Ignoring spatial dependence misestimates flood risk at the European scale

Flooding is a major problem worldwide causing many fatalities and economic losses. The quantification of flood risk can be difficult for large spatial scales due to its spatial variability. The traditional risk assessment approaches assuming unrealistic spatial homogeneity of flood return period for the entire catchment are often used and hence in many cases lead to misleading results especially for large-scale applications. In this study, we aim at investigating the influences of spatial dependence in flood risk estimation over national and continental scales by comparing the assessments under three spatial dependence assumptions: modelled dependence (MD), complete dependence (CD) and complete independence (CI) of flow return periods. In order to achieve the aim, we develop a copula-based model representing the dependence structure of annual maximum stream flow (AMS) at 507 stations (with basin area > 500km2) across Europe and use it to generate long-term (10000 years) spatially coherent AMS at these locations. The generated series at multiple sites are then used for estimating associated flood loss considering two levels (with and without) of flood protection. The flood risk is estimated and aggregated for the representative 3 regions (England, Germany and Europe) and for the three dependence assumptions considering also the role of tail dependence of the used copulas. The results highlight that ignoring spatial dependence misestimates flood risk. The deviation from the modelled risk (under-/over-estimation) depends differently on the assumptions of spatial dependence, tail dependence, flood protection level and spatial scales. For example, under CD assumption for 200-year return period and considering flood protection, approximately 2.5-, 3- and 3.5-fold overestimation of flood risk in England, Germany and Europe, respectively, is found.