Potential and limitations of increased radar observations assimilated in the 1.3 km AROME model at convective-scale

Despite the high horizontal resolution of the convective-scale AROME model, using radar data at high-density every hour in the 3D-Var assimilation system remains a sensitive issue. Diagnostics of the AROME assimilation system computed a posteriori have been done (as the well-known Desroziers's diagnostic or the backgroung-error method) at 1.3 km of resolution. Focus will be made on how the comparisons of these diagnostics have shown that observation errors were spatially highly correlated. However, the increased radar observations have been done in the 1.3 km AROME 3D-Var assimilation system also accompanied by changes and improvements of the assimilation method of radar reflectivity allowed by massively parallel computing capabilities. The impacts of such increased radar observations on the initialization, triggering and development of the convection in the short term forecasts will be illustrated with forecast scores but also through specific high-precipitating Mediterranean convective events. These developments currently benefit to the AROME WMED model that has been set-up for the HyMex experiment over the Western Mediterranean. In particular, some results from the second reanalysis of the first special observation period (SOP1) in autumn 2012 will be shown.