Past storms and future insurance risk
September 2018 - Windstorms result in some of the largest natural hazard impacts in Europe, both in terms of societal disruption and economic impacts.
Research between the Met Office and PERILS, an organisation that uses insurance industry information to improve understanding of natural hazard risk, is helping to quantify the impact of European windstorms in order to better assess potential losses.
Modelling of historic European windstorms
Windstorms – or extratropical cyclones – result in some of the largest natural hazard impacts in Europe, both in terms of societal disruption and economic impacts. Insured losses from major windstorm events, such as the storms that struck France over Christmas 1999, can reach several billion Euros. To ensure that the insurance industry can set premiums commensurate with the risk and maintain solvency to continue to provide protection to policyholders, it is vital to understand the likely frequency and severity of major storms. With digitisation, records of insured losses over the past decade are relatively complete; however, this period is not long enough to derive reliable conclusions about the likelihood of major damaging events which are relatively rare.
For this reason, the Met Office and PERILS have been working together to compile a catalogue of European windstorms dating back to 1979. This catalogue contains footprints of more than 200 events relevant for the re/insurance industry that describe the maximum gusts over a 72-hour period. To extend the catalogue back to 1979, the footprints have been derived from a reanalysis data set, ERA-Interim, which has been downscaled to higher spatial resolution. Using a reanalysis has three main advantages: it provides a modelled climatology of the past which is consistent through a period when forecast models have been developing rapidly; it incorporates all available observations from the period; it provides uniform coverage on a model grid.
Property damage and wind gusts have been shown to be closely related and, since wind gusts can vary significantly on short space and time scales, it is important to generate footprints with as much detail as possible. Downscaling using a higher resolution model (4km) embedded within the reanalysis provides a method for creating greater local detail while maintaining physical plausibility.
Translating wind speed into insured losses
The storm catalogue can be used to derive the relationship between insured losses and gust speeds. For each of the around 20 events for which PERILS has detailed insurance loss data a damageability function can be generated, relating actual damage as a percentage of insured values to the gust speeds. These functions can then be tested by applying them to the entire set of the historic events wind fields, i.e. replacing the gust values with the corresponding damage degrees and applying these damage degrees to the insured values. The result for each storm event is a modelled loss footprint which can be compared to actual loss footprints. By application of this scenario loss model, event losses for all 200 storm events over the past 39 years have been computed, indicating the losses that would have been suffered at today’s values and vulnerability relationships.
Key findings from history
According to PERILS, the biggest single event loss is from windstorm Lothar which occurred in December 1999 and which would have resulted in a loss of EUR 9.9bn if it occurred today. The worst season is 1989/1990 with a total loss of EUR 19.2bn, primarily from windstorms Daria, Herta, Judith, Vivian and Wiebke. The most benign season is 2012/2013, with total losses below EUR 100m. The average annual loss costs are EUR 2.6bn. There are 39 events which exceed EUR 510m, implying that this loss is reached or exceeded on average once a year. These numbers comprise the following countries: Austria, Belgium, Denmark, France, Germany, Ireland, Luxembourg, the Netherlands, Norway, Sweden, Switzerland and the United Kingdom.
These 200 event loss footprints can be used by insurers to generate for each market an exceedance-probability (EP) curve that enables return periods of loss levels to be determined and average annual losses to be calculated. Such EP curves can also serve to validate the results from commercially available natural catastrophe models.