Evaluation of GPM IMERG products and ERA5 reanalysis for flood modeling in a semi-arid watershed

Satellite-based and Reanalysis rainfall products could be a valuable source of data for precipitation for hydrological modeling over data-scarce regions. The objective of this study is to assess the suitability of ERA5 Reanalysis and GPM IMERG V06 data (GPM-Early, GPM-Late and GPM-Final) for flood modeling over a Moroccan semi-arid watershed (Rheraya) during 2013–2018.  Both statistical scores and an event-based model were used to evaluate the performance of these products to estimate precipitation and simulate flood events. The results showed that the four products often overestimate the observed precipitation. The highest bias (124% and 145%) was obtained with GPM-E and GPM-L while the bias for GPM-F and ERA5 was much lower (31% and 42%). However, the four products showed acceptable correlations with observed data. In terms of precipitation detection capability on the hourly time scale, the GPM-E and GPM-L products presented satisfactory performance. They were the most efficient for different rainfall thresholds. In addition, by comparison with observed rainfall, flood modeling results showed that the GPM-E and GPM-L were the most efficient for flood event simulation (Nash greater than 0.4 for the majority of events and correlation coefficients greater than 0.7). This study showed that the different precipitation products tested herein have satisfactory performance for the hydrological modeling of floods. These sources of precipitation could be an alternative in ungauged or poorly gauged basins for flood simulation.