Modelling extreme wind speeds at a German weather station as basic input for a subsequent risk analysis for high-speed trains

For high-speed trains there is a potential risk of derailment when driving very fast and being hit by an extraordinary strong gust at the same time. The risk depends on both the wind speed and the angle between train and gust. Several techniques have been established to minimize this risk to acceptable values. To decide which of these techniques at a given site is most appropriate, precise knowledge of the wind process at extreme levels is necessary. Therefore, methods adapted to the special requirements of the application are needed. We discuss directional modeling using an approach proposed by Coles and Walshaw (Appl. Statist. 43 (1994) 139). We focus on estimating extreme quantiles and their confidence intervals. Different types of confidence intervals are compared and we show how they can be used for probabilistic assessment of exceedance. We also consider the problem of masking by comparing daily maxima with those of a shorter time interval.