Extreme wave loads

This paper compares three main methods for estimating extreme wave loads with a view towards determining the sensitivity of estimates to the particular approach chosen. The approaches considered include: a) The generally used ad-hoc procedure of performing an extreme value analysis of the H_s data, trying to find a relationship between wave height and period at the storm peaks and then, once the return values of extreme wave heights are estimated, estimating the associated return value of the wave period by means of the relationship found, b) The 'structure variable method' in which the pairs wave height and period observations are converted into univariate loads to which univariate extreme value theory is applied to estimate the return value of the structural load, c) The multivariate extreme value approach suggested by [1] in which a 'multivariate return value', namely the most probable value of the wave period conditional on a return value of the wave height, is estimated. Our study is based on a 44-yr long timeseries of wave conditions created using the shallow water wave model SWAN and calibrated ERA-40 data. The results suggest that the three approaches yield similar estimates. However, the ad-hoc procedure a gives the least conservative estimates. Approach c provides results that apply to any choice of load function and which to a certain extent are independent of the location in which the estimates are obtained, for which reason it may generally be the preferred one.