Correlation models in flood risk analysis

Statistical dependence among random hydraulic variables generally stem from a common meteorological cause. Usually, this dependence increases the probability of occurrence of floods. Therefore, in many (potential) applications in flood risk analysis there is a need for techniques that properly describe the statistical dependence among random variables. This study makes an inventory of existing bivariate correlation models as available from international literature and/or ‘Dutch practice’ in the design and testing of flood defence structures. Differences in correlation structures are described and quantified. Subsequently, their practical applicability is tested in two case studies. of this ‘Dutch practice’ in testing flood defences is given. Without adequate flood defence structures, almost 60% of the Netherlands would be regularly flooded. To prevent floods by the sea, the large rivers and lakes, an extensive system of dikes and dunes has been constructed in the past. In the design and testing of dikes and dunes a safety level is defined of the order of 10–3 to 10–4 “failures” per year. This frequency is based on both the economic value of the protected area and the extent of the threat. To test whether the flood defences fulfil the required safety level, representative hydraulic conditions are derived. For this purpose a number of probabilistic models have been developed over the past decades. Different models have been developed for different water systems, such as Hydra-B for the tidal area of the Rhine-Meuse delta, Hydra-K for the coastal areas and Hydra-VIJ for the deltas of the rivers Vecht and IJssel. Furthermore, the model PC-Ring has been developed, which considers entire ring dikes and therefore integrates the various water systems. Currently, PC-Ring still has the status of “research tool”, whereas the Hydramodels are “officially approved” models. 3 MATHeMATICAL DeSCRIPTIoN oF BIVARIATe CoRReLATIoN MoDeLS