Continuous monitoring of glacier calving fronts is essential for sea level rise projections. This study benchmarks Deep Learning systems for front delineation in Synthetic Aperture Radar imagery. While Deep Learning systems exhibit errors up to 221 m, human annotators deviate by only 38 m, underscoring the need for further research.

Diagenetic snow facies represent distinct zones of snow and ice characterized by unique snow physical properties and attributes. These facies serve as indicators of changes in the surface mass balance of ice sheets, making them particularly relevant for monitoring the response of snow and firn cover to climatic changes. In this study, we propose a novel, fully unsupervised deep learning method based on convolutional neural networks (CNNs) to monitor snow zones on ice sheets using a decade of single-pass, bistatic interferometric synthetic aperture radar (InSAR) TanDEM-X data over Greenland. To do so, we develop an innovative iterative training approach to effectively manage a large variety of InSAR acquisition geometries with the goal of optimizing a robust, geometry-invariant model. The proposed approach achieves an average classification accuracy of 93.4% across varying acquisition geometries, segmenting the Greenland ice sheet into five distinct partitions. By analyzing these partitions in terms of elevation, snow density, and cumulative melt data, we link them to classical diagenetic snow facies and uncover trends correlated with climate events over the past decade, estimating a loss of over 454,000 km2 in dry snow extent on the Greenland ice sheet due to the 2012 extreme melt event, with a continuing trend of expanding percolation facies. The proposed method is adaptable to current and future single-pass SAR missions, such as the ESA 10th Earth Explorer Harmony mission, enhancing data robustness against varying acquisition geometries and demonstrating the potential of spaceborne single-pass InSAR missions for long-term monitoring of ice sheet dynamics.

Glaciers are losing ice mass at unprecedented rates, increasing the need for accurate, year-round monitoring to understand frontal ablation, particularly the factors driving the calving process. Deep learning models can extract calving front positions from Synthetic Aperture Radar imagery to track seasonal ice losses at the calving fronts of marine- and lake-terminating glaciers. The current state-of-the-art model relies on ImageNet-pretrained weights. However, they are suboptimal due to the domain shift between the natural images in ImageNet and the specialized characteristics of remote sensing imagery, in particular for Synthetic Aperture Radar imagery. To address this challenge, we propose two novel self-supervised multimodal pretraining techniques that leverage SSL4SAR, a new unlabeled dataset comprising 9,563 Sentinel-1 and 14 Sentinel-2 images of Arctic glaciers, with one optical image per glacier in the dataset. Additionally, we introduce a novel hybrid model architecture that combines a Swin Transformer encoder with a residual Convolutional Neural Network (CNN) decoder. When pretrained on SSL4SAR, this model achieves a mean distance error of 293m on the “CAlving Fronts and where to Find thEm” (CaFFe) benchmark dataset, outperforming the prior best model by 67 m. Evaluating an ensemble of the proposed model on a multi-annotator study of the benchmark dataset reveals a mean distance error of 75 m, approaching the human performance of 38 m. This advancement enables precise monitoring of seasonal changes in glacier calving fronts.

A crucial facet of ice sheet monitoring involves delineating distinct layers within the snowpack, known as snow facies, each characterized by unique physical properties. Variations in melt levels and snow properties across these facies exert influence on the radar wave penetration of spaceborne synthetic aperture radar (SAR) systems. This, in turn, affects the estimation of the radar mean phase center, a critical parameter for generating digital elevation models (DEMs), introducing penetration bias and leading to an underestimation of the surface topographic height. Accurate estimation of this bias is pivotal for reducing uncertainties in determining snow depth, ice thickness, and glacier mass balance through DEM differencing. In this paper, we explore the use of bistatic interferometric SAR features (InSAR) to monitor changes in the snow properties of the Greenland ice sheet (GIS), establishing links between these changes and anticipated variations in the radar penetration bias. We propose to combine machine learning-based models for snow facies segmentation and surface elevation bias estimation to achieve a more accurate estimation of the latter, while unveiling the importance of each feature for minimizing the prediction error. The surface elevation bias is estimated using a random forest baseline model based on TanDEM-X InSAR data and IceBridge laser-altimeter measurements acquired during the boreal winter season of 2010/11 in Greenland, achieving a coefficient of determination of R2= 84% and an RMSE of 0.70 m. Furthermore, we show that the derived snow facies are the most important feature for the final prediction.

Position changes of glacier calving fronts are important indicators for evaluating the health of ice sheet outlet glaciers and changes in ice dynamics. However, manual delineation of calving fronts in remote sensing imagery is a time-consuming task, resulting in potential large costs. Deep learning-based methods have made remarkable progress in automatically segmenting and delineating glacier calving fronts from remote sensing imagery. The relatively few remote sensing images and the limited geometric changes for glacier observations both reduce the diversity of the data and exacerbate the difficulty of accurate segmentation. Here we describe a novel automatic method for detecting glacier calving fronts in synthetic aperture radar (SAR) images, termed HookFormer. Our approach processes high-resolution (target) and low-resolution (context) inputs with a unified Transformer architecture. The global-local tokens from the context and the target branches are integrated purely by the proposed cross-attention mechanism and cross-interaction module to complement and enhance each other. Moreover, we redesign the HookFormer architecture based on the CNN model AMD-HookNet aiming to improve computational efficiency while achieving significant performance gains with only half of the model parameters/FLOPs. We conduct an in-depth analysis and make extensive comparisons based on the challenging glacier segmentation benchmark dataset CaFFe. As the first pure Transformer approach, HookFormer sets a new state of the art with a mean distance error of 353 m to the ground truth, outperforming the baseline, Swin-Unet, and AMD-HookNet by 53 %, 39 %, and 19 %, respectively.

Big advancements in the field of Deep Learning allow the automated extraction of glacier calving fronts from satellite imagery. However, current efforts on Synthetic Aperture Radar (SAR) imagery still produce coarse and partly spurious predictions. Therefore, in this study, a two-dimensional fully-connected Conditional Random Field (CRF) is incorporated into the post-processing of a deep learning-based calving front delineation pipeline. The CRF takes as input the artificial neural network prediction and the original SAR image. Experiments are undertaken using the newly introduced benchmark dataset CaFFe and results are compared to the associated baseline. By introducing the CRF into the post-processing of the baseline’s pipeline, the mean distance error of the front prediction is improved by an average of 27 meters. The code is available at https://github.com/EntChanelt/GlacierCRF.

Frontal dynamics of marine-terminating glaciers play a crucial role in glacier projections. To reduce manual effort, deep learning methods can be employed to extract calving front positions from satellite imagery automatically. The newly introduced benchmark dataset "CaFFe" (Calving fronts and where to find them: a benchmark dataset and methodology for automatic glacier calving front extraction from SAR imagery) [1] provides multi-mission Synthetic Aperture Radar (SAR) imagery along with manually annotated calving fronts. CaFFe establishes a standardized framework for comparing deep learning techniques in glacier calving front extraction. By utilizing CaFFe to benchmark forthcoming deep learning models, researchers can identify the most promising directions for future research. A leaderboard of models can be accessed at https://paperswithcode.com/sota/calving-front-delineation-in-synthetic.

Knowledge on changes in glacier calving front positions is important for assessing the status of glaciers. Remote sensing imagery provides the ideal database for monitoring calving front positions; however, it is not feasible to perform this task manually for all calving glaciers globally due to time constraints. Deep-learning-based methods have shown great potential for glacier calving front delineation from optical and radar satellite imagery. The calving front is represented as a single thin line between the ocean and the glacier, which makes the task vulnerable to inaccurate predictions. The limited availability of annotated glacier imagery leads to a lack of data diversity (not all possible combinations of different weather conditions, terminus shapes, sensors, etc. are present in the data), which exacerbates the difficulty of accurate segmentation. In this article, we propose attention-multihooking-deep-supervision HookNet (AMD-HookNet), a novel glacier calving front segmentation framework for synthetic aperture radar (SAR) images. The proposed method aims to enhance the feature representation capability through multiple information interactions between low-resolution and high-resolution inputs based on a two-branch U-Net. The attention mechanism, integrated into the two branch U-Net, aims to interact between the corresponding coarse and fine-grained feature maps. This allows the network to automatically adjust feature relationships, resulting in accurate pixel classification predictions. Extensive experiments and comparisons on the challenging glacier segmentation benchmark dataset CaFFe show that our AMD-HookNet achieves a mean distance error (MDE) of 438 m to the ground truth outperforming the current state of the art by 42%, which validates its effectiveness.

Glacier calving front position (CFP) is an important glaciological variable. Traditionally, delineating the CFPs has been carried out manually, which was subjective, tedious, and expensive. Automating this process is crucial for continuously monitoring the evolution and status of glaciers. Recently, deep learning approaches have been investigated for this application. However, the current methods get challenged by a severe class imbalance problem. In this work, we propose to mitigate the class imbalance between the calving front class and the noncalving front class by reformulating the segmentation problem into a pixelwise regression task. A convolutional neural network (CNN) gets optimized to predict the distance values to the glacier front for each pixel in the image. The resulting distance map localizes the CFP and is further postprocessed to extract the calving front line. We propose three postprocessing methods, one method based on statistical thresholding, a second method based on conditional random fields (CRFs), and finally the use of a second U-Net. The experimental results confirm that our approach significantly outperforms the state-of-the-art methods and produces accurate delineation. The second U-Net obtains the best performance results, resulting in an average improvement of about 21% Dice coefficient enhancement.

The vast majority of the outlet glaciers and ice streams of the polar ice sheets end in the ocean. Ice mass loss via calving of the glaciers into the ocean has increased over the last few decades. Information on the temporal variability of the calving front position provides fundamental information on the state of the glacier and ice stream, which can be exploited as calibration and validation data to enhance ice dynamics modeling. To identify the calving front position automatically, deep neural network-based semantic segmentation pipelines can be used to delineate the acquired SAR imagery. However, the extreme class imbalance is highly challenging for the accurate calving front segmentation in these images. Therefore, we propose the use of the Mathews correlation coefficient (MCC) as an early stopping criterion because of its symmetrical properties and its invariance towards class imbalance. Moreover, we propose an improvement to the distance map-based binary cross-entropy (BCE) loss function. The distance map adds context to the loss function about the important regions for segmentation and helps accounting for the imbalanced data. Using Mathews correlation coefficient as early stopping demonstrates an average 15% dice coefficient improvement compared to the commonly used BCE. The modified distance map loss further improves the segmentation performance by another 2%. These results are encouraging as they support the effectiveness of the proposed methods for segmentation problems suffering from extreme class imbalances.

Supervised machine learning requires a large amount of labeled data to achieve proper test results. However, generating accurately labeled segmentation maps on remote sensing imagery, including images from synthetic aperture radar (SAR), is tedious and highly subjective. In this work, we propose to alleviate the issue of limited training data by generating synthetic SAR images with the pix2pix algorithm [1]. This algorithm uses conditional Generative Adversarial Networks (cGANs) to generate an artificial image while preserving the structure of the input. In our case, the input is a segmentation mask, from which a corresponding synthetic SAR image is generated. We present different models, perform a comparative study and demonstrate that this approach synthesizes convincing glaciers in SAR images with promising qualitative and quantitative results.

Fluctuations of the glacier calving front have an important influence over the ice flow of whole glacier systems. It is therefore important to precisely monitor the position of the calving front. However, the manual delineation of SAR images is a difficult, laborious and subjective task. Convolutional neural networks have previously shown promising results in automating the glacier segmentation in SAR images, making them desirable for further exploration of their possibilities. In this work, we propose to compute uncertainty and use it in an Uncertainty Optimization regime as a novel two-stage process. By using dropout as a random sampling layer in a U-Net architecture, we create a probabilistic Bayesian Neural Network. With several forward passes we create a sampling distribution, which can estimate the model uncertainty for each pixel in the segmentation mask. The additional uncertainty map information can serve as a guideline for the experts in the manual annotation of the data. Furthermore, feeding the uncertainty map to the network leads to 95.24 % Dice similarity, which is an overall improvement in the segmentation performance compared to the state-of-the-art deterministic U-Net-based glacier segmentation pipelines.

An essential climate variable to determine the tidewater glacier status is the location of the calving front position and the separation of seasonal variability from long-term trends. Previous studies have proposed deep learning-based methods to semi-automatically delineate the calving fronts of tidewater glaciers. They used U-Net to segment the ice and non-ice regions and extracted the calving fronts in a post-processing step. In this work, we show a method to segment the glacier calving fronts from SAR images in an end-to-end fashion using Attention U-Net. The main objective is to investigate the attention mechanism in this application. Adding attention modules to the state-of-the-art U - N et network lets us analyze the learning process by extracting its attention maps. We use these maps as a tool to search for proper hyperparameters and loss functions in order to generate higher qualitative results. Our proposed attention U-Net performs comparably to the standard U-Net while providing additional insight into those regions on which the network learned to focus more. In the best case, the attention U-Net achieves a 1.5 % better Dice score compared to the canonical U-Net with a glacier front line prediction certainty of up to 237.12 meters.

The disintegration of various ice shelves on the Antarctic Peninsula has demonstrated their vulnerability and impacts on tributary glaciers. A satellite image of Wilkins Ice Shelf (WIS) from July of 2007 reveals the formation of a large new double fracture, accompanied by numerous small fractures. We show that bending stresses induced by buoyancy forces were responsible for fracture formation. On February 28-29, 2008, an area of about 425 km2broke up at a narrow connection of the WIS to one of its confining islands. In contrast to Larsen B Ice Shelf, melt ponds that drain into crevasses played no role in this breakup process. A further breakup of 160 km2in the same area occurred on May 30-31, 2008 and documented that breakup can occur during austral winter. Radar images reveal a frozen surface, which demonstrates that in this breakup, surface melt water did not play a role. We conclude that ice shelves with strong thickness contrasts carry potential for disintegration. The fact that the WIS experienced two breakup events under two widely contrasting surface conditions (one during the melt season and one during winter) reveals that there may be several reasons for the disintegration of ice shelves that operate under differing circumstances.

For Central Europe urbanization and urban sprawl are the major processes concerning land-use and land-cover change that alter the characteristic and functioning of the landscape permanently. Against the background of the current sustainable development debate there is an increasing demand of reliable information about the future trends of these land-use developments. Remote Sensing provides the required data and methods for an operational monitoring of land-use/-cover changes. However for the prediction of future trends additional information and a deep understanding of the driving forces of land-use/-cover change is needed. Here we show an exemplary pathway for obtaining the required future land-use information using accurate and stable techniques. We acquired historical land-use information using Landsat imagery. The measured changes (e.g. an increase of urban areas of about 37% between 1984 and 2005) are driven by a variety of socio-economic causes that have been evaluated statistically. We used this information to calibrate a spatial predictive model that has been implemented in the generic modeling framework XULU (eXtendable Unified LandUse Modeling Platform).

The German federal state of North Rhine-Westphalia (NRW) has experienced considerable land-use changes during the last decades. One key element of these changes is the increase in impervious surface that is driven by factors like suburbanization, extension of urban infrastructure, development of new industrial and commercial areas, village expansion, etc. The changes in the land-use have been detected with Landsat satellite data for a time period covering the last 30 years. For 4 steps (1975, 1984, 2001, and 2005) specific land-use maps have been generated. In this study we concentrated only on the increase in impervious areas, although other forms of land-use change clearly also appeared in the study area. To measure driving forces of urban sprawl in NRW logistic regression models have been developed in the spatial domain. Therefore probability maps were generated for the land-use configuration in 1984 and for the increase in impervious surface for the time span 1984-2001. For the initial land-use configuration in 1984 a statistical relationship between biogeophysical conditions, distance measures, and the appearance of impervious area could be established, while it was difficult to predict exact locations of urban sprawl between 1984 and 2001 with the explanatory factors used in this study. The statistical models were tested with the relative operating characteristic (ROC) as a measure for the goodness of fit.

The study investigates the possibilities to improve the performance of CERES-Wheat crop growth model by assimilating information derived by optical and SAR Earth observation data. Biophysical parameter retrieval was done with the water cloud model for SAR data and the CLAIR model was applied to multispectral imagery. The CERES -Wheat model was calibrated using ground truth information. The re-initialization method with an adjustable planting date was selected as assimilation strategy. Modelling results generally improved by using all different kind of remote sensing data. However, best results were achieved by using information of the optical sensors only and not by a synergetic time series of all available data.

In the proposed approach, the well-known enhanced Lee filter is modified to allow the integration of feature outlines-previously extracted from segmented optical images. The filter is applied to several ENVISAT ASAR images that cover urban, agricultural, and forest areas during different plant phenological stages. The performance of this segment-based speckle filter is compared to those of other filters using ratio images, visual interpretation, and statistical indexes. The approach reduces the loss of radiometry and spatial information. It performs comparable to more complex methods and outperforms common techniques

The Antarctic Peninsula is a region highly sensitive to climate change. For ice dynamic modelling and climatic impact studies several parameters have to be known. Radar interferometry is a feasible method in such remote areas to derive velocity fields. For King George Island, the available ERS-1/2 tandem pairs with reasonable baselines have been processed. Due to temporal decorrelation only 2 out of 19 pairs showed almost everywhere high coherence and an additional 6 have partial moderate coherence. A SAR DEM was constructed by double differencing interferometry. Ascending and descending scenes were combined to derive a velocity field for the main part of the Island. External elevation data were additionally incorporated in the processing. Glacier velocities range between 0 m and about 120 m per year. Validation with DGPS measurements from several field campaigns shows good agreement with the interferometric derived velocities. Inaccuracies exist for the edges of the ice cap as here decorrelation effects hampered the generation of an InSAR DEM and no external data were available.

The accuracy of supervised land cover classifications depends on variables like the chosen algorithm, adequate training data and the selection of features. It has been shown that classification results can be improved by classifier ensembles. In the present study decision trees have been generated with random selections of all available features and combined into such a multiple classifier. The influence of the number of selected features and the size of the multiple classifiers on classification accuracy is investigated using a set of 14 SAR images. Results of multiple classifiers are always better than those of a decision tree based on all available features. Maximum accuracies were achieved with multiple classifiers that use decision trees based on 70% of the available features. The visual inspection of produced maps underlines the high quality of the results. The area is classified into homogeneous fields with little noise, only.

In this paper, radiometric normalisation of multitemporal Landsat data is developed. A comparison of more classic methods with a Multivariate Alteration Detection (MAD) based transformation and an ordinary least square regression of no change pixels is presented.