利用1979—2017年青藏高原131个气象台站实测降水资料,分别从年、季尺度对CRU、ERA5和CMFD 3种再分析降水资料在青藏高原的适用性进行了评估。结果表明：（1） CRU、ERA5、CMFD 3种数据对青藏高原年降水的模拟能力都很强,与观测值的相关系数均超过了0.9,较观测降水量值均偏大;CRU和CMFD的春季降水较为接近观测值,CMFD夏、秋季降水与观测值最接近,3种数据对冬季降水模拟能力均较弱。（2） 从降水量分布来看,CMFD数据对青藏高原年、春季、夏季、冬季降水的分布模拟能力最好,3种数据对青藏高原秋季降水的模拟能力具有区域差异性,CRU、CMFD在青藏高原西部的降水与观测值较为接近。（3） 从变化趋势来看,青藏高原年、春季、夏季、秋季降水均呈增加趋势,其中夏季增加幅度较大,冬季降水整体呈减少趋势。（4） CRU数据对青藏高原年、春季、夏季、秋季降水的变化趋势与观测值较为一致,其次为ERA5,ERA5冬季降水与观测值较为一致。（5） 从偏差分析来看,CMFD数据与观测值的年、季降水偏差最小,最接近观测值。（6） 3种数据与站点平均的年、季降水的时间变化序列表明,CMFD的年、春季、夏季、秋季降水与观测值的变化最为接近,其次为CRU;CMFD冬季降水较观测值最为接近,但相关系数未通过95%的显著性检验。

In this paper, CRU, ERA5, and CMFD precipitation data for the Qinghai-Xizang Plateau were evaluated on annual and seasonal scales and compared with observed precipitation data from 131 weather stations. (1) The three data types all showed a strong ability to describe total annual precipitation on the plateau, and their correlation coefficients with the observed values were all >0.9. Results from all three models were generally higher than observed precipitation. CRU and CMFD results were closest to observed values in spring, while CMFD results were closest to the observed values in summer and fall; all three models showed the weakest correlations with observed values for winter precipitation. (2) In terms of precipitation distribution, CMFD data had the best ability to simulate plateau precipitation annually, and in spring, summer, and winter. The models showed regional differences in describing fall precipitation. CRU and CMFD results were closest to observed values on the western plateau. (3) Plateau precipitation increased during spring, summer, and fall, especially in the summer, and decreased in the winter. (4) The annual and seasonal variation (spring, summer, and fall) in the CRU data was consistent with that of the observed data, and the winter precipitation results from ERA5 were consistent with observed values. (5) The deviation between CMFD results and observed annual and seasonal precipitation was the smallest. (6) Time series analysis showed that CMFD annual, spring, summer, and fall precipitation results were closest to the observed values, followed by CRU. CMFD winter precipitation results were closest to the observed values, but the correlation coefficient failed the significance test of 95%.