Statistical Uncertainty in Long-Term Distributions of Significant Wave Height

Abstract This work examines some aspects involved in the estimation of the parameters of the probability distribution of significant wave height, in particular the homogeneity of the data sets and the statistical methods of fitting a distribution to data. More homogeneous data sets are organized by collecting the data on a monthly basis and by separating the simple sea states from the combined ones. A three-parameter Weibull distribution is fitted to the data. The parameters of the fitted distribution are estimated by the methods of maximum likelihood, of regression, and of the moments. The uncertainty involved in estimating the probability distribution with the three methods is compared with the one that results from using more homogeneous data sets, and it is concluded that the uncertainty involved in the fitting procedure can be more significant unless the method of moments is not considered.