Evaluating Seasonal Rainfall Forecast Gridded Models over Sub-Saharan Africa

Abstract The agricultural industry in Africa has recently been impacted by rainfall variability and long-term changes in amount and distribution. Reliable rainfall forecasts on a daily timescale are vital for in-season decision-making. This study evaluated the relative prediction abilities of the European Centre for Medium-Range Weather Forecasts (ECMWF-S5) and the NOAA Climate Prediction System (CFS) gridded rainfall models across Africa and three sub-regions from2012–2022. The results indicate that the performance of both models declines with increasing lead times and improves with aggregated or coarser temporal resolutions. Besides, the ECMWF-S5 data represents observed daily rainfall better than the CFS data at all lead times, particularly in West Africa. On dekadal timescales, ECMWF-S5 outperformed CFS in all sub-regions, confirming its superiority. Both models are great at capturing rainfall at low elevations than at high elevations. CFS tends to overestimate low- and high-intensity rainfall events, while the ECMWF-S5slightly underestimates low-intensity rainfall events and accurately captures high-intensity events over Africa. Overall, the accuracy of these models in forecasting rainfall patterns in Africa varies according to the lead time, region, intensity of rainfall, and elevation. As a result, it is vital to use effective bias-correction approaches on these models to increase their accuracy and dependability for use in several sectors. This study emphasized the potential and shortcomings of the CFS and ECMWF-S5 models for climate impact studies, particularly in West Africa and regions with low elevations.