Evaluation of satellite rainfall products through hydrologic simulation in a fully distributed hydrologic model

Abstract

The goal of this study is to evaluate the accuracy of four global high‐resolution satellite rainfall products (CMORPH, TMPA 3B42RT, TMPA 3B42, and PERSIANN) through the hydrologic simulation of a 1656 km 2 mountainous watershed in the fully distributed MIKE SHE hydrologic model. This study shows that there are significant biases in the satellite rainfall estimates and large variations in rainfall amounts, leading to large variations in hydrologic simulations. The rainfall algorithms that use primarily microwave data (CMORPH and TMPA 3B42RT) show consistent and better performance in streamflow simulation (bias in the order of −53% to −3%, Nash‐Sutcliffe efficiency (NSE) from 0.34 to 0.65); the rainfall algorithm that uses primarily infrared data (PERSIANN) shows lower performance (bias from −82% to −3%, Nash‐Sutcliffe efficiency from −0.39 to 0.43); and the rainfall algorithm that merges the satellite data with rain gage data (TMPA 3B42) shows inconsistencies and the lowest performance (bias from −86% to 0.43%, Nash‐Sutcliffe efficiency from −0.50 to 0.27). A dilemma between calibrating the hydrologic model with rain gage data and calibrating it with the corresponding satellite rainfall data is presented. Calibrating the model with corresponding satellite rainfall data increases the performance of satellite streamflow simulation compared to the model calibrated with rain gage data, but decreases the performance of satellite evapotranspiration simulation.