Satellite-based rainfall estimates evaluation using a parsimonious hydrological model in the complex climate and topography of the Nile River Catchments

Satellite-based rainfall products with high spatial and temporal resolutions can provide an excellent data source for regions where rain-gauge networks are missing or unevenly and sparsely distributed. The objective of this study was to examine the suitability of six high-resolution satellite-based rainfall products, Climate Hazards Group InfraRed Precipitation with Stations Version 8 (CHIRPSv8), African Rainfall Estimation Algorithm Version 2 (RFEv2), Tropical Rainfall Measuring Mission Multi-Satellite Precipitation Analysis version 7 (TMPA3B42v7), Integrated Multi-SatelliteE Retrievals for Global Precipitation Measurement version 6 (IMERGv6), Multi-Source Weighted Ensemble Precipitation version 2.8 (MSWEPv2.8) and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN), for the complex topography and climate of the Nile headwater catchments. The reliability of these products was assessed using rainfall estimates as an input to force a parsimonious and high-resolution hydrological model. A spatially distributed and parsimonious hydrological model was developed in Wflow-PCRaster/Python modelling framework to simulate streamflow using rain-gauge and satellite-based rainfall inputs. The model was calibrated and validated at three locations for each rainfall product and ground rainfall measurement. The accuracy of satellite-based rainfall products as input to the hydrological model for daily streamflow forecasting was dependent upon the product type. CHIRPSv8, IMERGv6 and RFEv2 demonstrated a reliable capability to simulate the streamflow at different spatial scales. In contrast, the TMPA3B42v7 and PERSIANN products failed to capture the ground streamflow measurements. All products performed inconsistently when simulating large floods and low flows. While TMPA3B42v7 and MSWEPv2.8 underestimated the streamflow with a large magnitude of variation in peak flow, the remaining products underestimated it throughout the study period. CHIRPSv8 and IMERGv6 product outperformed all other products consistently across all periods and locations followed by RFEv2. PERSIANN failed to capture the observed streamflow with lower performance indices. A better agreement of simulated streamflow with ground streamflow measurements was attained from CHIRPSv8 and IMERGv6 compared to the gauge-driven rainfall inputs. Performance of streamflow simulations improved when catchment area was increased from 481 to 46,000 km2. Overall, this stud