Significant Wave Height Cluster Analysis to Understand the Spatial Variation of Ocean Waves in Low Energy Systems

Abstract Understanding spatial variation in ocean waves is critical for erosion planning and infrastructure projects. The study is aimed at (1) performing a cluster analysis to categorize the wave statistics over space and (2) determining the important drivers affecting spatial variations of wave statistics in a low energy, fetch limited environment. In this study, 29 wave gauges were deployed in Back Bay Biloxi, Mississippi. Raw pressure and processed wave height were clustered using two algorithms: Euclidian and Dynamic Time Warping. The Euclidean algorithm was applied to raw and processed data. However, due to the computationally expensive nature of Dynamic Time Warping, this algorithm could not be used on raw pressure data and was only applied to processed wave data. Therefore, three combinations of distance algorithms and data were compared: (1) Euclidean algorithm on raw pressure data, (2) Euclidean algorithm on processed wave height data, and (3) Dynamic Time Warping algorithm on processed wave height data. The results showed similar clustering for Euclidean and Dynamic Time Warping on processed data, with most gauges falling into one cluster. Results from this study reveal that the dendrogram trees of the Euclidean and Dynamic Time Warping algorithms on processed data are similar, where most of the wave gauges fall in one cluster. Conversely, the Euclidian algorithm on the raw pressure data distributed the wave gauges more evenly between the clusters. Additionally, the Euclidean algorithm on the raw pressure data showed that water depth significantly affects wave clustering.