How well do gridded precipitation and actual evapotranspiration products represent the key water balance components in the Nile Basin?

Abstract

Study region: Nile Basin, Africa. Study focus: The accurate representation of precipitation (P) and actual evapotranspiration (ETa) patterns is crucial for water resources management, yet there remains a high spatial and temporal variability among gridded products, particularly over data-scarce regions. We evaluated the performance of eleven state-of-the-art P products and seven ETa products over the Nile Basin using a four-step procedure: (i) P products were evaluated at the monthly scale through a point-to-pixel approach; (ii) streamflow was modelled using the Random Forest machine learning technique, and simulated for well-performing catchments for 2009–2018 (to correspond with ETa product availability); (iii) ETa products were evaluated at the multiannual scale using the water balance method; and (iv) the ability of the best-performing P and ETa products to represent monthly variations in terrestrial water storage (ΔTWS) was assessed through a comparison with GRACE Level-3 data. New hydrological insights for the region: CHIRPSv2 was the best-performing P product (median monthly KGE’ of 0.80) and PMLv2 and WaPORv2.1 the best-performing ETa products over the majority of the evaluated catchments. The application of the water balance using these best-performing products captures the seasonality of ΔTWS well over the White Nile Basin, but overestimates seasonality over the Blue Nile Basin. Our study demonstrates how gridded P and ETa products can be evaluated over extremely data-scarce conditions using an easily transferable methodology.

Keywords

• Precipitation
• Evapotranspiration
• Water balance
• Remote sensing
• Random Forest
• GRACE