This paper describes the development of the APCC multi-model ensemble (MME)-based dynamical- statistical model for predicting tropical cyclone (TC) activity during the western North Pacific (WNP) TC season. The model is based on the statistical relationship between the interannual variability of the number of the WNP TCs during the active TCs season (July-October; JASO) and the APCC MME predicted variability of the key predictors for this season. Note that one or two key predictors are selected for each predict and (i.e., tropical storms, typhoons, and intense typhoons) by forward selection. This model provides a combination of statistical and dynamical methodologies, thus realizing better skill than the statistical as well as dynamical method. Further, MME prediction, which is used to derive the predictors for the statistical model, is more skillful in the seasonal forecasts than other individual models considered.

The leave-one-out cross validation results of the APCC MME-based retrospective dynamical- statistical model demonstrate its high skill in predicting TC number when using APCC MME-based predictors; the correlation coefficients range from 0.51 to 0.68. A comparison between the observed number of TCs and the APCC MME real-time TC forecasts also demonstrates that the APCC MME seasonal prediction has high skill in forecasting the seasonal TCs in the WNP during JASO. The information derived from the APCC MME-based model for predicting the TC number may be valuable, in particular to the people residing in the tropical cyclone-prone areas in the Asia-Pacific region.

1. INTRODUCTION

2. DATA AND METHODS

2.1. Observation and Reanalysis

2.2. Hybrid dynamical-statistical prediction method

3. RESULTS AND VALIDATION

3.1 APCC MME predictive skill

3.2 Variability of WNP TCs and its relation to potential predictors

3.3 APCC MME-based statistical prediction of seasonal TC activity

4. SUMMARY AND DISCUSSION

4.1 Summary

4.2 Discussion