NON-STATIONARITIES IN TIME SERIES AND ITS INTEGRATION IN EXTREME VALUE STATISTICS FOR RISK MANAGEMENT ISSUES

In this paper a non-stationary extreme value analysis approach is introduced, based upon the Generalized Extreme Value distribution (GEV) and the L-moment parameter estimation. It is shown, that for deriving present and future design water levels a parametric approach for extrapolation of the time-dependent distribution parameters is required. The results are compared to common stationary methods. In addition some definitions and methods for testing stationarity or non-stationarity are mentioned. All discussed methods are applied to the annual maximum time series from 1849 to 2007 at the German North Sea gauge Cuxhaven,