降水作为水文模型的重要气象输入，其精度直接影响流域水文模拟精度。气象站点数据精度较高，但由于空间分布不均，局部地形影响较大，在从点到面的过程中往往存在误差；卫星遥感降水数据能够在连续的时空上反映降水的特征，但由于间接观测和反演算法等，也存在一定的误差。基于此，论文以雅砻江流域为研究区，对卫星降水数据和插值结果数据的精度及特点进行对比分析，开展了卫星遥感降水与地面站点观测降水数据融合校正研究，并运用VIC水文模型，以卫星降水数据、插值数据和校正结果数据作为气象输入进行水文径流模拟，对多源降水数据在雅砻江流域水文模拟过程中的适用性进行了全面分析，为提高降水数据精度与水文过程模拟适用性提供了有效参考。论文的研究内容包括以下三个方面：

（1）对雅砻江流域四种卫星降水数据（TRMM 3B42V7、CMORPH、PERSIANN-CDR、APHRODITE）及三种插值方法（IDW、Cokriging和ANUSPLIN）结果数据的精度进行了综合分析评价。结果表明TRMM 4B42V7数据和APHRODITE数据与地面站点日降水数据具有较强的相关性，CMORPH在降水量较多的情况下高估严重，PERSIANN-CDR存在明显大幅度高估，精度较差。各个子区域最优数据集各有特点，区域I为TRMM 3B42V7数据集，区域II为APHRODITE数据集，区域III和区域IV插值结果普遍好于卫星降水数据，Cokriging相对最优。

（2）对TRMM 3B42V7卫星降水与地面站点实测降水数据进行四种方法（因子校正、贝叶斯校正、双核平滑校正和分位数映射校正）的融合校正处理。整体来看，几种方法均可以在一定程度上提高卫星降水数据精度。分位数映射校正方法和双核平滑校正方法总体最优，在三个水文站点处Corr均大于0.86，相较原始数据均提高了0.03以上。

（3）采用不同降水数据驱动VIC模型进行雅砻江流域水文过程模拟，TRMM 3B42V7能够相对较好地捕捉夏季的洪峰径流，插值数据不能很好地模拟洪峰径流，误差在40%以上，且在甘孜站NSE均小于0.3。基于分位数映射校正和双核平滑校正的结果降水数据在径流模拟精度较高，在桐子林和雅江水文站断面处NSE均在0.83以上，且在甘孜站NSE提高了0.1以上，因子校正和贝叶斯校正结果与TRMM 3B42V7数据相差不大。

Precipitation, as an important meteorological input of hydrological model, has a direct impact on the accuracy of hydrological simulation. The accuracy of meteorological station data is high, but due to the uneven spatial distribution and the influence of local terrain, there are often obvious errors. Satellite remote sensing precipitation data can reflect the spatial and temporal distribution characteristics of precipitation in the basin, but there are also some errors because of the indirect observation and inversion algorithm. Based on this, this paper uses the Yalong River Basin as the research area. The accuracy and characteristics of satellite precipitation datasets and interpolation result datasets are evalueated. The paper also suties the fusion correction by selecting four different correction methods and further uses the satellite precipitation datasets interpolation datasets and the fusion precipitation datasets to drive VIC hydrological model to simulate hydrological runoff. The applicability of multi-source precipitation datasets in the hydrological simulation process of Yalong River Basin are comprehensively analyzed, which provides an effective reference for improving the accuracy of precipitation data and the applicability of hydrological process simulation. The main research contents include the following aspects:

(1) The accuracy and applicability of four satellite precipitation data (TRMM 3B42V7, CMORPH, PERSIANN-CDR, APHRODITE and three interpolation methods (IDW, Cokriging and ANUSPLIN) in Yalong River Basin are analyzed and evaluated. The results show that TRMM 3B42V7 data and APHRODITE data have a strong correlation with the daily precipitation data of the ground station, and CMORPH overestimates seriously when there is a lot of precipitation. PERSIANN-CDR has a significant overestimation and poor accuracy. The optimal datasets of each sub region has its own characteristics. Region I is TRMM 3B42V7 data set, region II is APHRODITE data set. The interpolation results of region III and Region IV are  

generally better than those of satellite precipitation data, and Cokriging is relatively optimal.

(2) Four methods (Factor correction, Bayes correction, Double-kernel smoothing correction and Quantile mapping correction) are used for the fusion correction of TRMM 3B42V7 satellite precipitation. Several methods can improve the accuracy of satellite precipitation data to a certain extent. Among them, Quantile mapping correction method and Double-kernel smoothing correction method are the best, and the Corrs are both greater than 0.86, which is more than 0.03 higher than the original data.

(3) Using different precipitation data to drive VIC Model to simulate the hydrological process of Yalong River Basin, TRMM 3B42V7 can capture the peak runoff in summer relatively well, and the interpolation data can not simulate it well with the error more than 40%. The results based on Quantile mapping correction and Double-kernel smoothing correction have higher accuracy in runoff simulation, The NSE of Tongzilin and Yajiang hydrological stations is above 0.83, and the NSE of Ganzi station is more than 0.1 higher than the original data. The results of factor correction and Bayesian correction are similar to those of TRMM 3B42V7.