RealClimate logo

First successful model simulation of the past 3 million years of climate change

Filed under: — stefan @ 3 April 2019

Guest post by Matteo Willeit, Potsdam Institute for Climate Impact Research

A new study published in Science Advances shows that the main features of natural climate variability over the last 3 million years can be reproduced with an efficient model of the Earth system.

The Quaternary is the most recent geological Period, covering the past ~2.6 million years. It is defined by the presence of glacial-interglacial cycles associated with the cyclic growth and decay of continental ice sheets in the Northern Hemisphere. Climate variations during the Quaternary are best seen in oxygen isotopes measured in deep-sea sediment cores, which represent variations in global ice volume and ocean temperature. These data show clearly that there has been a general trend towards larger ice sheets and cooler temperatures over the last 3 million years, accompanied by an increase in the amplitude of glacial-interglacial variations and a transition from mostly symmetry cycles with a periodicity of 40,000 years to strongly asymmetric 100,000-year cycles at around 1 million years ago.  However, the ultimate causes of these transitions in glacial cycle dynamics remain debated.

Among others, the role of CO2 changes in shaping Quaternary climate dynamics is not yet fully understood, largely because of the poor observational constraints on atmospheric CO2 concentrations for the time before 800,000 years BP, beyond the period covered by high-quality ice core data.

In a paper published today in Science Advances (Williet et al., 2019), we were able to reproduce the natural climate variability of the whole Quaternary with an Earth system model of intermediate complexity. Besides ocean and atmosphere, the model includes interactive ice sheets for the Northern Hemisphere and a fully coupled global carbon cycle and was driven only by changes in orbital configuration and different scenarios for slowly varying boundary conditions, namely CO2 outgassing from volcanoes as a geologic source of CO2, and changes in sediment distribution over the continents.

The model simulations provide a self-consistent reconstruction of CO2, climate and ice sheets constrained by available observations, i.e. oxygen isotopes and reconstructions of sea surface temperature. The fact that the model can reproduce the main features of the observed climate history gives us confidence in our general understanding of how the climate system works and provides some constraints on the contribution of external forcings and internal feedbacks to climate variability.

Our results imply a strong sensitivity of the Earth system to relatively small variations in atmospheric CO2. A gradual decrease of CO2 to values below ~350 ppm led to the start of continental ice sheet growth over Greenland and more generally over the NH at the end of the Pliocene, beginning of Pleistocene. Subsequently, the waxing and waning of the ice sheets acted to gradually remove the thick layer of unconsolidated sediments that had been formed previously over continents by the undisturbed action of weathering over millions of years. The erosion of this sediment layer – it was essentially bulldozed away by moving glaciers – affected the evolution of glacial cycles in several ways. First, ice sheets sitting on soft sediments are generally more mobile than ice sheets grounded on hard bedrock, because ice slides more easily over sediments compared to bedrock. Additionally, glacial sediment transport to the ice sheet margins generates substantial amounts of dust that, once deposited on the ice sheet surface, increases melting of the ice sheets by lowering surface albedo. Our results show that the gradual increase in the area of exposed bedrock over time led to more stable ice sheets which were less responsive to orbital forcing and ultimately paved the way for the transition to 100,000 years cycles at around 1 million years ago.

The simulations further suggest that global temperature never exceeded the preindustrial value by more than 2°C during the Quaternary. Ice sheet evolution is very sensitive to temperature, and the initiation of NH glaciation at around 3 million years ago would not have been possible in the model if global temperature would have been higher than 2°C relative to preindustrial during the early Quaternary. Since the model has been shown to accurately reproduce the sea level variations over the last 400,000 years and also the spatial ice sheet distribution at the last glacial maximum (Ganopolski & Brovkin 2017), we are confident that the sensitivity of ice sheets to climate is well represented in the model.

Likewise, our results indicate that the current CO2 concentration of ~410 ppm is unprecedented over the past 3 million years. The climate sensitivity of the model is around 3°C global warming for a doubling of CO2 concentration, which is at the center of the range of current best estimates of climate sensitivity that range between 1.5 and 4.5°C. It is possible that the real climate sensitivity is lower than 3°C, in which case the modelled CO2 concentration needed to fit the oxygen isotope record during the early Quaternary would be higher than in the present model simulations, but it would still be unlikely to exceed the present day value. In the context of future climate change, our results imply that a failure to significantly reduce CO2 emissions to comply with the Paris Agreement target of limiting global warming well below 2°C will not only bring Earth’s climate away from Holocene-like conditions, but also push it beyond climatic conditions experienced during the entire current geological period.

Postscript (11 April): Breitbart has published a mind-blowing misrepresentation of the Willeit et al. study, in an article titled “Scientists Prove Man-Made Global Warming is a Hoax”. It’s that old non-sequitur “if climate has changed naturally in the past, humans can’t be causing the ongoing global warming.” RealClimate has asked Matteo Willeit to comment. Here is what he wrote:

Human CO2 emissions have already so massively disturbed the natural climate cycles that within the last hundred years they have more than undone the previous 5000 years of natural cooling due to the precession cycle, and global temperature is now already higher than at any time in the Holocene and thus the entire history of human civilization.

The same model that, in the present study, has been shown to be able to simulate the past 3 million years of natural climate variability also predicts that our CO2 emissions since the industrial revolution were so large that they will disturb the natural ice age cycles for tens of thousands of years to come because of the very long legacy of our fossil CO2 in the atmosphere (Ganopolski et al., 2016).


  1. M. Willeit, A. Ganopolski, R. Calov, and V. Brovkin, "Mid-Pleistocene transition in glacial cycles explained by declining CO2and regolith removal", Science Advances, vol. 5, pp. eaav7337, 2019.
  2. A. Ganopolski, and V. Brovkin, "Simulation of climate, ice sheets and CO<sub>2</sub> evolution during the last four glacial cycles with an Earth system model of intermediate complexity", Climate of the Past, vol. 13, pp. 1695-1716, 2017.

56 Responses to “First successful model simulation of the past 3 million years of climate change”

  1. 51
    Simon C says:

    BPL #44 and B Fagan #50 Thanks – yes, it has to be a very long-term trend so a protracted tectonic process – either volcanic activity changes or uplift causing erosion – could account for it. But perhaps also a gradual biological evolutionary refinement of techniques for grabbing carbon out of the ocean/atmosphere/biosphere. If so, a perturbation to the biological system caused by (eg) an anthropogenic mass extinction could reduce the drawdown, effectively increasing the atmospheric CO2 concentration. An analogy for the drawdown would be the Carboniferous ice ages, which may have been caused or enabled by carbon storage by the abundant forests of that time period.

  2. 52
    patrick027 says:

    Re 51 Simon C – I learned from watching Cosmos (the 2014 series) that when plants first developed lignin, there was a period of time before decomposers evolved to eat it, thus lignin contributed to the coal build-up.

  3. 53
    Dr. rer. nat. Florian Diehl says:

    Hello all,
    I am no climate expert, and have not been working as a scientist for almost 7 years now, but I am interested and active in the FFF (fridays for future) movement and thus also searching for arguments and current debates on the topic.

    Anyway, there are a lot of folks out there that believe CO2 is not the main driver for climate change. In googling the current state of research on this, I stumbled upon this paper by W. Davis, 2017:

    He argues, that CO2 and Temperature correlate very little at a time-scale of 450 million years and that the earth tilt-shifts and ellypsoid rotation around the sun have a far greater impact.
    Is there data suggesting otherwise? Did Davis just look at it at a far too large time scale?
    What does make sense however is, that the impact of CO2 decreases the more CO2 is already in the atmosphere…which also means, that any emission reductions applied by us humans will have a far too little impact in the beginning…

    Anyway, thanks so much for your work and thanks for reading my comment this far.

    Sincerely and all the best,
    F. Diehl

  4. 54
    Ray Ladbury says:

    If you want climate science, go to climate journals. Davis is just another idiot that can’t comprehend that there could be multiple drivers to climate. More proof that you can send stupid to college.

  5. 55

    FD 53,

    The impact of CO2 in terms of radiative forcing is a logarithmic relation, so in that sense “the impact of CO2 decreases the more CO2 is already in the atmosphere.” On the other hand, the increase in CO2 every year is exponential, so the two effects cancel out. We can achieve a lot by ceasing to emit CO2 as soon as possible.

  6. 56

    #53, Dr. Diehl–

    Yes, I think the bottom line take is indeed that timelines of hundreds of millions of years are not very helpful in the current context. (We talked about this paper a bit on RC previously, I think, but I don’t recall which thread it was–probably Unforced Variations.)

    One thing is certain, though, which is that the logarithmic nature of the CO2 forcing has been recognized in the literature at least since the days of Svante Arrhenius. So Davis is not presenting anything new in that regard.

    In fact it was implicated in a very interesting observation made here a number of years ago by Ray Pierrehumert*, who remarked that the key to the greater short-term warming potential of methane versus CO2 was precisely that the former exists in the atmosphere in much lower concnetrations.