Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imagery
<p>The fracturing of glaciers and ice shelves in Antarctica influences their dynamics and stability. Hence, data on the evolving distribution of crevasses are required to better understand the evolution of the ice sheet, though such data have traditionally been difficult and time-consuming to...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2023-10-01
|
| Series: | The Cryosphere |
| Online Access: | https://tc.copernicus.org/articles/17/4421/2023/tc-17-4421-2023.pdf |
| _version_ | 1797655956138491904 |
|---|---|
| author | T. Surawy-Stepney A. E. Hogg S. L. Cornford D. C. Hogg |
| author_facet | T. Surawy-Stepney A. E. Hogg S. L. Cornford D. C. Hogg |
| author_sort | T. Surawy-Stepney |
| collection | DOAJ |
| description | <p>The fracturing of glaciers and ice shelves in Antarctica influences their dynamics and stability. Hence, data on the evolving distribution of crevasses are required to better understand the evolution of the ice sheet, though such data have traditionally been difficult and time-consuming to generate. Here, we present an automated method of mapping crevasses on grounded and floating ice with the application of convolutional neural networks to Sentinel-1 synthetic aperture radar backscatter data. We apply this method across Antarctica to images acquired between 2015 and 2022, producing a 7.5-year record of composite fracture maps at monthly intervals and 50 <span class="inline-formula">m</span> spatial resolution and showing the distribution of crevasses around the majority of the ice sheet margin. We develop a method of quantifying changes to the density of ice shelf fractures using a time series of crevasse maps and show increases in crevassing on Thwaites and Pine Island ice shelves over the observational period, with observed changes elsewhere in the Amundsen Sea dominated by the advection of existing crevasses. Using stress fields computed using the BISICLES ice sheet model, we show that much of this structural change has occurred in buttressing regions of these ice shelves, indicating a recent and ongoing link between fracturing and the developing dynamics of the Amundsen Sea sector.</p> |
| first_indexed | 2024-03-11T17:22:13Z |
| format | Article |
| id | doaj.art-06450cd245a3453f8dc97e8a40339e53 |
| institution | Directory Open Access Journal |
| issn | 1994-0416 1994-0424 |
| language | English |
| last_indexed | 2024-03-11T17:22:13Z |
| publishDate | 2023-10-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | The Cryosphere |
| spelling | doaj.art-06450cd245a3453f8dc97e8a40339e532023-10-19T08:41:21ZengCopernicus PublicationsThe Cryosphere1994-04161994-04242023-10-01174421444510.5194/tc-17-4421-2023Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imageryT. Surawy-Stepney0A. E. Hogg1S. L. Cornford2D. C. Hogg3School of Earth and Environment, University of Leeds, Leeds, United KingdomSchool of Earth and Environment, University of Leeds, Leeds, United KingdomSchool of Geographical Sciences, University of Bristol, Bristol, United KingdomSchool of Computing, University of Leeds, Leeds, United Kingdom<p>The fracturing of glaciers and ice shelves in Antarctica influences their dynamics and stability. Hence, data on the evolving distribution of crevasses are required to better understand the evolution of the ice sheet, though such data have traditionally been difficult and time-consuming to generate. Here, we present an automated method of mapping crevasses on grounded and floating ice with the application of convolutional neural networks to Sentinel-1 synthetic aperture radar backscatter data. We apply this method across Antarctica to images acquired between 2015 and 2022, producing a 7.5-year record of composite fracture maps at monthly intervals and 50 <span class="inline-formula">m</span> spatial resolution and showing the distribution of crevasses around the majority of the ice sheet margin. We develop a method of quantifying changes to the density of ice shelf fractures using a time series of crevasse maps and show increases in crevassing on Thwaites and Pine Island ice shelves over the observational period, with observed changes elsewhere in the Amundsen Sea dominated by the advection of existing crevasses. Using stress fields computed using the BISICLES ice sheet model, we show that much of this structural change has occurred in buttressing regions of these ice shelves, indicating a recent and ongoing link between fracturing and the developing dynamics of the Amundsen Sea sector.</p>https://tc.copernicus.org/articles/17/4421/2023/tc-17-4421-2023.pdf |
| spellingShingle | T. Surawy-Stepney A. E. Hogg S. L. Cornford D. C. Hogg Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imagery The Cryosphere |
| title | Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imagery |
| title_full | Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imagery |
| title_fullStr | Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imagery |
| title_full_unstemmed | Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imagery |
| title_short | Mapping Antarctic crevasses and their evolution with deep learning applied to satellite radar imagery |
| title_sort | mapping antarctic crevasses and their evolution with deep learning applied to satellite radar imagery |
| url | https://tc.copernicus.org/articles/17/4421/2023/tc-17-4421-2023.pdf |
| work_keys_str_mv | AT tsurawystepney mappingantarcticcrevassesandtheirevolutionwithdeeplearningappliedtosatelliteradarimagery AT aehogg mappingantarcticcrevassesandtheirevolutionwithdeeplearningappliedtosatelliteradarimagery AT slcornford mappingantarcticcrevassesandtheirevolutionwithdeeplearningappliedtosatelliteradarimagery AT dchogg mappingantarcticcrevassesandtheirevolutionwithdeeplearningappliedtosatelliteradarimagery |