Phytoplankton bloom is closely tied to the seasonal evolution of Antarctic coastal polynyas, yet a phenology-based framework linking polynya dynamics to bloom spatiotemporal patterns and phase-specific environmental drivers has been missing. This study uses satellite-derived sea ice concentration and chlorophyll-a (Chl-a) data from 2002 to 2023 to examine the polynyas evolution in the Ross Sea and their relations with phytoplankton blooms. We characterize spatiotemporal polynya dynamics quantitatively based upon double-logistic functions and phenological metrics. Environmental drivers including photosynthetically active radiation (PAR), sea surface temperature (SST), wind speed (WS), and wind direction (WD) are integrated to assess their influence on different bloom phases. Results show significant increases in polynya areas (up to 1240.57 km2 per year) from 2002 to 2023. Polynyas exhibit a spatial pattern of slow northward advance and rapid southward retreat. Over the past two decades, they have tended to advance earlier, retreat later, and have longer open water durations (∼6 d per decade). A strong positive correlation is found between polynya advance and bloom initiation, with regionally varying regression slopes exceeding 2 d per decade. Environmental drivers of Chl-a show distinct regional and phenological differences: in the Ross Sea Polynya, PAR is the dominant positive factor during both growth and decline phases, whereas SST and WD show significant negative effects during the decline phase. In the Terra Nova Bay Polynya, growth-phase Chl-a is significantly influenced by environmental variables. These findings enhance our understanding of polynya‒bloom interactions in the Ross Sea.

A profound understanding of unique underwater terrain characteristics is crucial for the Internet of Underwater Things (IoUT). Currently, using autonomous platforms equipped with multibeam echosounder systems (MBESs) for underwater topographic scanning has become a key method and future trend to understand seabed topography. However, nonideal integration between MBES transducers and motion sensors induces dynamic integration errors (time delay, motion scale, yaw misalignment, and lever arm errors), which manifest as high-frequency ship-track orthogonal bathymetric undulations, severely limiting the accurate generation of high-resolution seabed topographic maps. To address this, this study innovatively proposes a multibeam dynamic error detection method based on time-series neural networks (TSNNs). Its core innovations are: first, leveraging the characteristic that dynamic errors' ship-track orthogonal bathymetric undulations are highly correlated with attitude; second, focusing on distinct manifestations of beam-angle-dependent and beam-angle-independent dynamic errors across different beam reception angles, laying a key basis for accurate error inversion. The method involves two core modules: the "multibeam bathymetric sequence trend extraction module" separates seabed terrain trends from undulations to obtain error-driven fluctuation signals; the "dynamic error regression module" realizes quantitative prediction of dynamic integration errors. Experimental results show that the method can effectively eliminate high-frequency undulations in shallow- and medium-water MBES data, with processing efficiency meeting autonomous platforms' real-time requirements-providing a new technical approach for high-resolution marine surveying and mapping to support IoUT's reliable deployment and operation.

The offshore wind power industry is undergoing rapid expansion, with installed capacity growing exponentially. Scour-induced erosion around turbine foundations remains the foremost challenge to operational stability, as the efficacy and adequacy of scour protection measures directly impact operational costs and safety risks in offshore wind farms. However, the lack of robust methodologies for precise scour monitoring and quantification has hindered practical solutions in engineering applications. This study proposes a multiperiod monitoring and quantitative scour analysis framework utilizing multibeam technology. First, a comprehensive point cloud segmentation method combining Z-axis normal vector thresholds with Euclidean clustering was developed to isolate turbine pile structures from surrounding seabed terrain while eliminating multibeam bathymetric outliers. Subsequently, two critical alignment challenges were addressed: 1) Planar reference inconsistencies across multitemporal data sets were resolved through random sample consensus algorithm-based registration enhanced by prior knowledge of offshore wind turbine geometries; 2) Vertical benchmarks were aligned by histogram cluster analysis of depth variance. Finally, surface fitting algorithms enabled quantitative scour evaluation for individual turbine foundations. The proposed methodology solves practical problems in two offshore wind farm projects and provides an important reference for quantitative analysis of scour in other offshore wind farm projects and other underwater structures.

High-precision seabed topographic information is the foundation for reliable Internet of Underwater Things (IoUT) deployment. Unmanned platforms equipped with Multibeam Echosounder Systems (MBES) have become the primary means for large-scale underwater topographic sensing in IoUT networks. However, autonomous multibeam surveys conducted with little or no direct human intervention face a common environmental-observation constraint: accurate and spatially representative in-situ sound speed profiles are often difficult or costly to acquire without interrupting continuous survey operations. Conventional equivalent sound speed profile (ESSP) inversion algorithms suffer from three inherent drawbacks: reference depth dependence, vertical structure mismatch with realistic ocean sound fields, and misleading false-consistency bathymetric matching. Targeting these bottlenecks, this paper develops a prior ocean sound speed structure constrained layered ESSP inversion framework utilizing redundant overlapping bathymetric observations from MBES. After systematic classification of vertical SSP characteristics across shallow, medium and deep oceans, four terrain-adaptive layered ESSP parametric models are established, alongside a customized multi-algorithm collaborative optimization scheme matched to different model dimensions. By jointly introducing overlapping sounding consistency constraint and prior structural information from global remote sensing reanalysis datasets, the proposed method eliminates false consistency and drastically reduces inversion parameter space. Comprehensive validations based on 50–5000 m full-depth field measurements demonstrate that the proposed approach limits bathymetry standard deviation below 1% of local water depth over rugged seabed without apparent acoustic refraction distortion. This technique eliminates the need for manual in-situ SSP measurements, enabling accurate ESSP inversion with only sparse remote sensing or distant historical profiles, supporting fully autonomous high-precision bathymetric sensing for IoUT unmanned platforms.

Rapid Arctic warming has accelerated sea ice drift, deformation and fragmentation, reshaping the Arctic ice environment and creating an urgent need for dynamic ice floe monitoring to ensure safe Arctic navigation. Marine radar (MR), with its all-weather, high-temporal-resolution capabilities, uniquely captures rapid small-scale ice deformation in fragmented ice fields. However, blurred floe boundaries and weak surface texture in MR imagery impede the accurate extraction of ice dynamic parameters. To address this challenge, we developed a Floe Motion and Deformation (FMD) framework specifically for MR data. It treats individual ice floes as discrete objects and employs point-set registration to estimate their displacement vectors and rotation angles. Hence, it enables a quantitative characterization of both motion and deformation at the floe scale. We first evaluated FMD on synthetic datasets, where it shows a high accuracy in retrieving kinematic and deformation parameters. Next, we applied it to MR image sequences acquired during the R/V Zhong Shan Da Xue Ji Di FACE2024 (Following Arctic/Antarctic iCE 2024) expedition, where it successfully tracked Lagrangian trajectories of 2425 individual ice floes. Validation against GPS buoys deployed on ice floes shows that FMD achieves a linear velocity error of 0.25 cm s-1 and an angular velocity error of 9.5 & times; 10-6 rad s-1 over 10-minute intervals. The FMD framework thus provides a high-precision, minute-scale solution for monitoring large floe populations, directly enhancing a vessel's ability to perceive dynamic ice environments and supporting the development and application of autonomous navigation systems in ice-covered waters.

Real-time seabed sediment classification (SSC) is crucial for underwater navigation, operations, and habitat assessment. Conventional methods relying on post-mission multibeam-echosounder (MBES) data processing impede in situ decision-making. We propose a novel, real-time SSC method deployable on both shipborne and Autonomous Underwater Vehicle (AUV) platforms, integrating three core components. Primarily, an efficient preprocessing pipeline comprising georeferencing, radiometric normalization, noise suppression, and incidence-angle correction enables rapid conversion of raw MBES backscatter into geometry-consistent tiles, supporting real-time operation with sub-second responsiveness. Afterwards, the system extracts multi-modal descriptors by combining entropy-regularised angular-response fitting for acoustic backscatter, object-level texture analysis using adaptive graph segmentation, and curvature-aware terrain metrics derived from quadratic surface fitting under entropy constraints by considering the physical responses and spatial distribution of MBES images and point clouds. Finally, a Dynamic Optimal Random Forest with Entropy-Adaptive Subnetwork Selection (DORF-EASNet) dynamically selects between a global classifier and lightweight domain-specific sub-models to match local acoustic complexity, achieving a balance between inference efficiency and physical interpretability. Field experiments conducted in Jiaozhou Bay and the South China Sea demonstrate the proposed framework’s robustness across platforms and sensing configurations, achieving macro-F1 scores of 0.881 and 0.913, respectively, while maintaining real-time processing capability exceeding that of conventional offline methods.

In recent years, the extent of Antarctic sea ice has frequently reached historically low levels. Melt ponds significantly affect the sea ice heat balance and mass balance while posing challenges to ice surface transportation. Melt ponds in the Arctic have been extensively monitored using satellite remote sensing. However, the Antarctic has received very little attention, and small-scale melt pond distribution patterns remain virtually unexplored. This study combines high-resolution uncrewed aerial vehicle (UAV) imagery with satellite remote sensing to investigate the characteristics of melt ponds on landfast ice near Zhongshan Station, East Antarctica. UAV-derived data reveal that the melt ponds in this region are small (median area = 0.34 m(2)) and irregular in shape (mean roundness = 4.97). Validation against the UAV data suggests poor performance of the traditional melt pond fraction (MPF) retrieval model in the Antarctic region. Therefore, we develop an MPF retrieval algorithm by integrating UAV and Landsat 8 images. The correlation coefficient is as high as 0.82, substantially outperforming the traditional method. Dynamic MPF monitoring indicates that melt ponds concentrate near the coastline, especially to the west of bare rock areas, as the persistent summer easterly winds transport low-albedo dust westward. The integrated melt-induced risk assessment, combining MPF and Normalized Difference Water Index for ice, shows that high-risk areas are concentrated in Nella Fjord, west of the Mirror Peninsula. Ice surface transport should avoid this area and be completed before mid-November.

Against the backdrop of a rapid trajectory toward a seasonally ice-free Arctic, late-summer Chukchi Sea ice recovered over 2020–2024, culminating in an abrupt retention that obstructed the Northeast Passage by September 2024 for the first time since 2001. Here we show that this recovery coincides with a persistent poleward-shifted Pacific–Arctic cyclonic anomaly near the Bering Strait. This shifted circulation produces a coupled dynamic–thermodynamic response, with Arctic cold advection, enhanced cloud-induced cooling, reduced oceanic heat, and delayed seasonal thinning acting together to sustain late-summer ice retention. Furthermore, we demonstrate that the Pacific–North American (PNA) mode is linked to summer Chukchi Sea ice variations, with its influence expressed through the regional position of the associated Pacific–Arctic cyclonic anomaly. These findings identify the Chukchi Sea as a sensitive Pacific Arctic sector where natural regional circulation variability generates abrupt, localized reversals in summer sea-ice loss, thereby offering a fresh perspective on the role of summer PNA-related circulation variability in Pacific-sector Arctic sea ice. The recovery of late-summer Chukchi Sea ice over 2020–2024 was driven by a poleward-shifted Pacific–Arctic cyclonic anomaly that enhanced cooling and delayed seasonal thinning, according to a study combining satellite observations, sea-ice thickness estimates and atmospheric-oceanic reanalyses.

Influenced by the measurement mechanism, the marine environment, and other factors, side-scan sonar images often exhibit high noise levels and low resolution. This presents significant challenges for high-resolution underwater topography imaging and small-scale target detection. To address this, this paper proposes a super-resolution reconstruction method for side-scan sonar images based on a diffusion model and texture consistency. Firstly, based on the imaging mechanism of side-scan sonar, the seafloor reverberation model is introduced to establish the degradation mechanism for underwater acoustic images. Next, a feature bootstrap module is developed to integrate low-resolution image and texture features, projecting them into a potential semantic space. Finally, through multiple iterations of the diffusion model, guidance information at different scales is incorporated into the generation process. This enables the reconstruction of target contours, edge features, and high-frequency details at various iteration stages. As a result, texture-consistent, high-quality side-scan sonar images are obtained. Experimental results demonstrate that the proposed algorithm outperforms the comparison methods in both subjective visual effects and objective evaluation metrics. Among them, the FID scores reached 106.66, 117.27, and 148.41, respectively. In addition, we conducted SSS pipeline target segmentation experiments and large-area georeferenced image reconstruction experiments, further verifying the feasibility and effectiveness of the proposed method in practical tasks such as SSS target detection, large-area topographic surveys, and high-resolution imaging. The source code is available at: https://github.com/Yang-Code98 4/Sonar-super-resolution-main.

During the 2024 Arctic summer, an unexpected sea ice patch (∼100 000 km ^2 ) survived in the western Chukchi Sea, an event that has not occurred in the past 20 years. Combining satellite data, PIOMAS model, and ERA5 reanalysis data, we analyze this anomaly’s evolution and its dynamic and thermodynamic drivers in the 2000–2024 series. The analysis reveals that July–August are key months; ice concentration and thickness during this time determines the fate of the sea ice survival in September. Before July 2024, sea ice near Wrangel Island has experienced rapid and sustained thickening, with its distribution closely resembling the levels observed in 2000 and 2001 (the only other summers with ice survival). Persistent northerly winds east of Wrangel Island during the melt season contributed to sea ice thickening and played a crucial role in its survival. At the same time, a cooler-than-normal air temperature near Wrangel Island supported the stable presence of sea ice in this region. The study underscores that the ‘New Arctic’ conditions, namely thinning and increasingly mobile sea ice, raise the likelihood of unpredictable events of both unexpected sea ice survival and sudden loss.

Abstract. Synthetic Aperture Radar (SAR)-based sea ice classification faces challenges due to the similarity among surfaces such as wind-driven open water (OW), smooth thin ice, and melted ice surfaces. Previous algorithms combine pixel-based and region-based machine learning methods or statistical classifiers, yet struggle with hardly improved accuracy arrested by the fuzzy surfaces and limited manual labels. In this study, we propose an automated algorithm framework by combining the semantic segmentation of ice regions and the multi-stage detection of ice pixels to produce high-accuracy and high-resolution ice-water classification data. Firstly, we used the U-Net convolutional neural networks model with the well processed GCOM-W1 AMSR2 36.5 GHz H polarization, Sentinel-1 SAR EW dual-polarization data, and CIS/DMI ice chart labels as data inputs to train and perform semantic segmentation of major ice distribution regions with near-100 % accuracy. Subsequently, within the U-Net semantically segmented ice region, we redesigned the GLCM textures and the HV/HH polarization ratio of Sentinel-1 SAR images to create a combined texture, which served as the basis for the Multi-textRG algorithm to employ multi-stage region growing for retrieving ice pixel details. We validated the SAR classification results on Landsat-8 and Sentinel-2 optical data yielding an overall accuracy (OA) of 84.9 %, a low false negative (FN) of 4.24 % indicating underestimated low backscatter ice surfaces, and a higher false positive (FP) of 10.8 % reflecting their resolution difference along ice edges. Through detailed analyses and discussions of classification results under the similar ice and water conditions mentioned at the beginning, we anticipate that the proposed algorithm framework successfully addresses accurate ice-water classification across all seasons and enhances the labelling process for ice pixel samples.

Efficient acquisition of high-quality underwater geomorphological information forms the foundation of real-time marine mapping. However, the inherent complexity of the underwater environment, coupled with the measurement and imaging characteristics of side-scan sonar (SSS), poses significant challenges to real-time high-quality SSS image generation. To bridge this gap, this study presents a comprehensive methodology for fast processing and quality enhancement of SSS raw data. Our approach encompasses four pivotal technologies: real-time data quality assurance, bottom line automatic tracking, radiometric distortions correction, and slant range adjustment. Quality control is achieved through rigorous preprocessing, including a sliding window-based outlier suppression algorithm for backscatter intensities and Kalman filtering for navigation data. Bottom-line segmentation leverages a customized U-Net architecture tailored to SSS characteristics, enabling accurate delineation of seabed-water boundaries while incorporating symmetry priors inherent in sonar imagery. In addition, radiometric correction via Time-Varying Gain (TVG) and geometric correction based on the spatial configuration of the towfish, seabed, and echo paths further improve image fidelity. Overall, the proposed pipeline not only facilitates real-time SSS image processing but also enhances the efficiency and reliability of downstream tasks such as target recognition and seafloor interpretation.

The limited availability of in situ ocean observations poses significant challenges to real-time oceanographic applications, particularly in hydroacoustic measurements where accuracy critically depends on the spatiotemporal variability of sound speed. To address the sparsity of sound-speed profile (SSPs), this study proposes an advanced modeling framework for constructing regional sound speed fields by integrating temporal and spatial dynamics. Specifically, the variation mechanisms of temperature and salinity are analyzed, and empirical orthogonal function decomposition is used to extract compact SSP representations. A novel multimodel temporal prediction architecture, combining seasonal-trend decomposition using LOESS, long short-term memory, multivariate unsupervised domain adaptation, and inverted Transformer, captures complex seasonal and adaptive patterns. Meanwhile, spatial modeling adopts a particle swarm optimization least squares support vector machine approach to enhance interpolation across diverse marine environments. Experiments show that the model outperforms existing methods, achieving a root-mean-square error of 0.812 m/s and a mean absolute percentage error of 0.037%. Its robust prediction capability supports accurate multibeam bathymetric processing even without direct SSP observations, confirming its practical value for real-time ocean mapping.

High-salinity shelf water (HSSW) acts as a precursor to the Antarctic Bottom Water and plays a critical role in regulating the global ocean circulation system. This study employs a high-resolution coupled ocean–sea ice–ice shelf model to analyze the interannual variation in HSSW formation in the Ross Sea, which is one of the major production sites of HSSW. We are particularly focused on anomalously high HSSW production during the winter of 2007. The results indicate that, in this winter, there were frequent passages of synoptic-scale cyclones that were centered near the front of the Ross Ice Shelf. The western flanks of these cyclones significantly enhanced offshore winds over the western Ross Ice Shelf Polynya, a major origin site of HSSW in the Ross Sea, leading to a sharp increase in ice production within this polynya. The HSSW formation resulting from brine rejection during ice freezing reached the highest volume of 16 000 km3 in 2007. However, the salinity and density of the Ross Sea during this period exhibited unexpectedly low values. This inconsistency was due to a rapid increase in ice shelf melting over the Amundsen Sea and Ross Sea during 2006–2007, with annual cumulative melt rates reaching a peak in recent decades. Meanwhile, the resulting large amount of meltwater was transported westward into the Ross Sea by notably strong slope and coastal currents in 2007, leading to large fluxes of freshwater flux into the Ross Sea. The interaction between enhanced HSSW formation driven by ice production and the large influx of meltwater highlights the complex dynamics that shape hydrographic variability in the Ross Sea.

Against the backdrop of global climate change, the continued decline in Arctic sea ice extent and thickness has intensified the dynamic evolution of the marginal ice zone (MIZ). As a critical transitional region between the open ocean and pack ice, the MIZ plays a pivotal role in mediating ocean-atmosphere interactions, influencing sea ice dynamics, and supporting polar ecosystems. This study investigates the dynamic variability and morphological evolution of the Arctic MIZ from 1979 to 2023 using the Bootstrap sea ice concentration (SIC) product. Results reveal that while the overall MIZ extent has remained relatively stable over the long term, the MIZ fraction (i.e. the ratio of MIZ extent to Arctic sea ice extent) has increased significantly, as the total sea ice extent has decreased over time. The seasonal cycle is pronounced, with minimum extents observed in March or April and maximum extents in August or September. From June to September, the SIC values within the MIZ showed a significant downward trend in spatial distribution, indicating that the SIC in this region generally decreased during summer. Furthermore, the MIZ has experienced a northward shift over the past four decades, with an accelerated rate of migration post-2000. This shift is accompanied by morphological changes, characterized by a smoother ice edge and more compact ice during late summer. A significant change point was detected in 2006, signaling a structural shift in MIZ dynamics. Post-2006, the frequency of MIZ occurrence increased in high-latitude regions, particularly across the Beaufort, Chukchi, East Siberian, and Laptev Seas. These findings provide critical insights into Arctic sea ice dynamics, highlighting the evolving nature of the MIZ and its role in shaping the future Arctic ice regime under continued climate change.

Antarctic coastal polynyas play a vital role in atmosphere–ocean interactions and local ecosystems. This study investigates the interannual variability of springtime coastal polynyas over the Ross Sea based on satellite-retrieved sea-ice concentration (SIC) data from 1992 to 2021. Firstly, the springtime coastal polynya areas display large interannual variability as well as a positive trend of about 2000 km2 (10 yr)−1 over the 30 years. Secondly, based on composite analysis, in spring, we find that a deepened Amundsen Sea Low (ASL) induces stronger meridional winds over the eastern Ross Sea, leading to stronger sea-ice advection and expansion of coastal polynya areas. This is accompanied by more solar radiation absorption in early summer (about 16 W m−2), resulting in upper-ocean warming (∼0.4°C) and significant sea-ice loss in late summer (∼50

Leads are linear fractures formed by the deformation of the ice cover and are crucial areas for Arctic heat exchange between the ocean and the atmosphere. Accurately identifying leads is essential for studying climate change, polar ecosystems, and shipping navigation. Recent research on leads detection methods developed using synthetic aperture radar (SAR) tend to ignore the developmental stage of the leads and the distribution context of leads [first-year ice (FYI) or multiyear ice (MYI)], which hinders the generalizability of these methods. In this article, we propose a stacking approach for lead classification (SALC) integrating five specialized learners for five typical lead conditions. Ablation experiments reveal the effectiveness of each component of SALC: 1) the optimal preprocessing method is to input the incidence angle as a feature and apply the advanced thermal noise removal. In addition, the effect of inputting the incidence angle is significantly stronger than the incidence angle correction for original images; 2) the SALC performs best when the data features include polarization, texture, and upscaling features; 3) XGBoost is the most suitable meta classifier as compared to support vector machine (SVM) and random forest (RF). The results show that SALC outperforms both single learners and traditional classifiers across various conditions and quality criteria. In addition, SALC can tolerate at least 15% of erroneous samples.

The application of side-scan sonar in underwater target detection plays a significant role in marine engineering construction and ocean resource exploration. However, existing methods for small object detection often suffer from limited accuracy and robustness. To address this issue, we propose a feature super-resolution-based approach for detecting and segmenting small targets in side-scan sonar images. During network training, a feature super-resolution branch is first constructed. Both low-level and high-level features from the backbone of the You Only Look Once (YOLO) network are fed into this branch. After processing through an encoder-decoder architecture, a super-resolved feature map is reconstructed, and the network is optimized via backpropagation to enhance the ability to extract features of small targets. Furthermore, an exponential decay strategy is adopted to define the loss weights of different branches, establishing a branch-aware training mechanism to improve training effectiveness. During inference, the super-resolution branch is discarded to balance detection accuracy and inference efficiency. Experimental results demonstrate that the proposed method achieves superior performance in detecting and segmenting small-scale targets in side-scan sonar imagery, achieving state-of-the-art results on two public side-scan sonar small object datasets. Additionally, this approach can be extended as a training strategy for side-scan sonar target detection and segmentation networks. The source code is available at https://github.com/Yang-Code984/FSR_Sonar.

Poor feature representation, confusing background topography, and excessive data volume render detecting sparse targets in large-size acoustic imagery challenging. Especially when conducting real-time processing tasks, accuracy and speed are required to be optimized with limited computational resources. Therefore, this paper proposes an efficient method for real-time side-scan sonar (SSS) image processing and detection of sparse targets in large-scale images. Primarily, an intelligent real-time processing method is proposed for the raw SSS data to acquire high-quality SSS images. Aiming at the characteristics of large-size SSS images and sparse targets, we propose an innovative two-stage inference method: The SSS image slices are pre-classified based on the MobileViTv3-XXS model, and then the optimized detection model of RepVGG+YOLOv5m is employed for target detection of image slices containing targets. Experiments show that real-time preprocessing yields SSS images with an average PSNR of 27.112 and SSIM of 0.816, comparable to the post-processing methods. Meanwhile, it maintains high efficiency and achieves 88.2% mAP, significantly outperforming the slice-only method in detection accuracy and efficiency.