The Dutch Water Defenses are designed and maintained according to hydraulic load estimates determined from extreme value analysis of wind and wave data from the North Sea. Commonly, this analysis involves selection of extremes via peaks-over-threshold which are then fit to a generalized Pareto distribution. This process assumes the data are independent and identically distributed (iid). The first can be achieved through declustering, but the second is more difficult to confirm and may be violated by seasonal variation and other temporal trends.
In this study, the importance of this violation will be investigated by applying a time-stratified (seasonal) model and a time-varying model (here a non-homogeneous Poisson process) for full-year return value estimates which can be compared to the results of current, non-seasonal methods. In addition, an effort will be made to estimate seasonal return values (eg. Separate summer and winter estimates), since the critical failure modes may be different in different seasons.
Info: MSc Student, CoMEM
