On the basis of the monthly mean temperature data of 19 meteorological stations from 1961 to 2008 in the Yili River Valley, the correlation between mean annual temperature and elevation was analyzed to get access to the temperature lapse rate of 0.564 ℃/100 m. The main purpose of this study was to apply a spatial interpolation method based on DEM and compare with the traditional ones of inverse distance weight (IDW) and Kriging methods. According to the cross-validation results, it was concluded that the accuracy of interpolation method based on DEM was superior to the other two methods. Combined with DEM data, the mean annual temperature distribution map with 90 m resolution was established and it expressed the main trend of temperature changing with terrain and reflected the spatial variability of temperature.