Ice jam events can be devastating for the environment, human infrastructure, and local population. During breakup season, it is of great importance to be informed about the river ice cover condition in order to mitigate breakup flood risk. The Athabasca River near FortMcMurray, located in Alberta, is particularly prone to ice jam events and subsequent floodings. Satellite remote sensing techniques provide the necessary means to monitor the ice cover. Because of the wide availability most research and operational services for SAR river ice classification are based on single- or dual-polarized images. However, such imagery is limited in its ability to distinguish certain river ice types and open water states. The research presented examines how SAR polarimetry influences the detecting possibilities of specific ice types. Sentinel-1 (dual-polarization), RADARSAT-2 (quad-polarization) and RADARSAT Constellation Mission (compact-polarization) data were used to classify river ice during breakup. This study was about analysing a stretch of the Athabasca River which is prone to ice jam formation. First, SAR images from the 2018-2019 and 2019-2020 breakup were studied to find the temporal and spatial patterns of the radar backscatter. Next, sample areas with known ice stage (sheet ice, ice jam or open water) were selected. The sample areas of each ice stage were compared to assess the influence of SAR characteristics, as incidence angle and overpass time. In the last part of this study, a Random Forest classification was implemented in which intensity, texture and polarimetric features were used. Results show that classification accuracies increase with the inclusion of polarimetric decomposition features and GLCM mean texture features by enhancing between class separability and reducing the misclassification. Accuracies of 85.6% (Kappa = 0.78), 91.2% (Kappa = 0.87) and 91.0% (Kappa = 0.87) were obtained for Sentinel-1, RADARSAT-2 and RCM, respectively. The majority of the confusion between classes was due to similarities at backscatter signatures in very small incidence angles, mainly between open water and sheet ice under melting conditions. Also sheet ice early in the breakup season was confused with ice jams. To reduce the likelihood of misclassification, it is recommended to only use images with incidence angles higher than 30º and to include polarimetric and texture features in a classifier. Additional improvements can be achieved when using expert knowledge for tracking, since previous SAR images can provide added information when one understands the temporal patterns of river ice breakup. Further research should be directed at the development of an automatic classification approach that should be able to detect ice jams during the entire ice covered season. Having more knowledge about river ice breakup may help to eventually develop a river ice forecasting system, which may significantly reduce flood risk.

Surface melt plays a vital role in impacting the polar mass balance and global sea level rise. Over the past decades, synthetic aperture radar (SAR) imagery has garnered considerable attention due to its capacity to provide high-precision and long-term information. However, the traditional SAR-based large-scale surface melt detection methods utilizing co-orbit normalization predominantly depend on reference images and the precise spatial registration to mitigate geometric distortions arising from diverse incidence angles. Consequently, both the absence of reference imagery and the movement of ice sheets and shelves present challenges to the method. In this study, we address this issue by developing a reference-free deep learning network integrating the Convolutional Block Attention Module (CBAM) into DeepLabv3+ to automatically detect surface melt and establishing the surface melt dataset based on multi-temporal Sentinel-1 SAR imagery, encompassing diverse surface conditions of the Antarctic. Our model achieves an accuracy of 95.67%, surpassing the reference-based method and an advanced deep learning-based approach by 4.23% and 4.67%, respectively. Moreover, compared to 500 m resolution UMelt product and the kilometer-level results obtained from Advanced Scatterometer (ASCAT) and Special Sensor Microwave Imager Sounder (SSMIS), our model demonstrates the capability to accurately capture the small-scale melting patterns of ice shelves with a higher spatial resolution of 40 m. Notably, our findings underscore the dispensability of reference imagery in traditional methods through the formidable information extraction capabilities of deep learning. We finally applied the proposed method to extract and analyze the spatiotemporal characteristics of surface melt on the Larsen C Ice Shelf during the 2019/2020 period. The corresponding code of this study is at https://github.com/Tangyu35/Surface-melt-detection.@en

Surface melt plays a vital role in impacting the polar mass balance and global sea level rise. Over the past decades, synthetic aperture radar (SAR) imagery has garnered considerable attention due to its capacity to provide high-precision and long-term information. However, the traditional SAR-based large-scale surface melt detection methods utilizing co-orbit normalization predominantly depend on reference images and the precise spatial registration to mitigate geometric distortions arising from diverse incidence angles. Consequently, both the absence of reference imagery and the movement of ice sheets and shelves present challenges to the method. In this study, we address this issue by developing a reference-free deep learning network integrating the Convolutional Block Attention Module (CBAM) into DeepLabv3+ to automatically detect surface melt and establishing the surface melt dataset based on multi-temporal Sentinel-1 SAR imagery, encompassing diverse surface conditions of the Antarctic. Our model achieves an accuracy of 95.67%, surpassing the reference-based method and an advanced deep learning-based approach by 4.23% and 4.67%, respectively. Moreover, compared to 500 m resolution UMelt product and the kilometer-level results obtained from Advanced Scatterometer (ASCAT) and Special Sensor Microwave Imager Sounder (SSMIS), our model demonstrates the capability to accurately capture the small-scale melting patterns of ice shelves with a higher spatial resolution of 40 m. Notably, our findings underscore the dispensability of reference imagery in traditional methods through the formidable information extraction capabilities of deep learning. We finally applied the proposed method to extract and analyze the spatiotemporal characteristics of surface melt on the Larsen C Ice Shelf during the 2019/2020 period. The corresponding code of this study is at https://github.com/Tangyu35/Surface-melt-detection.@en

Surface melt plays a vital role in impacting the polar mass balance and global sea level rise. Over the past decades, synthetic aperture radar (SAR) imagery has garnered considerable attention due to its capacity to provide high-precision and long-term information. However, the traditional SAR-based large-scale surface melt detection methods utilizing co-orbit normalization predominantly depend on reference images and the precise spatial registration to mitigate geometric distortions arising from diverse incidence angles. Consequently, both the absence of reference imagery and the movement of ice sheets and shelves present challenges to the method. In this study, we address this issue by developing a reference-free deep learning network integrating the Convolutional Block Attention Module (CBAM) into DeepLabv3+ to automatically detect surface melt and establishing the surface melt dataset based on multi-temporal Sentinel-1 SAR imagery, encompassing diverse surface conditions of the Antarctic. Our model achieves an accuracy of 95.67%, surpassing the reference-based method and an advanced deep learning-based approach by 4.23% and 4.67%, respectively. Moreover, compared to 500 m resolution UMelt product and the kilometer-level results obtained from Advanced Scatterometer (ASCAT) and Special Sensor Microwave Imager Sounder (SSMIS), our model demonstrates the capability to accurately capture the small-scale melting patterns of ice shelves with a higher spatial resolution of 40 m. Notably, our findings underscore the dispensability of reference imagery in traditional methods through the formidable information extraction capabilities of deep learning. We finally applied the proposed method to extract and analyze the spatiotemporal characteristics of surface melt on the Larsen C Ice Shelf during the 2019/2020 period. The corresponding code of this study is at https://github.com/Tangyu35/Surface-melt-detection.@en