TETRA-VARIATE EXTREME VALUE STATISTICS MODELING OF STORM SURGE DEVIATION, WAVE HEIGHT AND THE SOURCE FACTORS

Abstract

 Sea level deviation due to storm surge increases in response to the blowing effect by wind stress, the suction effect by atmospheric pressure and the wave setup by wave breaking. These four factors are mutually dependent and by making the use of this relation as a linear statistical regression, the deviation from the astronomical sea level has been predicted in a simple method. We propose here to construct the sparse structure of the four factor's extreme value dependence by removing redundant paths as conditional independence. For example, though the wave height is inherently generated by wind blowing on sea surface, it will be better to be connected to the storm surge deviation without the link to the wind velocity, which depends on the case. Such pattern of the dependence structure will be one of the important regional properties of the local coast, and it is shown to be classified by the hierarchical relations of the partial correlations of the four and three factors.