Assessment of K-Means Algorithm to Evaluate Nearshore Wave Climate

Accidents near ports have increased due to the ongoing expansion of maritime trade. These accidents have various causes, including adverse weather conditions. Accurate wave climate forecasts can help mitigate the risks associated with marine accidents. While numerical models are commonly used for obtaining nearshore wave climate forecasts, their high computational cost makes them impractical for wave climate forecasting and nowcasting. Artificial neural networks (ANNs) offer a potential solution to this limitation. However, existing ANNs have primarily focused on specific single points within the study areas, such as piers and port entrances. Enhancing early-warning strategies requires a broader understanding of the wave climate across larger areas. Thorough examinations of extensive areas with varying physical attributes can result in significant computational time requirements. The main objective of this study is to evaluate a clustering technique able to identify homogeneous areas to improve future applications of ANNs to assess nearshore wave characteristics in actual situations. The area around the port of Augusta (Sicily), one of the most important ports in Italy, serves as a case study in this article. Results show an optimal performance by applying the clustering algorithm K-means, capable of capturing the wave climate characteristics of the study area.