CO2 sequestration during glacial maxima

Over the last ~800000 years, the Earth has undergone a series of cycles in which it has switched from being in a glacial, with large ice sheets, into an interglacial, with small ice sheets. Known as glacial cycles, these are associated with orbital variations called Milankovitch Cycles, and typically involve global temperature changes of ~5oC, and ~100ppm changes in atmospheric CO2 (changing from ~180ppm to ~280ppm, and then back again).

One of the puzzles associated with these cycles is how atmospheric CO2 can have a period where it settles to ~180 ppm, and then another period where it settles to ~280ppm. My understanding of why this is an issue is illustrated by the figure on the right. If you consider the basics of carbonate chemistry (which I discuss in these posts) then the regions of the ocean that exchange CO2 with the atmosphere should have substantially different pH and DIC (Dissolved Inorganic Carbon) levels during these glacial and interglacial periods.

Given that low atmospheric CO2 is associated with high pH and low DIC, what the above implies is that during the glacial periods, a large amount of CO2 must be sequestered in a region that doesn’t exchange with the atmosphere. It can’t be sequestered in the lithosphere because the timescales for doing so is too long. It also can’t be the biosphere. What seems likely is that, during glacial cycles, a large amount of CO2 is sequestered into the deep ocean in such a way that it is not able to exchange with the atmosphere. When we move into an interglacial, this CO2 is released, and atmospheric CO2 rises.

Credit: Rae et al. (2018)

The reason I’m writing about this is that I came across an interesting paper by James Rae, who I met when I went to see Katharine Hayhoe speak in Edinburgh. The paper is about CO2 storage and release in the deep Southern Ocean on millennial to centennial timescales. The basic idea, as illustrated by the Figure on the left, is that during a glacial maximum, extensive sea ice in the Southern Ocean essentially prevents portions of the deep ocean from exchanging with the atmposphere, sequestering CO2 in these regions of the ocean. As the sea ice retreats, these regions can then exchange with the atmosphere, and atmospheric CO2 rises.

Credit: Rae et al. (2018)

What Rae et al. (2018) did is to look at deep-sea coral boron isotope data that track the pH — and thus the CO2 chemistry — of the deep Southern Ocean over the past forty thousand years. What they found is shown in the figure on the right. What it indicates is that as we moved from the Last Glacial Maximum (LGM) into the current inter-glacial, the pH of the deep Southern Ocean increased. This is consistent with the deep Southern Ocean sequestering CO2 during the LGM and then releasing it as we moved into the inter-glacial. As I mentioned at the beginning of this post, this is also the opposite of what we’d have expected if this region had been exchanging CO2 with the atmosphere during the LGM.

So, this new work may have resolved one of the puzzles associated with the glacial cycles and demonstrates that a large amount of CO2 was indeed probably sequestered in the deep Southern Ocean during the glacial periods.

My explanation at the beginning of the post may be wrong, but it’s getting late so I will think about this a bit more if I get a chance tomorrow.

Just to add that I think my explanation is broadly correct (although maybe not entirely). As this article, highlighed by BBD, points out you can’t explain the ~100ppm change as simply due to enhanced uptake by the oceans when it cools. I played around a bit with my carbonate chemistry code and to get a 100ppm reduction in atmospheric CO2 with a 5oC reduction in upper ocean temperature would seem to require the ocean pH goes up and the DIC goes down, which doesn’t really make sense if the upper ocean is taking up CO2. Additionally (as the article highlighted points out) the changes to the biosphere and changes to ocean salinity (due to ice sheet changes) further inhibits this. Therefore, there must be somewhere else that the CO2 is being sequestered during glacial maxima.

21/11/2018: I’ve edited the first paragraph to clarify that the climate variations are typically referred to as glacial cycles, while Milankovitch cycles are the orbital variations that are thought to be involved with triggering these glacial cycles.

CO2 storage and release in the deep Southern Ocean on millennial to centennial timescales, Rae et al., Nature, 562, 569-573, 2018.
St Andrews boffins discover what triggered the end of the Ice Age (Courier).
Antarctic Ocean carbon dioxide helped end the Ice Age (

This entry was posted in Climate change, Global warming, Research, Science and tagged , , , , , , . Bookmark the permalink.

38 Responses to CO2 sequestration during glacial maxima

  1. BBD says:

    Summer sea ice extent in the Southern Ocean as a modulator of deep ocean ventilation (Ferrari et al. 2014) has been on the up as a hypothesis for a while. It does look persuasive, and the supporting evidence is clearly still coming in.

  2. BBD,
    Thanks, I hadn’t seen that paper. Something I’m still not sure about is whether or not there is still some CO2 being sequestered in some parts of the deep Southern Ocean that could be released as we warm further, or if it is pretty much all released as we move from the glacial to the inter-glacial.

  3. BBD says:

    I think it’s a work in progress, specifically being investigated by the DIMES project. Not aware of any key finding as yet, though.

  4. BBD,
    Thanks again. Still lots to learn, as usual 🙂

  5. BBD says:

    You should try being me 🙂 I’d love to stand on the shoulders of giants but I first need to surmount the little toe of the left foot… 😉

  6. Mitch says:

    People have been proposing a Southern Ocean link to carbon exchange between the atmosphere and oceans for at least 20 years. To advance the problem there needs to be data, which the Rae et al paper adds to nicely.

    As for increasing the CO2 exchange–CO2 flux into the ocean occurs near the polar front, about 50 deg S, while CO2 fluxes out of the ocean near 65 deg S. It is the dynamic balance between these processes that determines the net.

  7. Nick Stokes says:

    “As I mentioned at the beginning of this post, this is also the opposite of what we’d have expected if this region had been exchanging CO2 with the atmosphere during the LGM (at least, I think this is right).”
    It doesn’t seem right to me. The air got colder (at LGM); more CO2 dissolved in the ocean high and low, DIC increased, and pH decreased (pH scale in last fig is upside down). Air then got warmer, CO2 went to the air, DIC decreased, and pH increased.

    “a large amount of CO2 was indeed probably sequestered in the deep Southern Ocean during the glacial periods”
    And still is. What is notable about Fig 1 is not that DIC rises with ppm CO2, but how little it changes proportionately. And that is expected. DIC represents about 30000 Gtons of dissolved C, and the fluctuation of 100 ppmv in air represents about 200 Gtons C. So the whole ocean is barely perturbed by these changes, including our FF burning, provided there is enough mixing. It all just depends on time scale. I haven’t seen this well quantified in discussion of the paper – just mention of centennial to millennial, which is hand wavy. But I’ll read the paper when I get to the lab, where there is a subscription.

  8. Nick,

    The air got colder (at LGM); more CO2 dissolved in the ocean high and low, DIC increased, and pH decreased (pH scale in last fig is upside down). Air then got warmer, CO2 went to the air, DIC decreased, and pH increased.

    I was worried that I’ve got this wrong, so you may have a point. However, if you can simply explain the changes as due to temperature changes alone, why is there a need to find where CO2 is sequestered during the glacial maxima? I’ll have to think about this a bit more.

    So the whole ocean is barely perturbed by these changes, including our FF burning, provided there is enough mixing. It all just depends on time scale.

    Except given the Revelle factor, there will be a residual enhancement in atmospheric CO2, and corresponding changes in ocean pH, that will persist for millenia.

  9. Nick Stokes says:

    ” why is there a need to find where CO2 is sequestered during the glacial maxima?”
    Is there a need? CO2 is sequestered in cold water, wherever it can get to. It happens every winter in upper levels. AFAICS, all that is new in the paper is information about the rates or transport. Or could be, but I haven’t seen much described.

    An interesting question is whether deep water can actually get much colder during a glacial. It isn’t that much above freezing now.

  10. I seriously do not understand the significance of naming the term “Revelle factor”. This to me should be classified as ordinary physical chemistry that was known long before climate scientists were around, and would be called a dissociation constant in any college chemistry curriculum. More than likely that generations of Coca-Cola bottlers have never heard of the Revelle factor.

    Maybe I am missing something that has to do with the particular sea-water environment. But then again this may fall under the category of terminology hijacking that always occurs across disciplines.

  11. Nick,
    I just ran my carbonate chemistry code again, but with fixed DIC. If you consider how pCO2 changes with temperature, then you do get a reduction as temperature drops, but it’s not quite as high as 100ppm for a 5oC change. Also, the pH should increase as pCO2 gets lower, rather than decrease.

  12. Paul,
    As I understand it, it was originally thought that the oceans could simply take up all our emissions on a relatively short timescale. Roger Revelle showed that this was not the case because of this property of carbonate chemistry that the fractional change of DIC in seawater would be about 10 times bigger than the fractional change in atmospheric CO2. This is essentially where the residual in the Bern model comes from.

  13. Rob Painting says:

    The idea of the Southern Ocean storing and releasing carbon during glacial/interglacials and therefore facilitating most of the change in atmospheric CO2 has been around for a while. As has the idea of a small, but not negligible, contribution by increased coral reef calcification with rising sea levels.
    In a global context, the pH of the upper ocean will increase as pCO2 falls and decrease as pCO2 rises. That these deep sea coral suggest the Southern Ocean pH was rising as atmospheric CO2 was increasing, and global surface ocean pH was decreasing, does give some credence to the hypothesis.
    And ATTP, thankfully the ocean doesn’t globally see 5 degree changes in temperature during Milankovitch Cycles!

  14. Everett F Sargent says:

    Very interesting stuff! Now, if someone could explain this stuff to an old hobo like me, as I didn’t do any better than average in chemistry class.

  15. Ed Davies says:

    Nick Stokes says: “An interesting question is whether deep water can actually get much colder during a glacial.”

    Isn’t the deep ocean around 4 °C because that’s the densest it can get? I.e., if it cooled any further it would become less dense so convection would set in; just that it’d be the opposite of most convection in that it would be the colder water which would rise. So we wouldn’t expect it to get any colder.

    I might be putting 2 and 2 together and making π but my understanding is that as the world cooled from the PETM the oceans cooled in step until the deep ocean reached 4 °C at which point further cooling had to happen elsewhere so Antarctica started to ice over.

  16. Everett F Sargent says:

    Sea water has no maxima rho at ~4C, my own half baked idea is that fresh water (ice melt) is subducted at ~4C (i. e. mixed with salt water when the fresh water is at it’s maximum rho) to become the abysmal deep ocean temperature.

    I’m more curious in what would happen once all the ice melts, which hasn’t happened for a very long time (i.e. tens of millions of years). We are in an ice age which is currently in an interglacial.

  17. Everett F Sargent says:


    Fresh water has a bulk modulus of ~300 ksi (units of pressure or stress), so I’d assume similar for sea water, but see …
    Physical Properties of Seawater

    Click to access TALLEY_9780750645522_chapter3.pdf

    or TEOS-10 …

  18. Ed Davies says:

    Everett F Sargent says: “Sea water has no maxima rho at ~4C…”.

    Interesting. Thanks.

  19. BBD says:

    There’s a good (and short) article here about the relationship between glacial cycles and the carbon cycle. It’s a little out of date (2010) but makes the point that ocean temperature alone plays only a minor part in the process, which is much more complex.

  20. Mitch says:

    There seems to be a little confusion about deep ocean circulation. The CO2 is not stored in the southern ocean during glacials. The southern ocean is just the valve to the atmosphere. Probably most of the excess DIC is stored in the abyssal Atlantic basin where there is increased carbonate dissolution from the sea floor during glacials.

    Abyssal ocean temperatures are around 1-2 deg C, and are thought to have cooled by 1-2 deg C during the glacial periods.

  21. Rob Painting says:

    Most storage in the Southern Ocean would explain the slow draw-down of atmospheric CO2 heading toward a glacial maximum (as the CO2 accumulates there) and a much faster rise heading out of one (if the bulk of the CO2 is in the Southern Ocean it can quickly reach the atmosphere when the sea ice retreats and vigorous wind-driven mixing returns). As you point out though, the entire abyssal ocean which is part of the wind-driven/thermohaline circulation would have seen similar trends in pH.

  22. Thanks for all the comments. There have been a number of really interesting comments. It’s been a long day, so I’ll try to respond to them, when I get a chance, tomorrow.

  23. David B. Benson says:

    Milankovitch cycles run into difficulty in explaining MIS, Marine Isotope Stage, 5, 11, and 13. For MIS 5 it appears that the effect precedes the cause.

    I have yet to see a satisfying explanation of the past about 800,000 years. Southern sea ice might be a part of the answer.

  24. Rob P.,
    Thanks, and – yes – I wasn’t suggesting that the whole ocean cooled by 5K.

    As far as seawater goes, the density increases as temperature decreases for all temperatures above freezing. Here is a good comment from Ray Pierrehumbert about what sets the temperature of the deep ocean.

    Thanks, so that article says that temperature changes alone can alone account for about a third of the change in atmospheric CO2.

    Thanks, I was probably being a lax there.

  25. Chubbs says:

    I’m guessing that biology is playing a role in sequestration, i.e., the biological carbon pump of sinking and decomposing organic material. With less mixing and a steady pump, the balance would tip toward deep ocean accumulation.

  26. Ed Davies says:

    ATTP: thanks, so I got it the wrong way round; ice forming at the poles limits the drop in deep ocean temperatures rather than the end of the drop in the deep ocean temperatures causes the ice to form.

  27. Ed,
    All I was really referring to is the fact that the density water is at 4oC only applies to freshwater. For the typical salinity of seawater, the density increases as temperature decreases for all temperatures above freezing. Hence, as I understand it, the temperature in the deep ocean is set by the temperature of the coldest water (typically at the poles) that then sinks.

  28. The strongest structural instability in climate is probably due to density stratification (due to temperature and salinity differences) and the reduced gravity differential this leads to. This results in widespread effects such the seasonal mixing patterns (turnover) and thermocline sloshing.

    The Milankovitch cycles are possibly prone to this same structural instability as slightly different models of the earth’s orbital parameters makes a difference according to here:

  29. paulski0 says:

    why is there a need to find where CO2 is sequestered during the glacial maxima?

    I think it’s about magnitude. AR5 Chapter 6 sez:

    All of the major drivers of the glacial-to-interglacial atmospheric CO2 changes (Figure 6.5) are likely to have already been identified. However, Earth System Models have been unable to reproduce the full magnitude of the glacial-to-interglacial CO2 changes.

    Figure 6.5

    The sum of the median of all terms on that figure comes to about 20ppm compared to the 100ppm change, although the text does state that some of the processes are not independent so that big discrepancy is not necessarily hugely meaningful. The sea ice effect would come under ocean circulation in the figure, for which there is a wide spread. On sea ice AR5 also sez:

    A long-standing hypothesis is that increased LGM sea ice cover acted as a barrier to air–sea gas exchange and hence reduced the ‘leakage’ of CO2 during winter months from the ocean to the atmosphere during glacial periods (Broecker and Peng, 1986). However, concurrent changes in ocean circulation and biological productivity complicate the estimation of the impact of increased sea ice extent on LGM atmospheric CO2 (Kurahashi-Nakamura et al., 2007). With the exception of the results of an idealised box model (Stephens and Keeling, 2000), ocean carbon models are relatively consistent in projecting a small effect of higher sea ice extent on maintaining atmospheric CO2 lower during LGM (Archer et al., 2003).

  30. Chubbs says:


    Thanks. Would be interesting to see same for Pliocene–>Pleistocene since that is what we are reversing in the upcoming decades. Would guess that land processes and permafrost/arctic biome in particular become more important.

  31. Dave_Geologist says:

    For MIS 5 it appears that the effect precedes the cause.

    What aspect David? Orbital vs. temperature, orbital vs. CO2, CO2 vs. temperature?
    Note that there are correlation issues when you’re getting down to fine detail (closure to diffusion of ice or in radiometric or trace-element methods, land-sea correlation between different proxies, blurring, time-lags and palaeogeographic factors in marine records).

    For example, a transition defining the end of a substage or stage may progress in time from a low-latitude marine sedimentary record to mid-latitude stalagmite records (each with different time lags imposed by different rates of groundwater movement) to a high-latitude pollen record or an alpine ice-core record. Even within one type of data, benthic foraminiferal δ18O records, Skinner and Shackleton (2005) found a 4-kyr Atlantic lead over the Pacific for the last deglaciation, caused by a local or basin-restricted component of this signal.

    Even Cesare Emiliani himself referred to the isotopically-defined intervals as “climatic stages” (Gartner and Emiliani, 1976). His usage is a reminder that all numbered Quaternary stages, whether identified in marine isotopic records, spelean isotopic records, pollen records, etc., have potentially diachronous transitions controlled by individual responses to changing climate, and they are therefore “climatic stages” in the climatostratigraphic paradigm.

  32. BBD says:

    You’ve been reading that book again, haven’t you?

  33. Dave_Geologist says:

    The quotes above were from An optimized scheme of lettered marine isotope substages for the last 1.0 million years, and the climatostratigraphic nature of isotope stages and substages, here or a non-paywalled ms. here.

  34. BBD says:

    Thanks for the reference Dave G. I often wish it were easier to convey gentle humour via text. Perhaps I should have 🙂 but even that runs the risk of appearing snarky, sometimes.

  35. BBD says:

    @ Dave – I’ve just bought a new hardback copy of Summerhayes from AbeBooks for a little over £60. Good tip, thanks.

  36. Dave_Geologist says:

    Hope you find it useful BBD. No offence taken. Could have saved money on Kindle – although there are quite a lot of diagrams with lengthy captions which you need to read on a large screen, which kinda defeats the object. And of course some people prefer proper books.

    I did try to embed the link in the OP – but lack of a preview here causes errors. Actually that’s wrong, it’s always user error 😉 . I must discipline myself to putting the title of the paper into comments, especially when I link to an author’s copy whose link is probably ephemeral. Even some of the publishers’ permalinks aren’t always permanent, but with the title you can always search on Google Scholar.

  37. BBD says:

    Well, at least it wasn’t the joke money they were asking at Amazon 😦

    I do prefer books for book-length texts – articles and papers are fine on a screen, but books… it’s probably illogical but there it is. Won’t use Twitter, either. Harumph.

    And it will be an improvement on a Christmas-themed knitted jumper and a set of nylon socks, each marked with the day of the week 🙂 Or possibly a block of coal, wrapped in a page from the Daily Mail 🙂

  38. Dave_Geologist says:

    Incidentally, the issues discussed by Railsback are one reason I can’t get excited about exactly which date we choose for the start of the Anthropocene. Not only is the geological record compressed, with a metre representing thousands or even millions of years, but it is inevitably blurred by diffusion and mixing, and asynchronous within and between oceans because of time lags as currents mix atmospheric changes into deep water and they’re incorporated into sediments. Hundreds of years is often as good as it gets, even for relatively recent events.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.