RealClimate logo

The underestimated danger of a breakdown of the Gulf Stream System

Filed under: — stefan @ 4 January 2017

A new model simulation of the Gulf Stream System shows a breakdown of the gigantic overturning circulating in the Atlantic after a CO2 doubling.

A new study in Science Advances by Wei Liu and colleagues at the Scripps Institution of Oceanography in San Diego and the University of Wisconsin-Madison has important implications for the future stability of the overturning circulation in the Atlantic Ocean. They applied a correction to the freshwater fluxes in the Atlantic, in order to better reproduce the salt concentration of ocean waters there. This correction changes the overall salt budget for the Atlantic, also changing the stability of the model’s ocean circulation in future climate change. The Atlantic ocean circulation is relatively stable in the uncorrected model, only declining by about 20% in response to a CO2 doubling, but in the corrected model version it breaks down completely in the centuries following a CO2 doubling, with dramatic consequences for the climate of the Northern Hemisphere.

The potential instability of the Atlantic Meridional Overturning Circulation or AMOC – commonly known as the Gulf Stream System – has been a subject of research since the 1980s, when Wallace Broecker warned in an essay in Nature of Unpleasant Surprises in the Greenhouse. The reason for this was growing evidence of abrupt climate changes in the history of the Earth due to instability of Atlantic currents.

Fig. 1 Schematic of the Atlantic ocean circulation (simplified). In red the relatively warm surface flow is seen, in blue the cold deep water flow. The northward surface flow and southward deep flow together make up the Atlantic Meridional Overturning Circulation (AMOC), popularly dubbed Gulf Stream System. Image by S. Rahmstorf (Nature 1997), Creative Commons BY-SA 4.0. 

Why the AMOC has a tipping point

The basic physical mechanism of this instability was already described by Henry Stommel in 1961. The freshwater balance (precipitation minus evaporation), which determines the salt content, is central to this. Freshwater continually flows into the northern Atlantic through precipitation, rivers and ice-melting. But supply of salty waters from the south, through the Gulf Stream System, balances this. If however the current slows, there is less salt supply, and the surface ocean gets less salty. This fresher water is lighter than saltier water and therefore cannot sink into the depths so easily. Since this sinking – the so-called deep water formation – drives the Gulf Stream System, the current continues to weaken. There is a critical point when this becomes an unstoppable vicious circle. This is one of the classic tipping points in the climate system.

However, it’s still unclear where exactly this tipping point is. Most models show a significant slowdown in the Gulf Stream System by 20% to 50% in typical global warming scenarios up to the year 2100, but do not exceed the tipping point that would lead to its collapse. However, there is a large spread between different models – which is not surprising since the stability of the Atlantic flow depends on a subtle balance in the salinity and thus also in the freshwater budget, which is only inaccurately known. In addition, there have long been serious indications that the models are not only inaccurate, but perhaps all systematically biased towards an exceedingly stable AMOC. We discussed these papers in a review article in PNAS in 2009.

What makes the new study different?

According to lead author Wei Liu, the starting point of the new study was my paper from 1996 on the relationship between the freshwater balance and stability of the flow. Back then I showed how to determine the stability of the AMOC from an analysis of the freshwater transport in the Atlantic at 30° south. The decisive factor is whether the AMOC brings freshwater into the Atlantic basin or whether it exports it (in the latter case, working to increase salinity in the Atlantic). My article ended with the suggestion to clarify this from observational data. That was later done by colleagues from Holland (Weijer et al. 1999). Several studies followed which performed this diagnosis for different climate models (e.g., Pardaens et al. 2003, de Vries and Weber 2005Dijkstra 2007, Drijfhout et al. 2010, Hawkins et al. 2011). According to the observational data, the AMOC is exporting freshwater, which is why freshwater will accumulate in the Atlantic when the AMOC breaks down. That is precisely the instability described by Stommel 1961 and Broecker 1987. In the models, on the other hand, the AMOC in most cases imports freshwater, so the flow is fundamentally stable there. The differences in AMOC stability between different models cannot be understood without the fundamental criterion of whether the AMOC imports or exports freshwater, and by what amount. Liu et al. 2014 have identified a known common bias in all coupled climate GCMs without flux adjustments, the “tropical bias”, which makes them import freshwater in contrast to what observations show for the real ocean. A model bias towards stability is also consistent with the fact that most models underestimate the cooling trend observed in the subpolar Atlantic, which is indicative of an ongoing significant AMOC weakening, as we have argued (Rahmstorf et al. 2015).

The new study attempts to correct this model deficit by modifying the freshwater exchange at the sea surface in a model by using a so-called flux correction (which also involves the heat exchange, but this should be secondary). As a result, the salinity distribution in the ocean of the model for today’s climate is in better agreement with that of the real ocean. This is an important criterion: while precipitation and evaporation over the oceans are difficult to measure and therefore only very imprecisely known, we have detailed and precise information about the salinity distribution from ship measurements. Apart from the improved salinity distribution, this correction has no significant influence on the model climate for the present.

And now the result …

With both model variants – with and without the subtle correction of the salinity distribution – an experiment was performed in which the amount of CO2 in the air was doubled. The reaction of the Atlantic circulation is shown in the following graph. Without correction, the AMOC once again proves to be very stable against the massive disturbance. With the correction, in contrast, the flow breaks down in the course of about 300 years. It has lost a third of its strength after 100 years.

Fig. 2 Time series of the Atlantic flow (AMOC) in the two model variants: without correction (blue) and with correction (orange). In model year 201, the CO2 concentration in the model is doubled and then left at this level. Source: Liu et al., Science Advances 2016.

As expected, the breakdown of the heat-bringing Gulf Stream System leads to a cooling in the northern Atlantic, as shown in Figure 3. Land areas are also affected: besides Greenland and Iceland mainly Great Britain and Scandinavia.

Fig. 3 Temperature change in the winter months (DJF), 300 years after CO2 doubling in the experiment. Due to the almost completely extinct Atlantic flow, the northern Atlantic region has cooled significantly. Source: Wei Liu, with permission.

This new study is certainly not the last word on this important question. Compared to the measured data the correction appears to be somewhat too strong – the adjusted model version might therefore be too unstable. As computing time is scarce and expensive, the CO2 concentration in the experiments was abruptly doubled, rather than gradually ramped up in a more realistic emission scenario. The experiment was carried out with only one climate model; for robust conclusions, one usually waits until a series of models shows consistent results. (However, consistent with the new results two earlier climate GCMs and a number of simpler models have shown an AMOC that exports freshwater and is bistable, i.e. could potentially pass a tipping point and break down, as discussed by Liu et al. 2014.)

Also, no meltwater influence from the dwindling continental ice on Greenland was taken into account, which could additionally weaken the flow. On this topic, only three weeks ago a new study was published (Bakker et al. 2016) comparing future warming scenarios, once with and once without consideration of the influx of Greenland meltwater. (An emulator was used for this study; that is a highly simplified computer model that reproduces the results of complex circulation models in a time-saving way, so that many experiments can be performed with it.) With unmitigated emissions (RCP8.5 scenario), the Gulf Stream System weakens on average by 37% by the year 2300 without Greenland melt. With Greenland meltwater this doubles to 74%. And a few months ago, a study with a high-resolution ocean model appeared, suggesting that the meltwater from Greenland is likely to weaken the AMOC considerably within a few decades (Böning et al. 2016 – as we reported).

There are, therefore, two reasons why thus far we could have underestimated the risk of a breakdown of the Gulf Stream System. First, climate models probably have a systematic bias towards stable flow. Secondly, most of them do not take into account the melting ice of Greenland. As the new studies show, each of these factors alone can lead to a much stronger weakening of the Gulf Stream system. We now need to study how these two factors work together. I hope these worrying new results will encourage as many other research groups as possible to pursue this question with their own models!


Washington Post: Scientists say the global ocean circulation may be more vulnerable to shutdown than we thought

Climate Central: Potential for Collapse of Key Atlantic Current Rises

The Verge: Climate change may shut down a current that keeps the North Atlantic warm

The Atlantic: The Atlantic Ocean and an Actual Debate in Climate Science

Video lecture on the Gulf Stream System

More on the Gulf Stream System slowdown at RealClimate

Q & A about the Gulf Stream System slowdown and the Atlantic ‘cold blob’

AMOC slowdown: connecting the dots

What’s going on in the North Atlantic?

A hypothesis about the cold winter in eastern North America

Blizzard Jonas and the slowdown of the Gulf Stream System


  1. W.S. Broecker, "Unpleasant surprises in the greenhouse?", Nature, vol. 328, pp. 123-126, 1987.
  2. H. STOMMEL, "Thermohaline Convection with Two Stable Regimes of Flow", Tellus, vol. 13, pp. 224-230, 1961.
  3. T.M. Lenton, H. Held, E. Kriegler, J.W. Hall, W. Lucht, S. Rahmstorf, and H.J. Schellnhuber, "Tipping elements in the Earth's climate system", Proceedings of the National Academy of Sciences, vol. 105, pp. 1786-1793, 2008.
  4. M. Hofmann, and S. Rahmstorf, "On the stability of the Atlantic meridional overturning circulation", Proceedings of the National Academy of Sciences, vol. 106, pp. 20584-20589, 2009.
  5. S. Rahmstorf, "On the freshwater forcing and transport of the Atlantic thermohaline circulation", Climate Dynamics, vol. 12, pp. 799-811, 1996.
  6. W. Weijer, W.P.M. de Ruijter, H.A. Dijkstra, and P.J. van Leeuwen, "Impact of Interbasin Exchange on the Atlantic Overturning Circulation", Journal of Physical Oceanography, vol. 29, pp. 2266-2284, 1999.<2266:IOIEOT>2.0.CO;2
  7. A.K. Pardaens, H.T. Banks, J.M. Gregory, and P.R. Rowntree, "Freshwater transports in HadCM3", Climate Dynamics, vol. 21, pp. 177-195, 2003.
  8. P. de Vries, "The Atlantic freshwater budget as a diagnostic for the existence of a stable shut down of the meridional overturning circulation", Geophysical Research Letters, vol. 32, 2005.
  9. H.A. Dijkstra, "Characterization of the multiple equilibria regime in a global ocean model", Tellus A, 2007.
  10. S.S. Drijfhout, S.L. Weber, and E. van der Swaluw, "The stability of the MOC as diagnosed from model projections for pre-industrial, present and future climates", Climate Dynamics, vol. 37, pp. 1575-1586, 2010.
  11. E. Hawkins, R.S. Smith, L.C. Allison, J.M. Gregory, T.J. Woollings, H. Pohlmann, and B. de Cuevas, "Bistability of the Atlantic overturning circulation in a global climate model and links to ocean freshwater transport", Geophysical Research Letters, vol. 38, pp. n/a-n/a, 2011.
  12. W. Liu, Z. Liu, and E.C. Brady, "Why is the AMOC Monostable in Coupled General Circulation Models?", Journal of Climate, vol. 27, pp. 2427-2443, 2014.
  13. S. Rahmstorf, J.E. Box, G. Feulner, M.E. Mann, A. Robinson, S. Rutherford, and E.J. Schaffernicht, "Exceptional twentieth-century slowdown in Atlantic Ocean overturning circulation", Nature Climate Change, vol. 5, pp. 475-480, 2015.
  14. P. Bakker, A. Schmittner, J.T.M. Lenaerts, A. Abe-Ouchi, D. Bi, M.R. van den Broeke, W. Chan, A. Hu, R.L. Beadling, S.J. Marsland, S.H. Mernild, O.A. Saenko, D. Swingedouw, A. Sullivan, and J. Yin, "Fate of the Atlantic Meridional Overturning Circulation: Strong decline under continued warming and Greenland melting", Geophysical Research Letters, vol. 43, pp. 12,252-12,260, 2016.
  15. C.W. Böning, E. Behrens, A. Biastoch, K. Getzlaff, and J.L. Bamber, "Emerging impact of Greenland meltwater on deepwater formation in the North Atlantic Ocean", Nature Geoscience, vol. 9, pp. 523-527, 2016.

70 Responses to “The underestimated danger of a breakdown of the Gulf Stream System”

  1. 51
    Matt Skaggs says:

    “Response: But your marks have no connection to anything else until it’s calibrated – which requires some kind of model. It’s the same way that a proxy measurement of temperature is used. – gavin”

    You have really pushed this argument – including a couple full posts – and it is frankly bizarre. The question here was “whether the AMOC brings freshwater into the Atlantic basin or whether it exports it.” The claim was made that the original model output was corroborated by output from a different model, with the latter model being informed by observational data. Now if I went out and deployed a zillion sensors across all boundaries of the basin, I could directly measure whether the model results fit with the measured results. (Arguing whether or not the interpretation of the sensor results requires a model is merely semantics.) If the direct measurement results showed that the flux has the opposite sign from what the models showed, in any scientific field in which rationality reigns, the model output would be rejected in favor of the measurement results. Can we at least agree on that?

    [Response: Yes. But the thing here is that any direct measurements we can make are point measurements of salinity and (though that’s much more difficult) currents. The freshwater transport by the AMOC is not a measurable quantity, it is a computed quantity. It requires computing an integral over the salinity times the flow speed – and to complicate things, not just the directly measurable flow speed but that which is part of the AMOC (rather than the wind-driven subtropical gyre in the South Atlantic). So there is no way to measure directly “whether the AMOC brings freshwater into the Atlantic basin or whether it exports it.” All you can do is compute this – using the best observational data you have in this complex calculation. -stefan]

  2. 52
    Hank Roberts says:

    Carbon isotope ratios at 21:00 in the Peter Ward video recommended above:
    Those not comprehending how this tells us something should watch.

  3. 53
    Michael Eby says:

    [Response: Hi Michael, good that you stop by! Have you had a look at Fig. 1 of Wei Liu’s paper? It shows that compared to observational estimates, all CMIP5 models have an AMOC that is a lot too stable, judging from the AMOC stability indicator. See also his previous 2014 paper. So I don’t think it is correct to just treat all models as equally likely and say that if only one model shows a collapse than this is unlikely to be correct. Rather, I think there is good evidence that all CMIP5 models are biased towards a too stable AMOC so I’d say it is likely that these are all wrong on this particular issue. -stefan]

    I am not so sure that the metric they use is a particularly reliable indicator of AMOC stability. Several studies show that it is not. However, even if you do believe the present day strength of this metric is reliable indicator of the likelihood of a collapse (which I do not), at present day their model is more unstable than the uncertain “data” would indicate (their Figure 1). After you add in the physics distorting use of flux adjustments, I would say it is more likely that this single model is wrong. I think more caution is required when interpreting this study.

  4. 54
    Geoff Beacon says:

    At #16 I asked

    How do short term climate forcing agents (like methane and black carbon) contribute to the danger of triggering a tipping point in the Gulf Stream System?

    I’ve just found Centuries of thermal sea-level rise due to anthropogenic emissions of short-lived greenhouse gases by Ramanathan et. al.

    Seems relevant but I haven’t yet paid to access it. The “significance” includes:

    Our study shows that short-lived GHGs contribute to thermal expansion of the ocean over much longer time scales than their atmospheric lifetimes. Actions taken to reduce emissions of short-lived gases could mitigate centuries of additional future sea-level rise.

    A bit from the abstract:

    For example, at least half of the TSLR due to increases in methane is expected to remain present for more than 200 y, even if anthropogenic emissions cease altogether, despite the 10-y atmospheric lifetime of this gas. Chlorofluorocarbons and hydrochlorofluorocarbons have already been phased out under the Montreal Protocol due to concerns about ozone depletion and provide an illustration of how emission reductions avoid multiple centuries of future TSLR. We examine the “world avoided” by the Montreal Protocol by showing that if these gases had instead been eliminated in 2050, additional TSLR of up to about 14 cm would be expected in the 21st century, with continuing contributions lasting more than 500 y.

    Any thoughts?

  5. 55
    Geoff Beacon says:

    Correction: The authors of the paper I mentioned were Kirsten Zickfeld, Susan Solomon, and Daniel M. Gilford

  6. 56
    Pat says:

    So simple a Caveman can do it..These… added to these in the Gulfstream powers Earth .. .Any questions???

  7. 57
    Pat says:

    So simple a Caveman can do it.. These .. added to these in the Gulfstream powers Earth .. .Any questions???

  8. 58
    Simon C says:

    Thomas @ 42 I think that Gavin makes a totally valid point that, while it may be uncomfortable to some, is worth recalling from time to time: all scientific insights and methods depend on models that are more or less closely related to practical experience, and the whole point of science is to refine these models. For example, the use of the thermometer depends on the understanding (ie the model or hypothesis) that mercury will react in the same way to ambient temperature conditions, other things being equal, and the use of a thermometer measurement to establish local temperature depends on the understanding (ie the model) that the temperatures being taken are a reasonable representation of conditions over a particular area over a particular time. The assumption (ie the model) that plumblines would always point towards the centre of the Earth led to considerable confusion in surveying the Himalayas. Newtonian assumptions about time led to errors in the prediction of astronomical events. I could go on, but I would think this is probably enough. It might be uncomfortable for certain philosophical positions, but models are all we have (actually all that anyone has, but that is another story …) We like science and trust it because it generally works and has a habit of getting better; not a bad record for a human activity.

  9. 59
    Thomas says:

    This topic is reaching the MSM

    There are active discussions in climate science—they’re just not about this. So before we all have to talk about a topic on which there is near total scientific agreement, I thought it might be fascinating to examine a real area of dispute in the field. And one of the most consequential disagreements is about something called the Atlantic Meridional Overturning Circulation, or AMOC.

  10. 60
    Ray Ladbury says:

    Thomas: “Like burrowing down into that proverbial rabbit hole, it’s not always helpful in the long term. :-)”

    I’m sorry you don’t find the truth helpful.

  11. 61
    t marvell says:

    Why are the AMOC models based on push only, rather than pull? The huge upwelling in the Pacific should draw water from the THC deep currants. Any pull mechanism would make the AMOC more stable.

    [Response: What makes you think they are based “on push only”? That’s wrong. -stefan]

  12. 62
    Lynn says:

    I think I got my answer re #13,#16 & #24 question about the Gulfstream slowdown making the tropics, incl the Gulf of Mexico?, warmer.

    If the SSTs in those already cyclone-prone areas get warmer, wouldn’t that also increase the risk of stronger and/or more freq cyclones in those areas (heat energy turning into kinetic energy) — at least providing a necessary if not sufficient cause?

  13. 63
    Susan Anderson says:

    Thanks Pat (@~57) for the videos on tidal and other forms of ocean energy. I’d like to see references to these efforts early and often!

  14. 64
    Nicholas K says:

    So, what are the actual likely effects of this happening then. What I’ve got so far is:
    – Where I live (the north of England), the climate becomes a bit more Scandinavian. Not sure what the costs would be of adjusting to that, although the Scandivians do seem to cope.
    – Life in Iceland may not be viable.
    – It will also get stormier, which will obviously cause some damage.
    – Sea level will rise in the Eastern US, threatening e.g. New York with more flooding.
    – Somebody mentioned effects on the ocean ecosystem from oxygen not getting circulated, and also whether this might have an effect on methane getting released and stuff?
    – Could warmer Canada and Russia cause other positive feedbacks?

  15. 65

    for those who like to say that we should not over-estimate or otherwise say scary things about global warming because reality is scary enough: look for a headline on this or any other credible climate science blog that talks about the danger of overestimating some aspect of climate change. The truth is that we should at least slightly overestimate the danger of things so that our engineered solutions include a little extra strength. Engineers don’t build things to stand the average storm or stress. Connect the dots, kemosabe. If you are bringing the “reality is scary enough” message you are discouraging appropriate action, you are not that different from the average science denier in terms of response to the situation that we have created.

    Daily CO2

    January 17, 2017: 406.40 ppm
    January 17, 2016: 402.78 ppm

    read’m and weep.


  16. 66
    Pat says:

    @ Susan #63 I would be glad to give a blackboard tutorial and explain how the idea works and can be combined to a board of scientists if they are interested in ending fossil and nuclear fuels…. Pascal and Bernoulli had it figured out many moons ago..

  17. 67
    Student says:

    This is a blog?? What the hell?! It looks stupid and it doesn’t help me complete my homework for Honors Earth and Space Science. Re-think your format please. It’s so boring that I almost fell asleep five times while trying to reading it.

  18. 68
    Fergus Brown says:

    @ 67 : You are a student? WTH? If you’re doing an honours in E&SS you should be able to distinguish spit from shinola. Good luck with the homework.

    Note also: Brunnabend et. al: is actually quite interesting on the AMOC. Anyone have any thoughts on this work?

  19. 69
    Ric Merritt says:

    Hey is that the same Student who introduced the T test?? Darn, we all had better take this criticism to heart, from such a distinguished source!

  20. 70
    Digby Scorgie says:

    Student @67

    This website is not a blog in the usual sense. It is run by climate scientists who post articles on their research from time to time. They don’t often comment themselves. The commenters are just people interested in climate science, some of whom are scientists but not climate scientists. You want the MOOC at