Anticipating Arctic Sea ice variability on multiyear timescales is critical for near-term climate prediction and risk assessment. Here we assess winter Atlantic-sector Arctic Sea ice predictability using both perfect-model framework and initialized decadal hindcasts. Combining average predictability time analysis with machine learning, we identify the dominant predictable modes and their physical drivers. Externally forced anthropogenic warming dominates predictability beyond a decade, where internal variability associated with the Atlantic Meridional Overturning Circulation (AMOC) controls shorter timescales. A mature AMOC-related mode, characterized by broad-wide negative sea ice anomalies, is predictable up to four years. In contrast, a transitional AMOC mode exhibiting a dipole sea ice pattern retains skill for approximately two years. Independent machine learning predictabilities corroborate these results, underscoring the key role of slowly evolving ocean circulation, particularly AMOC variability, in shaping multiyear Arctic Sea ice predictability and near-term climate forecasts. Distinct Arctic sea-ice modes linked to the AMOC remain predictable for two to four years, while anthropogenic warming dominates longer-term predictability, according to a study combining average predictability time analysis with machine learning and decadal hindcasts
Li et al. (Tue,) studied this question.