Abstract The test-area of 256 × 256 pixels sub-image within Huangyan county, Zhejiang province, was selected for application of landsat TM data to landuse classifcation and thematic information extraction studies. A method of integrated supervised classifier(dynamical clustering) with unsupervised classifier (minimal distanse) is used, the results are very satisfactory. The information of citrus trees distribution is also extracted successfully, the accuracy is 95. 3%. The research provided a efficent approch to invistigate the economic forest resource and detect its change rapidly in subtropical area.