Abstract: A decision tree classifier was used for desertified land classification using Landsat ETM+ data for a Cele Oasis, a typical ecotone in the south of Taklamakan. The decision tree classification algorithm was developed with the analysis of the spectrum reflecting the characteristic of various ranges of desertified lands and the integrated features of TM image data. Each node of the tree isassociated with MSAVI, NDWI and the Brightness, Greenness, Wetness of the soil.The results suggest that the decision tree classifier performs well, and could decrease and avoid some of the interference while extracting the information on desertified lands. The decision tree classifier method could serve as an effective measure for automatically extracting information on desertified lands.