Deep learning (DL)-based fully supervised approaches have demonstrated remarkable performance in sea ice classification, showcasing their potential for highly accurate results. However, their reliance on high-resolution labels poses a formidable challenge, as obtaining such data can be a difficult task. In contrast, our method based on weakly supervised learning excels by operating with lower-resolution polygon labels while still achieving outstanding performance. This approach enables precise pixel-level classification of ice stage of development (SOD) by learning from region-based labels embedded within expert-annotated ice charts. During training, region-based loss functions are introduced to quantify the disparity between predicted tensors describing SOD distributions and label tensors derived from ice charts. We leverage the AI4Arctic Sea Ice Challenge Dataset, comprising over 500 Sentinel-1 synthetic aperture radar (SAR) images, ancillary multisource data, and corresponding ice charts, for model training and evaluation. Visual interpretation and numerical analysis reveal that our weakly supervised method outperforms the fully supervised U-Net benchmark. It yields more accurate SOD predictions, significantly enhancing mapping resolution and class-wise accuracy. This methodology marks a critical step forward in the quest for automated operational sea ice mapping.

The space environment has become more congested and competitive than ever before, making collision-free satellite operation a top priority for both public and private sectors. Specifically, effective collision avoidance involving Residence Space Objects (RSOs) stands out as a critical challenge within the domain of Space Situational Awareness (SSA). This paper addresses this issue by exploring the application of Reinforcement Learning (RL) methods, which shows promising results in addressing the complexities of satellite path planning and collision avoidance. This paper focuses on single-object maneuver planning, where the primary satellite aims to avoid collisions with secondary objects using the Proximal Policy Optimization (PPO) algorithm. The obtained results demonstrate the promising performance of the trained model, establishing a proof of concept for further investigation of RL for maneuver planning and collision avoidance in space.

Fully supervised deep learning approaches have demonstrated impressive accuracy in sea ice classification, but their dependence on high-resolution labels presents a significant challenge due to the difficulty of obtaining such data. In response, our weakly supervised learning method provides a compelling alternative by utilizing lower-resolution regional labels from expert-annotated ice charts. This approach achieves exceptional pixel-level classification accuracy by introducing regional loss representations during training to measure the disparity between predicted and ice chart-derived sea ice type distributions. Leveraging the AI4Arctic Sea Ice Challenge Dataset, our method outperforms the fully supervised U-Net benchmark in mapping resolution and class-wise accuracy, marking a significant advancement in automated operational sea ice mapping.

High-resolution sea ice concentration (SIC) maps are critical to support various applications, e.g., climate modeling, ship navigation, and activities in Northern communities. However, operational mapping of SIC based on expert annotations is coarse in spatial resolution and time-consuming to prepare. Although many convolutional neural network (CNN)-based methods have been proposed for automated sea ice mapping from synthetic aperture radar (SAR) imagery in recent years, the lack of pixel-based labels for model training hinders them from producing high-resolution reliable mapping results. To overcome this challenge, this letter presents a novel weakly supervised learning approach that generates pixel-level SIC prediction using coarse region/polygon-level SIC ground truth. Specifically, a novel region-level loss function is designed to enable direct use of regional/polygon SIC values in ice charts for the training of a U-Net-based model. This avoids the errors in transferring region-level SIC values to pixel-level ground-truth SIC values effectively and allows the generation of pixel-level SIC and sea ice extent (SIE) estimates. The proposed approach is evaluated on the recently published AI4Arctic Sea Ice Challenge Dataset with over 500 Sentinel-1 SAR scenes, ancillary data, and associated ice charts. The results demonstrate the effectiveness of the weakly supervised model in producing pixel-level high-resolution SIC maps that are consistent with ice charts and visual interpretation.

This study investigates the relationship between gray-level co-occurrence matrix (GLCM) texture features and synthetic aperture radar (SAR) incidence angle (IA) for sea ice classification. We analyzed dual polarization RADARSAT-2 C-band SAR data comprising 29 scenes. GLCM features were extracted from the radar cross-section (σo) in dB and categorized by sea ice class. To assess IA dependence per sea ice class, we used linear interpolation and the coefficient of determination (R2). We evaluated separability among ice classes using the Jeffries–Matusita distance and confirmed the improved separability with a Bayesian classifier. The results reveal a significant IA dependence of GLCM features. Notably, GLCM features from the HV band display stronger IA dependence and higher separability among ice classes compared to those from the HH band. These findings emphasize the significance of considering IA in the utilization of GLCM features for sea ice classification.

This paper presents an investigation into the use of a heteroscedastic Bayesian neural network (HBNN) for predicting sea ice concentration (SIC) using both passive microwave (PM) and atmospheric data. The primary objective is to provide accurate estimates for downstream services that require uncertainty estimates. To achieve this, HBNNs are implemented using a multilayer perceptron (MLP) architecture with methods for uncertainty quantification based on the Bayes by backprop (BBB) algorithm and a heteroscedastic loss function. The models are trained and tested using data collected from the Eastern Arctic regions. The results of numerical analysis demonstrate that the HBNNs are able to significantly reduce estimation error compared to deterministic NNs. The study also investigates the spatial and seasonal variation of uncertainty in detail.

This paper investigates the influence of input image scale on deep learning-based Beluga whale detection from aerial remote sensing imagery. Beluga whales in the Arctic are jeopardized due to increased coastal activities and climate change. Aerial survey is a common population counting method, and it can be laborious and exhausting to count the number of whales manually. Convolutional neural networks (CNNs) have greatly improved the performance of detecting and counting whales. Since most remote sensing images are very high in resolution, it is a common practice to slice the image into small patches. In this work, we input the full image (after resizing) into an object detection model and compare its performance with the sliding window approach. Experimental results suggest that increasing the input image size helps improve the model’s performance, and the model is able to learn the contextual information.

Bayesian neural networks (BNNs) have been demonstrated to be effective in accurate retrieval of sea ice concentration (SIC) from multi-source data, while providing estimates of uncertainty, which are essential for downstream services. However, uncertainty obtained by BNNs are intrinsically uncalibrated, which indicates that it may not correlate well with model error. To address this issue, we investigate a new approach that combines an auxiliary prediction interval (PI) estimator with the BNN-based SIC mean estimator to develop a well-calibrated SIC retrieval model that is both accurate and reliable. We adopt a training strategy called “uncertainty matching" to train the model, which ensures that the estimated uncertainties match the estimated PIs. We use a subset of AMSR2 brightness temperature data and ERA5 atmospheric data collected from 2014 to 2015 in the Baffin Bay area as input features of the model. Comparison between model inference and SIC labels obtained from the enhanced NASA Team (NT2) algorithm shows that the proposed approach is able to produce well-calibrated uncertainty with more accurate predictions in marginal ice zones.

Algorithms designed for ice–water classification of synthetic aperture radar (SAR) sea ice imagery produce only binary (ice and water) output typically using manually labeled samples for assessment. This is limiting because only a small subset of labeled samples are used, which, given the nonstationary nature of the ice and water classes, will likely not reflect the full scene. To address this, we implement a binary ice–water classification in a more informative manner considering the uncertainty associated with each pixel in the scene. To accomplish this, we have implemented a Bayesian convolutional neural network (CNN) with variational inference to produce both aleatoric (data-based) and epistemic (model-based) uncertainty. This valuable information provides feedback as to regions that have pixels more likely to be misclassified and provides improved scene interpretation. Testing was performed on a set of 21 RADARSAT-2 dual-polarization SAR scenes covering a region in the Beaufort Sea captured regularly from April to December. The model is validated by demonstrating: 1) a positive correlation between misclassification rate and model uncertainty and 2) a higher uncertainty during the melt and freeze-up transition periods, which are more challenging to classify. By incorporating the iterative region growing with semantics (IRGS) segmentation algorithm and an uncertainty value-based thresholding algorithm, the Bayesian CNN classification outputs are improved significantly via both numerical analysis and visual inspection.

In recent years, the adoption of deep learning (DL) techniques for predicting sea ice concentration (SIC) given both passive microwave (PM) data and reanalysis data has seen a growing interest. For use in downstream services, these SIC estimates should be accompanied by uncertainty estimates. To provide these estimates, we utilize a heteroscedastic Bayesian neural network (HBNN), which can estimate both model (epistemic) and data (aleatoric) uncertainty. We use both PM and atmospheric data as our input features and demonstrate that both are needed for accurate SIC estimates. Results show that, over an annual cycle, the months of melt onset, such as April, May, and June, produce the highest uncertainties relative to other months, with total (epistemic + aleatoric) uncertainties of approximately 20%, while areas in the marginal ice zone contributed highest total uncertainty of 25% spatially. When considering an average over the test year, the level of uncertainty due to the data (aleatoric) is consistent with other studies, at 10%–15%. The advantage of our approach is that the uncertainties are specific to the data instance, and both model and data uncertainties are estimated.

In this paper, sea ice concentration along the east coast of Canada is predicted using a multilayer perceptron regression model, which has the AMSR-E passive microwave 36.5 GHz brightness temperatures and numerical weather prediction data as its inputs. Derived sea ice concentrations are evaluated against those from the NT2 algorithm, ASI algorithm, ESA algorithm, and CIS ice chart. Results show that the multilayer perceptron regression model estimates sea ice concentration in the marginal ice zones with a reasonably low bias in comparison to CIS ice charts and, in general, its capability of estimating the SIC is comparable to that of the other models.

A method to extend categorical triple collocation (CTC) to five data sets, three of which must have conditionally independent errors, is presented. The method is shown to enable comparison of three passive microwave sea ice concentration data sets, and can easily be extended to for use with more data sets. The method is evaluated in the Gulf of Saint Lawrence, on the east coast of Canada, during the freeze-up period. It is found that the ice concentration data set from the European space agency sea ice climate change initiative (ESA-SICCI) and the NASA Team 2 (NT2) algorithms have higher balanced accuracy than that from the Artist sea ice (ASI) algorithm.

Synthetic aperture radar (SAR) sea ice imagery is a promising source of data for sea ice data assimilation. Classification of SAR sea ice imagery into ice and water is of particular relevance due to its relationship with ice concentration, a key variable in sea ice data assimilation systems. With increasing volumes of SAR data, automated methods to carry out these classifications are of particular importance. Although several automated approaches have been proposed, none look at the impact of including an estimate of uncertainty of the model parameters and input features on the classification output. This article uses an established database of SAR image features to train a multilayer perceptron (MLP) neural network to classify pixel locations as either ice, water, or unknown. The classification accuracies are benchmarked using a recently developed logistic regression approach for the same database. The two methods are found to be comparable. The MLP approach is then enhanced to allow uncertainty to be estimated at each pixel location. Following methods proposed in the deep learning community, two kinds of uncertainty are considered. The first, epistemic uncertainty, is that due to uncertainty in the MLP weights. The second kind of uncertainty, aleatoric uncertainty, is that which cannot be explained by the model, and is therefore associated with the input data. It is found that including these uncertainties in the MLP models reduces their accuracies slightly, but also reduces misclassification rates. This is of particular importance for data assimilation applications, where misclassifications could severely degrade the analysis.

Snow-depth retrieval from passive microwave observations without a priori information is a highly undetermined problem. Achieving accurate snow-depth retrievals requires a priori information on the snowpack properties, such as grain size, density, physical temperature, and stratigraphy. On a practical level, however, retrieval algorithms must consider prior information, while minimizing the dependence on it, as accurate ancillary data are not globally available. In this study, we build on the previously published Bayesian Algorithm for Snow Water Equivalent Estimation (BASE) to retrieve snow depth using an airborne passive microwave data set over the tundra snow in the Eureka region. The method computes the optimal estimates of snow depth, density, grain size, and other variables, given the brightness temperature observations and prior information, using Markov chain Monte Carlo (MCMC). The airborne data set includes passive microwave brightness temperature (Tb) at 18.7 and 36.5 GHz. The in situ measurements of the snow depth provide validation data for 464 sensor footprints. The microwave radiative transfer (RT) model used is the Dense Media RT-Multilayered (DMRT-ML) model. We use a two-layer wind slab and depth hoar assumption based on the local snow cover knowledge from the previous research on the study area. To improve our understanding of the results using the airborne Tbs, the inversion was also applied using the synthetic observations, where Tbs were generated from the RT model. For the case with synthetic observations, the snow-depth RMSE was 0.07 cm. When the airborne Tbs are used, the snow-depth RMSE was 21.8 cm. This discrepancy is due to the large spatial variability in the MagnaProbe snow-depth measurements and the fact that not all physical processes affecting the airborne Tbs are represented in the RT model. Our work verifies the feasibility and applicability of the proposed methodology regionally for the airborne retrievals and reinforces the tractable applicability of a physics-based RT model in the SWE retrievals.

Shipping traffic has grown steadily in the Arctic in recent years. One of the reasons for this increased traffic is the lengthening of open water season, which is accompanied by increases in the area covered by intermediate ice concentrations' or marginal ice zones (MIZs). These regions are difficult to detect using passive microwave data. In this paper, we propose the use of deep learning for automatic detection of MIZs in the RADARSAT-2 satellite images. A synthetic aperture radar (SAR) dataset is manually annotated to train, test and refine the method. Various convolutional neural network (CNN) models are evaluated as fixed feature extractors for the task of classification. To aid the classification accuracy we use a weighted binary cross-entropy loss criterion. Finally, to refine the segmentation process, we used a multi-scale patch technique. The analysis of the results demonstrates that CNN model predictions obtained with multiple sizes of spatial windows is able to detect MIZs in SAR images.

Sea ice ridging presents significant danger to ships navigating Hudson Strait in the winter months. Every winter, ships spend a significant portion of their time beset in sea ice ridges and pressured ice, hindering the advancement of resource development in the Arctic. Manually identifying ridges in satellite imagery is a very laborious process. Combining an extensive Synthetic Aperture Radar (SAR) data set of Hudson Strait with labelled ridge locations spanning nine winters, an automated system can be developed to detect ridging. In this study we test three transfer learning approaches using a popular convolutional neural network (CNN) architecture; DenseNet-161. We find that when some layers of DenseNet-161 are unfrozen, the model can determine if an area is ridged or non-ridged with a receiver operating characteristic curve (ROC-AUC) score of 92.3. This approach outperforms previous work where statistical feature based classification was used to predict ridging on the same data set.

Historically, sea ice concentration (SIC) has been measured through the use of passive microwave sensors, as well as human interpretation of synthetic aperture radar (SAR). Although passive microwave data are processed automatically, it suffers from poor spatial resolution and the higher frequency channels are sensitive to weather conditions. Deep learning has demonstrated its ability to perform complex and accurate analysis of images; here, we apply deep learning to estimate ice concentration from SAR scenes. We developed a deep convolutional neural network (CNN) that predicts SIC from SAR, trained upon passive microwave data. The model achieves a 5.24%/7.87% error on its train and test set, respectively. To assess the real-world applicability, we performed an independent validation on 18 SAR scenes (from two distinct geographical regions), not previously seen during training or test. Comparing against human-generated ice analysis charts, we achieved an L1 error of 0.2059, competitive with passive microwave (EL1= 0.1863) for the Canadian Arctic Archipelago. For the Gulf of Saint Lawrence region, we achieved an L1 error of 0.2653, significantly better than the passive microwave result (EL1= 0.3593). By using novel techniques for model training, as well as training entirely upon passive microwave data, we present an accessible and robust method of developing similar systems for processing SAR.1Our results suggest that with further postprocessing, CNNs are accurate and robust enough to be used for operational tasks.

Monitoring the sea ice cover is important for both climate studies and ice operations, such as shipping. It is challenging to validate even basic essential variables, such as the sea ice extent, due to a lack of appropriate validation data. Instead of focusing on validation, this paper looks at the use of categorical triple collocation (CTC) for the task of quantitatively comparing three colocated data sets. CTC has been developed and used in earlier studies to rank binary data sets. In this paper, we extend earlier studies and bring in recent results from the binary classification community to estimate the class imbalance (the relative proportion of each class, ice or water). We then use this class imbalance to obtain quantitative estimates of the proportion correct of ice (sensitivity) and the proportion correct of water (specificity). The methodology is first tested using toy data, after which three data sets from the Gulf of Saint Lawrence, on the east coast of Canada, are used. These data sets are from an ice-ocean model, a passive microwave sea ice concentration retrieval, and a sea ice concentration retrieval from synthetic aperture radar (SAR). By looking at both the sensitivity and the specificity, it is found that the passive microwave data have difficulty in recognizing ice during freeze-up, but they perform well at obtaining the correct water observations. This distinction cannot be made by ranking the data sets. The CTC method is compared with, and found to be complementary to, a validation using ice/water states from the interactive multisensor snow and ice mapping system.

We propose a fully convolutional neural network (FCNN) model for ice concentration estimation from dual-polarized SAR images. Our network contains 5 convolutional layers. Tested in the Gulf of Saint Lawrence during freeze-up, the proposed model is demonstrated to generate improved ice concentration estimates compared to a CNNs with similar structure.

Mapping sea ice and open water in the oceans is significant for many applications. Accurate and robust classification methods of sea ice and open water are in demand by ice services. Convolutional neural networks (CNNs) are becoming increasingly popular in many research communities due to availability of large image datasets and high-performance computing systems. As Convolutional networks (ConvNets) have achieved great success on many image classification tasks, we pursue this method for the classification of image patches from synthetic aperture radar (SAR) imagery into ice and water. In this study we use image patches with three dimensions (HH polarization, HV polarization, and incidence angle) in a transfer learning method with CNNs: extracting features of the patches from AlexNet and applying a softmax classifier. Our method achieves an overall classification accuracy of 92.36% based on the held-out test data.

Observations of ice thickness are limited in high latitude regions, at a time when they are increasingly being requested by operational ice centers. This study aims to improve the retrieval of lake ice thickness using data from the Moderate Resolution Imaging Spectroradiometer (MODIS) on board NASA's Aqua (P.M.) and Terra (A.M.) satellites. The accuracy of ice thickness retrievals based on MODIS lake ice surface temperature (LIST) is investigated using a commonly used heat balance equation and the retrieved ice thicknesses are compared to in situ measurements from the Canadian Ice Service. The accuracy of ice thickness estimates is improved when using snow depth from the 1-D thermodynamic lake ice model Canadian Lake Ice Model (CLIMo) rather than an empirical relationship between snow depth and ice thickness utilized in the recent investigations. Taking into account all data over the study period (2002-2014), the mean bias error and the root-mean-square error are reduced from -0.42 to 0.07 m and 0.58 to 0.17 m, respectively, with the novel approach proposed herein. However, this approach is limited to ice thickness estimations of less than ca. 1.7 m.

High-resolution ice concentration maps are of great interest for ship navigation and ice hazard forecasting. In this case study, a convolutional neural network (CNN) has been used to estimate ice concentration using synthetic aperture radar (SAR) scenes captured during the melt season. These dual-pol RADARSAT-2 satellite images are used as input, and the ice concentration is the direct output from the CNN. With no feature extraction or segmentation postprocessing, the absolute mean errors of the generated ice concentration maps are less than 10% on average when compared with manual interpretation of the ice state by ice experts. The CNN is demonstrated to produce ice concentration maps with more detail than produced operationally. Reasonable ice concentration estimations are made in melt regions and in regions of low ice concentration.

Synthetic Aperture Radar (SAR) images of sea ice have proven to be very useful toward classification of ice cover into ice types. However, using SAR images to separate the marginal ice zone (MIZ) from consolidated ice and open water has not been explicitly considered before. One typical feature of MIZ is that it is more dynamic than consolidated ice, and includes floes, fast and thin ice or ice eddies. The current paper utilizes the dynamic feature of MIZ to investigate a curvelet-based feature extraction method in order to classify a SAR image into open water, dynamic ice and consolidated ice, as a first step toward using SAR imagery to identify the MIZ. An experiment of 10-fold cross validation is conducted to demonstrate that the proposed feature extraction method is effective. Finally, an SVM classifier is used on a SAR image to test the performance of the curvelet-based feature. The result shows that curvelet-based feature can classify the dynamic ice accurately.

Accurate and detailed information of sea ice is crucial for navigation in ice-covered waters. Synthetic Aperture Radar (SAR) has been used as an effective tool in remote sensing to collect information for estimating the sea ice state. Image analysis experts produce manual image analysis charts based on their interpretation of the SAR imagery. Since the manual analysis is very time-consuming automatic methods can be used to discriminate ice-ocean and accelerate the task. Image texture features from Grey Level Co-occurrence Matrix (GLCM) are used as a source of information for ice-water discrimination. In this work an ice-concentration from RIPS is used as the background state which is updated using probabilities of ice and water calculated using GLCM texture features from the SAR image.

A two-layer feed forward neural network is used to estimate ice concentration from SAR images directly in this research. SAR image patches are used as input. The CIS (Canadian Ice Service) ice concentration image analyses are used to train the neural network. The experiment shows that the simple neural network can be used to generate a reasonable ice concentration with no preprocessing to the SAR images.

In this paper, sea-ice thickness values are calculated along the Labrador coast using data from two sensors representative of those available for operational data assimilation. Data from the first sensor, the Moderate Resolution Imaging Spectroradiometer (MODIS), are used to calculate the ice thickness using a heat balance equation. Relationships between the MODIS ice thickness and polarization ratio from the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) are used to calculate the thickness of thin ice (less than 0.2 m) from the AMSR-E data. This is done for each frequency on the AMSR-E sensor in the range of 6.9-36.5 GHz. Through comparison with data from ice charts, it is found that the errors are lowest for thickness values calculated from low-frequency AMSR-E data. The accuracies of the ice thickness from MODIS, AMSR-E, operational ice charts, and two moored upward looking sonars are further assessed using the triple collocation method. It is found that the error associated with ice thickness from AMSR-E is the lowest and the error associated with ice thickness from MODIS is the highest. While the MODIS data represent the small-scale variability of the sea-ice thickness better than the AMSR-E data, the MODIS data can produce spurious values of ice thickness due to unmasked clouds. To use ice thickness from MODIS in an automated algorithm, quality control would need to be applied to the MODIS data to remove unmasked clouds which lead to spurious values of thick ice. The errors calculated for the ice thickness from AMSR-E, which are calculated based on a relationship calibrated with MODIS ice thickness from a clear-sky day, indicate that these data would be useful for operational data assimilation.

Estimates of sea ice thickness are important for shipping and weather forecasting applications. Sea ice thickness can be estimated using data from the thermal channels on the Moderate Resolution Imaging Spectroradiometer (MODIS). However, using this data for studies of surface conditions is significantly hampered by cloud cover. This is particularly problematic for studies of the marginal ice zone, where atmospheric conditions often lead to persistent cloudy conditions. In this study a new method is proposed in which data from a passive microwave sensor is used to guide the estimation of surface temperature in cloud-covered regions. The impact of the method is verified by checking sea ice thickness values calculated using the guided surface temperature against values from operational sea ice charts.