Some 34 million years ago, there was a rapid growth of ice on Antarctica. A modelling study indicates that the ultimate cause of this glaciation was a decrease in the concentration of atmospheric carbon dioxide. See Letter p.574
On page 574 of this issue, Goldner et al.1 tackle a long-standing debate in palaeoclimate science: the causes and effects of the largest climate transition of the past 50 million years, which occurred about 34 million years ago. It was characterized by rapid cooling and growth of Antarctic ice, marking a change from the warm 'greenhouse' climates of the Eocene epoch to the 'icehouse' of the Oligocene epoch.
Using a numerical climate model, the authors investigated the effects of inserting a continental ice sheet onto Antarctica, and found that spatial patterns of ocean cooling predicted by the model agree well with cooling patterns inferred from the geological record of this time period. This cooling, and associated ocean-circulation changes, had previously been attributed to geographical changes in ocean straits and seaways, but Goldner and colleagues conclude that the cooling is better explained as a response to Antarctic ice growth, itself caused by a decrease in the concentration of atmospheric carbon dioxide.
Before the cooling at the Eocene–Oligocene transition (EOT), much of Antarctica was vegetated and was home to flora and fauna that today is found nearer the Equator, including flowering plants, beech forests and marsupials2. Two main hypotheses have emerged to explain the cooling and the growth of ice in this period, which occurred over about 300,000 years3,4. The first proposal, called the gateway hypothesis, posits that gradual movements of continental plates over millions of years gradually widened the Drake Passage and Tasman Gateway in the Southern Ocean, allowing increased ocean flow around Antarctica. This led to decreased poleward heat transport, which resulted in cooling of the Antarctic continent and growth of the Antarctic ice sheet. The second proposal, called the CO2 hypothesis, postulates that a decreased concentration of atmospheric greenhouse gases, in particular CO2, led directly to cooling of the Antarctic continent and growth of the ice sheet.
One line of evidence previously used in favour of the gateway hypothesis concerns the spatial pattern of EOT ocean-temperature change. This pattern is derived by drilling deep into the modern ocean floor and extracting cores of ancient ocean sediments. The cores are analysed for their chemical and isotopic composition, providing insights into changes in ocean temperature over time. By drilling cores at various locations and depths in the Atlantic, such analysis has revealed increasing cooling down to a depth of 2 kilometres, and increasing cooling from the Equator southwards. Until now, it had been suggested that this signature was best explained by the gateway hypothesis5, being similar to that predicted as a response to changes in ocean circulation associated with evolving ocean gateways6.
However, Goldner et al. suggest that such a signature could be equally, or better, explained by the CO2 hypothesis. When the researchers included an enlarged Antarctic ice sheet in their numerical climate model of the EOT, the predicted temperature change in the Atlantic was very similar to that inferred from the ocean sediment cores. This was not the case when they imposed modifications to the ocean gateways in line with the gateway hypothesis (in contrast to previous work6, which the authors argue used modelling tools that are less advanced than their own). As such, Goldner et al. conclude that their work provides support for the CO2 hypothesis.
One of the strengths of this paper is that Goldner and colleagues have carefully analysed their model results to understand the climatic mechanisms (Fig. 1) that give rise to the model-predicted patterns of ocean cooling and circulation. However, in my view, several interesting issues still remain.
First, although the authors make inferences about the causes of Antarctic ice-sheet growth, they cannot tackle this explicitly because their modelling does not include a full representation of the interactions between ice and climate. It is possible that a change in gateways caused cooling that led to the growth of the Antarctic ice sheet, and it is these effects that are seen in the geological record. The causes of ice-sheet growth (or retreat) are best understood through the use of integrated climate and ice-sheet models. Ice-sheet models have recently undergone a period of rapid development (see ref. 7, for example), so the time is ripe to apply these to palaeoclimate events such as the EOT.
Second, models are by definition approximations of the real world, and it will be crucial for other groups to verify these findings with their own models. This is particularly important because other recent modelling work8 has indicated surface warming in the Atlantic sector in response to an increased Antarctic ice sheet — although that study focused on the more recent Middle Miocene climate transition, which occurred about 14 million years ago. This is in contrast to the earlier cooling found by Goldner and colleagues.
Third, the question also remains as to why greenhouse gases changed at this time, and by how much. Indirect estimates of CO2 concentration indicate a concurrent drop9,10, although the uncertainties for these estimates are currently large. Possible causes include changes in the balance of sources (for example, decreased volcanism) and/or sinks (such as increased weathering of silicate rocks), and/or changes to reservoirs of carbon (for instance, an increase in the residence time of carbon in the ocean, owing to changes in ocean circulation). Picking apart these possible causes is a crucial challenge.
Goldner and colleagues conclude their paper with a word of warning, noting that a complex web of positive and negative feedbacks means that the climate system can often behave unexpectedly. This can be interpreted as a strong note of caution regarding humanity's own current CO2 'experiment' with the climate system.
Goldner, A., Herold, N. & Huber, M. Nature 511, 574–577 (2014).
Francis, J. E. et al. in Antarctic Climate Evolution (eds Florindo, F. & Siegert, M.) 309–368 (Elsevier, 2008).
Lear, C. H., Bailey, T. R., Pearson, P. N., Coxall, H. K. & Rosenthal, Y. Geology 36, 251–254 (2008).
Liu, Z. et al. Science 323, 1187–1190 (2009).
Cramer, B. S., Toggweiler, J. R., Wright, J. D., Katz, M. E. & Miller, K. G. Paleoceanography 24, PA4216 (2009).
Sijp, W. P., England, M. H. & Huber, M. Paleoceanography 26, PA4207 (2011).
Cornford, S. L. et al. J. Comput. Phys. 232, 529–549 (2013).
Knorr, G. & Lohmann, G. Nature Geosci. 7, 376–381 (2014).
Pearson, P. N., Foster, G. L. & Wade, B. S. Nature 461, 1110–1113 (2009).
Pagani, M. et al. Science 334, 1261–1264 (2011).