CLASSIFYING WESTERN NORTH PACIFIC TROPICAL CYCLONES BY PHYSICAL INDEX SYSTEM  

作　　者：YAN Dong-yi XU Kui MA Chao MA Man-tang 颜东谊;徐奎;马超;马满仓(State Key Laboratory of Hydraulics Engineering Simulation and Safety, Tianjin University)

机构地区：State Key Laboratory of Hydraulics Engineering Simulation and Safety,Tianjin University,Tianjin 300072 China

出　　处：《Journal of Tropical Meteorology》2018年第2期142-150,共9页热带气象学报（英文版）

基　　金：National Key Research and Development Program of China(2016YFC0401903);National Natural Science Foundation of China(51722906,51679159,51509179);Tianjin Research Program of Application Foundation and Advanced Technology(15JCYBTC21800)

摘　　要：The classification of tropical cyclones(TCs) is significant to obtaining their temporal and spatial variation characteristics in the context of dramatic-changing global climate. A new TCs clustering method by using K-means clustering algorithm with nine physical indexes is proposed in the paper. Each TC is quantified into an 11-dimensional vector concerning trajectory attributes, time attributes and power attributes. Two recurving clusters(cluster A and E)and three straight-moving clusters(cluster B, C and D) are categorized from the TC best-track dataset of the western North Pacific(WNP) over the period of 1949-2013, and TCs' properties have been analyzed and compared in different aspects. The calculation results of coefficient variation(CV) and Nash-Sutcliffe efficiency(NSE) reveal a high level of intra-cluster cohesiveness and inter-cluster divergence, which means that the physical index system could serve as a feasible method of TCs classification. The clusters are then analyzed in terms of trajectory, lifespan, seasonality, trend,intensity and Power Dissipation Index(PDI). The five classified clusters show distinct features in TCs' temporal and spatial development discipline. Moreover, each cluster has its individual motion pattern, variation trend, influence region and impact degree.The classification of tropical cyclones（TCs） is significant to obtaining their temporal and spatial variation characteristics in the context of dramatic-changing global climate. A new TCs clustering method by using K-means clustering algorithm with nine physical indexes is proposed in the paper. Each TC is quantified into an 11-dimensional vector concerning trajectory attributes, time attributes and power attributes. Two recurving clusters（cluster A and E）and three straight-moving clusters（cluster B, C and D） are categorized from the TC best-track dataset of the western North Pacific（WNP） over the period of 1949-2013, and TCs＇ properties have been analyzed and compared in different aspects. The calculation results of coefficient variation（CV） and Nash-Sutcliffe efficiency（NSE） reveal a high level of intra-cluster cohesiveness and inter-cluster divergence, which means that the physical index system could serve as a feasible method of TCs classification. The clusters are then analyzed in terms of trajectory, lifespan, seasonality, trend,intensity and Power Dissipation Index（PDI）. The five classified clusters show distinct features in TCs＇ temporal and spatial development discipline. Moreover, each cluster has its individual motion pattern, variation trend, influence region and impact degree.

关 键 词：tropical cyclone physical index K-means clustering Nash-Sutcliffe efficiency inter-cluster divergence intra-cluster cohesiveness power dissipation index 

分 类 号：P444[天文地球—大气科学及气象学]