Extreme wind statistics for the hydraulic boundary conditions for the Dutch primary water defences : SBW-belastingen: phase 2 of subproject "Wind Modelling"

Updated extreme potential wind statistics were computed for 21 KNMI wind stations with long time series available. The updated wind statistics are to be made available for the inference of the Hydraulic boundary conditions for the Dutch primary water defences by the WTI team. The period of data considerd in the analyses was the 1970-2008 period. Both omni-directional and directional estimated were obtained. The sensitivity of the results to different periods are also analysed and found to be reasonable l.e. the found differences do not exceed the uncertainty associated with the estimates. The data was analysed using the standard AM/GEV and POT/GPD approaches. The hypothesis of a type l tail for the potential wind data was extensively tested. Power 2 and power k, with k being the shape parameter of the Weibul fit to the whole dataset, data transformations were considerd in order to accelerate the convergence to a type l tail, nor do they seem to be needed. The assumption of a type I tail seems to be valid for the considered potential wind data. Furthermore, the estimates obtained from expotential fits to the POT data were found to be realistic and reliable and are given as final/best estimates. Mainly due to type l tail assumption, the curvature problem does not seem to strongly affect the computed estimates. Furthermore, these new estimates do not differ much from the currently used estimates of Wieringa and Rijkoort (1983). More preciselym the 10,000 yr return value estimates of this study differ by less than 3% from those of Wieringa and Rijkoort (1983) i8n 10 of the 13 stations considered by them. The two stations for which the differences are large – about 10% - the estimates were adjusted ‘by hand’by Wieringa and Rijkoort (1983) and discrepancies would even be larger if they had not been adjusted. A number of caveats apply to the given estimates: the confidence intervals given are underestimated and long term trends and inhomogeneities in the data were ignored.