Mohammed Dabboor (Science and Technology Branch, Environment and Climate Change Canada, Canada); Saeid Homayouni (Centre Eau Terre Environnment, Institut National de la Recherche Scientifique, Canada)

Sea ice mapping in the Canadian Arctic is essential for monitoring climate change impacts and supporting safe navigation. Synthetic Aperture Radar (SAR) missions such as RADARSAT Constellation Mission (RCM) and Sentinel-1 have proven effective for ice typing since they provide high spatial resolution and weather independence. Conversely, the Surface Water and Ocean Topography (SWOT) satellite offers complimentary altimetric measurements with the potential to provide two-dimensional surface elevation maps.
This study proposes a machine learning approach for sea ice classification using Ka-band SAR imagery from the SWOT mission. We investigate the influence of SWOT’s radar incidence angle on classification performance. Several locations in the Canadian Arctic are selected as case studies. In addition, classification results are compared with sea ice types derived from co-located imagery from the RCM and Sentinel-1 satellites. Through this study, we aim to support the ice flags in SWOT products by incorporating information on different sea ice types.