Global Stochastic Tropical Cyclone Model Based on Principal Component Analysis and Cluster Analysis

Abstract

global stochastic tropical cyclone model was developed as a means for preparing a large number of artificial tropical cyclone (TC) samples with different values for parameters such as track, minimum sea level pressure, and translation speed. In this paper, the model and the results of its verification are presented in detail. The proposed stochastic model is sensitive to approximations of the joint probability distribution functions (PDFs) of TC parameters and temporal correlations. A newly introduced accurate method for approximating joint PDFs by using principal component analysis and cluster analysis resulted in improved reproducibility of TC parameters. The simulation results were compared with historical observational data from the northwestern Pacific, southwestern Pacific, and North Atlantic Oceans. The grid-averaged mean values and distribution patterns of PDFs of TC parameters were in agreement with observational data.