Background / research gaps
In Switzerland and many other locations on the planet, hail causes extensive damage. Furthermore, the setting for hail research in Switzerland is unique, thanks to its complex topography and to the variety of hail-related data sources. Radar-based hail algorithms and thunderstorm nowcasting systems already supply a basis for hail warning systems today. However, new hail ground observation data sets, simulation methods and radar products provide new opportunities for improving and extending the knowledge on convective systems, hail and its prediction. Concretely, in collaboration with the Mobiliar Lab for Natural Risks, MeteoSwiss has implemented a hail crowd-sourcing function in its app and has gathered several ten thousand reports from the Swiss Population. Furthermore, 80 automatic hail sensors are being placed in the Swiss hail hot spot regions and will measure the hail kinetic energy and momentum for 8 years. MeteoSwiss has also implemented a new hail diagnostic in the operational numerical weather prediction model and is testing it further. This project has the goal of jointly analysing these data sources and extending the understanding of convective systems to optimise hail prediction in Switzerland.
Approach and expected outcome
In the first approach, we pre-processed the crowd-sourced reports and used them to successfully verify the existing radar-based hail algorithms. The results provided new insights about the algorithms and we have a better idea on how they can be developed further. As a next step, we will conduct regional analyses of past hailstorms and their convective environment. As a tool, we use the already developed Thunderstorm Radar Tracking (TRT) which automatically detects and tracks thunderstorms in radar images. Along these tracks, we store the radar values, observations and measurements at the ground and model simulations. Consequently, statistical analyses and Machine Learning techniques can recognise patterns in the thunderstorms, from their initiation until their decay and thus simplify the prediction of hail. We expect that by the end of the project, we will have a better understanding of regional differences in the risk of hail and meteorological processes leading to hail in different synoptic weather situations. Furthermore, we expect these analyses and the machine learning models to improve hail nowcasting and increase the hail predictability.
