Global estimates of damaging hail hazard

Abstract

A single hailstorm can cause losses in the billion-dollar range if it occurs over a densely populated area. Property losses from hailstorms are rising with time mainly due to a combination of increases in population density and wealth. Report based observational hail data alone are highly inhomogeneous and unable to discriminate between climate and societal changes. Here we present a statistical approach that estimates hail hazard from large-scale environmental conditions. Using daily ERA-Interim reanalysis data and large hail observations (diameter larger than 2.5 cm) from the conterminous United States (CONUS) we show that four predictors enable skillful discrimination of large hail frequencies on a regional scale.The predictors include atmospheric instability, freezing level height, and 0–3 km wind shear and storm relative helicity. These variables are used to develop a hail algorithm, which provides the probabilities for large hail occurrence from regional to global scales and from daily to climate timescales. The algorithm skill is tested over the CONUS and with independent hail observations from Australia and Europe. It skillfully captures the frequency, annual cycle, spatial patterns, and interannual variability of observed large hail records in a large variety of climate regions. Deficiencies are found in regions with strong orographic forcing and low shear environments. The algorithm outperforms established severe convection indices in terms of more accurately predicting absolute hail frequencies and the annual cycles of large hail in all tested regions. The code is open-source and is applicable to a variety of tasks including daily to seasonal forecasting and assessing climate change influences on hail hazard.