Item

Abstract

With the opening of Arctic passage for maritime transportation under global warming, more accurate prediction of Arctic sea ice on subseasonal-to-seasonal (S2S) time scales becomes crucial for both economy and society but challenging. This study examined the S2S hindcast skill of Arctic sea ice during 1992-2019 using Flexible Global Ocean-Atmosphere-Land System, Finite-Volume version 2 (FGOALS-f2), a global coupled model including an interactive dynamical sea ice component. First, the prediction system can well capture the seasonal cycle with minimum in September and maximum in March and the long-term trend of annual range for Arctic sea ice extent (SIE). Second, high skills are found in predicting detrended anomalies of interannual SIE at one-to-six-month lead for each target month, particularly for summer and autumn with significant correlation scores above 0.4. Interestingly, the one-to-six-month lead prediction skills in April drop significantly by 26%-46% when the decadal variation is removed, because the change of April SIE is mainly dominated by the decadal components associated with Pacific decadal oscillation. Unlike several previous results, FGOALS-f2 shows higher spatial correlation scores in predicting minimum sea ice in September at lead times of one-to-six-month for extreme years compared to normal years. This study suggests that the SIE anomalies removing the decadal variation are more representative for characterizing interannual variations compared to the linearly detrended SIE anomalies, and September Arctic sea ice is not necessarily less predictable in extreme years than in normal years.