• Annual maximum water levels;
  • coastal structures;
  • extreme values;
  • generalized extreme value distribution;
  • nonstationarities;
  • North Sea;
  • parameter estimation;
  • time series


In this paper, a nonstationary extreme value analysis approach is introduced in order to determine coastal design water levels for future time horizons. The nonstationary statistical approach is based on the generalized extreme value (GEV) distribution and an L-moment parameter estimation. It is shown that to derive future design water levels, a parametric approach is required for the extrapolation of the time-dependent parameters. The results are compared with common stationary methods. In addition, some definitions and methods for testing on stationarity or nonstationarity are discussed. All the methods introduced are applied to the annual maximum water levels from 1849 to 2007 at the German North Sea gauge at Cuxhaven. The nonstationary results for the year 2007 are in the range of NN+507 cm (NN: German reference datum), and for the time horizon, 2100 the results vary from NN+542 to NN+546 cm for the 100-year event. The results show that the nonstationary GEV approach is suitable for determining coastal design water levels.