gas age–ice age differences and the chronology of the...

Gas age–ice age differences and the chronology of the Vostok ice core,

0–100 ka

M. L. Bender,1 G. Floch,2 J. Chappellaz,3 M. Suwa,1 J.-M. Barnola,3 T. Blunier,2

G. Dreyfus,1,4 J. Jouzel,4 and F. Parrenin3

Received 13 July 2005; revised 30 March 2006; accepted 17 July 2006; published 14 November 2006.

[1] Gas is trapped in polar ice at depths of �50–120 m and is therefore significantlyyounger than the ice in which it is embedded. The age difference is not well constrainedfor slowly accumulating ice on the East Antarctic Plateau, introducing a significantuncertainty into chronologies of the oldest deep ice cores (Vostok, Dome Fuji, and DomeC). We recorrelate the gas records of Vostok and Greenland Ice Sheet Project 2 (GISP2)cores in part on the basis of new CH4 data and use these records to construct six Vostokchronologies that use different assumptions to calculate gas age–ice age differences. Wethen evaluate these chronologies by comparing times of climate events at Vostok withcorrelative events in the well-dated Byrd ice core (West Antarctica). From this evaluationwe identify two leading chronologies for the Vostok core that are based on recent modelsof firn temperature, firn densification, and thinning of upstream ice. One chronologyinvolves calculating gas age–ice age differences from these models. The second, new,approach involves calculating ice depths in the core that are contemporaneous with depthsin the same ice core whose gas ages are well constrained. This latter approach circumventsproblems associated with highly uncertain accumulation rates in the Vostok core. Theuncertainty in Vostok chronologies derived by correlating into the GISP2 gas recordremains about ±1 kyr, and high-precision correlations continue to be difficult.

Citation: Bender, M. L., G. Floch, J. Chappellaz, M. Suwa, J.-M. Barnola, T. Blunier, G. Dreyfus, J. Jouzel, and F. Parrenin (2006),

Gas age–ice age differences and the chronology of the Vostok ice core, 0–100 ka, J. Geophys. Res., 111, D21115, doi:10.1029/

2005JD006488.

1. Introduction

[2] Ice cores record environmentally significant proper-ties both in the gas phase and the ice phase. Propertiesrecorded in the ice include aerosols content and the isotopiccomposition of precipitation. Properties registered in the gasphase include the concentration and isotopic composition ofbiogenic greenhouse gases. Gases are trapped �50–120 mbelow the surface, where the ice is compressed to a densityof about 0.82 gm cm�3 (density of pure ice = 0.917). Firn isopenly porous and permeable down to the trapping depth.Air mixes rapidly to this depth, and trapped air is youngerthan the enclosing ice.[3] Examining the timing of environmental changes

recorded in ice and in trapped gas requires correcting forthe age difference in the two phases. In the case of slowly

accumulating East Antarctic ice cores, this difference is verylarge, up to 7 kyr during glacial periods, and the timing ofclimate changes recorded in the two phases will not beaccurate unless the gas age–ice age difference can be wellconstrained. The gas age– ice age difference (Dage) isusually computed from estimates of surface temperature,accumulation rate, and an empirical or mechanistic modelthat expresses close-off depth and Dage in terms of theseproperties [e.g., Herron and Langway, 1980]. This approachworks very well in high-accumulation-rate sites where it hasbeen tested, notably in central Greenland [Goujon et al.,2003]. However, there is a problem concerning East Ant-arctica. A property of firn related to the close-off depth isregistered in the d15N of trapped N2. The registrationmechanism is gravitational fractionation, which causesheavy isotopes and gases to be enriched at a specified ratewith depth [Craig et al., 1988; Schwander, 1989; Sowers etal., 1989] (Figure 1). The problem is that the 15N enrich-ment in glacial samples is up to 1/3 less than predicted bythe empirical and mechanistic models of densification[Sowers et al., 1992; Blunier et al., 2004; Kawamura,2000] (Figure 2). If close-off depth is correctly recordedby d15N, the models overestimate Dage by up to 2 kyrduring glacial times. We can recognize two possible sourcesof error in model estimates of Dage: the model representa-tion of the physical mechanisms leading to firn metamor-phism and bubble close-off, and the temperature and

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, D21115, doi:10.1029/2005JD006488, 2006ClickHere

for

FullArticle

1Department of Geosciences, Princeton University, Princeton, NewJersey, USA.

2Climate and Environmental Physics, Physics Institute, University ofBern, Bern, Switzerland.

3Laboratoire de Glaciologie et Geophysique de l’Environnement,CNRS, St. Martin d’Heres, France.

4Laboratoire des Sciences du Climat et de l’Environnement, InstitutPierre Simon Laplace, Commissariat a l’Energie Atomique, CNRS,Universite de Versailles Saint-Quentin-en-Yvelines, Gif-sur-Yvette, France.

Copyright 2006 by the American Geophysical Union.0148-0227/06/2005JD006488$09.00

D21115 1 of 10

http://dx.doi.org/10.1029/2005JD006488

accumulation histories used to drive the models. Resolvingthe discrepancy between the model and empirical estimatesof Dage is important for accurately characterizing phasingof climate changes in different geographic regions, as wellas for determining the phasing of climate events recorded inthe gas and ice of a single core (e.g., deglacial CO2 rise andEast Antarctica warming). With the advent of a magnificentnew ice core record from East Antarctica extending back toabout 800 ka [EPICA, 2004], resolving this problem hasbecome particularly important.

[4] Models of firn densification and close-off either arecompletely empirical [e.g., Herron and Langway, 1980] orare mechanistic but tuned to/validated by observations[e.g., Arnaud et al., 2000]. In principle, one could testthe applicability of these models to conditions of the lastice age on East Antarctica by examining densification atmodern sites with extremely low temperatures and accu-mulation rates characteristic of glacial periods. However,no such modern sites have been studied. In the absence ofsuch data, there are two contradictory lines of evidencebearing on the question of whether the densificationmodels or d15N of N2 gives a more accurate estimate ofclose-off age. Studies of d15N of N2 in firn air, summarizedbelow, show that shallow convection is minimal at mostmodern sites: today, therefore, d15N generally gives anaccurate measure of the close-off depth. On the other hand,densification models capture the compelling systematicincrease in close-off depth as temperatures fall [e.g.,Herron and Langway, 1980]. One expects that this trendwill continue such that the close-off depth will deependuring glacial times rather than become shallower. Furthersupport for the densification models comes from the workof Caillon et al. [2003], who showed that these modelssuccessfully predict depths where gravitational enrichmentswould begin changing during glacial termination 3 atVostok, if the convective zone thickness were monotoni-cally related to surface temperature.[5] In this paper we first present two new firn air profiles.

We use these results, together with data from the literature,to assess the extent of shallow convection. Shallow con-vection would diminish the d15N enrichment, which ismanifested only where convection is absent and moleculardiffusion can establish isotopic gradients, but would notchange the close-off depth. In addition, we recorrelate thegas records of the Greenland Ice Sheet Project 2 (GISP2)and Vostok ice cores. This correlation uses 17 controldepths, based on the d18O of O2 or the methane concentra-tion, at which the gas age of Vostok is taken to be the sameas the gas age of GISP2. We then derive three Vostok icechronologies by calculating ice ages at the control pointsfrom gas age–ice age differences computed in three recent

Figure 1. Firn air d15N profiles from GISP2 and TaylorDome. The solid lines show the expected increasecalculated using the barometric equation (equation (1)).The line on the GISP plot, drawn through the two deepestpoints in the diffusive zone (57 and 70 m), extrapolates tozero d15N at 2 m depth, indicating a convective zone of thisthickness. The d15N values are reported in per mil.Analytical errors are smaller than the symbols.

Figure 2. The d15N of N2 (per mil) versus depth in Vostok as predicted by a glaciological model[Goujon et al., 2003] and as measured.

D21115 BENDER ET AL.: VOSTOK GAS AGE–ICE AGE DIFFERENCES

2 of 10

D21115

studies. We assign ice ages between control points assumingthat relative accumulation rates in each interval vary accord-ing to the Vostok GT4 chronology [Petit et al., 1999]. Wealso present a new approach to calculating correlationchronologies based on the correlation of control depths.This approach starts by estimating the close-off depth ateach gas control depth either from d15N of N2, or from adensification model. For each gas control depth, we calcu-late the depth of contemporaneous ice, on the basis of thecalculated close-off depth, the average density of the firncolumn, and the degree to which ice has been thinned at thecontrol depth. Again, we interpolate between ice age controlpoints on the basis of the relative accumulation rates of theGT4 chronology.[6] We evaluate these approaches by identifying 19

depths in the Vostok ice core marked by distinctive climateevents (Figure 3). Of these 19 points, we identify 15 atwhich common features appear to be present in the isotopictemperature records of the Vostok and Byrd ice cores. Foreach Vostok chronology we calculate the mean and stan-dard deviation between ages of temperature featuresrecorded in the two cores. Byrd, with small, well-constrained gas age–ice age differences, has been wellcorrelated to GISP2, and we take its chronology as areference [Blunier and Brook, 2001]. The degree of agree-ment with Byrd indicates the accuracy with which eachapproach constrains the gas age–ice age difference or

close-off depth. In this approach, we are implicitly assum-ing that gas age–ice age differences at Byrd are correctlycalculated. Two facts justify this assumption. First, Dagevalues at Byrd are modest (<1.5 kyr), and are unlikely tohave large errors. Second, glacial conditions at Byrd fallwithin the range of temperatures and accumulation rates atwhich current day close-off has been characterized.

2. Firn Air Studies

[7] Studies of the composition of air in polar firn werepioneered by Schwander et al. [1993]. Samples are acquiredby drilling a hole to the shallowest sampling depth, sealingthe bottom of the hole by lowering a long rubber bladderand inflating it against the sides, and pumping air throughnylon tubing inside the bladder to sampling flasks at thesurface [Schwander et al., 1993]. We sampled firn air atGISP2 (72�360N, 38�300W) and Taylor Dome (77�47.780N,158�43.430E, 2365 m elevation) using the method ofSchwander et al. as modified by Battle et al. [1996]. GISP2is located at the Summit site, central Greenland. TaylorDome is located in the Dry Valleys region, Antarctica. InFigure 1, d15N of N2 is plotted versus depth for these twopreviously unpublished firn air profiles.[8] The d15N data allow one to define three zones of the

firn, characterized by different modes of transport [Sowerset al., 1992]. The shallowest is the convective zone, where

Figure 3. The d18Oice (per mil) for Byrd, d18Oice for GISP2, and dD (per mil) for Vostok plotted versusage. The GISP2 chronology is the standard Meese-Sowers chronology [Meese et al., 1997], and the Byrdchronology is derived by correlation to GISP2 using the CH4 stratigraphy [Brook et al., 1999; Blunierand Brook, 2001]. The Vostok chronologies are derived as explained in the text and in Table 3. TheVostok chronologies plotted here include the CH4 control point at 719 m depth and exclude the d18Oatm

control point at 778 m depth. The vertical axes on the left-hand side are Vostok dD values of ice in per mil;the vertical axes on the right-hand side are GISP2 and Byrd d18O values of ice in per mil.


3 of 10

D21115

air is rapidly mixed by convection, and chemical andisotopic gradients are absent. Below is the diffusive zone,where convection is absent and gas transport proceeds onlyby atomic/molecular diffusion. In this zone, heavier gasesand isotopes are enriched according to the barometricequation [Craig et al., 1988; Schwander, 1989]. For d15N,

d15N ¼ eDmgz=RT � 1h i

* 1000; ð1Þ

where Dm is the mass difference between 15N and 14N(0.001 kg mole�1), g = 9.80 m s�2, z = depth (meters), R =8.314 J mol�1 K�1, and T = temperature (Kelvin). The solidlines in Figure 1 show values calculated from this equation.Below the diffusive zone lies the lock-in zone. Here,alternating layers of open and closed ice preserve someopen porosity. Concentrations (or isotopic compositions)remain constant within individual layers, but open porosityallows firn air sampling. The depth where all porosity isclosed corresponds to the bottom of the lock-in zone.[9] Figure 1 shows the expected d15N increase with depth

in the nondiffusive zone (clearest at Taylor Dome), andconstant values in the lock-in zone (clearest at GISP2).GISP2 shows a shallow maximum in the upper 10 m,followed by a minimum around 20 m depth. This patternreflects thermal fractionation, in which heavy isotopes areconcentrated at the cold end of the temperature gradient inthe firn that results from warm surface temperatures duringthe summertime sampling periods [Severinghaus et al.,2001]. Because of seasonal thermal effects, the thicknessof the convective zone is best diagnosed by fitting a linewith the theoretical slope through the points in the non-diffusive zone (the slope, d15N/dz, is calculated fromequation (1)). The thickness of the convection zone is thenthe x-axis intercept of this line. The derived value dependson the choice of data points. For GISP2, it is �2 m, and forTaylor Dome, �0, according to the lines in the figure.[10] Thin or zero convection zones have been diagnosed

at most other firn air study sites. On the basis of d15N,the convection zone was deduced to be 2–4 m at H72[Kawamura, 2000], 4 m at Law Dome [Trudinger et al.,2002], �5 m at Tunu (Greenland) and Siple Dome [Butler etal., 1999], and zero at South Pole [Battle et al., 1996]. Inaddition, recent studies also using d15N find shallow con-vection zones at Dome C (2 m), BAS Depot in DronningMaud Land (<5 m), Berkner Island (2 m) [Landais et al.,2006], and North GRIP (1 m) [Kawamura et al., 2006].Thus convection is clearly minimal at most sites, and d15Nof trapped gases faithfully records the depth of gas lock-in.[11] Four sites are important exceptions to this general-

ization, and three are located in the region of slow accu-mulation on the East Antarctic Plateau. The d15N profilesshow that the convective zone is 13m thick at Vostok [Benderet al., 1994], 8–10 m thick at Dome Fuji [Kawamura,2000], and�20 m thick at the Megadunes site [Severinghauset al., 2004]. Although the convective zone is small at DomeC [Landais et al., 2006], it appears that the slow rates ofaccumulation and extremely cold temperatures of the EastAntarctic Plateau may be conducive to thick convectivezones. A surprise is the rapidly accumulating Antarctic siteYM85, where the convection zone extends to 11 m[Kawamura et al., 2006]. It is plausible that convectionwould have been deeper during the last ice age, when

temperatures were colder and accumulation was lower.Nevertheless, it is difficult to imagine convection down to40 m below surface [Caillon et al., 2003; Blunier et al.,2004], as is required to reconcile d15N and densificationestimates of the close-off depth.

3. Comparing Ice Core PaleotemperatureRecords of GISP2, Byrd, and Vostok Back to100 ka

3.1. CH4 Studies on Vostok Ice Core Samples

[12] To improve the correlation of the Vostok and GISP2gas records, we report here 103 new CH4 concentrationmeasurements made on the Vostok core by the Laboratoirede Glaciologie et Geophysique de l’Environnement. CH4

measurements were made by the method of Chappellaz etal. [1997] using an automated extraction line allowinganalysis of 11 samples at a time. The line has a lower blankcorrection than the manual extraction line, 11.6 ± 5.0 ppb.The overall analytical uncertainty (blank + chromatography)is ±10 ppb (1s). Results are summarized in Table 1. Thesamples, from 845 to 1180 m depth, span the age rangefrom about 50 to 80 ka, and greatly improve the correlationof Vostok and GISP2 in this time frame.

3.2. Correlating the Vostok and GISP2 Gas Records

[13] We use the correlations of the Vostok gas record toGISP2 based on CH4 [Blunier et al., 1998; Blunier andBrook, 2001] and d18O of O2 [Bender et al., 1999] to derivegas chronologies for Vostok. Control points for CH4 arebased on published data [Brook et al., 1996; Blunier andBrook, 2001; Petit et al., 1999; Delmotte et al., 2004]supplemented by new results reported here (Table 1). Gasage correlation points between Vostok and GISP2 (Figure 4),derived from these CH4 records, are listed in Table 2. Thed18Oatm correlation points, also listed in Table 2, are asgiven by Bender et al. [1999] with one exception. Thed18Oatm control point at 47.6 ka (788 m depth at Vostok)represents only a small minimum, and is therefore highlyuncertain. It is incompatible with the new CH4 control pointat 719 m depth (45.7 ka), and is eliminated.[14] We then use these gas chronologies to derive 6 ice

chronologies for Vostok (Table 3). For each chronology wecalculate an ice control point corresponding to each gascontrol point. We then derive a continuous timescale byinterpolating between ice control points on the basis of therelative age-depth scale of the GT4 chronology [Petit et al.,1999].[15] We derive three chronologies using the traditional

approach of calculating Dage (gas age–ice age difference)for each control points (Table 4). The first of these(Pimienta/GT4) is simply based on the GT4 ice chronology[Petit et al., 1999] and the firn densification model ofPimienta [1987] that was used to generate the originalGT4 gas chronology. The two inputs to this model are firntemperature and accumulation rate; both properties areestimated from dD of the ice, which constrains surfacetemperature. The second chronology (Goujon/GT4) is basedon the GT4 timescale, the semimechanistic firn densifica-tion model of Arnaud et al. [2000], and a thermal model ofthe firn [Goujon et al., 2003]. Firn temperatures are relevantbecause they influence the rate of snow metamorphism and,


4 of 10

D21115

hence, the close-off depth. The third chronology (Goujon/Parrenin Dage) differs from the second in that it is based onthe ice chronology of Parrenin et al. [2004]. These authorsmodify the GT4 chronology by improving the thinningfunction of ice in the context of the upstream flow, andby decreasing the glacial accumulation rates. Both changeslead to important revisions in the chronologies. For eachchronology, we calculate Dage at each gas age controldepth to get an ice age at that depth. We then interpolateages between control depths adopting the relative accumu-lation rates of the GT4 chronology.[16] The GT4 chronology is based on a model of ice flow

tuned to give ages consistent with deep sea sedimentchronologies at depths of 1534 m (110 ka) and 3254 m(390 ka). The Pimienta densification model invokes empir-ical relationships between density, close-off depth, andmean annual temperature and accumulation. The firn modelof Arnaud et al. [2000] interprets the empirical data in termsof mechanistic equations describing grain growth and den-sification. The thermal model of Goujon et al. [2003]recognizes the role of geothermal heat in warming firnand promoting densification. Finally, the ice flow modelof Parrenin et al. [2004] takes into account the thickness ofice and the nature of basal topography along the flow pathwhen evaluating how ice at Vostok is thinned with depthand age.[17] We derive three additional chronologies using a new

approach in which we calculate Ddepth (the current depthdifference between a gas control point and its correlative icedepth). In each case, we adopt a value of 0.7 for the ratio ofthe height of the fully compacted ice column (density =

Table 1. CH4 Concentrations in Newly Analyzed Samples From

the Vostok Ice Core

Depth in Vostok CH4, ppb

845.50 420.1848.88 414.8851.18 434.7854.35 476.4857.19 507.2860.37 481.7863.04 497.0866.55 455.5869.17 459.6872.77 459.6875.18 503.3878.29 520.1881.18 556.1884.04 549.7887.16 522.9890.03 532.4900.70 508.0908.49 513.6911.18 484.0914.30 458.2917.18 422.1921.23 423.8923.15 448.7926.33 448.2929.57 412.7932.13 422.7937.19 428.9940.66 449.8943.44 422.1946.70 433.4949.34 416.4952.18 414.4956.17 430.9958.33 423.0961.97 411.0964.66 417.7967.03 421.5970.18 424.8973.18 426.7976.18 436.4979.03 451.4982.58 449.1984.03 465.0987.68 425.6990.67 422.5993.19 418.8996.17 434.2999.65 435.31002.03 448.91005.03 437.71008.52 443.91011.18 446.21013.35 446.31016.15 439.11019.29 450.61022.57 470.91025.53 434.91028.13 432.51031.35 438.41033.03 454.21037.46 444.21040.19 453.61043.32 449.81046.03 462.71048.33 443.51051.58 448.21055.12 430.21057.66 457.21060.71 448.91063.03 449.71066.03 452.51070.03 450.6

Depth in Vostok CH4, ppb

1072.18 465.21075.18 464.11078.28 432.51081.17 426.91085.53 424.31087.33 419.61090.28 424.41093.18 436.01096.67 433.81099.63 426.81102.17 440.51104.63 443.21108.27 435.01111.03 435.01114.47 444.31117.17 447.01120.85 455.21123.12 444.31126.17 457.41132.03 449.61135.03 391.41138.18 401.91141.26 425.01159.68 448.61162.19 457.61165.56 460.11168.34 437.01171.16 467.11174.64 471.11177.22 476.81180.68 491.9

Table 1. (continued)


5 of 10

D21115

0.91 gm cm�3) to the original uncompacted firn, from thesurface to the close-off depth. This value of 0.7 is typical ofmodern conditions on the East Antarctic Plateau [e.g.,Herron and Langway, 1980]. This compacting factorchanges by no more than 5% for conditions in the past[Blunier et al., 2004]. The firstDdepth chronology (d15N) isbased on d15N of N2, previously measured in Vostok(available from the National Snow and Ice Data Center).The d15N values are corrected for thermal diffusion so thatthey represent the result of gravitational fractionation alone(the magnitude of this correction is order 0.02%). Surfacetemperature is taken from Petit et al. [1999]. The temper-ature gradient in the firn, which influences d15N because ofthermal fractionation [Severinghaus et al., 2001], is calcu-lated by a newly constructed thermal model similar to thatof Goujon et al. [2003]. The original close-off depth iscalculated from d15N according to equation (1), given thesurface temperature record inferred from dD of the ice. Thecurrent Ddepth value is calculated from the close-off depth,the compaction factor (0.7), and the thinning function[Parrenin et al., 2004]. The second Ddepth chronology(d15N + 13 m) is identical, except there is assumed to be aconvective zone of 13 m, the current condition at Vostok[Bender et al., 1994]. The third Ddepth chronology (Gou-jon/ParreninDdepth) calculates the close-off depth from thedensification model of Arnaud et al. [2000], the firn thermalmodel of Goujon et al. [2003], and the ice chronology(including thinning function) of Parrenin et al. [2004].[18] The Goujon/Parrenin Dage chronology invokes the

same gas control ages and gas trapping depths as theGoujon/Parrenin Ddepth chronology. These ice chronolo-gies differ, however, because they use different approachesfor deriving ice ages from each of the gas age control points.The Dage chronology adds Dage to calculate an ice age foreach gas age control point. Dage depends on accumulationrate, which may be poorly known and therefore introducesconsiderable uncertainty. The Ddepth chronology subtractsthe densified and thinned firn column from a gas depth toget the correlative ice depth. The degree of densification iswell known and, at shallow depths, the thinning is generally

small and accurately characterized. Deep in the core,however, thinning is extensive and less accurately known.[19] Vostok ice originates along the Ridge B/Vostok ice

flow line, and past variations in the ice thickness along thisflow line affect the thinning function of ice that is now at theVostok site. In the section of the Vostok core of interest here(last 100 ka), the thinning function varies between 0.8 and1.0, and has recently been improved by taking into accountbedrock topography upstream of Vostok [Parrenin et al.,2004]. Still, there remain sources of uncertainty in thismodeled thinning function. First, Parrenin et al. [2004]did not take into account the temporal variations in icethickness, which have a significant impact on the thinningfunction. Second, Parrenin et al. [2004] assumed a path forthe flow above Lake Vostok that does not coincide exactlywith the path induced by examining isochronous layers[Tikku et al., 2004]. Third, it is not clear that the trajectoryof the ice flow line has been stable in the past. Fourth, amechanical model that is more complex than used byParrenin et al. [2004], and that takes into account thedifferent mechanical properties of the ice from different

Table 2. Control Points for Correlating Vostok and GISP2 Gas

Chronologies

Source GISP2 Depth, m Vostok Depth, mGISP2 GasAge, ka

GT4 150.00 2.36CH4 1699 323.10 11.73CH4 1818.5 374.00 14.84d18Oatm 2047.00 498.00 26.60CH4 2161.00 581.80 33.67CH4 2190.00 596.45 35.17CH4 2244.00 629.20 38.52CH4 2370.00 718.85 45.55CH4 2461.00 840.15 52.43CH4 2512.00 911.18 58.32d18Oatm 2566.00 1046.00 65.45CH4 2587.00 1076.73 68.44d18Oatm 2594.00 1082.00 69.74d18Oatm 2617.00 1133.50 72.94CH4 2679.00 1261.20 84.24d18Oatm 2699.00 1327.00 88.65d18Oatm 2736.00 1422.00 97.75

Figure 4. CH4 records for GISP2 and Vostok versus age. Vertical lines indicate gas age correlationpoints based on CH4. CH4 concentrations are in units of ppb.


6 of 10

D21115

time periods, might give different results. In total, weestimate that the uncertainty in the thinning function maynot be better than ±0.1 for the upper �1500 m of the Vostokcore that is relevant to this paper. In general, the Dagemethod may be preferable for the lower part of an ice core(where the ice is more thinned and the relative uncertaintyin thinning function is higher), and the Ddepth approachmay be better for the shallower part, where there is lessthinning and use of the thinning function introduces lessuncertainty.

3.3. Intercomparison of Chronologies

[20] For the purpose of evaluating our 6 chronologies, wedefine 19 depths in the Vostok core marked by localmaxima or minima in dD (Figure 3). Seventeen of theseextrema correspond to events of the same direction andapproximate age in the Byrd core (previously correlated toGISP2 by detailed CH4 stratigraphy) [Blunier and Brook,2001], and we assume these events are synchronous. Foreach correlative event, we calculate the Vostok-Byrd agedifference according to each of our Vostok chronologies

Table 4. Ages of Gas and Ice Control Pointsa

VostokDepth, m

Vostok CorrelationGas Age,

kaVostok Ice Age,Pimienta/GT4, ka

Vostok Ice Age,Goujon/GT4, ka

Vostok Ice Age,Goujon/Parrenin DAge, ka

150.00 2.36 5.68 5.66 5.59323.10 11.73 15.30 15.81 15.96374.00 14.84 19.36 20.09 20.54498.00 26.60 33.35 32.93 33.44581.80 33.67 40.03 39.37 39.78596.45 35.17 41.29 40.75 41.14629.20 38.52 43.83 43.92 44.27718.85 45.55 50.63 50.94 51.24840.15 52.43 57.70 57.54 57.82911.18 58.32 63.22 63.55 63.751046.00 65.45 70.85 70.55 71.011076.73 68.44 73.63 73.37 73.681082.00 69.74 75.00 74.66 75.021133.50 72.94 77.90 77.65 77.931261.20 84.24 88.52 88.68 88.881327.00 88.65 93.09 92.93 93.171422.00 97.75 101.98 101.75 101.93

Vostok CorrelationDepth,m

Vostok CorrelationGas Age,

ka

Vostok Ice Depthd15N,m

Vostok Ice Depthd15N + 13 m,

m

Vostok Ice Depth,Goujon/Parrenin DDepth,

m

150.00 2.36 75.5 75.5 86.1323.10 11.73 261.1 253.4 256.2374.00 14.84 322.1 313.8 303.3498.00 26.60 447.7 440.2 424.3581.80 33.67 528.9 520.2 507.5596.45 35.17 545.0 535.9 521.9629.20 38.52 576.8 568.5 554.3718.85 45.55 673.1 665.0 642.6840.15 52.43 785.8 777.2 765.0911.18 58.32 866.3 857.6 835.81046.00 65.45 996.7 989.2 974.01076.73 68.44 1027.3 1018.7 1005.31082.00 69.74 1035.5 1026.9 1010.51133.50 72.94 1084.3 1076.0 1063.91261.20 84.24 1209.8 1202.0 1195.41327.00 88.65 1286.4 1279.2 1264.81422.00 97.75 1364.7

aUpper part of table shows chronologies based on gas age– ice age differences. Bottom part of table shows chronologies based on gas depth– ice depthdifferences. Ages are in ka (thousands of years before present).

Table 3. Models Used To Calculate Ice Ages From Gas Ages

Model Characteristics

Models Calculated by Deriving Ice Ages at Depths of Gas Age Control PointsPimienta/GT4 GT4 temperature, accumulation rate, thinning function [Petit et al., 1999];

Pimienta [1987] close-off modelGoujon/GT4 GT4 chronology; Arnaud et al. [2000] close-off model taking into effect firn temperature

as calculated with the thermal model of Goujon et al. [2003]Goujon/Parrenin Dage same as Goujon/GT4 but calculated with the thinning function of Parrenin et al. [2004]

Models Calculated by Deriving Ice Depths of Gas Age Control Pointsd15N close-off depth calculated from d15N of N2 assuming no convective zoned15N + 13 m same as d15N, except that we assume a 13 m convective zone at the top of the firnGoujon/Parrenin Ddepth close-off depth calculated as in Goujon/Parrenin model; current gas-ice depth difference

calculated from Parrenin et al. [2004] thinning function


7 of 10

D21115

(Figure 5 and Table 5). The differences, ideally zero, give ameasure of the validity of the respective chronologies andthe approaches used to derive them. They also give a usefulway to intercompare the various Vostok chronologies.[21] We start by intercomparing the three Vostok chro-

nologies in which gas age–ice age differences are calculatedfrom glaciological models. Ages of Vostok events cal-culated with the Goujon Dage values (Goujon/GT4) are,on average, slightly younger than ages calculated withDage values from the densification model of Pimienta[1987], used in the GT4 chronology of Petit et al. [1999](Pimienta/GT4). The values can differ by up to 0.8 kyr atspecific depths in the core. These differences of coursereflect the different approaches used to calculate Dage.These variations are reflected in the Vostok-Byrd age differ-ences calculated for the two chronologies. On average, thereis very little difference in ages calculated using the GT4close-off depths and those of Pimienta or Goujon. Agescalculated with the close-off depths (and accumulationrates) of Parrenin (Parrenin/Goujon) are older on averageby 0.3–0.4 kyr. The Vostok-Byrd comparison does notidentify any of these three chronologies as superior. Parre-nin/Goujon ages are older than Byrd by 0.1–0.2 kyr, whilethe other two chronologies are younger by 0.1–0.3 kyr(depending on which chronology and whether one excludesthe comparison for event 18, which is poorly dated to±0.8 kyr in Byrd). The standard deviations of Vostok-Byrddifferences are similar, falling in the range of ±0.7–0.9 kyr.[22] We next intercompare the three Vostok chronologies

derived on the basis of Ddepth differences. CalculatingDdepth from the glaciological model (Goujon/ParreninDdepth), rather than d15N (no convection zone), increasesice age by 1.5–1.6 kyr. Compared with Byrd, ages ofcorrelative events are 1.2 kyr older in Vostok when calcu-lated using d15N and assuming no convective layer. This isthe largest mean age difference between any Vostok chro-nology and Byrd, and we accordingly reject d15N alone asthe basis for calculating Dage. Vostok ages are 0.6 kyryounger than Byrd when Ddepth is calculated using d15Nand a 13 m convection depth, and 0.4 kyr older whencalculated using the Ddepth derived from the glaciological

model (Goujon/Parrenin Ddepth). Age differences showless variance with the glaciological Ddepth values (±0.5–0.6 kyr) than with any other model.[23] On the basis of this comparison, we believe that the

methods of choice for transferring gas to ice chronologiesare the Goujon/Parrenin Dage model and the Goujon/Parrenin Ddepth model. Both invoke the glaciologicalmodel of Arnaud et al. [2000], the firn air temperaturemodel of Goujon et al. [2003], and an appropriate chronol-ogy (that of Parrenin et al. [2004] in the case of Vostok).The Goujon/Parrenin Dage model includes more physicalprocesses than Pimienta/GT4 or Goujon/GT4.[24] With respect to the approaches using d15N, we judge

the one based on d15N alone as unsatisfactory, because itgives a much larger difference between ages of Vostok andByrd events than any other chronology. We downgrade theDdepth approach based on d15N and a 13 m convectionzone because there is no reason to believe that the modernconvection zone thickness has always prevailed. Our resultsimply that the convection zones at sites of deep EastAntarctic ice cores were indeed consistently of order 40 mduring glacial times, as required to rationalize the differencebetween d15N values and close-off depths estimated byglaciological models [e.g., Sowers et al., 1992].[25] The underlying physics is similar for our two selected

chronologies: each invokes the same glaciological model,thermal model, and chronology (including thinning func-tion). However, as noted above, the chronology derived forthe Dage approach is different from that for the Ddepthapproach. The reason is that Dage depends strongly onaccumulation rate and is independent of the thinning func-tion, while Ddepth is only weakly dependent on accumula-tion rate but depends strongly on the thinning function. Thesedistinctions lead to significantly different chronologies. TheDage chronology is preferable when accumulation rate isbetter constrained than the thinning function, whereas theDdepth chronology is preferable when thinning rate is betterconstrained.[26] Finally, we note that there are sometimes systematic

changes among all chronologies in Vostok-Byrd age differ-ences for some events. For example, for all chronologies,

Figure 5. Vostok-Byrd age differences at comparison points for the various chronologies. Pimienta/GT4refers to the original chronology of Petit et al. [1999]. Other chronologies are summarized in Table 3.


8 of 10

D21115

the age difference between Vostok and Byrd is greater forevent 3 than for events 2 and 4. Such systematic differencescould be due to errors in correlations of the isotopictemperature records between Byrd and Vostok, to errors inthe gas age control points established between Vostok,Byrd, and GISP2, or to errors in the interpolations betweengas age control points. These uncertainties represent funda-mental limitations to our approach and prevent us fromvalidating a delta age method to better than about ±1.0 kyr.

4. Summary and Conclusions

[27] We have presented two new firn air profiles of d15Nof N2, and used these results, together with data in theliterature, to assess the likelihood for deep convection in thefirn that would attenuate the normal gravitational enrich-ments. The 103 new analyses of CH4 in the Vostok ice core,

together with existing data from the literature, lead to a newVostok timescale for the gas records that is more accuratelycorrelated with GISP2.[28] Ice chronologies, calculated from gas chronologies

on the basis of different assumptions, lead to ages forVostok climate events that can be compared with ages ofcorrelative events in the Byrd core. We conclude that d15Nseriously underestimates the close-off depth of glacialVostok ice, presumably because of deep convection.Kawamura and Severinghaus [2005] recently discoveredthat, at the Megadunes Site, Antarctica, low d15N values areassociated with deep cracks in the ice. Their result lendsplausibility to our conclusion and suggests a mechanism fordeep convection. Our leading chronologies are both based onthe Arnaud et al. [2000] firn densification model, calculatedvalues of temperature versus depth in the firn [Goujon et al.,2003], and the Vostok accumulation rate or thinning function

Table 5. Summary of Vostok Ages and Vostok-Byrd Age Differences for Vostok-Byrd Comparison Pointsa

EventVostokDepth

Goujon/ParreninDage 15N 15N + 13 m Pimienta/GT4 Goujon/GT4

Goujon/ParreninDdepth

Ages of Events at Vostok According to Chronology1 263 11.9 11.8 12.2 11.6 11.9 12.12 284 13.3 12.8 13.2 12.8 13.2 13.53 309 15.0 14.2 14.6 14.4 14.9 15.34 356 18.8 17.6 18.3 17.8 18.4 19.35 444 27.8 26.3 27.0 27.3 27.3 28.36 498 33.4 31.0 31.7 33.4 32.9 32.97 570 39.0 37.9 38.7 39.2 38.6 39.88 595 41.1 39.9 40.4 41.2 40.7 41.79 604 41.9 40.6 41.1 41.9 41.5 42.510 632 44.5 42.6 43.2 44.0 44.1 44.711 656 46.3 44.4 44.9 45.8 46.0 46.312 704 50.1 47.5 48.0 49.5 49.8 49.013 772 54.1 51.7 52.1 53.7 53.8 53.114 844 58.1 56.8 57.4 58.0 57.8 58.715 920 64.2 61.2 61.6 63.7 64.0 62.516 1008 69.1 66.6 67.4 68.8 68.7 69.217 1066 72.8 71.8 72.3 72.7 72.4 73.218 1176 81.6 81.5 82.1 81.4 81.3 82.819 1255 88.3 86.9 87.3 88.0 88.1 88.0

Age Differences, Vostok-Byrd, kyr1 263 0.2 0.1 0.4 �0.2 0.1 0.42 284 0.0 �0.5 0.0 �0.5 �0.1 0.23 309 0.9 0.0 0.5 0.3 0.7 1.24 356 �0.2 �1.4 �0.7 �1.2 �0.6 0.37 570 0.5 �0.5 0.2 0.7 0.1 1.38 595 �0.2 �1.4 �0.8 0.0 �0.6 0.59 604 �0.3 �1.6 �1.1 �0.3 �0.7 0.310 632 0.4 �1.5 �0.9 �0.1 0.0 0.611 656 0.5 �1.5 �0.9 0.0 0.2 0.512 704 1.5 �1.1 �0.6 0.9 1.2 0.414 844 �0.4 �1.8 �1.1 �0.6 �0.7 0.216 1008 �0.8 �3.3 �2.4 �1.0 �1.2 �0.617 1066 �0.4 �1.5 �0.9 �0.5 �0.8 0.018 1176 �2.0 �2.1 �1.4 �2.1 �2.2 �0.719 1255 1.1 �0.3 0.1 0.8 0.9 0.9

All SamplesAverage deviation, kyr 0.1 �1.2 �0.6 �0.3 �0.2 0.4SD of deviation, kyr 0.9 0.9 0.8 0.8 0.9 0.6RMS of deviation, kyr 0.9 1.5 1.0 0.9 0.9 0.7

Excluding Event 18Average deviation, kyr 0.2 �1.1 �0.6 �0.1 �0.1 0.4SD of deviation, kyr 0.7 0.9 0.8 0.7 0.7 0.5RMS of deviation, kyr 0.7 1.5 1.0 0.7 0.7 0.7

aAges are in ka (thousands of years before present), and age differences are in kyr. Depths are in meters. SD, standard deviation; RMS, root-mean-square.


9 of 10

D21115

of Parrenin et al. [2004]. The uncertainty in the Vostok gasage–ice age difference is still �1 kyr, complicating anaccurate assessment of climate phasing between Greenlandand East Antarctica during the last ice age. Our analysisshould be repeated for deep East Antarctic ice cores drilledon domes. Uncertainties may be smaller at such sitesbecause, relative to Vostok, the thinning function and per-haps other glaciological properties are better constrained.

[29] Acknowledgments. Financial support to Princeton came fromthe National Science Foundation, Office of Polar Programs (AntarcticDivision), and the Gary Comer Family Foundation. CH4 measurements atLGGE were supported by the EC project Pole-Ocean-Pole (EVK2-2000-22067). Additional financial support in France was provided by theProgramme National d’Etudes de la Dynamique du Climat (INSU-CNRS)and the CEA. The University of Bern was supported by the Swiss NationalScience Foundation. We thank ice drillers and all participants in thefieldwork and ice sampling, the Russian Antarctic Expeditions (RAE),and the Institut Polaire Francais Paul Emile Victor (IPEV).

ReferencesArnaud, L., J.-M. Barnola, and P. Duval (2000), Physical modeling of thedensification of snow/firn and ice in the upper part of ice sheets, inPhysics of Ice Core Records, edited by T. Hondoh, pp. 285–305, Hok-kaido Univ. Press, Sapporo, Japan.

Battle, M., M. Bender, T. Sowers, P. P. Tans, J. H. Butler, J. W. Elkins, J. T.Ellis, T. Conway, N. Zhang, P. Lang, and A. D. Clarke (1996), Atmo-spheric gas concentrations over the past century in air from firn at theSouth Pole, Nature, 383, 231–235.

Bender, M. L., T. Sowers, J.-M. Barnola, and J. Chappellaz (1994),Changes in the O2/N2 ratio of the atmosphere during recent decadesreflected in the composition of air in the firn at Vostok Station, Antarc-tica, Geophys. Res. Lett., 21, 189–192.

Bender, M. L., B. Malaize, J. Orchardo, T. Sowers, and J. Jouzel (1999),High precision correlations of Greenland and Antarctica over the last100 kyr, in Mechanisms of Global Climate Change at Millennial Time-scales, Geophys. Monogr. Ser., vol. 112, edited by P. Clark, R. Webb, andL. Keigwin, pp. 165–175, AGU, Washington, D. C.

Blunier, T., and E. J. Brook (2001), Timing of millennial-scale climatechange in Antarctica and Greenland during the last glacial period,Science, 273, 1087–1091.

Blunier, T., et al. (1998), Asynchrony of Antarctic and Greenland climatechange during the last glacial period, Nature, 394, 739–743.

Blunier, T., J. Schwander, J. Chappellaz, F. Parrenin, and J. M. Barnola(2004), What was the surface temperature in central Antarctica during thelast glacial maximum?, Earth Planet. Sci. Lett., 218, 379–388.

Brook, E. J., T. Sowers, and J. Orchardo (1996), Rapid variations in atmo-spheric methane concentration during the past 110,000 years, Science,273, 1087–1091.

Brook, E. J., S. Harder, J. Severinghaus, and M. Bender (1999), Atmo-spheric methane and millennial-scale climate change, in Mechanisms ofGlobal Climate Change at Millennial Timescales, Geophys. Monogr. Ser.,vol. 112, edited by P. Clark, R. Webb, and L. Keigwin, pp. 165–175,AGU, Washington, D. C.

Butler, J. H., et al. (1999), A record of atmospheric halocarbons during thetwentieth century from polar firn air, Nature, 399, 749–755.

Caillon, N., J. P. Severinghaus, J. Jouzel, J.-M. Barnola, J. C. Kang, andV. Y. Lipenkov (2003), Timing of atmospheric CO2 and Antarctic tem-perature changes across termination III, Science, 299, 1728–1731.

Chappellaz, J., T. Blunier, S. Kints, A. Dallenbach, J. M. Barnola,J. Schwander, D. Raynaud, and B. Stauffer (1997), Changes in theatmospheric CH4 gradient between Greenland and Antarctica duringthe Holocene, J. Geophys. Res., 102, 15,987–15,997.

Craig, H., Y. Horibe, and T. Sowers (1988), Gravitational separation ofgases and isotopes in polar ice caps, Science, 242, 1675–1678.

Delmotte, M., et al. (2004), Atmospheric methane during the last fourglacial-interglacial cycles: Rapid changes and their link with Antarctictemperature, J. Geophys. Res. , 109 , D12104, doi:10.1029/2003JD004417.

EPICA Community Members (2004), Eight glacial cycles from an Antarc-tic ice core, Nature, 429, 623–628.

Goujon, C., et al. (2003), Modeling the densification of polar firn includingheat diffusion: Application to close-off characteristics and gas isotopicfractionation for Antarctica and Greenland sites, J. Geophys. Res.,108(D24), 4792, doi:10.1029/2002JD003319.

Herron, M. M., and C. C. Langway Jr. (1980), Firn densification: Anempirical model, J. Glaciol., 25, 373–385.

Kawamura, K. (2000), Variations of atmospheric components over the past340,000 years from Dome Fuji deep ice core, Antarctica, Ph.D. disserta-tion, Tohoku Univ., Sendai, Japan.

Kawamura, K., and J. P. Severinghaus (2005), Krypton and xenon as in-dicators of convective zone thickness in firn at Megadunes, Antarctica,Eos Trans. AGU, 86(52), Fall Meet. Suppl., Abstract PP33C-1590.

Kawamura, K., J. P. Severinghaus, S. Ishidoya, S. Sugawara, G. Hashida,H. Motoyama, Y. Fujii, S. Aoki, and T. Nakazawa (2006), Convectivemixing of air in firn at four polar sites, Earth Planet. Sci. Lett., 244, 672–682.

Landais, A., J. M. Barnola, K. Kawamura, N. Caillon, M. Delmotte, T. VanOmmen, G. Dreyfus, J. Jouzel, V. Masson-Delmotte, and B. Minster(2006), Firn-air d15N in modern polar sites and glacial-interglacial ice:A model-data mismatch during glacial periods in Antarctica?, Quat. Sci.Rev., 25, 49–62.

Meese, D., A. J. Gow, R. B. Alley, G. A. Zielinsk, P. M. Grootes, M. Ram,K. C. Taylor, P. A. Mayewski, and J. F. Bolzan (1997), The Greenland IceSheet Project 2 depth-age scale: Methods and results, J. Geophys. Res.,102, 26,411–26,423.

Parrenin, F., F. Remy, C. Ritz, M. J. Siegert, and J. Jouzel (2004), Newmodeling of the Vostok ice flow line and implication for the glaciologicalchronology of the Vostok ice core, J. Geophys. Res., 109, D20102,doi:10.1029/2004JD004561.

Petit, J. R., et al. (1999), Climate and atmospheric history of the past420,000 years from the Vostok ice core, Antarctica, Nature, 399, 429–436.

Pimienta, P. (1987), Etude du comportement mecanique des glaces poly-cristallines aux faibles contraintes: Applications aux glaces des carottespolaires, Univ. des Sci. et Technol. Med. de Grenoble, Grenoble, France.

Schwander, J. (1989), The transformation of snow to ice and the occlusionof gases, in Environmental Record in Glaciers and Ice Sheets, edited byH. Oeschger and C. C. Langway Jr., pp. 53–67, John Wiley, Hoboken,N. J.

Schwander, J., J.-M. Barnola, C. Andrie, M. Leuenberger, A. Ludin,D. Raynaud, and B. Stauffer (1993), The age of the air in the firn andthe ice at Summit, Greenland, J. Geophys. Res., 98, 2831–2838.

Severinghaus, J. P., A. Grachev, and M. Battle (2001), Thermal fractiona-tion of air in polar firn by seasonal temperature gradients, Geochem.Geophys. Geosyst., 2(7), doi:10.1029/2000GC000146.

Severinghaus, J. P., M. Fahnestock, M. Albert, and T. Scambos (2004), Dodeep convective zones exist in low-accumulation firn?, Eos Trans. AGU,84(47), Fall Meet. Suppl., Abstract C31C-07.

Sowers, T., M. Bender, and D. Raynaud (1989), Elemental and isotopiccomposition of occluded O2 and N2 in polar ice, J. Geophys. Res., 94,5137–5150.

Sowers, T., M. Bender, D. Raynaud, and Y. S. Korotkevitch (1992), Thed15N of N2 in air trapped in polar ice: A tracer of gas transport in the firnand a possible constraint on ice age–gas age differences, J. Geophys.Res., 97, 15,683–15,697.

Tikku, A. A., R. E. Bell, M. Studinger, and G. K. C. Clarke (2004), Ice flowfield over Lake Vostok, East Antarctica inferred by structure tracking,Earth Planet. Sci. Lett., 227, 249–261.

Trudinger, C. M., et al. (2002), Reconstructing atmospheric histories frommeasurements of air composition in firn, J. Geophys. Res., 107(D24),4780, doi:10.1029/2002JD002545.

��J.-M. Barnola, J. Chappellaz, and F. Parrenin, Laboratoire de Glaciologie

et Geophysique de l’Environnement (LGGE), CNRS, BP 96, F-38402 St.Martin d’Heres Cedex, France.M. L. Bender and M. Suwa, Department of Geosciences, Princeton

University, Princeton, NJ 08540, USA. ([email protected])T. Blunier and G. Floch, Climate and Environmental Physics, Physics

Institute, University of Bern, Sidlerstrasse 5, CH-3012 Bern, Switzerland.G. Dreyfus and J. Jouzel, Institut Pierre Simon Laplace/Laboratoire des

Sciences du Climat et de l’Environnement, CEA-CNRS-UVSQ, Batiment701, CE Saclay, Orme des Merisiers, F-91191 Gif-sur-Yvette, France.


10 of 10

D21115

gas age–ice age differences and the chronology of the...

Documents