Evolution of Arctic sea ice in CMS reanalyses

Arctic sea ice has recently experienced rapid changes, indicating a transition toward a new sea ice regime dominated by the marginal ice zone (MIZ) during summer. Modifications in extent, distribution, and volume of the MIZ have significant implications for polar and global climate, as the physical processes in the marginal ice largely differ from those in the pack ice, including air/sea exchanges, dynamic interactions with waves and currents, fast thermodynamic changes, and impact on marine ecosystems.Copernicus Marine Service (CMS) provides a wide range of products capable of detecting the evolution of Arctic sea ice. Interestingly, not only the Arctic-focused regional products (such as TOPAZ reanalysis and satellite observations) can be used for this purpose, but the Global ocean Reanalysis Ensemble Product (GREP) has also proven its effectiveness in capturing the recent-past state of the Arctic sea ice.Here, we study the temporal and spatial variability of Northern Hemisphere sea ice area and thickness over the past three decades, assessing their representation across a range of CMS products. The reanalyses are examined against CMS remote sensing observations as well as other Arctic reanalysis products.We propose metrics at the pan-Arctic scale while also emphasizing the different responses of MIZ and consolidated pack ice to climate change. The results show that GREP and TOPAZ provide reliable estimates of present-day and recent past Arctic sea ice states and accurately reproduce the space/time variability of the MIZ area. In recents summers, the MIZ across both products has accounted for up to 40% of the total Arctic sea ice area, with its position (computed as monthly averaged latitude) experiencing a northward shift due to the contraction of pack ice in the central Arctic. Notably, despite an increased ensemble spread for sea ice thickness compared to that of sea ice area, GREP displays coherent interannual variability and trend. The proportion of GREP sea ice thinner than 2m has increased from 40% to 80% over the last 30 years. Additionally, TOPAZ demonstrates a significant impact of data assimilation updates on its outputs.Overall, this study confirms that CMS reanalysis products are adequate tools for understanding  the mean state and variability of ice classes in the Arctic region. Furthermore, these products hold significant potential for training machine learning model emulators for new predictions and supporting climate-related applications.