Based on the monthly mean temperature from 105 meteorological observation stations in Xinjiang from 2007 to 2016, five geographical factors  including longitude, latitude, slope, aspect and elevation were extracted from the digital elevation model (DEM),  and the correlation between the mean temperature  and the geographical factors was analyzed. Ordinary kriging, ordinary co-kriging and multiple linear regression spatial interpolation methods were selected to make air temperature interpolation in the north and the south region of the Tianshan moutains. The precisions of interpolated temperature were compared in the north and south regions of the Tianshan Moutains by using error analysis and correlation analysis. The results show that the ordinary kriging method was the best in summer and the ordinary co-kriging method was the best in winter, while the interpolation accuracy of the multiple linear regression method was the lowest in the north and south regions of the Tianshan Mountains. Because of climate difference between the north and south regions of the Tianshan Mountains, the interpolation accuracy of  ordinary kriging and ordinary co-kriging methods was better in the north and south regions of the Tianshan Mountains. Because of the influence of seasonal characteristics, the best method was the ordinary kriging from April to September, and the ordinary co-kriging method was the best from October to March in the north and south regions of the Tianshan Mountains. But in the south region of the Tianshan Mountains, ordinary co-kriging method was better than the others from March to September, and ordinary co-kriging method was the best from October to February.