In January 2014, Chinese National Antarctic Research Expedition (CHINARE) 30th cruise raised public concern since the Xuelong, the Chinese polar research vessel, was trapped in the sea-ice zone (66°39'20.88"S, 144°25'2.28"E) in the vicinity of the Adélie Depression area on the east Antarctic continent. This event highlighted the importance of an operational sea-ice classification map for ice routing to serve ship navigation. In this paper, unprecedented Antarctic sea-ice classification algorithms from RADARSAT-2 satellite dual-polarization synthetic aperture radar (SAR) images were developed using the conditional random fields (CRF) approach by including multiple features from sea-ice concentration, gray-level cooccurrence matrix textures, polarization ratio, backscatter coefficients, and intensity data. Coincident RADARSAT-2 Satellite SAR datasets with five scenes were collected for ice classification into categories such as open water, thin ice, smooth first year ice, deformed first year ice, and old ice during the CHINARE-30th cruise. The effects of deformation, rafting, and ridging during the spring-summer transition period were overwhelmed by the spatial and contextual CRF models in combination with the rich features extracted for sea-ice classification. Four strategies including statistical distribution and region connection, multiple features and support vector machine (SVM) integrated into the CRF model are proposed to describe the sea-ice-type relationships among pixels. By conducting comparative experiments between the proposed methods and state-of-the-art sea-ice classification based on the SVM algorithm, the best was obtained from the SVM-based CRF (SVM-CRF) algorithm for sea-ice classification with respect to the three scenes from the Indian Ocean sector and two scenes of Pacific Ocean sectors including medium-resolution dual-polarization SAR imagery with a pixel spacing of 50 m and higher resolution dual-polarization SAR imagery with a pixel spacing of 6.25 m. Results indicate that the SVM-CRF approach has the capacity for improving sea-ice classification, which can provide accurate and reliable sea-ice class information for sea-ice analysis.

High spatial resolution sea ice concentration (SIC) is crucial for global climate and marine activity. However, retrieving high spatial resolution SIC from passive microwave sensors is challenging due to the trade-off between spatial resolution and atmospheric contamination. Our study develops the Ice-WaterNet framework, a novel superpixel-based deep learning model that integrates Conditional Random Fields (CRF) with a dual-attention U-Net to enhance ice–water classification in Synthetic Aperture Radar (SAR) imagery. The Ice-WaterNet model has been extensively tested on 2735 Sentinel-1 dual-polarized SAR images from 2021 to 2023, covering both winter and summer seasons in the Fram Strait. To tackle the complex surface features during the melt season, wind-roughened open water, and varying ice floe sizes, a superpixel strategy is employed to efficiently reduce classification uncertainty. Uncertain superpixels identified by CRF are iteratively refined using the U-Net attention mechanism. Experimental results demonstrate that Ice-WaterNet achieves significant improvements in classification accuracy, outperforming CRF and U-Net by 3.375% in Intersection over Union (IoU) and 3.09% in F1-score during the melt season, and by 1.96 in IoU and 1.75 in F1-score during the freeze season. The derived high-resolution SIC products, updated every two days, were evaluated against Met Norway ice charts and compared with ASI from AMSR-2 and SSM/I, showing a substantial reduction in misclassification in marginal ice zones, particularly under melting conditions. These findings underscore the potential of Ice-WaterNet in supporting precise sea ice monitoring and climate change research.

This study provides the first long-time series of spatial and temporal distributions for small lakes in the Larsemann Hills (69°23′S, 76°20′E) in the East Antarctic. In the Larsemann oasis, there is a significant number of over 150 small lakes, which can be observed with high spatial resolution in remote sensing imagery. However, accurately identifying and analyzing these small water bodies and elongated rivers has been challenging due to the mixed pixels effect and limitations in available middle spatial resolution imagery. In our study, we propose a data-driven approach within the Conditional Random Fields (CRF) framework, which considers three scales: superpixel, pixel, and subpixel, to refine the boundaries of small water bodies efficiently. The superpixel level quickly identifies the main water body and NDWI provides a buffer region, while the pixel level employs SVM to obtain a more precise boundary. Subpixel mapping technology within the pixel level further reduces mixed pixel effects for improved accuracy. The waterbodies were extracted from Sentinel-2 images with a spatial resolution of 10 m. The lake boundaries derived from the proposed algorithm in this study showed good agreement with in-situ measurements of the lake shoreline delineated by aerial images from the 39th Chinese Antarctic Scientific Expedition. The analysis revealed distinct seasonal patterns across the Larsemann Hills while the lake areas achieved their peak extents earlier, specifically in February before 2020 and in January after 2020. The water body mapping based on the proposed algorithm can contribute to Antarctic remote sensing hydrological observations, particularly in the monitoring of outburst events. These findings demonstrate the potential of extending this method to other Antarctic oases to enhance intra-annual lake observations. Moreover, Sentinel-2 images provide valuable remote sensing data for studying the seasonal cycles of water bodies, including those of varying sizes in the Larsemann Hills, based on long-term time series imagery.

In recent years, the range and impact of atmospheric rivers (ARs) landing in Greenland have been increasing. Extreme precipitation brought by ARs can cause a short-term surge in regional ice and snow mass, resulting in “transient” signal in the coordinate time series of Greenland GNSS stations. However, previous studies have mainly focused on ARs detection and corresponding climate impact factors, lacking research on the contribution of extreme AR precipitation to crustal deformation. This article utilizes MERRA-2 reanalysis data and the vIVT algorithm to statistically analyze the frequency, duration, and spatiotemporal distribution of ARs landings in Greenland from 2010 to 2019. We quantitatively analyze the extreme precipitation events caused by ARs and their spatiotemporal evolution characteristics across Greenland and its basins. Then, using coordinates time series from 46 GNSS stations in Greenland, we extract the transient crustal deformation signals caused by AR-induced precipitation based on the multichannel singular spectrum analysis method. The results show that from 2010 to 2019, ARs landed in Greenland a total of 186 times, with a duration of 3174 h, and they showed an overall increasing trend. AR-induced precipitation in Greenland accounted for approximately 13.42% of annual precipitation, with the western region contributing more than 16%. Overall, the contribution rate of AR-induced rainfall was lower than that of AR-induced snowfall. Transient crustal deformation signals caused by AR-induced precipitation were found in some GNSS stations, with vertical deformation reaching up to 1 mm. Coastal areas showed greater deformation than inland areas, and low-latitude areas showed greater deformation than high-latitude areas.

Supraglacial lake depth is a key variable for quantifying surface meltwater storage and assessing ice-shelf stability, yet spatially continuous and reliable bathymetric information remains difficult to obtain in polar regions because in situ measurements are scarce and optical imagery cannot directly provide water depth. This study develops an integrated framework for supraglacial lake identification and bathymetry retrieval by combining ICESat-2 ATL03 photon-counting lidar data with Sentinel-2 multispectral imagery. ICESat-2 lake photons were used to constrain lake-region extraction from Sentinel-2 imagery, and the photon-derived along-track depths were corrected for scattering and refraction before being converted into Sentinel-2 pixel-level depth labels. Based on these labels, four retrieval models were constructed and evaluated, including an empirical model, CatBoost, a convolutional neural network (CNN), and a residual dense network (RDN). CatBoost generated initial depth estimates, while CNN and RDN further incorporated the CatBoost-derived depth prior and Sentinel-2 multispectral features for pixel-level depth prediction. Experiments over four investigated supraglacial lakes showed that RDN achieved the best average performance across the investigated lakes, with mean R², RMSE, and MAE values of 0.927, 0.187 m, and 0.144 m, respectively. For the investigated lakes, the integration of ICESat-2 and Sentinel-2 extended discrete along-track reference-depth observations to spatially continuous bathymetry maps. Because the training and validation samples were obtained from different spatial blocks within the same four lake scenes, the reported performance primarily reflects within-lake spatial generalization under the investigated conditions, and transferability to unseen lakes remains to be evaluated. These maps may provide inputs for future lake-volume estimation and ice-shelf hydrological analyses, while their applicability to lakes with different morphological and optical conditions requires further evaluation.