The AMOC Question: What the New Evidence Actually Changes

For years, the possibility of a collapse of the Atlantic ocean circulation system occupied a familiar mental category: serious, but distant. It was the kind of climate risk that resurfaced periodically and could still be filed away as too uncertain, too remote or too dependent on disputed models to influence present decisions.

That framing is becoming harder to maintain. Scientists have not suddenly agreed that collapse is imminent, and no new study has produced a reliable countdown. The more important change is that severe weakening now appears more plausible than the average of earlier model projections suggested, while the argument increasingly concerns how close such weakening would bring the system to a tipping threshold.

A 2026 study used present-day observations to constrain climate-model projections of the Atlantic Meridional Overturning Circulation, or AMOC. Its strongest-performing method estimated that the circulation could weaken by about half under an intermediate-emissions scenario by 2100. That is not a forecast that the AMOC has a fifty per cent chance of collapsing. It is still a result large enough to move the issue from distant catastrophe towards practical risk planning.

A system that appears stable—until it isn’t. Editorial image generated by the author.

What the New Study Actually Constrains

Climate models have long projected that the AMOC will weaken as greenhouse warming changes ocean temperature, rainfall, runoff and high-latitude freshwater input. The size of that weakening has varied enormously between models, making the direction of change more certain than its magnitude.

Under the intermediate SSP2-4.5 scenario, the unconstrained CMIP6 ensemble used in the 2026 observational-constraint study projected an average weakening of 32% by 2100, but with a 90% range extending 37 percentage points in either direction. That spread was broad enough to include both limited change and exceptionally severe decline.

The researchers applied four methods intended to identify relationships between how well models reproduce the present climate and how strongly they project the AMOC will weaken. The best-performing method incorporated a wide set of observable variables, including Atlantic surface temperature and salinity, and produced a projected weakening of 51% with an eight-percentage-point 90% range—approximately 43–59%.

The estimate is therefore not a statement that collapse is 51% likely. Nor is it a direct measurement of the future ocean. It is a narrower model projection produced by giving greater weight to present-day characteristics that appear relevant to future AMOC behaviour.

Its importance lies in which part of the model range survives the constraint. The models pointing towards stronger weakening appear to reproduce certain observed features of the Atlantic better than the more reassuring models. In particular, correcting a bias in South Atlantic surface salinity moves the constrained estimate towards a substantially weaker circulation.

Weakening, Tipping and Collapse Are Different Questions

The new result sits beside research pointing towards greater resilience. A Met Office-led modelling study published in 2025 concluded that complete collapse before 2100 was unlikely, even under severe warming and freshwater forcing. In those simulations, wind-driven upwelling in the Southern Ocean ensured that some sinking continued in the North Atlantic.

That finding did not suggest that the AMOC would remain healthy or unchanged. The circulation still weakened considerably, with consequences for Europe and other regions. The disagreement concerns whether the weakening crosses into a complete shutdown during this century, not whether greenhouse warming places the system under pressure.

The two studies can therefore be true at the same time. The AMOC could weaken much more than the average climate model previously suggested while retaining enough overturning to avoid a formally defined collapsed state by 2100. It could also cross a tipping threshold during this century while taking decades or longer to settle into its eventual weak state.

This distinction explains why AMOC coverage can appear to alternate between catastrophe and reassurance. Different studies may estimate the amount of weakening by a fixed date, the probability of crossing a threshold, the beginning of an irreversible transition or the time at which the circulation finally stabilises at very low strength. Those are related questions, but they are not interchangeable.

“Collapse” itself is not always used identically. Some researchers apply the term to the onset of a self-sustaining transition; others reserve it for the later condition in which overturning and heat transport have fallen to a very weak level. A threshold can be crossed before the most visible consequences have arrived.

What Is Actually Well Established

The IPCC’s baseline assessment remains more cautious than the strongest recent warnings. It concludes that the AMOC will very likely weaken during the twenty-first century under every assessed emissions pathway. It expresses medium confidence that this weakening will not involve an abrupt collapse before 2100, while assigning low confidence to the precise magnitude of the decline.

Three conclusions are therefore more secure than any collapse date. Continued weakening is expected. The amount of weakening remains uncertain. A complete collapse is not the central projection for this century, but it remains physically plausible and potentially severe enough to justify attention before its probability can be pinned down.

Direct measurement has improved the evidence without making it simple. The RAPID array has monitored the circulation continuously at approximately 26.5 degrees north since 2004. Other arrays, including OSNAP and SAMBA, now observe the system at different Atlantic latitudes.

The RAPID record shows large variation across days, seasons and years, as well as a downward trend over its first two decades. A recent synthesis of the observational evidence reports a weakening of roughly one sverdrup per decade over 2004–23, but also stresses that the record remains too short to separate the human-caused signal confidently from decadal variability.

This is not a contradiction. A decline can be present in the measured period without its causes being fully attributable from that record alone. Longer observations, indirect reconstructions and climate models remain necessary because the system varies substantially over timescales approaching the length of the available direct record.

Why Weakening Matters Before Collapse

The AMOC is often described as an ocean conveyor belt carrying warm surface water northwards and colder deep water southwards. The metaphor is useful, but the circulation is not one current driven by one mechanism. Winds, temperature, salinity, density and exchanges with the Southern Ocean all contribute to a system that transports heat, freshwater, carbon and nutrients through the Atlantic.

A weaker AMOC would not simply make Europe cold while the rest of the planet continues warming normally. Its regional effects would be superimposed on greenhouse warming. North-western Europe might still warm, but less than it would under a stronger circulation, while a major shutdown could produce pronounced regional cooling, especially in winter.

Substantial weakening could also shift tropical rainfall patterns, alter European precipitation and storm tracks, affect North Atlantic marine productivity and raise relative sea level along parts of the North American Atlantic coast. The scale and distribution of those consequences depend on how far the circulation declines and how the atmosphere and ocean respond.

These effects do not begin only after a formally defined collapse. Agriculture, water management, coastal infrastructure, energy demand and ecosystems respond to changes in temperature, rainfall, storms and sea level, not to the terminology researchers use for the circulation’s final state.

Nor would a transition resemble the switch in a disaster film. Crossing a tipping threshold means that the system may no longer return readily to its previous state, even if the original forcing is reduced. The subsequent movement towards a much weaker circulation could continue over decades or longer.

The policy difficulty is that the point of irreversibility may precede the moment when the full consequences become undeniable. Waiting for a completed collapse would therefore provide clarity only after the most useful period for prevention had passed.

Uncertainty Is Part of the Risk

Uncertainty around the AMOC has often been treated as though it mainly supported delay. The observational record was short, models disagreed and reconstructions of the past remained imperfect. That was sometimes translated into the more comfortable claim that the risk itself was probably remote.

Scientific uncertainty does not favour the safer outcome by default. It means that the system may prove more resilient than pessimistic projections—or less resilient than models that systematically stabilise it too strongly.

The 2026 review of AMOC collapse onset reflects that change in emphasis. It asks whether current Earth-system models adequately represent the circulation’s present stability and surveys attempts to estimate the probability of crossing a tipping threshold before 2100. It does not produce a universally accepted probability, but it treats the adequacy of reassuring model projections as part of the uncertainty rather than as its solution.

The responsible conclusion is not that collapse has become certain or even that one study has overturned the wider literature. It is that severe weakening has become harder to dismiss as an extreme edge case, while the probability and timing of a full transition remain unresolved.

That changes the decision problem. Society does not need a precise collapse date before it can prepare for weakening that is already expected, improve observation or reduce the forcing that makes the dangerous outcomes more likely.

From Distant Risk to Decision Horizon

There is no established engineering intervention capable of directly stabilising the AMOC at planetary scale. The practical lever remains limiting greenhouse warming and the associated changes in high-latitude heat and freshwater balance. Lower-emissions pathways do not guarantee that the circulation remains unchanged, but they reduce the magnitude and duration of the pressure applied to it.

Greenland melt belongs to the same problem rather than forming a separate policy category. Additional freshwater can lower North Atlantic salinity and density, making deep-water formation more difficult. Slowing that contribution means limiting the warming that drives the ice loss.

Monitoring also matters. RAPID, OSNAP, SAMBA and related programmes cannot prevent weakening, but they can test projections, distinguish regional mechanisms and eventually separate long-term change from natural variability more confidently. A circulation that varies across decades cannot be understood through intermittent measurements or short funding cycles.

Preparation should not wait for proof of complete collapse. European energy systems, agriculture, water management and infrastructure can be tested against scenarios involving altered winter temperatures, drier conditions in some regions, shifting rainfall and different storm patterns. Coastal planning in North America can incorporate the regional sea-level effects associated with a weaker circulation.

The new evidence does not turn the AMOC into a known catastrophe with a scheduled date. It changes something more practical. A roughly fifty per cent weakening by the end of the century is now a serious constrained projection under an intermediate pathway, not merely the alarming edge of a very broad model range.

The AMOC may remain functional throughout this century. It may cross a threshold whose full consequences unfold later. It may prove more resilient than the most pessimistic simulations or less resilient than the models that currently appear reassuring.

Uncertainty remains. What has changed is the legitimacy of using that uncertainty as a synonym for distance or safety.