← Back to the dashboard

How we track the Atlantic overturning circulation

What the dashboard shows, where the numbers come from, and what they can’t tell you.

The short answer

The Atlantic overturning circulation (AMOC for short, “the conveyor” on the dashboard) carries warm surface water north. In the far North Atlantic that water gives off its heat, becomes denser, sinks, and flows back south deep below the surface. It is not the Gulf Stream. The Gulf Stream is driven mainly by wind and will not switch off. The overturning is driven by differences in temperature and saltiness, which is why a warmer, fresher North Atlantic can weaken it.

It has weakened. Direct measurements across the Atlantic began in April 2004. In the first five years the overturning moved about 18.4 million cubic metres of water per second; in the last five years of the record, to November 2024, about 16.6, a tenth less. (Oceanographers call a million cubic metres per second a sverdrup, Sv, and the dashboard uses that unit.) The flow swings a lot: single months in the record range from 7 to 25 Sv, and the last twelve months were strong at 17.9, so no single year settles anything. The RAPID team’s own trend for the twenty years is −0.9 ± 0.7 Sv a decade.

The UN’s climate science panel, the IPCC, judged in 2021 that the overturning will very likely weaken further this century, and that a collapse before 2100 is unlikely but cannot be ruled out. Several studies published since then argue the risk is higher than that.

How we know

Direct measurement at 26°N (RAPID). Since 2004, a line of moorings across the Atlantic, from Florida and the Bahamas to the coast of Africa, has recorded temperature, saltiness, pressure and currents from the surface to the seafloor. From those readings, the UK–US RAPID programme calculates how much water the overturning moves. It is the longest continuous direct record of the AMOC, and it is the series in the dashboard’s main chart. The British Oceanographic Data Centre publishes each release openly (UK Open Government Licence); the dashboard uses release v2024-2 of October 2026, which runs to November 2024 and revised the calculation slightly. We average its monthly values by calendar year and compare the last five years with the first five.

Farther north (OSNAP). Since 2014, a second array called OSNAP has measured the overturning between Canada, Greenland and Scotland, where the sinking actually happens. It sees a different part of the same circulation, so its numbers are not directly comparable with RAPID’s. The dashboard shows its August 2014 to June 2020 average (16.7 ± 0.6 Sv, from the team’s own paper, Fu et al. 2023) as context. OSNAP has since published its series to 2022; we do not use the data files, because the dataset asks users to contact the investigators first.

Indirect evidence (the “cold blob”). Over the past century almost the whole planet has warmed, but a patch of ocean south of Greenland has warmed far less, and at times cooled. Many scientists read that as a sign that less heat is being carried north. It is a fingerprint, not a measurement of flow: winds and the exchange of heat with the air affect it too. The dashboard measures it with NOAA’s daily satellite and ship sea-surface temperatures (OISST): the average of a box south of Greenland (45–60°N, 20–50°W) minus the average of the ocean between 60°S and 60°N, both against 1971–2000, over the last 12 months. Caesar and colleagues (2018) define their index over the subpolar gyre region, for November to May only, from a record back to 1870; ours is a simpler rectangle over all months with three years of daily data, so it tracks the same pattern but is not their index. A negative value means the region is lagging the warming ocean. Until November 2026 the card showed NOAA’s Atlantic Multidecadal Variability index, which covers the whole North Atlantic and ended in January 2023.

The longer record (North Atlantic temperature index). A separate chart shows the average sea-surface temperature of the whole North Atlantic since 1856, minus a straight line fitted through the record, which scientists call the Atlantic Multidecadal Variability index. NOAA’s Physical Sciences Laboratory built it from the Kaplan temperature analysis by exactly that recipe (we checked: applying it to their undetrended averages rebuilds their index to a thousandth of a degree), and stopped in January 2023 because the Kaplan analysis stopped. We keep those values as they are. To carry the record on we use the North Atlantic (0–60°N) index that NOAA’s National Centers for Environmental Information publish every month from their ERSSTv5 analysis (Huang et al. 2017, doi:10.1175/JCLI-D-16-0836.1), the series NOAA’s own AMO page points to. It is not detrended, so we subtract a straight line fitted over the same span as the Kaplan record, January 1856 to January 2023. Over 2000 to 2022, when both exist, their yearly averages correlate at 0.92 (0.88 for single months), but ERSSTv5 runs 0.13°C warmer on average, so part of the step up after 2022 on the chart comes from the change of analysis. The chart draws both lines over the years they share and says which part comes from which. We also recomputed three months of NCEI’s index (May 2010, August 2024, October 2026) from NOAA’s gridded ERSSTv5 temperatures, with our own rough outline of the basin, and came within 0.03°C of the published values.

How current each part is

Where scientists disagree

The question is less whether the overturning can weaken than how fast, and whether it could pass a point beyond which it does not recover. The papers marked on the dashboard’s chart take different positions:

The IPCC’s 2021 assessment came before most of this work, so its next report will be the first to weigh it.

What the dashboard can’t tell you

It cannot tell you that a collapse is coming. Twenty years of direct measurement are invaluable but short for a circulation that varies over decades. Anything before 2004 relies on indirect evidence, such as ocean sediments and temperature fingerprints, with wider uncertainty. A decline since 2004 shows weakening over that period. It is not a countdown.

It cannot forecast your local weather either. A weaker overturning could make winters in parts of northwest Europe colder than they would otherwise be, shift tropical rain belts, and raise sea level along parts of the US East Coast. How much, when and where depends on emissions, melting ice, winds and the climate models used. The “near you” section describes the kinds of effects expected in your region, not a forecast for your town.

Further reading

Credits

The direct record exists because the RAPID and OSNAP teams have kept instruments in the open Atlantic for years, and the British Oceanographic Data Centre publishes their data. The sea-surface temperature index comes from NOAA’s Physical Sciences Laboratory, and the research list from OpenAlex. This dashboard collects and presents their work.

← Back to the dashboard