Charles Rotter
I spend much of my time here taking the climate press apart, so it is worth saying clearly when one of them gets it right. Paul Voosen’s feature in Science, “Shifting Currents,” is the best piece of mainstream climate reporting I have read this year. It went to sea on the British research ship Discovery, watched the crew haul up two years of ocean data from 2,000 meters down off the Canaries, and came back with the un-dramatic truth about the Atlantic Meridional Overturning Circulation. The un-dramatic truth is the story, and Voosen did not flinch from it. Credit where it is due.
I want to walk through what the piece actually found, because it lands at a useful moment. We are heading into an El Niño, and an El Niño is exactly the weather in which this subject gets loud.
On June 11, NOAA’s Climate Prediction Center moved the equatorial Pacific to an El Niño Advisory. The event is expected to strengthen through the coming winter, and the temperature datasets are already lining up 2026, or 2027, as a serious contender for the warmest year in the instrumental record. I wrote last week about the Godzilla-class adjectives being deployed for the El Niño itself. The relevant point here is the second-order effect. When the global thermometer spikes during an El Niño year, every slow-moving climate-tipping story gets a fresh news peg, and the AMOC is the most marketable tipping story there is.
The timing is sharpened by something else. As regular readers know, the IPCC’s preferred worst-case emissions pathway, RCP8.5 and its successor SSP5-8.5, was formally retired from the next assessment this spring. The doomcasting supply chain runs on inventory, and a record-hot El Niño year is the moment to restock. So the AMOC-collapse headline has been arriving on schedule:
CNN reported a “cold blob” in the North Atlantic as a sign the world is “hurtling toward one of the most alarming climate tipping points.”
Science News ran the proposal to dam the Bering Strait to stabilize the current, quoting a warning that the tipping point may arrive “as soon as the 2040s” and that there is “no time to lose.”
Phys.org headlined a “vital Atlantic current fading far faster,” with a projected 51 percent slowdown by 2100.
A widely shared Substack explainer announced that we have “already passed the atmospheric threshold for recovery,” and that collapse, once it arrives, will be permanent on any human timescale.
Iceland, for its part, has declared AMOC weakening an existential security threat.
Set that against what the people who actually measure the thing have to say.
The AMOC has been monitored continuously since 2004 by the RAPID array, a line of instrument-packed moorings strung across the Atlantic at 26.5 degrees north between the Bahamas and the Canaries. That is more than twenty years of direct observation, which in this field is a long and valuable record. The first thing it showed, and has shown ever since, is that the circulation swings wildly from year to year, and that those swings swamp any long-term trend. Voosen’s reporting is candid about this. The wild swings, he writes, have become a recurring source of both alarm and reassessment, which is a polite way of saying the press grabs the dips and ignores the rebounds.
The clearest example is the one that started the whole modern scare. When RAPID recorded a sharp AMOC decline beginning in 2009, the new analyst on the team assumed he had made a mistake. He had not. The drop was real, but it was not caused by climate change. It was weather, a winter of unusual air-pressure swings that shifted the wind patterns. The circulation recovered. Late last decade it surged again, erasing the earlier decline outright. In the past few years the readings have slipped once more, leaving a net drop of roughly 2 sverdrups since 2004. A sverdrup is about the combined flow of all the world’s rivers, so 2 of them is not nothing. But here is the sentence that did not make it into the collapse headlines: that decline is not yet statistically significant, and it is not clearly tied to global warming. Ben Moat, the cruise’s chief scientist, says another decade of measurement is needed before anyone can make that call. He is openly wary of the words “collapse” and “shutdown.”
The second mooring array, OSNAP, has been watching the high North Atlantic since 2014, precisely where the deep-water formation that drives the AMOC is supposed to happen. Its first decade of data, due to be published shortly, shows a slight increase in the circulation’s strength. It also blew up the textbook picture. The AMOC turns out not to be a single conveyor belt at all, but what one of the field’s leaders calls a belt of belts, with different parts speeding up and slowing down semiautonomously. The famous “ocean conveyor” image, it seems, has been actively misleading people about how the system behaves.
If the direct measurements are this equivocal, where do the mid-century-collapse forecasts come from? They come from two places, and it is worth knowing which is which.
The first is a statistical argument. In 2023, Peter and Susanne Ditlevsen of the University of Copenhagen published a paper estimating that the AMOC would collapse around 2057, with a 95 percent confidence range running from 2025 to 2095. That is the source of most of the “mid-century” headlines. The estimate was not built on the RAPID measurements. It was built on a single proxy, a fingerprint extracted from sea-surface temperatures in the subpolar Atlantic, run through statistical early-warning indicators borrowed from the math of tipping systems. The trouble, as Voosen reports and as the wider literature has noted, is that those indicators have a habit of pointing toward collapse regardless of what the underlying data are doing. A confidence interval that spans 2025 to 2095 is not a forecast. It is an admission that the method cannot tell you the answer.
The second is modeling. René van Westen and Henk Dijkstra at Utrecht have produced the most-cited collapse simulations, including the 2024 “early-warning signal” paper and the 2026 Gulf Stream “precursor” study. Van Westen puts the tipping risk at around 2.5 degrees of warming and a buoyancy flip around 2063. These are real results from serious people, and I am not going to wave them away. But Voosen does the thing good reporters do, which is to ask what it takes to get those outcomes. The answer is freshwater inputs well beyond what Greenland is actually expected to deliver.
And when modelers feed in realistic Greenland melt instead, the picture changes. Two preprints in the past year, one from the Danish Meteorological Institute and one from Utrecht, ran emissions out to the year 2250 with warming of up to 7 degrees. In both, the AMOC weakened by about 40 percent. In neither did it collapse. Both found the weakening reversible: stop the emissions, and the circulation comes back. A third, higher-resolution simulation held up even under quadrupled CO2. Tellingly, when the same model was run at low resolution, the AMOC collapsed, which suggests the collapses in the older models may be partly an artifact of coarse grids rather than a feature of the ocean. A 2025 paper in Nature reached a similar conclusion from a different direction, finding the AMOC stabilized by wind-driven upwelling in the Southern Ocean and unlikely to collapse this century.
Even the deep past has stopped cooperating with the collapse story. The classic evidence held that the AMOC switched off entirely at the end of the last ice age when meltwater flooded the Atlantic. New work using marine sediment cores, presented this February, suggests it may not have collapsed at all during that period. If anything, the paleoceanographer who did the work says, it may have been stronger.
Listen to the people on the boat
What makes Voosen’s piece valuable is not that a skeptic blog approves of it. It is that the restraint is coming from the scientists who run the arrays. Gerard McCarthy, a RAPID scientist, has publicly warned that exaggerated collapse headlines generate seesawing narratives that confuse policymakers rather than equip them. Eleanor Frajka-Williams, who previously led RAPID, notes that some people “may be inclined to be more inflammatory,” and adds that there is no unified simple theory of the AMOC because one does not exist. Fiamma Straneo of Harvard says the dire collapse scenarios can be a distraction from the climate impacts we are actually sure are happening. These are not contrarians. They are the field, asking the press to calm down.
The institutional response, predictably, runs the other way. Iceland has declared AMOC weakening a security and existential threat, the Nordic Council of Ministers has published a tipping-risk assessment, and a joint European initiative has convened in Brussels to produce a full AMOC review modeled on the UN climate reports. The observed circulation refuses to do anything alarming, and the policy apparatus is being built on top of it anyway.
So here is where we are. The El Niño will arrive, the thermometers will spike, and the AMOC-on-the-brink stories will spike along with them, because a record-hot year is when the genre sells best. The collapse forecasts will lead with the statistical estimator that cannot distinguish a real signal from noise, and with the model runs that require more meltwater than Greenland can supply, and they will leave out the twenty years of moorings showing a system that swings hard and trends gently and refuses to be called a shutdown by the man who measures it. Then, a while later, the readings will report another swing, somebody will quietly extract a smoothed line, and the reassessment will follow the alarm the way it always does. Rinse and repeat.
