Abstract:Crevasses are formed by glacier movement and the stresses within glacier ice. Knowledge of the crevasses’ distribution is critical for understanding the glacier and ice shelf stability. In this study, we propose an automated crevasse extraction framework based on Sentinel-1 SAR imagery and an improved U-Net network. The spatial distribution of crevasses on Antarctic ice shelves in 2020 was mapped with a spatial resolution of ~40 m, and the characteristics of crevasses on the Nickerson Ice Shelf, Jelbart Ice Shelf, Amery Ice Shelf, Thwaites Glacier, and Shackleton Ice Shelf were analyzed. The results indicated the extraction accuracy of our method was 84.2% and the F1 score was 72.5%. Compared with previous published studies, the identification of the crevasse areas had good visual consistency. However, in some scenes, the recall rate was relatively lower due to the quality of the SAR image, terrain surrounding the crevasses, and observation geometry. The crevasses on different ice shelves had different characteristics in terms of length, density, type, and spatial pattern, implying the different stress structures of ice shelves. The Thwaites Glacier and the Nickerson Ice Shelf in the West Antarctica Ice Sheet (WAIS) had shorter ice crevasses, whereas the lengths of ice crevasses on the Jelbart Ice Shelf and the Amery Ice Shelf in the East Antarctica Ice Sheet (EAIS) were relatively long. Nevertheless, there are more closely spaced crevasses on the ice shelf in WAIS compared to that in the EAIS. For the distribution of crevasse types, the Nickerson Ice Shelf and the Shackleton Ice Shelf had various forms of crevasses. There were mainly transverse crevasses developed on the Jelbart Ice Shelf and the Amery Ice Shelf. This study provides a helpful reference and guidance for automated crevasse extraction. The method proposed by this study manifests great application potential and the efficacy of producing a time-series crevasse data set with higher spatial resolution and larger coverage. In the future, more Sentinel-1 SAR imagery will be applied and the effect of temporal and spatial variations in crevasses on the stability of ice shelves will be investigated, which will contribute to project the ice shelf stability and explore the sea level rise implications of recent and future cryosphere changes.

Three-dimensional dynamic monitoring of crevasses based on deep learning and surface elevation reconstruction methods

Detection of Surface Crevasses over Antarctic Ice Shelves Using SAR Imagery and Deep Learning Method

Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imagery

Automatic fracture detection based on Terrestrial Laser Scanning data: A new method and case study

CREVNet: A Transformer and CNN-Based Network for Accurate Segmentation of Ice Shelf Crevasses

Ice plate deformation and cracking revealed by an in situ-distributed acoustic sensing array

A Novel Deep Learning-Based Approach for Rift and Iceberg Recognition From ICESat-2 Data

Efficient Location and Extraction of the Iceberg Calved Areas of the Antarctic Ice Shelves

Deep learning reveals seasonal patterns of Antarctic ice shelf front fluctuations

The distribution and evolution of surface fractures on pan-Antarctic ice shelves

Automated surface melt detection over the Antarctic from Sentinel-1 imagery using deep learning

3D vision technologies for a self-developed structural external crack damage recognition robot

Environment‐Modulated Glacial Seismicity Near Dålk Glacier in East Antarctica Revealed by Deep Clustering

UAV-Based Photogrammetry and LiDAR for the Characterization of Ice Morphology Evolution

Extraction and Analysis of the Antarctic Ice Shelf Basal Channel

AscDAMs: Advanced SLAM-based channel detection and mapping system

Automatic crack detection and 3D reconstruction of structural appearance using underwater wall-climbing robot

Multiple Defects Inspection of Dam Spillway Surface Using Deep Learning and 3D Reconstruction Techniques

Segmentation of Glacier Area Using U-Net through Landsat Satellite Imagery for Quantification of Glacier Recession and Its Impact on Marine Systems

Automated Extraction of Antarctic Glacier and Ice Shelf Fronts from Sentinel-1 Imagery Using Deep Learning

Three-dimensional deformation monitoring of landslides based on combination of two-track InSAR observations and pixel-level surface-parallel flow model