Science Direct - Landscape and Urban Planning - Modelling Dynamic Spatial Processes - Simulation of Urban Future Scenarios Throug

ScienceDirect - Landscape and Urban Planning : Modelling dynamic spa...
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V91...
Hub
ScienceDirect
Scopus
SciTopics
Applications
Register
Login
Go to SciVal Suite
Home
Articles Images
Browse
Search
All fields
My settings
My alerts
Author Volume Issue Page
Help
Advanced search Search tips
Journal/Book title
Find more full-text articles: Your search for "Modelling dynamic spatial simulation of urban future scenarios through cellular automata " would return 391 results on ScienceDirect. View Results
Font Size:
PDF (288 K)
Export citation
E-mail article
Highlight keywords on
Related Articles
Stochastic cellular automata modeling of urban land use... Computers, Environment and Urban Systems Modeling urban dynamics through GIS-based cellular auto... Computers, Environment and Urban Systems Errors and uncertainties in urban cellular automata Computers, Environment and Urban Systems Simulating complex urban development using kernel-based... Ecological Modelling Cost-benefit analysis and the use of urban land for tra... Journal of Urban Economics View more related articles
Article
Figures/Tables (6)
References (38)
Thumbnails | Full-Size images
Landscape and Urban Planning Volume 64, Issue 3, 15 July 2003, Pages 145-160
doi:10.1016/S0169-2046(02)00218-9 | How to Cite or Link Using DOI Copyright 2002 Elsevier Science B.V. All rights reserved.
Permissions & Reprints
Modelling dynamic spatial processes: simulation of urban future scenarios through cellular automata
Jos I. Barredo
[Author vitae] ,
, Marjo Kasanko, Niall McCormick and Carlo Lavalle
My Applications
Add Table Download
European Commission, Joint Research Centre, Institute for Environment and Sustainability, CCR-TP 261, 21020, Ispra (VA), Italy Received 15 October 2002; accepted 22 October 2002. ; Available online 2 December 2002.
Abstract
One of the most potentially useful applications of cellular automata (CA) from the point of view of spatial planning is their use in simulations of urban growth at local and regional level. Urban simulations are of particular interest to urban and regional planners since the future impacts of actions and policies are critically important. However, urban growth processes are usually difficult to simulate. This paper addresses from a theoretical point of view the question of why to use CA for urban scenario generation. In the first part of the paper, complexity as well as other properties of digital cities are analysed. The role of the urban land use allocation factors is also studied in order to propose a bottom-up approach which integrates the land use factors with the dynamic approach of the CA for modelling future urban land use scenarios. The second part of the paper presents an application of an urban CA in the city of Dublin. A simulation for 30 years has been produced using a CA software prototype. The results of the model have been tested by means of the fractal dimension and the comparison matrix methods. The simulation results are realistic and relatively accurate confirming the effectiveness of the proposed urban CA approach. Author Keywords: Cellular automata; Land use dynamics; Urban and regional planning; Scenario simulation; GIS Hide Applications
Cited by (75)
Landscape metrics in the analysis of urban land use pat... Landscape and Urban Planning A technique for rapidly forecasting regional urban grow... Environment and Planning B: Planning and Design Modelling the future coastal zone urban development as ... Landscape and Urban Planning View details of all 75 citing articles in Scopus
Provided by Scopus
Article Outline
1. Introduction 1.1. Characteristics and factors of the urban land use dynamics 1.2. How urban land use factors work: an approach
Related reference work articles e.g. encyclopedias

2. Modelling dynamic spatial systems 2.1. A bottom-up approach for urban land use simulation 3. Methods 3.1. The cell space 3.2. The cell neighbourhood 3.3. The cell states 3.4. The neighbourhood effect 3.5. The transition rules 3.6. Land use demands 4. Results and discussion 4.1. Model results testing 4.1.1. Assessment of accuracy using the radial dimension 4.1.2. Assessment of accuracy using comparison matrices 5. Concluding remarks
View Record in Scopus Cellular Automata International Encyclopedia of Human Geography Spatially Explicit Analysis in Environmental Studies International Encyclopedia of the Social & Behavioral S... Urban Modeling International Encyclopedia of Human Geography COMPUTER SIMULATION MODELING Encyclopedia of Archaeology Computational Human Geography International Encyclopedia of Human Geography More related reference work articles
1 sur 12
08/03/2011 08:38

Acknowledgements References Vitae
1. Introduction
Cities can be understood as complex systems considering their intrinsic characteristics of emergence, self-organising, self-similarity and non-linear behaviour of land use dynamics. The use of tools designed for systems that show the aforementioned characteristics will help us to get a better knowledge of the drivers behind urban land use dynamics. Furthermore, these tools can support the development of models for urban land use scenario generation. In the last few years, cellular automata (CA) have gained popularity as modelling tools for urban process simulation. Since the pioneering work of Tobler (1979), several approaches have been proposed for modifying standard CA in order to make them suitable for urban simulation ( White and Engelen, 1993a; White and Engelen, 2000; Itami, 1994; White et al., 1997; White et al., 1999; Clarke and Gaydos, 1998; Semboloni, 2000 and Sui and Zeng, 2001). Researchers can already answer the question of how to build urban CA, or how to modify standard CA in order to make it suitable for urban simulation. However, why to build urban CA is an issue which is still to be satisfactorily addressed (Torrens and OSullivan, 2001). The first part of this paper addresses the question of why. The drivers of urban dynamics are analysed from a theoretical point of view in order to propose an approach for the modelling of urban land use dynamics. In the second part of the paper, results of an urban simulation for the city of Dublin (Ireland) are reported. The CA has been designed and developed in the framework of the ECs Directorate General Joint Research Centre (DG JRC) EUROLANDSCAPE/MOLAND Project (Monitoring Land Cover/Use Dynamics) in collaboration with the Research Institute of Knowledge Systems (RIKS) (White et al., 1999 and Lavalle et al., 2001; MOLAND web-site: http://moland.jrc.it). The aim of this model is to predict future land use development under existing spatial plans and policies, and to compare alternative planning and policy scenarios in terms of their effects on future land use development.
1.1. Characteristics and factors of the urban land use dynamics

Cities are among the most complex structures created by human societies. Furthermore, cities may be different from several points of view, but despite these differences, cities have some characteristics that make them similar. Dynamism and growth are two of the elements which characterise most urban areas. However, modelling dynamism and growth may in some instances be a difficult or almost intractable task without tools which embrace their complexity. Cities are characterised by complex patterns of land use. Indeed, cities can be defined as complex objects from the point of view of their intricate mix of urban activities (White and Engelen, 1993a). Urban areas represent a mix of related units, but the degree and nature of the relationships are sometimes hard to establish. In addition, cities complexity can be seen from two points of view: the complexity represents the information-rich nature of the system, and the complexity is necessary for the successful functioning of the city as system ( White et al., 1997). Considering the previous ideas about cities complexity, two questions arise. Firstly, considering that the complexity of urban land use patterns has its own logic, should patterns of land use respond to certain measurable factors? Secondly, how do these factors contribute to the dynamics of the land use in cities? The answer of these questions may help us to understand how to model urban dynamics. Everything is related to everything else, but near things are more related than distant things. The meaning of Toblers (1970, p. 236) first law of geography is of central significance for understanding urban dynamics. Toblers assertion offers the basis for some applications in spatial analysis in which the neighbourhood space of a feature, even beyond the adjacent space, can influence the feature as a function of distance. For example, a new industrial area in a city can influence not only the areas in its immediate neighbourhood but in a buffer around it in a distance decay way. On the other hand, the place where the industrial area is situated, before the creation of the industrial area, has also been influenced by a buffer with several land uses which have attracted or repulsed other land use types in a distance decay way. Nevertheless, it does not seem that Toblers law alone is enough to understand urban dynamics, some more factors would have to be taken into account. Urban land use dynamics are the direct consequence of the action of individuals, public and private corporations acting simultaneously in time over the urban space. As a consequence, cities are the spatial result in time of all these influences, which continuously contribute to shape the city. If we consider all these influences acting simultaneously the messiness of cities is not surprising. From a practical point of view, several land use allocation factors have been identified for urban activities in the science of spatial decision-making (see Eastman et al., 1993; Voogd, 1983 and Carver, 1991). Five groups of factors can be identified: environmental characteristics; local-scale neighbourhood characteristics; spatial characteristics of the cities (i.e. accessibility); urban and regional planning policies; factors related to individual preferences, level of economic development, socio-economic and political systems. The first group is related to environmental characteristics. It may be represented as constraints for urban growth. For example, slopes, prone areas to natural hazards and natural barriers belong to the first group. It should be noted that the second factor is related with Toblers first law of geography. It can be defined as the present and past land use patterns and their dynamics. Land use patterns usually represent the strongest influence for the dynamics of land use. Distance from new features to existing land uses and the type of these land uses drive the urban dynamics at local scale.
2 sur 12
08/03/2011 08:38
It is logical to think that new residential areas usually grow near or adjacent to existent residential areas. However, they are influenced by other land uses. For example, in this case, the industrial land use could represent a repulsive factor. As a result a sort of equilibrium is reached between all actual land uses and their dynamics in a defined neighbourhood. The third group of factors is related to the spatial characteristics of cities. Factors such as distance to the centre, accessibility, flows or transport networks, are included in this group. For example, new links in the road network might contribute enormously to urban dynamics as an attractor for urban land uses. The fourth group is related to urban and regional planning policies. From a practical point of view, this group is represented by land use zoning status. Through land use zoning plans the city is regulated to be occupied by land uses in space and time. The fifth group comprises factors related to individual preferences, level of economic development and socio-economic and political system. These are the most complicated to understand and model. This group of factors is also related to human decision-making processes, which in most cases are qualitative, evolve in time and can be intransitive and therefore difficult or almost impossible to predict. For example, a new residential area could be located in a place because it is more beautiful than other places. Usually human decision-making processes include some level of unpredictability. Couclelis (1988) defined human systems as terribly complex. From a practical point of view the related complexity of human systems could be modelled as some degree of stochasticity in a probabilistic schema. Therefore, it can be considered as a stochastic factor in urban dynamics modelling. The problem arises in how it can be defined and calibrated. The sum of all the factors which participate in urban dynamics, plus the human decision component, generates a complex dynamic system whose behaviour is influenced by some degree of stochasticity. Where and when some features will change in a city is a spatio-temporal multi-factor process which necessarily includes some stochastic degree.
1.2. How urban land use factors work: an approach

Empirically, White et al. (1997) have shown that the transportation network and land use suitabilities are the determinant factors of the visual urban form. In addition, land use zoning status is also a factor which participates in the land use dynamics in a city, since it defines the legal regulations for future land uses. The factors so far used in urban CA correspond with the factors usually referred to by spatial decision-making studies for land use suitability and allocation (see Eastman et al., 1993 and Voogd, 1983). Here, suitability is understood as a linear deterministic function of several factors which participate in the evaluation of each point of the territory. At local scale, the process of urban dynamics can be understood in three main phases. From the set of the available land in a city represented by the bigger cube in Fig. 1, a set of factors behaving in a linear deterministic way produce a subset of areas prone to be occupied by some land use. In this phase, the factors are not very dynamic and remain stable for some time period until some external action modifies it. For example, the creation of a new motorway will modify the accessibility parameter. Until this step the process of urban dynamics can be modelled using a linear deterministic and more or less static function. In Fig. 1 the subset delimited by the smaller cube contains those places in a city most likely to be urbanised. These places show simultaneously the best accessibility, the best suitability and the right land use zoning status.
Full-size image (5K) Fig. 1. Three-dimensional representation of the factors which actuate in urban land use dynamics in phase 1.
In the second phase, another group of urban land use factors come into play. Those related with the land use dynamics at local scale. These factors initiate a non-linear dynamic process in which the current land use pattern and the local-level interactions between land uses combine to create the distribution of new built-up areas and changes from one urban land use to another. It is in this phase where the system actuates in a very dynamic way using as input the subset of areas generated in the first phase. The land uses interact at local scale as non-linear feedbacks producing a dynamic system. Unlike in the first phase, in the second one the factors are dynamic in time and space, iterative, and show a non-linear behaviour. Each change in urban land use affects future land uses at local scale, changing the local equilibrium every time a land use change takes place. All this makes the process very dynamic and interactive. Finally, the third phase gives the system its stochastic character. Cities as most social and economic processes show some degree of stochasticity. Because of the stochastic nature of the system some places that have been highly ranked in the first phase and established to be likely occupied by some land use in the second phase may be discarded or can be occupied by a less proper land use due to human-related decisions. To understand these three phases are useful as a rationale in order to analyse the process of urban land use dynamics. However, one might reasonably assume that in real situations the three phases probably work simultaneously, putting together the elements of the linear deterministic factors of the first phase, with the non-linear dynamics of local-scale factors of the second, and with the stochastic nature of social processes. All these factors working together produce what is known as a complex system. Complex systems are characterised by collective properties which define the behaviour of the system as a whole. However, much of the behaviour of the constituent parts can be different from the whole
3 sur 12
08/03/2011 08:38
and sometimes partially unknown (Meiss, 2000). On the other hand, self-organisation is a characteristic of complex systems like cities. Self-organisation in cities is the result of local-scale iterations which produce partially ordered large-scale patterns ( Torrens and OSullivan, 2001). Such behaviour deals with systems whose structure is emergent. In the proposed schema of urban land uses dynamics self-organisation is highly dependent of the second phase. In the second phase, local-scale interactions actuate over the existent patterns in a dynamic non-linear sequence generating new patterns that participate in the next iterations of the process. From a practical point of view, the process of urban dynamics can be defined as an iterative probabilistic system (White et al., 1999) in which the probability (p) that a place (i) in a city is occupied by a land use (K) in a time (t), is a function of the concerned factors measured for that land use: suitability (S), accessibility (A), land use zoning status (Z), neighbourhood influence (N) and a stochastic perturbation (v):
t
(1) Considering this approach, the probability that an area changes its land use is a function of the aforementioned factors working together in time, plus a stochastic degree. In this schema the neighbourhood factor (N) makes the system works like a non-linear system. Their dynamism and interactive behaviour can be understood as the basis of the land use dynamics.
piK=f(tSiK,tAiK,tZiK,tNiK,v)
2. Modelling dynamic spatial systems

Issues related to the complexity of cities have been extensively addressed (Batty and Longley, 1994; White et al., 1997; Torrens, 2000; Batty, 2000 and Torrens and OSullivan, 2001). Complexity deals with processes, phenomena, featuresin a word, systemscharacterised by collective properties associated with the system as a whole, that are different from the behaviour characteristics of the constituent parts. However, the degree and nature of the relationships between the parts of the system are sometimes imperfectly known. But it is from these relationships that the systems tend to organise themselves into patterns, through the process of self-organisation. Although some attempts in defining complexity have been cited by Couclelis (1988), informationtheoretical complexity, the one closest linked with urban dynamics, does not seem to be yet well defined. However, there are some elements that can help us to understand it. The idea of complexity deals with systems whose structure is emergent. This is a property of systems such as cities, economies and ecologies. All these systems evolve their structure from a set of rules and the structure itself, producing new and surprising elements (Batty, 2000). As has been pointed out by Batty (2000) taken from von Bertalanffy (1972), the whole is greater than the sum of its parts. Studying cities as complex systems the critical point is that the sum of the parts corresponds to urban morphology whilst the whole is represented by urban patterns dynamics. From this perspective, dynamics becomes the keyword in order to study cities from a holistic perspective. Within this approach urban models must be able, among other capabilities, to deal with feedbacks to be able to model the self-organising properties of emergent systems. In complex emergent systems, like cities, a small number of rules applied at local level are capable of generating surprising complexity in aggregate form (Torrens, 2000). What is likely to happen when the urban land use factors interact simultaneously in time over the urban land use patterns? The ordered large-scale patterns are developed from local-scale interactions. Depending on the nature of these local-scale interactions, the large-scale ordered patterns will take their form, structure, shape and/or behaviour. It is surprising that these aggregate patterns emerge apparently independently of the dynamics driving the individual components of the system ( Torrens and OSullivan, 2001). Complex emergent systems are frequent in nature. Simple processes produce complex structures without chance. Non-linearity and iterativity are the code keys for the development of very rich structures like the human brain (see Duadi, 1986). Clearly, there is a confrontation between some usually open ecological and social systems and the second law of thermodynamics. This law states that closed systems progressively evolve to a state of maximum disorder in which the entropy of the system grows in time. Conversely, some social open systems like cities tend to be ordered, at least in the sense of urban land use patterns. This kind of systems have been defined as self-organising. Self-organisation in dynamic systems establishes the tendency for system structures to develop ordered patterns, often on a large scale (Torrens, 2000 and Krugman, 1996). The ordered properties of cities have been studied by White and Engelen (1993a), White et al. (1997) and Frankhauser, 1991 and Frankhauser, 1994 by using fractal dimensionor radial dimension ( Frankhauser, 1991 and Frankhauser, 1994)as a measure of order. They define cities as bifractal objects based on the obtained fractal dimensions into two zones of the cities. The inner zone is more organised and the outer zone less organised or still evolving. In the inner zone, the transformation process has reached or it is near to the maximum level of organisation accessible to the system. It means that the system has reached a sort of equilibrium and the urban pattern is relatively stable. In the outer zone, the process of urbanisation still continues, thus it shows a less degree of organisation than the inner zone. Here the system is still evolving. The fact that cities do have fractal structure (Batty, 1991; Batty and Longley, 1994; Frankhauser, 1991; Frankhauser, 1994; Frankhauser and Sadler, 1991; White and Engelen, 1993b and White et al., 1997) introduces the concept of self-similarity. By definition fractal objects are self-similar. New areas in cities show patterns which often are indistinguishable from the previous patterns and from the whole of the city, moreover the structure of the pattern is independent from scale ( Torrens, 2000 and Wolfram, 1994). As was already stated, these patterns are fractal structures and can be characterised through fractal dimensions ( Torrens, 2000). As has been shown by Torrens (2000) and Torrens and OSullivan (2001), cities exhibit several of the characteristics of complexity: fractal dimensionality, self-similarity, self-organisation and emergence. As a result, cities can be seen as fine examples of complex emergent systems ( Torrens and OSullivan, 2001). As with most complex systems cities are non-linear systems, therefore their
4 sur 12
08/03/2011 08:38
dynamics and the rules for pattern generation and evolution require well suited tools for their modelling and understanding. At this point a question emerges: why is it important to study complexity in order to understand cities dynamics? From a practical point of view, as is argued by Couclelis (1988), understanding complexity can result in a useful and practical way to predict complex systems. In addition, complexity and the tools which deal with complex emergent systems can help us to understand and model the future of cities. In order to generate these future scenarios, urban land use patterns dynamics can be simulated taking into account the initial state of the system, the participating factors for land use dynamics, and the rules that produce the dynamics that drive the evolution of actual cities. Modelling city dynamics deals with a problem which shows non-linear behaviour, self-organising and self-similarity properties. It is also very sensible to initial conditions and rules and to determine causal relations becomes a hard task since the system presents a degree of stochasticity. In addition, as have been pointed out by Couclelis (1997), there are also laws of complexity at work within the urban and regional land use dynamics, which makes the prediction of the system enormously difficult. We can now point out a new question. What are the tools designed for dealing with this kind of complex spatial dynamic systems?
2.1. A bottom-up approach for urban land use simulation

CA are a joint product of the science of complexity and the computational revolution (Couclelis, 1986). Although CA are models which deal with complex systems, CA have been defined as very simple dynamic spatial systems. In CA the state of each cell in an array depends on the previous state of the cells within a neighbourhood, according to a set of transition rules ( White et al., 1999). Despite their simplicity some classes of CA are capable of universal computation ( Wolfram, 1984a and Wolfram, 1988), which implies that some types of CA can mimic or reproduce behaviours with high level of complexity, such as of physical, biological or social complex systems. CA have a remarkable potential for modelling complex spatio-temporal processes, and very simple CA can produce surprisingly complex forms through a set of basic rules. An understanding of CA leads to the conclusion that the description of complex systems need not be themselves complex, let alone complicated ( May, 1976 and Couclelis, 1988). Many processes in nature and in social systems can be approached trough non-linear differential equations. In these kinds of equations a magnitude (X) in a time (t+1) is the consequence of the magnitude in the preceding time (t). This configuration defines a basic non-linear differential equation: Xt+1=f(Xt)
(2) These equations, although fully deterministic, can produce a very dynamic behaviour, from stable points and limit cycles to chaotic regimes (strange attractors) (Wolfram, 1984a and Wolfram, 1984b; May, 1976). Furthermore, the behaviour of non-linear differential equations may be indistinguishable from the one produced by a random process. CA have been considered as spatial idealisations of partially differential equations with discrete space and time, thus it is not strange that CA show behaviours analogous to non-linear ordinary differential equations (Wolfram, 1984a and Wolfram, 1984b). From this point of view it is not surprising that CA are capable of producing and simulating complex spatial processes showing non-linear dynamics such as some socio-spatial processes (i.e. spatial segregation of socio-economic groups), and moreover CA produce spatial patterns that show chaotic behaviour in the sense of irregular dynamics in a deterministic system. In these kinds of systems the behaviour depends on its own internal logic, and not for the reason that the system is stochastic per se. An elementary CA is comprised by several elements (Couclelis, 1988 and Torrens, 2000): the space represented by an array of cells, each residing in a state at any one time; a discrete number of states (qualities referred to the cells); the neighbourhood template; transition functions, which define what the state of any given cell is going to be in the next time period based in the present state of the cells itself and that of its neighbouring cells. The simple mechanism for defining the next state of a cell based on its actual state and the actual state of the cells in the neighbourhood, made CA very simple mechanism with the following main characteristic: spatial-interactivity. It represents the key to some spatial dynamic processes like urban land use evolution (Couclelis, 1997). We can assume that CA are systems capable of modelling complex emergent systems, but what do CA have to do with cities? It should be noted that urban land use dynamics can be defined as a complex process, and CA can produce or mimic complex spatial processes. However, this is not the only link between them. Some characteristics of cities such as fractal dimensionality, self-similarity across scales, self-organisation and emergence ( Torrens, 2000; White et al., 1997; Batty and Longley, 1994; Allen, 1997 and Portugali, 2000), can also be generated by CA ( Torrens, 2000; Wolfram, 1994; Batty and Longley, 1994 and Batty, 1997). The question arises as to how to get in use all the modelling potential of CA, in order to model urban land use evolution in a realistic way. Two major requirements have been stated in order to produce useful urban land use models using CA: interactivity and realism. Serious qualitative forecasting, through urban and regional modelling, should respond to these two aspects in order to differentiate gameplaying from serious prediction ( Couclelis, 1997).
3. Methods
From the described approach an urban CA model has been designed and developed for modelling urban and regional systems through constrained CA-based simulation tools (White et al., 1999). It has the following specificities.
3.1. The cell space

The digital space in the CA consists of a rectangular grid of square cells each representing an area of 100 m100 m. This is the same size as the minimum area mapped in urban areas in the land use
5 sur 12
08/03/2011 08:38
datasets. The cell space is assumed not to be homogeneous. Each cell is characterised by a vector of suitabilities, one suitability for each land use taking part in the dynamics. The suitabilities are defined as a weighted linear sum of a series of physical, environmental and institutional factors characterising each cell. They are normalised to values in the range of 01, and represent the inherent capacity of a cell to support a particular activity or land use. The suitability maps remain constant during the simulation unless the user interrupts the run and edits them manually. Secondly, each cell is associated with a vector of accessibility factors, again one for each land use type. This factor represents the importance of access to transportation networks for various land uses for each cell, again one for each land use type. Some activities, like commerce, require better accessibility than others, such as residential discontinuous sparse. Accessibilities are calculated as a function of distance from the cell to the nearest point in the transport network as follows:
(3) In Eq. (3), tAr,K,x,y is the accessibility of the cell (x, y) to infrastructure element r for land use K at time t; Dr is the distance between cell (x, y) and the nearest cell (x, y) on infrastructure element r; and ar,K is a calibrated distance decay accessibility coefficient expressing the importance of good access to infrastructure element r for land use K. Finally, each cell is associated with a set of codes representing its zoning status for various land uses, and for various periods. Due to the combined effect of suitabilities, accessibilities, neighbourhood effect, and zoning, every cell is essentially unique in its qualities with respect to possible land uses. It is on this highly differentiated digital space where the dynamics of the cellular automata take place.
3.2. The cell neighbourhood

In this urban cellular automaton, the neighbourhood space is defined as a circular region around the central cell with a radius of eight cells. The neighbourhood thus contains 172 cells that are arranged in 30 discrete distance zones. The neighbourhood radius represents 0.8 km; this distance delimits an area that can be defined as the influence area for urban land uses. That is similar to what residents of a city commonly perceive to be their neighbourhood, and thus should be sufficient to allow local-scale spatial processes to be captured in the CA transition rules.
3.3. The cell states

The model uses 22 cell states. Since our purpose is to model urban land uses the most detailed level of classes is used for these, while agricultural and natural land uses are represented by more highly aggregated classes. Six of the classes: road and rail networks, airports, mineral extraction sites, dump sites, artificial non-agricultural vegetated areas, and water bodies, represent fixed features in the model, that is, features which are assumed not to change and which therefore do not participate in the dynamics. They do, however, affect the dynamics of the active land uses, since in the neighbourhood effect they may represent an attractive or repulsive effect. The passive functions are: arable land, pastures, forests, shrub, sparsely vegetated areas, and wetlands. That is, functions that participate in the land use dynamics, but the dynamics are not driven by an exogenous demand for land; they appear or disappear in response to land being taken or abandoned by the active functions. The active functions are the nine urban land uses which are forced by demands for land generated exogenously to the cellular automaton in response to the growth of the urban area: residential continuous dense, residential medium dense, residential discontinuous, residential discontinuous sparse, industrial, commercial, public and private services, port areas, and abandoned land. Construction site represents a transitional state between one function and another.
3.4. The neighbourhood effect

In the urban CA the neighbourhood space effect is calculated for each of the 15 function states (passive and active) for each cell. Essentially, this effect defines the state to which the cell could be converted or the permanence in their original state. The effect represents the attraction (positive) and repulsion (negative) effects of the various land uses and land covers within the neighbourhood of 172 cells in relation to the central cell. In general, cells that are more distant in the neighbourhood will have a smaller effect in the evaluated central cell. On the other hand, a positive weight of a cell on itself (zero-distance weight) represents an inertia effect due to the implicit and monetary costs of changing from one land use to another. Each cell in the neighbourhood receives a calibrated weight according to its state and distance from the central cell. The neighbourhood effect is thus calculated as:
(4) In Eq. (4), tNK,x,y is the contribution of the CA transition rules in the calculation of the transition potential of cell (x, y) for land use K at time t; wK,L,c is the weighting parameter expressing the strength of the interaction between a cell with land use K and a cell with land use L at a distance c in the neighbourhood; tIc,l is the Dirac delta function (inertia effect): tIc,l=1 if cell l at a distance c at time t is in the state L, otherwise tIc,l=0.
3.5. The transition rules

Following the theoretical probabilistic function of Eq. (1), a vector of transition potentials (one potential for each function) is calculated for each cell from the suitabilities, accessibilities, zoning status, and neighbourhood effect, and the deterministic value is then given a stochastic perturbation using a modified extreme value distribution, such that most values are changed very little but a few are changed significantly:
t
PK,x,y=(1+tAr,K,x,y)(1+SK,x,y)(1+tZK,x,y )
(5)
6 sur 12
08/03/2011 08:38

t t
( NK,x,y) v where PK,x,y is the CA transition potential of the cell (x, y) for land use K at time t; tAr,K,x,y the accessibility of the cell (x, y) to infrastructure element r for land use K at time t; SK,x,y the intrinsic suitability of the cell (x, y) for land use K; tZK,x,y the zoning status of the cell (x, y) for land use K at time t; NK,x,y the neighbourhood space effect on the cell (x, y) for land use K at time t; and v is the scalable random perturbation term at time t, it is defined as: v=1+[ln(rand)], where (0<rand<1) is a uniform random variable, and is a parameter that allows the size of the perturbation to be adjusted. Thus, the transition rule works changing each cell to the state for which it has the highest potential, subject to the constraint that the number of cells in each state must be equal to the number demanded in that iteration. Cell demands are generated outside the CA. In each iteration, all cells are ranked by their highest potential and cell transitions begin with the highest ranked cell and proceed downwards until a sufficient number of cells of a particular land use has been achieved. Each cell is subject to this transition algorithm at each iteration. However, most of the resulting transitions are from a state to itself, consequently the cell remains on its current state. When a transition to another state does occur, the actual transition to the second state occurs after one iteration (1 year) delay, during this interval the cell is assigned the state construction site.
t t t
3.6. Land use demands

In constrained CA such as this case the land use demands are generated exogenously to the cellular model (White et al., 1997). Demands reflect the growth of a city rather than the local configurational dynamics captured by the urban CA. However, currently an integrated model of regional spatial dynamics which include standard non-spatial models of regional economics and demographics is being developed to generate estimates of activity levels which are then converted via productivity or density relations into demands for cell space in an urban cellular automaton.
4. Results and discussion

A simulation for Dublin has been produced for the period 19681998 using the urban CA model. The simulation begins using the historical datasets for the year 1968, in order to test the simulation results using the reference datasets for the year 1998. In this way, the simulation has been tested by comparison with actual land use datasets. The calibration of the model was carried out over the period 19681998 using the MOLAND land use datasets. The present simulations make no use of suitabilities or zoning, since these data are not yet available. Nevertheless, the preliminary results are a useful demonstration of the extent to which the urban CA can produce realistic simulations using the available data. The built-up area of Dublin has expanded considerably over the past 30 years. Fig. 2a and b show the actual urbanised area in 1968 and 1998, respectively. Fig. 2c shows the transport network used in the accessibility calculations and Fig. 2d shows the simulated built-up area in 1998. Notice that in order to make the maps clearer, the large number of 22 land use classes have been grouped into residential, industrial and other urban built-up areas.
Full-size image (40K) Fig. 2. Dublin maps: (a) urbanised area in 1968; (b) urbanised area in 1998; (c) road network used in the simulation; (d) simulated urbanised area in 1998.
Accurate simulations in the present urban CA depend on several factors. Among the most important of these is the calibration of the weighting parameters of Eq. (4), which define the neighbourhood effect. The weighting parameters are calibrated in order to minimise the differences between the simulated land use map for 1998 and the actual land use map for that year. The schema of weighting parameters for any pair of land uses is based on a rational evaluation of the land use pattern in the city and their evolution. The pattern of weighting parameters for any one pair of land uses is not surprising. For example, residential discontinuous sparse land use is attracted to itself, much more so at close distances, and also, but less strongly, to commerce and to artificial non-agricultural vegetated areas, while it is repelled by industry, especially at close distances (Table 1). From a practical point of view, the calibration of the model is based on an interactive procedure in which each state (active and passive) is calibrated against each of the land uses classes. The weighting parameters are thus assigned for each pair of land uses through interactive windows incorporated in the software prototype. Table 1. The weighting parameters for the neighbourhood effect applied on the residential discontinuous sparse land use
7 sur 12
08/03/2011 08:38
Notice that one matrix of weight has been produced for each of the passive and active states. The weighting assignment is done by verifying visually the spatial effects of the weights in the urban CA prototype. The simulations can be produced in a few seconds using a good PC, offering the possibility to modify the weight values until the results fit with the reference land use map. Once all the functions have been calibrated, the model is re-run several times in order to verify if the land use transitions work in a logical way. Comparison matrices and more structural procedures like the fractal dimension are also used for producing a fine-tuned version of the simulation which gives accurate and realistic results. Another calibration factor is the random perturbation. This was set at =0.6 following a trial and error approach. This value reproduces the radial (fractal) dimension of the urbanised area measured for the actual land use map. In this case, the value was also fine-tuned to generate a sufficient number of new seed cells of various land uses in new locations such as rural areas, which will subsequently develop into, for example, new industrial, commercial or residential areas. The random perturbation parameter produces also non-continuous (i.e. leap-frog) growth of urban land uses based on a stochastic function. If the perturbation is increased, then the stochasticity of the simulation increases too. The problem remains, however, as to how to make the stochasticity of both systemsthe simulation and the actual citymatch.
4.1. Model results testing

Although more research is needed in order to generate new appropriate testing methods for urban CA models (White et al., 1999) several approaches are feasible. The methods that we considered are based on the comparison of the simulation map with the actual city. In this case the comparison has been done using the reference land use dataset of MOLAND for Dublin (year 1998) against the obtained simulation. The urban CA model was tested in three ways: (1) a visual comparison of the maps; (2) comparison of land use pattern distribution through relatively abstract measures like fractal dimension; and (3) a quantitative evaluation of the degree of coincidence between the two land use maps using the comparison matrices method. The first, mainly intuitive, approach for testing the simulation results is a visual comparison between the simulated land use map and actual 1998 land use map. The main result of this analysis is that there is a good visual similarity between the maps, and moreover, the simulation actually looks like a city. However, one noticeable difference is that the simulation shows a somewhat less fragmented urban land use pattern than the actual city. On the other hand, some clusters of land use, mainly in the peripheral areas, show a different shape in comparison with the actual city map. Although the cellular model simulates reasonably well the pattern distribution of land uses of the actual city, it is noticeable that the urban CA is not able to reproduce, for example, a new residential area in a place where there were originally pastures and where no residential areas were in the neighbouring zones, unless the random perturbation (i.e. stochastic perturbation) puts a seed cell in the same place, which is not very likely. However, a good simulation can be produced if the place is near to some link in the transport network or, if the datasets are available, the development takes place in a place with high suitability and the right zoning conditions. In this case, the place for the new residential area could be predicted. Yet it is important that the land use pattern distribution in the whole area is similar to or resembles the actual city pattern, what actually happened in this simulation. 4.1.1. Assessment of accuracy using the radial dimension The fractal or radial dimension has already been used for studying the fractal structure of urban areas (Frankhauser, 1991; Frankhauser, 1994 and Frankhauser and Sadler, 1991) and even for testing urban CA models ( White and Engelen, 1993a and White et al., 1997). Formally, the radial dimension can be interpreted as the slope of the relationship between the size of an object, as measured by the number of its constituent cells, and its diameter. This is a practical measure of the radial dimension for irregular or stochastic spatial distributions ( White and Engelen, 1993a) or patterns as in the case of urban areas. In this study, the fractal dimension was obtained by calculating
8 sur 12
08/03/2011 08:38
the total area occupied by individual land uses within a given radius from the centre of the city, using a set of increasing radii, in 1000 m increments, which resulted in 21 radius zones for each of the cities. The slope of the simple regression line in the arearadius plot for each land use corresponds to the radial dimension of that land use (Fig. 3).
Full-size image (11K) Fig. 3. Arearadius plots for Dublin ( ) and simulation (): (a) industrial; (b) commercial; (c) residential; (d) urbanised area.
The radial dimension is a measure of how sparse a Euclidean bi-dimensional object is. For example, if some land use increases radially its number of cells in a relatively low rate from the centre of the city, the land use becomes relatively sparser on each radius unit. Thus, their slope in the arearadius plot and hence their radial dimension will be relatively low. On the other hand, if the growth rate is relatively high, there will be a high number of cells occupied on each radius unit. This makes that the slope become relatively high and consequently the radial dimension increases. In this case, the land use will be relatively less sparse. The actual city and the simulation display a bifractal structure in the evaluated four land use classes: commercial, industrial, residential (the sum of residential dense, residential medium dense, residential continuous, and residential discontinuous sparse) and urbanised (the sum of residential dense, residential medium dense, residential continuous, residential discontinuous sparse, industrial, commercial, and public and private services). In the arearadius plots in Fig. 3ad, the inner zone (represented by a steep slope) is evident for the four land use classes. The outer zone, on the other hand, is represented by a flatter slope. The border of the two zones is within a radius of 810 km. These results match with the experiments carried out by Frankhauser, 1991 and Frankhauser, 1994 and White and Engelen, 1993a and White and Engelen, 1993b in which several cities of the world have been studied showing similar bifractal structures. From this point of view cities are structured in two zones: the first one is a fully urbanised inner zone in which the urbanisation process has reached a sort of equilibrium. In this area, the urbanisation process is complete and radial dimension values are usually higher than in the outer zone. The outer zone is an area where land use is less intensive and the urbanisation process continues to evolve: there is still a number of vacant areas that can be taken over by urban land uses, and consequently the urban structure is dynamic and the radial dimension is therefore smaller. Despite some slight differences in the arearadius plot and the corresponding simulation, the general agreement of the plots is an evidence of the similarity of the pattern distribution in both maps. The clear bifractal structure is a consequence of the concentric zonation of land uses in the simulation and in the actual city. In addition, the simulation match with the actual city not only in the inner zone, where few cells were available at the beginning of the simulation, but also in the outer zone, where most of changes in land use took place in the 30-year simulation period. Individual analyses of the arearadius plots show that the land uses are distributed bifractally. For example, up to the radius of 10 km the radial dimension for the industrial land use is 2.45, whilst in the outer zone the dimension is 1.61. These values of radial dimension represent a well developed central area in which the industrial land use is structured in a compact way. Inversely, in the outer zone, the industrial land use is sparser revealing a less urbanised area, and likely it is still evolving to become a more compact structure. In Fig. 3 each individual plot reveals the high similarity between the simulation and the actual city. Although the plots are not identical they are very similar. In addition to this, it is also important that the simulation resembles the concentric distribution of land use of the actual city as it is. In general commercial, residential and the urbanised arearadius plots show a similar structure as industrial: a well identified inner zone and the evolving outer zone, showing all of them the referred bifractal structure. The commercial land use in Fig. 3b shows a plateau in the radius 35 km, this may be caused by the vicinity of the sea and other water bodies. In this radius range the simulation overestimates the area occupied by commercial land use in the actual city. Values of radial dimension were calculated considering the overall 21 km radius zone in the real city and in the simulation. These are shown in Table 2. By comparing the radial dimension values between the actual city and the simulation, the similarity between them is striking. The similar values of radial dimension mean that the simulation reproduces accurately the spatial pattern for the evaluated land uses during the simulated period. But how is the urban CA able to reproduce patterns without any long-range iteration procedure? The answer lies in the characteristics of systems with self-organising properties, such as standard CA: in the urban CA the pattern structure is a consequence of the local-level iterations, which produce the global structure of the simulation map. Table 2. Radial dimension for individual land use classes
4.1.2. Assessment of accuracy using comparison matrices
9 sur 12
08/03/2011 08:38
The third approach for testing the model results is to evaluate the degree to which the two maps (the simulation and the actual land use map) are in line with each other, by means of a coincidence matrix and the associated k indices. This approach is useful for identifying the cells that are identical in both maps, considering their state in a cell-by-cell comparison. Although this approach generates quantitative measures of coincidence for the two maps, there are some weaknesses in their implementation in urban CA models (White et al., 1997 and Torrens and OSullivan, 2001). The coincidence matrix procedure is not able to evaluate patterns, since the procedure is based on independent comparisons between pairs of cells, and therefore is unable to take account of patterns or distributions per se. This means that small displacements are identified as discordances, and the same discordance will be stated if the displacement is of 100 cells instead of 1. On the other hand, in land uses with a low number of cells, the k-value will not yield a useful statistical indicator. Regardless of these drawbacks the technique is applied for testing the simulations. The comparison matrix was made using the simulation map and the actual city map for 1998 (Table 3). The matrix has been summarised in order to simplify the extensive comparison matrix produced for the 22 land use classes. The table shows the results only for active states; the remaining land uses were included as other land uses. Total cell numbers have been included only in columns since the number of cells is the same in the simulation and in the actual city for each of the active states. The k-statistics have been also included in the matrices for the active land uses. The last rows of the table show the numbers of cells added for each active land use during the simulated period. Notice that residential medium dense, residential discontinuous and port areas show a low number of added cells in the simulated period, therefore the k-value for these land uses should not be taken into account in the validation process since they do not represent a consistent indicator. Table 3. Coincidence matrix and k-values for the simulation 19681998
Full-size table (16K) Note: the numbers in the coincidence matrix are cell counts. The residential discontinuous sparse land use shows a k-value of 0.73, indicating a good match between the simulation and the actual city. This land use type covers a large area in Dublin, and 71% of the area of the active states. Considering the simulation it is noticeable that 249 cells have been incorrectly assigned to industrial instead of residential discontinuous sparse. This fact is not surprising considering that both land uses compete for open vacant land in the peripheral areas of Dublin due to their very similar allocation criteria. Noticeable errors related to the simulation for Dublin are the 3152 cells incorrectly assigned by the simulation to other land uses instead of residential discontinuous sparse. Most of these cells have been assigned by the simulation to artificial non-agricultural vegetated areas (1450 cells) and pastures (1298). In the first case, artificial non-agricultural vegetated areas are usually integrated in urban areas with residential land uses. Considering that artificial non-agricultural vegetated areas has not been regarded as an active state, the assignment of the residential land use instead is not surprising in view of their very similar locational factors. In the case of the industrial land use, the 259 cells incorrectly assigned as residential discontinuous sparse confirms the aforementioned ideas about the similarity of the locational requirements of both land uses. On the other hand, most of the 692 cells incorrectly assigned to other land uses instead of industrial, have been assigned to pastures (339 cells) and arable land (153 cells). In this case, the zoning datasets would improve substantially the results. The rest of land uses with significant numbers of cells show k-values from 0.72 for commercial, to 0.92 for residential discontinuous, which can be considered as good or very good.
5. Concluding remarks
The results of the simulation for Dublin have demonstrated that the urban CA prototype can provide reasonable representations of the future evolution of cities, even when, as in this study, the amount of supporting data is limited, and the simulated period is quite long (30 years). Clearly, the inclusion of suitabilities and zoning status and a more realistic modelling period of, say, 20 years could yield more reliable results, in a time scenario that is adequate for planning purposes. Despite these constraints, the capacity of the model to reproduce the actual urban form through large-scale patterns maintaining the resolution of the spatial data included in the model is an advantage. This is a property of some CA that are capable to produce fractal objects from self-organisation, universal computation and other types of behaviours pertaining to systems balanced between order and chaos. As opposite to cell-by-cell basis analysis urban CA are intrinsically spatial models. Therefore, these kinds of models are capable to produce dynamic spatial analyses since their analytic basis is not only the cell values on each case. It includes as well the neighbourhood cell states. In addition, the interactivity of the model produces an approach that meets the realism needed for simulating actual
10 sur 12
08/03/2011 08:38
urban processes. On the other hand, the model structure is open. Thus, as the simulation tool is moved closer to end-users such as urban planners who have access to a wealth of data, it will be possible to extend the richness of the modelling framework to provide better estimates of future conditions producing several future scenarios. Thanks to the experimental potential of the model it could be used by planners as a simulation box, in which a number of spatial conditions if then can be tested easily in a realistic way within a known degree of accuracy. In addition, the use of the arearadius plots for comparing the model results may be useful for measuring the spatial consequences of political decisions and in general the land use evolution, since this model is capable to simulate single land use classes individually. The link between the cellular model and GIS is important considering that the spatial data used by the model was developed and managed using a raster-based GIS. On the other hand, some current GIS applications provide a set of operators enough to develop urban CAs: user defined filters, dynamic operations, overlay, reclassification, and a scripting language to put the operations in a logical order. Some researchers have recently proposed ways of building CA functionality into GIS, or conversely, GIS functionality into CA. This model has been developed following the second approach, in which the GIS is used for handling the models output in the form of predictions. This type of model-GIS approach is called loose coupling (Clarke and Gaydos, 1998). Within this approach the possibility of delivering the model to end-users as a stand-alone software is a clear advantage.
Acknowledgements
This research was funded by the European Commission, Directorate General Joint Research Centre, postdoctoral grant #17646-2001-03 P1B30 ISP IT Monitoring sustainable urban and regional development using spatial dynamics models awarded to Jos I. Barredo; and undertaken as part of the ECs Directorate General Joint Research Centre (DG JRC) EUROLANDSCAPE/MOLAND (Monitoring Land Cover/Use Dynamics) project. The cellular automata software has been developed by RIKS in the framework of the contract #14518-1998-11 F1ED ISP NL awarded by the European Commission, Directorate General Joint Research Centre.
References
Allen, P.M., 1997. Cities and Regions as Self-Organizing Systems: Models of Complexity. Gordon and Breach, Amsterdam. Batty, M., 1991. Cities as fractals: simulating growth and form. In: Crilly, A.J., Earnshaw, R.A., Jones, H. (Eds.), Fractals and Chaos. Springer, Berlin, pp. 4369. Batty, M., 1997. Editorial: Urban systems as cellular automata. Environ. Plan. B 24, pp. 159164. Batty, M., 2000. Editorial: Less is more, more is different: complexity, morphology, cities, and emergence. Environ. Plan. B 27, pp. 166168. Batty, M., Longley, P., 1994. Fractal Cities. Academic Press, London. Carver, S.J., 1991. Integrating multi-criteria evaluation with geographical information systems. Int. J. Geogr. Inf. Syst. 5, pp. 321339. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (205) Clarke, K.C. and Gaydos, L., 1998. Loose coupling a cellular automata model and GIS: long-term growth prediction for San Francisco and Washington/Baltimore. Int. J. Geogr. Inf. Sci. 12, pp. 699714. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (259) Couclelis, H., 1986. Artificial intelligence in geography: conjectures on the shape of the things to come. Prof. Geogr. 38, pp. 111. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (6) Couclelis, H., 1988. Of mice and men: what rodent populations can teach us about complex spatial dynamics. Environ. Plan. A 20, pp. 99109. View Record in Scopus | Cited By in Scopus (50) Couclelis, H., 1997. From cellular automata to urban models: new principles for model development and implementation. Environ. Plan. B 24, pp. 165174. Duadi, A., 1986. Julia sets and the Mandelbrot set. In: Peitgen, H.O., Richter, P.H. (Eds.), The Beauty of Fractals, Images of Complex Dynamical Systems. Springer, Berlin. Eastman, J.R., Kyem, P.A., Toledano, J., Jin, W., 1993. GIS and Decision Making. United Nations Institute for Training and Research, UNITAR, Geneva. Frankhauser, P., 1991. Aspects fractals des structures urbaines. Espace Gogr. 1, pp. 4569. Frankhauser, P., 1994. La Fractilit des Structures Urbaines. Economica, Paris. Frankhauser, P., Sadler, R., 1991. Fractal analysis of agglomerations. In: Proceedings of the 2nd International Colloquium of the Sonderforshungsbereich 230: Naturliche Konstruktionen, University of Stuttgart, Stuttgart, pp. 5765. Itami, R.M., 1994. Simulating spatial dynamics: cellular automata theory. Landscape Urban Plan. 30, pp. 2747. Abstract | PDF (5928 K) | View Record in Scopus | Cited By in Scopus (55) Krugman, P., 1996. The Self-Organizing Economy. Blackwell Scientific Publications, Malden, MA. Lavalle, C., Demicheli, L., Turchini, M., Casals-Carrasco, P. and Niederhuber, M., 2001. Monitoring megacities; the MURBANDY/MOLAND approach. Dev. Prac. 11, pp. 350357. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (9) May, R.M., 1976. Simple mathematical models with very complicated dynamics. Nature 261, pp. 459467. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (1240) Meiss, J.D., 2000. Frequently Asked Questions About Nonlinear Science. The Internet: http://amath.colorado.edu/faculty/jdm. Portugali, J., 2000. Self-Organization and the City. Springer, Berlin. Semboloni, F., 2000. The dynamic of an urban cellular automata model in a 3-D spatial pattern. In: Proceedings of the XXI National Conference, Aisre: Regional and Urban Growth in a Global Market, Palermo. Sui, D.Z. and Zeng, H., 2001. Modeling the dynamics of landscape structure in Asias emerging desakota regions: a case study in Shenzhen. Landscape Urban Plan. 53, pp. 3752. Article | PDF (534 K) | View Record in Scopus | Cited By in Scopus (57) Tobler, W., 1970. A computer movie simulating urban growth in the Detroit region. Econ. Geogr. 46, pp. 234240. Full Text via CrossRef Tobler, W., 1979. Cellular geography. In: Gale, S., Olson, G. (Eds.), Philosophy in Geography.
11 sur 12
08/03/2011 08:38

Reidel, Dordrecht, pp. 379386.
Torrens, P.M., 2000. How cellular models of urban systems work. WP-28. Centre for Advanced Spatial Analysis (CASA), University College, London. Torrens, P.M. and OSullivan, D., 2001. Editorial: Cellular automata and urban simulation: where do we go from here. Environ. Plan. B 28, pp. 163168. von Bertalanffy, L., 1972. General System Theory. Penguin, Harmondsworth, Middx. Voogd, H., 1983. Multicriteria Evaluation for Urban and Regional Planning. Pion, London. White, R. and Engelen, G., 1993. Cellular automata and fractal urban form: a cellular modelling approach to the evolution of urban land use patterns. Environ. Plan. A 25, pp. 11751199. View Record in Scopus | Cited By in Scopus (258) White, R., Engelen, G., 1993b. Complex dynamics and fractal urban form. In: Nijkamp, P., Reggiani, A. (Eds.), Nonlinear Evolution of Spatial Economics Systems. Springer, Berlin, pp. 223247. White, R. and Engelen, G., 2000. High-resolution integrated modelling of the spatial dynamics of PDF (354 urban and regional systems. Comput. Environ. Urban Syst. 24, pp. 383400. Article | K) | View Record in Scopus | Cited By in Scopus (133) White, R., Engelen, G. and Uljee, I., 1997. The use of constrained cellular automata for high-resolution modelling of urban land use dynamics. Environ. Plan. B 24, pp. 323343. View Record in Scopus | Cited By in Scopus (157) White, R., Engelen, C., Uljee, I., Lavalle, C., Erlich, D., 1999. Developing an urban land use simulator for European cities. In: Proceedings of the 5th EC-GIS Workshop, Stresa, Italy. Wolfram, S., 1984. Preface. Physica D 10, pp. viixii. View Record in Scopus | Cited By in Scopus (13) Wolfram, S., 1984. Universality and complexity in cellular automata. Physica D 10, pp. 135. Abstract | PDF (3706 K) | View Record in Scopus | Cited By in Scopus (141) Wolfram, S., 1988. Complex systems theory. In: Emerging Syntheses in Science, Proceedings of the Founding Workshops of the Santa Fe Institute. Addison-Wesley, Reading, MA, pp. 183189. Wolfram, S., 1994. Cellular Automata and Complexity. Addison-Wesley, Reading, MA. Corresponding author. Tel.: +39-332-789429; fax: +39-332-789085.
Vitae
Barredo: Jos I. Barredo is a postdoctoral researcher at the European Commission, DG Joint Research Centre, Institute for Environment and Sustainability (Italy). He has done PhD in GIS and remote sensing in 1995 at University of Alcal (Spain). He has worked in the field of GIS applied to urban and regional planning studies. Currently, his major research topic is the application of cellular automata for the study of urban processes. Kasanko: Marjo Kasanko is a researcher at the European Commission, DG Joint Research Centre, Institute for Environment and Sustainability (Italy). She has a Ph. Lic. in impact assessment methodology in the University of Helsinki in 1999. She has worked in the field of various impact assessments (mainly environmental and employment) and urban and regional development. At the moment she is developing spatial indicators for measuring urban sustainability. McCormick: Niall McCormick is a researcher at the European Commission, DG Joint Research Centre, Institute for Environment and Sustainability (Italy), where he has carried out extensive research into the use of advanced GIS and remote sensing techniques for wide-area environmental surveys, such as forest inventories and urban growth assessments. He is currently working on the development and application of automated methods for measuring and modelling the spatial impacts of urban expansion, including landscape fragmentation. Lavalle: Carlo Lavalle works at the European Commission, DG Joint Research Centre, Institute for Environment and Sustainability (Italy), as Head of the Urban and Regional Development sector. He is working in the following areas of applications: sustainable urban and regional development, environmental impact of human activities, environmental protection and impact assessment of transport infrastructure.
Landscape and Urban Planning Volume 64, Issue 3, 15 July 2003, Pages 145-160
Home
Browse
Search
My settings
My alerts
About Elsevier About Elsevier About SciVerse About SciVal Terms and Conditions Privacy policy Information for advertisers
Help
About ScienceDirect What is ScienceDirect Content details Set up How to use Subscriptions Developers
Contact and Support Contact and Support
Copyright 2011 Elsevier B.V. All rights reserved. SciVerse is a registered trademark of Elsevier Properties S.A., used under license. ScienceDirect is a registered trademark of Elsevier B.V.
12 sur 12
08/03/2011 08:38

Science Direct - Landscape and Urban Planning - Modelling Dynamic Spatial Processes - Simulation of Urban Future Scenarios Throug

Uploaded by

Copyright:

Available Formats

Science Direct - Landscape and Urban Planning - Modelling Dynamic Spatial Processes - Simulation of Urban Future Scenarios Throug

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Science Direct - Landscape and Urban Planning - Modelling Dynamic Spatial Processes - Simulation of Urban Future Scenarios Throug

Uploaded by

Copyright:

Available Formats

ScienceDirect - Landscape and Urban Planning : Modelling dynamic spa...

Thumbnails | Full-Size images

Permissions & Reprints

, Marjo Kasanko, Niall McCormick and Carlo Lavalle

Related reference work articles e.g. encyclopedias