Abstract: Based on the CMIP5 historical simulations and RCP4.5 datasets, the ability of the CMIP5 models in simulating the spatial pattern and interannual variability of precipitation over Hunan Province are evaluated using the statistical downscaling method of transform cumulative distribution function (CDF-t). The future precipitation change is then projected. The results show that due to the low resolution of GCM models, characteristics of precipitation related to the terrain and atmospheric circulation over Hunan Province were not exactly reproduced. The downscaled precipitation agrees well with observations, and the capacity of CMIP5 models for simulating the spatial pattern and interannual variability of precipitation can be effectively improved by the downscaling approach. The multi-model ensemble projection of future changes in precipitation shows that in the 21 century, precipitation over Hunan Province will slightly increase by a rate of 0.95% perdecade. Precipitation was projected to increase by 4.6%, 5% and 5.2% with respect to that during 1986-2005 for Hunan Province in 2016-2035, 2046-2065 and 2081-2100 respectively. The spatial pattern of the future change in the three periods displays a good consistency, with the most obvious change occurring in the northeast, northwest and southeast of Hunan Province. It should be noted that there exist differences among models. With increases in radiative forcing, the differences among various model situations tend to become larger.