Regional Arctic Sea Ice Prediction Skill and Predictability in Dynamical Forecast Systems

Research over the past decade has demonstrated that dynamical forecast systems can skillfully predict Pan-Arctic sea ice extent (SIE) on the seasonal timescale, however there have been fewer assessments of prediction skill on user-relevant spatial scales. In this work, we evaluate regional Arctic SIE predictions made with two dynamical seasonal forecast systems developed at the Geophysical Fluid Dynamics Laboratory. We consider suites of retrospective initialized seasonal predictions performed with the FLOR and recently-developed SPEAR_MED dynamical models. Compared to FLOR, we find that SPEAR_MED generally displays improved skill in predicting regional detrended SIE anomalies, partially owing to improvements in sea ice initial conditions. In both systems, we find skillful predictions of detrended winter SIE up to 11 months in advance associated with persistence of upper ocean heat content anomalies. In the Labrador Sea and the Sea of Okhotsk, SIE prediction skill is attributable to the skillful prediction of large-scale modes of sea-surface temperature variability associated with the North Atlantic Oscillation (NAO) and North Pacific Gyre Oscillation (NPGO), respectively. Confirming earlier work, we find that SIE persistence provides the key source of summer sea ice predictability at short lead times (0-1 months), whereas sea ice thickness persistence provides the key source of predictability at longer lead times of 2-3 months. We also identify a third predictability regime, relevant in the Chukchi and Kara seas, characterized by a trade-off between ocean-based predictability and thickness-based predictability that occurs as the sea ice edge position evolves seasonally.