Prediction of extreme significant wave heights using maximum entropy

This paper presents maximum entropy (MaxEnt) as a powerful tool for the prediction of extreme significant wave heights comparing it with models within the extreme value theory (EVT) framework, i.e. the Generalized Pareto distribution (GPD) and the Generalized Extreme Value distribution (GEV). The MaxEnt method produces the least biased probability density function by maximizing the Shannon's form of entropy, which is applied to samples generated using the peak over threshold approach. The aforementioned methods have been tested on a data set from the northern North Sea. The results showed that return levels associated to high return periods obtained with MaxEnt are very stable through the whole threshold range, implying that MaxEnt is more robust to the variation of threshold, i.e. sample size than the GPD, leading to the conclusion that MaxEnt is the more suitable model than the GPD model, for this particular data set.

► This paper presents maximum entropy for the prediction of extreme significant wave heights. ► Generalized Pareto distribution (GPD) and Generalized Extreme Value (GEV) models are also used. ► The methods were tested on a data set from northern North Sea. ► The POT approach is adopted. ► The results showed that high return periods obtained with MaxEnt are more stable than with other models.