Abstract: To solve the problem that there exists large bias between simulated precipitation from global climate models (GCMs) and measured precipitation, a bias correction method based on separated three gamma distributions (STG) was proposed. This method divides the precipitation sequence into three parts, i.e., the minimum values, the normal values and the maximum values, and then the cumulative distribution functions (CDFs) of these three parts are respectively bias corrected. In addition, based on the thought of quantile deviation accumulation, a comprehensive evaluation index C is also put forward to evaluate the bias correction effects of STG method. The STG method is applied to correct the bias between the simulated precipitation from 18 GCMs and measured precipitation for 17 meteorological stations in Yarlung Tsangpo River basin, and is further compared with the existing two gamma-related bias correction methods, i.e., Single Gamma (SG) and Double Gamma (DG). The results show that the STG method can effectively remove the bias between GCMs simulated precipitation and observed precipitation, and the evaluation indexes C are mostly above 0.85. Moreover, STG performes better than SG and DG in the validation period which indicates STG is more robust for bias correcting GCMs simulated precipitation.