AEWs and weather prediction in West Africa

As compared with midlatitudes the improvement of Numerical Weather Forecast (NWP) in the Tropics is very slow even when using metrics better adapted to low latitudes characterized by a large Rossby radius. For instance a recent study shows that the skill of AEW forecast within different operational NWP systems is less than 48 h, and the synoptic variability and the coupling between AEWs and precipitation are very different between those NWP systems.

The basic reasons for such biases and poor skills at short forecast ranges are multiple and depend on the quality of observations, of the data assimilation, as well of the models, but also on the level of predictability in West Africa. The African Monsoon Multidisciplinary Analyses (AMMA) international project contributed to reactivate and modernize the radiosonde network over West Africa and provided complementary observations to document all components of the West African Monsoon (WAM) system. The assimilation of such improved observation networks and of alternative data such as satellite microwave channels over land improves significantly initial fields as compared with independent observations provided by AMMA such as driftsondes or GPS precipitable water. Nevertheless forecasts initialized from those analyses lost this advantage within the first 24 h. An ensemble-based sensitivity analysis confirms this rapid decay of the positive impact of the initial fields for the dynamics. Nevertheless the impact of the initial moisture increases with the forecast range in agreement with an increasing sensitivity to the model representation of diabatic convective terms.

We conclude that the major difficulty to forecast the synoptic activity over West Africa roots in the strong interactions between the AEWs and convection that are not properly represented by current NWP models especially for the triggering stage. The convection is very sensitive to humidity field due numerous non-linear and small scales processes (source of energy, surface fluxes, detrainment of convection, evaporation of precipitation generating convective wakes…) that are not well represented in models. The difficulties are magnified over West Africa due to the mean dry conditions and to the large variability of the humidity field prevailing in that region. Cloud resolving models are a way to study this coupling between AEWs, convection and the surface, and to improve the representation of the key processes involved in the NWP parameterization. In parallel it is crucial to further the effort to improve the operational observation networks over Africa and the data assimilation for both surface and atmosphere using more satellite data.

Complementary to this NWP research and development, forecasters have a major role to adapt and improve the forecast at short and medium ranges based on their experience, on NWP products and on real time observations. AMMA helped to develop the WASA/F method to guide their work by integrating their experience and methods with our present understanding of processes. New diagnostics are currently developed and tested to monitor and forecast the synoptic activity and the intra-seasonal variability, showing a potential of predictability to explore.