Cellular automata for simulating land use changes based on support vector machines

Cellular automata (CA) have been increasingly used to simulate urban sprawl and land use dynamics. A major issue in CA is defining appropriate transition rules based on training data. Linear boundaries have been widely used to define the rules. However, urban land use dynamics and many other geographical phenomena are highly complex and require nonlinear boundaries for the rules. In this study, we tested the support vector machines (SVM) as a method for constructing nonlinear transition rules for CA. SVM is good at dealing with nonlinear complex relationships. Its basic idea is to project input vectors to a higher dimensional Hilbert feature space, in which an optimal classifying hyperplane can be constructed through structural risk minimization and margin maximization. The optimal hyperplane is unique and its optimality is global. The proposed SVM-CA model was implemented using Visual Basic, ArcObjects^®, and OSU-SVM. A case study simulating the urban development in the Shenzhen City, China demonstrates that the proposed model can achieve high accuracy and overcome some limitations of existing CA models in simulating complex urban systems.     

Applicability and calibration of an irregular cellular automata model for land use change

Cellular automata (CA) models of spatial change have been developed and applied in the context of large regional or metropolitan areas and usually use regular cells, with spatial interactions and transition rules operating within fixed-size neighbourhoods. Model calibration has also been an area of intensive research with many models still using expert-based input to ensure visual calibration of modelled land use maps. In this paper, we present an innovative CA model where irregular cells and variable neighbourhoods are used to better represent space and spatial interaction. Calibration is based on an optimisation procedure that uses particle swarm (PS) to determine the optimal set of parameters of the CA model. Hypothetical test instances are used to assess the CA model and its calibration to small urban areas. Our conclusion was that the use of PS ensures calibration results for the CA model that compare very well with results obtained through other approaches reported in the literature.     

Modelling multi-regional urban growth with multilevel logistic cellular automata

Simulation models based on cellular automata (CA) are useful for revealing the complex mechanisms and processes involved in urban growth and have become supplementary tools for urban land use planning and management. Although the urban growth mechanism is characterized by multilevel and spatiotemporal heterogeneity, most existing studies focus only on simulating the urban growth of singular regions without considering the heterogeneity of the urban growth process and the multilevel factors driving urban growth within regions that consist of multiple subregions. Thus, urban growth models have limited performance when simulating the urban growth of multi-regional areas. To address this issue, we propose a multilevel logistic CA model (MLCA) by incorporating a multilevel logistic regression model into the traditional logistic CA model (LCA). In the MLCA, multilevel driving factors are considered, and the multilevel logistic model allows the transition rules to not only vary in space, but also change when the subregional level factors change. To verify the MLCA's validity, it was applied to simulate the urban growth of Tongshan County, located in China's Xuzhou Prefecture. The results were compared with three comparative models, LCA1, which only considered grid cell-level factors; LCA2, which considered both grid cell- and subregional-level factors; and artificial neural network CA. Urban growth data for the periods 2000–2009 and 2009–2017 were used. The results show that the MLCA performs better on both visual comparison and indicators for accuracy verification. The Kappa of the results increased by <5%, but the improvement was significant, while increases for the accuracy of urban land and figure of merit were much higher than 5%. In addition, the results of MLCA had the smallest mean absolute percentage error when allocating new urban land areas to the various subregions. The results reveal that higher-level (e.g., town level) factors either strengthened or weakened the effects of grid cell-level factors on urban growth, which indirectly affected the spatial allocation of new urban land. The MLCA model is an effective step towards simulating nonstationary urban growth of multi-regional areas, using the comprehensive effects of multilevel driving factors.     

Modeling urban growth by coupling localized spatio-temporal association analysis and binary logistic regression

Understanding and forecasting the dynamics of urban growth can be helpful for making sustainable land-use policies. Computing models can simulate urban growth but many require extensive data input, which cannot be always met. Here we proposed coupling localized spatio-temporal association (LSTA) analysis and binary logistic regression (BLR) to model urban growth from historical land cover configurations. An indicator called neighborhood aggregation index (NAI) was defined first to measure configuration enrichment for any land cover type under spatial-and-temporal contexts. Multiple NAIs for different land cover types were taken into the proposed LSTA-BLR model to project future urban growth. A case study was selected in Wuhan, China where land covers were classified for each year during 2014–2017 based on the Landsat Imagery from Google Earth Engine. Urban growth from the year 2016 to 2017 was extracted from the classified land cover maps as the dependent variable which was modeled by the LSTA-BLR using predictors of the NAIs from the previous years. The LSTA-BLR models were tested under different neighborhood sizes (3 × 3, 5 × 5, 7 × 7, 9 × 9, and 11 × 11) and time windows (2016, 2015–2016, and 2014–2016). Results indicated that the best accuracy of the modeled urban growth reached 72.9% under the setting of 5 × 5 neighborhood size and time window 2014–2016. Urbanization was most likely to occur close to previously urbanized areas and unlikely near the neighborhood of enriched forest and water bodies. The neighborhood size affected the modeled result and the time window defining the NAIs had the most significant impact on model performance. We conclude that prior land cover configurations should be integrated for mapping future urban growth and the proposed LSTA-BLR model can serve as a useful tool to understand the near-future urbanization process based on the historical land cover configurations.     

A method to analyse neighbourhood characteristics of land use patterns

Neighbourhood interactions between land use types are often included in the spatially explicit analysis of land use change. Especially in the context of urban growth, neighbourhood interactions are often addressed both in theories for urban development and in dynamic models of (urban) land use change. Neighbourhood interactions are one of the main driving factors in a large group of land use change models based on cellular automata (CA).This paper introduces a method to analyse the neighbourhood characteristics of land use. For every location in a rectangular grid the enrichment of the neighbourhood by specific land use types is studied. An application of the method for the Netherlands indicates that different land use types have clearly distinct neighbourhood characteristics. Land use conversions can be explained, for a large part, by the occurrence of land uses in the neighbourhood.The neighbourhood characterization introduced in this paper can help to further unravel the processes of land use change allocation and assist in the definition of transition rules for cellular automata and other land use change models.     

A comparative analysis of cellular automata models for simulation of small urban areas in Galicia,NW Spain

Urban growth models developed in the second half of the 20th century have allowed for a better understanding of the dynamics of urban growth. Among these models, cellular automata (CA) have become particularly relevant because of their ability to reproduce complex spatial and temporal dynamics at a global scale using local and simple rules. In the last three decades, many urban CA models that proved useful in the simulation of urban growth in large cities have been implemented. This paper analyzes the ability of some of the main urban CA models to simulate growth in a study area with different characteristics from those in which these models have been commonly applied, such as slow and low urban growth. The comparison of simulation results has allowed us to analyze the strengths and weaknesses of each model and to identify the models that are best suited to the characteristics of the study area. Results suggest that models which simulate several land uses can capture better land use dynamics in the study area but need more objective and reliable calibration methods.     

A stochastically constrained cellular model of urban growth

Recent approaches to modeling urban growth use the notion that urban development can be conceived as a self-organizing system in which natural constraints and institutional controls (land-use policies) temper the way in which local decision-making processes produce macroscopic patterns of urban form. In this paper a cellular automata (CA) model that simulates local decision-making processes associated with fine-scale urban form is developed and used to explore the notion of urban systems as self-organizing phenomenon. The CA model is integrated with a stochastic constraint model that incorporates broad-scale factors that modify or constrain urban growth. Local neighborhood access rules are applied within a broader neighborhood in which friction-of-distance limitations and constraints associated with socio-economic and bio-physical variables are stochastically realized. The model provides a means for simulating the different land-use scenarios that may result from alternative land-use policies. Application results are presented for possible growth scenarios in a rapidly urbanizing region in south east Queensland, Australia.     

Patch-based cellular automata model of urban growth simulation: Integrating feedback between quantitative composition and spatial configuration

Urban land use change modeling can enhance our understanding of processes and patterns of urban growth that emerge from human-environment interactions. Cellular automata (CA) is a common approach for urban land use change modeling that allows for discovering and analyzing potential urban growth pathways through scenario building. Fundamental components of CA such as neighborhood configuration, transition rules, and representation of geographic entities have been examined in depth in the literature. However, trade-offs in the quantitative composition that urban gains from different non-urban land types and their dynamic feedback with the spatial configuration of urban growth are often ignored. The urban CA model proposed in this study links the quantitative composition with the spatial configuration of urban growth by incorporating a trade-off mechanism that adaptively adjusts the combined suitability of occurrence for non-urban land types based on analysis of transition intensity. Besides, a patch growing module based on seeding and scanning mechanisms is used to simulate the occurrence and spreading of spontaneous urban growth, and a time Monte Carlo (TMC) simulation method is employed to represent uncertainties in the decision-making process of urban development. Application of the model in an ecologically representative city, Ezhou, China, reveals improvement on model performance when feedback between the quantitative composition and spatial configuration of urban growth is incorporated. The averaged figure of merit and K-fuzzy indices are 0.5354 and 0.1954 with respect to cell-level agreement and pattern similarity, indicating the utility and reliability of the proposed model for the simulation of realistic urban growth.     

Modeling urban growth using spatially heterogeneous cellular automata models: Comparison of spatial lag,spatial error and GWR

Many methods can be used to construct geographical cellular automata (CA) models of urban land use, but most do not adequately capture spatial heterogeneity in urban dynamics. Spatial regression is particularly appropriate to address the problem to reproduce urban patterns. To examine the advantages and disadvantages of spatial regression, we compare a spatial lag CA model (SLM-CA), a spatial error CA model (SEM-CA) and a geographically-weighted regression CA model (GWR-CA) by simulating urban growth at Nanjing, China. Each CA model is calibrated from 1995 to 2005 and validated from 2005 to 2015. Among these, SLM and SEM are spatial autoregressive (SAR) models that consider spatial autocorrelation of urban growth and yield highly similar land transition probability maps. Both SAR-CA and GWR-CA accurately reproduce urban growth at Nanjing during the calibration and validation phases, yielding overall accuracies (OAs) exceeding 94% and 85%, respectively. SAR-CA is superior in simulating urban growth when measured by OA and figure-of-merit (FOM) while GWR-CA is superior regarding the ability to address spatial heterogeneity. A concentric ring buffer-based assessment shows OA valleys that correspond to FOM peaks, where the ranges of valleys and peaks indicate the areas with active urban development. By comparison, SAR-CA captures more newly-urbanized patches in highly-dense urban areas and shows better performance in terms of simulation accuracy; whereas, GWR-CA captures more in the suburbs and shows better ability to address spatial heterogeneity. Our results demonstrate that spatial regression can help produce accurate simulations of urban dynamics featured by spatial heterogeneity, either implicitly or explicitly. Our work should help select appropriate CA models of urban growth in different terrain and socioeconomic contexts.     

Uncertainty in hydromorphological and ecological modelling of lowland river floodplains resulting from land cover classification errors

Land cover maps provide essential input data for various hydromorphological and ecological models, but the effect of land cover classification errors on these models has not been quantified systematically. This paper presents the uncertainty in hydromorphological and ecological model output for a large lowland river depending on the classification accuracy (CA) of a land cover map. Using four different models, we quantified the uncertainty for the three distributaries of the Rhine River in The Netherlands with respect to: (1) hydrodynamics (WAQUA model), (2) annual average suspended sediment deposition (SEDIFLUX model), (3) ecotoxicological hazards of contaminated sediment for a bird of prey, and (4) floodplain importance for desired habitat types and species (BIO-SAFE model). We carried out two Monte Carlo (n = 15) analyses: one at a 69% land cover CA, the other at 95% CA. Subsequently we ran all four models with the 30 realizations as input.The error in the current land cover map gave an uncertainty in design water levels of up to 19 cm. Overbank sediment deposition varied up to 100% in the area bordering the main channel, but when aggregated to the whole study area, the variation in sediment trapping efficiency was negligible. The ecotoxicological hazards, represented by the fraction of Little Owl habitat with potential cadmium exposure levels exceeding a corresponding toxicity threshold of 148 μg d⁻¹, varied between 54 and 60%, aggregated over the distributaries. The 68% confidence interval of floodplain importance for protected and endangered species varied between 10 and 15%. Increasing the classification accuracy to 95% significantly lowered the uncertainty of all models applied. Compared to landscaping measures, the effects due to the uncertainty in the land cover map are of the same order of magnitude. Given high financial costs of these landscaping measures, increasing the classification accuracy of land cover maps is a prerequisite for improving the assessment of the efficiency of landscaping measures.     

Modeling urban growth in Atlanta using logistic regression

This study applied logistic regression to model urban growth in the Atlanta Metropolitan Area of Georgia in a GIS environment and to discover the relationship between urban growth and the driving forces. Historical land use/cover data of Atlanta were extracted from the 1987 and 1997 Landsat TM images. Multi-resolution calibration of a series of logistic regression models was conducted from 50 m to 300 m at intervals of 25 m. A fractal analysis pointed to 225 m as the optimal resolution of modeling. The following two groups of factors were found to affect urban growth in different degrees as indicated by odd ratios: (1) population density, distances to nearest urban clusters, activity centers and roads, and high/low density urban uses (all with odds ratios < 1); and (2) distance to the CBD, number of urban cells within a 7 × 7 cell window, bare land, crop/grass land, forest, and UTM northing coordinate (all with odds ratios > 1). A map of urban growth probability was calculated and used to predict future urban patterns. Relative operating characteristic (ROC) value of 0.85 indicates that the probability map is valid. It was concluded that despite logistic regression’s lack of temporal dynamics, it was spatially explicit and suitable for multi-scale analysis, and most importantly, allowed much deeper understanding of the forces driving the growth and the formation of the urban spatial pattern.     

Learning cellular automata rules for binary classification problem

This paper proposes a cellular automata-based solution of a binary classification problem. The proposed method is based on a two-dimensional, three-state cellular automaton (CA) with the von Neumann neighborhood. Since the number of possible CA rules (potential CA-based classifiers) is huge, searching efficient rules is conducted with use of a genetic algorithm (GA). Experiments show an excellent performance of discovered rules in solving the classification problem. The best found rules perform better than the heuristic CA rule designed by a human and also better than one of the most widely used statistical method: the k-nearest neighbors algorithm (k-NN). Experiments show that CAs rules can be successfully reused in the process of searching new rules.     

Linking satellite images and hand-held infrared thermography to observed neighborhood climate conditions

While satellite images effectively show surface urban heat islands in urbanized areas, linking surface temperatures to actual ambient temperatures remains a research challenge. Microclimates in urbanized settings can vary tremendously in very short distances, making adequate climate interpolations across a large metropolitan area difficult, at best. This study links the coarse scale of satellite (ASTER) images to the fine scale of hand-held thermography as part of an in-depth suburban neighborhood climate study to determine if ASTER imaging can be used to adequately estimate neighborhood climate conditions in an urbanized area. The study utilizes day and night remotely-sensed and ground data from June, 2004 for Phoenix, Arizona. Microclimate conditions of three urban fringe neighborhoods with varying amounts of natural vegetation and development density were studied, along with ASTER remote sensing data, mobile climate transects, and spot infrared thermographic images.These neighborhoods, though variable, showed only minor differences, and the study indicates that daytime images (11:20 am) do not adequately rank observed conditions within these neighborhoods — the highest ASTER surface temperatures were recorded for the least-dense neighborhood with a natural desert landscaping, though lowest ambient temperatures were measured there. Daytime mean surface temperatures versus air temperatures were 50.4 °C (30.8 °C air temp); 53.5 °C (29.7 °C); and 50.6 °C (31.9 °C). It was found that nighttime (10:40 pm LST) differences among neighborhoods of surface and air temperatures were relatively consistent, with the most densely developed neighborhood having the highest ASTER surface temperatures (29.0 °C) and transect-derived air temperatures (30.0 °C). Issues of view angle, shadowing, emissivity, resolution, and wind conditions for daytime results with their relatively small mean differences observed across the neighborhoods may explain why the rank of ASTER thermal conditions versus observed ambient conditions was poor. However, following sunset, these issues of view angle, etc., are much less problematic.     

A linguistic cellular automata simulation approach for sustainable land development in a fast growing region

Rapid rural-urban land conversion as a consequence of economic growth has raised serious concerns over sustainable development. There is an urgent need to understand what possible urban scenarios can result from different policies towards land conversions. In many ways, the question resembles the exploration of a self-organising phenomenon which generates macroscopic patterns upon microscopic and local decision-making processes. In this paper a linguistic simulation approach has been developed. As a prototype, the study integrates cellular automata (CA) with heuristically-defined transition rules to simulate land use conversions in the rural-urban fringe of a fast growing metropolis. Fuzzy set theory has been applied to capture uncoordinated land development process. An innovative feature of the integrated approach lies in its definition of transition rules through a “natural language interface”, thus being more realistic and transparent. The model can simulate development scenarios in a gaming style. By providing a series of scenarios, it reveals risks inherent in certain development strategies which may jeopardize sustainable development of the city.     

A segment derived patch-based logistic cellular automata for urban growth modeling with heuristic rules

Cellular automata (CA) models are extensively applied in urban growth modeling in different forms (i.e., pixel or patch). Studies have reported that the patch-based approach can achieve a more realistic urban landscape. However, they are subjected to uncertainties due to a variety of stochastic processes involved, which weakens their effectiveness on urban planning or decision making. Here, we propose a new patch-based urban growth model with heuristic rules that employed logistic CA model with a watershed segmentation algorithm (Segmentation-Patch-CA). The segment objects derived from features of urban CA model were regarded as potential patches for conversion, through defining a utility function that considered both the suitability and heterogeneity of pixels within the patch. Thereafter, two different urban growth types, i.e., organic growth and spontaneous growth, were identified and simulated separately by introducing a landscape expansion index (LEI) that built on neighborhood density analysis. The proposed Segmentation-Patch-CA was applied to Guangzhou City, China. Our results revealed that the proposed model produced a more realistic urban landscape (96.00% and 97.38%) than pixel-based (45.14% and 74.82%) for two modeling periods 2003–2008 and 2008–2012, respectively, when referring to an assembled indicator that closely related to urban patterns (e.g., shape, size, or distribution). Meanwhile, it also achieved a good performance when comparing to other patch-based urban CA models but with less uncertainty. Our model provided a very flexible framework to incorporate patches using segments or self-growth based on pixels, which is very helpful to future urban planning practices.     

Comparing three global parametric and local non-parametric models to simulate land use change in diverse areas of the world

This paper compares one global parametric land use change model, the artificial neural network – based Land Transformation Model, with two local non-parametric models: a classification and regression tree and multivariate adaptive regression spline model. We parameterized these three models with identical data from different regions of the world; one region undergoing extensive agricultural expansion (East Africa), another region where forests are re-growing (Muskegon River Watershed in the United States), and a third region where urbanization is prominent (South-Eastern Wisconsin in the United States). Independent training data and testing data were used to train and calibrate each model, respectively. Comparisons of simulated maps from the three kinds of land use change patterns were made using conventional goodness-of-fit metrics in land use change science. The results of temporal and spatial comparison of the data mining models show that the artificial neural network outperformed all other models in a short-time interval (East Africa; 5 years) and for coarse resolution data (East Africa; 1 km); however, the three data mining models obtained similar accuracies in a long-time interval (Muskegon River Watershed; 20 years) and for fine resolution data with large numbers of cells (Muskegon River Watershed; 30 m). Furthermore, the results showed that the probability of agriculture gain was more likely in locations closer to towns and large cities in East Africa, urbanization was more likely in locations closer to roads and urban areas in South-Eastern Wisconsin and the probability of forest gain was more likely in locations closer to the forest and shrub land cover and farther away from roads in Muskegon River Watershed.     

Collaborative optimization of rural residential land consolidation and urban construction land expansion: A case study of Huangpi in Wuhan,China

The synergetic development of urban and rural construction land is always an important issue. We propose a collaborative optimization model (COMRU) of rural residential land consolidation and urban construction land expansion, which is a coupling model of cellular automata (CA), genetic algorithms (GA), and the Lewis turning point theory. This model regards the rural population transfer as a scenario and generates a new quantity and space allocation system for the population and the land-use types in the study area. The optimized result will balance the development of urban and rural construction lands to ultimately reduce the income gap between urban and rural areas and promote the rationality of the spatial distribution of urban and rural construction lands. We applied COMRU to Huangpi District in the city of Wuhan, the capital of Hubei Province, People's Republic of China and obtained three important results: (1) After optimization, the scattered rural settlements were effectively consolidated and large amounts of land resources were released, thereby supplementing cultivated and urban construction lands; (2) The urban–rural income ratio decreased significantly, indicating a considerable reduction in the income gap between the urban and rural areas; (3) The structure and function of the construction lands were improved, leading to the improved equity of urban and rural public services. The final space optimization allocation program generated by COMRU provides a reference for the sequence of rural settlement consolidation and urban spatial planning.     

Calibration of cellular automata urban growth models from urban genesis onwards - a novel application of Markov chain Monte Carlo approximate Bayesian computation

Cellular Automata (CA) models are widely used to study spatial dynamics of urban growth and evolving patterns of land use. One complication across CA approaches is the relatively short period of data available for calibration, providing sparse information on patterns of change and presenting problematic signal-to-noise ratios. To overcome the problem of short-term calibration, this study investigates a novel approach in which the model is calibrated based on the urban morphological patterns that emerge from a simulation starting from urban genesis, i.e., a land cover map completely void of urban land. The application of the model uses the calibrated parameters to simulate urban growth forward in time from a known urban configuration.This approach to calibration is embedded in a new framework for the calibration and validation of a Constrained Cellular Automata (CCA) model of urban growth. The investigated model uses just four parameters to reflect processes of spatial agglomeration and preservation of scarce non-urban land at multiple spatial scales and makes no use of ancillary layers such as zoning, accessibility, and physical suitability. As there are no anchor points that guide urban growth to specific locations, the parameter estimation uses a goodness-of-fit (GOF) measure that compares the built density distribution inspired by the literature on fractal urban form. The model calibration is a novel application of Markov Chain Monte Carlo Approximate Bayesian Computation (MCMC-ABC). This method provides an empirical distribution of parameter values that reflects model uncertainty. The validation uses multiple samples from the estimated parameters to quantify the propagation of model uncertainty to the validation measures.The framework is applied to two UK towns (Oxford and Swindon). The results, including cross-application of parameters, show that the models effectively capture the different urban growth patterns of both towns. For Oxford, the CCA correctly produces the pattern of scattered growth in the periphery, and for Swindon, the pattern of compact, concentric growth. The ability to identify different modes of growth has both a theoretical and practical significance. Existing land use patterns can be an important indicator of future trajectories. Planners can be provided with insight in alternative future trajectories, available decision space, and the cumulative effect of parcel-by-parcel planning decisions.     

Understanding the drivers of sustainable land expansion using a patch-generating land use simulation (PLUS) model: A case study in Wuhan,China

Cellular Automata (CA) are widely used to model the dynamics within complex land use and land cover (LULC) systems. Past CA model research has focused on improving the technical modeling procedures, and only a few studies have sought to improve our understanding of the nonlinear relationships that underlie LULC change. Many CA models lack the ability to simulate the detailed patch evolution of multiple land use types. This study introduces a patch-generating land use simulation (PLUS) model that integrates a land expansion analysis strategy and a CA model based on multi-type random patch seeds. These were used to understand the drivers of land expansion and to investigate the landscape dynamics in Wuhan, China. The proposed model achieved a higher simulation accuracy and more similar landscape pattern metrics to the true landscape than other CA models tested. The land expansion analysis strategy also uncovered some underlying transition rules, such as that grassland is most likely to be found where it is not strongly impacted by human activities, and that deciduous forest areas tend to grow adjacent to arterial roads. We also projected the structure of land use under different optimizing scenarios for 2035 by combining the proposed model with multi-objective programming. The results indicate that the proposed model can help policymakers to manage future land use dynamics and so to realize more sustainable land use patterns for future development. Software for PLUS has been made available at https://github.com/HPSCIL/Patch-generating_Land_Use_Simulation_Model     

Neighborhood detection and rule selection from cellular automatapatterns

Using genetic algorithms (GAs) to search for cellular automation (CA) rules from spatio-temporal patterns produced in CA evolution is usually complicated and time-consuming when both, the neighborhood structure and the local rule are searched simultaneously. The complexity of this problem motivates the development of a new search which separates the neighborhood detection from the GA search. In the paper, the neighborhood is determined by independently selecting terms from a large term set on the basis of the contribution each term makes to the next state of the cell to be updated. The GA search is then started with a considerably smaller set of candidate rules pre-defined by the detected neighhorhood. This approach is tested over a large set of one-dimensional (1-D) and two-dimensional (2-D) CA rules. Simulation results illustrate the efficiency of the new algorithm