A new ice-core record from Antarctica provides the best evidence yet of a link between climate in the northern and southern polar regions that operates through changes in ocean circulation.
Over the past 20 years, the analysis of ice cores has been transforming our understanding of past climate. Most notably, the Vostok core from Antarctica1 provided remarkable evidence of the correspondence between temperature and atmospheric carbon dioxide concentrations over the past 420,000 years. And the GISP2, GRIP and NGRIP cores from Greenland2,3 offered a view in unprecedented detail of climate change over the past 100,000 years (including the revelation that abrupt warming events of 10 °C or more have taken place in Greenland).
More recently, the European Project for Ice Coring in Antarctica (EPICA) obtained the longest ice-core record yet4, one spanning 800,000 years of climate history, from Dome C, in the same sector of Antarctica as Vostok. The EPICA team has now pulled off another feat. As it reports on page 195 of this issue5, it has completed analysis of 2,500 metres of an ice core from Dronning Maud Land in the Atlantic sector of Antarctica (Fig. 1).
The significance of this 'EDML' core is not its sheer length; at the site concerned, 2,500 m takes us back 150,000 years. Rather, it is its resolution: this is the highest-resolution record obtained outside Greenland that extends well beyond the last glacial maximum (about 20,000 years ago). In consequence, the EPICA team has been able to place the EDML record with high precision on the same timescale as the records from Greenland. This allows us to compare the Greenland and Antarctic records over time intervals as short as a few centuries.
At this point, an analogy may help. On a year-to-year basis, much global climate variability is dominated by the El Niño–Southern Oscillation (ENSO). Understanding ENSO has required comparison of climate records in the main ENSO region (the tropical Pacific) with records elsewhere. This could not have been achieved without placing different time-series data on the same timescale. We would otherwise have little confidence, for example, in the observed correspondence between ENSO and rainfall variations in southern California. Nor would it be meaningful to try to understand the ocean and atmospheric dynamics that give rise to that relationship, because we could not reject the null hypothesis that it was merely due to chance. The Greenland and Antarctic ice-core records are likewise measures of climate variability, but on timescales of centuries, millennia and longer. Indeed, the high-resolution measures of climate afforded by ice-core records show unambiguously that climate varies on these longer timescales much more widely than one would expect from simple extrapolation of the power spectrum of observed (modern) climate6.
The usual explanation for the millennial-scale variability is that it is due to changes in the deep meridional overturning circulation (MOC) in the Atlantic Ocean. Put simply, a vigorous MOC is thought to deliver heat to the North Atlantic at the expense of the Southern Ocean. Increases and decreases in MOC strength should thus result in a climate 'see-saw' between the Southern and Northern Hemispheres7. An observation cited in support of this idea is that there is an out-of-phase relationship between Antarctic and Greenland ice-core records of temperature (or rather, of oxygen and deuterium isotope ratios, which are well-established proxies for temperature)8. The problem has been that — unlike in the ENSO analogy — there is considerable uncertainty in the dating. Analyses of the existing records have generally shown that the relationship between the Greenland and Antarctic records is weak and not statistically significant, except on the very longest timescales associated with well-understood astronomical factors (the Milankovich forcing of ice ages)9. So it has been difficult to rule out the null hypothesis that the variability in these records largely reflects regional phenomena such as variations in wind patterns or sea-ice extent.
This is where the EDML results come in. They show that the Antarctic and Greenland ice-core records are meaningfully related, and on quite short timescales. In particular, comparison of the oxygen-isotope records shows that one can make a direct link between the distinctive temperature maxima in the Antarctic record (at least going back 60,000 years) and the unambiguous abrupt warmings in Greenland (Fig. 2). Not every Antarctic temperature maximum is as distinct as the Greenland warmings; for example, it is not clear why the small maxima labelled 6 and 10 in Fig. 2 should count as 'events', but the similarly sized bumps between maxima 1 and 2 should not. But the relationship is too strong to be due to chance.
In fact, for the interval 20,000 to 90,000 years ago a remarkable 40% of the variance in the Greenland records can be explained by the EDML time series. A more rigorous estimate of the spectral coherence between the records shows that this significant relationship extends to timescales as short as a few centuries. Furthermore, there is a consistent out-of-phase relationship between the records. They are not strictly 'antiphased', as the term see-saw would imply. Rather, the average phase relationship is about 90°. Although cold conditions in Greenland tend to be associated with warming in Antarctica, and vice versa, the peak warmth in both records actually occurs at about the same time.
Does the EDML record demonstrate the dominant influence of the MOC on climate variability? This is not just an academic question. Variations in this circulation have been invoked to explain everything from the abrupt climate changes observed in the Greenland records to the Little Ice Age (a period of cooling between about AD 1400 and 1900 in the North Atlantic region). And the possibility of a sensitive MOC has been proposed as a 'tipping point' in future human-influenced climate change10.
One objection to these ideas is that the MOC plays a minor role in the heat budget of the polar regions11. Heat transport in the atmosphere is much more important, and the atmosphere might simply compensate for any changes in MOC12. Furthermore, the causes of the purported changes in MOC are not understood. The conventional answer — flooding of the North Atlantic Ocean by ice and meltwater from the Laurentide ice sheet in northern North America (so-called Heinrich events) — is not very convincingly supported by the evidence13.
The EDML data do not directly address these concerns. But they are nonetheless compatible with the idea that the MOC has a central role in millennial-scale variability. What is particularly compelling is that there is a strong linear relationship between the magnitude of warming in Antarctica and the duration of the warm period that follows each abrupt event in Greenland (see Fig. 3 of the paper5 on page 197). The authors' explanation is simple: the duration of the warm periods in Greenland reflects the duration of reduced MOC, and hence the amount of heat retained in the Southern Ocean. This is consistent with a model14, proposed a few years ago, in which the magnitude of Antarctic temperature change is controlled by the effective size of the Southern Ocean heat reservoir (including both dynamic and thermodynamic effects). We may have to wait some time before we see whether these results can be reproduced by more sophisticated ocean–atmosphere climate models, because realistically encapsulating the dynamics of the Southern Ocean in such models remains a problem. But we can hope that these new results5 will inspire the relevant work to be done.
About this article