The Atlantic's Hidden Heartbeat
A system of ocean currents that shapes climate for billions of people may be approaching a point of no return.
The animated flow lines trace the three-dimensional path of the Atlantic Meridional Overturning Circulation, derived from ocean velocity data in the ECCO V4R4 reanalysis. The surface branch (red, 0–600 m) carries warm, salty water northward from the tropics. The sinking branch (white, 200–2,000 m) shows where cooled, dense water plunges at high latitudes. The deep return branch (blue, below 2,000 m) flows southward along the ocean floor, completing the loop. Together they form the "conveyor belt" that transports 1.26 petawatts of heat toward Europe — equivalent to the output of roughly a million nuclear power plants. The particle paths are computed from real velocity fields, not drawn by hand.
Data: ECCO V4R4 ocean reanalysis, 2010 annual mean.1
The Great Conveyor
Deep beneath the Atlantic's surface, a vast river of water circulates in a pattern that has persisted for millennia. Warm, salty water from the tropics flows northward along the surface, carried by the Gulf Stream and the North Atlantic Current. As it reaches the high latitudes near Greenland, Iceland, and Norway, it releases its heat to the atmosphere, warming the winds that blow across western Europe. Then, cooled and dense, it sinks thousands of metres to the ocean floor and begins a slow return journey southward through the deep abyss.
This is the Atlantic Meridional Overturning Circulation, or AMOC. It is one of the largest heat-transport systems on the planet, and it operates on a scale that defies easy intuition. The volume of water it moves is measured in sverdrups: one sverdrup equals one million cubic metres per second, roughly equal to the combined flow of all the world's rivers.
The AMOC is not merely a current. It is a thermohaline engine: driven by differences in temperature (thermo) and salinity (haline) that create density gradients in the water. Where the water is cold and salty enough, it becomes dense enough to sink. This sinking — deep convection — is the pump that drives the entire system. Without it, the conveyor stalls.
Named after Norwegian oceanographer Harald Sverdrup. One sverdrup is one million cubic metres of water per second — picture a cube of water 100 m on each side (roughly the height of the Eiffel Tower's first floor) sliding past every single second. Or think of it this way: it would fill an Olympic swimming pool in about 2.5 milliseconds.
The AMOC moves about 13 times more water per second than all of Earth's rivers combined. The Gulf Stream, fed by the whole subtropical Atlantic gyre, carries roughly 9× more than the AMOC alone — ocean currents operate on a scale that dwarfs anything on land.
/cent.
Temperature alone tells you where the climate is; rate of change tells you how fast it is moving. One °C per century means a region warms or cools by one degree Celsius over a hundred years — slow enough for ecosystems and societies to partially adapt. The higher the rate, the less time there is to respond.
An AMOC collapse would cool northern Europe at up to 20°C per century — roughly ten times faster than current global warming is heating the planet. The Dansgaard-Oeschger events of the last ice age were even more extreme: Greenland warmed 10–15°C within a single decade, a rate so fast it is literally off this chart.
Power measures how fast energy moves. One petawatt = 10¹⁵ watts — that is one quadrillion watts, or the output of about a million large nuclear power plants running simultaneously. A single power plant produces ~1 GW; 1 PW is a billion times that.
The AMOC transports about 70 times more energy as heat than all of human civilisation — every power station, car, plane, and heating system on Earth — produces. A collapse would cut that heat delivery by 70%, with the remainder equivalent to less than 22 times humanity's total energy use.
What Drives the Circulation
The AMOC's engine relies on a delicate balance. In the Nordic Seas and the Labrador Sea, winter storms strip heat from the surface water. The water cools, its density increases, and it plunges to depths of 1,000 to 4,000 metres. This process — deep convection — renews the supply of North Atlantic Deep Water, the cold, dense mass that flows slowly back toward the Southern Hemisphere along the ocean floor.
But this engine is vulnerable to a subtle and dangerous feedback loop. As warm surface water flows northward, it also carries salt. Salt increases density, which helps maintain the sinking that drives the circulation. If freshwater is added to the surface — from melting ice sheets, increased rainfall, or river discharge — it dilutes the salt, reduces the density, and weakens the sinking. A weaker circulation carries less salt northward, which further reduces density, further weakens sinking, and so on. This is the salt-advection feedback, and it is the mechanism by which the AMOC can tip.
In 2025, van Westen, Kliphuis, and Dijkstra demonstrated for the first time in a strongly eddying ocean model (0.1-degree horizontal resolution) that the AMOC can collapse entirely through this mechanism. Previous concerns that ocean eddies — the swirling turbulence that pervades the real ocean — might stabilise the circulation proved unfounded. The salt-advection feedback is fundamental to the physics of the thermohaline circulation, not a model artefact.
Tens of thousands of virtual particles trace ocean surface currents in real time, advected by velocity fields from the Mercator GLORYS12V1 global ocean reanalysis. Each particle's colour encodes sea surface temperature — blue for polar water near −2°C, orange and red for tropical water approaching 32°C. Brightness reflects speed: the Gulf Stream blazes as it detaches from Cape Hatteras, while the sluggish tropical gyres drift in near-darkness. The Antarctic Circumpolar Current encircles the Southern Ocean as an unbroken ring — the largest current system on Earth. This is a snapshot of one month's ocean state, not a long-term average.
Data: Mercator GLORYS12V1 (Copernicus Marine Service). June 2021 monthly mean surface velocity and temperature at 1/12° resolution.
The globe above reveals the ocean's surface currents as a living, turbulent system. Tens of thousands of particles trace the velocity field measured by ocean reanalysis, showing the Gulf Stream as a powerful jet off the US East Coast, the Antarctic Circumpolar Current as an unbroken ring around the Southern Ocean, and countless smaller gyres and eddies that stir the ocean at every scale.
Notice how the brightness of each particle encodes its speed: the Gulf Stream burns bright where it detaches from Cape Hatteras, while the deep tropical oceans drift slowly by comparison. This is the surface expression of the circulation — but the AMOC's true engine lies below, in the deep convection zones of the high-latitude North Atlantic, where cooling surface water plunges thousands of metres to form the return current that completes the loop.
The colour overlay shows the linear trend in sea surface temperature from 1993 to 2021 — where the ocean has cooled (blue) and where it has warmed (red) over nearly three decades of satellite observations. Two signals stand out. First, the "cold blob" south of Greenland: a region that has actually cooled against the global warming trend, because reduced northward heat transport leaves the subpolar ocean starved of warmth. Second, above-average warming along the US East Coast, where the Gulf Stream is shifting northward and closer to shore as the deep boundary current beneath it weakens. Both signals are physically linked — two faces of the same AMOC weakening — and cannot be explained by aerosol cooling alone.
Data: Copernicus Marine Service — Global Ocean SST trend (1993–2021). Caesar et al. (2018).14
The Evidence
The signs are converging. Since 2004, the RAPID monitoring array — a line of instruments spanning the Atlantic at 26.5°N — has been measuring the AMOC's strength directly. The data show a decline of approximately 0.8 Sv per decade.1 But the RAPID record spans only two decades. To see the full picture, Caesar and colleagues (2018)14 developed a sea surface temperature fingerprint — a method for reconstructing AMOC strength from ocean temperature patterns going back to 1870. Their calibrated estimate: the AMOC has weakened by 3 ± 1 sverdrups, roughly 15 percent, since the mid-twentieth century.14 A broader comparison of eleven independent proxy records reaching back to AD 400 shows that the modern slowdown is unprecedented: the AMOC is now at its weakest point in over a thousand years.15
Perhaps the most striking evidence is what Caesar et al. called the AMOC fingerprint — a pair of linked signals visible in sea surface temperature trends.14 The first is the cold blob: while the planet has warmed by more than 1.2°C since pre-industrial times, a patch of the North Atlantic south of Greenland has actually cooled. The second is less widely known — an above-average warming along the US East Coast, caused by the Gulf Stream shifting northward and closer to shore as the deep western boundary current weakens beneath it. Both signals are two faces of the same coin, and cannot be explained by aerosol cooling — high-resolution climate models with no aerosol forcing reproduce the full pattern.14
Caesar and colleagues (2021)15 assembled eleven independent proxy records — from sediment cores and deep-sea corals to ice cores and marine productivity — to reconstruct the AMOC's evolution over the past 1,600 years. The picture that emerges is remarkably consistent across all proxies:
The Gulf Stream itself offers another clue. Observations from satellite altimetry (1993 to 2024) show a statistically significant northward trend in the Gulf Stream's path near Cape Hatteras — 0.16 degrees per decade (p < 0.05). Subsurface temperature data spanning six decades (1965 to 2024) corroborate this shift.
van Westen and Dijkstra (2026)4 showed that in high-resolution ocean simulations these path changes are not random drift — they are a precursor to collapse. The same model places an abrupt Gulf Stream shift approximately 25 years before AMOC collapse, though the precise lead time is model-dependent and should not be taken as a real-world forecast.
The trajectory itself is telling — but there is a deeper signal hidden in the data, visible only through the mathematics of a system losing its grip.
Those warning signs emerged in a 2,200-year model simulation. The critical question is whether the real ocean shows the same pattern.
The three markers show the AMOC's deep convection sites — Labrador Sea, Irminger Sea, and Nordic Seas — where winter storms strip heat from the surface water, making it dense enough to plunge 1,000 to 4,000 metres to the ocean floor. This sinking is the pump that drives the entire circulation. The vertical beams represent the maximum depth of winter mixing at each site: deeper beams indicate a healthier pump. Observational data show that maximum mixing depths have been declining across all three sites over the past five to ten years. The "Projected Collapse" toggle shows what happens when mixing depth drops below 250 metres — the threshold below which deep convection effectively fails. In CMIP6 climate models, this deep mixing collapse precedes the shutdown of the northern AMOC by approximately 30 years — a potential early warning signal.
Data: Drijfhout et al. (2025); observational mixing depths from EN4, ARMOR, GLORYS, and Argo.3
The Tipping Threshold
The AMOC is a bistable system. Like a ball resting in a shallow valley, it can exist in one of two stable states: "on" (the current strong circulation) or "off" (a collapsed, near-stagnant mode). Pushing the ball over the ridge between the two valleys — crossing the tipping point — means the system falls into the collapsed state. And the valley of the collapsed state is deep: once there, the AMOC does not recover on any timescale relevant to human civilisation.
The physics of this bistability are well understood, rooted in the Stommel bifurcation theory developed in the 1960s. What remains uncertain is precisely where the ridge lies — how much freshwater forcing, how much warming, is needed to push the system past the point of no return. The AMOC-induced freshwater transport at 34°S (FovS), identified by van Westen and colleagues as a physics-based early warning signal,9 shows increasing variance — a classic precursor to tipping in dynamical systems. Critically, current reanalysis data show FovS already on a statistically significant negative trend, consistent with the AMOC heading toward its tipping point.9
That gauge captures a snapshot — where the AMOC stands today. The chart below traces the trajectory, and projects where it may be heading under different emission pathways.
What would these declining numbers mean for the people living downstream of the Atlantic's heat delivery?
Those bars represent the end state — the new climate once the AMOC has fully stopped. The chart below shows the journey: how fast the cooling unfolds once the tipping point is crossed.
Deep mixing collapse may provide another early warning. Before the AMOC itself weakens catastrophically, the maximum depth of winter mixing in the Labrador, Irminger, and Nordic Seas drops below 200 metres — a sign that deep convection is failing. Drijfhout and colleagues (2025)3 showed that this deep mixing collapse precedes northern AMOC shutdown by approximately 30 years in CMIP6 models. Observational data (EN4, ARMOR, GLORYS, Argo) already show downward trends in deep mixing over the past five to ten years across all available data products.
The colour overlay shows the speed of ocean currents averaged over the upper 250 metres of the water column. Three views are available.
Healthy AMOC shows a simulation in which the overturning circulation is running at full strength (~27 Sv at 26.5°N). The Gulf Stream appears as a bright band of fast flow detaching from Cape Hatteras and sweeping northeast across the Atlantic — exactly what we observe today.
Collapsed AMOC shows the same model after freshwater forcing has pushed the circulation past its tipping point: AMOC strength drops to ~5 Sv. The Gulf Stream shifts 219 km northward within just two years, the Labrador Current nearly vanishes, and ocean heat transport at 26.5°N falls from 1.26 to 0.38 petawatts — a 70% reduction.
Observed shows real ocean currents from the GLORYS12V1 reanalysis (Copernicus Marine Service), year by year from 1993 to 2024. Satellite altimetry over this period shows a statistically significant northward trend in the Gulf Stream's path near Cape Hatteras (0.16° per decade, p < 0.05) — a pattern the model identifies as a precursor to collapse.
Both model views come from the Parallel Ocean Program (POP) at 0.1° eddy-resolving resolution, run as a stand-alone ocean simulation with slowly increasing freshwater forcing. These are not calendar years — they are model years from a controlled experiment designed to isolate the AMOC's response to freshwater input. The simulation is not a forecast for any specific date; it reveals what the physics permits if the tipping point is crossed.
Data: van Westen & Dijkstra (2026).4 Observational data: GLORYS12V1 (1/12°, Copernicus Marine Service).
If the AMOC Collapses
In high-resolution ocean simulations, the collapse unfolds with startling speed. The Gulf Stream, normally separated from the coast near Cape Hatteras, abruptly shifts northward by 219 kilometres — in just two years.4 The continental slope along the eastern United States warms by 4°C as the weakening of the Labrador Current allows warmer subtropical water to flood northward. At 26.5°N, the AMOC's heat transport drops from 1.26 petawatts to just 0.38 petawatts — a reduction of 70 percent.2
What does this reshaping of the ocean look like? The globe that follows shows the sea surface temperature difference between a collapsed and a healthy AMOC, simulated at 0.1-degree resolution in the HR-POP model — fine enough to resolve individual eddies. The pattern is stark: a band of intense warming (red) off the US east coast where the Gulf Stream shifts northward, and deep cooling (blue) across the subpolar North Atlantic where the conveyor's heat delivery fails.
The colour overlay shows the difference in sea surface temperature between a collapsed AMOC state and a healthy control run in the same model — blue where the ocean cools, red where it warms. The pattern is hemispheric: the North Atlantic cools dramatically (by up to 10°C in places) as the conveyor stops delivering tropical heat to high latitudes. Meanwhile, the Southern Hemisphere warms slightly as heat that would normally be exported northward accumulates south of the equator. This "bipolar seesaw" is a fundamental signature of AMOC shutdown, and matches patterns seen in paleoclimate records of past collapse events. The simulation uses a 0.1° eddy-resolving ocean model — fine enough to resolve the swirling eddies that transport heat and salt in the real ocean.
Data: van Westen, Kliphuis & Dijkstra (2025), Geophysical Research Letters. HR-POP model, 0.1° eddy-resolving resolution.2
The consequences cascade beyond the ocean surface. The atmospheric model below reveals what collapse means at street level: several European cities experience temperature drops of 5 to 15°C from their present-day baseline, locked in over decades of collapse.9 In Bergen, Norway, the CESM simulation records temperatures falling at roughly 3.5°C per decade during the collapse phase — a rate of change ten times faster than current anthropogenic warming.9 Sea ice advances as far south as 50°N. Agriculture across the British Isles, France, and Scandinavia would face conditions for which no modern infrastructure is prepared.
The Intertropical Convergence Zone, the band of tropical rainfall that follows the warmest ocean surface, shifts southward, reorganising monsoon systems that hundreds of millions of people depend on for water and food. The coupled climate model (CESM, ~1° resolution) below reveals the hemispheric seesaw in full: the Northern Hemisphere cools by up to 20°C per century while the Southern Hemisphere warms.9 This is no longer just an ocean story — it is a global reorganisation of heat.
Global shows the rate of atmospheric temperature change during the century-long AMOC collapse (model years 1750–1850 in a CESM simulation — not calendar years). Blues mark the dramatic cooling across the Northern Hemisphere; reds show compensating warming in the south. This is not a forecast for a specific date — it shows what would happen if the AMOC crosses its tipping point.
European Detail zooms into Europe and compares two stable climate states in the same model: one with a healthy AMOC and one without. It reveals how profoundly Europe's climate depends on Atlantic heat transport. Five variables capture different facets of the impact:
• Winter — change in December–February mean temperature (up to ~34°C colder in the far north)
• Annual — change in the yearly average (up to ~20°C colder in northwestern Europe)
• Summer — change in June–August mean (the mildest shift, up to ~10°C)
• Cold ext. — change in the coldest day expected once per decade (up to ~43°C more extreme — the single most dramatic metric)
• Frost — additional days per year below freezing (up to 290 extra frost days — a transformation of growing seasons and agriculture)
Data: van Westen, Kliphuis & Dijkstra (2024)9 (Global); van Westen & Baatsen (2025, Geophysical Research Letters) (European Detail).
The atmosphere is only part of the story. An AMOC collapse would also reshape the ocean itself — raising coastlines, shifting currents, and reaching far beyond the Atlantic.
Sea levels along the American Atlantic coast are projected to rise by 15 to 20 centimetres from the AMOC effect alone by 2100, on top of rises from thermal expansion and ice sheet melt.1 The HR-POP model simulation shows dynamic sea-level rise of up to 66 cm near the US East Coast in the collapsed state — reflecting the Gulf Stream's northward shift redirecting ocean mass.4
Yet the picture is not uniformly catastrophic, and one recently discovered interaction offers a counterintuitive nuance. Hogner and colleagues (2025)7 found that AMOC weakening actually increases dry-season precipitation in the Southern Amazon, offsetting about 17 percent of the observed drying trend through a teleconnection involving the Caribbean Low-Level Jet. This stabilising feedback between two tipping elements — the AMOC and the Amazon rainforest — suggests that cascade risks may be partially tempered by compensatory feedbacks that are only now being mapped.
A Cascade of Tipping Points
The AMOC does not exist in isolation. It sits at the centre of a web of interconnected tipping elements — each capable of crossing its own threshold independently, but also capable of nudging others toward theirs. The emerging science of tipping cascades suggests the real risk may not be any single collapse, but of one collapse compressing the remaining distance to the next — a chain that no decarbonisation programme can outrun once it begins.
The Greenland Ice Sheet and the AMOC are locked in mutual reinforcement. Greenland melt adds freshwater to the North Atlantic, diluting surface density and weakening deep convection — precisely the mechanism that drives the AMOC toward its tipping point. In the HR-POP model, a freshwater forcing of just 0.125 Sv — seventeen times the current Greenland meltwater discharge rate — was sufficient to initiate collapse.2 As warming accelerates Greenland's contribution, the gap between the current rate and the critical threshold narrows.
The West Antarctic Ice Sheet presents a striking paradox. Sinet and colleagues (2025)12 found that meltwater from West Antarctica, routed northward through thermohaline exchange, can actually counteract the freshwater-driven collapse of the AMOC — because Southern Ocean freshwater tends to reinforce, rather than weaken, the Atlantic salinity gradient that sustains overturning. One potential collapse may temporarily delay another. But this protection is finite: the underlying trajectory of greenhouse warming continues regardless, and the stabilising effect diminishes as ice loss accelerates.
The Amazon rainforest interaction is equally counterintuitive. Rather than amplifying collapse risk, AMOC weakening appears to partially buffer the Amazon against its own tipping point. Hogner and colleagues (2025)7 traced a causal pathway through the Caribbean Low-Level Jet: AMOC weakening intensifies dry-season moisture transport into the southern Amazon, offsetting roughly 17 percent of the observed drying trend. But 17 percent offset leaves 83 percent of the drying signal intact. Once Amazon deforestation crosses approximately 20–25 percent of the basin, the forest's own internal moisture recycling collapses — an outcome that oceanic teleconnections cannot prevent.
Permafrost and boreal forests add further momentum. Deutloff and colleagues (2025)11 modelled the probability of triggering 16 tipping elements under current policy trajectories — a pathway consistent with roughly 2.5–3°C of warming. They found a 62 percent probability of triggering permafrost thaw and a 40 percent probability of triggering boreal forest dieback. Critically, each of these events releases additional carbon — roughly 0.3–0.5°C of additional global warming per triggered element — compressing the remaining distance to the next threshold, including the AMOC's own.
What science cannot yet specify is the speed of any such cascade. The AMOC responds over decades to centuries; ice sheets over centuries to millennia; the Amazon, under deforestation pressure, could shift within decades. A cascade would unfold not as simultaneous collapse but as a slow-motion sequence in which each event narrows the options for what follows. Dijkstra and van Westen (2026)13 stress that current Earth system models are not yet fit to estimate these cascade probabilities reliably — because the same freshwater biases that cause models to underestimate AMOC bistability also distort their representation of cross-system feedbacks. Rare-event simulation applied to bias-corrected models is, as of 2026, the most promising path to a defensible cascade risk estimate.
The Timeline of Risk
When could this happen? The honest answer is that no one knows with precision. Ditlevsen and Ditlevsen (2023)8 used statistical early-warning signals in sea surface temperature data to estimate a tipping window between 2025 and 2095 — though this methodology has been contested by other researchers. What the latest models do reveal is the sensitivity of the outcome to our choices.
Drijfhout and colleagues (2025)3 analysed the extended CMIP6 ensemble — climate models run to the year 2300 and beyond. Under high emissions (SSP5-8.5), 67 percent of models show northern AMOC shutdown by 2300 — a cessation of deep overturning driven by North Atlantic Deep Water formation, not a complete collapse to zero flow. Under intermediate emissions (SSP2-4.5), this drops to 30 percent. Under the most ambitious mitigation scenario (SSP1-2.6), it falls to 21 percent — still far from negligible.
The duration of temperature overshoot matters enormously. Ritchie and colleagues (2026)5 showed that both the height and the duration of a temperature overshoot above 1.5°C determine tipping risk. The constraint is stringent: doubling the exceedance above 1.5°C requires cutting its duration by a factor of four to maintain the same tipping risk — a trade-off that is very difficult to satisfy given realistic decarbonisation rates.
The AMOC, as a slow-responding system (50 to 300 years), is more resilient to brief temperature exceedances than fast systems like coral reefs — but sustained overshoot erases this advantage. Once deep mixing collapses and the salt-advection feedback takes hold, the process becomes self-reinforcing and effectively irreversible on human timescales.
What We Know and What We Don't
The AMOC has collapsed before. During the last ice age, a series of events known as Dansgaard-Oeschger oscillations saw abrupt warmings of 10–15°C in Greenland within a single decade, driven by shifts in North Atlantic deep convection — more than 20 such events have been documented in ice core records.1 Larger Heinrich events — massive iceberg calving episodes that flooded the surface ocean with freshwater — triggered full AMOC shutdowns that persisted for centuries. The paleoclimate record makes clear that this system has more than one mode, and that transitions between modes can be sudden and violent.
What remains genuinely uncertain is how close we are to the threshold. The central estimate of 4.0°C for the tipping temperature carries the largest uncertainty range of any tipping element assessed: 1.4 to 8.0°C.5 But the AMOC is not the only ocean tipping element to watch. The North Atlantic Subpolar Gyre may constitute a separate and more imminent tipping system: best-estimate threshold of ~1.8°C (range 1.1–3.8°C), with a collapse timescale of roughly a decade5 — far faster than the AMOC itself. Some models show SPG collapse as early as the 2040s under moderate warming of 1–2°C, with a 36–44% probability across the model ensemble.6 An SPG collapse would cause regional surface air temperature drops of 2–3°C around the UK and northwest Europe within years, disrupt the fisheries and ecosystems of the high-latitude North Atlantic, and likely hasten AMOC weakening by suppressing deep convection. These caveats apply: not all models show SPG bistability, and why some models produce it while others do not remains unresolved.6
Across all tipping elements, current climate models systematically underestimate AMOC collapse risk due to biases in ocean salinity distributions that make the simulated circulation artificially stable.10 The observational record (70 years of comprehensive data, 20 years from the RAPID array) may also be too short to distinguish irreversible weakening from natural multidecadal variability.3 But the proxy record now extends the baseline to over a millennium — and it shows that the current weakening has no precedent in that entire span.15
The monitoring infrastructure is improving. The RAPID array continues its vigil at 26.5°N. Argo floats profile the ocean in three dimensions. Satellite altimetry tracks the Gulf Stream's path with centimetre precision. Freshwater transport at 34°S — the SAMBA transect — is being monitored as a physics-based early warning signal.9 These observational systems, combined with the next generation of high-resolution coupled climate models, may soon resolve whether the AMOC's current trajectory leads to recovery, gradual decline, or abrupt collapse.
What the science makes unambiguous is this: the AMOC is weakening, multiple early warning signals are flashing, and the consequences of crossing the tipping point would reshape the climate of the Northern Hemisphere for centuries. The uncertainty about exactly when this might happen is not a reason for complacency. It is a reason for urgency.
References
- Rahmstorf, S. (2024). Is the Atlantic Overturning Circulation Approaching a Tipping Point? Oceanography (Early Online Release). doi:10.5670/oceanog.2024.501
- van Westen, R. M., Kliphuis, M. & Dijkstra, H. A. (2025). Collapse of the Atlantic Meridional Overturning Circulation in a Strongly Eddying Ocean-Only Model. Geophysical Research Letters, 52, e2024GL114532. doi:10.1029/2024GL114532
- Drijfhout, S., Angevaare, J. R., Mecking, J., van Westen, R. M. & Rahmstorf, S. (2025). Shutdown of northern Atlantic overturning after 2100 following deep mixing collapse in CMIP6 projections. Environmental Research Letters, 20, 094062. doi:10.1088/1748-9326/adfa3b
- van Westen, R. M. & Dijkstra, H. A. (2026). Abrupt Gulf Stream path changes are a precursor to a collapse of the Atlantic Meridional Overturning Circulation. Communications Earth & Environment, 7, 197. doi:10.1038/s43247-026-03309-1
- Ritchie, P. D. L., Steinert, N. J., Abrams, J. F. et al. (2026). The implications of overshooting 1.5 °C on Earth system tipping elements — a review. Environmental Research Letters, 21, 043001. doi:10.1088/1748-9326/ae3cad
- Loriani, S., Aksenov, V., Armstrong McKay, D. L. et al. (2025). Tipping points in ocean and atmosphere circulations. Earth System Dynamics, 16, 1611–1653. doi:10.5194/esd-16-1611-2025
- Hogner, A., Di Capua, G., Donges, J. F., Donner, R. V., Feulner, G. & Wunderling, N. (2025). Causal pathway from AMOC to Southern Amazon rainforest indicates stabilising interaction between two climate tipping elements. Nature Geoscience. doi:10.1038/s41561-025-01234-w
- Ditlevsen, P. & Ditlevsen, S. (2023). Warning of a forthcoming collapse of the Atlantic meridional overturning circulation. Nature Communications, 14, 4254. doi:10.1038/s41467-023-39810-w
- van Westen, R. M., Kliphuis, M. & Dijkstra, H. A. (2024). Physics-based early warning signal shows that AMOC is on tipping course. Science Advances, 10, eadk1189. doi:10.1126/sciadv.adk1189
- Liu, W., Xie, S.-P., Liu, Z. & Zhu, J. (2017). Overlooked possibility of a collapsed Atlantic Meridional Overturning Circulation in warming climate. Science Advances, 3, e1601666. doi:10.1126/sciadv.1601666
- Deutloff, J., Held, H. & Lenton, T. M. (2025). High probability of triggering climate tipping points under current policies modestly amplified by Amazon dieback and permafrost thaw. Earth System Dynamics, 16, 565–583. doi:10.5194/esd-16-565-2025
- Sinet, S., von der Heydt, A. S. & Dijkstra, H. A. (2025). West Antarctic Meltwater can Prevent an AMOC Collapse. arXiv:2502.17104v1. doi:10.5281/zenodo.14800555 ⚠ Preprint.
- Dijkstra, H. A. & van Westen, R. M. (2026). The Probability of an AMOC Collapse Onset in the Twenty-First Century. Annual Review of Marine Science, 18, 23–46. doi:10.1146/annurev-marine-040324-024822
- Caesar, L., Rahmstorf, S., Robinson, A., Feulner, G. & Saba, V. (2018). Observed fingerprint of a weakening Atlantic Ocean overturning circulation. Nature, 556, 191–196. doi:10.1038/s41586-018-0006-5
- Caesar, L., McCarthy, G. D., Thornalley, D. J. R., Cahill, N. & Rahmstorf, S. (2021). Current Atlantic Meridional Overturning Circulation weakest in last millennium. Nature Geoscience, 14, 118–120. doi:10.1038/s41561-021-00699-z
Suggested citation
GLOBAÏA (2026). The Atlantic's Hidden Heartbeat — AMOC [interactive exploration]. globaia.org/amoc/. Accessed .