Empirical evaluation of windstorm losses and meteorological variables over the Netherlands

This study investigates windstorm impacts by combining high-resolution wind hazard data with a unique asset-level insurance loss dataset, specifically focusing on the Netherlands. We conduct statistical analyses to associate wind hazard characteristics with spatial data on windstorm losses at various spatial aggregation levels (four-digit to nationwide postal codes). Different wind hazard intensities (e.g. maximum wind gust, maximum hourly wind speed) are derived using meteorological data from 2017 to 2021 (the same period as the loss data). This data is based on station and downscaled ERA5 reanalysis data. Results show that the recorded gust has a good correlation with damage components (r = 0.41–0.61). The downscaled reanalysis data on gust and daily maximum (hourly mean) wind speed also have a good correlation (r = 0.38–0.59), albeit a bit smaller than the observed gust. When comparing different levels of aggregated data (PC4—four-digit postal code, PC2—two-digit postal code, and NL—national level), the correlation between claim and loss ratios becomes more pronounced as the level of aggregation increases. In addition, at the aggregated data level of two-digit postal codes, we see a wind speed threshold (around the 98th percentile of the records, ~ 22 m/s), where both losses and reported claims begin to rise as wind speed increases. Nevertheless, with lower wind speeds, damages and reported claims become meaningful using more aggregated data (NL). Our findings highlight the complex link between hazard and damage variables for windstorm losses, offering valuable insights for insurance portfolios, risk assessment, and management.