Performance evaluation of satellite-based rainfall products on hydrological modeling for a transboundary catchment in northwest Africa

The scarcity of rainfall data is one of the main problems affecting the use of hydrological models. Several model satellite-based rainfall estimates (SREs) have been developed to provide an alternative to poorly or ungauged basins. The aim of this work was to evaluate the suitability of SREs for hydrological modeling using a semi-distributed model in the transboundary basin of Medjerda, shared by Tunisia and Algeria. Two satellite-based rainfall products (PERSIANN-CDR and CHIRPSv2) were first compared to rain gauge observations based on sub-basin and point-to-pixel analysis. The selected SREs products were then used as inputs to simulate discharge at a daily time-step over the 1996–2016 period. The simulated streamflows were compared to data measured at four runoff gauging stations and at the outlet of two dams. It was first shown that both SRE products perform weakly at daily scale but that the CHIRPSv2 product performs better at monthly scale. Second, comparison at sub-basin scale led to a better correlation with rain gauge observations than point-to-pixel analysis. Third, direct sampling can be reliably used to fill gaps in discharge time series by using auxiliary stations highly correlated with the target station. Finally, the CHIPRSv2 daily satellite rainfall product is more efficient and more suitable than the PERSIANN-CDR product for hydrological modeling. Thus, CHIRPSv2 can be used as an alternative or as a complementary source of information to simulate hydrological models in arid and semi-arid regions and can successfully solve the issue of missing rainfall data in transboundary catchments.