The drivers of AMOC’s destabilization
And what developing more comprehensive prediction, preparedness, and risk reduction and prevention capacity for the future requires from us today.
Carbon Brief recently identified a 74-fold increase in the number of news articles that mention “Atlantic Meridional Overturning Circulation” (AMOC) between 2006 to 2024. For decades, scientists have known AMOC is sensitive to climate changes and global warming, and since the 1980s, climate models have forecast that AMOC could slow down as global warming mounts and climate change intensifies. Until recently, these risks were not, however, discussed widely outside climate science circles.
That’s now changed. Last year, Iceland declared the potential collapse of AMOC a national security threat, the first instance of a national government elevating a specific climate-related dynamic to a national security threat at the national-security-council level. In May of this year, the Finnish government similarly made a formal call for greater AMOC preparedness, adaptation planning, and for the creation of a related fund for better monitoring and early‑warning systems. Some of the largest financial institutions, like JP Morgan, are also paying more attention to systemic and potentially catastrophic climate risks generally, including tipping point risk, and tipping point risks inherent to AMOC.
Most coverage of AMOC readily identifies climate change as the main driver of risk but rarely extends to an analysis of what we might actually do about that, beyond hoping that conventional climate mitigation and adaptation efforts accelerate. This absence is remarkable considering the stakes, and betrays the fact that there are few dedicated efforts, whether in academia, government, or otherwise, to advance prediction, preparedness, or prevention efforts for AMOC specifically. This article picks up that thread to examine what a comprehensive risk response effort for AMOC should entail, arguing that four pillars of advancement, focused on the science and understanding of AMOC itself, monitoring and measurement capacity and infrastructure, translational R&D for potential risk-reduction strategies, and governance architectures to manage future decisionmaking, must be developed in parallel, starting now, despite the presence of high uncertainty.
What AMOC is and why it matters
AMOC is a massive system of oceanic currents that acts like a conveyor belt, transporting warm surface water northward through the Atlantic. For millennia, it has regulated much of Earth’s climate, keeping north‑western Europe much more mild and temperate than its latitude would otherwise indicate and shaping rainfall patterns across the tropics. A core component of its function lies in a process called deep-water formation (also “overturning,” as in the “overturning circulation”), which is driven predominantly by surface cooling and evaporation of warm, salty subtropical waters:
Warm, tropical water flows along the ocean’s surface from the equator towards the North Atlantic.
As this water travels, it releases heat to the atmosphere and some of it evaporates, making it cooler and more saline (saltier).
When water is colder and saltier, it sinks more easily.
At northern latitudes, waters become cold and salty enough to sink several kilometers to the ocean floor.
This denser water then travels southward again, creating a pull (“deep return flow”) that draws more warm surface water north to replace it.
In time, the deep water rises back to the surface through upwelling in the Southern Ocean, where it warms and rejoins the northward flow, completing the cycle.
The manner in which deep-water formation and deep return flow reciprocally reinforce each other creates a self-sustaining loop that contributes to AMOC.
AMOC exerts a profound influence on Earth’s climate and many of its subcomponents. If it were to weaken considerably, the potential impacts would be severe and global, both across Earth’s physical climate system and many societal systems. Specifically, a significant loss of strength in AMOC, or its complete collapse, could yield:
Increased temperature difference between northern and southern Europe of up to 4° C (7° F), which in turn would make extreme weather more likely and disrupt agricultural production and other sectors like energy and power grid infrastructure.
An additional 0.5 meters of sea level rise, potentially imperiling trillions of dollars in real estate and infrastructure value (the costs of sea level rise scale nonlinearly).
Disrupted global monsoon patterns, weakening the West African and Asian monsoons and potentially intensifying the South American monsoons by up to 40%, threatening agricultural production and food security for billions.
Accelerated global migration, as the above impacts, whether on their own or in combination, would drastically alter the habitability of regions, many of which are currently densely populated.
Exacerbated geopolitical conflict, as resource scarcity, global migration, and economic instability yield both intra- and inter-regional conflicts.
Any of these consequences in itself is certainly cause for considerable concern. The confluence of all of them could match or exceed any and all historical precedents in catastrophic risks. Laybourn et al. (2023) specifically analyzed an AMOC collapse scenario and its consequences to define the concept of “derailment risk,” namely, a scale and complexity of disruption that completely overwhelms society’s capacity to respond to it.
The drivers of AMOC’s destabilization
The main sources of pressure on AMOC currently are those that cause waters to warm and freshen (reduced salinity). In contrast to cooler waters, which are conducive to overturning given that they sink more easily, warmer, fresher waters are buoyant, disrupting deep-water formation. The main categories of drivers disrupting AMOC include:
Changing thermodynamic conditions: As polar and subpolar air temperatures warm, northward-flowing surface waters don’t cool as much before reaching deep-water formation regions, disrupting densification.
Freshwater influx: Fresh water input, whether from increased precipitation and river runoff, melting of sea ice, the Greenland ice sheet melt, and other cryospheric elements in the Arctic, and the influx of fresh Pacific water through the increasingly open Bering Strait, dilutes surface salinity in the North Atlantic, again, disrupting densification.
Reduced reflectivity: As the Arctic warms (four times faster than global mean averages, with summer sea ice loss of ~12% since 1979), the extent of sea ice formation in the Labrador and Nordic Seas is diminishing and glaciers are melting faster. This translates into lost reflectivity, which contributes to both above categories of drivers as well
Nor do these sources of pressure on AMOC necessarily operate in isolation. Much in the same way that AMOC’s current function includes self-sustaining and reinforcing components and dynamics, there’s a risk that if and as it slows down, some of the factors that contribute to its weakening could also reciprocally strengthen one another. For instance, if AMOC slows down meaningfully, it will transport less salty water northwards from the tropics, which would reduce the North Atlantic Ocean’s salinity, further disrupting densification dynamics in a process known as the salt-advection feedback.
To be sure, it is also possible that some perturbing factors impacting AMOC could also counterbalance versus reinforcing one another. Sea ice retreat, for instance, may shift deep-water formation northward, partially offsetting its destabilizing influence. Further, there is considerable debate and uncertainty regarding which of the proximate drivers of AMOC’s weakening are strongest (and how this varies and will vary depending on which timelines are in question), whether and to what extent certain drivers’ contribution is overstated, as well as whether some of the factors thought by some to be causes of AMOC weakening might actually be consequences of it. Really, high uncertainty is characteristic of many if not most of the components of study and discussion of AMOC, which is a point that we’ll now examine further and discuss the practical implications of.
Highly uncertain and hotly debated: The landscape of AMOC risk analysis
While there’s broad consensus on the main sources of risk facing AMOC, as well as the magnitude and direction of what risks a weakening AMOC poses to the world, there are many specific details regarding AMOC that are hotly debated and highly uncertain.
For one, estimates concerning to what degree AMOC will weaken in the future vary considerably and there’s a steady stream of new studies that offer new estimates using different methodologies. As part of its sixth assessment report, in 2021, the IPCC estimated AMOC could lose ~24-30% of its strength by 2100, with a wider range of a 4-46% weakening across all its various models and future emissions trajectories and warming scenarios. Recent scientific studies and meta-analyses, however, suggest AMOC could weaken more over the same timeframe: In a study released in April of this year, Portmann et al. integrated a larger set of observable variables to estimate a weakening of 51% by 2100. That said, several other analyses contend both older and new estimates are too high.
Debate and uncertainty also extends to the question of whether AMOC is already weakening. Many studies and observations suggest it has and is. There are also concerning present-day signals, such as a specific patch of ocean near Greenland, one of the only parts of the Earth’s ocean that is cooling rather than warming, which some interpret as signs that AMOC’s redistributive function is already faltering. Similarly, readings from the RAPID array, which uses vertical moorings anchored to the seabed to monitor AMOC’s strength, have indicated a noticeable downward trend in AMOC’s strength since 2004 (as shown in the graph below).
That said, not everyone interprets the presence of cooling patches in the ocean as indicative of the same thing, and there are other observational and monitoring systems for AMOC that tell a different story in terms of its present and historical strength. OSNAP, another system of vertical moorings used to monitor AMOC, has not detected a statistically significant change in it since 2014. Moreover, other recent studies suggest that AMOC has weakened less since 2010 than it did previously. Further still, some researchers posit AMOC may, in general, be more resilient to climate change than previously thought. All of this points in part to the size of AMOC and the difficulty of monitoring and measuring it comprehensively. Many current studies feature significant limitations. For example, in a recent study by Xing et al., which concluded AMOC is already weakening, the study’s focus was on AMOC’s western-boundary, rather than its full basin-width.
Finally, debate and uncertainty surrounding AMOC also extends to a third category of risk, namely whether AMOC faces tipping point risk. Earth system tipping points (ESTPs) are large-scale climate systems, such as AMOC (as well as, for another example, the Amazon) that could undergo irreversible and potentially self-reinforcing changes if and when certain thresholds to which they’re sensitive are crossed. In analyses of ESTPs, AMOC is almost always included as one example; there’s ample research supporting the idea that, under certain conditions, it could enter an irreversible transition into a permanently weakened state. Indeed, some scientists already point to statistical signatures that are consistent with proximity to a tipping threshold across various instrumental records. But, again, there’s deep uncertainty about questions such as whether AMOC faces tipping risk in general, whether its transition to a weaker state beyond certain thresholds would necessarily be irreversible, not to mention where key threshold(s) lie, how long it might take to cross them, and how abruptly AMOC would transition to an alternate stable state if thresholds are crossed. In fact, some of these questions may feature irreducible uncertainty, i.e., uncertainty that cannot fully be resolved. Certainly, there are uncertainties that cannot be resolved with currently available data and existing monitoring infrastructure.
Clearly, there is no shortage of debate about the severity, timing, and characteristics of the risks facing AMOC and what exactly will happen if AMOC does weaken. All in all, the degree and diversity of uncertainty makes the case for the first pillar of action needed as part of a more comprehensive AMOC response plan, namely improving monitoring and measurement. A concerted push to scale-up monitoring and measurement would help improve studies, science, and research on AMOC generally, and could help assuage the degree to which uncertainties and debate characterize much of the ongoing AMOC conversation. Addressing open scientific questions will also be critical such that in the future, statistical early warning systems can be developed to track system vulnerability, risk levels, and the relevant time horizons. These systems, in turn, could be used (if not needed) to help inform future decision-making efforts via pre-established protocols, integrating into governance architectures designed to manage future decisionmaking.
Before predictions from statistical early-warning systems for AMOC can be trusted, however, more of the foundational science must be settled. That doesn’t mean work on such systems can’t start now, whether to raise funding for them, to establish who will be responsible for building and maintaining them, or to discuss how to incorporate them into governance and other decision-making discussions and frameworks. It just makes the importance of settling more of the science all the more important and urgent.
Nor should the presence of significant uncertainty inherent to AMOC constrain R&D on other risk response options. The absence of certainty doesn’t preclude doing more than just reducing uncertainties. Whether there is a risk of AMOC’s weakening generally is not in dispute, nor is the fact that if it does weaken, it will cause significant global challenges. Hence, assessments of other preparedness and risk reduction options as well as governance structures to manage how these might be deployed must advance in parallel.
Assessing potential interventions to reduce and prevent risks
Currently, the risk prevention options for AMOC are almost solely focused on conventional climate mitigation efforts. To be sure, reducing greenhouse gas emissions and addressing losses of reflectivity, which increasingly contribute to Earth’s energy imbalance, will always remain essential to reduce AMOC-related risks. That said, the stakes of AMOC’s stability likely also warrant evaluating whether there are other risk reduction and prevention options that are scientifically, technoeconomically, politically, and socially feasible.
One reason for this stems from the timelines in question. If, for instance, further research establishes more consensus that AMOC is indeed at risk of crossing a tipping threshold within policy relevant timeframes (for instance, if there are also discernible signals that it could cross such a threshold this century), then mitigation alone is unlikely to suffice to adequately address that risk. Carbon dioxide is a long-lived atmospheric gas; it lingers in the atmosphere for centuries. As a result, even if the world reaches net zero on a faster timeline than currently seems likely, it won’t cool quickly. If meaningfully reducing acute pressure on AMOC becomes a nearer-term prerogative, more targeted measures designed to address the proximate drivers of its destabilization may be desired. De-risking potential options for such a scenario will take time, making it critical that research and evaluation on these fronts accelerate and continue in sustained fashion in coming decades.
Yet, even though many scientific assessments indicate a tipping risk for AMOC cannot be excluded within policy-relevant time horizons, a vanishingly small amount of capital is currently allocated to assessments of targeted interventions and response options, whether for AMOC or for other subcomponents of Earth’s climate that may face tipping point risks.
There are a variety of intervention categories that could help reduce pressure on AMOC, though most of them sit upstream of AMOC itself. “Direct” interventions in AMOC are theoretically imaginable—for instance, dumping massive amounts of salt in targeted areas to try to promote more overturning—but are likely far more costly and lower leverage than targeted efforts to, for instance, cool specific Arctic regions to preserve ice and glaciers. The interventions for which most existing research and development exists span atmospheric strategies, “direct-to-glacier” efforts, sea ice seeding, albedo enhancement, as well as large-scale ocean-based strategies, including proposals such as damming the Bering Strait. While no intervention-focused programs are expressly focused on AMOC, interventions being scoped for other risks could be used with AMOC’s stabilization in mind.
Let’s look at some examples. ARIA, also the funder of a general program focused on early-warning infrastructure for tipping points, has funded a £56.8m program designed to assess a range of interventions that includes both atmospheric interventions and on-land approaches such as ice-thickening experiments. The Arctic Stabilization Initiative (ASI), a program incubated by Advanced Research for Climate Emergencies (ARC) and Renaissance Philanthropy, is assessing interventions to reduce systemic and nonlinear climate risks that overlap with many of the sources of pressure on AMOC, such as the Greenland ice sheet and Arctic sea ice. Specifically, the first phase of ASI’s research program focuses on mixed-phase cloud thinning (MCT), an atmospheric intervention. ~60% of Arctic clouds are mixed-phase; they contain both ice crystals and supercooled liquid water. For most of the year, these clouds trap more heat than they reflect. MCT aims to enhance ice formation in these clouds to allow more surface heat to radiate back into space, thereby reducing warming and climate change.
Several teams and research programs are also actively exploring interventions designed to stabilize the cryosphere. One such program is the Arête Glacier Initiative, which is evaluating subglacial interventions to combat global sea level rise. Like ASI, Arête isn’t expressly focused on designing solutions for AMOC; it is focused on predicting ice loss from Antarctica and researching related interventions. But, if it determines there are viable approaches to glacier and ice sheet restabilization generally, they may be applicable for Arctic sea ice and the Greenland ice sheet, which are relevant to AMOC’s stability.
Other examples of relevant programs and efforts include Real Ice, a privately-funded company supported by the U.K.’s Advanced Research and Invention Agency. In recent tests, it has shown that drilling holes in ice and pumping ocean water up onto its surface can help rethicken ice. It ran its pumps for a total 1,080 hours between January and February this year, adding about 30 centimeters of thickness, which they estimate could enhance the ice’s durability during warmer temperatures for 7-10 days. Then there’s also the Ice Conservation Engineers’ work to assess localized albedo modification options, Swiss Academia Engiadina’s work to produce artificial snow on glaciers, and multiple teams led by John Moore that are working on underwater curtains in the Amundsen Sea to block warm Circumpolar Deep Water from reaching glacier grounding lines as well as a sill at the entrance to Ilulissat Icefjord to anchor a curtain that limits the flow of warm Atlantic water.
While the progress on all these programs is welcome, many more programs, funds, and research-efforts are warranted given the scope of what intervention assessments must cover and the breadth of possible interventions. As illustrated by this brief survey, there are many potential interventions, and the idiosyncrasies of deploying them in different geographies and towards different ends introduces more complexity. Dedicated programs for AMOC are also needed to coordinate ongoing research on AMOC-related scientific questions, synthesize research on interventions from elsewhere that could be applicable to AMOC, and to create shared agendas for what research and development is most needed going forward, whether with respect to early warning systems, interventions, or as we’ll soon turn our attention to, on questions of governance or social and institutional legitimacy.
Moreover, intervention assessments will likely take years even to advance to field tests. Researchers focused on atmospheric interventions, for instance, estimate it will take decades to adequately de-risk them, let alone to develop the technological capacity for deployment. Hence, it’s imperative that this field scales quickly.
Governance capacity; a parallel prerogative for better preparedness and risk reduction
Many programs, including ASI, are structured explicitly with future handoffs to governments and multilateral institutions in mind. These are the actors with the requisite technical, resource, and governance capacity, as well as the ability to develop the coalitions of stakeholders who will be needed to cultivate legitimacy for deployments and to adjudicate potential conflicts. AMOC influences countless systems worldwide, heightening the need for buy-in from many different stakeholders and the complexity of achieving that buy-in.
Still, development of governance for AMOC-related risks is perhaps even more nascent than research development efforts for interventions. There is functionally no operational governance architecture to guide under what circumstances and how interventions might be deployed. This gap introduces its own risks. The vacuum created by the absence of governance architecture may make it more likely that future deployment decisions will be made unilaterally, which is likely to drive geopolitical conflicts. Even if we assume robust future governance, some geopolitical disagreements are still likely. In the absence of governance, the resultant geopolitical conflicts could undermine response efforts entirely.
Outlining everything that’s required on the governance front to address AMOC-related risks is well beyond the remit of this article. What we can establish is that work on governance structures and capabilities for AMOC-risk response efforts must also accelerate, as is true for the other pillars of the risk preparedness and reduction plan we’ve suggested. As with intervention assessments, it will take time to design governance structures and systems, to negotiate and ratify them, and to implement them. The High Seas Treaty was first proposed in 2011 and wasn’t ratified and operationalized until 2025. And implementation takes time, too. While intergovernmental negotiations to structure the Sendai Framework required less than a year, it still features a 15-year implementation period (2015 to 2030).
All that said, it would be inaccurate to conclude that all risk reduction response and intervention-related governance work must start entirely from scratch. For one, Seddon and Milkoreit have proposed a dedicated tipping-point governance structure. Their Tipping Element Monitoring and Response Facilities (TEMRFs) would constitute purpose-built institutions designed to monitor, assess, and respond to the risks posed by tipping elements in various Earth systems, including AMOC. Each system potentially exposed to tipping point risk would get its own TEMRF, which would provide targeted monitoring, specialized expertise, and coordinated international collaboration for each at-risk system. Existing frameworks like the aforementioned Sendai Framework or other disaster risk response structures may also offer elements that are worth adapting in hybrid structures.
Regardless of the specific approach, governance development would also benefit from involving the communities most exposed to the potential consequences of AMOC’s weakening and of any potential interventions. Engaging stakeholders like Indigenous populations in the Arctic to work on response options would help avoid familiar failure modes in which options are developed that, while perhaps technically viable, lack social license. On this front, models like the Arctic Council’s Permanent Participant, which gives Indigenous organizations a formal role alongside member states, can offer a starting point.
AMOC: A case study for climate stabilization
Making decisions under deep uncertainty is never easy, and no one imagines that scaling monitoring and measurement infrastructure for AMOC, assessing risk reduction and prevention interventions, and developing governance capabilities, let alone doing all of this in parallel, will be easy. But the alternative amounts to leaving future generations exposed to scenarios in which they’ll need to make even more consequential decisions under similar if not greater amounts of uncertainty, which is more likely to lead to underinformed, unilateral, and ultimately less effective responses. Nor is waiting for more certainty before scaling prediction, preparedness, and prevention options an exercise in caution. It’s a decision in its own right, one that could increase risks more than it reduces them.
Acknowledgements: We’d like to thank Dr. Colin Meyer, Geoff Vallis, and Spencer Jones, whom we interviewed as part of an overarching effort to survey and better understand the AMOC landscape. That said, this opinion and analysis is our own; their contributions to our research should not be read as endorsements of any part of this article per se.
How useful would continuous sea surface temperature and sea surface salinity measurements from ships transiting the North Atlantic be for monitoring AMOC strength and/or reducing uncertainty in models?