Seasonal prediction of tropical cyclone frequency over the East China Sea through a Bayesian poisson-regression method

In this study, a Bayesian method has been used to predict the seasonal number of tropical cyclones (TCs) over the East China Sea (25°N-35°N, 120°E-130°E). The method considers two periods each year and provides several different predictions of the number of TCs that will enter the target area during a typhoon season, assigning a probability value to indicate the likelihood of each prediction. The method was used to forecast the number of TCs that would occur over the extended season (June-September) issued by June 1 and over the peak season (July-September) issued by July 1. The three parameters of sea surface temperature (SST), outgoing longwave radiation (OLR), and 850-hPa vorticity (VOR850) were used as predictors, based on lag correlation coefficients with the TCs over the target region for 1979-2003. For the extended season prediction, the SST for February-April, the VOR850 for February 16-May 15, and the OLR for May 1-May 15 were chosen as predictors in the TC forecast system. The three predictors for the peak season prediction are delayed one month relative to those for the extended season prediction. The observed TCs over the target region mainly fall into the prediction range within 25% probability from the median during the period from 1979 to 2007. The method predicts the mean TC frequency with a high level of accuracy, yielding a correlation coefficient of 0.69 and 0.71 and a root mean square error of 1.48 and 1.20 between the predicted and observed TCs for the extended and peak season forecasts, respectively. This TC prediction method may soon be used for operational purposes.