ORIGINAL RESEARCH article
Front. Mar. Sci.
Sec. Physical Oceanography
Volume 12 - 2025 | doi: 10.3389/fmars.2025.1648021
Automatic Eddy Detection in Antarctic Marginal Ice Zone Using Sentinel-1 SAR Data
Provisionally accepted
- 1Moskovskij fiziko-tehniceskij institut nacional'nyj issledovatel'skij universitet, Dolgoprudny, Russia
- 2UiT The Arctic University of Norway, Tromsø, Norway
- 3FGBUN Institut fiziki atmosfery imeni A M Obuhova Rossijskoj akademii nauk, Moscow, Russia
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Studying oceanic eddies in the Antarctic marginal ice zone (MIZ) is essential due to their unique characteristics and their significant influence on polar climate systems. However, the automated detection of such features remains largely underexplored in general. Moreover, even manual eddy detection has been practically neglected within the Antarctic MIZ specifically. This work presents the first study on the implementation of the machine learning approach for automatic eddy identification in the Antarctic MIZ. We investigate the potential of YOLOv11, a state-of-theart deep learning model, to detect and classify Antarctic eddies using high-resolution synthetic aperture radar imagery. By fine-tuning YOLOv11 on a specialized dataset representing the dynamic Antarctic MIZ, we achieved robust detection of submesoscale and mesoscale eddies.Special significance was placed on distinguishing between cyclonic and anticyclonic eddies, providing essential insights for compiling statistical datasets. Moreover, YOLOv11 architecture was evaluated through a variety of quantitative metrics and visual inspection. The integration of SAHI module with YOLOv11 demonstrated its capability to improve detection of small eddies and increased the mAP 0.5 0.95 by 50 % in comparison with the baseline YOLOv11 model.Experimental results highlight the model's capability to reliably identify eddies across diverse scales and environmental conditions. Overall, this study addresses a significant gap in Antarctic eddy research and sets the stage for advancing automated oceanographic studies in polar regions.
Keywords: mesoscale eddies, Submesoscale eddies, Eddy detection, marginal ice zone, deep learning, YOLOv11
Received: 16 Jun 2025; Accepted: 30 Jul 2025.
Copyright: © 2025 Sandalyuk, Khachatrian and Marchuk. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Eduard Khachatrian, UiT The Arctic University of Norway, Tromsø, Norway
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.