Studying monsoonal convective extremes at high resolution in the Dakar region

Sub-Saharan West Africa is among the regions with the highest climate risk, especially the semi-arid Sahel region which is characteristically a climate hotspot. The Sahel lies at a large latitudinal gradient of mean temperature and precipitation. Therefore, even small variations in the seasonal meridional migration of the Intertropical Convergence Zone (ITCZ) lead to drastic hydro-climatic shifts, such as pronounced drought or increased flood risk --- with severe socio-economic consequences. Densely populated and rapidly growing urban areas, such as Dakar, Senegal, with many informal settlements that have limited adaptation capacity, are disproportionately affected by intense rainfall events. These often lead to flash floods and consequently pose a threat to human lives and infrastructure.

In a future climate, research suggests that the frequency of extreme weather associated with mesoscale thunderstorm, or convective, systems (MCS) will increase and hence improved warnings are required. However, the sparseness of observational data in the region makes reliable prediction of the initiation and evolution of MCS near impossible. Also, forecasts from numerical weather models often exhibit low skills in this region. For the improvement of risk preparedness, short-term prediction based on statistical inference from observational data referred as 'nowcasting' at a lead time of 1-6 hours, is a promising option that can outperform dynamical models. In the current study a new high-resolution observational network for MCS is described. The automatic weather station (AWS) network, currently consisting of 14 multi-variable stations including atmospheric and soil variables, sends data at 1-min resolution to a data cloud through the local mobile network. Within the project "High-resolution weather observations east of Dakar (DakE)" additional 60 low cost weather sensors and 12 flood sensors have been installed. This network of stations will contribute to the understanding of sub-mesoscale (100m-10km) features that are typically under-resolved by a typical operational network and under-represented in numerical models.