Discriminating Developing versus Nondeveloping Tropical Disturbances in the Western North Pacific through Decision Tree Analysis

Abstract

his study investigates the classification of developing and nondeveloping tropical disturbances in the western North Pacific (WNP) through the C4.5 algorithm. A decision tree is built based on this algorithm and can be used as a tool to predict future tropical cyclone (TC) genesis events. The results show that the maximum 800-hPa relative vorticity, SST, precipitation rate, divergence averaged between 1000- and 500-hPa levels, and 300-hPa air temperature anomaly are the five most important variables for separating the developing and nondeveloping tropical disturbances. This algorithm also unravels the thresholds of the five variables (i.e., 4.2 ? 10?5 s?1 for maximum 800-hPa relative vorticity, 28.2°C for SST, 0.1 mm h?1 for precipitation rate, ?0.7 ? 10?6 s?1 for vertically averaged convergence, and 0.5°C for 300-hPa air temperature anomaly). Six rules are derived from the decision tree. The classification accuracy of this decision tree is 81.7% for the 2004?10 cases. The hindcast accuracy for the 2011?13 dataset is 84.6%.