Improving Numerical Weather Forecasts of Rainfall Over Central Africa Using a Three‐Dimensional Variational Data Assimilation Technique and the Weather Research and Forecasting Model

The impact of satellite radiance data assimilation using the three‐dimensional variational (3DVAR) technique within the WRFDA system was investigated to assess potential improvements in numerical weather forecasts over Central Africa. Eight experiments grouped into two sets were conducted for four extreme rainfall events, including the catastrophic flood that occurred in Douala on August 21, 2020: a control simulation without data assimilation (WRF) and a simulation including radiance assimilation (WRFDA). Results indicate that the assimilation of satellite radiances can lead to modest improvements in the representation of rainfall and atmospheric variables. For rainfall, the correlation coefficient remains relatively low (around 0.43–0.44) due to the strong spatial variability associated with convective systems. However, the WRFDA simulation slightly reduces forecast errors, with RMSE decreasing from about 7.5 mm in WRF to about 7.1 mm and MAE decreasing from approximately 3.8 to 3.6 mm. For near‐surface temperature, both simulations show relatively high correlations (generally above 0.85) with similar error levels, indicating that the assimilation produces only limited changes in temperature fields. Relative humidity exhibits moderate correlations (around 0.70–0.82), with comparable performance between WRF and WRFDA. For precipitable water, correlations are generally high, reaching values close to unity for some events, suggesting that the large‐scale atmospheric moisture content is reasonably well represented. Overall, the results suggest that satellite radiance assimilation in WRFDA can slightly improve the representation of rainfall and atmospheric moisture over Central Africa, although the differences between the WRF and WRFDA simulations remain relatively moderate. These findings highlight the potential value of satellite data assimilation for improving short‐term weather forecasts in data‐sparse and convectively active tropical environments.