Improved Real-Time Hail Damage Estimates Leveraging Dense Crowdsourced Observations

Abstract

Severe hail storms are a leading cause of building damages in Switzerland, yet accurately observing hail using weather radar remains challenging. Opportunely, Switzerland benefits from a uniquely dense network of crowdsourced hail reports, providing an additional data source. Since 2021, over 50,000 reports were submitted each hail season through the national weather service's mobile application, including some false reports. In this study, we apply a rigorous filtering approach to these reports, including the implementation of a 4D-DBSCAN clustering algorithm, to develop a gridded hail size product. Using 65,000 hail damage claims from August 2020 to September 2023, an impact function is calibrated and used to model hail damage to buildings. The new crowdsource-based hail size product improves hail damage estimates in comparison to the radar-based data, largely due to an improved distinction of severe and sub-severe hail within a storm. The model can approximate the number and cost of hail damages to any user-provided building portfolio in real time, facilitating the management of the aftermath of a hail storm.