Abstract Reliable seasonal predictions of Atlantic hurricanes heavily rely on the predictions of sea surface temperature anomalies (SSTA) in the main development region (SSTA MDR ) and its relative changes against the SSTA over the whole tropical region (RSST). By analyzing the prediction skill of Community Earth System Model ensemble mean hindcasts, we identified significant multidecadal variations in the seasonal prediction skill of both SSTA MDR and RSST during hurricane seasons. Both RSST and SSTA MDR are more predictable in the beginning and ending periods of the 20th century and less predictable in between. The multidecadal variations of the seasonal prediction skill of RSST are dominated by that of SSTA MDR , which can be explained mostly by its variability changes in July. Further, mixed layer heat budget analyses based on SODA subsurface oceanic temperature and NOAA‐20C reanalysis data show that the multidecadal variations of the seasonal prediction skill of SSTA MDR and its variability changes in July are closely linked to the mixed layer depth fluctuations, which is caused by turbulent heat flux changes induced by surface wind anomalies.
Yan et al. (Wed,) studied this question.