Uncertainties in extreme surge level estimates from observational records

Ensemble simulations with a total length of 7540 years are generated with a climate model, and coupled to a simple surge model to transform the wind field over the North Sea to the skew surge level at Delfzijl (NL). The 65 constructed surge records, each with a record length of 116 years, are analyzed with the Generalized Extreme Value (GEV) and the Generalized Pareto distribution (GPD) to study both the model and sample uncertainty in surge level estimates with a return period of 104 years, as derived from 116-year records. The optimal choice of the threshold, needed for an unbiased GPD-estimate from Peak-Over-Threshold (POT) values, cannot be determined objectively from a hundred-year data set. This fact, in combination with the sensitivity of the GPD-estimate to the threshold, and its tendency towards too low estimates, leaves the application of the GEV distribution to storm-season maxima as the best approach. If the GPD-analysis is applied, the exceedance rate $lambda$ should be chosen not larger than 4. % The climate model hints at the existence of a second population of very intense storms. As the existence of such a second population can never be excluded from a hundred-year record, the estimated 104-year wind speed from such records has always to be interpreted as a lower limit.