1. Introduction

The 3623 m long Vostok 5G core was drilled in the East Antarctic plateau between 1991 and 1997. The Vostok core contains a record of climate over at least the past ∼400 kyr (Reference PetitPetit and others, 1999). However, the stratigraphic integrity of the section below ∼3310 m has been questioned (Reference PetitPetit and others, 1999). In fact, stratigraphic disturbance of the bottom part of the ice sheet has been observed in some of the other long ice cores from polar regions, such as the GRIP and GISP2 ice cores from Summit, Greenland (Reference Chappellaz, Brook, Blunier and MalaizéChappellaz and others 1997; Reference LandaisLandais and others, 2003; Reference Suwa, von Fischer, Bender, Landais and BrookSuwa and others, 2006). More recently, Reference RaynaudRaynaud and others (2005) suggested that the section of the Vostok 5 G core between 3321.0 and 3344.9 m might contain ice from Termination V, but in reversed stratigraphic order. They reached the conclusion by visually comparing CH₄ and CO₂ measurements of samples from this section of the Vostok core with those of the EPICA Dome C (Antarctica) core, which now extends back to >650 kyr (Reference Siegenthaler, Stocker, Monnin, Lüthi, Schwander and StaufferSiegenthaler and others, 2005; Reference SpahniSpahni and others, 2005).

Considering that glacial terminations are periods of intense scientific interest, and the precious nature of ice samples this old (to date, only the EPICA Dome C and Dome Fuji cores include Termination V ice), it is important to reconstruct the stratigraphy of this section of the Vostok core. Therefore, this paper aims to constrain the age–depth relationship of the Vostok ice core for the section between 3300 and 3347 m using gas properties. In addition to the CH₄ and CO₂ records considered by Reference RaynaudRaynaud and others (2005), we incorporate measurements of δ¹⁸O of atmospheric O₂ (δ¹⁸O_atm) into the age reconstruction. We use the recently published ∼800 kyr long δ¹⁸O_atm record of the EPICA Dome C core for the reference (Reference DreyfusDreyfus and others, 2007) and new δ¹⁸O_atm measurements of samples from Vostok for our analysis. We note that there is no indication that the three gas properties considered in this study had been significantly modified or contaminated during ice-core extraction, core transport and gas extraction. In fact, CO₂ and CH₄ are within the expected range (Reference PetitPetit and others, 1999) and so are δ¹⁸O and total gas content (Reference SuwaSuwa, 2007).

2. Reconstructing the Chronology Using CH₄–δ¹⁸O_atm–CO₂

The method used in this study is similar to that reported by Reference Suwa, von Fischer, Bender, Landais and BrookSuwa and others (2006), who reconstructed the chronology for the bottom sections of the GISP2 and GRIP ice cores. The basis of their method is that individual samples from the stratigraphically disturbed section are dated based on their gas concentrations, which can be compared with concentration–time profiles of a stratigraphically intact ice core (i.e. Dome C). The difference between their method and the method developed here is that, in addition to the two gas properties they used, δ¹⁸O_atm and CH₄, we use CO₂ as a third constraint. CO₂ was not useful in reconstructing the GISP2/GRIP chronologies because it cannot be measured reliably in those cores (Reference AnklinAnklin and others, 1997; Reference Smith, Wahlen, Mastroianni and TaylorSmith and others, 1997).

2.1. CH_4–δ¹⁸O_atm–CO₂ records of the Vostok and EPICA Dome C ice cores

For the Vostok CH₄ and CO₂ records, we use values reported by Reference PetitPetit and others (1999). For δ¹⁸O_atm we combine measurements by Reference PetitPetit and others (1999) and the new set of δ¹⁸O_atm measurements. The method for determining δ¹⁸O_atm for the new dataset is similar to the method described in Reference Sowers, Bender and RaynaudSowers and others (1989). We use a ∼15g sample of ice for each measurement, and apply the double melt refreeze method to extract air trapped in bubbles and clathrate hydrates in an ice sample. We then correct for the gravitational enrichment by subtracting 2 × δ¹⁵N from δ¹⁸O of the same sample to give δ¹⁸O_atm (Reference Craig, Horibe and SowersCraig and others, 1988; Reference Schwander, Oeschger and LangwaySchwander, 1989). Unfortunately, δ¹⁸O_atm (and in some cases CO₂) is not measured in exactly the same sample as CH₄. Therefore, we first choose 14 samples where CH₄ and CO₂ are measured at the same depths. Values for δ¹⁸O_atm are then interpolated for these depths. We note that δ¹⁸O_atm changes gradually in this section, which indicates that small-scale mixing is not dominant here. Figure 1 shows CH₄, CO₂, δ¹⁸O_atm and δD_ice of the Vostok ice core for the section between 3200 and 3400 m. Black solid circles indicate where original measurements are located. Red squares indicate ‘interpolated’ depths that we use for the following analysis. δD_ice is the temperature proxy for Vostok and Dome C.

Fig. 1.

The CH₄, CO₂, δ¹⁸O_atm and δD_ice records for the section between 3200 and 3400 m of the Vostok ice core. Black solid circles indicate where original measurements are located. Red squares indicate values at interpolated depths used for the analysis.

Next, we derive the ‘reference’ line, which we assume is the true atmospheric history, based on the undisturbed record of EPICA Dome C. The ‘reference line’ in Figure 2 refers to the trajectory of δ¹⁸O_atm vs CH₄ vs CO₂ concentration between 400 and 650 kyr on the EPICA Dome C EDC2 timescale (EPICA Community Members, 2004). We use the published CO₂ (Reference Siegenthaler, Stocker, Monnin, Lüthi, Schwander and StaufferSiegenthaler and others, 2005), CH₄ (Spahni and others, 2007) and δ¹⁸O_atm (Reference DreyfusDreyfus and others, 2007) records to define the reference line.

Fig. 2.

(a) An x-y-z plot of CO₂ vs δ¹⁸O_atm vs CH₄. The green curve traces the trajectory of CO₂, CH₄ and δ¹⁸O_atm points between 400 and 650 kyr in the EPICA Dome C core. Closed circles colored in shades of red and blue are Vostok data from the disturbed section of the core; the color represents δD_ice at the depth of sampling. The black curve connects Vostok samples in order of increasing depth. (b) As for (a) but shown only for CH₄ and δ¹⁸O_atm. Gray ellipses surrounding each data point are uncertainty ranges. Black letters next to closed circles indicate the depths (m) of the Vostok sample. Orange labels next to yellow dots indicate ages (kyr) of Dome C samples.

The pooled standard deviations for EPICA Dome C and Vostok are: ±0.028‰ (1σ) (Reference DreyfusDreyfus and others, 2007) and ±0.05‰ (1σ), respectively, for δ¹⁸O_atm; ±15 ppb (1σ) (Spahni and others, 2007) and ±20 ppb (1σ) (Reference PetitPetit and others, 1999), respectively, for CH₄; and ±1.5 ppm (1σ) (Reference Siegenthaler, Stocker, Monnin, Lüthi, Schwander and StaufferSiegenthaler and others, 2005) and ±3 ppm (1σ) (Reference PetitPetit and others, 1999), respectively, for CO₂. These errors are added quadratically for each gas property. We adopt 2σ uncertainties, so that errors associated with δ¹⁸O_atm, CH₄ and CO₂ are ±0.11‰, ±50 ppb and ±6.7 ppm, respectively, in the following analysis.

2.2. Method

We first identify all compatible ages for each sample, which are our estimation of age uncertainties. This is done by seeking values for time t on the EPICA EDC2 timescale that meet the condition,

(1)

where CH_{4 EDC}(t) is the CH₄ concentration, δ¹⁸O_{atm EDC}(t) is the δ¹⁸O_atm value and CO_{2 EDC}(t) is the CO₂ concentration in EPICA Dome C at time t. Likewise, CH_4Vtk(z), δ¹⁸O_{atm Vtk}(z) and CO_{2 Vtk}(z) are CH₄ concentration, δ¹⁸O_atm value and CO₂ concentration at depth z in a sample from the section between 3300 and 3347 m of the Vostok core, and σ(CH₄), σ(δ¹⁸O_atm) and σ(CO₂) are one standard deviation for CH₄, d¹⁸O_atm and CO₂, respectively, as described in section 2.1.

Second, we derive the ‘best-estimate age’ within all compatible age ranges derived according to Equation (1). The best-estimate age is defined as t which minimizes d, the distance between a sample and the reference line, calculated as

(2)

where the subscript n indicates that the term is normalized to a mean of 0 and standard deviation of 1. Normalization weights each of the three properties equally.

We first derive gas ages of samples, based on Equations (1) and (2), which are several thousand years younger than ages of ice at the same depths. This gas-age/ice-age difference (Δ_age) arises from the fact that the air is trapped in ice at ∼100 m below the surface of the ice sheet, and thus Δ_age should be taken into account when plotting proxy signals recorded in ice vs time. We compute Δ_age using the empirical model of Reference Herron and LangwayHerron and Langway (1980), which requires temperature and accumulation rate as input parameters. We estimate temperature from a simple linear regression of ΔT (deviation from modern Vostok temperature of −57.4°C) on δD_ice using values reported in Reference PetitPetit and others (1999), and we estimate accumulation rate from the equation used in Reference PetitPetit and others (1999). We understand that there are relatively large uncertainties associated with this model, especially for low-accumulation sites, but these uncertainties do not have a large impact on our results.

3. Results and Discussion

The derived ages for 14 samples examined in this study produce an excellent fit between the two ice-core records, for all four geochemical parameters. The age–depth relationship so derived (Fig. 3) allows us to draw the following five conclusions. First, 9 of 14 samples have only one compatible age range which is consistent with EPICA Dome C ages between 400 and 650 kyr. Samples from 3328, 3331 and 3334 m have multiple compatible age ranges, but all ages fall between 416 and 424 kyr. The bottommost two samples of our analysis, the samples from 3343 and 3346 m, also have two compatible age ranges. One is ∼440 kyr and the other is ∼460 and ∼450 kyr, respectively. Second, our age reconstruction indicates that the stratigraphic disturbance starts somewhere between 3316 and 3319 m. Third, four or five samples between 3319 and ∼3330 m are in reversed order. In this depth interval, one finds younger age as one goes deeper in the ice core. Fourth, it appears that there is a second stratigraphic disturbance between 3340 and 3343 m. Although the exact mechanism of development of these folds remains to be examined, layered rheological contrasts across the interglacial–glacial ice might have contributed to the process. Such folds are similar to the ones observed in the horizontal ice core retrieved from the Pâkitsoq ice-sheet margin site in Greenland (Reference Petrenko, Severinghaus, Brook, Reeh and SchaeferPetrenko and others, 2006). Lastly, the oldest sample in the section between 3300 and 3347 m is dated to at least ∼440 kyr on the EPICA Dome C EDC2 timescale.

Fig. 3.

The age–depth relationship reconstructed in this study. Horizontal black bars indicate all compatible age ranges. The ‘best-estimate ages’ are plotted as solid circles. Shades of red and blue indicate dD of ice at the same depth.

Figure 4 shows the comparison of CO₂, CH₄, δ¹⁸O_atm and δD_ice records between EPICA Dome C and Vostok between 400 and 650 kyr. Black curves show the EPICA Dome C records, and blue circles are Vostok values plotted on their best-estimate ages. Our analysis supports the earlier suggestion by Reference RaynaudRaynaud and others (2005) that the temperature history during Termination V is similar at the two sites.

Fig. 4.

CO₂, CH₄, δ¹⁸O_atm and δD records between 400 and 650 kyr. Black curves show EPICA Dome C. Blue circles indicate Vostok samples plotted on the best-estimate ages. Note that the y axis for δ¹⁸O_atm is reversed, and δD_ice is plotted on offset scales for EPICA Dome C and Vostok.

As stated above, we limited our age search between 400 and 650 kyr. Therefore, we cannot completely dismiss the possibility that these samples are from much older age intervals. However, we note that the EPICA Dome C record shows δ¹⁸O_atm of ∼1.5‰ during Termination V. This heavy δ¹⁸O_atm value is noteworthy as it is not reached for the rest of the published record which goes back to ∼800 kyr (Reference DreyfusDreyfus and others, 2007). Therefore, the heaviest δ¹⁸O_atm value we observed in the Vostok samples, which is ∼1.5‰, indicates that this sample is uniquely dated over the whole 800 kyr range.