Untersuchungen zur Ermittlung von hydrologischen Bemessungsgrößen mit Verfahren der instationären Extremwertstatistik : Methoden und Anwendungen auf Pegelwasserstände an der Deutschen Nord- und Ostseeküste

Many coastal regions are faced with the global climate change, which has effects on sea level rise and partially on increasing storm surge frequency and heights.
The design of coastal defence structures is mainly based upon water levels to which a certain occurrence probability (return period) is assigned. Nowadays, design levels with a return period between 100 and 10,000 years are common. The determination of the occurrence probabilities is done by an analysis of observed water level data based upon extreme value statistics. Thereby, not only statements of the actual occurrence probabilities are required, but also for future time horizons. The common statistical methods postulate a stationary behaviour (no changes through time) of the water level data, which needs to be analysed. However, in the context of climate change, it is possible, that this assumption would no longer hold and the results become doubtful. If a significant non-stationary behaviour of the water level data exists, the common statistical stationary methods may not be used any more. Either the non-stationarities have to be removed or the non-stationarities need to be integrated directly into the extreme value analysis. The present study introduces a method, where trends in the observed data can be integrated directly into the extreme value analysis, in order to determine future design water levels.
The projection of future design water levels can be realised by using a non-stationary approach of the Generalized Extreme Value (GEV) distribution. Thereby, a parametric approach for the parameter estimation have been applied in order to extrapolate the results to future time horizons. The parameter estimation has been realised by L-moments, based on time windows. The non-parametric processes of the L-moments have been described by functional relations in order to extrapolate them. This is called a parametric approach. By use of the introduced method it is possible to specify a T-year design level at any point of time.
Applications of the non-stationary method to the gauges Cuxhaven (North Sea) and Travemünde (Baltic Sea) show, that the location parameters of the analysed water level time series have a significant increasing trend. For both gauges the 100-year design levels have been calculated up to the time horizon 2100 by using the introduced non-stationary extreme value approach. Additionally, effects of projected climate scenarios on future design water levels have been analysed.
On basis of the presented results, the non-stationary extreme value approach may be recommended to determine future design water levels.