Signatures of Climate Change in Weather Metrics Important for Catastrophe Model Development

The Insurance Industry is under increasing pressure to disclose to regulators how they are accounting for climate change. Many in the industry rely on Catastrophe Models to provide an accurate view of their climate risk. Despite the Industry focus on the short term (the next year), that view still has to include how climate change may have altered frequency, intensity, and other important metrics over the historical record without short-changing the climate variability which has occurred over the long term. A necessary exercise in creating catastrophe models is therefore analyzing temporal trends and spatial shifts, etc that may have occurred for relevant weather phenomena variables and/or large-scale parameters (ingredients).

In this presentation, temporal changes in ingredients important for the AIR Worldwide US Winter Storm, US Severe Convective Storm, and US Wildfire Catastrophe Models are evaluated using a combination of severe weather reports, other observations, and reanalysis data. Evaluation of reanalysis data for US winter storms shows statistically significant increases in the relative frequency of strong lows (< 960 hPa) for certain parts of the US. The results are consistent with documented changes in the strength and evolution of the Polar Vortex. Trends in parameters from reanalysis data relevant for severe weather (e.g., hail, tornado, damaging wind) show statistically significant changes both temporally and spatially. Maximum values of SHiP (Significant Hail Parameter), STP (Significant Tornado Parameter), and EHI (Energy Helicity Index) have been increasing steadily since 1979. First and last dates for each have been coming earlier and ending later respectively. Maximum areal coverage of values above threshold for each parameter have also been growing. Agreement with storm reports are somewhat consistent but difficult to assess given population biases and other reporting inconsistencies. For the wildfire peril, significant increases in various metrics of Diablo and Santa Ana Winds in California are evident from reanalysis data and observations.

Using long-term historical data that exhibit climate trends can therefore dilute/contaminate the view of current risk. More information on the historical trends found as well as some synoptic perspective will be presented as well as ideas for constructing catastrophe models that more faithfully represent both the current mean climate state, the appropriate interannual variability, and hence the current risk.