Enhanced hydrological modeling with the WRF-Hydro lake–reservoir module at a convection-permitting scale: a case study of the Tana River basin in East Africa

East Africa frequently experiences extreme hydrological events, such as droughts and floods, underscoring the urgent need for improved hydrological simulations to enhance prediction accuracy and mitigate losses. A major challenge lies in the limited quality of precipitation data and constraints on model capabilities. To address these challenges, the upper and middle Tana River basin, characterized by its sensitivity to drought, vulnerability to flooding, and data availability, was selected as a case study. We performed convection-permitting (CP) regional climate simulations using the Weather Research and Forecasting (WRF) model and conducted hydrological simulations with a lake–reservoir-integrated WRF Hydrological modeling system (WRF-Hydro) driven by the CPWRF outputs. Our results show that the CPWRF-simulated precipitation outperforms ERA5 when benchmarked against Integrated Multi-satellite Retrievals for GPM (Global Precipitation Measurement) (IMERG), with evident bias reduction in seasonal precipitation mainly over the Mount Kenya region and with a probability of light rainfall (1–15 mm d$^{−1}$) during the dry season. Improved precipitation enhances the hydrological simulation, significantly reducing false peak occurrences and increasing the Nash–Sutcliffe efficiency (NSE) by 0.53 in the calibrated lake-integrated WRF-Hydro model (LakeCal) driven by CPWRF output compared to ERA5-driven simulations. ... mehrAdditionally, the lake–reservoir module increases the sensitivity of river discharge to spin-up time and affects discharge through lake–reservoir-related parameters, although adjustments to the parameters (i.e., the runoff infiltration rate, Manning's roughness coefficient, and the groundwater component) have minimal effects on discharge, particularly during the dry season. The inclusion of the lake–reservoir module effectively reduces the model-data bias in WRF-Hydro simulations, particularly for the dry-season flow and peak flow, resulting in an NSE increase of 1.67 between LakeCal and LakeNan (model without the lake–reservoir module). Notably, 24 % of the NSE improvement is attributed to CPWRF and 76 % is attributed to the lake–reservoir module. These findings highlight the enhanced capability of hydrological modeling when combining CPWRF simulations with the lake–reservoir module, providing a valuable tool for improving flood and drought predictability in data-scarce regions like East Africa.