Data Assimilation of Satellite-Derived Arctic Sea-Ice Thickness During Boreal Summer

Reliable estimation and initialization of Arctic sea ice thickness (SIT) through data assimilation (DA) during the summer melt season were previously hampered by the lack of available observations owing to limitations in satellite retrieval algorithms. Recently, successful satellite-derived Arctic SIT measurements from CryoSat-2 (CS2) and AMSR2 during the boreal summer have been achieved using advanced retrieval algorithms. This study compares the impacts of CS2 and AMSR2 SIT datasets by individually assimilating each dataset using the ensemble optimal interpolation DA technique with CICE 5 dynamical sea ice model in 2019 and 2020. The underestimated sea ice extent (SIE) in the control simulation without DA during summer was effectively corrected in the reanalysis assimilating AMSR2. However, the degree of correction was less pronounced in the reanalysis assimilating CS2. A sensitivity experiment confirmed that the weak correction degree when using CS2 was not due to its low spatiotemporal resolution, suggesting that the issues may arise from a systematic negative bias related to ice roughness over the central Arctic Ocean in CS2. During the summer and subsequent sea-ice growing seasons, the simulated SIT in the DA of AMSR2 shows greater similarity with independent reanalysis and satellite data than that of CS2. Validations against SIT observations measured by ice mass balance and upward-looking sonar indicate that the DA of AMSR2 effectively enhances the day-to-day variability compared to CS2 and control simulations during both the summer and subsequent winter seasons. This study underscores the response of the model to assimilating current satellite summer SIT data and highlights the factors to consider when utilizing these data.