{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,25]],"date-time":"2026-07-25T05:02:34Z","timestamp":1784955754420,"version":"3.55.0"},"reference-count":20,"publisher":"Springer Science and Business Media LLC","issue":"8","license":[{"start":{"date-parts":[[2026,7,2]],"date-time":"2026-07-02T00:00:00Z","timestamp":1782950400000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,7,2]],"date-time":"2026-07-02T00:00:00Z","timestamp":1782950400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"name":"Universit\u00e0 del Salento"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Earth Sci Inform"],"published-print":{"date-parts":[[2026,8]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Accurate ocean forecasting is essential for supporting a wide range of marine applications. Recent advances in artificial intelligence have highlighted the potential of data-driven models to outperform traditional numerical approaches, particularly in atmospheric weather forecasting. However, extending these methods to ocean systems remains challenging due to their inherently slower dynamics and complex boundary conditions. In this work, we present MedFormer, a fully data-driven deep learning model specifically designed for medium-range ocean forecasting in the Mediterranean Sea. MedFormer is based on a U-Net architecture augmented with 3D attention mechanism, operating at a high horizontal resolution of 1\/24\u00b0. The model is trained on 20 years of daily ocean reanalysis data and fine-tuned with high-resolution operational analyses. It generates 9-day forecasts using an autoregressive strategy. The model leverages both historical ocean states and atmospheric forcings, making it well-suited for operational use. We benchmark MedFormer against the state-of-the-art Mediterranean Forecasting System (MedFS), developed at the Euro-Mediterranean Center on Climate Change (CMCC), using both analysis data and independent observations. The forecast skills, evaluated with the Root Mean Squared Difference and the Anomaly Correlation Coefficient, indicate that MedFormer demonstrates improved forecast skill over MedFS across key 3D ocean variables, with the most pronounced gains during summer and autumn. These findings underscore the potential of data-driven approaches like MedFormer to complement traditional numerical ocean forecasting systems, demonstrating competitive or improved accuracy for most evaluated variables, with the exception of Sea Surface Height alongside substantial gains in computational efficiency.<\/jats:p>","DOI":"10.1007\/s12145-026-02169-9","type":"journal-article","created":{"date-parts":[[2026,7,2]],"date-time":"2026-07-02T05:40:47Z","timestamp":1782970847000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["MedFormer: a data-driven model for forecasting the Mediterranean Sea"],"prefix":"10.1007","volume":"19","author":[{"given":"Italo","family":"Epicoco","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Davide","family":"Donno","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Gabriele","family":"Accarino","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Simone","family":"Norberti","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Alessandro","family":"Grandi","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Michele","family":"Giurato","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ronan","family":"McAdam","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Donatello","family":"Elia","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Emanuela","family":"Clementi","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Paola","family":"Nassisi","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Enrico","family":"Scoccimarro","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Giovanni","family":"Coppini","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Silvio","family":"Gualdi","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Giovanni","family":"Aloisio","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Simona","family":"Masina","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Giulio","family":"Boccaletti","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Antonio","family":"Navarra","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,7,2]]},"reference":[{"key":"2169_CR1","unstructured":"Aouni AE et\u00a0al (2025) Glonet: Mercator\u2019s end-to-end neural global ocean forecasting system. arXiv:2412.05454"},{"key":"2169_CR2","doi-asserted-by":"publisher","unstructured":"Bengio Y, Louradour J, Collobert R, Weston J, for Computing\u00a0Machinery A (ed) (2009) Curriculum learning. (ed.for Computing\u00a0Machinery, A.) Proceedings of the 26th Annual International Conference on Machine Learning, ICML \u201909, 41\u201348 (New York, NY, USA). https:\/\/doi.org\/10.1145\/1553374.1553380","DOI":"10.1145\/1553374.1553380"},{"key":"2169_CR3","doi-asserted-by":"publisher","first-page":"533","DOI":"10.1038\/s41586-023-06185-3","volume":"619","author":"K Bi","year":"2023","unstructured":"Bi K et al (2023) Accurate medium-range global weather forecasting with 3d neural networks. Nature 619:533\u2013538. https:\/\/doi.org\/10.1038\/s41586-023-06185-3","journal-title":"Nature"},{"key":"2169_CR4","doi-asserted-by":"publisher","unstructured":"Clementi E et\u00a0al (2021) Mediterranean sea physical analysis and forecast (cmems med-currents, eas6 system) (version 1) [data set]. https:\/\/doi.org\/10.25423\/CMCC\/MEDSEA_ANALYSISFORECAST_PHY_006_013_EAS6","DOI":"10.25423\/CMCC\/MEDSEA_ANALYSISFORECAST_PHY_006_013_EAS6"},{"key":"2169_CR5","doi-asserted-by":"crossref","unstructured":"Coppini G et\u00a0al (2023) The mediterranean forecasting system \u2013 part 1: Evolution and performance. Ocean Science 19:1483\u20131516. https:\/\/os.copernicus.org\/articles\/19\/1483\/2023\/","DOI":"10.5194\/os-19-1483-2023"},{"key":"2169_CR6","doi-asserted-by":"crossref","unstructured":"Dheeshjith S et\u00a0al (2025) Samudra: An ai global ocean emulator for climate. Geophysical Research Letters 52:e2024GL114318. https:\/\/agupubs.onlinelibrary.wiley.com\/doi\/abs\/10.1029\/2024GL114318. E2024GL114318 2024GL114318","DOI":"10.1029\/2024GL114318"},{"key":"2169_CR7","doi-asserted-by":"crossref","unstructured":"Escudier R et\u00a0al (2021) A high resolution reanalysis for the mediterranean sea. Frontiers in Earth Science 9. https:\/\/www.frontiersin.org\/journals\/earth-science\/articles\/10.3389\/feart.2021.702285","DOI":"10.3389\/feart.2021.702285"},{"key":"2169_CR8","doi-asserted-by":"crossref","unstructured":"Guo Z et\u00a0al (2024) Data-driven global ocean modeling for seasonal to decadal prediction. arXiv:2405.15412","DOI":"10.1126\/sciadv.adu2488"},{"key":"2169_CR9","doi-asserted-by":"publisher","first-page":"45051","DOI":"10.1038\/s41598-025-31177-w","volume":"15","author":"D Holmberg","year":"2025","unstructured":"Holmberg D, Clementi E, Epicoco I, Roos T (2025) Accurate mediterranean sea forecasting via graph-based deep learning. Sci Rep 15:45051. https:\/\/doi.org\/10.1038\/s41598-025-31177-w","journal-title":"Sci Rep"},{"key":"2169_CR10","doi-asserted-by":"publisher","first-page":"1416","DOI":"10.1126\/science.adi2336","volume":"382","author":"R Lam","year":"2023","unstructured":"Lam R et al (2023) Learning skillful medium-range global weather forecasting. Science 382:1416\u20131421. https:\/\/doi.org\/10.1126\/science.adi2336","journal-title":"Science"},{"key":"2169_CR11","doi-asserted-by":"publisher","unstructured":"Li S et al (2020) Pytorch distributed: experiences on accelerating data parallel training. Proc VLDB Endow 13:3005\u20133018. https:\/\/doi.org\/10.14778\/3415478.3415530","DOI":"10.14778\/3415478.3415530"},{"key":"2169_CR12","doi-asserted-by":"crossref","unstructured":"Liu Z et\u00a0al (ed) (2021) Swin transformer: Hierarchical vision transformer using shifted windows. In: 2021 IEEE\/CVF International Conference on Computer Vision (ICCV), 9992\u201310002 (IEEE)","DOI":"10.1109\/ICCV48922.2021.00986"},{"key":"2169_CR13","doi-asserted-by":"crossref","unstructured":"Liu Z et\u00a0al (ed) (2022) Swin transformer v2: Scaling up capacity and resolution. In: 2022 IEEE\/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 11999\u201312009 (IEEE)","DOI":"10.1109\/CVPR52688.2022.01170"},{"key":"2169_CR14","unstructured":"Loshchilov, I, Hutter F (2019) Decoupled weight decay regularization. arXiv:1711.05101"},{"key":"2169_CR15","unstructured":"Loshchilov I, Hutter F (2017) Sgdr: Stochastic gradient descent with warm restarts. arXiv:1608.03983"},{"key":"2169_CR16","unstructured":"Ronneberger O, Fischer P, Brox T Navab N, Hornegger J, Wells WM, Frangi AF (eds) (2015) U-net: Convolutional networks for biomedical image segmentation. In: Navab N, Hornegger J, Wells WM, Frangi AF (eds) Medical Image Computing and Computer-Assisted Intervention \u2013 MICCAI 2015, 234\u2013241 (Springer International Publishing, Cham)"},{"key":"2169_CR17","doi-asserted-by":"crossref","unstructured":"Storto A, Masina S (2016) C-glorsv5: an improved multipurpose global ocean eddy-permitting physical reanalysis. Earth System Science Data 8:679\u2013696. https:\/\/essd.copernicus.org\/articles\/8\/679\/2016\/","DOI":"10.5194\/essd-8-679-2016"},{"key":"2169_CR18","unstructured":"Vaswani A et\u00a0al (2023) Attention is all you need. arXiv:1706.03762"},{"key":"2169_CR19","unstructured":"Wang X et\u00a0al (2024) Xihe: A data-driven model for global ocean eddy-resolving forecasting. arXiv:2402.02995"},{"key":"2169_CR20","unstructured":"Xiong W et\u00a0al (2023) Ai-goms: Large ai-driven global ocean modeling system. arXiv:2308.03152"}],"container-title":["Earth Science Informatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s12145-026-02169-9.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s12145-026-02169-9","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s12145-026-02169-9.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,25]],"date-time":"2026-07-25T04:35:55Z","timestamp":1784954155000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s12145-026-02169-9"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,2]]},"references-count":20,"journal-issue":{"issue":"8","published-print":{"date-parts":[[2026,8]]}},"alternative-id":["2169"],"URL":"https:\/\/doi.org\/10.1007\/s12145-026-02169-9","relation":{"has-preprint":[{"id-type":"doi","id":"10.21203\/rs.3.rs-7899254\/v1","asserted-by":"object"}]},"ISSN":["1865-0473","1865-0481"],"issn-type":[{"value":"1865-0473","type":"print"},{"value":"1865-0481","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,2]]},"assertion":[{"value":"19 October 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"13 June 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"2 July 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"The authors declare no competing interests.","order":1,"name":"Ethics","label":"Competing interests","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"Not applicable.","order":2,"name":"Ethics","label":"Ethics approval and consent to participate","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"Not applicable.","order":3,"name":"Ethics","label":"Consent for publication","group":{"name":"EthicsHeading","label":"Declarations"}}],"article-number":"129"}}