Does Gaia end Ice Ages?

Orbital changes to solar radiation inside the Arctic circle alone are insufficient to end all ice ages.  A new paper by Ralf Ellis proposes an elegant solution to the mystery of how ice ages terminate. Life itself could provide the ultimate thermostat control.

Variations in insolation at high latitudes occur in tune with the 23Ky precession cycle, but their strength is strongly modulated by the 41Ky obliquity cycle and the 100Ky eccentricity cycle. Glaciations used to follow the 41Ky obliquity cycle, but for the last 800Ky the maximum summer insolation in the arctic has not been enough to fully terminate an ice age. Ice sheets in the Northern hemisphere have  grown so large that the ice-albedo feedback suppresses solar melting alone. Something else is now  needed to break the ice albedo feedback effect. The basic proposal is as follows.

As ice sheets expand and ocean temperatures fall so CO2 levels will eventually fall dangerously below 200ppm, threatening the survival of plants based on C3 photosynthesis. Boreal forests and  temperate grasslands will begin to die back exposing soil to weathering. Desertification ensues and strong winds transport dust storms over the ice sheets. With very little annual snowfall during a glacial maximum, this dust layer will build up and thereby reduce net ice albedo. The next  ‘Grand Summer’  precession cycle at maximum eccentricity will now be able to  rapidly melt back the ice sheets, because finally a lower albedo ensures that more heat is absorbed by the ice sheets.  Does  the data actually support this hypothesis ?

Figure 1: The blue curve is a calculation of the maximum insolation in summer at the north pole. The ice volume data is the benthic d18O stack. In red is the Epica temperature anomaly data from Antarctica. In yellow is the Epica CO2 data and in purple their dust data. The black curve is the change in the Earth's orbital eccentricity. The temperature, dust, and ice volume data have all been scaled for comparison purposes.

Figure 1: The blue curve is a calculation of the maximum insolation in summer at the north pole. The ice volume data is the benthic d18O stack. In red is the Epica temperature anomaly data from Antarctica. In yellow is the Epica CO2 data and in purple their dust data. The black curve is the change in the Earth’s orbital eccentricity. The temperature, dust, and ice volume data have all been scaled for comparison purposes. Click to expand

The last glacial cycle is particularly interesting because the Great Summer maxima are suppressed by low values of orbital eccentricity. This results in the classic sawtooth shape showing how two insolation maxima had very little effect on ice volume growth. However, when CO2 levels fall below ~220 ppm the deposition of dust increases dramatically. The onset of the next summer maximum triggered the current interglacial. We are very lucky that human civilization developed during an era of low orbital eccentricity since the current summer minimum insolation in the arctic is very weak. Otherwise we would already be heading for the next ice age. The most similar interglacial occurred 400,000 years ago and lasted ~30,000 years (see figure 2). That too shows a large dust peak prior to rapid warming.

I think Ellis’s proposal is attractive and it is likely to be the correct explanation for the last glacial termination, as no other explanation so far makes sense to me. However the evidence is less strong for the previous glacial cycle. What is clear is that whenever CO2 levels fall below 200ppm there is a steep increase in dust deposition. The lack of moisture and low CO2 levels kills off boreal forests and soil turns to dust. The previous glacial cycle shows a long period of dust but without a final peak. However, the final Great Summer insolation is very strong, reaching an average daily insolation at the pole of 600W/m2. Note however that  a similarly large excursion during the last ice age had no effect because CO2 levels were still high without any dust deposition on the ice sheets.

Figure 2. The last 800,000 years of glacial cycles. Curves are the same as in figure 1. The current interglacial is most similar to that 400,000 years ago. Click to expand

Figure 2. The last 800,000 years of glacial cycles. Curves are the same as in figure 1. The current interglacial is most similar to that 400,000 years ago. Click to expand

Figure 2 above shows the last 800k years and demonstrates why it is eccentricity that drives the long-term glacial cycle. What is also evident is that there are always peaks of dust deposition whenever CO2 levels fall below 200ppm. However, the data implies that it is only when eccentricity is small that the dust-albedo effect becomes critical. Otherwise the boost to a Great Summer Insolation caused by large eccentricity is usually sufficient to terminate a glaciation. Note that the glacial ‘cycle’ from -630Ky to -480Ky essentially consisted of three mini-cycles, each broken by an enhanced Great Summer Insolation, because eccentricity remained continuously above current values.

The IPCC AR5 report is closer to a  ‘CO2 controls everything’ scenario, because they see the feedback of CO2 forcing feedback as the ‘extra’ mechanism for ending ice ages. However cooler oceans absorb more CO2 and rises in CO2 follow those first in temperature. There clearly must be some enhanced CO2 greenhouse feedback, but rises in humidity(H2O) with warming are probably far more important. Taken together rising CO2 and H2O levels  boost the regrowth of boreal forests thereby ending dust storms. Life rapidly recovers following a deep shock. Gaia lives on.

In conclusion Ralf Ellis has made a very interesting proposal which I believe could well be the correct explanation for the rapid recovery from the last glacial minimum (LGM), and also that of 400,000 years ago. During the LGM CO2 levels reached dangerously low levels of ~180 ppm causing arid desertification  as temperate trees and savannah died off. The resulting dust storms then deposited huge amounts of dust on the ice sheets increasing its albedo. The consequent Great Summer Insolation, coinciding with maximum eccentricity, finally melted back the ice sheets through reduced albedo, perhaps aided by increasing CO2 and H2O. However, the evidence is far less strong for previous glacial cycles. A fall in dust induced ice-albedo is less likely to be the primary cause of interglacial warming. This is because the orbital eccentricity was much larger and  was sufficient to melt back ice sheets for the  Great summer maxima.

About Clive Best

PhD High Energy Physics Worked at CERN, Rutherford Lab, JET, JRC, OSVision
This entry was posted in Climate Change, climate science, Gaia, Ice Ages, Science and tagged . Bookmark the permalink.

40 Responses to Does Gaia end Ice Ages?

  1. Jerry says:

    This is interesting but there is a question. If it only explains the end of one glacial cycle then how is it of general value when searching for components of the glacial cycles in general?

    • Clive Best says:

      I think it explains the end of all glaciations when the earth’s orbit reaches a minimum of the 400,000 year eccentricity cycle. So that includes the ice age 400,000 years ago. When all else fails dust from dying forests comes to the rescue.

  2. ralfellis says:

    Thanks for this article Clive. And I love your graphs – better than mine I think. I thought a multiple-plot graph would be messy, but these are nice and clear.

    They do highlight the fact that dust storns arise at eccentricity minima. I have thought about this a lot, and I think this is semi-coincidental. There is no orbital or astronomic link between eccentricity and dust. However, maximum temperatures and full interglacials always happen on the approach to an eccentricity maxima, because the eccentricity maxima produces the greatest precessional insolation — a strong Great Summer (and the ice is dusty).

    Following this sudden interglacial warming, the world tries to stay at this temperature, via the cloud thermostat system, but largely fails and slowly slips back into a glacial climate. So the the timing of the next interglacial is dictated by this slow fall in temperature, and the corresponding slow decreace in CO2 concentrations. And it just happens to take about 80 kyr for the temperature and CO2 to fall to critical levels.

    Once criticality has been achieved, the world’s flora dies, deserts are created, dust storms rage, the ice sheets become dirty, and the world is then primed and ready for another interglacial warming. Which can only happen when approaching the next eccentricity maxima. But if the decrease in temperature and CO2 had taken longer (and therefore the dust did not occur until later) I think the world would have simply remained in glacial conditions until the next eccentricity maxima.

    So I do not think the eccentricity minima are linked with dust maxima, in any direct fashion, except for the fact that both have the same starting-point. The starting-point is approaching the eccentricity maxima, because this creates a strong Great Summer and initiates interglacials, and it is this eccentricity-generated start-point that keeps the dust maxima closely correlated with eccentricity minima.

    Best regards,
    Ralph Ellis

    • Clive Best says:

      Hi Ralf,

      I think your paper is really important and solves much of the mystery of inter-glacial timing. The effects of orbital eccentricity is very subtle because the annual insolation for a highly elliptical orbit and a circular orbit is almost the same. As you note in your paper eccentricity modulates the 23ky precession of the equinoxes. If the orbit was purely circular then the precession would have no effect at all. The magnitude of a Great Summer depends on the eccentricity. The larger eccentricity the stronger the Great summer insolation at the poles. Changes in obliquity are independent of eccentricity. Every 41Ky the earth reaches a maximum inclination with respect to the solar plane. This increases the size of the arctic circle and the angle of incidence of solar radiation, independent of eccentricity. For 2 million years glaciations followed only the obliquity cycle as summer insolation waxed and waned. It was only when long term cooling increased ice sheets beyond a critical size that the precession term ( amplified by eccentricity) became important. I have spent many happy hours calculating this effect !

      The mystery has always been why glaciations appear to follow the eccentricity maxima. In reality they don’t but need help to overcome the negative ice-albedo effect. Sometimes it is enough to wait for an eccentricity amplified Great Summer. Otherwise deep glaciations end with die back dust reducing ice albedo and a reduced Great summer – just as you describe.

      The reason why I don’t think Dust always ends ice ages is for example the one 300,000y ago where the dust peaked before the large Great Summer caused melt back. Of course I may be wrong.

      • ralfellis says:

        >>The reason why I don’t think Dust always ends ice ages is
        >>for example the one 300,000y ago where the dust peaked
        >>before the large Great Summer caused melt back.

        I think this interglacial is a good example of how the dust is a latent warming agent, and why the eccentricity cycle is so important.

        The Great Summer 280 ykr ago was strong, but since there had been no dust storms it had no effect. Just 15 kyr later there was an era of strong dust storms, and the world was ready for an interglacial. But the next Great Summer was during a low eccentricity period and was weak, and was unable to warm the ice sheets.

        So the world waited patiently until the eccentricity cycle built up to a maximum. Only now were all the conditions in place. The ice sheets were already dusty, the next Great Summer was strong, and only now could the ice sheets melt.


  3. ralfellis says:

    But I do think this theory accounts for all interglacial warming periods, not just the low eccentricity ones.

    The same set of requirements appear to be present before and during each interglacial … low CO2 … flora die-back … increased dust … decreased ice albedo … increasing eccentricity … increased Great Summer insolation … and presto, an interglacial. It is a complex series of events and interconnections, but the same series appears to repeat itself every time.


  4. lapogus says:

    I have no competing hypothesis but am not convinced by this. We have Antarctic ice core data going back more than 800k now – is there any evidence for these layers of dust in the cores?

    • ralfellis says:

      >>We have Antarctic ice core data going back more than 800k now – is
      >>there any evidence for these layers of dust in the cores?

      The dust data is from the ice cores. And the dust in the Greenland ice cores in at least 4x greater. Please see the link to my Academia. edu paper, at the beginning of Clive’s post.


  5. ralfellis says:


    I rather like your second graph. Could I use it in the paper, in the Summary at the end? It does, after all, summarise the entire paper very nicely.

    Would you have a pdf version? (I am presuming it was created as a pdf).


    • Clive Best says:

      Yes of course. I’ll contact you by email.

      I was thinking about developing a quantitative model to answer some of the above questions. Will let you know if I manage to get something working.



  6. ralfellis says:

    The paper has now gone for peer review at the Royal Society. Would you be willing and able to act as a reviewer? My email address is in the paper. Don’t worry if you are too busy.

    Many thanks,

  7. Pingback: Week in review – science edition | Climate Etc.

  8. Pingback: Week in review – science edition |

  9. etudiant says:

    This analysis is predicated on the idea that dust makes ice more susceptible to melting, because it reduces the albedo of the surface. Against that is the reality that dirt is a pretty good insulator.
    It is not obvious that the insulation benefit would be overcome by the albedo reduction. Certainly the permafrosts in the arctic suggest the albedo reduction is insufficient to thaw out deep deposits.

    • ralph ellis says:

      If you read the paper, you will see that the albedo reductions of small deposits of dust are very large. And the insulation effects of dust do not become significant until there is more than 5mm of dust-dirt on the surface. And that figure was derived by Warren et al by experiment, not by a model. And as Warren observed, even with dissipation concentration, the levels of Arctic dust should never reach 5mm.


  10. Javier says:

    Nice theory, but I am not convinced. Some of the dust peaks take place thousands of years before the warming and the ice gets a new layer every year. Anything you leave at the poles sinks pretty fast, as happened with those planes that landed in Greenland during WWII.

    There are competing theories. My personal opinion is that the ice melting is helped by a powerful feedback mechanism, a combination of insolation and rising sea levels, so when you have very low sea levels all the low lying ice is melted by rising sea levels, very, very fast, creating a powerful feedback mechanism. The reducing sea levels at a Glacial Maximum act as a spring pulling for the next obliquity cycle to attempt an interglacial.

    • Clive Best says:

      Hi Javier,

      Remember that what is important is not really the dust as measured in the ice cores but what occurred at the edges of the ice sheets. Unfortunately we don’t have any measurements for that because they have all melted. However, I was wondering if there could be any geological evidence left of such large deposits across North America and Eurasia.

      I think we agree that ice melting must be helped by a powerful feedback mechanism. Even if sea levels get very low they have to start rising somehow for your mechanism to start. I also once proposed that extreme lunar tides could help kick start an interglacial. The tidal forces acting on the ice sheets could be enormous and their maxima coincide with high eccentricity.

      Does the Moon trigger interglacials?



  11. ralfellis says:

    >>Some of the dust peaks take place thousands of years before the
    >>warming and … anything you leave at the poles sinks pretty fast,
    >>as happened with those planes that landed in Greenland during WWII.

    Indeed, but that is why this albedo feedback system works so well.

    The D-O events are also probably dust-soot events, but since they only have a few years of dust the warming is quick but temporary. However, interglacials are preceded by 15 kyr of dust, which is distributed through the top 1/3 of the ice sheets. So the more the ice sheet melts, the more dust is concentrated on the surface. So the interglacial is not a 10-year wonder, it continues all the way through to interglacial completion.

    I have assumed that the ice sheet termini have 3x the dust levels found at 75ºN. Read the full article, in the link above.


    • Javier says:

      Completely disagree, Ralph,

      The latest evidence points to D-O events originating in the Nordic seas under the sea ice due to the mixing of warm subsurface waters brought North by oceanic currents with cold surface waters disrupting the layer structure that is kept in place by the halocline, allowing the escape of the accumulated heat. You can see for example Dokken T.M. et al. 2013. Dansgaard-Oeschger cycles: Interactions between ocean and sea ice intrinsic to the Nordic seas. Paleoceanography 28 491-502.

      • ralfellis says:

        But all D-O events are coincident with an increase in wildfires. The FPF at the bottom of the graph is the Fire Peak Frequency, and each FPF matches a D-O warming.

        D-O events are not the subject of my paper. However, if my thesis is correct then the achilles heel of an ice age world is a lowering of albedo via surface dust or soot deposits. With soot being the stronger of the two by a factor of 50. And so it has to be asked if these global conflagrations during D-O events were causal rather than consequential.

        We are looking for a forcing or feedback that can warm the NH by 10ºc in ten years. Do you really think a change in ocean currents is up to that task? In such a short time? When these ocean currents have a cycle time of 1,000 years or more? Or is it more likely that the warming was caused by seriously blackened ice sheets, and an increase in insolation absorption of 250 w/m2 or more?

        But unlike a full interglacial, these were thin layers of soot that were easily washed away, and so glacial conditions resumed quite quickly. And the result is a transient D-O warming event.


        • Javier says:

          I see no problem with the alternative explanation to the coincidence between D-O interstadials and wildfires, that the warming increases wildfires and not the other way around.

          You should read the Dokken et al. 2013 article, as it is your line of research. They explain that the accumulation of warm water below the halocline takes place over the time of the D-O stadial, that can go over a millenium. The vertical mixing of waters of different temperatures is a very powerful forcing, that can quickly change temperatures by several degrees, towards cooling or warming, depending on the conditions and the area involved. After all the El Niño and La Niña are based on water temperature differences.

          ¿What is more likely? That is a strange question. Things are or are not regardless of what we think about them.

          Your data shows that there is a coincidence between very cold, low CO2 periods and dust. We already knew that deserts expand on cold periods, so it is consistent. We also knew that sea productivity increases during glacial periods due to desert dust increase. But all the rest is opinion. Whether interglacials come from increased dust or not, you have shown no evidence, and some interglacials come after little dust increase and others come several thousand years after the dust has fallen from peak levels, and that is not very convincing.

          Perhaps you could strengthen your case if you could confirm that there is a relationship today between soot levels and polar ice melting. After all the Chinese are making a lot of soot.

          • ralfellis says:

            >>Perhaps you could strengthen your case if you could
            >>confirm that there is a relationship today between
            >>soot levels and polar ice melting.

            There has been a study about industrial soot ending the Little Ice Age ice sheets in the Alps, when the climatic factors did not suit such a sudden change. The soot level they give are only 0.04 ppm vs 8 ppm of dust during the ice age warming. But since soot is 50x more effective than dust, that places the modern contamination not far behind the interglacial equivalent. But there would be less surface concentration in the Alpine case, because there was only a few years of deposition.


            >>What is more likely?
            >>That is a strange question.

            Not at all. If there are two competing theories, which do you give more credence to?


  12. Why did Earth plunge into the Pleistocene out of the Eocene in the first place. Is dust from dying forests all that’s keeping us from perpetual major glaciation?

    • ralfellis says:

      One likely for the Mid Pleistocene Transition (MPT)** was given by Kuhle. He suggested that the Himalaya was uplifted by 500m 800 kyr ago, and so its ice cap grew much larger, which depressed NH temperatures and created more severe ice ages. This led to the large ice ages that followed the MPT. The Himalaya, Kuhle suggests, has a temperature influence four times greater than the Arctic, area for area, because of its southerly latitude.

      Kuhle. The Pleistocene Glaciation of Tibet and the Onset of Ice Ages

      I don’t know if this rise in the Himalaya has been proven, but the hypothesis appears sound. Again the mechanism is albedo controlled, rather than Co2 controlled, and the resulting larger Arctic ice cap could well have resisted the warming cycle of obliquity. And the only thing that could challenge this larger ice cap, is dust-albedo combined with precession – hence the severe modern ice age cycle.

      The same process of Himalayan uplift could also be invoked to explain all the steady cooling of the Pleistocene era. And CO2 decreased at the same time, merely because of reduced temperatures and oceanic absorption. So again the role of CO2 is consequential rather than causal, and the real modulator of temperature is albedo.


      ** Where obliquity-modulated 40kyr cycles changed to precession-modulated 100kyr cycles.

  13. The aspect that concerns me is that the drop in CO2 sufficient to produced the necessary dust clouds having killed off plant life would be a global phenomenon so that the tropics would die off too so that all global plant and animal life would be extinguished yet that doesn’t seem to have happened.
    Any ideas?

    • ralfellis says:

      Plant life will begin to die off at high altutude and in already arid locations, long before it will be extinguished in moist lowlands. (Due to lower partial pressures at altitude, and the effects of CO2 on plant respiration.).

      Thus the die-off of plant life is actually exponential. The greater part of the dust was formed from the Gobi region becoming a shifting-sand CO2 desert, and that would have happened long before the tropics became a CO2 desert.

  14. J Martin says:

    I do like that first graph but wish you’d re-do it replacing the blue line 65° composite insolation line with a line for obliquity on its own. That would be far more revealing I think.

    • Clive Best says:

      The red curve below shows the obliquity component.

    • ralfellis says:

      Not really. It seems abundantly clear that where possible, the ice age cycle responds to the approx 22 kyr precessional Great Year cycle, The blue 65 lat plot (the Great Year) is actually a combination of both precession and obliquity, but is influenced more by precession than obliquity because the former is the stronger of the two.

      The only time that obliquity becomes the equal of precession, is during periods of low eccentricity. It so happens that we are currently in a low exentricity era and so current temperatures appear to be closely following the obliquity cycle to the exclusion of precession. But that is a bit of a false conclusion.

      But whether you like obliquity or precession, you still end up with the missing cycle problem. Precession is approx 22 kyr and obliquity approx 41 kyr, but the ice age cycle is between 90 and 110 kyr. So why are some orbital cycles missed? The only theory that can answer that perennial problem, is the dust-albedo theory.


  15. So no die off in tropical regions?
    C02 is supposed to be well mixed so how much tropical vegetation and animal life could remain to resupply the rest of the globe at a later date?

    • ralfellis says:

      Resupply with what? There was plenty of life still populating the planet. It was only the upland and arid regions that died off – including all of northern China-Mongolia. But that was enough to provide the dust that ended the last ice age. And we know this because of the evidence from the deep dust deposits on the Loess Plateau.


  16. ralfellis says:

    Not sure if you know, but this article was finally published as a peer-reviewed scince paper. It can be seen here on Sience Direct:


    P.S. Where did you see this article, as it is rather old….

  17. How would low CO2 selectively cause die off only in the regions you mention?
    Wouldn’t CO2 levels be equally low everywhere ?

    • I think there is no need to involve global CO2 or plant die off to get the raised dust levels.
      I recall a description whereby the cold,windiness and dryness equatorward of the ice front creates deserts that produce the observed dust.
      The CO2 variations could just be the response of the ocean’s absorption capability when temperatures change. Correlating but not necessarily causing.

      • ralfellis says:

        If you read my paper, you will see that precipitation remained largely the same during the LGM, and parts of the Gobi (where the dust came from) were wetter than now. The only reason the Gobi became a shifting-sand desert and produced so much dust, is that it became a CO2 desert.


    • ralfellis says:

      Please read my comments before replying. There is the same concentration of oxygen on Mt Everest as at sea level. So why do you run short of oxygen and possibly die, if you climb to the top of Everest?


  18. Ok, I’ll read it more carefully to see whether one really needs to bring in the CO2 aspect when simple cold, dryness and windiness could theoretically do the trick.

  19. andrewR says:

    Fascinating website and some great insights. I’d class myself as a ‘Luke warmer’ concerned about long term consequences but not alarmist. As you intimate here and elsewhere on your website, Gaia seems to have a way of dealing with these things.

    Re the Ralph Ellis paper, the thought struck me that if snow fall over the glaciated regions drops off during an ice age as most of the water vapor is sucked out of the atmosphere at high latitudes, I would expect dust to accumulate naturally over time – maybe you don’t need global desertification and forest die-off as the paper proposes. It is my understanding that during the ice ages, the Sahara (perhaps Gobi as well?) turns into Savannah and the paleo record seems to support this.

    Again, fabulous website – thank you

Leave a Reply