Retrieving Sea Ice Information in the Pan-Arctic Region from Synthetic Aperture Radar

The Arctic’s unprecedented transformation due to anthropogenic warming necessitates close monitoring of sea ice to understand and address climate change impacts. As the sea ice retreats and becomes thinner, increased human activity in the region emphasizes the urgent need for detailed, near real-time sea ice information as well as improved sea ice forecasts for maritime safety and planning.

Current methods of Arctic sea ice retrieval relies on passive microwave (PMW) sensors, which offer global coverage but struggle to capture fine-scale features and changes in the sea ice. Synthetic Aperture Radar (SAR) imagery, with its high spatial resolution and independence from sunlight and clouds, is pivotal in the year-round mapping of Arctic sea ice conditions that is carried out manually at the national ice services. Yet, automating SAR-based sea ice retrieval remains challenging due to inherent ambiguities in the observations.

Recent advances in deep learning vision methodologies show promise in SAR-based sea ice retrievals. A robust pan-Arctic SAR-based sea ice retrieval system can serve maritime sectors, national ice services, and local communities by providing timely, high-resolution sea ice information. Furthermore, SAR-based sea ice retrievals can be assimilated in numerical ocean and sea ice models, improving sea ice forecasts crucial for local communities and maritime sectors.

Here, we present a comprehensive deep learning approach to retrieve high-resolution sea ice concentration and calibrated uncertainties from Sentinel-1 SAR and AMSR-2 PMW observations at a pan-Arctic scale for all seasons. Daily pan-Arctic sea ice products based on our methodology will be operationally provided as part of the Copernicus Marine Service portfolio by the end of 2024. Further, we are in the process of producing daily pan-Arctic products for the entire Sentinel-1 era, which began with the launch of Sentinel-1A in 2014. 

Lastly, we present preliminary results for the impact of assimilating level-2 SAR-based sea ice concentrations gap-filled with level-2 PMW-based sea ice concentration in the HYCOM-CICE coupled ocean-sea-ice forecasting system for the pan-Arctic region.