Bellinghausen Sea, Antarctica, taken onboard the R/V Falkor (too) in 2025. (Credit: Laura Cimoli, University of Cambridge)
In A Nutshell
- A layer of relatively warm ocean water known as Circumpolar Deep Water has been shifting toward Antarctica at roughly 1.26 kilometers per year for at least two decades, according to a new study.
- Researchers combined ship-based ocean surveys with data from thousands of robotic floats and machine-learning models to track the movement across nearly the entire Southern Ocean.
- As this warm water moves closer to the continent, it poses a growing risk to Antarctic ice shelves, which hold back glaciers that could raise global sea levels if they collapse.
- Scientists identified two likely drivers: a shrinking layer of cold bottom water that normally acts as a buffer, and southward-shifting westerly winds that are pulling ocean circulation patterns toward the continent.
A massive layer of relatively warm water deep in the Southern Ocean has been moving closer to Antarctica for at least two decades, and a new study shows just how fast. Researchers found that the boundary of this warm-water layer has shifted toward Antarctica at roughly 1.26 kilometers per year. That may sound slow, but this shift could increase the amount of ocean heat available to melt ice shelves from below and contribute to rising seas around the world.
Circumpolar Deep Water, as scientists call it, circles Antarctica at middle ocean depths, acting as the main delivery system of ocean heat toward the underside of the continent’s floating ice shelves. When it reaches those edges, it eats away at ice from underneath, a process scientists consider one of the biggest uncertainties in sea-level rise projections. Researchers have now produced a detailed, wide-ranging picture of how this warm water has been redistributing itself, showing a southward shift across almost all longitudes of the Southern Ocean.
Two Data Sources, One Clear Signal in Antarctic Ice Shelf Warming
Published in Communications Earth & Environment, the team, led by Joshua Lanham at the University of Cambridge, identified different water types by their physical and chemical fingerprints, including temperature, saltiness, dissolved oxygen, and nutrient levels. It works a bit like identifying different brands of soup by their ingredients rather than their labels. Researchers used a global ocean chemistry dataset spanning 1972 to 2013 to map the long-term average distribution of Southern Ocean water masses, then compared ship surveys from 2005 to 2010 against later ones collected after 2015 to identify where water had shifted.
Ship surveys are highly accurate but cover only thin slices of the ocean at widely spaced intervals. To fill the gaps, the researchers turned to the Argo program, a global network of roughly 4,000 robotic floats that periodically dive to 2,000 meters and beam temperature and salinity readings back to shore via satellite. Because Argo floats lack the chemical sensors needed for traditional water identification, the team built machine-learning models trained on the richer ship data to recognize water types from Argo’s simpler measurements, then applied the system to monthly data covering January 2004 through January 2024.
Both the direct ship observations and the machine-learning-powered Argo analysis told the same story. In the upper 2,000 meters of the ocean, the warm-water layer was getting thicker near Antarctica and thinner farther north. The pattern held across almost all sections around the continent, though the speed of the shift varied by region.
Warm Water Advancing Fastest Near the Weddell Sea
Among the regional differences, the Weddell Sea stood out most sharply, with the warm water’s boundary migrating southward at 2.39 kilometers per year, nearly three times the rate in West Antarctica at 0.80 kilometers per year. East Antarctica fell in between at 1.31 kilometers per year, with the overall continent-wide average at 1.26 kilometers per year.
A statistical analysis supported the idea that this southward migration was not background noise: it was the single leading pattern of change in the warm-water layer, accounting for 22 percent of total variability.
Researchers also found clues about what is making room for the warm water’s advance. In the Weddell Sea and along East Antarctica, the warm layer’s expansion was largely offset by a shrinking of the cold, dense water that forms near the continental shelf and sinks to the ocean floor. This cold bottom water helps limit warm-water access near the shelf, and as it contracts, that barrier may weaken.
What’s Driving the Antarctic Warm Water Shift
Researchers outlined several possible causes but stopped short of pointing to any single one. One possibility involves the well-documented decline in cold bottom water production, which has been warming, freshening, and shrinking in recent decades, potentially weakening the barrier that previously kept the warm water at bay. Another involves shifts in the powerful westerly winds circling Antarctica, which have been moving southward and intensifying. Climate models consistently project those winds will continue creeping southward under future warming, which could mean the migration documented here persists or speeds up.
Heat carried by this warm water is a major driver of melting from below for vulnerable Antarctic ice shelves. Ice shelves hold back the massive glaciers behind them, and when they thin or collapse, those glaciers flow faster into the sea, raising global sea levels. Because this water mass also moves dissolved carbon and nutrients through the Southern Ocean, changes in its position could affect carbon exchange with the atmosphere, though the study flags that as an area for future work, not a confirmed outcome.
What gives the finding weight is the agreement across several lines of evidence: two independent data sources, multiple analytical approaches, and extensive testing all pointing to the same development. Warm water that Antarctica’s ice shelves are most vulnerable to is not staying put.
Paper Notes
Limitations
Several methodological constraints apply. The Argo dataset used to track water mass changes extends only to 65 degrees south, limiting observation of near-shelf dynamics and leaving out regions around Dense Shelf Water formation zones. The classification framework also assumes biogeochemical end members remain stable across the Argo period, a simplification the authors acknowledge. End member selection necessarily involves a degree of subjectivity, though the authors tested a range of alternatives and found their main conclusions remained robust. The GO-SHIP ship section data is sparse in both time and space, which is part of why the machine-learning extension to Argo data was needed. The Weddell Sea sector lacked complete post-2015 biogeochemical ship data, requiring the team to use earlier cruises as a proxy for the later period.
Funding and Disclosures
Joshua Lanham was supported by an EPSRC doctoral training grant (EP/T51780X/1). Sarah Purkey received support from the US Argo program through a NOAA-GOMO award. Kaushik Srinivasan was funded by Office of Naval Research grants. Matthew Mazloff received funding from NSF and NASA awards. Laura Cimoli received support through Schmidt Sciences, LLC, and the Advanced Research and Invention Agency (ARIA). Ali Mashayek was supported by an ONR grant. The authors declare no competing interests.
Publication Details
Authors: Joshua Lanham, Sarah Purkey, Kaushik Srinivasan, Matthew Mazloff, Laura Cimoli, and Ali Mashayek. | Affiliations: Department of Earth Sciences, University of Cambridge; Scripps Institution of Oceanography, University of California San Diego; Atmospheric and Oceanic Sciences, University of California Los Angeles; Department of Applied Mathematics and Theoretical Physics, University of Cambridge. | Journal: Communications Earth & Environment (a Nature Portfolio journal), Volume 7, Article 371 (2026). | Title: “Poleward migration of warm Circumpolar Deep Water towards Antarctica.” | DOI: https://doi.org/10.1038/s43247-026-03426-x | Received: October 2, 2025. Accepted: March 10, 2026. Published online: April 28, 2026.







