Professional Documents
Culture Documents
Modelling Urban Change With Cellular Automata: Contemporary Issues and Future Research Directions
Modelling Urban Change With Cellular Automata: Contemporary Issues and Future Research Directions
Yan Liu
University of Queensland, Australia
Michael Batty
University College London, UK
Siqin Wang
University of Queensland, Australia
Jonathan Corcoran
University of Queensland, Australia
Abstract
The study of land use change in urban and regional systems has been dramatically transformed in the last
four decades by the emergence and application of cellular automata (CA) models. CA models simulate
urban land use changes which evolve from the bottom-up. Despite notable achievements in this field,
there remain significant gaps between urban processes simulated in CA models and the actual dynamics
of evolving urban systems. This article identifies contemporary issues faced in developing urban CA
models and draws on this evidence to map out four interrelated thematic areas that require concerted
attention by the wider CA urban modelling community. These are: (1) to build models that compre-
hensively capture the multi-dimensional processes of urban change, including urban regeneration, den-
sification and gentrification, in-fill development, as well as urban shrinkage and vertical urban growth; (2)
to establish models that incorporate individual human decision behaviours into the CA analytic frame-
work; (3) to draw on emergent sources of ‘big data’ to calibrate and validate urban CA models and to
capture the role of human actors and their impact on urban change dynamics; and (4) to strengthen
theory-based CA models that comprehensively explain urban change mechanisms and dynamics. We
conclude by advocating cellular automata that embed agent-based models and big data input as the most
promising analytical framework through which we can enhance our understanding and planning of the
contemporary urban change dynamics.
Corresponding author:
Yan Liu, University of Queensland, School of Earth and Environmental Sciences, Brisbane, Queensland 4072 Australia.
Email: yan.liu@uq.edu.au
4 Progress in Human Geography 45(1)
Keywords
agent-based modelling (ABM), big data, cellular automata (CA), future research directions, human
behaviours, multi-dimensional urban change processes
2008; Václavı́k and Rogan, 2009; Adhikari and rules where the rules are invariant through time,
Southworth, 2012), iCity (Stevens et al., 2007), while ABMs are founded on agents acting
Metronamica (Stanilov and Batty, 2011; White within the cellular landscape, interacting purpo-
et al., 2015), and the more recently developed sively with one another and with the environ-
FLUS-CA model by Liu et al. (2017). ment defined by cellular space. Agents thus
Yet, contemporary practice in urban CA have the potential to change their own beha-
modelling continues to be limited by its persis- viours as they interact in the cellular space.
tent over-simplification of urban processes that Given the somewhat diverse practices in the
are both multidimensional and complex (Tor- development and application of urban CA mod-
rens and O’Sullivan, 2001; Salvati and Serra, els due to the lack of a holistic understanding of
2016). While models (in general) are abstrac- the multi-dimensional urban change processes
tions and thus simplifications of reality, most as well as the complex human decision beha-
urban CA models focus on simulating the viours, there is a pressing need for the CA mod-
spatio-temporal processes of urban expansion. elling community to coalesce around a series of
Few attempts to date have been made to capture research themes through which targeted prog-
the wider chaotic dynamics of the urban system ress can be achieved. The aim of this paper is to
that cover a broader spectrum of urban change address this need by summarising these contem-
processes, including urban regeneration, densi- porary issues and challenges, and then charting
fication, gentrification, inner-city decline, poly- a future research agenda towards which we can
centric formation, de-urbanisation, and urban orientate our collective effort in urban CA
shrinkage, to name but a few. Current urban modelling.
CA models are also limited in simulating the
influences of physical and economic factors,
and often overlook the summative impact of
II Contemporary issues in urban
human decision-making behavioural factors on CA modelling
urban development. This shortcoming may be A large number of urban models have been
due to the inability of CA models to represent developed for different policy fields at various
entities at the finest disaggregate levels, such as levels of spatial and temporal resolution (Spie-
individual decision behaviours that drive the kermann and Wegener, 2018), with the broad
evolution of social systems (O’Sullivan, aim of capturing the dynamics of space, time,
2002). Here, we argue that there is a need for and human choice in relation to urban change.
CA models to embrace a wide variety of urban According to King and Kraemer (1993), a
change processes that acknowledge the bottom- model should play at least three key roles in
up processes that are associated with fractal pat- a policy context: (1) to clarify the issues in a
terns and dynamics, which taken together offer debate, such as issues pertaining to the interac-
the capacity to model more realistically highly tions and conflicts between different urban
complex urban systems (Batty, 2007, 2013). We activities and land uses; (2) to enforce a disci-
also consider that the integration of CA with pline of analysis and discourse among
other individual-based models such as the stakeholders; and (3) to provide suggestive
agent-based modelling (ABM) provides unique feedback and advice primarily in the form of
opportunities for urban CA modellers to incor- what not to do – since the politics of any prac-
porate the underlying human decision factors tical application often conflicts with what a
that drive urban change. Here, CA is considered model suggests. By considering these three
as a passive system that provides a cellular objectives alongside the abilities of existing
space which evolves under different transition urban models, we identify the following four
6 Progress in Human Geography 45(1)
key issues in the contemporary literature of shrinkage has also been observed in some cities
urban CA modelling. These include: in China, Japan, and South Africa (Rieniets,
2009; Long and Wu, 2016).
the restrictions of CA modelling to urban From recent reviews of CA-based urban
expansion rather than the multi- modelling (Aburas et al., 2016; Musa et al.,
dimensional processes of urban change, 2017), it is evident that most existing urban
such as urban regeneration, densification models focus on simulating urban expansion
and gentrification, in-fill development, and sprawl (Liu, 2008, 2012; Liu et al., 2013;
sprawl, as well as urban shrinkage and Sakieh et al., 2015; Pérez-Molina et al., 2017),
vertical urban growth; whereas studies of other types of urban transfor-
the lack of factors representing individual mation such as gentrification, regeneration, and
human decision behaviours and their col- urban shrinkage have been rather limited.
lective implications for urban change; Recent CA modelling concerned with the
the minimal effort in drawing on emer- process of urban regeneration involves the
gent sources of ‘big data’ to calibrate and application of ABMs that employ cellular repre-
validate CA models and to capture the sentations and complex model calibration rou-
role of human actors and their impact tines. For instance, Jordan et al. (2014)
on urban dynamics; and developed an agent-based model to simulate
the absence of theories in CA modelling residential mobility and assess the impact of
that comprehensively explain urban urban regeneration policy on the housing choice
change mechanisms and dynamics. behaviours for a residential community in the
We discuss each of these in turn. United Kingdom. Another application by Zheng
et al. (2015) simulated land use change in an
urban renewal district in Hong Kong by com-
1 The restrictions of CA modelling to urban bining the conversion of land use and its effects
at the small area level (CLUE-S) model with a
expansion rather than the multi- Markov prediction model. Their work demon-
dimensional processes of urban change strates the utility of the modelling framework as
Following the processes of urbanisation and a policy tool for scenario analysis of urban
suburbanisation which came to dominate urban renewal, but lacked the capacity for capturing
growth in the post-war years, one extreme and the entire life-cycle of urban regeneration from
diametrically opposed trend – urban shrinkage – land evacuation to redevelopment.
has more recently emerged. This trend is char- Urban gentrification, as a process that is usu-
acterised by low fertility rates, outmigration of ally accompanied by urban regeneration, has
young families, declining productivity and the also gained attention in the CA modelling com-
lack of a skilled workforce (Martinez-Fernan- munity. O’Sullivan (2000, 2002) simulated
dez et al., 2012). These exist primarily in the urban gentrification at the micro level by apply-
older industrial regions of Europe (Northern ing an irregular graph-based CA architecture,
England, Scotland’s Clydeside, Lorraine, and drawing on the principle of proximal space and
the Rhine-Ruhr region), in large parts of the rent gap theory in residential property markets.
post-socialist countries in eastern Europe Following on from this, Diappi and Bolchi
(Großmann et al., 2008; Haase et al., 2016), and (2006) investigated the gentrification process
in north-eastern United States rustbelt cities by applying an urban spatial model of gentrifi-
such as Buffalo, Cleveland, and Pittsburgh cation also based on rent gap theory (Smith,
(Wiechmann and Pallagst, 2012). Recent 1987). Their model included behavioural rules
Liu et al. 7
for each type of agent such as homeowners, housing supply and the specific demands of the
landlords, tenants, and developers, with non- individual agents, both of which are key issues
linear interactions between agents at the local that should be considered. As such, there is a
level which can then produce different config- pressing need to develop models that address
urations of the system at the macro level. Simi- the question of how socio-spatial land use
larly, Torrens and Nara (2007) developed a change dynamics contribute to urban shrinkage.
hybrid cellular and agent-based automata model Urban development is a process that involves
that allowed for the representation of co- changes in urban form in both the horizontal and
interactions among fixed and mobile entities vertical geometric dimensions. Since the 1950s,
in urban settings as well as across multiple spa- planners have commonly perceived that vertical
tial scales. While this hybrid approach is useful growth in the form of tall buildings increases
in representing human behaviours in complex urban density and is desirable in eliminating
adaptive urban systems, it could also benefit urban sprawl, increasing housing affordability,
from considering more top-down factors such reducing energy costs for transportation, and
as urban planning and zoning, social biases, and distributing resources in a more compact way
cultural factors in the form of traditional cus- (Goetz, 2013). Therefore, vertical urban growth
toms and behaviours shared by certain ethnic represents one of the most important aspects of
communities. ‘smart growth’ and ‘sustainable development’,
Beyond urban expansion, some scholars have which further transforms the morphology and
also modified the strict CA model by adding functioning of cities (Palme and Ramı́rez,
constraints and processes that enable the simu- 2013). For example, the development of high-
lation of urban shrinkage (Santé et al., 2010; rise buildings has also been considered as a
Schwarz et al., 2010). For instance, Haase symbol of advancement, wealth, and efficiency,
et al. (2010) developed an agent-based model which resonates particularly with the politics in
that computes spatially explicit household pat- developing countries where the transformation
terns, housing demand, and residential vacan- of cities is primarily concerned with a large
cies. This model was applied to simulate urban population base (Palme and Ramı́rez, 2013).
shrinkage in Leipzig, Germany. An updated Vertical urban growth is also reflected in build-
version of a joint system dynamics (SD)-CA ings with various functions including commer-
model and an ABM was developed by Haase cial, residential, and industrial land uses (Lin
et al. (2012) using an integrated dataset of land et al., 2014). Some scholars also report the neg-
cover and cadastral data, with specific indica- ative effects of vertical urban growth; for exam-
tors of urban shrinkage including population ple, the densification of urban surfaces can
decline, change in household structure, housing decrease urban efficiency (Jaksch et al., 2016),
costs, proximity of growing and declining increase pollution (Aristodemou et al., 2018),
neighbourhoods, decline in land use density accelerate urban heat island effects (Santa-
through the expansion of brownfields and the mouris et al., 2015), amplify road traffic noise
emergence of unintended green spaces, and the (Tang and Wang, 2007), and adversely affect
concomitant consumption of land for new resident habits and lifestyles. In fact, there is
development. The integration of different mod- no longer widespread agreement that the
elling approaches enabled the inclusion of data increased compactness of vertical urban growth
on demographic, socioeconomic, housing, and reduces either urban sprawl or the use of energy
governance features of urban shrinkage. How- in cities.
ever, these models capture neither decision- However, despite the divergent views on ver-
making processes nor the relationships between tical urban growth and its impact on urban form,
8 Progress in Human Geography 45(1)
it is only in recent years that CA models of study, Lin et al. (2014) developed a 3D CA
vertical urban growth have emerged (Lin model that adopts a linguistic approach to simu-
et al., 2014; Koziatek and Dragićević, 2017), late building distribution patterns across space
though 3D CA models have for some years and time. This model combines a series of vari-
already been applied in other fields. Despite the ables such as population density, building
fact that CA can be easily extended to incorpo- height, and accessibility to transportation nodes.
rate 3D growth, most simulations have only It provides a modelling framework that captures
dealt with two-dimensional space, with the vertical urban growth patterns across space
exception of early ad hoc examples such as from the city centre through the fringe, periph-
those developed by Batty and Longley (1994), ery, and hinterland. More recently, Koziatek
Semboloni (2000), and Semboloni et al. (2004), and Dragićević (2017) developed the 3D iCity
who generated fractal growth in 3D when to model vertical urban development in the
exploring the space-filling properties of urban Canadian city of Surrey. To further develop
regions but did not pursue the simulations fur- applications of this type of 3D urban CA model,
ther. However, models that better account for we need to confront a series of challenges which
simulating the dynamics of spatio-temporal pro- include: (1) insufficient data to parameterise the
cesses and patterns associated with vertical model at the initial stage of model operation; (2)
urban growth need to be further developed. This the inability to validate the results over the
limitation is not just confined to CA models; all simulation period; and, (3) the inaccuracies in
types of urban simulation models focus exclu- simulating buildings with heterogeneous
sively on simulating horizontal expansion heights using representations based on varying
(Schwarz et al., 2010), despite the global trend neighbourhood cells (Lin et al., 2014; Koziatek
of increasing urban density through high-rise and Dragićević, 2017).
living. For instance, in CA-based urban models,
the state of an ‘urban’ cell is typically consid- 2 The lack of factors representing individual
ered as the final state that will not change
regardless of any further development on the
human decision behaviours and their
cell through multi-functional development collective implications for urban change
(such as high-rise buildings that have multiple Compared to natural or agricultural landscapes,
functions including retail and residential use), urban systems are strongly influenced by both
densification, or vertical urban growth. Further- human and environmental factors. In the con-
more, transportation and interactions between text of the urban environment, this can be inter-
cells are usually confined to two dimensions, preted as a concern for the degree of harmony
despite there being significant interaction between city residents and their everyday urban
effects in the third dimension. The advancement surroundings (Pacione, 1990). Land use
of LiDAR and building information models changes are the consequence of various drivers
(BIM) representing physical and functional that include the economy, technology, and
characteristics of places and the vertical mixed human behaviours and decisions (Agarwal
use of multi-level buildings has begun to fulfil et al., 2002). Among these drivers, people’s
an analytical need in urban CA modelling behaviours and decisions such as the housing
(Batty, 2000). Links between 3D representation choices of residents and investors, investment
and CA are being explored using generative decisions by land developers, and urban plan-
grammars such as those embodied in ERSI’s ning regulations and design have played an
CityEngine software (Koziatek and Dragićević, overriding role. Haase and Schwarz (2009) pro-
2017). As a practical example and a pioneer vide a general but comprehensive overview of
Liu et al. 9
the major components of an urban system, con- Nevertheless, the existing literature reveals a
sisting of three components collected through somewhat simplified capacity to model com-
four types of relationships: (1) the impact of the plexity in human decision-making using more
human sphere on land use; (2) feedback of land or less sophisticated measurements or concep-
use on the human sphere; (3) the impact of land tualisations of stochasticity (De Almeida et al.,
use on the environment (including ecosystems); 2003; Feng et al., 2016). These models are yet to
and (4) feedback of the environment on the capture the fundamental role of humans and
human sphere (Haase and Schwarz, 2009). The their interactions within the built environment
principal driving force for urban change is which shapes our cities, and there is no unique
the way in which the human sphere and the approach to represent an individual’s decision
pressure it exerts on urban land use can have behaviours and how collectively such individ-
environmental consequences, which can in turn ual decision behaviours impact on the change in
affect future human decisions and behaviours. urban form. The critical role of human beha-
Since cities are places where people deliber- viours in urban models has been generally over-
ately come together to interact, our understand- looked for two main reasons: (1) the challenges
ing of the evolution of cities must be enriched of developing realistic operational models to
by studies of networks, interactions, connec- incorporate the impact of decision behaviours
tions, and transactions between humans and the and other drivers of urban transformation
environment, ranging from individual, local, (Elliott and Kiel, 2002), and (2) the difficulty
and regional, to global scales (Batty, 2013). of even observing, but also collecting, verify-
In the process of coupling human and envi- ing, and validating data reflecting an individu-
ronmental systems with urban modelling, al’s decision behaviours as dynamic contextual
understanding how humans make decisions is parameters (Crooks et al., 2008). Subsequently,
of paramount importance (Gimblett, 2002; An, CA models rarely have had the capacity to
2012). Human decisions and subsequent incorporate the most comprehensive drivers of
actions change the structure and function of human behaviours with respect to land use
many environmental systems, which in turn change (Agarwal et al., 2002). The emergent
influence human decisions and actions. To sources of ‘big data’ in various forms at the
assess the extent to which current urban CA individual level could shed new light on how
models have captured the human–environmen- urban modellers might tackle this fundamental
tal dynamics, Table 1 presents a selection of issue in urban CA models, but this will require
urban CA models that considered (or not) dif- new forms of data concerning decision pro-
ferent levels of human-decision complexity cesses which currently barely exist. Moreover,
categorised by Agarwal et al. (2002), as illu- the lack of effort in using such data in current
strated in Figure 1. CA modelling practice needs to be addressed.
Table 1 shows that some CA models, partic-
ularly those in more recent studies, have com-
prehensively considered human–environment 3 The minimal effort in drawing on
interactions in the modelling process. Further- emergent sources of ‘big data’ to calibrate
more, the integration of CA and ABM would and validate CA models and to capture the
allow the decision behaviours of various
‘agents’ to be incorporated and simulated in the
role of human actors and their impact on
cellular space in order to understand the emer- urban change dynamics
gent spatial patterns through time (Batty, 2009; A persistent challenge in urban CA modelling
Waddell, 2002; Jjumba and Dragićević, 2012). concerns the fitting of urban CA models to data.
10 Progress in Human Geography 45(1)
Table 1. The levels of modelling human-decision complexity in urban CA models, adapted from Agarwal
et al. (2002: 6).
Human-decision
Level complexity Model components Examples
1 No human decision-making (only Land use (e.g. urban or non- Rafiee et al., 2009; Van Vliet
biophysical variables) urban, roads, different land et al., 2009; Feng et al., 2011;
use types) Liao et al., 2016; Liu and Feng,
Environment (e.g. topography) 2016.
2 Human decision-making Human sphere (e.g. demand Verburg and Overmars, 2007;
assumed to be related rules, population, economy, He et al., 2008; Lin et al., 2014.
deterministically to selected planning, accessibility via
human variables (such as transportation network)
population size, change or Land use (e.g. suitability rules,
density) land use functions)
3 Human decision-making seen Human sphere (e.g. population, Jantz et al., 2010; Haase et al.
as a probability function household, jobs, employment) 2012; Fuglsang et al., 2013;
depending on socioeconomic Land use (e.g. single-family Kamusoko and Gamba, 2015;
and/or biophysical variables residential, multi-family Rienow and Goetzke, 2015;
without feedback from the residential, commercial, Sakieh et al., 2015; Berberoğlu
environment to the choice industrial, transportation, et al., 2016; Tian et al., 2016.
function public)
Environment (e.g. undeveloped
land)
4 Human decision-making seen Human sphere (e.g. urban Dabbaghian et al., 2010.
as a probability function growth, policy simulation and
depending on socioeconomic evaluation)
and/or biophysical variables Environment (e.g. habitat change
with feedback from the and habitat fragmentation)
environment to the choice
function
5 One type of agent whose Human sphere (e.g. initial capital, Torrens and Nara, 2007; Jordan
decisions are modelled overtly lending amount) et al., 2014; Wahyudi et al.,
with regard to choices made Environment (e.g. distance to 2019a.
about variables that affect CBD/toll road)
other processes and Land use (e.g. land cover/values)
outcomes Interaction: developer-
environment relation (e.g. land
find, assess, and decide)
6 Multiple types of agents whose Human sphere (e.g. multiple O’Sullivan, 2002; Semboloni
decisions are modelled overtly agents, socioeconomic et al. 2004; Diappi and Bolchi,
with regard to choices made change) 2006; Liu et al., 2013; Liu et al.,
about variables that affect Environment (e.g. climate 2017; Wahyudi et al., 2019b.
other processes and change)
outcomes Land use (e.g. historical land use
change)
Interactions between variables
via system dynamics and ABMs
Liu et al. 11
use across unappropriated purposes if without mathematical or logical models, with computers
protection (Perera et al., 2015). acting as the laboratory in which the simulation
When data are ready for analysis, the chal- of reality takes place (Haase et al., 2012). Given
lenge becomes how to organise existing data in a good theory, an urban model would be con-
meaningful ways to allow modelling and com- structed, validated, and then used as a vehicle
parison of hidden patterns and relationships for refining the theory through ‘what if’ style
among dynamic urban systems. Data-driven experiments and sensitivity testing (Batty,
urban modelling, particularly where the data 1976; Crooks et al., 2008). However, the overt
tend to be big, is considerably more sophisti- role of theory has faded in many contexts as
cated than the simpler skills we used in the past urban models embed theory within themselves
for locating, identifying, analysing, and citing (Crooks et al., 2008); that is, theory is often
data (Wu et al., 2014). What needs to be primar- derived as the model is constructed. For exam-
ily kept in mind is the purpose of modelling, ple, UrbanSim as a simulation platform for sup-
rather than adapting studies to data that are porting planning and analysis of urban
available. An opportunistic way of conducting development has been developed by adopting
data-driven research may lead to interesting a micro-simulation strategy that directly repre-
observations, but often bypasses ideas not sents the choices of households, businesses,
meaningful to tackling real-world issues developers, and governments (representing pol-
(Hashem et al., 2015). Past studies have icy inputs) in the real estate market in a way that
revealed how big data can help to improve the is ‘behaviourally natural and intuitive’ and can
planning of smart cities in at least four respects be understood by non-technical stakeholders
(Lim et al., 2018): preventive administration (Waddell, 2011: 217). Because models come
(e.g. civil complaint and crime prevention), to control all of the inputs and parameters, most
operational management (e.g. trash collection social systems cannot be represented in the form
and traffic control), network development (e.g. of a theory that guarantees any measure of clo-
bus service scheduling and Wi-Fi hotspot opti- sure. Earlier generations of urban models have
misation), and information diffusion (e.g. pollu- demonstrated the extent to which they can be
tion monitoring and intelligent navigation). validated in terms of the goodness-of-fit of the
Thus, changing city operations and improving model to an existing system, but the difficulties
the lives of city residents should be the ultimate and failures in model validation are features that
goal of using big data in urban modelling, where exist for both the theory involved and data from
urban research should combine both academic the observed reality (Crooks et al., 2008).
rigor and practical knowledge. A large number of urban models have been
developed over the last half century, and most
4 The absence of theories in CA modelling can generally be classified into two categories:
top-down aggregate models or bottom-up disag-
that comprehensively explain urban change gregate models (Tan et al., 2014). Top-down
mechanism and dynamics models, such as economic equilibrium models,
Contemporary sciences usually invoke theory are often constructed by breaking down a sys-
that enables scientific predictions to be com- tem to gain insight into its compositional sub-
pared with reality, with respect to the variations systems in a reverse engineering fashion; such a
in the phenomena of interest across time and modelling approach is typically based on tradi-
space (Batty, 2009). To deal with the complex- tional macroeconomic theories and is unable to
ity of such realities, theories are translated into a deal with micro-level decision-making or with
form that enables them to be represented as small-scale social and environmental problems
14 Progress in Human Geography 45(1)
(Itami, 1994). On the other hand, the bottom-up reflected in the equilibrating micro-economics
modelling approach is formulated by piecing of the individual and the firm. This theory pro-
together sub-systems or its components to for- vides important insights into residential develop-
mulate more complex systems; this modelling ment, and there is clear potential for invoking
approach has become more dominant in the such theory in any CA or ABM that captures
field of urban modelling with the ongoing human behaviours. In short, the scope of urban
development of algorithms to represent the models and theory is now considerably wider
dynamics of urban systems. CA models, as one than in the past, and urban CA models must
of the most popular bottom-up approaches, can respond to this increasing complexity by expli-
capture changes in urban morphology through citly embracing the most appropriate theories
simple and flexible transition rules (Santé et al., pertaining to economic and social decision-
2010; Feng and Liu, 2016). However, CA mod- making. This is a major challenge given that
els inevitably have the common weakness of all it requires theory that deals not only with the
bottom-up models: modellers have done little to static structure of the space but also its tem-
link their models to urban theories which exist poral dynamics, though this has been slow in
at a more aggregate scale (Torrens and O’Sulli- progressing over the recent decades.
van, 2001). They have also faced difficulties in Developing urban models based on explicit
incorporating human decision behaviours theory has two benefits for modellers. First, it
(Haase et al., 2010; Arsanjani et al., 2013) and simplifies the process of identifying driving fac-
in capturing the macro-scale social and eco- tors – physical/environment, institutional, or
nomic driving forces of urban change (Han human decision behaviour factors – on urban
et al., 2009). change dynamics, particularly with regard to
The theoretical orientations of many CA and modelling the complexity of human decision-
ABMs remain implicit and hidden, often covered making behaviours. Second, the theory contri-
by ad hoc assumptions about modelling struc- butes to stakeholders’ involvement with respect
ture, process, and software interfacing (Crooks to an in-depth understanding of causes and feed-
et al., 2008). Moreover, their processes, although backs of changes in human behaviours. Such
explicit in these models, are almost invisible with model applications could well be used to cumu-
respect to observation and data. In many cases, latively modify human behaviours (Davis et al.,
the development of an urban CA model is only an 2015). This benefit is critical when stakeholders
additional application of some simple structures need to encourage city residents to behave in a
that are adjusted for a local context (Couclelis, certain way. A good example is the positive
2002). In the increasingly diverse array of CA influence of transit-oriented development that
modelling applications, such models are consid- increases the amount of residential, business,
ered generic; they can be applied and fitted to and leisure space within walking distance of
data and processes in any particular field, subject public transport, thereby enhancing accessibil-
to use for particular purposes, and hence largely ity and promoting healthy community. It is thus
independent of theory and practice (Batty, 2007). necessary that urban modellers pay more atten-
Although a theory is not necessarily required to tion to develop models in accordance with the-
guide this kind of urban modelling, many facets ories of human behaviour. On the other hand,
of theorising and thinking should be brought to some stakeholders such as politicians may also
bear on model construction, and to understand prefer a ‘black box’ approach, whereby a result
the mechanisms of urban change dynamics. For is generated from a model that fully explores a
instance, location theory, which has long been theory. In this case, CA models engaging Arti-
used in the study of urban spatial structure, is ficial Neural Networks (ANNs) as a machine
Liu et al. 15
learning approach, for example, would appear First, it is critical that models incorporate
more appropriate (Li and Yeh, 2002). various types of drivers and constraints associ-
ated with urban transformation so that these can
be applied across multiple metropolitan con-
III Towards a future research texts. By doing so, complex transition rules
agenda need to be considered to determine how,
Responding to current challenges in urban CA when, and to what extent a given land parcel
modelling, we now map out four interrelated might be developed through time. For exam-
thematic areas that we argue require concerted ple, when modelling urban regeneration and
attention by the urban CA modelling commu- gentrification, one often needs to change the
nity. These are: (1) to build models that com- order in which the transition rule sets are
prehensively capture the multi-dimensional implemented so that the various effects of
processes of urban change, including urban regeneration and gentrification can be incor-
regeneration, densification and gentrification, porated over time. We argue that urban
in-fill development, as well as urban shrinkage growth and decline as simulated in CA models
and vertical urban growth; (2) to establish is not symmetric; the way the transition rules
models that incorporate individual human are implemented differs for modelling growth
decision behaviours into the CA analytic and modelling decline.
framework; (3) to draw on emergent sources Second, the selection of parameters should
of big data to calibrate and validate urban CA consider qualitative factors such as land owner-
models and to capture the role of human actors ship and land lease, land use density change, and
and their impact on urban change dynamics; land parcel shape by adjusting the CA model’s
and (4) to strengthen theory-based CA models parameter settings. For instance, the choice of
that comprehensively explain urban change the neighbourhood type and size, which has sig-
mechanisms and dynamics. nificant impact on the global behaviour of a CA
model, can be adjusted contingent on varying
perceptions of the relative merits of other neigh-
1 Modelling multi-dimensional processes bourhoods in the wider urban system.
of urban change Third, 3D urban modelling needs to be
As discussed in the previous sections, few urban enhanced, especially for modelling vertical
CA models can simulate multiple urban forms growth. To this end, LiDAR datasets are a pro-
and processes. Although some of these gaps are mising source of information for 3D urban mod-
being addressed to a certain extent by a number elling, but collecting and processing these
of researchers, such as O’Sullivan (2002), Dia- datasets is both time consuming and costly.
ppi and Bolchi (2006), and Haase et al. (2010, Land use attributes at the parcel level are being
2012), the challenges faced by developers of collected slowly by various national mapping
urban CA models producing single urban form agencies, but policymakers should provide
outputs lie in how to combine the multiple back- more facilities for universities and the private
ground processes to produce a range of urban sector to collect and process the data so that it
forms and phenomena – from urban regenera- can be used for urban modelling – without the
tion, gentrification, densification, vertical need for intensive pre-processing. As a result,
growth, to urban shrinkage. In order to over- creating a platform for data sharing can provide
come the weaknesses and limitations in current new opportunities for urban modellers to model
scholarship, future research needs to consider in 3D. Scholars also need to develop more com-
the following. plex transition rules for 3D modelling; to
16 Progress in Human Geography 45(1)
achieve this, close collaboration between pro- in ‘Metronamica’ (Research Institute for
grammers and urban modellers is critical. Knowledge Systems, 2013) and the more recent
‘GeoDynamiX’ (Flemish Research and Tech-
nology Organisation, 2018), but these tend to
2 Incorporating human decision behaviours reflect a loose coupling across different spatial
into the CA modelling framework scales, and further work remains to be done.
The integration of CA with other types of mod- The development of urban models that can
els has been suggested by various scholars capture human decision behaviours and their
(Hewitt et al., 2014; Musa et al., 2017). This interactions with the built environment requires
integration has only been partially addressed micro-scale spatial and social survey data,
through participatory modelling by obtaining which can be furnished in part by the use of
feedback from stakeholders to calibrate the public participatory GIS mapping (McCall,
CA model and enhance its performance (Hewitt 2003; Aburas et al., 2016) which incorporates
et al., 2014). To close this gap, ABMs are an the knowledge of stakeholders – residents, land
unparalleled tool for modelling human deci- developers, and urban planners – who are key
sions. However, the main challenges of ABMs drivers in the urban transformation process. The
arise from their complexity in implementing emergent sources of big data from government,
and designing the real-world rules for the rele- social media, citizen science, and other
vant agents, as noted earlier. Often, these rule location-based services and devices can also
sets are generated using plausible hypotheses serve as excellent input for urban modellers to
but are never tested due to lack of observational understand and model urban change dynamics.
data (Wahyudi et al., 2019a). The CA-ABM
hybrid approach needs to consider top-down
(as well as bottom-up) concepts to address 3 Drawing on emergent sources of big data
issues such as urban planning and zoning, trans- to calibrate and validate urban CA models
portation, social biases, and cultural factors,
which may all be reflected in micro-scale sur-
and capture the role of human actors and
vey data or other primary data sources (Torrens their impact on urban change dynamics
and Nara, 2007). The emergence of open and new data available
Moreover, both CA and ABM have limited from various sources has presented significant
geographic functionalities when considered in opportunities for research in the urban sciences.
isolation (Torrens and Benenson, 2005). In con- Entering into the new era of big data, ever-
trast, the integration of CA and ABM offers a increasing quantities at near real time will ulti-
powerful spatial approach to modelling com- mately change the ways in which human agents
plex geographic systems that are affected by interact with each other and with the urban
physical and human factors at multiple scales space they occupy and transform; these pose
ranging from the individual to the metropolitan new challenges to urban modellers and
region, as tested by Batty (2007) and Torrens researchers (Batty, 2018), and much effort
and Benenson (2005). An integrated CA-ABM should be devoted to conquer the aforemen-
model would provide unique opportunities for tioned big data challenges. Considering that
urban modellers to address various types of these challenges are interrelated since informa-
urban transformation with regard to human tion creation, data collection, manipulation,
decisions and preferences. The integration with analysis, and modelling are interdependent
other types of models to develop extended suites activities, we propose the following four
of urban models has been attempted particularly considerations for future work of modelling
Liu et al. 17
used in ABM, including micro-economic mod- decision behaviours; (3) drawing on emergent
els, space theory-based models, psychosocial sources of big data to calibrate and validate
and cognitive models, institution-based models, urban CA models and to capture the role of
experience- or preference-based decision mod- human actors and their impact on urban change
els (rules of thumb), participatory agent-based dynamics; and (4) strengthening theory-based
models, empirical- or heuristic-based models, CA models that comprehensively explain
and evolutionary programming methods (An, urban change mechanisms and dynamics.
2012). Some machine learning models such as We suggest that the continued growth in CA
ANNs have also been developed (see, for exam- modelling is in part contingent on tackling these
ple, Li and Yeh, 2002), but these models need to four challenges in order to remain at the van-
avoid the many traps of autocorrelation and guard of urban modelling. In tandem with pro-
multi-collinearity, along with a host of ad hoc gressing these four themes is the need to
pattern-matching features that are essentially develop, implement, and train urban planners
spurious in terms of the way spatial systems and policymakers in the use of CA-based model
function and evolve. Furthermore, CA models outputs. This is indeed most important to ensur-
based on human behaviour theory need to incor- ing that urban CA modelling moves from an
porate the probability-of-occurrence with the approach that is arguably the purview of a rela-
conversion cost, neighbourhood conditions, and tively select academic community to one that
competition among the different agents (Liu has a place in mainstream policy and practice.
et al., 2017; Wahyudi et al., 2019b). The inter- The first component of setting urban CA mod-
active coupling of the top-down system elling on this path towards mainstream use is the
dynamic demands and the bottom-up CA development of user-friendly tools that are
approach would likely enhance the model’s embedded within familiar computing environ-
capability for long-term stochastic simulation, ments, of which the iCity model (Stevens et al.,
reflecting the real city comprised of chaotic 2007) is an excellent example in this regard.
human behaviours. To sum up, the current global interest in sus-
tainable urban development has highlighted the
need to deepen our understanding of the pro-
IV Conclusion cesses that underpin urban transformations –
This article acknowledges the fundamental and land use and urban models here play a vital
transformations which have been brought to role. However, there persists a major void
urban modelling via the bottom-up perspective between the real features of our urban systems
through cellular automata models. Despite and the relevant representativeness of these fea-
notable achievements in this field over a num- tures with the current array of urban CA models.
ber of decades, we argue that there remain at The purpose of this study has been to chart these
least four pressing issues faced by CA model- deficits and to map out a future research agenda
lers in contemporary urban modelling for both urban CA models and the CA model-
practices. We draw out these issues into a dis- ling community. To this end, we encourage
cussion mapping out four interrelated thematic scholars to concentrate their efforts on develop-
areas that require concerted attention by the ing multi-dimensional, dynamic, and vector-
urban CA modelling community, which based models to simulate the holistic processes
include: (1) building models that comprehen- of urban change. We also highlight the need to
sively capture the multi-dimensional processes better capture the role of human behaviours and
of urban change; (2) establishing urban CA decisions by drawing on theories that can facil-
models that incorporate individual human itate more rigorously grounded modelling
Liu et al. 19
Batty M and Longley P (1994) Fractal Cities: A Geometry De Almeida CM, Batty M, Monteiro AMV, Câmara G,
of Form and Function. San Diego, CA: Academic Press. Soares-Filho BS, Cerqueira C and Pennachin CL
Batty M, Couclelis H and Eichen M (1997) Urban systems (2003) Stochastic cellular automata modeling of urban
as cellular automata. Environment and Planning B: land use dynamics: Empirical development and estima-
Planning and Design 24(2): 159–164. tion. Computers, Environment, and Urban Systems
Berberoğlu S, Akın A and Clarke KC (2016) Cellular 27(5): 481–509.
automata modeling approaches to forecast urban Diappi L and Bolchi P (2006) Gentrification waves in the
growth for Adana, Turkey: A comparative approach. inner-city of Milan. In: Leeuwen JP and Timmermans
Landscape and Urban Planning 153: 11–27. HJ (eds) Innovations in Design & Decision Support
Bithell M and Macmillan WD (2007) Escape from the cell: Systems in Architecture and Urban Planning. Dor-
spatially explicit modelling with and without grids. drecht: Springer, 187–201.
Ecological Modelling 200(1–2): 59–78. Dong XL and Srivastava D (2013) Big data integration. In:
Chaudhuri G and Clarke K (2013) The SLEUTH land use 2013 IEEE 29th International Conference on Data
change model: A review. Environmental Resources Engineering (ICDE), 1245–1248.
Research 1(1): 88–105. Dong X, Li R, He H, Zhou W, Xue Z and Wu H (2015)
Chen H, Chiang RH and Storey VC (2012) Business Secure sensitive data sharing on a big data platform.
intelligence and analytics: From big data to big impact. Tsinghua Science and Technology 20(1): 72–80.
MIS Quarterly 36(4): 1165–1188. Dyson G (2012) Turing’s Cathedral: The Origins of the
Clarke KC, Hoppen S and Gaydos L (1997) A self- Digital Universe. New York: Penguin Books.
modifying cellular automaton model of historical urba- Elliott E and Kiel LD (2002) Exploring cooperation and
nization in the San Francisco Bay area. Environment competition using agent-based modeling. Proceedings
and Planning B: Planning and Design 24(2): 247–261. of the National Academy of Sciences 99(3): 7193–7194.
Couclelis H (2002) Modeling frameworks, paradigms, and Feng YJ and Liu Y (2013) A heuristic cellular automata
approaches. In: Clarke KC, Parks BE and Crane MP approach for modelling urban land-use change based
(eds) Geographic Information Systems and Environ- on simulated annealing. International Journal of Geo-
mental Modelling. Upper Saddle River, NJ: Prentice- graphical Information Science 27(3): 449–466.
Hall. Feng YJ and Liu Y (2016) Scenario prediction of emerging
Croitoru A, Crooks AT, Radzikowski J and Stefanidis A coastal city using CA modeling under different envi-
(2013) GeoSocial gauge: A system prototype for ronmental conditions: a case study of Lingang New
knowledge discovery from geosocial media. Interna- City, China. Environmental Monitoring and Assess-
tional Journal of Geographical Information Science ment 188: 540. DOI: 10.1007/s10661-016-5558-y.
27(12): 2483–2508. Feng YJ, Liu Y and Batty M (2016) Modeling urban
Crooks A, Castle C and Batty M (2008) Key challenges in growth with GIS based cellular automata and least
agent-based modelling for geo-spatial simulation. squares SVM rules: A case study in Qingpu-
Computers, Environment and Urban Systems 32(6): Songjiang area of Shanghai, China. Stochastic Envi-
417–430. ronmental Research and Risk Assessment 30(5):
Crooks AT, Croitoru A, Jenkins A, Mahabir R, Agouris P 1387–1400.
and Stefanidis A (2016) User-generated big data and Feng YJ, Liu Y, Tong X, Liu M and Deng S (2011) Mod-
urban morphology. Built Environment 42(3): eling dynamic urban growth using cellular automata
396–414. and particle swarm optimization rules. Landscape and
Dabbaghian V, Jackson P, Spicer V and Wuschke K (2010) Urban Planning 102(3): 188–196.
A cellular automata model on residential migration in Flemish Research and Technology Organisation VITO
response to neighborhood social dynamics. Mathemat- (2018) GeoDynamiX land use solutions. Available at:
ical and Computer Modelling 52(9–10): 1752–1762. http://www.geodynamix.eu/ (accessed 9 June 2019).
Davis R, Campbell R, Hildon Z, Hobbs L and Michie S Fuglsang M, Münier B and Hansen HS (2013) Modelling
(2015) Theories of behaviour and behaviour change land-use effects of future urbanization using cellular
across the social and behavioural sciences: A scoping automata: An Eastern Danish case. Environmental
review. Health Psychology Review 9(3): 323–344. Modelling & Software 50: 1–11.
Liu et al. 21
Gimblett HR (2002) Integrating Geographic Information potential model with cellular automata. Landscape and
Systems and Agent-Based Modeling Techniques for Urban Planning 86(1): 79–91.
Simulating Social and Ecological Processes. Oxford: Hewitt R, Van Delden H and Escobar F (2014) Participa-
Oxford University Press. tory land use modelling, pathways to an integrated
Glaeser EL, Kominers SD, Luca M and Naik N (2018) Big approach. Environmental Modelling & Software 52:
data and big cities: The promises and limitations of 149–165.
improved measures of urban life. Economic Inquiry Ilachinski A (2001) Cellular Automata: A Discrete Uni-
56(1): 114–137. verse. Singapore: World Scientific.
Goetz A (2013) Suburban sprawl or urban centres: tensions Itami RM (1994) Simulating spatial dynamics: Cellular
and contradictions of smart growth approaches in automata theory. Landscape and Urban Planning
Denver, Colorado. Urban Studies 50(11): 2178–2195. 30(1–2): 27–47.
Großmann K, Haase A, Rink D and Steinführer A (2008) Jaksch S, Franke A, Österreicher D and Treberspurg M
Urban shrinkage in East Central Europe? Benefits and (2016) A systematic approach to sustainable urban den-
limits of a cross-national transfer of research sification using prefabricated timber-based attic exten-
approaches. In: Nowak M and Nowosielski (eds) sion modules. Energy Procedia 96: 638–649.
Declining Cities/Developing Cities: Polish and Ger- Jantz CA, Goetz SJ, Donato D and Claggett P (2010)
man perspectives. Poznan, Poland: Instytut Zachodni, Designing and implementing a regional urban model-
77–99. ing system using the SLEUTH cellular urban model.
Haase A, Bernt M, Großmann K, Mykhnenko V and Rink Computers, Environment and Urban Systems 34(1):
D (2016) Varieties of shrinkage in European cities. 1–16.
European Urban and Regional Studies 23(1): 86–102. Jjumba A and Dragićević S (2012) High resolution urban
Haase D and Schwarz N (2009) Simulation models on land-use change modeling: Agent iCity approach.
human-nature interactions in urban landscapes: a Applied Spatial Analysis and Policy 5(4): 291–315.
review including spatial economics, system dynamics, Jordan R, Birkin M and Evans A (2014) An agent-based
cellular automata and agent-based approaches. Living model of residential mobility: Assessing the impacts of
Reviews in Landscape Research 3(2): 1–45. urban regeneration policy in the EASEL district. Com-
Haase D, Haase A, Kabisch N, Kabisch S and Rink D puters, Environment and Urban Systems 48: 49–63.
(2012) Actors and factors in land-use simulation: The Kamusoko C and Gamba J (2015) Simulating urban
challenge of urban shrinkage. Environmental Model- growth using a Random Forest-Cellular Automata
ling & Software 35: 92–103. (RF-CA) model. ISPRS International Journal of Geo-
Haase D, Lautenbach S and Seppelt R (2010) Modeling Information 4(2): 447–470.
and simulating residential mobility in a shrinking city King JL and Kraemer KL (1993) Models, facts, and the
using an agent-based approach. Environmental Model- policy process: the political ecology of estimated truth.
ling & Software 25(10): 1225–1240. In: Goodchild MF, Parks BO and Steyaert LT (eds)
Hammi B, Khatoun R, Zeadally S, Fayad A and Khoukhi L Environmental Modeling with GIS. New York: Oxford
(2017) IoT technologies for smart cities. IET Networks University Press, 353–360.
7(1): 1–13. Koziatek O and Dragićević S (2017) iCity 3D: A geosi-
Han J, Hayashi Y, Cao X and Imura H (2009) Application mualtion method and tool for three-dimensional mod-
of an integrated system dynamics and cellular automata eling of vertical urban development. Landscape and
model for urban growth assessment: A case study of Urban Planning 167: 356–367.
Shanghai, China. Landscape and Urban Planning Labrinidis A and Jagadish HV (2012) Challenges and
91(3): 133–141. opportunities with big data. Proceedings of the VLDB
Hashem IAT, Yaqoob I, Anuar NB, Mokhtar S, Gani A and Endowment 5(12): 2032–2033.
Khan SU (2015) The rise of ‘big data’ on cloud com- Li XC and Gong P (2016) Urban growth models:
puting: Review and open research issues. Information Progress and perspective. Science Bulletin 61(21):
systems 47: 98–115. 1637–1650.
He C, Okada N, Zhang Q, Shi P and Li J (2008) Modelling Li X and Yeh AG-O (2002) Neural-network-based cellular
dynamic urban expansion processes incorporating a automata for simulating multiple land use changes
22 Progress in Human Geography 45(1)
using GIS. International Journal of Geographical spatial planning. Habitat International 27(4):
Information Science 16(4): 323–343. 549–573.
Liao J, Tang L, Shao G, Su X, Chen D and Xu T (2016) Moreno N, Wang F and Marceau DJ (2009) Implemen-
Incorporation of extended neighborhood mechanisms tation of a dynamic neighborhood in a land-use
and its impact on urban land-use cellular automata vector-based cellular automata model. Computers,
simulations. Environmental Modelling & Software Environment and Urban Systems 33(1): 44–54.
75: 163–175. Musa SI, Hashim M and Reba MNM (2017) A review of
Lim C, Kim KJ and Maglio PP (2018) Smart cities with big geospatial-based urban growth models and modelling
data: Reference models, challenges, and considera- initiatives. Geocarto International 32(8): 813–833.
tions. Cities 82: 86–99. O’Sullivan D (2000) Graph-based Cellular Automaton
Lin J, Huang B, Chen M and Huang Z (2014) Modeling Models of Urban Spatial Processes. PhD thesis, Centre
urban vertical growth using cellular automata: for Advanced Spatial Analysis (CASA), University
Guangzhou as a case study. Applied Geography 53: College London.
172–186. O’Sullivan D (2002) Toward micro-scale spatial modeling
Liu X, Liang X, Li X, Xu X, Ou J, Chen Y, Li S, Wang S of gentrification. Journal of Geographical Systems
and Pei F (2017) A future land use simulation model 4(3): 251–274.
(FLUS) for simulating multiple land use scenarios by Pacione M (1990) Urban liveability: A review. Urban
coupling human and natural effects. Landscape and Geography 11(1): 1–30.
Urban Planning 168: 94–116. Palme M and Ramı́rez J (2013) A critical assessment and
Liu Y (2008) Modelling Urban Development with Geogra- projection of urban vertical growth in Antofagasta,
phical Information Systems and Cellular Automata. Chile. Sustainability 5(7): 2840–2855.
New York: CRC Press. Perera C, Ranjan R, Wang L, Khan SU and Zomaya AY
Liu Y (2012) Modelling sustainable urban growth in a (2015) Big data privacy in the internet of things era. IT
rapidly urbanising region using a fuzzy-constrained Professional 17(3): 32–39.
cellular automata approach. International Journal Pérez-Molina E, Sliuzas R, Flacke J and Jetten V (2017)
of Geographical Information Science 26(1): Developing a cellular automata model of urban growth
151–167. to inform spatial policy for flood mitigation: A case
Liu Y and Feng YJ (2016) Simulating the impact of eco- study in Kampala, Uganda. Computers, Environment
nomic and environmental strategies on future urban and Urban Systems 65: 53–65.
growth scenarios in Ningbo, China. Sustainability Pinto NN and Antunes AP (2010) A cellular automata
8(10): 1045. model based on irregular cells: Application to small
Liu Y, Corcoran J and Feng YJ (2019) Cellular automata. urban areas. Environment and Planning B: Planning
In: Kobayashi A (ed.) International Encyclopedia of and Design 37(6): 1095–1114.
Human Geography, 2nd Edition. Elsevier. Pontius RG Jr and Petrova SH (2010) Assessing a predic-
Liu Y, Kong X, Liu Y and Chen Y (2013) Simulating the tive model of land change using uncertain data. Envi-
conversion of rural settlements to town land based on ronmental Modelling & Software 25(3): 299–309.
multi-agent systems and cellular automata. PLoS One Rabari C and Storper M (2014) The digital skin of cities: Urban
8(11): e79300. theory and research in the age of the sensored and metered
Long Y and Wu K (2016) Shrinking cities in a rapidly city, ubiquitous computing and big data. Cambridge Jour-
urbanizing China. Environment and Planning A: Econ- nal of Regions, Economy and Society 8(1): 27–42.
omy and Space 48(2): 220–222. Rafiee R, Mahiny AS, Khorasani N, Darvishsefat AA and
Martinez-Fernandez C, Audirac I, Fol S and Cunningham- Danekar A (2009) Simulating urban growth in Mashad
Sabot E (2012) Shrinking cities: Urban challenges of City, Iran through the SLEUTH model (UGM). Cities
globalization. International Journal of Urban and 26(1): 19–26.
Regional Research 36(2): 213–225. Research Institute for Knowledge Systems (2013) Metro-
McCall MK (2003) Seeking good governance in namica documentation. Maastricht, The Netherlands.
participatory-GIS: A review of processes and gov- Available at: http://www.metronamica.nl/ (accessed 9
ernance dimensions in applying GIS to participatory June 2019).
Liu et al. 23
Rieniets T (2009) Shrinking cities: Causes and effects of Spiekermann K and Wegener M (2018) Multi-level urban
urban population losses in the twentieth century. models: Integration across space, time and policies.
Nature and Culture 4(3): 231–254. Journal of Transport and Land Use 11(1): 67–81.
Rienow A and Goetzke R (2015) Supporting SLEUTH – Stanilov K and Batty M (2011) Exploring the historical
enhancing a cellular automaton with support vector determinants of urban growth patterns through cellular
machines for urban growth modeling. Computers, automata. Transactions in GIS 15(3): 253–271.
Environment and Urban Systems 49: 66–81. Stevens D, Dragićević S and Rothley K (2007) iCity: A
Sakieh Y, Amiri BJ, Danekar A, Feghhi J and Dezhkam S GIS–CA modelling tool for urban planning and deci-
(2015) Simulating urban expansion and scenario predic- sion making. Environmental Modelling & Software
tion using a cellular automata urban growth model, 22(6): 761–773.
SLEUTH, through a case study of Karaj City, Iran. Jour- Tan R, Liu Y, Liu Y, He Q, Ming L and Tang S (2014)
nal of Housing and the Built Environment 30(4): 591–611. Urban growth and its determinants across the Wuhan
Salvati L and Serra P (2016) Estimating rapidity of urban agglomeration, central China. Habitat Interna-
change in complex urban systems: A multidimen- tional 44: 268–281.
sional, local-scale approach. Geographical Analysis Tang UW and Wang ZS (2007) Influences of urban forms
48(2): 132–156 on traffic-induced noise and air pollution: Results from
Santamouris M, Cartalis C, Synnefa A and Kolokotsa D a modelling system. Environmental Modelling & Soft-
(2015) On the impact of urban heat island and global ware 22(12): 1750–1764.
warming on the power demand and electricity con- Tian G, Ma B, Xu X, Liu X, Xu L, Liu X, Xiao L and Kong
sumption of buildings: A review. Energy and Buildings L (2016) Simulation of urban expansion and encroach-
98: 119–124. ment using cellular automata and multi-agent system
Santé I, Garcı́a AM, Miranda D and Crecente R (2010) model: A case study of Tianjin Metropolitan Region,
Cellular automata models for the simulation of China. Ecological Indicators 70: 439–450.
real-world urban processes: A review and analysis. Tobler WR (1975) Linear operators applied to areal data.
Landscape and Urban Planning 96(2): 108–122. In: Davis JC and McCullaugh MJ (eds) Display and
Schwarz N, Haase D and Seppelt R (2010) Omnipresent Analysis of Spatial Data. New York: John Wiley &
sprawl? A review of urban simulation models with Sons, 14–37.
respect to urban shrinkage. Environment and Planning Tobler WR (1979) Cellular geography. In: Gale S and
B: Planning and Design 37(2): 265–283. Olsson G (eds) Philosophy in Geography. Dordrecht:
Semboloni F (2000) The dynamic of an urban cellular Springer, 379–386.
automata model in a 3D spatial pattern. National Con- Torrens PM and Benenson I (2005) Geographic automata
ference Aisre, Proceedings, XXI, Regional and Urban systems. International Journal of Geographical Infor-
Growth in a Global Market. Palermo, Italy. mation Science 19(4): 385–412.
Semboloni F, Assfalg J, Armeni S, Gianassi R and Marsoni Torrens PM and Nara A (2007) Modeling gentrification
F (2004) CityDev, an interactive multi-agents urban dynamics: A hybrid approach. Computers, Environ-
model on the web. Computers, Environment and Urban ment and Urban Systems 31(3): 337–361.
Systems 28(1–2): 45–64. Torrens PM and O’Sullivan D (2001) Cellular automata
Shirley LJ and Battaglia LL (2008) Projecting fine resolu- and urban simulation: Where do we go from here?
tion land-cover dynamics for a rapidly changing terres- Environment and Planning B: Planning and Design
trial–aquatic transition in Terrebonne Basin, Louisiana, 28(2): 163–168.
USA. Journal of Coastal Research 24(6): 1545–1554. Václavı́k T and Rogan J (2009) Identifying trends in land
Silva EA and Clarke KC (2005) Complexity, emergence use/land cover changes in the context of post-socialist
and cellular urban models: Lessons learned from apply- transformation in central Europe: A case study of the
ing SLEUTH to two Portuguese metropolitan areas. greater Olomouc Region, Czech Republic. GIScience
European Planning Studies 13(1): 93–115. & Remote Sensing 46(1): 54–76.
Smith N (1987) Gentrification and the rent gap. Annals of Van Vliet J, White R and Dragićević S (2009) Modeling urban
the Association of American Geographers 77(3): growth using a variable grid cellular automaton. Comput-
462–465. ers, Environment and Urban Systems 33(1): 35–43.
24 Progress in Human Geography 45(1)
Verburg PH and Overmars KP (2007) Dynamic simulation Queensland, Australia. She is a Quantitative
of land-use change trajectories with the Clue-S model. Human Geographer with research focus on cities
In: Koomen E, Stillwell J, Bakema A and Henk JS (eds) and urban analytics, spatial modelling and geo-
Modelling Land-Use Change: Progress and Applica- simulation, and the application of GIS in spatial
tions. Dordrecht: Springer, 321–337. planning, policy analysis and spatially integrated
Waddell P (2002) UrbanSim: Modeling urban develop- social geographical studies. She convenes the Spa-
ment for land use, transportation, and environmental tio-Temporal Analytics Research Laboratory
planning. Journal of the American Planning Associa- (STAR Lab) at UQ and is the co-editor of the Jour-
tion 68(3): 297–314. nal of Spatial Science.
Waddell P (2011) Integrated land use and transportation
planning and modeling: Addressing challenges in Michael Batty is Bartlett Professor of Planning at
research and practice. Transport Reviews 31(2): the Centre for Advanced Spatial Analysis, Univer-
209–229. sity College London, UK. His research work
Wahyudi A, Liu Y and Corcoran J (2019a) Simulating the involves the development of computer models of
impact of developers’ capital possession on urban cities and regions, and he has published many books
development across a megacity: An agent-based and articles in this area. His books Cities and Com-
approach. Environment and Planning B: Urban Analy- plexity (MIT Press, 2005) and The New Science of
tics and City Science. DOI: https://doi.org/10.1177/ Cities (MIT Press, 2013) both won the Alonso Prize
2399808319875983. of the Regional Science Association in 2010, and
Wahyudi A, Liu Y and Corcoran J (2019b) Generating 2017. His most recent book is Inventing Future Cit-
different urban land configurations based on heteroge- ies (MIT Press, 2018). He is editor of the journal
neous decisions of private land developers: An agent- Environment and Planning B: Urban Analytics and
based approach in a developing country context. ISPRS City Science.
International Journal of GeoInformation 8: 229. DOI:
10.3390/ijgi8050229. Siqin Wang holds a PhD in Human Geography
White R, Engelen G and Uljee I (2015) Modeling Cities within the School of Earth and Environmental
and Regions as Complex Systems: From Theory to Sciences at The University of Queensland, Australia.
Planning Applications. Cambridge, MA: MIT Press. Her research interests focus on urban geography,
Wiechmann T and Pallagst KM (2012) Urban shrinkage in migration, 3D visualisation, spatial analysis and urban
Germany and the USA: A comparison of transforma- modelling. Her publications cover topics including
tion patterns and local strategies. International Journal migration and residential mobility, residential path-
of Urban and Regional Research 36(2): 261–280. ways, big data application in transport planning,
Wu F and Webster CJ (2000) Simulating artificial cities in neighbour complaints, and ageing issues in the inter-
a GIS environment: urban growth under alternative disciplinary field of urban geography and sociology.
regulation regimes. International Journal of Geogra-
phical Information Science 14(7): 625–648. Jonathan Corcoran is Professor in Human Geo-
Wu X, Zhu X, Wu GQ and Ding W (2014) Data mining graphy within the School of Earth and Environ-
with big data. IEEE Transactions on Knowledge and mental Sciences at The University of Queensland,
Data Engineering 26(1): 97–107. Australia, and the director of the Queensland Cen-
Zheng HW, Shen GQ, Wang H and Hong J (2015) Simu- tre for Population Research. His research interests
lating land use change in urban renewal areas: A case lie in the fields of population geography, spatial
study in Hong Kong. Habitat International 46: 23–34. science and regional science. His publications
cover topics including human mobility and
migration, human capital, and social problems,
Author biographies
each of which has a focus on quantitative meth-
Yan Liu is Associate Professor in Geographical ods. He is the co-editor of Australian Population
Information Science within the School of Earth and Studies and Secretary of the Applied Geography
Environmental Sciences at The University of Commission.