Influence of neighborhood distance on simulation accuracy of cellular automata model in land use

Simulation of land use change is often based on cellular automata (CA) modeling theory. The aspect of distances which is in authors' opinion the major parameter in a cellular automata has not been sufficiently discussed. In this paper, neighborhood module of CLUE-S model was used to simulate the land use change of Miyun County, Beijing from 1991 to 2004 and the simulating precision variation was analyzed due to the parameter of neighborhood distance change. Two types of neighborhood with six different distances, 200, 400, 600, 800, 1 000, 1 200 m, were applied to the model CLUE-S to simulate land use changes. The kappa indexes investigation of different simulation results depicted that the model precision was sensitive to the neighborhood distance and different land types had different sensitivities. The increase of distance in circle neighborhood will decrease the model precision, in the mean while the increase of distance in rectangle neighborhood will increase the precision first then decrease. In future researches of CA model applications a prior validation of neighborhood distances is recommended according to the above conclusion.