Evaluating the Performance of Merged Multi-Satellite Precipitation Products Over a Complex Terrain

The aim of this paper is to investigate the effectiveness of merging satellite precipitation products on rainfall estimates over different terrains in Iran. Four widely used satellite rainfall estimates (SREs) namely PERSIANN, TMPA-3B42, TMPA-3B42RT, and CMORPH were evaluated and blended in order to generate new combinations in which we expect to estimate rainfall more accurately than with using each individual product. For this purpose, first, the daily bias of each satellite algorithm was calculated at each pixel by comparing with gauge observed data for the time period of 2003–2008. Then, for each pixel, a linear combination of SREs was selected which resulted in the least bias for that pixel. This combination named as merged satellite rainfall estimate (MSRE). In order to test the proposed method, several statistical indices including bias, root mean square error (RMSE), and correlation coefficient (CC) were employed. The results indicate that the merged product used in this study can improve the accuracy of rainfall estimates compared with each individual satellite estimate over all the study regions for different seasons. However, the efficiency of improvement depends highly on the number of rain gauges distributed over the study area.