Abstract: Cellular Automata (CA) model is an efficient model to simulate land use/cover change (LUCC) process. However, the majority of geographic cellular automata researches now focus more on transition rules than on spatial scale selection. As a result, the spatial scale decisions of geographic cellular automata are often made arbitrarily. When using CA model in geographic context, nontrivial questions have to be answered about the choice of spatial scale and neighborhood configuration. This article takes Huadu district,Guangzhou City as the study area, to investigate the effect of changing spatial scale and neighborhood configuration on the result prediction accuracy of the CA-Markov model and the morphology of land use change simulation result. The result shows that (1)the spatial scale has great impact on the simulation results of CA-Markov model. Coarse cell size will reduce the result prediction accuracy of the model. There is a threshold of cell size in the model. When the cell size exceeds this threshold, the result prediction accuracy dropped rapidly. (2) Neighborhood configuration also has impact on the simulation results of CA-Markov model. The patch number and patch density of simulation result using 3?3 Von Neumann neighborhood are higher than simulation result using 3?3 and 5?5 Moore neighborhood, but there is little difference of Kappa coefficients of the simulation results.