Cluster Analyses of Tropical Cyclones with Genesis in the South China Sea Based on K-Means Method

Tropical cyclone (TC) with genesis in the South China Sea (SCS) has been a major concern because of their high landfall frequency and associated serious hazards to the surrounding coastal areas. The classification of TCs from records of historical tracks is an important way to obtain their characteristics and to help predict their future behavior. According to the generation location, intensity, direction, and track length of TC, TCs with genesis in the SCS from 1950 to 2020 are classified into four clusters by the K-means clustering method, including northwestward track cluster A, westward track cluster C and two long northeastward track clusters B and D. The landfall probability, peak season, climate trend, lifespan, maximum wind speed, and power dissipation index show a significant distinction for each cluster. All clusters had a landfall probability exceeding 50%, with the highest probability in cluster A (90.44%), followed by cluster C, cluster B, and cluster D with the lowest probability (54.55%). The clustering results indicate that tracks of TCs are strongly affected by the distribution pattern of the Western Pacific Subtropical High. When the WPSH moves southward, the southwesterly anomalies provide a significantly favorable steering flow for TC northeastward. Conversely, the WPSH located northward in July-September, the strong southeasterly anomaly favoring the northwestward movement of TC. From October to November, the WPSH shrinking in size gives way to the prevailing anomalous easterlies that steer the TCs westward. Further concerning the influence of TCs in the different clusters by the WPSH movement will be helpful for prediction in terms of the occurrence, track and landfall probability of TCs in the SCS.