Ocean eddy automatic detection in satellite optical images of the Norwegian Sea using deep machine learning.
https://doi.org/10.30758/0555-2648-2026-72-1-6-18
Abstract
Ocean eddies are an important factor in the large-scale dynamics of the global ocean, including polar and subpolar regions. A robust statistical information on the number and characteristics of mesoscaleand submesoscale eddies will yield new insights on their effect on dynamics of large-scale currents, ice-edgevariability, and other dynamic and biochemical processes in the ocean. Optical images complement the results of the eddy identification study in radar and satellite altimetry images, each of which has its inherent limitations. In optical images, eddies are often observed as spiral or mushroom-shaped structures in chlorophyll distribution, which are formed through the effect of eddy rotation and convergence/divergence patterns. Massive studies of characteristics of ocean eddies require algorithms for their automatic identification. Although several such algorithms have been suggested for satellite altimetry and radar data, no such algorithm exists for satellite optical images. It this study we propose a machine deep learning algorithm for efficient automatic eddy detection in Sentinel-3 optical images. The Lofoten Basin of the Norwegian Sea, an area with a small Rossby deformation radius of less than 10 km, but densely populated with eddies, was selected as a region for algorithm training and validation. Even though the study area is known as one of the cloudiest areas of the northern polar latitudes, 52 mostly cloud-free images were collected over the 9 years of Sentinel-3 data, where 938 eddies were detected. For automatic eddy identification we used SegFormer neural network architecture with an AdamW optimizer, applied for 512×512 pixel tiles. In the course of validation high quality metrics were obtained: Precision = 0.94, Recall = 0.91, Intersection of Union = 0.87 and Dice = 0.93. This demonstrates high efficiency of the algorithm developed. The algorithm additionally identified several eddies missed during visual image inspection. The results of the study are particularly relevant to polar ocean regions, where the predominant eddy sizes are significantly smaller than in the tropics. The robust identification of eddies in optical images is a promising step forward in understanding mesoscale and sub-mesoscale eddy dynamics.
Supporting agencies: The research is funded by the SPbSU project № 129659573.
About the Authors
V. V. Kulak
Scientific Foundation “Nansen International Environmental and Remote Sensing Centre”
Russian Federation
Russian Federation
D. M. Demchev
Scientific Foundation “Nansen International Environmental and Remote Sensing Centre”; Lomonosov Moscow State University Marine Research Center (LMSU MRC); State Scientific Center of the Russian Federation Arctic and Antarctic Research Institute
Russian Federation
Russian Federation
F. A. Gnevashev
Saint Petersburg State University; Scientific Foundation “Nansen International Environmental and Remote Sensing Centre”
Russian Federation
Russian Federation
T. A. Alekseeva
State Scientific Center of the Russian Federation Arctic and Antarctic Research Institute; Space Research Institute Russian Academy of Science
Russian Federation
Russian Federation
I. L. Bashmachnikov
Saint Petersburg State University; Scientific Foundation “Nansen International Environmental and Remote Sensing Centre”
Russian Federation
Russian Federation
References
1. Wunsch C., Ferrari R. Vertical mixing, energy, and the general circulation of the oceans. Annual Review of Fluid Mechanics. 2004;36(1):281–314. https://doi.org/10.1146/annurev.fluid.36.050802.122121
2. Koshlyakov M.N., Tarakanov R.Y. Water transport across the subantarctic front and the global ocean conveyer belt. Oceanology. 2011;51:721–735. https://doi.org/10.1134/S0001437011050110
3. Meneghello G., Marshall J., Cole S.T., Timmermans M.L. Observational inferences of lateral eddy diffusivity in the halocline of the Beaufort Gyre. Geophysical Research Letters. 2017;44(24):12– 331. https://doi:10.1002/2017gl075126
4. Meneghello G., Marshall J., Campin J.M., Doddridge E., Timmermans M.L. The ice-ocean governor: Ice-ocean stress feedback limits Beaufort Gyre spin-up. Geophysical Research Letters. 2018;45(20):11–293. https://doi.org/10.1029/2018GL080171
5. Bashmachnikov I.L., Raj R.P., Golubkin P., Kozlov I.E. Heat transport by mesoscale eddies in the Norwegian and Greenland seas. Journal of Geophysical Research: Oceans. 2023;128(2):e2022JC018987. https://doi.org/10.1029/2022JC018987
6. Klein P., Lapeyre G. The oceanic vertical pump induced by mesoscale and submesoscale turbulence. Annual Review of Marine Science. 2009;1(1):351–375. https://doi.org/10.1146/annurev.marine.010908.163704
7. Gaube P., Chelton D.B., Strutton P.G., Behrenfeld M.J. Satellite observations of chlorophyll, phytoplankton biomass, and Ekman pumping in nonlinear mesoscale eddies. Journal of Geophysical Research: Oceans. 2013;118(12):6349–6370. https://doi.org/10.1002/2013JC009027
8. Chelton D.B., Schlax M.G., Samelson R.M. Global observations of nonlinear mesoscale eddies. Progress in Oceanography. 2011;91(2):167–216. https://doi.org/10.1016/j.pocean.2011.01.002
9. Kubryakov A., Kozlov I., Manucharyan G. Large mesoscale eddies in the Western Arctic Ocean from satellite altimetry measurements. Journal of Geophysical Research: Oceans. 2021;126(5):e2020JC016670. https://doi.org/10.1029/2020JC016670
10. Atadzhanova O.A., Zimin A.V., Romanenkov D.A., Kozlov I.E. Satellite radar observations of small eddies in the White, Barents and Kara Seas. Physical Oceanography. 2017;2:75–83. https://doi.org/10.22449/1573-160X-2017-2-75-83
11. Bashmachnikov I.L., Kozlov I.E., Petrenko L.A., Glok N.I., Wekerle C. Eddies in the North Greenland Sea and Fram Strait from satellite altimetry, SAR and high-resolution model data. Journal of Geophysical Research: Oceans. 2020;125(7):e2019JC015832. https://doi.org/10.1029/2019JC015832
12. Morozov E.A., Kozlov I.E. Eddies in the Arctic Ocean revealed from MODIS optical imagery. Remote Sensing. 2023;15(6):1608. https://doi.org/10.3390/rs15061608
13. Ivanov A.Y., Ginzburg A.I. Oceanic eddies in synthetic aperture radar images. Journal of Earth System Science. 2002;111(3):281–295. https://doi.org/10.1007/BF02701974
14. Johannessen J.A., Kudryavtsev V., Akimov D., Eldevik T., Winther N., Chapron B. On radar imaging of current features: 2. Mesoscale eddy and current front detection. Journal of Geophysical Research: Oceans. 2005;110(C7). https://doi.org/10.1029/2004JC002802
15. Лаврова О.Ю., Костяной А.Г., Лебедев С.А., Митягина М.И., Гинзбург А.И., Шеремет Н.А. Комплексный спутниковый мониторинг морей России. М.: ИКИ РАН; 2011. 480 с.
16. Kozlov I.E., Artamonova A.V., Manucharyan G.E., Kubryakov A.A. Eddies in the Western Arctic Ocean from spaceborne SAR observations over open ocean and marginal ice zones. Journal of Geophysical Research: Oceans. 2019;124(9):6601–6616. https://doi.org/10.1029/2019JC015113
17. Zhurbas V., Väli G., Kuzmina N. Rotation of floating particles in submesoscale cyclonic and anticyclonic eddies: a model study for the southeastern Baltic Sea. Ocean Science. 2019;15(6):1691–1705. https://doi.org/10.5194/os-15-1691-2019
18. Ginzburg A.I., Krek E.V., Kostianoy A.G., Solovyev D.M. Evolution of mesoscale anticyclonic vortex and vortex dipoles/multipoles on its base in the south-eastern Baltic (satellite information May–July 2015). Okeanologicheskie issledovaniya = Journal of Oceanological Research. 2017;45:10–22. (In Russ.). https://doi.org/10.29006/1564-2291.JOR-2017.45(1).3
19. Новикова Ю.С., Башмачников И.Л. Вихри Лофотенской котловины по данным спутниковой альтиметрии, изображениям в радиолокационном и видимом диапазонах. В: Моря России: Год науки и технологий в РФ — Десятилетие наук об океане ООН. Тезисы докладов Всероссийской научной конференции. Севастополь, 2021. Севастополь: ФИЦ МГИ; 2021. С. 291–292.
20. Федоров К.Н., Гинзбург А.И. Грибовидные течения (вихревые диполи) — одна из наиболее распространенных форм когерентных движений в океане. В: Когерентные структуры и самоорганизация океанских движений. Баренблатт Г.И. (ред.). М.: Наука; 1992. С.12–20.
21. Munk W., Armi L., Fischer K., Zachariasen F. Spirals on the sea. Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences. 2000;456(1997):1217–1280. https://doi.org/10.1098/rspa.2000.0560
22. Xu G., Dong C., Liu Y., Gaube P., Yang J. Chlorophyll rings around ocean eddies in the North Pacific. Scientific Reports. 2019;9(1):2056. https://doi.org/10.1038/s41598-018-38457-8
23. Башмачников И.Л., Белоненко Т.В., Куйбин П.А. Приложение теории колоннообразных Q-вихрей с винтовой структурой к описанию динамических характеристик Лофотенского вихря Норвежского моря. Вестник Санкт-Петербургского университета. Науки о Земле. 2017;(3):221–236. https://doi.org/10.21638/11701/spbu07.2017.301
24. Siegelman L., Klein P., Rivière P., Thompson A.F., Torres H.S., Flexas M., Menemenlis D. Enhanced upward heat transport at deep submesoscale ocean fronts. Nature Geoscience. 2020;13(1):50–55. https://doi.org/10.1038/s41561-019-0489-1
25. Calil P., Richards K. Transient upwelling hot spots in the oligotrophic North Pacific. Journal of Geophysical Research: Oceans. 2010;115(C2). https://doi.org/10.1029/2009JC005360
26. Xia L., Chen G., Chen X., Ge L., Huang B. Submesoscale oceanic eddy detection in SAR images using context and edge association network. Frontiers in Marine Science. 2022;9:1023624. https://doi.org/10.3389/fmars.2022.1023624
27. Xu M., Li H., Yun Y., Yang F., Li C. End-to-end pixel-wisely detection of oceanic eddy on SAR images with stacked attention network. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 2023;16:9711–9724. https://doi.org/10.1109/JSTARS.2023.3322404
28. Zhang D., Gade M., Wang W., Zhou H. EddyDet: A deep framework for oceanic eddy detection in synthetic aperture radar images. Remote Sensing. 2023;15(19):4752. https://doi.org/10.3390/rs15194752
29. Zi N., Li X.M., Gade M., Fu H., Min S. Ocean eddy detection based on YOLO deep learning algorithm by synthetic aperture radar data. Remote Sensing of Environment. 2024;307:114139. https://doi.org/10.1016/j.rse.2024.114139
30. Sandalyuk N., Khachatrian E. Automatic eddy detection in the MIZ based on YOLO algorithm and SAR images. Science of Remote Sensing. 2025;100228. https://doi.org/10.1016/j.srs.2025.100228
31. Raj R., Johannessen J., Eldevik T., Nilsen J., Halo I. Quantifying mesoscale eddies in the Lofoten Basin. Journal of Geophysical Research: Oceans. 2016;121(7):4503–4521. https://doi.org/10.1002/2016JC011637
32. Zhang M., Xie S., Liu X., Lin W., Zheng X., Golaz J.C., Zhang Y. Cloud phase simulation at high latitudes in EAMv2: Evaluation using CALIPSO observations and comparison with EAMv1. Journal of Geophysical Research: Atmospheres. 2022;127:e2022JD037100. https://doi.org/10.1029/2022JD037100
33. Schirber M. The full palette of photosynthesis. Astrobiology at NASA. October 25, 2013. Available at: https://astrobiology.nasa.gov/news/the-full-palette-of-photosynthesis/ (accessed 20.09.2025).
34. Koszalka I., LaCasce J., Andersson M., Orvik K., Mauritzen C. Surface circulation in the Nordic Seas from clustered drifters. Deep Sea Research Part I: Oceanographic Research Papers. 2011;58(4):468–485. https://doi.org/10.1016/j.dsr.2011.01.007
35. Belonenko T.V., Travkin V.S., Koldunov A.V., Volkov D.L. Topographic experiments over dynamical processes in the Norwegian Sea. Russian Journal of Earth Science. 2021;21(1):ES1006. https://doi.org/10.2205/2020ES000747
36. Xie E., Wang W., Yu Z., Anandkumar A., Alvarez J.M., Luo P. SegFormer: Simple and efficient design for semantic segmentation with transformers. Advances in Neural Information Processing Systems. 2021;34:12077–12090.
37. Maier-Hein L., Eisenmann M., Reinke A. et al. Why rankings of biomedical image analysis competitions should be interpreted with care. Nature Communications. 2018;9(1):5217. https://doi.org/10.1038/s41467-018-07619-7
38. Dice L.R. Measures of the amount of ecologic association between species. Ecology. 1945;26(3):297–302. https://doi.org/10.2307/1932409
39.
Review
For citations:
Kulak V.V., Demchev D.M., Gnevashev F.A., Alekseeva T.A., Bashmachnikov I.L. Ocean eddy automatic detection in satellite optical images of the Norwegian Sea using deep machine learning. Arctic and Antarctic Research. 2026;72(1):6-18. (In Russ.) https://doi.org/10.30758/0555-2648-2026-72-1-6-18
Views:
541
JATS XML
Email this article 
