The sea is losing its breath, without a sound. On 2026-06-30, a review paper led by the Scripps Institution of Oceanography appeared in the journal Limnology and Oceanography. Its message is heavy: the amount of oxygen dissolved in water is falling fast enough to push Earth outside its "safe operating space," and some of those changes may persist for centuries and prove irreversible within a human lifetime [source: Limnology and Oceanography, 2026].
We are used to visible threats such as air pollution or rising seas. But aquatic deoxygenation — the loss of dissolved oxygen in oceans, lakes and rivers — advances quietly, below the surface. This article sorts out what has actually been measured, what is being blamed for it, and why scientists have begun to describe it in the language of "planetary boundaries." And it draws a clear line between what has been observed and what is still only proposed or projected.
A note on how this article is built. The 2026 review's full text sits behind a paywall, so its claims are taken from the paper's own published record and from the official releases of the institutions that published it [source: Scripps/UC San Diego, 2026]. Every figure below is quoted in the form its source reports it, with the baseline year attached, and nothing is converted into a different unit. Where a statement is a measurement, it is presented as one; where it is a scenario projection or a researchers' proposal, it is labeled as such.
In this article
- The decline, in numbers
- Why the oxygen is leaving — two main drivers
- The planetary-boundary lens
- What is at stake
- Can it be undone?
- Conclusion — what to watch
The decline, in numbers
The global average, and what it hides
Start with the measured values. Observations show that over the past half-century, oxygen in the water has genuinely declined. Across the global open ocean, dissolved oxygen has fallen by roughly 2% since 1960 [source: Nature Ecology and Evolution, 2024]. Two percent sounds small, but remember it is an average across the entire world ocean. Zoom in to specific regions and the losses are far larger.
That figure is also unusually well corroborated. Three independent primary sources report a global open-ocean decline of about 2%, though each attaches it to a slightly different starting point: the 2024 analysis dates it from 1960 [source: Nature Ecology and Evolution, 2024], Scripps dates it from the 1960s [source: Scripps Institution of Oceanography, 2026], and the IUCN dates it from the 1950s [source: IUCN, 2019]. The baselines differ; the direction and the magnitude do not. That convergence is why percent declines, each carrying its own baseline year, do the work in this article.
Where the losses run deepest
Fresh water and regional seas show steeper numbers. According to the same analysis, lakes have lost more than 5% of their oxygen since 1980, reservoirs about 18%, and deep water off central California has dropped by as much as roughly 40% in recent decades [source: Nature Ecology and Evolution, 2024]. Scripps likewise notes that some ocean regions have already seen oxygen declines of 20-50% [source: Scripps Institution of Oceanography, 2026].
These are not four measurements of the same thing. Lakes and reservoirs are fresh water, the California figure is deep water off one stretch of coast, and the windows differ — 1960 for the open ocean, 1980 for lakes and reservoirs, recent decades for the California deep water [source: Nature Ecology and Evolution, 2024]. The reported forms differ too: lakes are given as more than 5%, a floor, while the California figure is given as up to about 40%, a ceiling. Read as a list, they show how uneven the loss is; read as a ranking, they would mislead. The analysis behind them was led by Kevin Rose of Rensselaer Polytechnic Institute, who is also a coauthor of the 2026 review [source: Limnology and Oceanography, 2026].
The spreading zones, and how they are counted
The low-oxygen zones themselves are widening. The ocean naturally contains a low-oxygen band called the oxygen minimum zone (OMZ) at depths of roughly 100-1,000m, and deoxygenation is enlarging it. In the open ocean, low-oxygen water has expanded by about 4.5 million km² [source: Scripps Institution of Oceanography, 2026]. Along coasts, the number of sites where oxygen has fallen too low to sustain life has passed 500 [source: Science, 2018]. By the count of the International Union for Conservation of Nature (IUCN), low-oxygen sites grew from about 45 before the 1960s to roughly 700 by 2011, and the volume of completely oxygen-free (anoxic) water has quadrupled since the 1960s [source: IUCN, 2019].
The two site counts are not in competition; they count different things. The 500-plus figure tracks coastal sites where oxygen has fallen too low to sustain life, the kind driven by nutrient loading [source: Science, 2018], while the IUCN's roughly 700 is a wider tally of low-oxygen sites as of 2011 [source: IUCN, 2019]. The 4.5 million km² is a third kind of quantity again — an area of open-ocean water, neither a count nor a percentage [source: Scripps Institution of Oceanography, 2026]. Keeping the three apart matters, because adding them together is exactly how a real trend acquires a false precision.
Measured, and projected
One distinction is worth fixing here. The figures above are, for the most part, measured observations. Projections of the future are a different kind of statement. The IUCN estimates that if warming continues on a business-as-usual path, oxygen could fall by a further 3-4% by 2100 [source: IUCN, 2019]. That number should be read for what it is — a scenario-based projection, not a measurement.
The distinction has a practical form. A measurement comes with a baseline year — 2% since 1960 — and a projection comes with a scenario and a horizon: a further 3-4% by 2100, if warming continues on a business-as-usual path [source: IUCN, 2019]. The projected loss is additional, stacked on top of what has already been measured, not a restatement of it. And business-as-usual is a named assumption about the emissions path, not a forecast of what will happen. In the rest of this article, a figure without a scenario attached is an observation.
Why the oxygen is leaving — two main drivers
Scientists point to two principal causes: warming and nutrient pollution [source: Scripps/UC San Diego, 2026]. What matters is that the two operate on different stages. Warming works mainly across the vast open ocean; nutrient pollution works mainly along the coasts.
Warming — the open-ocean problem
The first cause is physics. Warmer water holds less oxygen. As temperature rises, solubility drops, so the surface layer can simply carry less oxygen to begin with [source: Scripps Institution of Oceanography, 2026]. A second mechanism compounds this. When the surface warms, the water column resists mixing — a strengthening of stratification — and the path that carries surface oxygen down into the deep sea is blocked. The movement and ventilation of deep water slows [source: Limnology and Oceanography, 2026].
This is a passage to read carefully for the difference between correlation and causation. That warm water holds less oxygen is a law of physics confirmed in the laboratory. Yet attributing a precise share of any given region's observed oxygen loss to warming is a harder problem, one that must also weigh natural variability. The review names warming as a "main driver" because this physical mechanism and broad observations point the same way — not because every local change can be pinned on a single cause [source: Limnology and Oceanography, 2026].
One more mechanism belongs on the list. Warming does not only reduce how much oxygen water can hold; the IUCN notes that it also raises the demand side, because oxygen consumption in the water increases as it warms [source: IUCN, 2019]. And the 2026 review names a third driver alongside warming and nutrient pollution: changes in the movement and ventilation of deep water, the process that carries oxygen from the surface into the ocean's interior [source: Limnology and Oceanography, 2026]. Supply falling, demand rising, and the delivery route slowing are three separate levers pushing in the same direction.
Eutrophication — the coastal problem
The second cause is pollution. When nutrients such as nitrogen and phosphorus — washed in from farm fertilizer and sewage — flood coastal waters, phytoplankton bloom explosively. This eutrophication carries a reckoning. When the mass of plankton dies, the microbes that decompose it consume large amounts of oxygen from the water, creating low-oxygen "dead zones" where life struggles to survive [source: Scripps Institution of Oceanography, 2026]. A 2018 synthesis in Science likewise concluded that open-ocean declines are driven by warming, while coastal declines are led by agricultural runoff and sewage pollution [source: Science, 2018].
It is worth walking the chain slowly, because each link is a different kind of process. Nutrients arrive from fertilizer and sewage; phytoplankton bloom on them; the bloom dies; microbes decompose the dead mass; that decomposition draws oxygen out of the water; and what is left behind is a dead zone [source: Scripps Institution of Oceanography, 2026]. Nothing in that sequence requires the ocean to be warmer, which is why the coastal problem has its own geography and its own remedy. The 2018 Science paper that reported the coastal site count was itself a synthesis produced through the GO2NE network [source: Science, 2018].
The planetary-boundary lens
What a planetary boundary is
What makes this review new is that it reframes the oxygen problem beyond a single phenomenon, through the lens of "planetary boundaries." A planetary boundary is a concept that defines the limits within which humanity can live safely, expressed as nine Earth-system processes. In a 2023 update, six of the nine boundaries — climate change, biosphere integrity, the nitrogen and phosphorus cycles, land-system change, freshwater change, and novel entities — were assessed as already outside the safe range [source: Science Advances, 2023].
Two features of the framework matter for reading the oxygen argument. It is not new — the boundaries were first proposed in 2009 by Rockström and colleagues, and the 2023 assessment is the third major revision, the first to put a quantified value on all nine [source: Science Advances, 2023]. And crossing a boundary is not a declaration that catastrophe has arrived. The safe operating space marks the range within which humanity can act safely; beyond it lies a zone of increasing risk [source: Science Advances, 2023]. The framework describes rising danger, not a switch that has been flipped.
The case for a tenth boundary
Aquatic deoxygenation is not yet on that list of nine. But a 2024 study in Nature Ecology & Evolution was the first to formally propose deoxygenation as a candidate planetary boundary [source: Nature Ecology and Evolution, 2024], and the 2026 Scripps review takes that argument forward. Its core point is that deoxygenation does not act alone. Oxygen loss is entangled with other boundaries — climate change, biogeochemical cycles, ocean acidification — trading influence through feedbacks, and the review synthesizes evidence that this interplay pushes the planet toward the "unsafe" side [source: Limnology and Oceanography, 2026].
The proposal has a lineage worth stating plainly. The 2024 Nature Ecology & Evolution paper, led by Kevin Rose, was the first to put deoxygenation forward as a planetary boundary [source: Nature Ecology and Evolution, 2024]; the 2026 review is the follow-on synthesis, with Erica M. Ferrer — a Scripps alumna, now a postdoctoral researcher at NCEAS at UC Santa Barbara — as lead author, and the Scripps biological oceanographer Lisa A. Levin as senior author [source: Limnology and Oceanography, 2026]. The same research community is making the case twice, two years apart, with more evidence assembled the second time.
What has not been settled
Here one must separate a proposal from a verified fact. Deoxygenation has not been officially adopted as a "tenth planetary boundary." This is the researchers' proposal and argument. Moreover, there is as yet no agreed quantitative safe threshold for oxygen loss — no number such as "danger begins past X percent." The "unsafe space" the review describes is therefore not a reading on a precise gauge but a qualitative warning grounded in accumulated observations and in the links to other boundaries [source: Limnology and Oceanography, 2026]. It is also worth stating that this paper is not a study of new field measurements but a review synthesizing existing literature [source: Scripps/UC San Diego, 2026].
Erica Ferrer, the review's lead author, sums up the intent this way: the study is "designed to elevate the profile of aquatic deoxygenation as a global threat and show that it does not operate in isolation," and "adding aquatic deoxygenation to the Planetary Boundaries framework will help us understand its impacts on Earth system stability" [source: Scripps/UC San Diego, 2026].
Ferrer states the underlying premise more simply elsewhere in the same release: "The health and stability of our planet depends on the health and stability of aquatic ecosystems, which need oxygen to function normally" [source: Scripps/UC San Diego, 2026]. That is a claim about why the question is worth asking. Whether the answer belongs inside the planetary-boundary framework, and at what numerical threshold, is the part still open.
What is at stake
From microbes to sharks
Oxygen is the basic condition of life in the water. So when it falls, the ripple runs through the entire food web, from microbes to fish to sharks [source: Scripps/UC San Diego, 2026]. At the level of the individual organism, low oxygen slows growth, impairs reproduction, and raises vulnerability to disease. In oxygen-starved water, animals are driven toward the layers they can still survive in, and this habitat compression crowds species toward the surface, exposing them more readily to predators and fishing fleets [source: Scripps Institution of Oceanography, 2026].
The reach extends past the animals that feel the oxygen directly. The 2026 review notes that marine mammals are affected indirectly, through changes in their prey and their habitat, and that oxygen loss disturbs the biological and chemical processes that help regulate Earth's climate [source: Limnology and Oceanography, 2026]. Habitat compression is a useful illustration of how the harm actually arrives: the water does not become uninhabitable everywhere at once, it becomes thinner as a habitat, and the squeeze itself is the damage [source: Scripps Institution of Oceanography, 2026].
Winners, losers, and the catch
The makeup of ecosystems shifts, too. As oxygen falls, species tolerant of low oxygen — jellyfish, some microbes, squid — gain an edge, while fish that need a lot of oxygen are pushed out [source: IUCN, 2019]. This is not unrelated to the human dinner plate. Naturally low-oxygen upwelling regions support about one-fifth of the world's wild marine catch, so oxygen shifts in these waters are a direct threat to fisheries [source: IUCN, 2019].
Two things are worth separating in that picture. A shift in species composition is a redistribution rather than a simple subtraction — jellyfish and squid do not vanish, they gain ground [source: IUCN, 2019]. But upwelling regions are where redistribution meets people directly, because those waters are naturally low in oxygen to begin with and already carry about a fifth of the world's wild marine catch [source: IUCN, 2019]. The 2018 Science synthesis framed the long-run risk in similar terms: continued decline is not sustainable, and the costs fall on ecosystems and on the societies and economies attached to them [source: Science, 2018].
The loop back to climate
The problem does not stop there; it loops back to the climate. Deep water stripped of oxygen produces more greenhouse gases such as nitrous oxide, carbon dioxide and methane [source: IUCN, 2019]. Warming takes oxygen away, and the oxygen-depleted ocean in turn releases greenhouse gases that stoke further warming. This is precisely why the review stresses that deoxygenation must not be treated as an isolated problem [source: Limnology and Oceanography, 2026].
That is what a feedback means in practice, and it is the structural reason the review resists treating oxygen as a standalone environmental issue. If oxygen loss is driven in part by warming, and the oxygen-depleted water in turn releases gases that drive warming, then the size of the problem cannot be read off the oxygen numbers alone [source: Limnology and Oceanography, 2026].
Can it be undone?
The slow clock of the deep ocean
The heaviest passage concerns recovery. The review warns that some of today's changes could persist for centuries and may not be reversible within a human lifetime [source: Limnology and Oceanography, 2026]. Warming-driven oxygen loss in the open ocean in particular follows the slow clock of deep-water circulation, so it will not rebound the moment the cause is halted.
Two timescales are stacked here. Nutrient inflows can be turned down within years; the circulation that carries oxygen into deep water turns over on a far longer cycle. So the same word — recovery — describes two very different waits, depending on which body of water is being talked about. The review's warning about changes persisting for centuries applies to the slower of the two [source: Limnology and Oceanography, 2026].
Where the levers still work
Not all waters are alike, though. Unlike the warming-driven changes of the open ocean, coastal low-oxygen caused by eutrophication is comparatively more tractable. There are reported cases where cutting the inflow of nutrients has locally improved oxygen conditions. The response therefore splits in two. For the open ocean, the key is to cut greenhouse-gas emissions and slow warming itself; for the coasts, it is to manage the nutrient load from fertilizer and sewage. Neither is a quick fix, but because the causes divide in two, so do the levers for acting on them.
The balance is worth keeping in both directions. The reported coastal improvements are real, and they are local; they are not evidence that open-ocean, warming-driven loss can be reversed on the same schedule, and they do not soften the review's warning about irreversibility within a human lifetime [source: Limnology and Oceanography, 2026]. What they do show is that one of the two causes responds to management on a human timescale — which makes it the nearer of the two levers, not the more important one.
Conclusion — what to watch
What the evidence supports
To sum up: that oxygen in the water has genuinely fallen over the past half-century is an observed fact. Warming is named as the main driver in the open ocean, nutrient pollution along the coasts. And the 2026 Scripps review recasts the phenomenon as an Earth-system-scale threat, entangled with the other planetary boundaries [source: Limnology and Oceanography, 2026]. But the status of a "tenth boundary," or the verdict of irreversibility, should be remembered as the researchers' proposal and assessment rather than as settled fact.
It helps to hold the three tiers apart one last time. Measured: oxygen in the open ocean, in lakes, in reservoirs and in deep water off central California has fallen, each figure dated to its own baseline year. Projected: a further 3-4% by 2100, conditional on a business-as-usual warming path [source: IUCN, 2019]. Proposed: deoxygenation as a tenth planetary boundary — argued from synthesis rather than settled by adoption, with no agreed quantitative threshold attached to it yet [source: Limnology and Oceanography, 2026].
Three things to watch
What should you watch from here? First, whether deoxygenation is actually incorporated into the planetary-boundary framework, and whether the discussion of a quantitative threshold for oxygen loss advances. Second, how global observing networks such as Argo update the oxygen trend. Third, whether the two policy levers — coastal nutrient management and greenhouse-gas reduction — actually translate into recovered oxygen. How fast the ocean's breathing grows labored will, in the end, depend on how hard we pull those two levers.