Impact of data assimilation on Arctic sea‐ice thickness variability and its coupling with atmospheric forcing

Abstract

Sea‐ice thickness (SIT) is a key state variable of the Arctic climate system, yet its representation in coupled ice–ocean models remains uncertain due to sparse observations and complex coupling with atmospheric forcing. Satellite‐based observations have improved the accuracy of SIT in models significantly through data assimilation (DA). This study assesses the impact of DA on the statistical characteristics of Arctic sea‐ice variability by comparing two reanalysis products for the period 2014–2017, with and without the assimilation of SIT observations, both from a coupled ocean–sea‐ice model system. On the basin scale, DA reduced the thick bias in the Beaufort Sea and the thin bias in the central Arctic ice pack, improving the overall model climatology. In terms of intraseasonal variability, DA reduced the excessive SIT variability between the Beaufort Sea and the Chukchi Sea, while increasing the SIT variability substantially in the central Arctic ice pack and exit flow through the Fram Strait. DA also increased the de‐correlation time‐scale of SIT north of Greenland, indicating a higher potential for predictability there. The coupled modes of variability between SIT and sea‐level pressure are revealed by a joint principal‐component analysis. The leading mode features a dominating Arctic gyre, while the subsequent modes feature the movement of the Arctic gyre and weather systems. With the additional constraint of SIT by DA, a large share of variability is shifted from the first mode to the subsequent modes, indicating the increased importance of weather‐driven ice variability. However, a large amount of variability for ice thickness and drift is still unexplained, which implies the importance of processes other than the atmospheric forcing.