Using the example of retrospective analysis of storm activity since 1950 in Northern Europe, we apply the same strategy to tropical storm activity in E Asia – namely post-processing NCEP re-analyses with a spectrally nudged regional climate model (RCM), which is run for many years and even decades of years. It turns out that the regional model indeed generates TCs, which are considerably deeper and stronger than in the NCEP re-analysis. They are formed in thee interior of the model domain (without bogussing); most observed storms are simulated, albeit usually somewhat too weak (even if the numerical resolution is 0.165 deg). When the same season is done in an ensemble of simulations and no spectral nudging is administered, then different numbers and tracks of storms are simulated. This emphasizes the significance of the previously identified RCM-phenomenon “intermittent divergence in phase space” when dealing with TCs.

The skill of the method is examined in a series of case studies; also the two rather different seasons of 1994 and 1998 are studied - and it turns out that the different characteristics are indeed recovered by our dynamical downscaling method.