Integrating machine learning and physical models for rainfall-runoff prediction in the Upper Baro Akobo River Basin, Ethiopia

This study aims to enhance runoff prediction accuracy in the data-scarce Upper Baro Akobo Basin, Ethiopia, by integrating the Hydrologic Engineering Center’s Hydrologic Modelling System (HEC-HMS) with support vector regression (SVR). Performance was evaluated using Nash-Sutcliffe efficiency (NSE), root mean square error (RMSE), and coefficient of determination (R²). The hybrid HEC-HMS-SVR model demonstrated superior performance with NSE of 0.989, RMSE of 29.78 m³/s, and R² of 0.99. In contrast, the HEC-HMS model showed an NSE of 0.85, RMSE of 113.4 m³/s, and R² of 0.89, while the SVR model had an NSE of 0.987, RMSE of 32.067 m³/s, and R² of 0.99. The hybrid SVR model enhances the understanding of key hydrological components, such as peak flow, baseflow, hydrograph dynamics, and total runoff volume. These components are vital for effective flood management and play a crucial role in the thoughtful design and implementation of engineering solutions to address hydrological challenges.