黄河源区位于青藏高原的东北部， 该流域气象站点分布稀疏， 水文过程研究存在很大的局限性， 研究多种降水数据产品在流域的适用性对水文模型研究具有重要的推动作用。本文以黄河源区作为研究区域， 基于中国气象同化驱动数据集（China Meteorological Assimilation Datasets for SWAT model Version1.1， CMADS V1.1）、 热带降雨测量卫星（Tropical Rainfall Measurement Mission， TRMM）降水数据集（3B42 Version7）和气象站观测降水数据分别驱动土壤水文评估工具分布式水文模型（Soil and Water Assessment Tool， SWAT）， 同时利用SWAT-CUP（SWAT Calibration and Uncertainty Program）和SUFI-2（Sequential Uncertainty Fitting2）算法对27个敏感性参数进行率定， 模拟黄河源区多年月平均径流量的变化规律， 并将模拟结果与观测值进行对比分析， 评估了CMADS和TRMM 3B42降水数据产品在该流域的精度以及SWAT模型在黄河源区的适用性。结果表明： （1）由3种降水数据获得的降水空间分布为由西向东递增的趋势， TRMM 3B42与实测降水在年和月变化上一致性好于CMADS数据集。（2）参数敏感性分析结果表明： SCS（Soil Conservation Service）径流曲线数、 地下水滞后系数、 土壤蒸发补偿系数对径流模拟的敏感性程度较强。（3）利用CMADS和TRMM 3B42降水数据集模拟的径流结果均优于实测降水数据， 流域内3个水文站在率定期的相关系数R分别是0.93， 0.92和0.88； TRMM 3B42模拟结果次之， 率定期和验证期的相关系数R均在0.80以上， 纳什系数（Nash-Sutcliffe efficiency coefficient， NSE）在0.50以上。本研究证明了CMADS数据集和SWAT模型在地貌类型复杂、 气候变化敏感的高海拔地区的径流模拟方面具有较好的适用性， 为气象站点稀缺的地区提供了建立水文模型的替换方案。

The Source Area of the Yellow River is located in the northeastern part of the Qinghai-Xizang Plateau， and the meteorological stations are sparsely distributed in this basin， the study of the applicability of various precipitation data products has an important values in promoting the hydrological modeling in the basin.Based on the China Meteorological Assimilation Datasets for SWAT model Version1.1 （CMADS V1.1）， the Tropical Rainfall Measurement Mission （TRMM） precipitation datasets （3B42 Version7） and the Soil and Water Assessment Tool （SWAT） driven by these precipitation data， respectively， and the SWAT-CUP （SWAT Calibration and Uncertainty Program） and SUFI-2 （Sequential Uncertainty Fitting2） algorithm 27 sensitivity parameters were rate in simulating the variation of multi-year monthly average runoff， the simulated results were compared with the observations to evaluate the accuracy of CMADS and TRMM 3B42 precipitation data products and the applicability of SWAT model were evaluated in the Source Area of the Yellow River source area.The results show that： （1） The distribution of all three precipitation datasets showed an increasing trend from the west to the east， and TRMM 3B42 was in better agreement with the measured precipitation than CMADS data set in terms of annual and monthly variation.（2） The sensitivity analysis of the parameters showed that the sensitivity degree of SCS （Soil Conservation Service） runoff curve number， groundwater lagging coefficient， and soil evaporation compensation coefficient were stronger than that of the others.（3） The simulated runoff by using the CMADS and TRMM 3B42 precipitation datasets had better results than that by using the measured precipitation data， with the correlation coefficients R ² of 0.93， 0.92 and 0.88 for the rate period at the three hydrological stations， respectively， while the results of the TRMM 3B42 simulation were the next best， with the coefficients of correlation （R） of the rate-period and validation-period of above 0.80， and the Nash-Sutcliffe efficiency coefficient （NSE） of the simulations is above 0.50.This research demonstrates the applicability of CMADS datasets and SWAT model for runoff simulation in high-altitude areas with complex landscape types and sensitive to climate change， and provides a replacement solution for improving the hydrological models in areas where there are sparely meteorological stations.