Exposure and vulnerability estimation for modelling flood losses to commercial assets in Europe

Commercial assets comprise buildings, machinery and equipment, which are susceptible to floods. Existing damage models and exposure estimation methods for this sector have limited transferability between flood events. In this study we introduce two methodologies aiming at broader applicability: (1) disaggregation of economic statistics to obtain detailed building-level estimates of replacement costs of commercial assets; (2) a Bayesian Network-based (BN) damage model to estimate relative losses to the aforementioned assets. The BN model is based primarily on six post-disaster company surveys carried out in Germany after flood events that had occurred between 2002 and 2013, which is a unique source of microdata on commercial losses. The model is probabilistic and provides probability distributions of estimated losses, and as such quantitative uncertainty information. The BN shows good accuracy of predictions of building losses, though overestimates machinery/equipment loss. To test its suitability for pan-European flood modelling, the BN was applied to validation case studies, comprising a coastal flood in France (2010) and fluvial floods in Saxony (2013) and Italy (2014) are presented as well. Overall difference between modelled and reported average loss per company was only 2–19% depending on the case study. Additionally, the BN model achieved better results than six alternative damage models in those (except for one model in the Italian case study). Further, our exposure estimates mostly resulted in better predictions of the damage models compared to previously published pan-European exposure data, which tend to overestimate exposure. All in all, the methods allow easy modelling of commercial flood losses in the whole of Europe, since they are applicable even if only publicly-available datasets are obtainable. The methods achieve a higher accuracy than alternative approaches, and inherently provide uncertainty information, which is particularly valuable for decision making under high uncertainty.