Compared with total account of basin wide tropical cyclones (TC) genesis, the prevailing tracks of TC activity and its potential of landfall is more important for disaster prevention. Despite its relatively lower predictability, a hybrid statistical-dynamical model was developed based on the relationship between leading empirical orthogonal function (EOF) modes of western North Pacific (WNP) TC track patterns and large scale environmental fields. Due to the significant difference in both variability and associated mechanisms on different time scale as revealed in previous studies, the temporal variations of three leading principal components (PCs) are separated into decadal and inter-annual component and forecasted respectively. Potential predictors for multi-linear regression (MLR) model was selected through stepwise regression analysis based on the correlation maps between each components of leading PCs and environmental fields in both observations and Beijing Climate Center climate system model version 1.1 (BCC_CSM1.1) hindcast. The forecast map of anomalous WNP TC track density was obtained through weighted composite of forecasted leading PCs and EOF modes according to its explained variance. One-year-out cross validation test shows that the hybrid model well captures the inter-annual variations of WNP TC track patterns. The hybrid model also shows significant improvement of prediction skill compared with ENSO reference forecast, indicates its potential of application in operational prediction.

Tropical cyclones,Track density,Western North Pacific,Seasonal prediction,ENSO