<b>ASSESSMENT OF GLOBAL HYDROLOGICAL MODEL PERFORMANCE AGAINST OBSERVED DISCHARGE DATA</b><b><i>Application to Africa River Basins</i></b>

Reliable river discharge information is essential for flood forecasting, drought monitoring, and climate adaptation planning, yet across much of Africa, sparse hydrometric observations severely constrain the ability to validate the global hydrological models that are increasingly used to fill these data gaps. This thesis presents a spatially matched evaluation of global discharge products across African river basins, with comparisons to better-monitored regions in Europe and the United States. Discharge data from Global Flood Awareness System (GloFAS) and Community Earth Systems Model version 2 (CESM2) are compared with Global Runoff Data Centre (GRDC) observations, alongside a secondary monthly analysis incorporating the Community Land Model (CLM). A key feature is the explicit handling of spatial mismatch between model grids and gauge locations through watershed-based evaluation and area-weighted discharge adjustment. Model performance is assessed using efficiency metrics, hydrograph characteristics, and flow distribution behavior, across climate zones and time periods to account for data limitations in Africa.

The results show that GloFAS outperforms CESM2 across all regions, with strongest performance in the United States (KGE of 0.12), moderate performance in Europe (KGE of -0.04), and reduced skill in Africa (KGE of -0.15). CESM2 exhibits systematic overestimation of discharge, particularly in larger basins. The CLM analysis demonstrates that improved precipitation forcing alone is insufficient for accurate discharge simulation, as errors in other model components can still drive flow overestimation. Across all models, performance is strongly controlled by hydroclimatic conditions, with the lowest skill occurring in water-limited environments that dominate much of Africa.

Overall, this study establishes practical limits on the reliability of current global discharge products and provides a framework for their evaluation in data-scarce regions, highlight the need for targeted bias correction and improved hydrological monitoring across Africa.