Data

Grounded Icebergs around Antarctica: A High-Resolution Dataset Derived from Deep Learning and Sentinel-1 Synthetic Aperture Radar

University of Tasmania, Australia
Jiao, Kaihong ; Fraser, Alexander D. ; Lohse, Johannes ; Wongpan, Pat ; Adams, Caitlin ; Bradley, Alexander C.
Viewed: [[ro.stat.viewed]] Cited: [[ro.stat.cited]] Accessed: [[ro.stat.accessed]]
ctx_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rfr_id=info%3Asid%2FANDS&rft_id=info:doi10.25959/54SX-PT47&rft.title=Grounded Icebergs around Antarctica: A High-Resolution Dataset Derived from Deep Learning and Sentinel-1 Synthetic Aperture Radar&rft.identifier=10.25959/54SX-PT47&rft.description=This dataset (provided as a GeoPackage file) consists of a high-resolution, continent-wide map of grounded icebergs around the Antarctic continental shelf, derived from Sentinel-1 Synthetic Aperture Radar (SAR) imagery. The data capture conditions during the annual landfast sea ice minimum, spanning from late February to early April 2025. The dataset comprises 39,619 stationary iceberg targets, providing vector polygons, surface area, centroid coordinates, and temporal tracking information for each target. The data are provided in a polar stereographic projection (EPSG:3031) and include a classification for each iceberg based on its spatial relationship with contemporary landfast sea ice.Maintenance and Update Frequency: notPlannedStatement: The dataset was generated using Sentinel-1 Synthetic Aperture Radar (SAR) Ground Range Detected (GRD) imagery acquired in Extra-Wide (EW) and Interferometric-Wide (IW) swath modes. Images were pre-processed to a radiometrically terrain-corrected gamma-nought backscatter product using pyroSAR and GAMMA software, and projected to the Antarctic Polar Stereographic coordinate system (EPSG:3031). Iceberg detection was performed via a custom ResUNet deep learning architecture that integrates Convolutional Block Attention Modules (CBAM) to extract targets from complex coastal backgrounds. Stationary targets were then identified across multi-temporal image sequences (typically 12-day temporal baselines) using a tracking algorithm based on the Hungarian method, which evaluated centroid displacement, morphological similarity (Hu invariant moments), and Intersection over Union (IoU). To mitigate environmental false positives, post-processing quality control filters were applied utilizing IBCSO version 2 bathymetry and AMSR2 daily sea ice concentration data. Finally, the identified stationary icebergs were subjected to a spatial overlay analysis using a contemporary, high-resolution circum-Antarctic landfast sea ice mask. This enabled the partitioning of targets into 'high-confidence grounded' icebergs (located outside fast ice) and those potentially immobilized by fast ice entrapment.&rft.creator=Jiao, Kaihong &rft.creator=Fraser, Alexander D. &rft.creator=Lohse, Johannes &rft.creator=Wongpan, Pat &rft.creator=Adams, Caitlin &rft.creator=Bradley, Alexander C. &rft.date=2026&rft.coverage=westlimit=-179.99; southlimit=-80.00; eastlimit=179.99; northlimit=-60.00&rft.coverage=westlimit=-179.99; southlimit=-80.00; eastlimit=179.99; northlimit=-60.00&rft_rights=Creative Commons Attribution 4.0 International License https://creativecommons.org/licenses/by/4.0/&rft_rights=Cite data as: Jiao, K., Fraser, A., Lohse, J., & Wongpan, P. (2026). Grounded Icebergs around Antarctica: A High-Resolution Dataset Derived from Deep Learning and Sentinel-1 Synthetic Aperture Radar [Data set]. Institute for Marine and Antarctic Studies. https://doi.org/10.25959/54SX-PT47&rft_rights=Data, products and services from IMAS are provided as is without any warranty as to fitness for a particular purpose.&rft_subject=oceans&rft_subject=geoscientificInformation&rft_subject=Grounded icebergs&rft_subject=ICEBERGS&rft_subject=Physical Oceanography&rft_subject=EARTH SCIENCES&rft_subject=OCEANOGRAPHY&rft_subject=Glaciology&rft_subject=PHYSICAL GEOGRAPHY AND ENVIRONMENTAL GEOSCIENCE&rft_subject=Marine Geoscience&rft_subject=GEOLOGY&rft.type=dataset&rft.language=English Access the data

Licence & Rights:

Open Licence view details
CC-BY

Creative Commons Attribution 4.0 International License
https://creativecommons.org/licenses/by/4.0/

Cite data as: Jiao, K., Fraser, A., Lohse, J., & Wongpan, P. (2026). Grounded Icebergs around Antarctica: A High-Resolution Dataset Derived from Deep Learning and Sentinel-1 Synthetic Aperture Radar [Data set]. Institute for Marine and Antarctic Studies. https://doi.org/10.25959/54SX-PT47

Data, products and services from IMAS are provided "as is" without any warranty as to fitness for a particular purpose.

Access:

Open

Contact Information



Full description

This dataset (provided as a GeoPackage file) consists of a high-resolution, continent-wide map of grounded icebergs around the Antarctic continental shelf, derived from Sentinel-1 Synthetic Aperture Radar (SAR) imagery. The data capture conditions during the annual landfast sea ice minimum, spanning from late February to early April 2025. The dataset comprises 39,619 stationary iceberg targets, providing vector polygons, surface area, centroid coordinates, and temporal tracking information for each target. The data are provided in a polar stereographic projection (EPSG:3031) and include a classification for each iceberg based on its spatial relationship with contemporary landfast sea ice.

Lineage

Maintenance and Update Frequency: notPlanned
Statement: The dataset was generated using Sentinel-1 Synthetic Aperture Radar (SAR) Ground Range Detected (GRD) imagery acquired in Extra-Wide (EW) and Interferometric-Wide (IW) swath modes. Images were pre-processed to a radiometrically terrain-corrected gamma-nought backscatter product using pyroSAR and GAMMA software, and projected to the Antarctic Polar Stereographic coordinate system (EPSG:3031). Iceberg detection was performed via a custom ResUNet deep learning architecture that integrates Convolutional Block Attention Modules (CBAM) to extract targets from complex coastal backgrounds. Stationary targets were then identified across multi-temporal image sequences (typically 12-day temporal baselines) using a tracking algorithm based on the Hungarian method, which evaluated centroid displacement, morphological similarity (Hu invariant moments), and Intersection over Union (IoU). To mitigate environmental false positives, post-processing quality control filters were applied utilizing IBCSO version 2 bathymetry and AMSR2 daily sea ice concentration data. Finally, the identified stationary icebergs were subjected to a spatial overlay analysis using a contemporary, high-resolution circum-Antarctic landfast sea ice mask. This enabled the partitioning of targets into 'high-confidence grounded' icebergs (located outside fast ice) and those potentially immobilized by fast ice entrapment.

Notes

Credit
Australian Research Council (ARC) (Grants: FT230100234, LP170101090, LE220100103, DP240100325)
Credit
Antarctic Science Collaboration Initiative (ASCI), Australian Government
Credit
Antarctic Science Foundation (Harris Charitable Trust)
Credit
National Computational Infrastructure (NCI Australia) (Project jk72)
Credit
Geoscience Australia

Issued: 06 03 2026

Data time period: 2025-02-22 to 2025-04-21

This dataset is part of a larger collection

Click to explore relationships graph

-179.99,-80 0,-80 179.99,-80 179.99,-60 0,-60 -179.99,-60 -179.99,-80

0,-70

text: westlimit=-179.99; southlimit=-80.00; eastlimit=179.99; northlimit=-60.00

Other Information
Jiao, K., Fraser, A. D., Lohse, J., Wongpan, P., Adams, C., and Bradley, A. C. (2026). Grounded icebergs around Antarctica: a high-resolution dataset derived from deep learning and Sentinel-1 synthetic aperture radar. Earth System Science Data, 18, 6017–6037 (Associated publication)

doi : https://doi.org/10.5194/essd-18-6017-2026

Explore data in Seamap Antarctica

url : https://seamapantarctica.org/map/#079c3622-389f-4dcf-91f3-b091db6b613b

Identifiers
ACN 633 798 857