Abstract (eng)
The COSMO-D2 EPS is the very high resolution, limited-area ensemble prediction system (L-EPS) maintained at the German Weather Service (DWD) and has an horizontal resolution of 2.2 km. At such spatial scales, which lie at the lower end of the mesoscale, deep convection does not need to be parametrized and can instead be resolved directly in the model. At the same time, the development of innovative parameters which combine synoptic scale forcings and intra-cloud physics, like the Lightning Potential Index (LPI), significantly increased the potential accuracy when forecasting heavy showers and thunderstorms. However, such improvements in spatial resolution and modeling also need a proper verification approach in order to put into perspective grid-point related issues such as the double-penalty effect. The probabilistic approach of an EPS applied to high resolution models could nonetheless help increasing the accuracy and the predictability also in case of very localized convective phenomena. The first part of this work is dedicated to the analysis of the two datasets used (the LPI from the COSMO-D2 EPS and the observed lightning activity from the LINET observation network). A preliminary verification based on a conventional measure such as the Symmetric Extremal Dependence Index (SEDI) has also been conducted. In the second part, fuzzy and object based verification methods such as the dispersion Fractions Skill Score (dFSS) and the ensemble-SAL (eSAL) has been used to analyze the COSMO-D2 EPS forecasts of the LPI. This second part is focused on better understanding the spread-errpr relationship in the model, thus investigating possible positive effects on the predictability of convection. In general, the COSMO-D2 EPS tends to generate too little dispersion in its members if compared to the actual model error. Specifically, the ensemble mean generates useful lightning activity forecasts at a spatial scale of around 200 km for the afternoon hours, while the spatial spread of the ensemble members lies at more or less 100 km.