利用站点实测资料、GCMs 月数据对 GCMs 进行秩评分评估排序, 从 21 种 GCMs 模式优选出的 6 种 GCM模式的日数据、6 种 GCM 集成的气候模式、站点实测资料和 NCEP 再分析资料构建统计降尺度模型 SDSM, 预估泾河上游流域的未来降水变化。结果表明: 构建的降尺度模型对降水模拟较为可靠, 率定期各模式决定系数 R2 为 0.228~ 0.324, 标准误差为 0.354~ 0.450, 率定期和验证期模拟月均降水与实测值年内分布相近。在降尺度性能评价中集成模式表现最好。在 RCP 4.5 情景下, 泾河上游流域未来降水大多数模式和集成模式呈增加趋势, 到 2030 年泾河上游流域降水量将增加 4.8% , 且当地的春季雨量会增加, 夏季雨量会减少。

The prediction of precipitation changes in the future can provide a basis for research on water resource changes in the upper Jinghe Riv er basin. The GCMs are ranked according to the site measured data and monthly GCMs data. The statistical downscaling model SDSM is constr ucted based on daily data of 6 GCM models selected from 21 GCMs, the climate models integrated by 6 GCMs, in situstations data, and NCEP reanalysis data, to predict the future precipitation change in the upper reaches of Jinghe Riv2 er. The results show that the SDSM is reliable for precipitation simulation. The R2 of each model is between 0.228 and 0.324 , the standard error is between 0.354 and 0.450, respectively. The simulated monthly average precipitation in the periodic and verification periods is similar to the measured v alue and the distribution is similar within a year. The integr ated model performs best in the downscaling performance evaluation perio d. Under the RCP 4.5 scenario, most future precipitation models and integrated models in the upper Jing he River show an increasing tr end. By the 2030s, precipitation in the upper Jinghe River may increase by 4.8% , and local rainfall in spring also exhibits an incr easing trend, and summer rainfall may decrease.

国家自然科学基金( 51679061) ; 宁夏重点研发计划( 2018BEG02010)