Data driven analysis on the extreme wave statistics over an area

In this paper we analyse ocean wave crest statistics over different sized areas using data driven methods. We use second order numerical simulations to generate extreme crest data. We consider a simplistic Gumbel distribution fit as well as using a Random Forest Model to map the sea-state parameters to extreme crest values. Our simulations are compared with the existing distributions in the literature. We find that existing distributions perform well for more straightforward cases but that as more parameters are introduced the data science approach can capture features other methods cannot. Our approach also highlights the importance of different parameters such as steepness or length in the mean wave direction. We conclude that machine learning model is promising approach to predicting wave crest distributions in complex scenarios.