Scientists Just Recalculated the AMOC Collapse. The New Number Is Far Worse.

The ocean current system responsible for keeping Western Europe warm, regulating African rainfall, and stabilising sea levels along the entire US East Coast is collapsing at twice the speed climate planning assumed, according to new ocean measurements published this week.

Findings published in Science Advances in April 2026 calculate an AMOC slowdown of 51 percent by 2100, against the 32 percent figure currently used in international climate assessments, with the margin of error narrowed from plus or minus 37 percentage points to plus or minus 8.

The Atlantic Meridional Overturning Circulation, the AMOC, is the name for a vast system of connected ocean currents running the length of the Atlantic. Warm, salty water flows northward near the surface, carrying heat from the tropics toward Europe and the North Atlantic. As that water reaches the cold seas around Iceland, Greenland, and Norway, it releases its heat into the atmosphere above, cools rapidly, becomes denser than the surrounding ocean, and sinks to the seafloor. That cold, dense water then travels southward again along the ocean floor, completing a loop that has operated continuously for thousands of years. The heat released during the northward journey is what keeps Ireland, Britain, Scandinavia, and coastal Europe several degrees warmer than their latitudes should allow. It is also what drives the rainfall patterns that feed hundreds of millions of people across sub-Saharan Africa.

Previous projections put the weakening at 32 percent, but that number came with a problem. Climate models disagree with each other on the AMOC more than on almost any other variable. For most climate questions, the biggest source of uncertainty is which emission pathway the world follows. For the AMOC, emission pathways barely matter. By 2100, disagreements between individual climate models account for 78 percent of all uncertainty in AMOC projections, with emission scenarios contributing just 14 percent and natural ocean variability the remaining 8 percent. Averaging across models that disagree so fundamentally produces a central estimate with a margin of error so large it is useless for planning. A range of plus or minus 37 percentage points around a 32 percent central figure is compatible with outcomes ranging from modest slowdown to near-total collapse.

The new approach cuts through that problem by anchoring projections to what the ocean actually measures rather than what models predict. Instead of averaging all models equally, the method tests each model against 19 real-world ocean measurements covering sea surface temperature and sea surface salinity across nine regions of the Atlantic, Indian Ocean, and equatorial Pacific. Models that accurately reproduce those 19 observed values get more weight in the final projection. Models that do not get less. The result is a constrained central estimate of 51 percent weakening, with an uncertainty range of just plus or minus 8 percentage points. Every value within that range points to weakening severe enough to cross the threshold the IPCC formally classifies as substantial collapse.

Two specific errors in standard climate models explain most of the gap between the old number and the corrected one. The first involves salt. The South Atlantic in standard models runs consistently fresher than the real ocean, meaning models assign it lower salt concentrations than measurements record. Salt is what makes surface water dense enough to sink when it reaches the cold North Atlantic, driving the overturning. When models underestimate South Atlantic salinity, they underestimate how much salt the current carries northward, a process called the salt-advection feedback. A circulation with a weaker salt-advection feedback resists slowdown more than the real ocean does. It looks stable in the model when the actual ocean is not. South Atlantic salinity errors alone account for 47 percent of the gap between the old projection and the corrected figure. The second error is temperature. Models run the North Atlantic consistently colder than it actually is across the subtropical, subpolar, and Nordic Sea regions. A colder-than-real North Atlantic produces a circulation that responds less to warming, again generating a projected AMOC that slows less than real ocean physics justifies. North Atlantic temperature errors account for a further 36 percent of the correction, with salinity errors in the Labrador Sea, Tropical North Atlantic, and Eastern Subpolar zone making up the remaining 17 percent.

The physical consequences of a 51 percent weakening are not evenly distributed and they are not distant. The AMOC currently moves approximately 16.9 sverdrups of water past 26 degrees North latitude, with one sverdrup equal to one million cubic metres per second, a flow volume roughly 100 times greater than the Amazon River. Each sverdrup of that flow carries heat northward. Cutting the circulation to 8.1 sverdrups, which is what the corrected projection calculates for the period 2091 to 2100, removes approximately 0.44 petawatts of thermal energy from the northward heat delivery system. That heat no longer reaches the atmosphere above the North Atlantic. Average temperatures across Britain, Ireland, Norway, and the Atlantic coast of Europe fall as a direct result, with no compensating mechanism available on century timescales. Growing seasons across northern Europe shorten. Crop yields fall in agricultural zones already managing tight margins. Winter storms reach coastlines that no longer have the oceanic warmth to moderate incoming Atlantic weather systems.

Across sub-Saharan Africa, the consequences arrive through rainfall. The AMOC controls the position of the intertropical convergence zone, the band of intense tropical rainfall that circles the Earth near the equator and determines wet and dry seasons across Africa and northern South America. When the AMOC weakens, that rainfall band shifts southward. In the Sahel, the semi-arid strip stretching from Senegal to Ethiopia across the southern edge of the Sahara, that southward shift reduces wet season rainfall at precisely the scale that pushes agricultural systems past their failure points. The Sahel supports several hundred million people through smallholder farming with virtually no infrastructure buffer against sustained rainfall failure. Crop failure at the scale driven by a 51 percent AMOC weakening does not produce manageable shortfalls. It produces food and water insecurity across a population the size of the United States, in a region where no aid network or resettlement infrastructure operates at anything close to the required scale.

Along the US East Coast, sea level rises independently of whatever global average sea level does. The AMOC normally pulls surface water away from the eastern seaboard, holding coastal sea levels lower than they would otherwise sit. As the circulation weakens and that pull diminishes, sea level from Miami to Boston rises by an additional margin on top of contributions from melting ice and thermal ocean expansion. Storm surges that currently reach a certain inland point reach further. Tidal flooding that currently occurs at a given frequency occurs more often and at greater depth. The drainage systems, road networks, and building foundations of every major East Coast city were designed around historical sea level and historical storm frequency. At 8.1 sverdrups of AMOC flow, both parameters sit outside the design envelopes of that existing infrastructure.

Under the moderate SSP2-4.5 emissions scenario, the weakening reaches 51 percent. Under the high-emissions SSP5-8.5 pathway, the corrected projection rises to 57.7 percent, with a margin of error of just 5.4 percentage points. At that level, the entire uncertainty range sits above 52 percent. Across every emission scenario examined in the corrected projections, the AMOC crosses the IPCC’s substantial collapse threshold before 2100. The corrected numbers also exclude one additional factor entirely. Greenland ice melt delivers fresh water directly into the North Atlantic sinking zones. Fresh water is less dense than salty water, and injecting it into the locations where the AMOC drives its downward flow suppresses that flow directly. Greenland melt is not captured in the 19 variables used to anchor this projection, because it is not consistently tracked across the model archive used. The 51 percent figure does not include Greenland’s contribution. A projection that did would produce a more severe number than the one currently on the table.

The RAPID monitoring array, moored across the Atlantic at 26 degrees North and running continuously since 2004, currently logs the AMOC at a 2005-to-2023 average of 16.9 sverdrups. The corrected projection targets 8.1 sverdrups for the final decade of this century. That array continues active measurement under a joint UK-US ocean monitoring programme, and any sustained departure from the current range over the coming decade will be the first real-world test of whether the corrected trajectory is accurate.