Calibration of a Clustering Algorithm to Improve the Nearshore Wave Prediction System

Accidents in the proximity of ports are rising due to the ongoing expansion of marine trade. Various factors can lead to them, including weather conditions. The risk of marine accidents can be reduced with accurate wave climate forecasts. Numerical models are typically used to analyse the weather forecast in port areas. However, these models have a high computational cost, rendering them unsuitable for forecasting and nowcasting wave conditions. Artificial Neural Networks (ANNs) can overcome this limit. However, the ANNs that had already been developed tended to concentrate on specific points of the study area, like piers and port entrances. Early-warning strategies would be improved with more knowledge of the wave climate over a large area. The primary objective of this work is to evaluate the clustering techniques that identify homogenous areas to build a novel approach for using ANNs to assess nearshore wave characteristics in actual situations. This paper uses the port of Augusta (SR), one of the most significant Italian ports, as a case study.