卫星与地面观测融合降雨产品精度与径流模拟评估

Accuracy and runoff simulation appraisal of merged satellite–station rainfall product

虽然近年来卫星反演降雨产品逐渐成熟,但精度仍较差,不能满足现代水文气象业务和科研的需求,因此需要地面观测资料对其进行订正。文中采用的卫星与地面站融合数据是基于中国3万多个自动站降雨观测数据和CMORPH卫星反演降雨资料,采用PDF(probability density function)和OI(optimal interpolation)两步融合方法生成的中国区域高时空分辨率(0.1°/1h)的降雨量融合产品。通过在汉江丹江口水库以上流域建立分布式水文模型,对2008~2012年的融合降雨数据进行水文模拟适用性分析。结果表明:在日尺度上,CMORPH卫星降雨与地面站点融合数据具有较好的相关性,但存在一定的系统偏差;该降雨产品能较好地捕捉到强度小于25 mm的中小降雨,其模拟的径流与流域下部高峰流量的观测结果有较好的一致性,但总水量比实测值低15.7%,NSE达0.723。研究成果为基于高时空分辨率的卫星雨量计融合降雨产品在分布式水文模型的水文适用性提供了新的见解和经验,并为该数据在流域洪水模拟的水文预报应用提供参考与借鉴。

Although satellite precipitation products have gradually matured in recent years,the accuracy is still poor and can not meet the needs of modern hydrometeorological operations and scientific research,so it is necessary to revise them with the ground observation data.The merged satellite–station rainfall data used in this paper are high temporal-spatial resolution(0.1°/1h)products by two-step merging method of PDF(probability density function)and OI(optimal interpolation)based on precipitation observation data from more than 30 000 automatic stations in China and precipitation inversion data by CMORPH satellite.By establishing a distributed hydrological model in the upstream basin of Danjiangkou Reservoir,Hanjiang River,the merged precipitation data from 2008 to 2012 were analyzed.The results show that:on the daily scale,there is a good correlation between CMORPH satellite precipitation and ground station data,but there is still some system deviation.The precipitation product can capture the small and medium precipitation with intensity less than 25 mm.The simulated runoff is in good agreement with the observed results of the flood peak flow in the lower reaches of the basin,but the total water quantity is 15.7%lower than the measured value,and the NSE is 0.723.This study provides a new opinion and experience for the hydrological applicability of high temporal-spatial resolution merged satellite–station rainfall and a reference for the application of this data in hydrological forecasting of watershed flood simulation.