Bad weather is an important factor in many ship accidents and a proper understanding of the wave climate is of paramount importance for maritime safety. Extreme sea states need to be taken adequately into account in design and operation of ships and other marine structures and would be an essential part of maritime risk management and ocean and coastal engineering.
This paper describes a spatio-temporal stochastic model for modeling monthly maximum significant wave height in space and time. The model consists of different components describing spatial variability and dependence structure, space-time dynamics, seasonal effects and long-term temporal trends, respectively, and has a hierarchical structure. A Bayesian approach has been applied for inference on the model parameters, utilizing prior knowledge where adequate, and MCMC techniques (Gibbs sampler with Metropolis-Hastings steps) have been employed in the implementation. The model has been applied to significant wave height data for a rather large area in the North Atlantic ocean over a period of more than 44 years. Various model alternatives are tried out with different alternatives for the long-term trend component.
The paper outlines the model with its main structure and various components and presents the main results, which suggest that the model performs rather well. The contributions from the various model components are discussed with particular focus on the temporal trend parts. In general, the various model alternatives agrees reasonable well and expected increasing trends in the order of about 70 cm are estimated over the period. Extrapolated into the future, such trends, if they were to persist, would correspond to an increase of about 1.5 meters over a century. Overall results are compared to previous studies reported in the literature and are found to agree reasonably well.
Possible extensions of the model are briefly discussed, e.g. including regression terms relating the wave climate to emission scenarios and meteorological parameters. It is believed that such components are needed for future projections, and this will be addressed in future work.