Study of tropical cyclone wave characteristics based on a hybrid track clustering method

Abstract

Understanding the activity and wave characteristics of different TC track types with different formation mechanisms is crucial for coastal and ocean engineering. In this study, a hybrid clustering method based on track indicators representing track shape and influence region is proposed to classify 721 TC tracks along the coast of China into three track types. Due to the intrinsic similarity of track indicators diagnosed by the self-organizing map method, the initial clustering centroid selection of K-means method is reasonable. 5000 clustering trials show that the clustering results are more robust than those of the traditional K-means method, and the average difference percentage decreases from 11.3% to 5.5%. By analyzing the robust clustering results, the activity characteristics of three TC types under the background of atmospheric circulation patterns and the characteristics of the waves generated by the three TC types are studied, especially the characteristics of extreme waves. The joint distribution pattern of extreme wave height and wave direction indicates the direction sector shift and frequency change caused by different TC types. These results provide important insights for risk analysis and resilience construction of coastal and ocean engineering structures.