A long-term proxy for sea ice thickness in the Canadian Arctic: 1996-2020

This study presents a long-term winter sea ice thickness proxy product for the Canadian Arctic based on a random forest regression model - applied to ice charts and scatterometer data, trained on CryoSat-2 observations, and applying an ice type-sea ice thickness correction using the Pan-Arctic Ice Ocean Modeling and Assimilation System (PIOMAS) - that provides 25 years of sea ice thickness in the Beaufort Sea, Baffin Bay, and, for the first time, the Canadian Arctic Archipelago. An evaluation of the product with in situ sea ice thickness measurements shows that the presented sea ice thickness proxy product correctly estimates the magnitudes of the ice thickness and accurately captures spatial and temporal variability. The product estimates sea ice thickness within 30 to 50 cm uncertainty from the model. The sea ice thickness proxy product shows that sea ice is thinning over most of the Canadian Arctic, with a mean trend of 0.82 cm yr(-1) in April over the whole study area (corresponding to 21 cm thinning over the 25-year record), but that trends vary locally. The Beaufort Sea and Baffin Bay show significant negative trends during all months, though with peaks in November (2.8 cm yr(-1)) and April (1.5 cm yr(-1)), respectively. The Parry Channel, which is part of the Northwest Passage and relevant for shipping, shows significant thinning in autumn. The sea ice thickness proxy product provides, for the first time, the opportunity to study long-term trends and variability in sea ice thickness in the Canadian Arctic, including the narrow channels in the Canadian Arctic Archipelago.