults and, if so, are additional routing or behaviour modifiers?
– base model changes justified (e.g. using survey data or observations)? Check for changes to the
pedestrian dynamics of each type, especially walk speed and aggression as these 4.8 Route assignment If several models are being audited, any changes between the base and option test
can be used to influence flow rates. – differences from models (or AM and PM models) should be mentioned. If there are no differences in
other models parameters between the various models, say so.
3.2 Demand matrices Discuss the number of matrices used and check which pedestrian types are assigned
and flow profiles to each matrix level, and the proportions of each type in each matrix. Check the 4.9 Pedestrian speeds/ Have the pedestrian speeds and/or object parameters been modified? If so, has a
– base model validity of the demand matrix for each time period. objects configuration/ reasonable justification been provided for the modifications?
throughput
Check for flow profiles. If the demand matrix has been flat profiled, is that appropriate, – base model
or would it be better to use an alternative profile instead? Explain the benefits of
using multiple profiles if they do not exist. 4.10 Pedestrian speeds/ If several models are being audited, any changes between the base and option test
objects configuration/ models (or AM and PM models) should be mentioned. If there are no differences in
Check the profiles have been applied correctly. It is common to encounter a model throughput parameters between the various models, say so.
with several profiles only to discover that only one of these is being used as the – differences from other
others have not been assigned properly. models

5. Outputs How many runs (sometimes referred to as ‘seeds’) of the model are the outputs
3.3 Time periods Does the model have multiple time periods? derived from? For example, are journey times averaged from a number of runs? Have
– base model Have warm-up and cool-down (i.e. shoulder periods) been included? upper and lower confidence intervals been derived and shown as part of model
Are they applied correctly? analysis output?
What is the purpose of each period?
6. Conclusion Write a conclusion. Comment on whether the model(s) in their present condition are
If there is more than one model, highlight the differences in all the above in each. fit for purpose. If not, the client should be able to make the model fit for purpose by
following the recommendations in the audit report.
40 | Pedestrian microsimulation modelling UK good practice guide Pedestrian microsimulation modelling UK good practice guide | 41

References

1
Network Rail (2021) Station Capacity Planning Design Manual
NR/GN/CIV/100/03

2
Transport for London, London Underground (2012) G371A
Station Planning Standards and Guidelines

3
Transport for London, London Underground (2019) S1371
Station Capacity Planning

4
Transport for London, London Underground (2021) Modelling
with Legion: TfL Best Practice Guide

5
Sports Grounds Safety Authority (2018) Guide to Safety at
Sports Grounds. 6th Edition

6
National Highways (2022) Guidelines for the Use of Microsim-
ulation Software

7
Department for Transport (2014) TAG UNIT M1 Principles of
Modelling and Forecasting

8
SIAS Limited (2012) The Microsimulation Consultancy Good
Practice Guide

9
Royal Institute of British Architects (2020) RIBA Plan of Work
Template. http://www.ribaplanofwork.com/

Department for Transport (2020) TAG UNIT M3.1 Highway

10

Assignment Modelling

11
Fruin JJ (1971) Pedestrian Planning and Design. New York
Better outcomes through
innovation and excellence.

Engineering. Management. Development.

© Mott MacDonald Limited

All rights reserved. This document is
subject to copyright protection.

No portion may be copied without the

prior written consent of the authors.

mottmac.com