By: Jon Robson
A number of recent high-profile studies have strongly suggested that an important part of the North Atlantic Ocean circulation – the AMOC – has declined and that it is edging closer to a tipping point. Such a long-term decline would have important implications for regional weather and climate for Europe and North America, and a collapse of the AMOC could have serious consequences globally. However, in the sixth assessment report for the Intergovernmental Panel On Climate Change (IPCC) working group 1, the confidence in a long-term 20th Century AMOC decline was assessed as low (down from medium confidence in the IPCC special report on Ocean and Cryosphere in a changing climate, SROCC). So what is going on, and how did we* come to that decision?
The AMOC – or specifically the Atlantic Meridional Overturning Circulation – is a system of currents that brings warm water from the lower latitude Atlantic to the higher latitude Atlantic (see schematic in figure 1). As such, it is a major player in the movement of heat and carbon through the climate system and is, hence, an important regulator of global climate.
Figure 1: Schematic of the AMOC circulation system. Red shows a simplification of warm upper ocean currents, including the gulf stream. Blue shows a simplification of denser (and colder) southward flowing water at depth. Also shown is the RAPID array, which has been observing AMOC at ~26N since 2004. From Srokosz and Bryden, 2015.
There is little doubt that we expect the AMOC to weaken due to increased greenhouse gas emissions. This is because a key driver of the AMOC is the formation of dense seawater as it gets colder and saltier in the northern North Atlantic and Arctic. So, as the world warms, the high-latitude ocean will warm and the melting of ice sheets will dump more freshwater into the ocean. This will decrease the rate at which dense water is formed and slow the AMOC.
However, although we have high confidence in a future AMOC decline, there remain very large uncertainties and many questions still exist. For example: How fast will the AMOC decline in the next few decades? When will the AMOC decline? or, has the AMOC already declined?
Unfortunately, that last question is difficult to assess as routine observations have only existed since the early 2000s. Therefore, we need to use a range of evidence – including observations, model simulations (including ocean-reanalysis, ocean-only and coupled), and indirect observations or “proxies” – to constrain what we think happened.
Over the recent periods (circa 1980-present), these different sources of evidence are all available in some form and, in a recent study**, we show that they generally agree that there has been significant variability in the AMOC and no discernable long-term trend. However, over the longer period of the 20th Century, we can only rely on coupled simulations and proxies.
One such “proxy”, or fingerprint, of an AMOC slowdown, is thought to be a cooling of the subpolar North Atlantic (that’s the bit roughly between 45-65°N) – at least once you make those temperatures relative to global surface temperatures. Such a “warming hole index” (as it is sometimes called) indicates that the subpolar North Atlantic has cooled significantly relative to the rest of the globe – indicating that the AMOC has declined. Furthermore, many other AMOC proxies have also suggested a similar decline and that the AMOC is at its weakest for thousands of years.
However, the results from the proxies are in contrast to the results from coupled models which indicate that the AMOC increased over the 20th Century due to external forcing. Indeed, historical simulations made for the CMIP6 show an increase in the AMOC from 1850–1985 (see top panel of figure 1). This increase is largely due to a competition between historical greenhouse gas and anthropogenic aerosol precursor emissions (see bottom panel of figure 2). Simply put, more models now include aerosol-cloud interactions and, thus, simulate a stronger anthropogenic aerosol forcing which counters the greenhouse gas-induced weakening.
Figure 2: Shows the evolution of the with varying historical external forcings. Top shows the comparison between simulations from CMIP6 and CMIP5. Bottom shows the changes in AMOC in CMIP6 models when only one external forcing is changed at in turn, including greenhouse gasses (hist-GHG, green), and anthropogenic aerosol precursors (hist-aer, blue), and natural changes (e.g. sun or volcanic eruptions, hist-nat, yellow). Taken from Menary et al, 2020.
But what line of evidence is more believable? Well, this is where the waters start to get a bit murkier.
Indeed, there are many reasons to be sceptical about the – usually low resolution – coupled model simulations. For example, there are many shortcomings in how ocean models represent the North Atlantic including the formation of the dense “headwaters” of the AMOC. CMIP6 historical simulations also struggle to simulate other aspects of the climate related to aerosol changes, including Northern Hemisphere temperatures and top of atmosphere shortwave radiation. So, shouldn’t we just trust the proxies?
Well, the problem is that, in the absence of AMOC observations, model simulations have been used to test and (in some cases) calibrate the AMOC proxies. In other words, the different lines of evidence are not fully independent. Furthermore, some studies suggest that the temperature based proxies may not work so well for picking out historically forced variability, and other studies have highlighted that other processes may be contributing to such AMOC “fingerprints”. Finally, there are many proxies, and not all of them agree.
Therefore, to reflect these counteracting lines of evidence we chose to reduce the certainty of a long-term 20th Century AMOC decline to “low confidence” in IPCC AR6.
However, it is important to underline that a long-term decline of the AMOC is a plausible interpretation of the evidence that we have. Furthermore, If the AMOC has declined significantly already, then this would be further – and worrying – evidence that current coupled models may systematically underestimate the sensitivity of the AMOC to greenhouse gasses and the likelihood of a rapid decline in the AMOC. A dangerous position to be in, indeed!
Therefore, there is an urgent need to better understand the AMOC and the current mismatch between model simulations and proxies. To make progress we need to continue to bring a range of observations, models, proxies, and other tools to understand the drivers of the AMOC variability and changes, and to understand the representation of the AMOC in models.
Ultimately, to predict the overall trajectory of the AMOC over the next few decades we still have more to do to understand the AMOC in the past.
*All IPCC WG1 authors who were involved in summarising the AMOC were involved in discussing the confidence statements. These covered Chapter 2 (Karina von Schuckmann (LA) and Gerard McCarthy (CA)), Chapter 3 (Shayne McGregor (LA) and myself (CA)) and Chapter 9 (Sybren Drijfhout (LA)).
** Unfortunately Jackson et al, 2022 is behind a paywall – please email me for a preprint!