A Markov Chain Monte Carlo Cellular Automata Model to Simulate Urban Growth

This paper investigates the potential of a cellular automata (CA) model based on logistic regression (logit) and Markov Chain Monte Carlo (MCMC) to simulate the dynamics of urban growth. The model assesses urbanization likelihood based on (i) a set of urban development driving forces (calibrated based on logit) and (ii) the land-use of neighboring cells (calibrated based on MCMC). An innovative feature of this CA model is the incorporation of MCMC to automatically calibrate the CA neighborhood transition rules. The MCMC based CA model is applied to Wallonia region (Belgium) to simulate urban growth from 1990 to 2000 using Corine Land Cover data (CLC). The outcome of logit model is evaluated by the relative operating characteristic (ROC). The simulated map of 2000 is then validated against 2000 actual map based on cell-to-cell location agreement. The model outcomes are realistic and relatively accurate confirming the effectiveness of the proposed MCMC-CA approach. Keywordscellular automata; Markov chain Monte Carlo; logistic regression.