Performance of different reanalysis rainfall datasets and hydrological models in gap-filling for Mpologoma catchment in Eastern Uganda

ABSTRACT Study region Mpologoma Catchment, Lake Kyoga Basin, Eastern Uganda. Study focus The rising frequency and magnitude of hydrological extremes has increased demand for their accurate prediction for infrastructure design and climate adaptation. However, historical datasets used for such predictions, are often limited by short records, gaps and sparse monitoring. In catchments with discontinuous river flow records, rainfall–runoff models can generate flows to fill gaps, but these rely on ground-based rainfall datasets, which are scarce in mountainous regions of developing countries. To address these limitations in Mpologoma catchment, reanalysis rainfall datasets, 5th generation of European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA5) and Princeton Global Forcing (PGF) were blended with ground-based rainfall data and used to drive hydrological models. The daily blended datasets were evaluated against daily ground-based observations using correlation (CORR), root mean square error (RMSE), Nash–Sutcliffe efficiency (NSE) and percentage bias (PBIAS). Hydrological performance was assessed with Hydrologiska Byråns Vattenavdelning (HBV) model using Kling–Gupta efficiency (KGE), coefficient of determination (R²) and visual comparisons. The best-performing dataset was then used to evaluate Nedbør-Afstrømnings-Model (NAM), Australian Water Balance Model (AWBM) and HBV models. New hydrologic insights Results show that bias-corrected ERA5 merged with ground observations performed well statistically (CORR = 0.766, RMSE = 5.87 mm, NSE = 0.583, PBIAS = 0.416). Both ERA5 and PGF yielded similar HBV model performance (KGE = 0.35–0.78; R² = 0.4–0.64), though PGF simulated river flows better, underscoring its suitability for gap-filling ground-based rainfall in the study area. Among the models, AWBM performed best (KGE = 0.5–0.83; R² = 0.5–0.79), especially for peak flows, underscoring its suitability for gap-filling river flows in the study area. This study highlights the potential of reanalysis datasets for hydrological modelling in data-scarce catchments of the Lake Kyoga basin.