2018 Volume 14 Pages 138-143
This study investigates prediction of TC intensity in the western North Pacific basin using a statistical-dynamical model called the Statistical Hurricane Intensity Prediction Scheme (SHIPS), with data sources in operations at the Japan Meteorological Agency (JMA) such as the JMA/Global Spectral Model forecast fields. In addition to predicting the change in the maximum wind (Vmax) as in the original SHIPS technique, another version of SHIPS for predicting the change in the minimum sea-level pressure (Pmin) has been developed. With 13 years of training samples, a total of 26 predictors were selected from among 52 through stepwise regression. Based on three years of independent samples, the root mean square errors of both Vmax and Pmin by the 26-predictor SHIPS model were found to be much smaller than those of the JMA/GSM and a simple climatology and persistence intensity model, which JMA official intensity forecasts are currently mainly based on. The prediction accuracy was not sensitive to the number of predictors as long as the leading predictors were included. Benefits of operationalizing SHIPS include a reduction in the errors of the JMA official intensity forecasts and an extension of their forecast length beyond the current 3 days (e.g., 5 days).