Risk assessment of building damage related to extreme winter windstorms in the canton of Zurich, Switzerland

We compare the risk assessment for storm related building damage based on three different foundations: (1) insurance claims data, (2) modelled building damages based on a historic event set of wind gust data, and (3) modelled building damages based on a probabilistic extension of the historic event set. Windstorms cause large socio-economic damages in Europe. In the canton of Zurich (Switzerland) they are responsible for one third of the building damages caused by natural hazards.

The Wind Storm Information Service (WISC) of the Copernicus Climate Change Service provides open wind gust datasets for the insurance sector to understand and assess the risk of windstorms in Europe. This is the first open climatological data set covering a longer time range than the insurance claims data of most small insurance companies. Our science-practice collaboration is a case study to illustrate how climatological data can be used in risk assessments in the insurance sector and how this approach compares to risk assessments based on proprietary claims data. We describe and use a storm damage model that combines wind gust data with exposure and vulnerability information to compute an event set of modelled building damages. These modelled damages are used to calculate relevant risk metrics for the insurance industry like the annual expected damage (AED) as well as the damage of rare events, with a return period of up to 250 years.

The AED calculated based on the insurance claims data (i.e. the mean damage over the observation period of 35 years) is 2.34 million Swiss Francs (CHF). This is almost double the value of the AED computed based on the storm damage model and historic event set (CHF 1.36 million). The storm Lothar/Martin in December 1999 is the most damaging event in the insurance claims data (CHF 62.4 million) as well as the historic event set (modelled building damage of CHF 62.7 million).

Both the insurance claims data and the modelled building damages based on historic events are not well suited to derive information about rare events with return periods considerably exceeding the observation period. To provide some information about rare events, we propose a new probabilistic event set, by introducing various perturbations, resulting in 4’200 events. This probabilistic event set results in an AED of CHF 1.45 million and a damage amount of CHF 75 million for a return period of 250 years. The probabilistic event set allows for testing the sensitivity of the risk to e.g. portfolio changes and changes in the insurance condition for events of a higher intensity than the historic events.

Our analysis is implemented in the GVZ’s proprietary storm damage model as well as the open-source risk assessment platform CLIMADA (https://github.com/CLIMADA-project/climada_python). This guarantees scientific reproducibility and offers insurance companies the opportunity to apply this methodology to their own portfolio with a low entry threshold.