Earth & climatePreprintSimulation4 min read

WHAT IF CLIMATE MODELS MAKE THE ATLANTIC CIRCULATION TOO FRAGILE?

The Atlantic Meridional Overturning Circulation, or AMOC, is the system of currents that carries warm, salty water north near the surface of the Atlantic and returns it south at depth. Its possible collapse is widely seen as one of the most severe consequences of continued warming, because it could start abruptly and be impossible to stop: a “tipping point”. For a long time, the working assumption among climate modellers was that their models made the AMOC too stable, underplaying that risk.

Models that now collapse

That picture has shifted. A recent analysis of the latest generation of climate models, known as CMIP6, extended their simulations to the year 2300 and found a partial collapse of the AMOC, in which the deep part of the circulation vanishes completely. At least one model produced it in every emission scenario, and every model did under the highest-emission scenario.

What survives is a weak, shallow overturning, but it no longer carries heat to the northern North Atlantic. For Europe, the result would have the same climate impact as an AMOC that disappeared entirely.

In a new perspective paper, Sybren Drijfhout, of the University of Southampton and the Royal Netherlands Meteorological Institute, and Alejandra Sanchez-Franks, of the UK’s National Oceanography Centre, ask the reverse question: could these models now be too unstable? Two observations motivate it. Measurements show no clear downward trend in the AMOC so far. And some CMIP6 models are known to be too sensitive to greenhouse gases and to the removal of aerosol pollution, which could push them towards a spurious collapse. The authors stress that they are not claiming the models are wrong — only that, if they are right, the simulated collapses are “very worrying if not threatening”, so the question deserves testing.

In the simulations, the collapse follows a sequence. Warming at the surface of the subpolar North Atlantic and the Nordic Seas first stops the deep winter mixing that forms dense water: the depth reached by this winter mixing (the “mixed layer”, deepest in March) shrinks from about 1,000 metres to less than 200, often 100, within 20 to 30 years. Without new dense water, the east–west pressure difference that drives the deep southward flow, between about 1 and 2 km down, and that is kept up by dense water piled against the western edge of the basin, slowly erodes, which takes about 50 years. The deep circulation fades over 50 to 100 years.

Several feedbacks reinforce the slide: a weaker AMOC brings less salt north, fresher surface water, being lighter, sinks and mixes less, and the atmosphere carries more heat north to compensate, further reducing the ocean’s heat loss. Sea ice that regrows in a cooler region adds salt back, because freezing seawater releases its salt into the water below, but only after mixing has already collapsed, and too weakly to stop it.

The authors list where each link could be exaggerated in models whose grid is too coarse to capture ocean eddies. In high-resolution models, for example, currents running along the western edge of ocean basins heat up three times faster than the ocean interior, because eddies concentrate heat there; something similar near the mixing regions could keep surface water from stratifying so easily, that is, from forming a light top layer that no longer mixes with the water below. They propose specific checks for each step: tracking water masses, comparing models that collapse with those that do not, and benchmarking all of them against observations.

What the ocean says so far

Observations cannot yet settle the matter. The OSNAP array in the subpolar Atlantic has only run since 2014, the RAPID array at 26° N since 2004. The Florida Current, measured by a submarine cable since 1982, has stayed broadly stable, and the full AMOC record shows a weak negative trend that is only marginally significant. Telling a climate trend apart from natural variability would likely take 30 to 70 years of continuous records.

Yet the North Atlantic is changing in unprecedented ways: a decline over two decades of the deep current flowing south along the western edge of the Atlantic at 26° N, a drop in salinity (a “freshening”) in 2012–2016 that was the largest in over a century, and temporary shutdowns of deep winter mixing (deep convection) in the Labrador Sea in 2021–2023. Whether these changes are linked to a weakening of the AMOC is still an open question.

The authors’ conclusion is a programme rather than a verdict. One proposed warning sign shows why the layering of the water matters: a “critical salinity”, the point at which surface water, even cooled to −1.9 °C, stays lighter than the deep water and can no longer sink. It is crossed at the very moment deep mixing collapses, so it gives no advance warning, but it shows that this layering decides the timing. They would start by checking whether today’s simulated stratification and salt feedbacks in the regions of deep winter mixing match measurements, then improve how models represent eddies and the dense water spilling south of Denmark Strait, which is missing from almost every climate model.