Orbital and Millennial Antarctic Climate Variability over the Past 800,000 Years

See allHide authors and affiliations

Science  10 Aug 2007:
Vol. 317, Issue 5839, pp. 793-796
DOI: 10.1126/science.1141038


A high-resolution deuterium profile is now available along the entire European Project for Ice Coring in Antarctica Dome C ice core, extending this climate record back to marine isotope stage 20.2, ∼800,000 years ago. Experiments performed with an atmospheric general circulation model including water isotopes support its temperature interpretation. We assessed the general correspondence between Dansgaard-Oeschger events and their smoothed Antarctic counterparts for this Dome C record, which reveals the presence of such features with similar amplitudes during previous glacial periods. We suggest that the interplay between obliquity and precession accounts for the variable intensity of interglacial periods in ice core records.

The European Project for Ice Coring in Antarctica (EPICA) has provided two deep ice cores in East Antarctica, one (EDC) at Dome C (1), on which we focus here, and one (EDML) in the Dronning Maud Land area (2). The Dome C drilling [fig. S1 and supporting online material (SOM) text] was stopped at a depth of 3260 m, about 15 m above the bedrock. A preliminary low-resolution δD record was previously obtained from the surface down to 3139 m with an estimated age at this depth of 740,000 years before the present (740 ky B.P.), corresponding to marine isotope stage (MIS) 18.2 (1). Other data, such as grain radius, dust concentration, dielectric profile, and electrical conductivity, as well as chemical data (3), are available down to this depth, and analyses of the entrapped air have extended the greenhouse gas record—i.e., CO2, CH4, and N2O—back to MIS 16, ∼650 ky B.P. (4, 5).

We completed the deuterium measurements, δDice, at detailed resolution from the surface down to 3259.7 m. This new data set benefits from a more accurate dating and temperature calibration of isotopic changes based on a series of recent simulations performed with an up-to-date isotopic model. In turn, this very detailed Antarctic surface temperature record sheds light on climate analyses in four ways: (i) It allows reliable extension of the climate record back to MIS 20.2 (∼800 ky B.P.); (ii) it resolves Antarctic millennial variability over eight successive glacial periods; (iii) it allows quantifiable comparison of the strengths of the successive interglacial and glacial periods; and (iv) the improved time scale allows more accurate investigation of the links between Antarctic temperature and orbital forcing.

This detailed and continuous δDice profile is shown as a function of time in Fig. 1 and on a depth scale in figs. S2 and S3. For our analysis, we adopted a more precise time scale (SOM text), in which EDC3 has a precision of ±5 ky on absolute ages and of ±20% for the duration of events (6, 7). This scale clearly indicates that the Antarctic counterpart of MIS 15.1 was too long by about a factor 2 in EDC2 (1), as already suggested from the comparison with the deep-sea core record (8), whereas the scale confirms the long duration of MIS 11.3 (1).

Fig. 1.

Comparison of the δD Dome C record on the EDC3 time scale (with all data points in light gray and a smoothed curve in black) with the benthic oxygen-18 record (blue) on its own time scale (8). The 3259.7-m δD record, which includes published results down to 788 m (38), benefits from an improved accuracy (1σ of ± 0.5‰) and a much more detailed resolution of 55 cm all along the core, whereas the previously published record was based on 3.85-m samples (1). The agreement between the two time series back to ∼800 ky B.P. justifies the use of oceanic sediment nomenclature (MIS) for describing the ice core record.

The deep-sea benthic oxygen-18 record (8) and the δDice Dome C record are in excellent overall agreement back to ∼800 ky B.P. (MIS 20.2), which suggests that our extended EPICA Dome C record now entirely encompasses glacial stage MIS 18 and interglacial MIS 19. This agreement does not hold true for the earlier part of the record below ∼3200 m, and we have strong arguments that the core stratigraphy has been disturbed over its bottom 60 m (SOM text). In contrast, the stratigraphic continuity of the record above ∼3200 m is supported by all available data, including preliminary CO2 and CH4 measurements performed along the transition between MIS 20.2 and 19 (SOM text). We are thus confident that the Dome C δD record provides an ∼800-ky-old reliable climatic record.

Results derived from a series of experiments performed with the European Centre/Hamburg Model General Circulation Model implemented with water isotopes (9) for different climate stages (SOM text) allowed us to assess the validity of the conventional interpretation of ice core isotope profiles (δD or δ18O) from inland Antarctica, in terms of surface temperature shifts (fig. S4). We inferred that the change in surface temperature (ΔTs) range, based on 100-year mean values, was ∼15°C over the past 800 ky, from –10.3°C for the coldest 100-year interval of MIS 2 to +4.5°C for the warmest of MIS 5.5 (Fig. 2). Despite some differences, the three long East Antarctic isotopic records, Dome C, Vostok (10, 11), and Dome F (12), show a very high level of similarity over their common part and the EDC temperature record is expected to be representative of East Antarctica. All glacial stages before 430 ky B.P. are warmer than MIS 2, by ∼1°C for MIS 12, 16 and 18 and by ∼2°C for MIS 14 (Fig. 2).

Fig. 2.

Dome C temperature anomaly as a function of time over the past 810 ky. Back to 140 ky B.P., we report 100-year mean values, whereas for earlier periods (middle and lower traces), ΔTs is calculated from 0.55-m raw data; a smooth curve using a 700-year binomial filter is superimposed on this detailed record. In the upper trace (which is plotted on a more highly resolved time axis), we show the correspondence between the DO events as recorded in the North Greenland Ice Core Project isotopic record (2, 15) and AIM events recorded in the EDC temperature record during the last glacial period and the last deglaciation. We have indicated the successive MIS, and the transitions are labeled from TI to TIX.

We confirm that the early interglacial periods, now including MIS 19, were characterized by less pronounced warmth than those of the past four climatic cycles (1). Whereas peak temperatures in the warm interglacials of the later part of the record (MIS 5.5, 7.5, 9.3, and 11.3) were 2° to 4.5°C higher than the last millennium, maximum temperatures were ∼1° to 1.5°C colder for MIS 13, 15.1, 15.5, and 17, reaching levels typical of interstadials, such as 7.1 and 7.3. MIS 19 shows the warmest temperature for the period before Tv (∼ –0.5°C). For MIS 11 to MIS 17, with the exception of MIS 15.1, peak warmth occurred at the end of the warm periods in contrast with the more recent interglacials for which earlier peak warmth was typical (Fig. 2).

Although isotopic records from Antarctica do not exhibit the rapid and large climate variability observed in Greenland records for the so-called Dansgaard-Oeschger (DO) events of the last glacial period (1315), they clearly exhibit millennial-scale variability with muted and more symmetrical events. Synchronization based on gas indicators unambiguously showed that large DO events have Antarctic counterparts (16, 17), and there were indications that shorter events also have such counterparts both from Vostok and Dome C cores (1820).

The recent high-resolution EDML isotopic profile over the last glacial period has unambiguously revealed a one-to-one correspondence between all these Antarctic Isotope Maxima (AIM) and DO events (2), which with a few exceptions holds true for the EDC core over the entire last glacial period back to DO 25 (Fig. 2 and fig. S5). At Dome C, the typical amplitude of larger events is ∼2°C, much lower than for corresponding DO warmings in Greenland, which are often larger than 8°C and as high as 16°C (21, 22). Although some AIM events are more prominent in one of the two EPICA sites (fig. S5), they record millennial variability of comparable magnitude (SOM text), despite the fact that EDML is situated in the Atlantic sector whereas Dome C is facing the Indo-Pacific Ocean. Atmospheric circulation and/or efficient circumpolar oceanic currents can contribute to distribute such climatic signals around Antarctica. This detailed comparison between EDC and EDML records further supports the thermal bipolar seesaw hypothesis (23), which postulates that abrupt shutdowns and initiations of the Atlantic meridional overturning circulation produce slow warmings and coolings in the Southern Ocean and Antarctic region.

Our record exhibits quite similar millennial climate variability during the past three glacial periods, in terms of both magnitude and pacing (fig. S5), suggesting this was also the case in the North Atlantic, as indicated by sediment data (24) and inferred from CH4 data from Antarctic cores (5, 25). Our lower temporal resolution prevents clear detection of small AIM for earlier glacials, but the amplitude of large AIM, thus presumably of large DO events, does not appear to be substantially influenced by the smaller extent of Northern Hemisphere ice sheets before Termination V. In particular, a very well featured sequence is displayed by the additional cycle provided by the extension of the core from 740 and 800 ky B.P. (Fig. 2 and fig. S5) with three well-marked oscillations that have not yet had counterparts identified in the MIS 18 ocean record (8). Finally, our record shows that during each glacial period, AIM events appear once Antarctic temperatures have dropped by at least 4°C below late Holocene temperature (Fig. 2). We suggest that decreases in Antarctic temperature over glacial inceptions modified the formation of Antarctic bottom waters and that the associated reorganization in deep ocean circulation is the key for the onset of glacial instabilities.

Obliquity changes were previously invoked to explain the change in amplitude between glacial and interglacial periods at the time of the Mid-Brunhes Event (MBE), ∼430 ky B.P. (1). This key characteristic of the EDC δD record is now fully supported by our 800-ky detailed temperature record and its improved EDC3 time scale (Fig. 3). Dominated by a periodicity of ∼100 ky, the power spectra of ΔTs (fig. S6) also reveal a strong obliquity component and point to the influence of the precession, at least for 0 to 400 ky. The relative strength of the obliquity and 100-ky components increases when going from past to present, which is consistent with the increasing amplitude of obliquity variations over the past 800 ky due to a 1.2-million-year modulation (26). The 40-ky component is particularly strong, accounting for one-third (4.3°C) of the total range of temperature in the 800-ky record (Fig. 3). Also noticeable are its strong coherency with 65°N summer insolation in the obliquity range (0.97) and its substantial ∼5-ky lag with respect to obliquity (fig. S7).

Fig. 3.

(A) Precession parameter displayed on an inversed vertical axis (black line). (B) EDC temperature [solid line, rainbow colors from blue (cold temperatures) to red (warm temperatures)] and its obliquity component extracted using a Gaussian filter within the frequency range 0.043 ± 0.015 ky–1 [dashed red line, also displayed in (D) as a solid red line on a different scaling]. Red rectangles indicate periods during which obliquity is increasing and precession parameter is decreasing. (C) Combined top-of-atmosphere radiative forcing due to CO2 and CH4 (solid blue) and its obliquity component [dashed blue, also displayed in (D) as a solid blue line on a different scaling]. (D) Obliquity (solid black line), obliquity component of EDC temperature (red line), and obliquity component of the top-of-atmosphere radiative forcing due to CO2 and CH4 (blue). Insolations were calculated using the Analyseries software (39).

Intermediate-complexity climate models indeed capture a high-latitude signature in annual mean temperature in response to extreme configurations of obliquity, albeit half of that observed here (27). With respect to the strong linear relationship between δD and obliquity, the link may be local insolation changes, which at 75°S vary by ∼8% up to 14 W/m2 (28). Such changes in high-latitude insolation may be amplified by associated changes in heat and moisture transport in the atmosphere (including water vapor and sea-ice feedbacks at high latitudes). They can thus generate changes in the density of ocean surface waters and therefore in ocean thermohaline circulation; such processes involve deep ocean heat storage with constants of millennia. Notably, the obliquity components of temperature records from the tropical Pacific and from Antarctica are in phase (29) within age scale uncertainties. They are thus in phase with the mean annual insolation at high latitude but out of phase with the obliquity component of the mean annual insolation in the tropics. This indicates that mechanisms transferring the high-latitude effect of obliquity toward the tropics may involve changes in heat export from the tropics.

Our EDC ice core shows no indication that greenhouse gases have played a key role in such a coupling. Not only does the obliquity component of the radiative forcing—calculated accounting for both CO2 and CH4 changes (30)—have a small amplitude over the past 650 ky (∼0.5 W/m2, Fig. 3) but it also seems to lag Antarctic and tropical temperature changes. Nor can this in-phase temperature behavior be explained by local insolation, given that this parameter is in antiphase between low and high latitudes. Rather than assuming that this is caused by greenhouse coupling, we suggest that it results from a transfer of the high-latitude obliquity signal to the tropics through rapid processes involving atmospheric circulation or intermediate oceanic waters, possibly linked, as documented from present-day climates (31) and examined for past climates (32), with changes in sea-ice around Antarctica. The amplitude of the radiative greenhouse forcing, however, is very important in the 100-ky band (∼2.5 W/m2 comparable to the additional greenhouse forcing due to anthropogenic activities). This points to a strong carbon-cycle feedback involved in the magnitude and possibly duration of ice ages (33) and to a global character of the Antarctic temperature record.

One key question in this frequency band concerns the relative role of the different orbital parameters in driving terminations. Some authors suggest that terminations occur at multiples of obliquity (34, 35) or precession cycles (36). The latter includes the insolation canon hypothesis that calls upon the interplay of precession and obliquity with considerations of total energy input and threshold effects (37). With our current age scale, the insolation canon approach works well for TI to TIV but not for earlier terminations. We support the view of combined effects of precession and obliquity in driving ice age dynamics but suggest that the role of obliquity is underestimated by this approach [e.g., high-latitude insolation should not be considered for mid-month but integrated over several months (35)].

The strength of interglacials is highly variable along the record. We suggest that this variation results from an interplay between obliquity and precession (Fig. 3). When 65°N summer insolation (or the inverted precession parameter) and obliquity changes peak in phase (within 5 ky), their combined effects induce strong interglacial periods (MIS 1, 5, 9, 11, and 19). When they are in antiphase, compensating effects induce weak interglacial intensities (MIS 13, 15, 17, and 7.3). In this line, we calculated for each interglacial the cumulative warmth defined with respect to a temperature threshold; we then explored the relationship between this index and insolation. This analysis accounts for the effects of both precession (in the timing of glacial-interglacial transitions) and obliquity (through the mean annual high-latitude insolation). The most robust result is obtained when comparing the cumulative warmth with the cumulative high-latitude insolation (above its average value and taking into account the phase lag of 5 ky). Whereas for small changes in insolation, there is no clear relationship between these two, a linear relationship is observed when the cumulative insolation is larger than ∼1700 GJ/m2 (fig. S8). In turn, we suggest a causal link between the change of amplitude observed in the EDC temperature record and the modulated amplitude of obliquity.

Our new high-resolution Antarctic climate record is able to resolve systematic long-term as well as millennial changes over the past 800,000 years. Whereas the former may be controlled by local insolation changes largely induced by the obliquity cycle, the latter are induced by changes in North Atlantic deep water formation through the thermal bipolar seesaw. Clearly shown for the last glacial cycle, this is also suggested for earlier glacial periods. Overall, our Antarctic temperature record points to an active role for high southern latitudes in the dynamics of climate change both at orbital and millennial time scales, rather than to a picture of these polar regions simply recording variability originating from other parts of the climate system. This climate record will now serve as a benchmark for exploiting the many properties that are, or will be in the near future, measured on the Dome C core, both in the ice (elemental and isotopic composition of dust and of chemical compounds) and in the gas phase (records of greenhouse gases, other atmospheric compounds, and their isotopic signatures).

Supporting Online Material

SOM Text

Figs. S1 to S8


References and Notes

View Abstract

Navigate This Article