A simple method is presented designed to assess uncertainties from dynamical downscaling of regional high impact weather. The approach makes use of the fact that the choice of the simulation domain for the regional model is to a certain degree arbitrary. Thus, a small ensemble of equally valid simulations can be produced from the same driving model output by shifting the domain by a few of grid cells. Applying the approach to extra-tropical storm systems the regional simulations differ with respect to the exact location and severity of extreme wind speeds. Based on an integrated storm severity measure, the individual ensemble members are found to vary by more than 25 % from the ensemble mean in the majority of episodes considered. Estimates of insured losses based on individual regional simulations and integrated over Germany even differ by more than 50 % from the ensemble mean in most cases. Based on a set of intense storm episodes, a quantification of winter storm losses under recent and future climate is made. Using this domain shift ensemble approach, uncertainty ranges are derived representing the uncertainty inherent to the used downscaling method.
