Based on daily mean temperature records from 1961 to 2007 at 20 meteorological sites in Southwest Yunnan, and the surface temperature simulated by IPCC AR4 Climate Models, a quantitative examination was undertaken into the characteristics of multi—timescale temperature (AMT, DMT and WMT) variation in Southwest Yunnan. The simulation abilities of the models were also evaluated with the normalized root mean square error (NRMSE) and Mann—Kendal test statistic methods. Temperatures show remarkable increasing trend from 1961 to 2007, with the Mann—Kendall test statistic passing 95% confidence verification. The result of the NRMSE analysis shows that the simulated temperature anomaly variations are more similar to observed ones especially for AMT and DMT, and the projected result (anomalies) of IPCC AR4 climate models can be used for predicting the trends in multi—timescale temperature variation in Southwest Yunnan in the next 40 years under the three emission scenarios, which has better simulating effect on AMT and DMT than WMT. Over the next 40 years the temperature will continue to rise, with annual mean temperature showing a more remarkable rising trend than that of the dry and wet seasons. Temperature anomalies exhibit different increasing rates under different emission scenarios: During the 2020s the rising rates of multi—timescale temperature anomalies in a high greenhouse gases emissions scenario (SRESA2) are smaller than those under a low emission scenario (SRESB1). Except that, the rate of increase in temperature anomalies are the highest in the intermediate emissions scenario (SRESA1B), followed by those in SRESA2, and those in low emissions scenario (SRESB1) are the lowest. The reason of different simulating effects on WMT from AMT and DMT was also discussed.