We will present an analysis of the performance of the UKV model for a number of severe events in 2011 which showed that many of the most serious model errors appear to come from errors in the boundary conditions provided by the lower resolution model that the model is nested in. We would hope that this issue will be addressed by the ensemble system which will, initially, simply be downscaling the regional ensemble.
Another class of problem with the 1.5km model is the character of convection represented. For example the 1.5km model tends to have too large convective cells and too much heavy rain and too little light rain. In order to understand this we are carrying out research with the model at a number of gridlengths between 1.5km and 100m. This work will also help to address future decisions about whether to use increases in computer resources on higher resolution as opposed to more ensemble members, larger domains etc. We will present a number of results of running the model for convective cases at several high resolutions including discussion of whether we see any signs of convergence and the effects of vertical resolution. This work is being carried out in conjunction with the DYMECS project which is discussed in another paper at this conference and we will also validate the performance of the high resolution models against this observational dataset.