Assimilating summer sea ice thickness enhances predictions of Arctic sea ice and surrounding atmosphere within two months

Abstract

Subseasonal prediction of Arctic sea ice and associated atmospheric conditions during the melting season remains challenging due to limited understanding of sea ice initial conditions. This study integrates sea ice assimilation into the coupled model FGOALS-f2 using the localized error subspace transform ensemble Kalman filter, and conducts subseasonal predictions starting from August 1st over 2004-2023. Results show that simultaneous assimilation of sea ice concentration (SIC) and thickness (SIT) significantly improves sea ice predictions for up to two months, while assimilating SIC alone primarily benefits one-month lead predictions. SIT assimilation provides added predictive value for surface air temperature (SAT) forecasts beyond SIC assimilation alone, effectively extending the atmospheric influence of sea ice initial conditions to two months. This improvement in SAT predictions is primarily attributed to a more realistic representation of the surface energy budget. These findings highlight the pivotal role of summer SIT assimilation to enhance subseasonal predictions in the Arctic and challenge the conventional view that initial conditions affect only short-term forecasts. This study underscores the necessity for better representation of ice-atmosphere interactions in models and advocates for enhanced observational capabilities for summer SIT to improve subseasonal predictions in the Arctic and surrounding regions.