An automatic classification of urban texture: form and compactness of

morphological homogeneous structures in Barcelona

Nicola Colaninno1, Josep Roca Cladera2, Karin Pfeffer3

Centre of Land Policy and Valuations (CPSV) Technical University of Catalonia (UPC)
Av. Diagonal, 649, 08028 Barcelona – SPAIN, e-mail: [email protected]

Keywords: urban texture, morphological structures, classification, built-up area

ABSTRACT

This paper investigates the morphology of the city of Barcelona; the main objective is
to identify and automatically classify morphological homogeneous categories and determine
the level of compactness of urban texture. The classification consists of several steps. A first
analysis was performed to identify different morphological types in the built-up area and
typify urban patterns by using geometrical and relational characteristics such as shape of
buildings and spatial interaction between them in terms of distance. This was followed by an
automatic classification of urban structures with homogeneous characters. Finally, the
morphological compactness was determined, calculated as “equilibrium” between size and
distances of buildings. In order to make the automatic classification, we calibrated a set of
indices, capable to describe specific formal and functional characters. Using GIS and
statistical analysis, such as factor and cluster analysis, we categorized several homogeneous
structures.
This classification process was based on geometric characteristics, assuming that, in
general, different functions require different forms. By aggregating subjects of similar
behaviour we could identify to which kind of urban configuration an area possibly belongs.
The outcome of this study can support urban planning and management in identifying target
areas concerning socio-economic interventions within urbanized territories.

1
Researcher at the Center for Land Policy and Valuations of the Universitat Politecnica de Catalunya
2
Professor, Director of the Center for Land Policy and Valuations of the Universitat Politecnica de Catalunya
3
Assistant Professor at the Department of Geography, Planning and International Development Studies of the
Universiteit van Amsterdam and member of AISSR
 1. INTRODUCTION

Cities can be classified according to their urban morphology. Depending on the spatial
configuration of built-up and empty spaces, blocks and streets (or squares) we can distinguish
compact or fragmented cities as well as continuous or discontinuous. Anas et al (1998), in
their review of the contributions to an understanding of urban spatial structure, pointed to the
diversity of the nature and dynamics of urban morphology. The principle behind these studies
is that a city, as a self-organism, has a unique identity that is preserved through the years,
despite the passage of time and the spatial and functional transformations that occur
(Rabie,1991); (Frenkel 2004). The configuration of the physical elements, with their own
functional dynamics, produces different “drawings” in the cities. By disaggregating the urban
texture into different components, it is possible to study the topological, geometrical and
dimensional relationships between the elements. In recent years, research on urban
morphology has investigated the way in which cities are created, renewed, and extended in
space. Studies have constructed random models to describe town growth (for example, Batty,
1998; Batty and Longley, 1994; Benguigui et al, 2001; Longley and Mesev, 1997); (Frenkel
2004).
The paper is structured in the following way. First we will briefly describe the
relevance of this work, the objectives, and the main characteristics of urban morphology.
This is followed by the methodology and a description of the typologies of urban texture in
Barcelona. Then comes a description of results of the morphological analysis produced by the
proposed methodology and we will conclude the paper with a discussion of the major
findings.

2. EFFECTIVENESS OF THE RESEARCH AND OBJECTIVE

The increasing amount and detail of digital spatial data is a useful resource for
automatic interpretation of urban phenomena such as urban morphology. Knowledge on urban
morphology can be a useful resource in urban planning and management, for example in
socio-economic targeting, or also for sustainable issues.
The main research objective is to design a methodology for automatic classification of
urban texture which produces homogeneous areas with respect to the shape of buildings, and
their spatial configuration and arrangement in space. This also involves the selection and
calibration of a set of indices which are suitable to describe certain geometric and relational
characteristics of urban structures and which can be used in the subsequent automatic
classification of urban structures. By evaluating characters such as the density of the city and
the efficiency of urban forms we also attempt to identify what variables have a higher impact
on the estimation of urban compactness.
The experiment will be performed for the municipality of Barcelona, in Spain, using
the footprint data of built-up area and constituting around twenty thousand objects in a shape
file format. Figure 1 shows a planimetric map of the built-up area items of the city.

Figure.1 Planimetric map and section of Barcelona: case study

 3. URBAN MORPHOLOGY

The morphological identification and the “quantification” of building groups is based

on geometric characteristics of the objects, assuming that, in general, different functions
require different forms of buildings.
The assumption is that different urban structures generate different economic and
social dynamics, and that in such complex framework of the urban areas, all these factors
interact and affect each other.
It is also important to underline that buildings with similar properties will be close to
each other or even form groups in feature space (Steiniger et al. 2007)
Morphology studies suggest parameters to measure the urban configuration that might
serve as tools in the hands of urban designers to assist the functioning of intra-urban systems.
Geometrical and topological indicators are used in investigating relationships between the
urban pattern and other systems that affect urban morphology. Steiniger et al. (2007) argue
that we can define to what type of urban structure a neighbourhood possibly belongs to by
measuring the character of the buildings in a neighbourhood, based on similarity and
proximity.
The new generation of models tries to recapture, based on past trends, the evolution of
the formation of the city's shape and its geometrical pattern. Understanding the urban system
and its components can be supported by a quantitative analysis of the relative share of the
various land uses (Frenkel 2004).

4. METHODOLOGY

The methodology consists of three parts. A first analysis concerns the identification of
different groups of building typologies and the analysis of patterns of urban structure based on
geometric characteristics such as the shape of the buildings and the spatial interaction
between them in terms of distances. We identified seven different kind of urban fabric, based
on basic concepts of urban planning theory which are presented in section 5.
The second analysis is to identify similarities in the morphological configuration of
structures, in order to automatically classify urban textures. Based on the fact that buildings
have particular geometric features, and that the zoning of a city may cause that nearby
buildings are similar (although exceptions occur), we concentrate on the application of
indicators that measure shape and the distances between buildings.
The automatic classification into homogeneous morphological is done by statistical
analysis, factorial and cluster, and the application of a filter of spatial correlation4 in a 200
meters buffer around each building.
The final part concerns the evaluation of the level of compactness of homogeneous
areas from the morphological point of view. Figure 2 shows a conceptual map of the applied
methodology.

Figure.2 A conceptual map of the methodology

Characterizing Urban Structures

Urban Texture Computing Urban Planimetric Footprint of

Typologies Structures Map Built-up Area

Morphological
Indexes
Data Source

Dense

Building Form
Filter of Characters
Spatial Correlation
200 m Buffer Compact

Buildings Distances
Relation

Complex
Automatic
Classification of
Buildings

Homogeneous Compactness
Morphological Areas

4
The filter of spatial correlation consists in making a buffer around each building and calculates the average
value of a specific index, for the objects intersected by the buffer. It provides a new value which takes into
account the major trend of nearby objects.
 5. URBAN FABRICS TYPOLOGIES

It has been widely asserted that the morphology of urban areas is a result of the
interactions of urban function and urban form. This has led to a number of studies to
postulate, either explicitly or implicitly, that a shaping exists between the physical form (land
cover) of the urban fabric and its corresponding function and activity (land use) (Barr et al.
2004). As stated in the introduction, studies of urban morphology investigate the urban
pattern and the structure of urban land uses from design perspectives.
The metropolis of today is a palimpsest, characterized by the continuous overlapping
of different architectural languages, the multiplicity of spatial forms, construction techniques
and functions (Secchi 2001). The city is the sum of smaller building blocks, which could be
delineated depending on their characteristics strictly correlated to the planning process and
resulting from the stratification of different epochs. Within the city of Barcelona we defined
seven classes of homogenous structures, also related to different historical periods, and named
these typologies of urban texture as follow:
• Old Town
• Enlargement
• XX Century city
• Fragmented City
• City of Seventies
• Suburb
• Industrial, Commercial, and Special Buildings
In order to better understand the characteristics of the typologies, we will first describe briefly
the substance of the structures which are illustrated in figure 3.

Figure.3 Examples of typologies of urban fabrics in Barcelona

 In Barcelona, the Old Town, mostly identified like the historical centre, originates
from the gothic period. It presents a continuous and highly dense urban system. The fabric is
very compact and the boundaries are well defined and form a closed circle.
The layout of the Enlargement, built around the end of the XIX century, in Barcelona
commonly known as “Eixample”, is a regular grid. It is composed of large grains and a
regular roads pattern, often oriented differently than the adjacent neighbourhoods.
The organization of this structure does not show special hierarchies and the public space is
represented by main roads, extensive and orthogonal, motivated mainly for reasons of
functionality (Mattogno 2007).
The XX Century City, together with the Enlargement and the Old Town, defines the
limits of the consolidated city. This type of structure consists of big blocks, bordered by
buildings with continuous fronts, alternating with a small discontinuous tissue which provides
the common spaces. The density is now decreasing, due to the less number of floors, but the
design is still clear due to the road network which testifies that the open space represents a
clear choice of planning.
In the Fragmented City land occupation is considerable lower than in the first three
types. It has more open space, almost leading to a switch in the order of things: it seems like
now the empty space is organizing the position of the buildings, while the hierarchies of roads
are slowly disappearing, together with the importance of the public space.
The City of Seventies (Mattogno 2007), marks the transition from the modern city to
the contemporary city. Lines and blocks are juxtaposed in an inconsistent manner, passing
from dense building structure to a progressive loss of density as we move away from the main
road infrastructure. It is almost impossible to detect any presence of public spaces, except for
the dominant space represented by the road, which however does not lead to any recognizable
urban design; in fact the urban morphology is only characterized by the shapes of wide open
spaces, high buildings and big infrastructures. The clear route of the highway, in most cases to
a higher level, highlights the status of suburbs and produces considerable impacts of physical
and visual barriers (Mattogno 2007).
In contrast to the contemporary residential structures characterized by intensive blocks
of buildings, the Suburb is a structure of houses for one or two families, fragmented into small
lots, with own gardens, completely dispersed and located at the periphery of the city. Any
identifiable configuration of facilities and types of occupation of the territory is disappeared
and the great value of the open space has lost all morphological and structural features, being
deprived of their intrinsic meaning (Mattogno 2007).
 The Industrial, Commercial, and Special Buildings area, represents groups with
plenty variety of forms, even if the blocks are mainly industrial. It is because an industrial
area is often constituted by objects with ample variety of characteristics in terms of
dimensions, form and distances, and it is common across a city to find buildings with
“special” characters such as highly isolated, or with big dimensions.

6. MORPHOLOGICAL INDICES FOR URBAN FABRICS

We identified nine indices, useful to quantify the morphology of the city, based on
several features, calculated depending on formal and relational characters of the buildings.
The nine indices were selected on the basis of a statistical and also qualitative analysis of the
originally 15 variables found in the literature and the concepts explained. The other 6 indices
were removed from the analysis because of too little explanatory contribution or redundancy.
Table 1 shows the indices.

Table.1 Indices of urban morphology

Form / Building Relation / System

Building Area Buildings Proximity

AREA = a i

∑
n
Core Area Index ai
PROX = 1
⋅n
∑
n
ac d ij2
CAI = i (100 ) 1
ai

Area to Perimeter Ratio Buildings to Buffer Area

ai
AP =
pi
∑
n
a iInt
BHAD = 1
Shape Index I a i Buff
Area/π
SSPI =
Longest Axis

Shape Index II Buildings to Convex Hull Area

pi
SHPI =
2 π ai
∑
n
a iInt
CHAD = 1

# Corners of the building a i Hull

CORN
 The core area index (CAI), like implemented in fragstats (McGarigal et al. 2002), is
the ratio between the core area of the building, with an inside offset of 10 meters, and the area
of the building. We also computed the ratio between area and perimeter of each object in
order to reach a value of building efficiency (AP).
Shape index I is the Schumm’s longest axis to area ratio (MacEachren 1985), while
shape index II is the index of Gravelius or coefficient of compactness, which compares the
profile of any polygon with the profile of the circumference, intended as the maximum
compact form.
The proximity (PROX) is calculated like the ratio between the sum of areas of the
buildings in a buffer of 200 meters around a building and the sum of the squares of the
distances, toward the central object, of all building in the buffer. The result is multiplied for
the number of buildings in the buffer. BHAD is computed as follow: 1) Buffering of buildings
and selection of all buildings intersected by buffer. 2) Calculation of intersection regions
(Buffer, Buildings). 3) Calculation of area of intersection to area of buffer ratio (Steiniger et
al. 2007). CHAD is computed as follow: 1) Buffering of buildings and selection of all
buildings intersected by the buffer. 2) Calculation of convex hull around these buildings. 3)
Calculation of area of buildings to area of convex hull ratio (Steiniger et al. 2007).
In order to identify similarities in the morphological configuration of structures
automatically classify urban textures, we also aimed to set synthetic indices by using our
basic indicators. By applying a factorial analysis and using principal component analysis with
the varimax rotation method we obtained three synthetic indexes which are explaining our
model at more than 83% (table 2).

Table.2 Components and total variance of the model explained through the factorial analysis

Total Variance Explained

Initial Eigenvalues Extraction Sums of Squared Loadings Rotation Sums of Squared Loadings
Component
Total % of Variance Cumulative % Total % of Variance Cumulative % Total % of Variance Cumulative %

1 4,333 48,147 48,147 4,333 48,147 48,147 2,856 31,729 31,729

2 1,648 18,313 66,460 1,648 18,313 66,460 2,793 31,034 62,763

3 1,530 17,004 83,464 1,530 17,004 83,464 1,863 20,701 83,464

Extraction Method: Principal Component Analysis.

The extracted indices refer to three characteristics which could summarize the degree
of compactness of urban forms. In fact by interpreting the components we identified some of
the main characteristics of compactness commonly used in morphology studies; these are
density of the urban texture, the formal efficiency of buildings and complexity of the shape.
The latter two are indicators of form. Efficiency, in fact, is referring to the effectiveness of
form, taking into account the dimension of buildings but related to the level of squareness of
the building. Complexity, instead, is evaluating how convoluted the outline is, with more
details. Table 3 shows the loadings of the indices on the three components, which we can
assign the following meanings: Dense, Efficient, and Complex. Figure 4 explains the position
of every basic index in the 3-dimensional space of the three principal components.

Table.3 The result of the factorial analysis and the interpretation of components

Rotated Component Matrixa

Component

1 2 3

Dense Efficient Complex

AREA ,150 ,868 ,223

CAI ,170 ,913 -,080

AP ,278 ,909 ,077

CHAD ,932 ,223 ,122

BHAD ,945 ,208 ,142

PROX ,947 ,152 ,088

SHPI ,199 ,213 ,887

SSPI ,004 -,206 ,729

CORN ,165 ,418 ,663

Extraction Method: Principal Component Analysis

a. Rotation converged in 4 iterations.

Figure.4 The three dimensions of the model in a 3D representation

 Considering that the concept of zoning in urban planning produces homogeneous
contiguous blocks, a filter of spatial correlation has been applied to the three synthetic indices.
Specifically, in order to implement the filter procedure, a buffer of 200m has been applied to
every building, followed by calculating the mean and the standard deviation of the synthetic
indices for the objects in that buffer. The result after applying the filter, provides our final
indices (Dense*, Efficient* and Complex*), as shown in the figure 5, and the standard
deviation of first synthetic indices as additional indicator for classifying urban structures.

Figure.5 The spatial behaviour of the Dense*, Efficient*, and Complex*

If we intend the compactness like high levels of density and efficiency, but also of the
complexity of buildings forms, it could be effectively hypothesize the relation between urban
structures and compactness, also related to distances, by observing the image upon.
In order to determine the distance of the buffer, we have taken into account several
studies in which the measure most commonly accepted and proved to be useful in the analysis
of urban dynamics is that of the 200 meters. Varying this parameter should have an influence
on the classification result, which can be concluded from the work of Le Gléau et al. (1997)
and Boffet (2001). When classifying the buildings it is very likely that a building will have
the same structure type as buildings in the neighbourhood owing to the underlying zoning
structure. Maximum distance values given by Le Gléau et al. (1997) vary between 50 m (e.g.
Scotland) and 200 m (e.g. France). The results show that a maximum of classification
accuracy and certainty is reached for a 200 m buffer radius. Improvements can still be reached
if the dataset has been characterised with density indices based on a 200 m buffer radius
(Steiniger et al. 2007).
 7. AUTOMATIC CLASSIFICATION OF URBAN FABRICS

In this section we demonstrate the suitability of the cluster analysis for identifying,
automatically, similar morphological structures across a city.
The cluster analysis was applied to the morphological components Dense*, Efficient*,
and Complex* as well as their standard deviation, calculated for every building. We have
chosen representative samples (‘training sites’) for each of the seven urban typologies,
qualitatively defined in section 4. For each sample, we calculated the mean value of the
morphological components and their standard deviation. The resulting values were used as
cluster centres for the seven categories defined by the typology. Additionally the analysis was
made without iterations, in order to "lock” the value of cluster centres. The outcome is
illustrated in the figure 6.

Figure.6 Result of Cluster analysis for seven urban typologies

 We propose to include spatial characteristics in the identification process for city
clusters such as quantitative data on the size, composition, and intensification of the built-up
area, as well as geometrical measures indicating the configuration and morphology of the
urban pattern (Frenkel 2004), however in some groups the homogeneity may disappear. For
example in the case of industrial areas, the variability of dimensions and isolation could be
quite sensitive. It means that, the values of standard deviation can make our model working
better, mostly using together the average and the standard deviation.
Figure 7 it is an example of the behaviour of the indices, across a section of Barcelona,
in relation with the distance from the centre of the city towards the periphery, while the red
zone can be assumed like the most compact urban textures, based on our indicators.

Figure.7 The trend of Density, Efficiency, and Complexity in a section of Barcelona

8. CONCLUSIONS

The aim of this work was to propose a methodology to investigate the complexity of
the actual city, ie its urban morphology, considering that a city is not only design, economic
movement or migration of people, but rather an outcome of their interaction. A proper
definition of urban structures, such as inner the city area or industrial sites, could support on
one hand map reading and on the other hand initial decision making processes in planning
(Steiniger et al. 2007)
 The goal of the experiment was to systematically investigate the complexity of urban
morphology by applying automatic classification of different typologies, which follow the
dynamics of the urban ambits, making it possible to assign standardized values to different
structures. It is really important, for example, to be able to calculate values of density or
compactness in the cities, and to evaluate whether it is a compact model or a dispersal model.
The methodology proposed, could also be implemented using high resolution remote
sensing imagery of urban areas, to automatically identify typologies of urban texture with the
corresponding values of compactness. This could be useful in cities in developing countries to
identify different types of settlements with differing demands in the provision of basic
services.
This result gives the possibility to assign values of compactness to those
neighbourhoods which we defined in the section 4, based on theoretical background, in order
to standardize the definition of urban homogeneous areas from the morphological point of
view.

ACKNOWLEDGEMENT

The authors of this paper gratefully acknowledge the research funding provided by the
Spanish Ministry of Education and Science (SEJ2006-09630), the Spanish Ministry of
Science and Innovation (CSO2009-09057), the Spanish Ministry of Development (E08/08),
and the Spanish Ministry of Housing. Acknowledgements are also due to the European Union
through the INTERREG IIIB Programme (South Western Europe).
For technical support the authors strongly acknowledge Sjoerd de Vos and Marco
Bontje, staff members at Amsterdam Institute for Social Science Research of the University
of Amsterdam, and Montserrat Moix, Carlos Marmolejo, Jorge Cerda, Bahaaeddin Al
Haddad, Malcolm Burns and Alejandro Marambio, staff members at Centre of Land Policy
and Valuations (CPSV) of the Technical University of Catalonia (UPC) (Barcelona TECH).
BIBLIOGRAPHY

Adolphe L. 2001. A simplified model of urban morphology: application to an analysis of the

environmental performance of cities. Environment and Planning B: Planning and
Design 28: 183-200
Anas A., Arnott R., Small K.A. 1988. Urban spatial structure. Journal of Economic
Literature 36: 1426-1463
Arentze T.A., Borgers A.W.J., Timmermans H.J.P. 1994. Multistop-based measurement of
accessibility in a GIS environment. International Journal of Geographical Information
Systems 8: 343-356
Bader M. 1999. Building Squareness. In Specifications of Internal Measures. AGENT Report
D C2: 18
Barr S. L., Barnsley M. J., Steel A. 2004. On the separability of urban land-use categories
in fine spatial scale land-cover data using structural pattern recognition. Environment
and Planning B: Planning and Design 31: 397-418
Batty M. 1998. Urban evolution on the desktop: simulation with the use of extended cellular
automata. Environment and Planning A 30: 1943-1967
Batty M., Longley P.A. 1994. Fractal Cities (Academic Press, London)
Benguigui L., Czamanski D., Marinov M. 2001. The dynamics of urban morphology: the
case of Petah Tikvha. Environment and Planning B: Planning and Design 28: 447-460
Boffet A. 2001. Methode de creation d’information multi-niveaux: pour la generalisation
cartographique de l’urban. Ph.D. thesis, Université de Marne-la-Vallée
Frenkel A. 2004. Land-use patterns in the classification of cities: the Israeli Case.
Environment and Planning B: Planning and Design 31: 711-730
Hillier B., Penn A., Hanson J., Grajewski T., Xu J. 1993. Natural movement: or,
configuration and attraction in urban pedestrian movement. Environment and
Planning B: Planning and Design 20: 29-66
Krafta R. 1994. Modelling intraurban configurational development. Environment and
Planning B: Planning and Design 21: 67-82
Le Gléau J-P., Pumain D., and Saint-Julien T. 1997. Towns of Europe: to each country its
definition. INSEE STUDIES No. 6. Originally published in French as: Villes
d’Europe: à chaque pays sa définition. Économie et Statistique. 1996, 4/5
Longley P., Mesev V. 1997. Beyond analogue models: space filling and density
measurements of an urban settlement. Papers in Regional Sciences 76: 409-427
MacEachren A. M. 1985. Compactness of geographic shape: comparison and evaluation of
measures. Geografiska Annaler 67: 53-67
McGarigal K., Cushman S. A., Neel M. C., and Ene E. 2002. FRAGSTATS: Spatial Pattern
Analysis Program for Categorical Maps. Computer software program produced by the
authors at the University of Massachusetts, Amherst. Available at the following web
site: http://www.umass.edu/landeco/research/fragstats/fragstats.html
Mattogno C. 2007. Urban forms and metropolitan spaces. Paper read at “Forme nella
città/Spazi nella metropoli”. 2-3 aprile 2007, at Università degli Studi di Roma
Sapienza. Facoltà di Ingegneria. Roma.
Rabie J. 1991. Towards the simulation ofurbanmorphology. EnvironmentandPlanning
B:Planning and Design 18: 57-70
Steiniger S., Lange T., Burghardt D., Weibel R. 2007. An Approach for the Classification
of Urban Building Structures Based on Discriminant Analysis Techniques.
Transactions in GIS 12(1): 31-59