Modeling urban sprinkling with cellular automata
Introduction
The quantitative and qualitative analysis and prediction of land use transformation dynamics plays an increasingly central role in earth system science (Feng, 2017; Feng & Tong, 2020). Among these transformation dynamics, urban sprawl is undoubtedly amongst the most studied dynamics. Urban sprawl is usually associated with the development of mono-functional, low-density urban settlements occupying a large territory around medium- and large-sized cities (Galster et al., 2001; Hasse & Lathrop, 2003; Nechyba & Walsh, 2004). A similar, though somehow distinct, phenomenon of territorial transformation is observed in the proximity of smaller cities. This has been defined as urban sprinkling (Romano, Zullo, Fiorini, Ciabò, et al., 2017). Literally speaking, according to Merriam-Webster’s definition, sprinkling is defined as: “a small amount falling into drops or scattered particles”. Urban sprinkling is typical of rural areas and involves the construction of low-density, small-scale residential settlements, scattered throughout the territory, far from existing public services and characterized by low levels of accessibility. It is characterized by very low population density and building density rates (Romano, Zullo, Fiorini, Marucci, et al., 2017). It is usually associated with territorial fragmentation (Saganeiti et al., 2018). The configuration of urban sprinkling is extremely dispersed. It develops through heterogeneous and very small-sized aggregates, composed of a single building or a small group of buildings. Aggregates may welcome different uses, from residential to industrial, agricultural or commercial uses, which are inserted within the rural matrix. These aggregates typically develop along the existing road networks, which are usually characterized by an organic scheme. While existing cities tend to attract urban sprawl, which progressively consolidates over time, newly built areas would rather tend to develop farther away from existing urban settlements in urban sprinkling.
On the Italian territory, urban sprinkling is usually associated with weak urban planning. In some cases, it is the consequence of abusive initiatives, encouraged by amnesties for the violation of building regulations. In Belgium it is largely related to deficiencies of land planning documents adopted in the 1970’s, which provided room for the so-called “ribbon development” along secondary roads without much congruency with existing settlements (Halleux et al., 2012; Mustafa, Van Rompaey, et al., 2018; Verbeek et al., 2014). Uncontrolled, poorly managed and fragmented urban transformation generates an economic and social impact on the population also defined as a social cost. The social cost includes both direct costs arising from the construction of new roads and new infrastructure services, as well as indirect costs arising from health costs owing to increased pollution and therefore increased travel, costs on account of loss of landscape quality and other costs associated with the daily life of the local population (Jan K. Brueckner, 2001; Carruthers & Ulfarsson, 2003; Freilich & Peshoff, 1997).
It is of great interest to study the dynamics of territorial transformation, in particular those linked to urban sprinkling, so as to understand which factors (drivers) have most contributed to the development of the phenomenon. Moreover, the projection into the future (even if not a very distant one) of the transformation dynamics may help control the social costs that the population bears because of the urban sprinkling phenomenon. As demonstrated in Manganelli et al., 2020, the social costs increase, as urban sprinkling increases.
A fundamental prerequisite for developing or applying an urban transformation prediction model is certainly the knowledge of the factors that influence the process, i.e. the causative factors (drivers). Many studies (G. Li et al., 2018; Mustafa et al., 2017; Puertas et al., 2014; Rienow & Goetzke, 2015; J. Yang et al., 2020) are dedicated to the analysis of the drivers that regulate the transformation dynamics of a territory, i.e. the urbanization processes. Table A1 in the appendix shows a collection of 19 scientific articles from 2014 to 2020. For each one, in addition to other information discussed in the next paragraph, the drivers taken into consideration have been reported. The processes of expansion are influenced by geophysical, socio-economic and legislative conditions. Most research shows that economic factors, including population growth, income and value of agricultural land, are of primary importance in setting the rules for urban expansion (J. K. Brueckner & Fansler, 1983; McGrath, 2005). As a matter of fact, urban prediction models are generally applied in contexts where the population growth rate is positive and settlement density is relatively high. Among the reviewed articles, only Rienow et al. (Rienow & Goetzke, 2015) analyses a rural context with scattered settlements and a negative demographic trend in a German region.
Land use models allow to project and simulate future urban patterns in order to act on the dynamics and mechanisms of urban expansion (Wahyudi & Liu, 2016). Urban land use change models are generally analyzed and applied to provide a decision support to policy makers in the implementation of new urbanization plans (Allen & Lu, 2003; Nasiri et al., 2019; Puertas et al., 2014). Furthermore, these models can support planning policies such as flood risk mitigation (Mustafa, Bruwier, et al., 2018), regulation of climate change and the provision of ecosystem services (Shoyama & Yamagata, 2014) as well as the development of scenarios for environmental impact assessment (Poelmans & Van Rompaey, 2009). In this study, our focus is to cross the literature gap concerning the correlation between demographic trends and urban expansion in order to arrive at the conclusion that a predictive model can be used to control urban transformations where they are not necessary and to preserve those areas with a particular environmental value.
In this research, we have applied a modeling approach that integrates multinomial logistic regression (MLR) and cellular automata (CA) to analyze and project urban sprinkling. The model was proposed by Mustafa, Bruwier et al. (2018), Mustafa, Heppenstall, et al. (2018) and used to simulate urbanization scenarios in Belgium.
Our case study, the Basilicata region in southern Italy, is affected by the urban sprinkling phenomenon and a decoupled growth between demographics and built-up areas in the last 30 years (Saganeiti et al., 2018). This phenomenon is representative of a number of internal Western areas which, even though their population is in decline and/or static, have the necessary resources to continue to invest in urbanization processes (Caselli et al., 2020; Martinez-Fernandez et al., 2012; Wiechmann & Pallagst, 2012). A simulation and projection model of urban expansion based on a multi-density approach (4 urban density classes) will be used. This approach appears to be fundamental and novel for a context characterised by urban sprinkling or an urban expansion in the absence of population growth. Therefore, for urban expansion modeling, built-up density maps were generated on the basis of three regional building datasets (1989, 1998 and 2013) with four density classes: non built-up, low density, medium density and high density and used for the calibration, validation, and simulation phases. The transition probability for calibration (1989–1998) was calculated with MLR for the built-up causative factors and with multi-objective genetic algorithm (MOGA) for CA neighborhood interactions. Among the causative factors considered, are those concerning the physical, socio-economic, proximity and constraints factors. The calibrated parameters were used for the simulation of the 2013 map which was compared with the actual map of 2013 (validation). The final objective is to simulate business as usual urban pattern in 2030.
Why model future urban expansion scenarios? Undoubtedly, to understand the spatial processes of urban development dynamics over time and to serve as tools to project future policies focused on the principles of sustainable development. Geographic Information System (GIS) and remote sensing techniques are used to feed and construct urban expansion models. Among the common approaches: cellular automata, logistic regression, geographic weighted regression, Markov Chain integrated with cellular automata methods and agent-based. The ability of the cellular automata approach to simulate and predict land use changes assumes that any previous urban expansion, influences future characteristics through local and regional interactions between different types of land use.
CA models are easy to apply, can simulate complex models, have an open structure, integrate with other models and can simulate spatial and temporal models (Aburas et al., 2016). They overcome their limit of not being able to include drivers for urban expansion simulation, by integrating with other quantitative and space-time methods such as Analytic Hierarchy Process (AHP) and MLR. Table A1 lists 19 scientific articles on urban expansion modeling from 2014 to 2020. Due to the presence of several recent reviews of prediction models (Aburas et al., 2016; Dang & Kawasaki, 2016; Poelmans & Van Rompaey, 2010; van Vliet et al., 2016; Wahyudi & Liu, 2016), articles published before 2014 have not been considered in this summary. The search was conducted on February 2020, in the Scopus website and was carried out considering one of these keywords: “Urban growth”, “Model prevision”, “Prediction”, “Cellular Automata”, “Logistic regression”, “Scenario”, “Urban expansion”. Only articles from scientific journals in the disciplines of urban and territorial sciences, environmental engineering and earth sciences were selected. A further selection was made by reading the abstracts and finally 119 papers were collected.
In conclusion, for each year, the papers with the most citations were analyzed in detail and synthetized. Table 1 shows a summary of the main elements collected in Table A1 in the appendix where the following elements are reported:
- -
Authors: reference is made to the first author of the paper and the year of publication;
- -
Study case: the area of study analyzed is reported;
- -
Application: the type of application that was carried out in the paper such as: simulation and forecasting of urban expansion, study of the drivers that regulate the dynamics of transformation, flood risk mitigation, modelling of urban densification processes and others;
- -
models and techniques: used in research. Among the most common are certainly the models with cellular automata, logistic regression and artificial neural networks;
- -
predictive variables: this includes all variables that affect the dynamics of urban transformation. They have been divided into 6 categories: physical factors, proximity factors, social factors, neighborhood interactions and urban policies. For each article the box corresponding to the driver taken into consideration is highlighted in grey and, in most cases, the drivers are specified (with text);
- -
population trends and urban dynamics: in this column the demographic growth and urban expansion dynamics of the various case studies have been included. Population growth rates and urban expansion dynamics are highlighted.
The articles analyzed (16 papers out of 19) show a positive demographic trend and high population density indices, which confirm that urban expansion is usually associated with population growth. Only in the article by Rienow et al. (Rienow & Goetzke, 2015) is there a negative growth rate, here the case study was analyzed to be compared with another case study with positive population growth in the same article. Other cases of non-significant population growth concern the work of Mustafa et al. (Mustafa et al., 2017) which analyzes a case study with a low population growth rate and the article by Halmy et al. (Halmy et al., 2015) which contains no information on population growth. Concerning drivers influencing urban transformation dynamics, physical factors have been considered in all papers except in the Ku (Ku, 2016) and Halmy et al. who consider only proximity factors and urban planning policies. Elevation and slope are the most used physical factors. The case of Yang et al. (J. Yang et al., 2020) is particular as it also considers soil quality (in terms of habitat quality) among the physical factors. With the exception of Hua et al. (LIU & MA, 2011) and Martellozzo et al (Martellozzo et al., 2018) who used the SLEUTH model (Slope, Land Use, Exclusion, Urban Extent, Transportation and Hillshade), all the others considered proximity as a factor. The proximity to road networks is certainly amongst the most utilized. Among those used less, one can find the proximity to highway entrances (J. Yang et al., 2020) to irrigation canals (Halmy et al., 2015). Social and economic factors are not always considered as drivers for urban transformation dynamics. Among those most frequently used are population density, access to jobs, gross domestic product (GDP) and richness rates. Eleven papers in all consider interactions between neighbors from different neighborhood sizes ranging from 3 × 3–5 × 5. Only 10 papers consider urban planning policies as a driving force. On the basis of this data collection, the study area was modelled using a logistic regression with driving forces belonging to each identified category of the predictive variables.
Section snippets
Study area
The urban expansion model was applied to the entire territory of the Basilicata region in the south of Italy. It is a region characterized by a low population density and settlement. With 56 inhabitants per square kilometer it is the second to last Italian region, followed only by the mountainous Aosta Valley region with 39 inhabitants per square kilometer (data from the 15th national census of National Institute of Statistics -ISTAT - 2011 (Istat.it, 2011)). The territory is affected by a
Methods
Built-up density maps were used for the calibration phase (1989–1998) and validation phase (1998–2013). Two components were considered for the calculation of the transition potential from one density class to another for the calibration phase. The first component concerns the built-up development causative factors, calibrated with the MLR. The second component was the CA neighborhood effects that were calibrated using a MOGA as in Mustafa et al. (Mustafa, Heppenstall, et al., 2018). The
Results
In this section the results of the MLR, calibration and validation processes are listed and discussed. The VIF index was used for the multicollinearity test among the independent variables, with values < 2.58 implying no collinearities, so that all the 11 variables presented in Table 4 were introduced in MLR. The X9 and X10 variables refer respectively to the population density and employment rate for 2001. This data is collected at a municipal level and, in particular, the employment rate was
Discussion
The model applied in this study addresses the dynamics of urban expansion through a multi-density approach. It was possible to assess the different relationships between built-up causative factors and built-up density classes for each transition, with specific factors derived from MLR and CA coefficients for each density class.
Specifically, it allows to read and analyze the different impacts of drivers on the transformation modalities, especially on the expansion and densification processes.
One
Conclusions
Simulating future urban expansion scenarios is fundamental to understand the spatial model in action in a given territory. It can help policy makers adopt urban policies aimed at containing expansion. The aim of this work has been to project future business-as-usual scenario in urban sprinkling contexts in order to control and regulate urban transformation processes.
An innovative aspect of this study consists in considering urban expansion not in a dual way (built-up/not built-up) but along
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (72)
- et al.
The simulation and prediction of spatio-temporal urban growth trends using cellular automata models: A review
International Journal of Applied Earth Observation and Geoinformation
(2016) - et al.
Improving the capability of an integrated CA-Markov model to simulate spatio-temporal urban growth trends using an Analytical Hierarchy process and Frequency Ratio
International Journal of Applied Earth Observation and Geoinformation
(2017) - et al.
Land use changes modelling using advanced methods: Cellular automata and artificial neural networks. The spatial and explicit representation of land cover dynamics at the cross-border region scale
Applied Geography
(2014) - et al.
Performance-based spatial monitoring. An interpretative model for long-term shrinking medium-small Italian towns
Sustainable Cities and Society
(2020) - et al.
Capturing the varying effects of driving forces over time for the simulation of urban growth by using survival analysis and cellular automata
Landscape and Urban Planning
(2016) - et al.
The adaptive efficiency of land use planning measured by the control of urban sprawl. The cases of the Netherlands, Belgium and Poland
Land Use Policy
(2012) - et al.
Land use/land cover change detection and prediction in the north-western coastal desert of Egypt using Markov-CA
Applied Geography
(2015) - et al.
Land resource impact indicators of urban sprawl
Applied Geography
(2003) - et al.
Modeling urban growth in Atlanta using logistic regression
Computers, Environment and Urban Systems
(2007) Incorporating spatial regression model into cellular automata for simulating land use change
Applied Geography
(2016)
Examining spatiotemporally varying effects of urban expansion and the underlying driving factors
Sustainable Cities and Society
The varying driving forces of urban expansion in China: Insights from a spatial-temporal analysis
Landscape and Urban Planning
Modeling spatial variations of urban growth patterns in Chinese cities: The case of Nanjing
Landscape and Urban Planning
Modelling the impact of urban growth on agriculture and natural land in Italy to 2030
Applied Geography
More evidence on the spatial scale of cities
Journal of Urban Economics
Effects of spatial planning on future flood risks in urban environments
Journal of Environmental Management
Coupling agent-based, cellular automata and logistic regression into a hybrid urban expansion model (HUEM)
Land Use Policy
Modelling built-up expansion and densification with multinomial logistic regression, cellular automata and genetic algorithm
Computers, Environment and Urban Systems
Detecting and modelling spatial patterns of urban sprawl in highly fragmented areas: A case study in the Flanders–Brussels region
Landscape and Urban Planning
Complexity and performance of urban expansion models
Computers, Environment and Urban Systems
Land-cover change model validation by an ROC method for the Ipswich watershed, Massachusetts, USA
Agriculture, Ecosystems & Environment
Assessing spatial dynamics of urban growth using an integrated land use model. Application in Santiago metropolitan Area, 2010-2045
Land Use Policy
Supporting SLEUTH - Enhancing a cellular automaton with support vector machines for urban growth modeling
Computers, Environment and Urban Systems
Land transformation of Italy due to half a century of urbanization
Land Use Policy
Coupling machine learning, tree-based and statistical models with cellular automata to simulate urban growth
Computers, Environment and Urban Systems
Predicting land-use change for biodiversity conservation and climate-change mitigation and its effect on ecosystem services in a watershed in Japan
Ecosystem Services
Assessment of decoupling between rural settlement area and rural population in China
Land Use Policy
A review of current calibration and validation practices in land-change modeling
Environmental Modelling & Software
Presence and trends of linear sprawl: Explaining ribbon development in the north of Belgium
Landscape and Urban Planning
Monitoring and predicting land use change in Beijing using remote sensing and GIS
Landscape and Urban Planning
Patch-based cellular automata model of urban growth simulation: Integrating feedback between quantitative composition and spatial configuration
Computers, Environment and Urban Systems
Modeling urban expansion in the transnational area of Changbai Mountain: A scenario analysis based on the zoned Land Use Scenario Dynamics-urban model
Sustainable Cities and Society
Modeling and prediction of future urban growth in the Charleston region of South Carolina: A GIS-based integrated approach
Ecology and Society
The effects of urban policies on the development of urban areas
Sustainability
Using spatiotemporal analysis in urban sprawl assessment and prediction
Log-linear models, extensions, and applications
Cited by (31)
Modeling gridded urban fractional change using the temporal context information in the urban cellular automata model
2023, CitiesCitation Excerpt :With the migration of the population from rural to urban areas, the urban lands in cities are dramatically expanded at the cost of encroaching surrounding natural lands (e.g., agricultural lands and forests), resulting in worldwide concerns about deforestation (DeFries, Rudel, Uriarte, & Hansen, 2010; Hu, Li, Gong, Yu, & Huang, 2020), ecosystem (Alberti et al., 2017), food security (Foley et al., 2005; Gong, 2011), thermal environments (Zhou, Zhao, Liu, Zhang, & Zhu, 2014), air pollutions and public health (Gong et al., 2012; Zhang, He, & Huo, 2012), energy consumption and carbon cycles (Güneralp et al., 2017; Karen C Seto et al., 2016; Zhang, Cai, & Hu, 2013). The latest report of the World Urbanization Prospects (United Nations, 2019) indicates the future urbanization pressure is relatively high in developing countries (or regions) in Africa and Asia, where the conservation of ecological environments deserves more attention than developed regions (Dinerstein et al., 2017; Gibb et al., 2020; Saganeiti, Mustafa, Teller, & Murgante, 2021). For the mitigation and adaptation of future climate change and global urbanization, the model-based approach is an extensively used tool to explore different urban growth scenarios for scientific decision-making and urban planning (Li & Gong, 2016; Santé, García, Miranda, & Crecente, 2010).
Modelling built-up land expansion probability using the integrated fuzzy logic and coupling coordination degree model
2023, Journal of Environmental ManagementUrban form simulation in 3D based on cellular automata and building objects generation
2022, Building and EnvironmentCitation Excerpt :Another stream of urban simulations adopt some established schemes of urban structure classification, such as the local climate zones (LCZ), and simulate the change under different scenarios [17]. Although traditional CA-based models feature the discrete representations of urban form [13], modifications have also been made to support the continuous representations in urban CA models. For instance, some models use ‘grey cells’ to represent urban land fractions [18,19], population density [12] or building height [20], thereby leveraging the simulation of the continuous change in urban form.
Quantitative assessment of local warming based on urban dynamics
2022, Multi-Chaos, Fractal and Multi-Fractional Artificial Intelligence of Different Complex SystemsIntegrating strategic planning intentions into land-change simulations: Designing and assessing scenarios for Bucharest
2022, Sustainable Cities and SocietyCitation Excerpt :Furthermore, while scenarios and simulations have proven useful in anticipating future land uses, it is often difficult to link these back to policies and management decisions. Results of scenario-based simulations further offer decision support for urban planning and policy makers (Feng et al., 2012; García-Ayllón, 2018), for instance through determining desired locations for urban development or minimizing negative impacts on other land uses (Karakus, Cerit & Kavak, 2015; Li & Yeh, 2000; Saganeiti et al., 2021; Zhai et al., 2020). Thus, simulations results could be further integrated into future plans as a way to perceive the future and to engage stakeholders in discussions and decision-making processes.