Probabilistic modelling of significant wave height using the extended generalized inverse Gaussian distribution

Abstract

It is common practice to represent wave climate and sea severity through significant wave height. The development of accurate probabilistic models of wave height is paramount in ocean-related activities (e.g., offshore structures, weather routing, wave energy assessment studies). In this work, several models for the probabilistic description of the significant wave height are examined and evaluated along with a new distribution that is introduced for the first time in wave height modelling, namely the Extended Generalized Inverse Gaussian model. The corresponding density and the cumulative distribution function are expressed through closed forms. All the examined models were applied on long-term measured wave data at four different locations of the Greek Seas and wave height data obtained from four buoys of the US NOAA National Data Buoy Centre. The coefficient of determination, the Bayesian information criterion and the average deviation criteria were used for the evaluation of the examined models’ performance. The final selection is based on the multiplicative criterion that combines the above criteria into a single expression. The proposed distribution outperforms the examined parametric models and describes in an optimal way the probabilistic structure of significant wave height.