How much should you worry about a collapse of the Atlantic conveyor belt?

Several high-profile research papers have brought renewed attention to the potential collapse of a crucial system of ocean currents known as the Atlantic Meridional Overturning Circulation, or AMOC, as we discussed in& part one& of this two-part post. Huge uncertainties in both the timing and details of potential impacts of such a collapse remain. Even so, scientists warned in a recent open letter (see below) that “such an ocean circulation change would have devastating and irreversible impacts.”

The AMOC is a vast oceanic loop that carries warm water northward through the uppermost Atlantic toward Iceland and Greenland, where it cools and descends before returning southward. The Gulf Stream, which carries warm water from the tropics toward Europe, is an important section of this loop. Studies of ancient climates reveal that in the distant climate past, AMOC has gone through cycles of collapse that lasted hundreds of years. As we saw in& part one, these “off” cycles of the AMOC brought massive environmental changes that would lead to tremendous havoc should they descend on today’s complex, interconnected world.

In& part one, we looked at observations from the North Atlantic that suggest a gradual weakening in AMOC strength over the last few decades, but only a marginally significant drop over the past 40 years of AMOC monitoring, including the largest near-surface component, the Gulf Stream.

By no means does this rule out the possibility of a forthcoming AMOC collapse. New research is leading to startlingly specific time frames for when the AMOC might collapse. These studies aren’t without controversy, as we’ll see below. But collectively, they’ve raised the profile of AMOC &-& and also raised fears that the initial impacts of AMOC collapse could manifest within the lifetimes of many of us.

Stefan Rahmstorf at the Potsdam Institute for Climate Impact Research is an eminent researcher who’s studied AMOC and its various modes for more than 30 years. In October 2024, discussing the specter of AMOC collapse, Rahmstorf& warned:

Even with a medium likelihood of occurrence, given that the outcome would be catastrophic and impacting the entire world for centuries to come, we believe more needs to be done to minimize this risk.

Hunting for advance notice

The end-stage impacts of a collapsing AMOC wouldn’t arrive until long after the causative wheels were set into motion. So several of the new studies have looked into ways to get early notice that a tipping point is at hand &- basically, a flashing red light that would imply a century-long AMOC collapse was about to get underway, which might, in turn, make some types of long-range preparation more feasible. It’s& roughly akin to a cardiac scan that might reveal adhesions of heart-dangerous plaque decades before any full blockage.

Papers from European research groups that have helped draw attention to AMOC this decade include the following:

Observation-based early-warning signals for a collapse of the Atlantic Meridional Overturning Circulation& (Boers, Nature Climate Change 2021). Led by the Potsdam Institute’s Niklas Boers, this study used eight independent indices gleaned from Atlantic Ocean sea surface temperature and salinity data. One goal was to piece together a broader picture of AMOC behavior since the late 1800s, given that organized observations of AMOC’s three-dimensional flow began only in the 2000s.

The overall conclusion of Boers and coauthors: “AMOC could be close to a critical transition to its weak circulation mode.”

Warning of a forthcoming collapse of the Atlantic meridional overturning circulation& (Ditlevsen & Ditlevsen, Nature Communications 2023). This study emerged from the brother-and-sister team of climate scientist Peter Ditlevsen and statistician Susanne Ditlevsen, both at the University of Copenhagen. Using a broad set of assumptions about how AMOC works, the team focused on a single variable to represent the complex beast that is AMOC. That variable was built from monthly global variations in sea surface temperature as compared to such variations in the subpolar Atlantic, where increased meltwater is already counteracting global warming’s influence.

This 150-year series of data points (1870-2020) was then analyzed for statistical elements that have been linked elsewhere to early warning signs of tipping behavior &-& especially increasing variance and increasing autocorrelation, two measures that each suggest movement away from steady-state behavior.

The Ditlevsens came to an arresting conclusion: “We estimate a collapse of the AMOC to occur around midcentury under the current scenario of future emissions.” Their analysis found a median tipping year of 2057, with an expansive 95%-confidence-level range from 2025 to 2095.

As with any forecast, the team acknowledged this one is only as solid as the multiple assumptions about AMOC on which it’s built. They added, “given the importance of the AMOC for the climate system, we ought not to ignore such clear indicators of an imminent collapse.”

Physics-based early warning signal shows that AMOC is on tipping course& (van Westen et al., Science Advances, 2024). In contrast to the simplify-where-possible approach of the Ditlevsens, postdoctoral researcher René van Westen of Utrecht University, Netherlands, who is working through spring 2025 at the Royal Netherlands Meteorological Institute, and colleagues used one of the world’s most sophisticated tools for depicting Earth’s climate in 3D detail: the Community Earth System Model. A number of previous studies using complex, high-end models that couple the atmosphere and ocean hadn’t typically produced the on-to-off switch of an AMOC collapse &- even for simulations that span many hundreds of years &- because they’re designed with various computational guardrails (including their handling of ocean salinity) that inhibit such a drastic change from occurring.

To get around this challenge, van Westen and colleagues carried out an idealized experiment, designed not as a literal forecast but to provide insights into what signals might precede an AMOC collapse. They did so by keeping greenhouse gas values at preindustrial levels but with a steadily increasing infusion of polar meltwater into the North Atlantic, building to dozens of times more than the current Greenland melt rate. After 1,750 years of model time, at which point six Greenlands’ worth of ice had entered the ocean, a classic AMOC collapse began.

By the time the collapse was done, the impacts included Arctic sea ice spreading far into the North Atlantic and wet and dry seasons trading places in the Amazon. The average February temperature plummeted by a bone-chilling 15 degrees Celsius (27 degrees Fahrenheit) in London and by around 3°C (5°F) across the mid-Mississippi Valley of the United States.

Figure 1. Trend in annually averaged surface (two-meter) air temperature, in degrees Celsius, over the century spanning AMOC collapse in the idealized model of van Westen et al., 2024. Circles denote where trends are below the 95% significance level. Warming spans most of the Southern Hemisphere, with extreme cooling toward higher latitudes of the Northern Hemisphere. The “1750-1850” at top refers to the number of years since the start of this idealized simulation of preindustrial climate, rather than to real-world years. (Image credit:& van Westen et al., 2024& / licensed under& CC BY 4.0)

The new modeling also lent support to a different early-warning indicator of imminent collapse, this one toward AMOC’s south end. As freshwater in the uppermost South Atlantic moves north and passes through the tropics, it gradually gets warmer and saltier. The authors found that a long-term decline in northward transport of uppermost South Atlantic water, as measured at latitude 34 degrees south, gave more reliable early warning than the indicators used by Ditlevsen and Ditlevsen. They also noted that in recent decades the transport at this latitude has been decreasing (i.e., not good), but that observations there aren’t yet sufficiently long to gauge the timing of a tipping point.

“Some journalists were very much angling for a statement about when the tipping point would occur exactly,” van Westen& told& the Utrecht University news site, “but I just kept on saying: ‘We don’t know.’”

Research& still in review, led by Utrecht University doctoral candidate Emma Smolders and including van Westen, draws on the physics-based South Atlantic index to come up with an AMOC-collapse timing range for a high-emissions scenario that’s similar to the range found by Ditlevsen and Ditlevsen, albeit somewhat narrower.

Uncertainties too large to predict tipping times of major Earth system components from historical data& (Ben-Yami et al., Science Advances, 2024). This research team &-& led by doctoral researcher Maya Ben-Yami at the Technical University of Munich and including Niklas Boers from the 2021 paper above &- issued several notes of caution about specific forecasts of a tipping time for the AMOC.

“Although the utility of such predictions, if robust, would be undeniable, the problem lies in the multiple levels of uncertainty inherent to such extrapolations from historical data,” they wrote. Among the issues explored by Ben-Yami and colleagues:

• Techniques optimized to predict tipping times are built on the assumption that a tipping point exists. As a result, they can be prone to what Ben-Yami and colleagues call “false positives.”

• Since the AMOC is a product of multiple flows with many complexities, it may not behave as expected from the strongly simplifying assumptions that underly tipping-time predictions.

• As observations in data-sparse regions become more plentiful over time, the century-scale analyses of past climate &-& which carefully fill in gaps across the earlier, more limited data &-& may show increased variance unrelated to the AMOC simply because of how the data are filled in.

None of these critiques guarantee that the AMOC won’t end up behaving as predicted in some studies. It just means we still lack enough information to predict with confidence when the AMOC will tip. (A response from Ditlevsen and Ditlevsen to the Ben-Yami et al. paper is now in review.) In an email, Ben-Yami said:

My hope is that future work is careful when trying to predict a tipping time for the AMOC from historical data. This means that new methods need to be rigorously tested, for example by applying them to test data from different scenarios, to different observational data sets, and to a large range of models. And personally I would rather see research focused on improving AMOC understanding, modeling, and observation rather than trying to predict an exact tipping time … The AMOC is going to weaken with future warming, regardless of whether or not it has already weakened in the last century. The question is more about how much and how fast it will weaken.

Similarly, van Westen emphasized in an email:

Moderate AMOC weakening (of about 30%) can already cause climate impacts, and these impacts become even larger when the AMOC declines furthers (or collapses). To limit this weakening and its impacts, we need to reduce our emissions. This will also mitigate other (disastrous) climate impacts, such as more intense heat waves, the frequency of extreme weather, and additional sea level rise.

The IPCC’s carefully parsed wording on AMOC

The& Intergovernmental Panel on Climate Change, or IPCC, issues major assessments only every six to 10 years. It’s a deliberate pace that hasn’t yet caught up with the latest AMOC research. The IPCC’s tilt toward consensus science makes its reports solid and authoritative and reduces the risk of conclusions that whipsaw back and forth &-& but it also means that it can take years for a paradigm shift to move the assessment needle.

The first several IPCC assessments made only general statements about AMOC. The fourth and fifth assessments (published in 2007 and 2013) concluded that during the 21st century, an AMOC slowdown was “very likely,” but a large or abrupt transition was “very unlikely.”

The sixth assessment, released in 2021, reiterated those “very likely” and “very unlikely” themes. The report hewed closely to the gradual decline portrayed in most of the global models used for the assessment, which can capture slow AMOC change but aren’t tailored to allow for collapse-style changes. An average decrease of roughly 30% in AMOC strength was deemed likely by 2100 under most emission scenarios. The amount of decline is largely independent of emission changes until around the 2060s; after that point, the scenarios with greater emissions lead to a more rapid decline.

Figure 2. AMOC transport relative to 1995&-2014 (defined as maximum transport at 26°N), in Sverdrup units (Sv). On average, the current AMOC strength is about 18 Sv. Assessed observational change from 2004 to 2018 (teal) is drawn from the RAPID array of undersea cables, smoothed with a 12-month running mean. Shading around the mean shows the 12-month running standard deviation around the mean.& A subsequent study, discussed in part one of this series, has reduced the downward trend in RAPID data between 2004 and 2022 by about 40% after accounting for shifts in the geomagnetic field that affected the cable data. (Adapted from Figure TS.11(b), “Climate Change 2021: The Physical Science Basis,”& IPCC Sixth Assessment Report)

This latest IPCC report also included some stronger bet-hedging. It announced there was only “medium confidence” that an AMOC collapse wouldn’t happen this century, with that lukewarm confidence level stemming from “known biases in the climate models used for this assessment.” The report added:

… such a collapse might be triggered by an unexpected meltwater influx from the Greenland Ice Sheet. If an AMOC collapse were to occur, it would& very likely& cause abrupt shifts in the regional weather patterns and water cycle, such as a southward shift in the tropical rain belt, and could result in weakening of the African and Asian monsoons, strengthening of Southern Hemisphere monsoons, and drying in Europe.

Sounding an& alarm& with renewed vigor

In the wake of the newly specific research on early-warning methods and AMOC collapse timing, several of the scientists involved have intensified their outreach efforts, especially in 2024.

The Potsdam Institute’s Stefan Rahmstorf weighed in 20 years ago during AMOC’s biggest moment in pop culture to date: the 2004 blockbuster film “The Day after Tomorrow.” That movie depicted an AMOC shutdown from hell that buried New York under immense, relentless snow drifts and flung tornadoes at Los Angeles, among other lurid outcomes. When the movie came out, Rahmstorf said:

Luckily it is extremely unlikely that we will see major ocean circulation changes in the next couple of decades … at least most scientists think this will only become a more serious risk towards the end of the century.

Two decades later, in early 2024, Rahmstorf& said in a comment& on the website he cofounded, RealClimate:

Indeed I have in recent years changed my mind about the likelihood [of an AMOC collapse], mainly based on the observations-based early warning signal papers … and on the increasing evidence that models have an AMOC which is systematically too stable, when compared to observations-based stability indicators.

In October 2024, Rahmstorf joined Niklas Boers, Peter Ditlevsen, and 41 other climate scientists from Europe, Australia, Asia, and North America to release a widely publicized open letter to the Nordic Council of Ministers (see& PDF). This letter pointed to what the scientists termed a “serious risk of a major ocean circulation change in the Atlantic. A string of scientific studies in the past few years suggests that this risk has so far been greatly underestimated. Such an ocean circulation change would have devastating and irreversible impacts especially for Nordic countries, but also for other parts of the world.”

Alluding to the latest IPCC assessment, they wrote:

The purpose of this letter is to draw attention to the fact that only “medium confidence” in the AMOC not collapsing is not reassuring, and clearly leaves open the possibility of an AMOC collapse during this century. And there is even greater likelihood that a collapse is& triggered& this century but only fully plays out in the next.

Signaling that even the latest findings are still works in progress, they added: “Despite significant research into the possibility and mechanisms of a collapse, the probability of such an occurrence remains highly uncertain.”

What do we do now?

It’s not yet game over for the AMOC. As we discussed in& part one& of this post, the actual observations of 3D AMOC flow through the North Atlantic are still too brief to separate natural variations from long-term trends. And a modest decrease observed over the last 40 years& has just been revised& in an even more modest direction.

More papers on AMOC evolution, including follow-up studies by several of the authors above, are already in the pipeline. And more forecasts of AMOC-collapse timing are surely on the way. Some 100 presentations and posters related to AMOC were featured in December at the American Geophysical Union’s& 2024 annual meeting& in Washington, D.C.

As for now, we’re left with a dystopian, world-changing possibility &- but one that’s still shrouded by enough observational and methodological uncertainty to make the answer to “When will it happen?” both less concrete and more worrisome than we’d like.

There are myriad ways to adapt to climate change that’s already unfolding or imminent, many of them spotlighted here at Yale Climate Connections, such as moving to higher ground and making communities more heat-resilient. It’s far less clear how one might prepare at this point for a possible AMOC collapse, given the huge uncertainties in both the timing and details of potential impacts.

Perhaps the best way to channel AMOC anxiety would be to work toward emission reductions that could help prevent or at least forestall an AMOC collapse in the first place. That can include everything from demanding climate action on the local, state, and federal levels to talking climate with neighbors and colleagues, as well as inspiring others by example through a lower-emissions lifestyle. Here too, you’ll find& plenty of ideas& at Yale Climate Connections.

As René van Westin put it in the& Utrecht University article& cited above:

Climate scientists have been saying for a long, long time that the climate is changing and that this change has harmful consequences, so we need to do something. I see our study as an extra page on top of this pile of scientific articles. The evidence is already overwhelming. But it would be great if it would nudge politicians in the right direction because there needs to be more and more urgent climate policy.

Jeff Masters contributed to this post.