Multiyear hybrid prediction of Atlantic tropical cyclone activity and the predictability sources

A hybrid statistical–dynamical model is developed to predict multiyear variability of Atlantic tropical cyclone (TC) activity. A Poisson model takes sea surface temperature (SST) averaged over the Atlantic main development region (MDR) and the Atlantic subpolar gyre region (SPG) from the initialized CESM prediction as predictors, and skillfully predicts the basinwide TC frequency, accumulated cyclone energy (ACE), landfalling TC frequency, and hurricane and major hurricane days. Further analysis shows that the SPG SST is a more important source of predictability than the MDR SST for multiyear Atlantic TC activity. The comparison between the uninitialized and initialized CESM predictions suggests that the SPG SST is better predicted by the initialized CESM owing to the better prediction of Atlantic meridional overturning circulation, which contributes to the overall more skillful TC predictions. On the other hand, the skillful prediction of the basinwide TC frequency by the uninitialized CESM suggests the role of external forcing in the variability of Atlantic TC activity. The dependence of the hybrid prediction skills on the dynamic model ensemble size is also explored, and an ensemble size of;20 is suggested as optimal. Further analysis shows that the SPG SST is associated with the variability of vertical wind shear and precipitable water over the tropical Atlantic even when the influence of the MDR SST is controlled. The spatial patterns of vertical wind shear and precipitable water suggest a strong modulation of ACE and hurricane frequency but a relatively weak influence on the basinwide TC frequency. The physical mechanisms between the SPG SST and Atlantic TC activity are discussed.