Abstract: Using the long term NDVI time series derived from SPOT VGT, the spatio-temporal feature of land use/land cover (LUCC) dynamic changes from 1999 to 2009 was investigated. First, the BISE model was improved to become a new pre-processing method for VGT time series processing. Second, a new classification model which can be named as SAM-MDM was reconstructed with SAM and MDM based on the phenological characteristics of annual NDVI time series; and then LUCC maps were retrieved from annual NDVI time series, and the post-classification recoding was performed on them. The classification accuracy was improved obviously and met the requirement of the trend analysis of land use/land cover changes (LUCC). Third, applying the model of LUCC dynamic change rate, the analysis was performed on the spatio-temporal feature of LUCC. Finally, applying the principles of Markov process, a model of LUCC class transition was constructed, and it was used to perform prediction for dynamic changes of LUCC in the next 20 years. These results indicate some conclusions: 1) the cropland, water body and grassland reduced continuously in the past decade; 2) the built-up area, bare land and woodland increased continuously; 3) there are different patterns of the spatio-temporal feature of LUCC for different land covers in different areas in china, including the south east, the north east, the north west and the south west; 4) the transition probability of LUCC which was discovered in the past decade should be continued in the next 20 years; 5) these results can provide information for regional socio-economic development decisions.