Climate science

Predictable ice ages on a chaotic planet

Statistical analysis has revealed a simple rule for the occurrence of warm periods during the Quaternary, whereas on much longer timescales geological data have confirmed that the Solar System is chaotic. See Article p.427 & Letter p.468

Ice ages are paced by astronomical changes. However, a persistent difficulty has been to explain the dominant 100,000-year climate cycle observed1 during the past million years. On page 427, Tzedakis et al.2 report a model that links variations in Earth's orbit to the observed sequence of interglacials. On page 468, Ma et al.3 demonstrate that astronomical imprints on Earth's past climate can be used to probe the long-term evolution of the Solar System.

In 1941, the Serbian physicist Milutin Milanković proposed4 that small changes in Earth's orbit and rotational axis alter the distribution of solar energy on the planet's surface, affecting its climate — an effect known as astronomical forcing. Milanković's theory predicts that ice sheets in the Northern Hemisphere should grow and melt every 41,000 years, following changes in Earth's obliquity, which is defined as the angle between the planet's rotational axis and orbital plane. But in 1976, observations revealed1 a dominant 100,000-year climate cycle during the later part of the Quaternary — the period that started about 2.6 million years ago — that seems to be driven by variations in the eccentricity of Earth's orbit.

Many different ideas have been put forward over the past four decades5 to explain the apparent link between Earth's orbital eccentricity and climate during the Quaternary. The concept of stochastic resonance was invented6 to explain how small external oscillations (such as those associated with orbital eccentricity) could be amplified by a noisy process (such as climate) to eventually become the dominant periodicity of a system. It was even suggested7 that there might be no link between eccentricity and ice ages at all — instead, the observations could be explained by statistical fluctuations around multiples of obliquity cycles.

There have been only about ten ice-age cycles over the past million years. Each of these varied in duration, severity or structure, and the astronomical forcing was never quite the same. Consequently, the relative role of simple deterministic processes versus more unpredictable, or even random, fluctuations is difficult to ascertain.

Although it is generally assumed that large ice sheets vary slowly, filtering out any short-term climate fluctuations, there is evidence for abrupt changes not only in climate or ocean circulation, but also in the ice sheets themselves. Furthermore, during the most recent deglaciation, atmospheric CO2 levels and global temperatures increased a few millennia before the melting of the northern ice sheets. This effect could be linked to abrupt ocean-circulation changes, suggesting that such abrupt variability had a role in the deglaciation process8. Therefore, although astronomical forcing played a crucial part in the evolution of the Quaternary ice ages, this is not the whole story. Without a full mechanistic understanding of ice-sheet–climate interactions, it might be difficult to determine whether or not ice ages are entirely predictable9.

Tzedakis and colleagues have made progress in this direction by discovering a rule that links astronomical forcing to the timing of the Quaternary interglacials. In the earlier part of the Quaternary, interglacials occurred when the amount of summer solar radiation was above a given threshold; in the later part, this threshold was a linearly decreasing function of the time elapsed since the previous interglacial. The authors' idea is not completely new10, but their demonstration is enlightening because it focuses on the key difficulty — the occurrence of interglacials. Their statistical analysis shows that the rule is robust, and that the observed sequence of interglacials is the most probable one, among a small set of possible histories. This is a strong indication that ice ages are indeed predictable, and that the underlying mechanisms behind the 100,000-year climate cycle are probably quite simple.

The time elapsed between two interglacials — the ice-age duration — is linked to the size of the ice sheets. Therefore, Tzedakis and colleagues' rule suggests that, above some critical size, the larger the ice sheet, the more easily it can be melted5. So far, at least three mechanisms have been proposed to explain this apparent paradox (Fig. 1).

Figure 1: Melting an ice sheet.

Tzedakis et al.2 report a simple rule to explain how changes in Earth's orbit controlled the timing of interglacials during the past million years, whereas Ma et al.3 confirm that these orbital mechanics were chaotic before 50 million years ago. Tzedakis and colleagues' rule suggests that, the larger the ice sheet, the more easily it can be melted. Shown here are three possible explanations for this finding. First, a larger ice sheet might develop a warmer base than a smaller one, inducing a faster rate of ice flow (blue arrow), and potentially causing instabilities and enhanced iceberg calving11. Second, the surface of a larger ice sheet could experience a colder and drier climate, accumulating less snow, but more dust. Less sunlight would be reflected (pale yellow arrows), resulting in increased ice-sheet melting12. Finally, the Antarctic Ice Sheet at its maximum extent could alter bottom-water formation, break down ocean stratification — the vertical gradient of water density — and release CO2 into the atmosphere (red arrows), favouring ice-sheet retreat5.

The first explanation involves instabilities in large ice sheets: for instance, the bases of such ice sheets might become warm, inducing faster ice flow and increased iceberg calving11. A second possibility is associated with ice-sheet albedo (reflectance) in a cold and dry glacial climate: for large ice sheets, less snow and more dust increases the surface absorption of solar heat and enhances ice-sheet melting12. Finally, the Antarctic Ice Sheet at its maximum extent would affect bottom-water formation and ocean carbon storage5, which might be directly linked to the rise in atmospheric CO2 that precedes deglaciations8. It is also possible that all three of these mechanisms act together to ultimately lead to the simple dynamics outlined in Tzedakis and colleagues' paper.

Unravelling connections between astronomical changes and Earth's climate on much longer timescales has far-reaching implications for our knowledge of the Solar System itself. Indeed, it has been shown13 from celestial mechanics that our Solar System is chaotic, which prevents the precise computation of planetary motions before about 50 million years ago. However, geological data could help us to extend this horizon. Ma and colleagues have used sedimentological data to show that, between 85 million and 87 million years ago, a switch occurred from 1.2-million-year to 2.4-million-year climate cycles. The authors find that this transition was caused by a shift in the secular resonance (the synchronization of the precession) between the orbits of Earth and Mars.

Such a constraint on the dynamics of the Solar System could pave the way for an improved chronology of Earth's geological past. It is somewhat ironic that climate and sedimentological processes seem to respond to astronomical forcing in a rather simple way, but that the planets are not so predictable after all.Footnote 1


  1. 1.

    See all news & views


  1. 1

    Hays, J., Imbrie, J. & Shackleton, N. Science 194, 1121–1132 (1976).

    ADS  CAS  Article  Google Scholar 

  2. 2

    Tzedakis, P. C, Crucifix, M., Mitsui, T. & Wolff, E. W. Nature 542, 427–432 (2017).

    ADS  CAS  Article  Google Scholar 

  3. 3

    Ma, C., Meyers, S. R. & Sageman, B. B. Nature 542, 468–470 (2017).

    ADS  CAS  Article  Google Scholar 

  4. 4

    Milanković, M. M. Canon of Insolation and the Ice-Age Problem (R. Serb. Acad., 1941).

    Google Scholar 

  5. 5

    Paillard, D. Quat. Sci. Rev. 107, 11–24 (2015).

    ADS  Article  Google Scholar 

  6. 6

    Benzi, R., Parisi, G., Sutera, A. & Vulpiani A. Tellus 34, 10–16 (1982).

    ADS  Article  Google Scholar 

  7. 7

    Huybers, P. & Wunsch, C. Nature 434, 491–494 (2005).

    ADS  CAS  Article  Google Scholar 

  8. 8

    Shakun, J. D et al. Nature 484, 49–54 (2012).

    ADS  CAS  Article  Google Scholar 

  9. 9

    Crucifix, M. Clim. Past 9, 2253–2267 (2013).

    Article  Google Scholar 

  10. 10

    Paillard, D. Nature 391, 378–381 (1998).

    ADS  Article  Google Scholar 

  11. 11

    Oerlemans, J. Tellus A 35, 81–87 (1983).

    ADS  Article  Google Scholar 

  12. 12

    Ganopolski, A. & Calov, R. Clim. Past 7, 1415–1425 (2011).

    Article  Google Scholar 

  13. 13

    Laskar, J. Nature 338, 237–238 (1989).

    ADS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Didier Paillard.

Related links

Related links

Related links in Nature Research

In retrospect: Forty years of linking orbits to ice ages

Climate science: Earth's narrow escape from a big freeze

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Paillard, D. Predictable ice ages on a chaotic planet. Nature 542, 419–420 (2017).

Download citation


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.