The impact of winter storms in Switzerland – prototyping decision-support tools
Winter windstorms are among the most destructive and costliest natural hazard events in Europe. Insured losses from the events Daria 1990, Lothar 1999, and Kyrill 2007 amount to 8.6 billion, 8.4 billion, and 7.1 billion USD respectively. In Switzerland, the most recent severe event Burglind/Eleanor caused 165 million CHF in building damages additional to other socio-economic impacts like forest damage and disruption in traffic and power supply. Societal decisions for the management of winter windstorm risk require information about those impacts to be able to reduce them or handle them sustainably and efficiently. This thesis documents the development and prototyping of decision-support tools for different questions of risk management. The tools are implemented in the CLIMADA framework, an open-source risk-modelling platform in the programming language python. It models risk as a combination of hazard, exposure and vulnerability. Meteorological and climatological research has culminated in hazard datasets describing the intensity, spatial distribution, and likelihood of storm events. This thesis implements, calibrates and evaluates risk information derived from the combination of these hazard datasets with exposure and vulnerability. The exposure describes the assets or value at risk: a spatial description of infrastructure, natural resources, population, or vulnerable groups. The vulnerability defines how the proportional impact to exposure, such as a damage degree, is linked to the hazard intensity. One tool describes the risk assessment of building damages from winter windstorms and is applied to the building insurance industry. The second tool forecasts building damages based on weather forecasts. It firstly supports decision-making for preventive actions in the building insurance industry, and secondly can be used in the context of incorporating impacts into weather warnings by national meteorological services. In chapter 1, the scientific context of the tools is introduced regarding winter windstorms, impact data, risk modelling, risk assessment and decision-making. In chapter 2, the socio-economic impacts of winter windstorms are illustrated by the example of a recent event. The impacts of the winter windstorm Burglind/Eleanor (3 January 2018) in Switzerland are collected from different government agencies and other organizations. The event is responsible for the largest infrastructure and forest damage from a winter storm in Switzerland since Lothar 1999. Burglind/Eleanor caused building damages of around 165 million CHF. There were also disruptions to road- and rail-traffic and in power supply. In Swiss forests, around 1.3 million cubic metres of wood were felled. Chapter 3 reviews two newly published hazard datasets of winter storms, one containing high-resolution gust speed intensities of more than 140 historical storms, and the other a synthetic event set with more than 7'660 events generated in climate model runs. A comparison with previously used datasets in industry and academia reveals that the historical dataset represents similar storm severity characteristics to the other datasets. Its high spatial resolution, the long historical time period covered, and its open-access and free availability recommend this dataset for use in risk assessments in industry and research. The synthetic event set shows different storm severity characteristics from industry and academic datasets as well as from the historical dataset, as especially the spatial extends of the events were smaller. The use of this particular synthetic event set in risk assessments is cautioned. In chapter 4, the first decision-support tool, i.e., risk-assessments for building damages, is presented in an applied context. The open-access historical event set from chapter 3 and a purpose-built probabilistic event set are used in a risk assessment for the insurance industry in a collaboration with the cantonal building insurance of the canton of Zurich GVZ. Insurance companies, with access to their claims data, have been in a good position to assess their risk from winter windstorms. The risk assessments based on claims data are compared with risk assessments from modelled building damages from the GVZ proprietary impact model and the open-source impact model within the CLIMADA platform. Insurance companies can benefit from complementing their claims-based risk assessments using the newly available events sets, especially concerning rare events. In a special focus, the uncertainties of the different approaches are discussed and illustrated. Chapter 5 presents the second decision-support tool, a newly developed impact forecasting system for building damages from winter windstorms in Switzerland. Since societal decisions on preventive actions are best supported with estimations of expected impacts of the weather events, national meteorological services aim to incorporate impacts into their warnings. This system's forecasted impacts support decision-making for specific users, e.g. the building insurance sector who need to pre-allocate additional resources for claims adjustments and claims handling. In a comparison with claims, the impact forecasts of building damages from winter windstorms are promising, but for other wind phenomena like thunderstorms and foehn storms they do not work as reliably. This is followed by a synthesis of the thesis and the four main chapters. The uncertainties in impact modelling for risk assessment and impact forecasts are discussed. In future endeavors, the uncertainties in exposure and vulnerability could be explicitly represented in an ensemble of opportunity of different model implementations. This would allow a comparison of the uncertainty of the hazard, exposure and vulnerability components, and a discussion of second order uncertainty. The role of exposure in impact warnings is highlighted: it provides the metric and spatial pattern of impact forecasts which are straightforward and relatable information. Suggestions for future topics and implications in research and applications are provided in an outlook. This thesis demonstrates how the decision-support tools within the CLIMADA framework in combination with climatological datasets, as well as open-access weather forecasts, can provide actionable risk assessments and impact warnings. The free, open-source methods lend their support for the suggested future developments.
