CFSR降水数据在干旱山区水文模型中的应用——以新疆开都河流域为例

西北干旱山区雨量站稀少且分布不均,降水资料难以满足水文模拟的需求。CFSR(climate forecast system reanalysis)再分析数据覆盖范围广、时空分辨率高,是很有潜力的降水数据替代品,但其在干旱山区水文模型中的适用性还有待深入探究。本文以开都河流域为研究区,采用LOCI法、Gamma分布映射法以及以上两种方法相耦合(gam LOCI法)对CFSR降水数据进行偏差校正,分别用校正前和校正后的降水数据驱动已根据观测数据率定好的水文模型,比较流域日径流模拟结果差异。结果表明:由于CFSR原始数据与观测数据偏差较大,导致模拟的径流量与实测径流相对偏差大于40%,并且春季洪峰提前、洪峰峰值提高;与雨量站实测数据相比,gam LOCI法兼具了LOCI法和Gamma分布映射法的优点,既能校正湿日频率和湿日降水强度,又能校正反映离散程度的标准差和分位数,在高海拔山区校正效果更优,但在平原区,3种校正方法无显著差别;与水文站实测数据相比,LOCI、Gam和gam LOCI法校正后的降水数据驱动水文模型得到的径流模拟结果有很大提升,3种校正方法虽各有特点,但径流模拟结果差别不大,都能获得较为满意的校正效果。

Application of CFSR Precipitation Dataset in Hydrological Model for Arid Mountainous Area:A Case Study in the Kaidu River Basin

Rain gauges are very scarce and unevenly distributed in the arid regions of northwest China,especially in mountainous regions,and such situation cannot meet the need of hydrological models.CFSR reanalysis precipitation dataset is a potential substitution of ground observations because of its wide coverage and high spatiotemporal resolutions,but its applicability in hydrological models for arid mountains needs to be further explored.This paper focused on the Kaidu-Kongque River Basin,one of the headwaters of the Tarim River,and compared the stationary observations with CFSR data,after applying three precipitation correction methods namely the local intensity scaling (LOCI),distribution mapping with gamma distribution (Gam) and the method coupling with the aforementioned two (gamLOCI).The raw CFSR data and the bias corrected data were used as the meteorological inputs of a distributed hydrologic model to ascertain the difference of streamflow.The results are as follows:① The relative bias between the simulated runoff volume data with raw CFSR data and the measured runoff volume data was high than 40％,and the peak of spring flood was earlier and increased because of the large deviation between the original CFSR data and the observed data;② Compared to the observed precipitation data,the advantages of both LOCI and Gam were taken in the gamLOCI method,which could be used to correct the wet-day frequency and intensity of precipitation,standard deviation and fractile,and performed well in mountainous regions;③ Compared to the observed streamflow data,all the correction methods could be used to effectively improve the daily streamflow simulations and equally performed well in SWAT model.