Clustering Large-Scale Atmospheric Patterns Associated with Hailstorms

Le résumé fourni par la source

Hailstorms are among the most destructive convective weather phenomena, causing extensive damage to agriculture, infrastructure, and property. Improving hail prediction and hindcasting requires a deep understanding of the atmospheric conditions under which hail forms. This study analyzes hail occurrence across Europe using ERA5 reanalysis data and ground-based hail observations to characterize the meteorological environments conducive to hail events.Hail-prone conditions are identified based on a dataset of key atmospheric variables - including convective parameters, moisture and temperature profiles, and dynamic indices - at 0.25° spatial resolution on a daily timescale. Feature selection is performed using statistical relevance, variance, and correlation criteria to select variables that best capture the thermodynamic and dynamic processes involved in hail formation while minimizing redundancy.To explore patterns linking atmospheric conditions to hail events, this study applies dimensionality reduction and clustering techniques. The resulting classification reveals distinct synoptic and mesoscale regimes associated with hailstorms, highlighting spatial and seasonal variability across Europe. These clusters expose region-specific hail environments and highlight the atmospheric states most likely to generate hail.These regime-based classifications offer direct insights for improving convective-scale numerical weather prediction. By associating each cluster with specific model physics, this framework provides guidance for optimizing convection-permitting simulations - such as those using the Weather Research and Forecasting (WRF) model - to more accurately hail events/processes. Ultimately, this work contributes to improve hail prediction capability and robust climate-scale risk assessment of hailstorms across a range of European weather regimes.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.