Added-Value of 3DVAR Data Assimilation in the Simulation of Heavy Rainfall Events Over West and Central Africa

This study aimed to evaluate the ability of a numerical weather prediction (NWP) model to capture the spatial distribution and magnitude of rainfall during three recent intense events (15–17 June 2011, and 23–25 August and 04–06 September 2012) observed over West and Central Africa, as well as the associated atmospheric and near-surface conditions. For each event, two numerical experiments were performed using the regional Weather Research and Forecasting (WRF) Model without (CNTL) and with (DA) data assimilation. Simulations were initialized using Global Forecast System data. The analyses were updated with the three-dimensional variational (3DVAR) technique using PrepBUFR and radiance observational data in a time window of ±3 h. The potential added value of data assimilation was addressed by comparing meteorological variables such as relative humidity, zonal and meridional wind components, 2 m temperature, and rainfall with the European Centre for Medium-Range Weather Forecasts reanalysis and the Tropical Rainfall Measuring Mission satellite-derived rainfall product datasets. WRF accurately simulated the spatiotemporal propagation and the zonally extended structure of rainfall as well as of relative humidity, 2 m temperature, and horizontal wind components. DA exhibited different biases, root mean square error, and spatial correlation, leading to mixed results in terms of outperforming CNTL. Results indicated that there was an increment in control variables, implying an added value from 3DVAR to the initial and boundary conditions. Rainfall forecasts were improved by 15–25 %. Uncertainties in the simulation of intense events in the study domain were noted, but improvement resulting from DA was limited due to lack of assimilated data for the region.