Toward the validation of a National Risk Assessment against historical observations using a Bayesian approach: application to the Swiss case

Abstract

National Risk Assessment (NRA) studies have recently received increased interest from governments, authorities and other involved stakeholders, e.g. civil protection agencies, emergency planners, etc. The NRA approach combines risk assessment methods and decision-making processes in a structured manner to support the prioritization and management of national hazards and threats. A NRA commonly builds upon an intensive information processing and judgmental exercise, therefore, in a qualitative or semi-quantitative way, requiring the cooperation of experts from various fields. The validation of a NRA is of great importance to ensure that the analysis of national hazard scenarios results in the implementation of adequate preparedness strategies and contingency planning. In this study, a first step toward a comprehensive framework for NRA validation is presented. The method is developed to evaluate the accuracy and consistency of establishing risk indicator values by expert elicitation by comparing it against historical experience. The proposed method is based on Bayesian analysis, and is tested in three hazard scenario case studies for the Swiss NRA. The results for the chosen Swiss NRA cases show that expert estimates and quantitative assessment of historical experience are overall in good agreement. However, scenario-specific assumptions and boundary conditions defined by the experts are not in all cases supported by evaluation of historical experience. Finally, the main drawback of the proposed approach is that it can only be applied to hazard scenarios with sufficient historical observations, while it has the advantage to be generic and thus applicable to other risk assessment contexts outside of NRA.