Quantifying and analyzing neighborhood configuration characteristics to cellular automata for land use simulation considering data source error

Cellular Automata (CA) simulation models have been increasingly used in land use studies. However, neighborhood configuration, an essential element of CA model, remarkably impacts the accuracy of simulated results. Moreover, errors from data source may propagate through the CA modeling process. The objective of this study is to analyze the effect of neighborhood configuration to CA model and further on to explore its capacity of resisting disturbance from data source error. With statistic-based CA model and several neighborhood configurations respectively, the land use changes of Wuhan, China were analyzed. It is demonstrated that there are significant differences on the simulated results produced by different neighborhoods. Besides, different neighborhoods respond differently to data source error. In light of these results, we find out that (1) neighborhood configurations with larger neighborhood size and planar neighborhood type, introduced in this paper, contribute to higher prediction accuracy; and (2) the neighborhood configurations above also have higher capacity of resisting disturbance from data source error and give rise to more stable simulated results. This study provides a comprehensive basis for scale selection of CA model with a meaningful consideration of data source error and thus will improve the research on land use change.