Variable Source Area (VSA) modeling for water management: streamflow analysis in Upper Gilgel Abay watershed, Ethiopia

Abstract

Reliable streamflow simulation is fundamental for effective water resources management in data-scarce watersheds, particularly where runoff is dominated by saturation-excess processes that are inadequately represented by many conventional hydrological models. This study evaluates the applicability of the variable source area (VSA) hydrological model for simulating streamflow dynamics in the Upper Gilgel Abay watershed, a humid tropical watershed in the Upper Blue Nile Basin of Ethiopia, an ecologically and socio-economically vital region. To improve parameter estimation, a two-stage calibration strategy was implemented, combining a Monte Carlo based Latin hypercube sampling (LHS)) for efficient exploration of the feasible parameter space and identification of behavioural parameter sets and Differential Evolution optimization (DEoptim) for refinement of the optimal parameter values. Model performance was evaluated using the Nash-Sutcliffe Efficiency (NSE), Kling-Gupta Efficiency (KGE), and Percent Bias (PBIAS). The calibrated VSA model achieved very good performance during calibration (NSE = 0.81, KGE = 0.86, PBIAS = − 4.87%) and good performance during validation (NSE = 0.69, KGE = 0.75, PBIAS = 5.13%). The model successfully captured the seasonal variability of streamflow, though peak flows were generally overestimated and low flows were underestimated during the validation period. Parameter analysis revealed substantial differences in parameter influence and uncertainty, with the groundwater recession coefficient (rec_coef) identified as the dominant parameter, while available water capacity (AWCper) and Depth showed secondary influence. Storage-related parameters, particularly Depth and SATper, remained comparatively uncertain, indicating limitations in constraining internal watershed storage from streamflow observations alone. Overall, the results demonstrate that the VSA framework, combined with LHS-DEoptim calibration and complementary sensitivity and uncertainty analyses, provides a physically meaningful and computationally efficient approach for representing saturation-excess runoff in data-limited Ethiopian highland watersheds, supporting hydrological evaluation and water management.