Flood exposure and vulnerability estimation methods for residential and commercial assets in Europe

Floods affect many types of tangible assets in Europe. Here, we present novel methods of estimating exposure and vulnerability of four important categories of assets: (1) residential buildings, (2) household contents, (3) non-residential buildings, (4) machinery and equipment. One set of models was developed for “residential” assets (1+2) and another one for “commercial” assets (3+4) within the activities of EIT Climate-KIC “SaferPLACES” project. All methods are applicable within Europe using only openly-available datasets at the level of individual buildings. Residential exposure is calculated through combination of estimated useful flood space of buildings and average national replacement costs of buildings and household contents. We devised a method for estimating building height using OpenStreetMap and pan-European raster datasets and transforming it into floor space area. Combining economic and demographic data, we calculated replacement costs of residential assets per m² of floor space in 30 European countries, covering years 2000–2017. Regarding commercial assets, we created a method to disaggregate economic statistics to obtain detailed building-level estimates of replacement costs of non-residential buildings and machinery/equipment. Finally, we developed two Bayesian Network-based, probabilistic, multivariate models to estimate relative losses to residential and commercial assets. Those combine post-disaster survey data from Germany with information on flood hazard and vulnerability; data from events in the Netherlands and Italy are also used in the residential damage model. The exposure and damage models were tested for a case study of the 2010 coastal flood in France, which is a challenge especially for the damage models built on the basis of fluvial and pluvial floods. Using water depths from hydrodynamic modelling of the event, we found that the models underestimated both residential and commercial losses in the two French departments most affected by the 2010 event, but shown a very close alignment for average losses recorded in the whole affected area. Good performance of the models was achieved in all investigated subcategories (residential buildings’ structure, household contents, agricultural establishments, companies in services & industry). Our models were compared with alternative damage models and exposure estimates, showing the best accuracy in most aspects analyzed.