The SDSM was applied to develop the quantitative statistical relationships between the predictands and the predictors based on daily maximum and minimum temperature data observed and the NCEP re - analysis data in Tianshan Mountains during the period of 1961 - 2000. The daily temperature for future periods (2020s,2050s and 2080s) was estimated by using the validated transfer function from output of the HadCM3 SERSA2 and B2 at 27 meteorological stations. Results show that SDSM is efficient for reproducing observed daily temperature at each station. There is an obvious increasing trend for daily temperature in the future. Both of them show a greater increasing range in A2 scenario that in B2 scenario,while daily maximum temperature shows a greater increasing range than and minimum temperature in both scenarios. The broadest range was found in spring while the lowest was in winter. The spatial distributions of daily temperature are similar in the future periods. The daily increments of temperature in A2 and B2 scenarios become smaller from the north to the south of Tianshan Mountains.