A Study On Tracks Clustering And The Variation Of Tropical Cyclones In The Western North Pacific

The Tropical Cyclones(TC) over the western North Pacific(WNP) are among the most destructive weather systems as they impose vast disasters to China each year. The research of their tracks has always been a key topic in the meteorological field, but the detailed and objective classification, the mechanism and characteristics of tracks need further study. Applying data mining technology, TC track and climate circulation data set, it can not only find the rules of TC movements in different climatic conditions, but also improve the defense ability of sudden weather and reduce the loss caused by the disaster. Based on the TC best-track data during summer and autumn (June-November)for the period 1951 -2015 from the STI of CMA, 6-hourly and monthly mean reanalysis data of wind and other fields from NCEP/NCAR, integrating the monthly meanreanalysis SST data from HadISST data set, we analyze the impacts of different ENSO phases on TC activities and circulation situations by applying FMM algorithm and SEOF. The research results are described in the following three aspects:(1) Using FMM algorithm, tracks of TC in WNP are classified into seven classes,and tracks of TC landing China are clustered into three classes. The results show that the clustering algorithm is relatively objective to the TC track.(2) Based on the change of TC tracks in different phases of ENSO, it is verified that the ENSO phases have different effects on the TC generation and intensity in WNP,which in turn cause the change of tracks. Cluster-6 (offshore recurving cluster) is the main track in El Nino developing years and TCs are active in the southeastern quadrant of the WNP. Cluster-2 (South China Sea westward cluster) and Cluster-7 (Japan Sea recurving cluster) are the two major tracks in El Nino decaying. In La Nina developing years, TCs are active over the western part of the WNP and Cluster-7 (Japan Sea recurving cluster) owns the highest proportion. In La Nina decaying year, TCs are active over the southwest quadrant of the WNP and Cluster-3 (coastal recurving cluster)is the mainly active track.(3) The TC tracks landing China are clustered into three categories by using FMM algorithm, and that correspond with the traditional classification result. Cluster-1 approximates landing north, Cluster-2 approximates landing westward and Cluster-3 approximates landing northwest. In summer, Cluster-1 and Cluster-2 appear more frequently than Cluster-3, but in autumn, the highest frequency of occurrence is Cluster-3. The west ridge point of the subtropical ridge line' s location of Cluster-1 is the most northward, and the convergence in genesis and disappearance is the strongest,while Cluster-2's point locates the most westward, and the convergence is the strongest in disappearance. The location of Cluster-3's west ridge point is the most eastward and southward and this cluster owns the strongest vertical zonal wind shear when occurring.The above work and conclusions not only enrich the research of tropical cyclonetracks, but also provide reference for climate diagnosis and new ideas for scientific research in the field of data mining and meteorology.