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

Abstract

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.

Introduction

Urban growth models allow for the analysis and extrapolation of the dynamics of city growth, which is highly beneficial for researchers and planners. Among urban growth models, cellular automata (CA) are particularly relevant because of their ability to reproduce complex spatial and temporal dynamics at a global scale using local rules. These rules operate in the neighborhood of the cells of a lattice that represents the space in which the simulated processes take place. Transition rules are applied at discrete time steps and determine the state of each cell in the lattice in every iteration of the model based on the state of its neighboring cells.

One of the main advantages of CA is their ability to reproduce emergent complex dynamics such as those found in cities, based on simple rules (Silva, 2010, White and Engelen, 1993). Moreover, because CA operate on a lattice, raster-format geographic data can be incorporated into the simulation and integrated in a GIS to facilitate the visualization and interpretation of results.

Santé, García, Miranda, and Crecente (2010) reviewed the main operational urban CA models applied to real-world urban development processes and confirmed the model applied by Xie (1996) in Amherst, New York, as one of the first applications of urban CA to the simulation of real-world cases. However, the first widespread empirical applications of CA were the model of White, Engelen, and Uljee (1997) and SLEUTH (Clarke, Hoppen, & Gaydos, 1997). The first one is based on the model developed by White and Engelen, 1993, White and Engelen, 1997. A number of models based on White and Engelen’s model, were applied to The Netherlands (Engelen, Geertman, Smits, & Wessels, 1999), San Diego (Kocabas & Dragicevic, 2006), Dublin (Barredo, Kasanko, McCormick, & Lavalle, 2003), Lagos (Barredo, Demichelli, Lavalle, Kasanko, & McCormick, 2004) or Tokyo (Arai & Akiyama, 2004). SLEUTH is a pattern-extrapolation model that considers four types of urban growth and has been frequently applied to North American cities such as San Francisco, Washington/Baltimore (Clarke & Gaydos, 1998), Sioux Fall (Goldstein, 2003), San Joaquin county (Dietzel & Clarke, 2006) or Phoenix (Berling-Wolff & Wu, 2004), but also to European (Silva & Clarke, 2002), South American (Leao, Bishop, & Evans, 2004) or Asiatic (Mahiny & Gholamalifard, 2007) regions.

Other well-known models are the models developed by Wu, 2002, Wu and Webster, 1998 and Wu and Martin (2002) focused on the calculation of the probability of development for every cell according to a number of factors, among which the neighborhood. To reduce subjectivity in the allocation of weights to factors, Wu and Webster (1998) used multicriteria evaluation techniques, whereas Wu (2002) used logistic regressions. The model built by Wu (2002) was used by other authors, who calibrated the CA using new methods such as genetic algorithms (Li, Yang, & Liu, 2007) or support vector machines (Yang, Li, & Shi, 2008). Although the aforementioned models are probably the most frequent, a wide variety of urban CA may be found in the literature based on neural-network (Li & Yeh, 2002a), statistical techniques (Li & Yeh, 2002b), probabilistic methods (Almeida et al., 2003), optimization algorithms (Liu, Li, Liu, He, & Ai, 2008), etc.

This paper assesses the feasibility of some of the best-known examples of urban CA for the simulation of urban growth in the town of Ribadeo, located in Galicia, a region in NW Spain. Ribadeo is a small urban settlement in an intermediate functional range between Galician large urban areas and rural areas. Ribadeo has experienced a slow urban growth process in the last 30 years which took place in relatively small scattered plots. This kind of urban growth is quite different from those which are commonly simulated with urban CA models. Most examples found in literature deal with regions which are experimenting high growth rates in large urban patches, where it is relatively easier to make generalizations and extrapolate processes than in slow growth areas because there is more information on urban processes.

In this paper, the theoretical basis for the urban CA models selected to perform the comparative analysis is presented, the study area and the methods are described, and the simulation results are discussed. Finally, the conclusions drawn from the analysis of the capability of the models in simulating urban growth in Ribadeo are presented.