Arctic Wintertime Sea Ice Lead Detection from Sentinel-1 SAR Images

Leads in sea ice cover are almost linear fractures within the pack ice, and are commonly observed in the polar regions. In winter, leads promote energy flux from the underlying ocean to the atmosphere. Synthetic aperture radar (SAR) can monitor leads with a fine spatial resolution, regardless of solar illumination and atmospheric conditions. In this paper, we present an approach for automatic sea ice lead detection (SILDET) in the Arctic wintertime using Sentinel-1 SAR images. SILDET is made up of four modules: 1) a segmentation module; 2) a balance module; 3) an optimization module; and 4) a mask module. The validation results presented in this paper show that SILDET has the capability of detecting open and frozen leads at different stages of freezing. The lead map obtained from SILDET was compared to a lead dataset based on Moderate Resolution Imaging Spectroradiometer (MODIS) data and validated by the use of Sentinel-2 images. This shows that SILDET can provide a more detailed distribution of leads and better estimation of lead width and area. Compared with visual interpretation of Sentinel-1 images, the overall detection accuracy is 97.80% and the Kappa coefficient is 0.88 (for all types). The pyramid scene parsing network (PSPNet) in the segmentation module shows a better performance in detecting frozen leads, compared with the deep learning methods of UNet and DeepLabv3+. The optimization module utilizing shape features also improves the precision in detecting frozen leads. SILDET was applied to present the Arctic lead distribution in January and April 2023 with a spatial resolution of 40 m. The Arctic-wide lead width distribution follows a power law with an exponent of 1.64 ± 0.07. SILDET can be expected to provide long-term high-resolution lead distribution records.