Storm surges present a substantial risk in marine cargo insurance. This thesis endeavors to model and understand the features of storm surges using extreme value theory. In essence, the research delves into both univariate and multivariate extreme value theory, highlighting their relevance in predicting and understanding real-world storm surge events. The initial phase focuses on univariate analysis about the generalized extreme value distribution, utilizing statistical techniques such as Block Maxima, Peaks Over Threshold, and various semi-parametric estimators for model estimation. The scope then broadens to multivariate analysis, introducing concepts like multivariate maxima, extreme value copulas, and both parametric and nonparametric estimation methods. These serve as tools to analyze dependencies within extreme events. The next section introduces a practical layer via a case study on storm surges, analyzing water levels at measuring stations in both Hamburg and Bremerhaven. This analysis determines return levels for storm surge events and evaluates dependencies between the two locations using bivariate extreme techniques, including bivariate POT and copulas. Time-dependent effects are also considered in this study. In the end, the research uses a parametric insurance strategy, leveraging a sigmoid loss function, to estimate potential insurance losses based on the derived water level models.     «

Storm surges present a substantial risk in marine cargo insurance. This thesis endeavors to model and understand the features of storm surges using extreme value theory. In essence, the research delves into both univariate and multivariate extreme value theory, highlighting their relevance in predicting and understanding real-world storm surge events. The initial phase focuses on univariate analysis about the generalized extreme value distribution, utilizing statistical techniques such as Block...     »