Sampling joint time series of significant wave heights and periods in the North Sea

Stochastic simulation and description of extreme hydraulic variables is essential for coastal risk assessment. Existing copula-based approaches are restricted to modeling the multivariate distribution of the maxima of a few of such variables during extreme events. In this article we propose a dependence tree to simulate joint time series of hourly data of significant wave heights and corresponding periods. The time series can represent both daily and stormy conditions. The tree connects a bivariate copula between wave heights and periods with one between wave heights of subsequent time steps, and can be extended to arbitrary many time steps. Both copulas belong to the skew-t family, which is very flexible due its four parameters and can capture tail dependence and asymmetry patterns. The storm wave scenarios could be used as input for models of subsequent processes in the flood risk chain, such as levee failure and inundation. As a next step, we suggest an extension of the tree to a vine to further improve accuracy of the artificial time series.