Hail Trend Estimation in Germany utilizing Radar-based Hail Tracks, Convective Parameters, and Machine Learning Techniques

Hail is among the most destructive hazards associated with severe convective storms. Nevertheless, direct observational records are limited. We therefore based our analysis of changes in hail frequency and intensity in response to climate change on severe convective cell tracks identified over 20 years from 3D radar data of the German Weather Service (DWD). These tracks are supplemented and validated with lightning detections, hail reports from the European Severe Weather Database, and insurance claims.

To establish a link between hail occurrence and its environmental drivers, convolutional neural networks (including U-Net architectures) are trained on convective parameters from ERA5 reanalysis to predict hail events. These models facilitate the simulation of hail climatologies and the estimation of hail occurrence for periods without observational data, thereby extending trend analyses into the more distant past.

Trend analyses of the radar-based tracks show a distinct spatial pattern: while a significant increase in hail activity in southern Germany is observed, no or slightly negative trends in northern Germany have been identified. The ML models demonstrate a high degree of success in replicating this distinct spatial pattern, thus indicating that the observed trends are likely driven by changes in large-scale convective environments rendered by the models. Current efforts are centered on the quantification of the relative importance of individual convective parameters for hail prediction. This will offer further insight into the physical drivers behind these changes.

The insights gained into the physical mechanisms behind hail formation enhance our ability to interpret and constrain hail projections in future climate scenarios. This bridges the gap between past climatologies and future risks.