No abstract

No abstract

No abstract

In June 1876, June 1910 and August 2005, northern Switzerland was severely impacted by heavy precipitation and extreme floods. Although occurring in three different centuries, all three events featured very 15 similar precipitation patterns and an extra-tropical storm following a cyclonic, so called Vb trajectory around the Alps. Going back in time from the recent to the historical cases, we explore the potential of dynamical downscaling a global reanalysis product from a grid size of 220 km to 3 km. We use the full, 56-member ensemble provided in the reanalysis and a regional weather model to investigate sensitivities of the simulated precipitation amounts to a set of differing model configurations. These setups are evaluated by combining spatial verification metrics, inter-20 subjective visual inspection and an objective similarity measure. The best-performing model setup, featuring a 1day initialization period and moderate spectral nudging, is then applied to assess the sensitivity of simulated precipitation totals to cyclonic moisture flux along the downscaling steps. The analyses show that cyclone fields and tracks are well defined in the reanalysis ensemble for the 2005 and 1910 cases, while deviations increase for the 1876 case. In the downscaled ensemble, the accuracy of simulated precipitation totals is closely linked to the 25 exact trajectory of the cyclone, with slight shifts producing erroneous precipitation, e.g., due to a break-up of the vortex if simulated too close to the Alpine topography. To reproduce the extreme events, continuous moisture fluxes of >200 kg m -1 s -1 from accurate directions are required. Misplacements of the vortex, in particular for the 1876 case, point to limitations of downscaling from coarse input for such complex weather situations and for the more distant past. On the upside, a well-reasoned selection of reanalysis members for downscaling may be adequate 30 in cases where the driving large-scale features in the atmosphere are well known.

No abstract