Abstract: The land use maps in 2020 and 2030 were simulated by using CA-Markov model(cellular automaton-markov)based on the land use data of Guizhou Province in 1990, 2000 and 2010, and the spatio-temporal patterns of the land use changes according to landform characteristics were characterized by using these land use data together, The result indicates that the characteristics of land use change in study area was an increase in forest lands, grasslands, water area and construction lands accompanying a decrease of farmland and unused land. Land use change of each landform in study area had the similar features. Land use degree composite index of basin firstly increased and then decreased. Land use degree composite index of other landforms decreased. There were significant differences in variation and dynamic degree of land use of different landform types. The area and ratio of the stable model in each landform were the biggest. The stable model and post-change model of basin were greater than other landforms. However, the pre-change model of hills, low mountains, middle mountains and plateau were higher than basin. The ratio of continuous change model and intermediate change model were small. There was distribution-advantage of post-change model in basin and hills. The pre-change model, intermediate change model and continuous change model were the distribution advantage of TUPU in low mountains, middle mountains and plateau, respectively. ③ The merits of natural conditions, the different pace of economic development and the policy implementation of grain for green project were the main factors affecting the difference of land use change in each landform.