In the northern slope of Tianshan Mountain

Xinjiang

with Landsat MSS data

the grassland classification and mapping were studied. Based on the analysis of some spectral properties of grasslands and the separability between the ground classes

with the topographic types

the distinct 42 classes （27 grassland classes） were classified by a maximum-likelihood classifier. The classification accuracy of 86 percent ground classes is above 80 percent. Computer plotting that draws the line map of grassland types （scale 1:100

000） from the classification result was proposed. It speeds up 200--300 times than manual mapping. The methods to obtain the Digital Terrain Model （DTM） from contour map by scan digitizing were studied. With the DTM the grassland classification accuracy in the mountain areas was improved efficiently.