A Statistical Model for Prediction of the Tropical Cyclone Activity over the Western North Pacific

A statistical model for the prediction of tropical cyclone activity has been developed and applied to the western North Pacific. It aims to predict the number of tropical cyclones and the overall tropical cyclone activity index over a certain period. The model is based upon regression analysis. The predictors for the regression model are obtained by examining the lag correlations between the predictands and the synoptic variables such as the sea surface temperature, the 850 hPa temperature, the 850 hPa meridional wind, and the 500 hPa geopotential. Even though the current model is a statistical model based on the regression analysis, there is a unique difference from the conventional regression model. The latter is simply the best fit of the past data, while in the former we select the combination of predictors that yielded the best prediction performance in the past among all possible combinations out of all potential predictors. We call them smart predictors. Predictands in this study are the number of total tropical cyclones over the whole year, over the three summer months (June-August), or the fall (September-November) period. It is found that the performance of the model has been successful in the seasonal prediction of the number of tropical cyclones over the western North Pacific. Especially the predicted results for the tropical cyclones that will affect Korea are delivered to the Korea Meteorological Administration and thereby serve as guidance when producing the official forecast for the tropical cyclones in the three-month seasonal outlook.