Abstract: Eddies in the marginal ice zone (MIZ) are critically important for understanding, monitoring, and predicting polar ocean hydrographic and thermodynamic processes. With sea ice serving as a tracer, MIZ eddies are visible in both optical and synthetic aperture radar (SAR) satellite imagery. Existing studies have primarily relied on Sentinel-1 SAR imagery and employed deep learning models for detection. However, due to eddy motion and sea ice interference, Sentinel-1 imagery is insufficient for further analysis of the three-dimensional structural parameters of MIZ eddies. To address this limitation, this study trains a deep learning model for automated detection of MIZ eddies based on the L2_LR_SSH product (Version D) from the SWOT (Surface Water and Ocean Topography) mission, using both the Unsmoothed and Expert files. Equipped with a Ka-band SAR interferometer (KaRIn), the SWOT mission provides not only imaging capability but also sea surface height (SSH) information, thereby enabling three-dimensional structural analysis of oceanic phenomena. This study develops an incidence-angle correction method based on the Geophysical Model Function (GMF), which effectively corrects the anomalously high values of newly formed frazil ice in SWOT σ₀ data. This study construct a dataset comprising more than 1,000 MIZ eddies from SWOT σ₀ and SSHA data, and train a YOLO26 instance segmentation model on this dataset, the model achieving an mAP50 of 0.71 on the validation set. This model was applied to detect MIZ eddies around Greenland over a one-year period spanning June 2025 to May 2026; the detection results are consistent with prior studies based on Sentinel-1 imagery, and the model successfully identifies some eddies with radii smaller than 1.5 km, demonstrating the credible capability of SWOT for detecting submesoscale oceanic phenomena. Finally, this paper presents a preliminary investigation of the imaging characteristics of KaRIn in the marginal ice zone and the signatures of MIZ eddies in SSHA data, confirming the strong potential of SWOT data for three-dimensional structural studies of oceanic phenomena in the marginal ice zone.
 

暂无