Deep learning detection and analysis of eddies in the East Greenland marginal ice zone from Sentinel-1 SAR imagery
文献类型：期刊论文
| 作者 | Jiang, Fei1,2,3,4; Li, Xiaofeng2,3,4; Liu, Yingjie2,3,4; Ren, Yibin2,3,4 |
| 刊名 | REMOTE SENSING OF ENVIRONMENT
 |
| 出版日期 | 2026-03-01 |
| 卷号 | 334页码:17 |
| 关键词 | Ocean eddy Marginal ice zone (MIZ) Deep learning Synthetic aperture radar (SAR) |
| ISSN号 | 0034-4257 |
| DOI | 10.1016/j.rse.2025.115177 |
| 通讯作者 | Ren, Yibin(yibinren@qdio.ac.cn) |
| 英文摘要 | Ocean eddies in the marginal ice zone (MIZ) play a crucial role in sea-ice dynamics and polar ocean-atmosphere interactions; however, their detection remains challenging due to their complex surface signatures. In this study, we developed MIZ-EDYOLO, a deep learning model customized for detecting MIZ eddies from dual-polarized Sentinel-1 SAR imagery. Built upon the YOLOv9-t architecture and enhanced with specialized modifications, the model was trained on a dataset containing over 20,000 slices extracted from 1370 SAR images. The MIZ-EDYOLO model achieves high detection accuracy (similar to 80 % F1-score) on the test set and performs reliably on 200 full-scene images, enabling efficient and automated eddy identification. Using this model, we constructed the first six-year (2018-2023) SAR-based MIZ eddy dataset for the East Greenland region, comprising over 10,000 eddy instances. Analysis of this dataset reveals that eddy distributions are related to boundary currents, topographic forcing, and seasonal variations of the MIZ. Cyclonic eddies (CEs) outnumber anticyclonic eddies (AEs) by a factor of 8.4, while AEs exhibited an average radius about 1.8 times larger than CEs. The observed asymmetries between CEs and AEs are linked to their rotational dynamics and the associated sea-ice responses. This study presents a scalable and operational framework for efficient eddy monitoring in the MIZ, providing new insights into multi-scale oceanographic processes in climate-sensitive polar regions. |
| WOS关键词 | SYNTHETIC-APERTURE RADAR ; MESOSCALE EDDIES ; FRAM STRAIT ; ARCTIC-OCEAN ; RESOLUTION ; DYNAMICS ; NORTH ; FRONT ; MODEL ; FIELD |
| 资助项目 | National Natural Science Foundation of China[42306194] ; National Natural Science Foundation of China[42576205] ; National Natural Science Foundation of China[42206202] ; National Natural Science Foundation of China[42221005] ; IOCAS Foundation[IOCASZZZX301] ; IOCAS Foundation[IOCASZZCG003] ; Laoshan Laboratory Science and Technology Innovation Project[LSKJ202202302] |
| WOS研究方向 | Environmental Sciences & Ecology ; Remote Sensing ; Imaging Science & Photographic Technology |
| 语种 | 英语 |
| WOS记录号 | WOS:001638897900001 |
| 出版者 | ELSEVIER SCIENCE INC |
| 源URL | [http://ir.qdio.ac.cn/handle/337002/204451]  |
| 专题 | 海洋研究所_海洋环流与波动重点实验室 |
| 通讯作者 | Ren, Yibin |
| 作者单位 | 1.Univ Chinese Acad Sci, Coll Marine Sci, Beijing, Peoples R China 2.Chinese Acad Sci, Inst Oceanol, Lab Ocean Circulat & Waves, Qingdao, Peoples R China 3.Qingdao Key Lab Artificial Intelligence Oceanog, Qingdao, Peoples R China 4.Chinese Acad Sci, Inst Oceanol, Key Lab Ocean Observat & Forecasting, Qingdao, Peoples R China |
| 推荐引用方式 GB/T 7714 | Jiang, Fei,Li, Xiaofeng,Liu, Yingjie,et al. Deep learning detection and analysis of eddies in the East Greenland marginal ice zone from Sentinel-1 SAR imagery[J]. REMOTE SENSING OF ENVIRONMENT,2026,334:17. |
| APA | Jiang, Fei,Li, Xiaofeng,Liu, Yingjie,&Ren, Yibin.(2026).Deep learning detection and analysis of eddies in the East Greenland marginal ice zone from Sentinel-1 SAR imagery.REMOTE SENSING OF ENVIRONMENT,334,17. |
| MLA | Jiang, Fei,et al."Deep learning detection and analysis of eddies in the East Greenland marginal ice zone from Sentinel-1 SAR imagery".REMOTE SENSING OF ENVIRONMENT 334(2026):17. |
入库方式： OAI收割
来源：海洋研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。