Master thesis defense by Morten Hallas

Title: Storm surges of the West Coast.
A statistical model of storm surges on the Danish West Coast.

Abstract: Motivated by the mean sea level rise as described in IPCC [2013] and the declining trend in storm surges for Eastern England as described in Menéndez and Woodworth [2010], this paper aims to examine the distribution of storm surges on the West Coast of Denmark and test for non-stationary behavior. Only data from the stations facing the Wadden Sea has been acquired. The peak over threshold method is used, the peaks modelled as a Generalized Pareto distribution and the rate of occurrences as Poisson distribution. Through teleconnection analysis a connection is found with the 3^rd EOF of the SST over the Northern Hemisphere, which Osborn [2010] identifies as a proxy for the NAO, although this is questioned at the end of the paper.  A range of different models, stationary and non-stationary, and predictors, such as NAO and the Global Temperature Anomaly, are fitted with Markov Chain Monte Carlo simulations and their performance intercompared with Bayes factors. No evidence is found in favor of a non-stationary model for the Generalized Pareto distribution, but there is very strong evidence that the rate parameter of the Poisson distribution is dependent on the predictor identified in the teleconnection analysis. The coefficient of this dependency is found to be negative, and by extrapolation with the CMIP5 ensemble, the rate of storm surges is found to be declining for all of the four emission scenarios, RCP2.6, 4.5,6 and 8.5.