The Influence of DACCIWA Radiosonde Data on the Quality of ECMWF Analyses and Forecasts

During the extensive field campaign of the DACCIWA (Dynamics-Aerosol-Chemistry-Cloud Interaction in West Africa) project ~900 radiosondes were launched from twelve stations in southern West Africa between 15 June and 31 July 2016. In order to assess the impact of these additional observations on the quality of both analysis and forecast data, data denial experiments were conducted using the Integrated Forecasting System of the European Centre for Medium-Range Weather Forecasts (ECMWF). An experiment with all radiosondes assimilated is referred to as “DACCIWA”, while the denial experiment “noDACCIWA” is more representative of the current state of the observational network in West Africa. Mean 6-hourly differences between the two analyses reveal largest observation impacts on low-level winds and temperatures during the night and early morning, in particular downstream, i.e. to the north, of the radiosonde stations. Weaker monsoon winds in the “DACCIWA” analysis over Ivory Coast and Ghana lead to higher nighttime temperatures due to weaker cold air advection from the Gulf of Guinea, which in turn reduces low-level relative humidity and cloud cover. Increased high-level cloud cover indicate an impact on deep convection. Low-level differences are generally much weaker during daytime, likely due to the effect of turbulent mixing in the boundary layer. Assimilating the radiosondes also causes a discernible increase in low-level and column moisture over and adjacent to Benin. Disappointingly, the positive impact is lost in the forecasts after a day or less. Improvements are seen in low-level wind and temperature but also in precipitation over the downstream Sahel, while impacts on outgoing longwave radiation are somewhat ambiguous. This suggests that the main limiting factors for better ECMWF forecasts in West Africa are model error and data assimilation. It is hoped that the wealth of DACCIWA observations and the improved physical understanding can contribute to reducing these errors in the future.