Sensitivity of the 2014 Pentecost storms over Germany to different model grids and microphysics schemes

Abstract

At Pentecost 2014, following a period of hot weather, northwestern Germany was affected by a series of severe convective storms leading to fatalities and significant damage from strong winds, heavy precipitation, hail, and lightning. We present convection‐permitting numerical simulations for 2 days of this event (8 and 9 June) using the COnsortium for Small‐scale MOdeling (COSMO) model. Whereas a control run in a nearly operational configuration successfully reproduced the convective events of the first day, it failed to adequately reproduce the events of the second day. To further assess the predictability of this event, sensitivity studies with an enlarged model domain, finer horizontal and vertical grid spacing, and a double‐moment microphysics scheme were performed. Results show that enlarging the model domain improved the results over France and the Netherlands by better resolving deep convection and secondary cell initiation in that area, but the model failed to simulate deep convection over Germany. The increase of model resolution from 2.8 km to 2.2 and 1 km had minor effects only. However, the errors of accumulated precipitation were diminished independently of the model domain and grid resolution on the first day when using the double‐moment microphysics scheme. A better result for 9 June was obtained by using later initialization times (0300 and 0600 UTC instead of 0000 UTC). The control run initialized at 0000 UTC produced one cell, which decayed while travelling over the Netherlands. In contrast, the runs initialized later produced thunderstorms over Germany as well, due to outflow triggering or cell splitting. Finally, different cloud condensation nuclei (CCN) assumptions revealed a systematic relationship for condensate amounts of cloud water, rain and ice with increasing CCN. However, the evaporation of raindrops at lower levels led to a non‐systematic response of accumulated precipitation to CCN. In addition, grid spacing effects on aerosol–cloud interactions were assessed.