ARTICLE IN PRESS
Quaternary Science Reviews 25 (2006) 49–62
Firn-air d15N in modern polar sites and glacial–interglacial ice: a model-data mismatch during glacial periods in Antarctica? A. Landaisa,b,, J.M. Barnolac, K. Kawamurad,1, N. Caillona, M. Delmottea, T. Van Ommene, G. Dreyfusa,f, J. Jouzela, V. Masson-Delmottea, B. Minstera, J. Freitagg, M. Leuenbergerh, J. Schwanderh, C. Huberh, D. Etheridgei, V. Morgane a IPSL/CEA-CNRS LSCE, Gif-sur-Yvette, France Institute of Earth Sciences, Hebrew University, Givat Ram, 93183 Jerusalem, Israel c LGGE, CNRS, 53 Rue Molie`re, 38 402 St. Martin d’Heres, France d Center for Atmospheric and Oceanic Studies, Tohoku University, Sendai 980-8578, Japan e Australian Antarctic Division and Antarctic Climate and Ecosystems Cooperative Research Centre, Private Bag 80, Hobart 7001, Australia f Department of Geosciences, Princeton University, Princeton, 08544 NJ, USA g AWI Bremerhaven, Germany h Physics Institute, University of Bern, Sidlerstrasse 5, CH-3012 Bern, Switzerland i Division of Atmospheric Research, CSIRO, Aspendale, Victoria, Australia b
Received 31 January 2005; accepted 14 June 2005
Abstract The phase lag between atmospheric composition (air bubbles) and temperature (water isotopes) can be quantified from ice cores provided that the age difference between entrapped air and the surrounding air can be correctly estimated. This difference depends on the lock-in depth (LID), when air no longer mixes with the atmosphere. The LID can be estimated from firnification models or from the air isotopic composition (d15N and d40Ar). Both methods give consistent results for Greenland and one coastal site in Antarctica (Byrd). New firn measurements in Greenland (NorthGRIP) and Antarctica (Berkner Island, BAS depot, Dome C) confirm that firnification models correctly reproduce the present LID over a large range of surface conditions. However, a systematic mismatch is observed for the Last Glacial Maximum (LGM) in East Antarctic sites (Vostok, Dome C, Dome F) questioning the model’s validity. Here we use new d15N measurements from two coastal Antarctic sites (Kohnen Station and Law Dome) providing depth estimates again distinct from firnification model calculations. We show that this discrepancy can be resolved by revising the estimate of past accumulation rates. d15N measurements can therefore help to constrain past accumulation rate and improve ice core dating. r 2005 Elsevier Ltd. All rights reserved.
1. Introduction Ice cores trap air bubbles that uniquely preserve past atmospheric composition. In addition, the water isotope composition of the ice itself provides a continuous and Corresponding author. Tel.: +1 972 2 65 84246; fax: +1 972 2 56 62581. E-mail address:
[email protected] (A. Landais). 1 Now at Scripps Institute of Oceanography, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92 037, USA.
0277-3791/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.quascirev.2005.06.007
high-resolution local temperature record. These past temperature and atmospheric composition records, however, cannot be directly compared because the air is trapped at 50–100 m depth, at the bottom of the firn where snow is transformed into ice. As a consequence, for a given depth, the air is always younger than the ice. This age difference, Dage, can be up to 7000 years during the glacial period in cold and low accumulation sites of Antarctica (e.g. Vostok, Arnaud et al., 2000). Precise determination of Dage requires a correct understanding of the firn densification and air entrapment processes.
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
50
The firn is usually divided into three different zones according to gas transport (Sowers et al., 1992). The air in the upper convective zone is mixed with atmosphere mainly by wind convection (Colbeck, 1989). Below, the air diffuses until the porosity of the firn becomes too low, at a depth defined from modelling as the lock-in depth (LID). The complete pore close-off is expected to occur a few metres below, at the bottom of a nondiffusive zone, i.e. at the close-off depth (COD). Hence, while the LID corresponds to the deepest depth where air can diffuse, the COD is defined by the depth where the complete bubble close-off physically isolates air from the atmosphere. In general, the convective and non-diffusive zones do not exceed 10 m and the diffusive zone is thus the main component of the firn. Firnification models calculate the firn density increase with depth as a result of a rearrangement of firn grains and of sintering. The COD is then defined as the depth where a critical density depending on the site temperature is reached (Martinerie et al., 1994). It mainly depends on surface conditions, i.e. accumulation rate and temperature (Schwander et al., 1997; Arnaud et al., 2000; Goujon et al., 2003). The LID is estimated from the isotopic composition (d15N, d40Ar) of stable gases trapped in the ice (Sowers et al., 1992; Schwander et al., 1993, 1997). In the absence of rapid temperature changes (Caillon et al., 2001), d15N and d40Ar values reached at the LID result from the gravitational fractionation in the diffusive zone. They can be expressed by the barometric equation d15 N ¼ d40 Ar=4 ¼
gz 1 kg mol1 , RT
where g is the gravitation acceleration, z the thickness of the diffusive column, R the gas constant, and T the
mean firn temperature (all parameters being expressed in SI). d15N or d40Ar measurements in the enclosed air relate directly to the thickness of the diffusive zone, which can be inferred from the above equation. d15N and d40Ar have been extensively measured in the air trapped in ice cores from GRIP (Schwander et al., 1997), GISP2 (Bender et al., 1994a; Severinghaus and Brook, 1999), Byrd (Sowers et al., 1992) and Vostok (Sowers, 1991; Caillon et al., 2003). Firnification models systematically predict shallower firn depths during interglacial periods than during glacial periods because of higher temperatures (despite a small opposite effect of higher accumulation rate). A data-model comparison is presented in Table 1, showing that both approaches are in broad agreement for GRIP, GISP2 and Byrd. By contrast, in East Antarctica sites characterised by extremely low temperatures and accumulation rates (around 55 1C, 2.5 cm eq ice yr1 today), d15N and d40Ar measurements reveal a thickening diffusive zone during deglaciations, whereas firn models predict a decrease of the COD. To explain such a discrepancy between models and measurements in central East Antarctica, several hypotheses have been proposed: (1) The thicknesses of the convective and non-diffusive zones vary with time. In this scenario, d15N would not be linearly related to COD. This implies that the combined thickness of the convective and nondiffusive zones must decrease from 40 to 0 m during the deglaciations (Caillon et al., 2003; Kawamura, 2000). (2) The firnification models are not appropriate for extreme LGM conditions in East-Antarctica because there is no present-day equivalent against which the empirical parameterisations of firn models could be
Table 1 Comparison of the different records of d15N evolution (increase or decrease indicated by ‘+’ or ‘’) over glacial–interglacial transitions for different polar sites and comparison with firn modelling (mainly the last deglaciation except for Vostok T3 which stands for the Termination III, 240 kyr BP) Polar sites a,b,c
Summit Byrdd Vostokd Vostok (T3)e Dome Ff Dome Cg
Model (%)
Data (%)
0.135 0.07 0.1470.04 0.1470.04 0.1470.04 0.1470.04
0.12 0.10 +0.04 +0.07 +0.06 +0.06
The error for East Antarctic sites comes from the currently accepted range of uncertainties for past temperature and accumulation rates. ‘‘Summit’’ gathers the records for GRIP and GISP2. LGM firn modelling was performed using the model of Goujon et al. (2003), which minimizes the mismatch between data and model by taking into account the heat diffusion (hypothesis 3 of the introduction). a Bender et al. (1994a). b Schwander et al. (1997). c Severinghaus and Brook (1999). d Sowers et al. (1992). e Caillon et al. (2003). f Kawamura (2000). g Dreyfus (2003).
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
calibrated. Indeed, the relationships linking COD to temperature, densification to accumulation rate and temperature cannot be extrapolated out of the current range of described polar sites. Such an hypothesis was previously raised to explain the unexpected air content evolution during glacial periods at Vostok (Martinerie et al., 1994). (3) The d15N fluctuations may result not only from gravitational settling but also from thermal fractionation which could be larger in glacial times than in interglacial times (Goujon et al., 2003; Caillon et al., 2003). In glacial times, the reduced accumulation rate alters the balance between thermal advection by mass transport and geothermal heat flux, leading to potential changes in firn temperature gradient. (4) The model-data mismatch may result from an inadequate forcing of the firn model: inaccuracies associated with the estimation of past polar temperatures and accumulation rates.
We aim here to review these different hypotheses using new sets of data and firnification modelling experiments. To explore hypothesis 1, we have compiled published (Bender et al., 1994b, Severinghaus et al., 2001, Kawamura, 2000, Huber et al., in revision a) and unpublished firn air isotopic measurements from Greenland and Antarctic sites. This database enables us to test the firn scheme for different polar sites and to quantify the differences between the modern measured diffusive column depth and the modelled COD. Hypothesis 2 is often invoked when large disagreement arises between model and data. To explore this possibility, we first check here if different firn models are really able to capture the measured COD on the large variety of available modern firn data. Secondly, two new d15N profiles spanning the last deglaciation have been obtained from the EPICA Dronning Maud Land (EDML) ice core (Graf et al., 2002; Oerter et al., 2003) and in the Dome Summit South (DSS) ice core from Law Dome (Morgan et al., 1997). The current surface conditions (EDML: 7.1 cm ice equivalent yr1, 44.6 1C; DSS: 70 cm ice eq yr1, 22 1C) combined with the available water isotopes measurements and other evidence (van Ommen et al., 2004) suggest that local glacial surface conditions at both of these sites are well within the range described by present-day sites. However, firn models fail again to reproduce the air isotopic profile and we reject hypothesis 2 because glacial conditions are not outside the region of model applicability. We therefore explore the possible influence of the firn thermal gradient (hypothesis 3) on the two d15N profiles as well as a revision of the accumulation rate scenario used for firn modelling under glacial conditions (hypothesis 4).
51
2. Firn structure 2.1. New firn air isotopic measurements Modern vertical structures of firn air isotopic composition have been intensively surveyed (e.g. Schwander et al., 1993; Bender et al., 1994b; Battle et al., 1996; Severinghaus et al., 2001). Recently, new data have been collected in the frame of the European projects FIRETRACC (FIrn REcord of Trace gases Relevant to Atmospheric Chemical Change over 100 years) and CRYOSTAT (CRYOSpheric STudies of Atmospheric Trends in stratospherically and radiatively important gases), expanding the firn data collection. d15N, d18O and d40A measurements are now performed at high precision at LSCE (Laboratoire des Sciences du Climat et de l’Environnement) using a Finnigan MAT 252 with associated analytical uncertainties, 1s, of 0.006%, 0.015% and 0.025%. This requires corrections for the influence of CO2 on d15N (through formation of CO+ with mass 28) and for that of O2/N2 on d15N and d18O (Severinghaus et al., 2001). d40Ar measurements are performed after destruction of interfering gases (N2, O2, CO2, etc.) by a Zr–Al alloy getter (Severinghaus et al., 2003). We present here modern firn air isotope measurements from Dome C, BAS depot on the Dronning Maud Land plateau, Berkner Island and NorthGRIP performed at LSCE (Table 2). The results presented in Fig. 1 depict the traditional evolution of air isotopic composition with depth (Bender et al., 1994b; Battle et al., 1996; Severinghaus et al., 2001). Because the air sampling was performed in summer, the kink in the top 30 m of the firn reflects a thermal fractionation signal arising from the seasonal temperature gradient in the firn surface. Downwards, d15N, d40Ar and d18O signals are enriched with respect to depth because of gravity. For 3 of the 4 firn profiles presented here, the shallowest sub-surface isotopic value (corresponding to a depth between 2.5 and 5 m depending on the site) differs from zero. This suggests a very shallow (i.e. less than 2.5–5 m deep) convective zone, at least during the sampling period. At BAS depot, we do not have samples covering the top of the firn, but separate measurements performed on the top 30 m of the same firn confirm that the seasonal convective zone is also less than 5 m deep (Huber et al., in revision a). The negligible convective zone observed at Dome C is surprising when compared to measurements performed at Vostok and Dome F, sites with similar surface characteristics, which indicate the presence of convective zones on the order of 10 m (Bender et al., 1994a; Kawamura, 2000). Indeed, at such sites, (i) the combination of surface wind and low accumulation rate and (ii) the relatively low surface snow density with no possibility of meltlayers (Delmotte
20.7 28.3 30.0 36.0 152.2
751100 N 421320 W 791340 S 45V420 W 751N 821W 691120 S 411070 E 661440 S 1121500 E 32 26 23 20 19
55.5 54.5 57 51 39 25 22
Mean surface temperature ( 1C)
2 o2 o15 o2 o10
12 2 9 o1 o5 o1.5 o15
Convective zone depth (m)
67 52 48 55 74
100 98 104 116 67 48 41
UD (measurements) (m)
5–11 13 10 5 12
2 3 0 7 7 8 11
Minimal nondiffusive zone depth (COD-LID) (m)
66 60 53 52 71
100 100 115 110 67 47.5 45
COD model (Arnaud et al., 2000) (m)
67 59 53 53 72
100 100 120 115 70 47 46
COD model (Barnola et al., (1991) (m)
52
The use of ‘‘o’’ signs for the convective zone indicates that no measurements were performed below that depth. We used, when possible, the direct borehole temperature measurements (~ 30 m depth, 2 years after drilling) as more representative of the mean temperature of the last centuries. For the two sites located on Law Dome, Berkner Island and Dome F, those measurements are not available; we therefore used the mean annual temperature. For H72, the temperature was estimated by a combination of 10 m depth measurements (Nishio et al., 2002). a Bender et al. (1994b). b This study. c Kawamura (2000). d Watanabe et al. (1999). e Battle et al. (1996). f Huber et al. (in revision a). g Severinghaus et al. (2001); Severinghaus, pers. comm. h Trudinger et al. (1997). i Trudinger et al. (2002).
2.4 2.7 2.3 7.6 7.6 8.7 16.3
781270 S 101510 E 751060 S 1231210 E 771190 S 391400 E 901S 001 7710210 S 101300 W 811400 S 1481460 W 661440 S 1121500 E
Vostoka Dome Cb Dome Fc,d South Polee BAS Depotb,f Siple Domeg DSSW20 K (Law Dome)h NorthGRIPb,f Berkner Islandb Devon Islandf H72c DEOB (Law Dome)i
Accumulation rate (ice eq yr1)
Location
Site
Table 2 Review of main firn characteristics and comparison with firnification modelling
ARTICLE IN PRESS
A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
δ15 N and δ40 Ar/4 (% ) 0.05
0.15
0.25
δ15 N (% )
0.35
0.45
0.55
-0.05 0
10
10
20
20
30
30
40
40 depth (m)
depth (m)
-0.05 0
50 60
80
80
90
90
0.30
0.50
0.70
0.90
18 δ O (% )
(a)
100 -0.10
1.10
0.15
0.10
0.30
0.25
0.35
0.45
0.55
0.55
0.50
0.70
0.90
1.10
-0.05
0.05
0.15
-0.1
0.1
0.3
0.25
0.35
0.45
0.55
0.5
0.7
0.9
1.1
0 10
20
20
30
30
40
40 depth (m)
depth (m)
0.15
0.45
δ15 N and δ40 Ar/4 (% )
10
50
50
60
60
70
70
80
80
90
90
100 -0.1 (c)
0.05
0.35
δ18 O (% )
(b)
δ15 N (% ) and δ40 Ar/4 (% ) -0.05 0
0.25
60 70
0.10
0.05
50
70
100 -0.10
53
100 0.1
0.3
0.5 18
δ O (% )
0.7
0.9
1.1 (d)
18
δ O (% )
Fig. 1. Firn air isotopic measurements (d15N, black circles; d18O, crosses; and d40Ar, open circles) performed on (a) Dome C, (b) BAS depot on the Dronning Maud Land plateau (without d40Ar measurements), (c) Berkner Island and (d) NorthGRIP. The scatter in surface BAS data is due to a storage problem on some bottles (those air bottles were avoided for latter sampling). For Dome C (a), we have extrapolated the line that best fits the d15N, d18O/2 and d40Ar/4 evolutions within 30 and 90 m depth (grey line). The slope obtained from the data is almost exactly (within 5%) the slope of the theoretical barometric line (d15N ¼ (gz/RT)1 kg mol1) and the intercept is 373 m. The dotted black line corresponds to d15N ¼ 0. The shaded areas stand for the non-diffusive zones.
ARTICLE IN PRESS 54
A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
et al., 1999; Albert and Shultz, 2002) suggest possible continuous removal and redeposition of the top part of the firn and wind pumping of the low-density upper part (Colbeck, 1989; Albert and Shultz, 2002) (Fig. 2). Further evidence for the absence of a convective zone at Dome C is given by the slope and intercept calculated for the points only affected by gravitational fractionation (between 30 and 90 m depth). The slope agrees with that derived from the barometric equation, and the d15N ¼ 0 intercept is for a firn depth of 373 m. For comparison, this intercept is +13 and +10 m at Vostok (Bender et al., 1994b) and Dome F (Kawamura, 2000), respectively.
High accumulation rate firn
In theory, a non-diffusive zone exists between the COD (where air can no longer be pumped) and the LID (where d15N stops increasing with depth). In this zone, impermeable layers inhibit vertical diffusion, preventing additional gravitational fractionation. The operational determination of the COD given above is approximate since it depends on the firn sampling resolution and on the probability of incorrectly identifying an impermeable layer as the COD. We therefore define a minimal non-diffusive zone from the experimental data with its deepest limit being the last sampling depth before high resistance to pumping. Over the large variety of sites presented in
Low accumulation rate firn + wind effect
wind
Upper impermeable layer COD Non-diffusive zone (middle = modeled COD) COD Close-off density
Close-off density
Fig. 2. Densification scheme for two polar sites with different surface characteristics. On the left, a high accumulation rate site (e.g. NorthGRIP, Berkner Island, etc.) depicts a variable density profile (black solid line) with an alternance of summer (low density, high accumulation, light grey) and winter (high density, low accumulation, dark grey) layers. The modelled density profile which does not account for such a seasonal layering is represented by the long dashed line for the high accumulation rate site. The first impermeable layer appears when the winter layer reaches the closeoff density and the close-off process is ended when the summer layer reaches this density. On such a site, there is a non-diffusive zone. On the right, a low accumulation rate site combined with surface wind (e.g. Vostok) has a more homogeneous surface density. The measured and modelled density profiles are identical (solid line). In this case, the non-diffusive zone is totally absent but a convective zone is observed.
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
Fig. 1, the minimal non-diffusive zone varies between 0 m in Dome F and 13 m in Berkner Island. The amplitude of this difference is significant. Indeed, in combination with a deep convective zone, a nondiffusive zone on the order of 10 m could explain some of the mismatch between the measured and the modelled d15N evolution. 2.2. Compilation of present-day firn data To further explore this hypothesis, we present a compilation of new and existing firn data spanning the full range of modern accumulation rates and temperatures at both poles, and compare the measurements with model predictions (Table 2). Model simulations were performed assuming that the mean temperature and the accumulation rate were roughly constant during the past 1000 years. This assumption is in general agreement with temperature reconstructions by Goosse et al. (2004) and van Ommen and Morgan (1997) over different polar sites showing that the range of variations is within 70.5 1C. Such orders of magnitude would be associated with a maximum error of 3% in the COD estimate when taking into account variations in both temperature and accumulation rate, this latter being related to temperature through seasonal calibrations (Van Ommen and Morgan, 1997). We use the Barnola et al. (1991) and Arnaud et al. (2000) (Fig. 3) steady-state firn models, which essentially differ by their representation of firn densification: the first one is based on semi-empirical equations, while the second one integrates the physical processes of pressure sintering. As depicted in Table 2,
55
these different approaches lead to similar COD estimate within 5%. In general, the modelled COD is in good agreement with the measured LID (Table 2). At the South Pole, the relatively large disagreement could arise from uncertainties on the surface accumulation rate (between 7 and 9 cm eq ice yr1 (Mosley-Thompson et al., 1995)). For very cold sites (Dome C, Dome F and Vostok), variations of up to 15 m in the modelled COD can be obtained when considering (1) the 2 1C uncertainty in the modern surface temperature arising from the different approaches, e.g. using the annual mean temperature or the temperature at 30 m under the ice sheet surface (Fig. 3 and caption of Table 2), and (2) the 10% uncertainty of modern surface accumulation rate estimate. For sites characterised by high accumulation rates, differences between LID estimates from d15N and modelled COD can be explained. For high accumulation rate sites (4 7 cm eq ice yr1), the firn is characterised by a seasonal layering, with density maxima for winter layers (Fig. 2). As a consequence, at the bottom of the firn, such impermeable winter layers prevent any diffusion, hence d15N increase with depth, while the air from summer layers can still be sampled: this is the nondiffusive zone. As the current firnification models do not account for such seasonal layering, the critical density that defines the modelled COD should be somewhere in the middle of the non-diffusive zone (Fig. 2, Schwander and Stauffer, 1984) and probably overestimates the LID defined by d15N. This explanation is supported by firn density profiles (Kawamura, 2000) and by data
40 H72
accumulation rate (cm eq ice.yr-1)
Devon Island
30
70 Berkner
NorthGRIP
20
50
DSSW20K 160
10
100
140
40 120 South Pole
Dome F
0 -60
Dome C
80
Siple Dome
BAS depot 60
30
Vostok
-50
-40 temperature (deg C)
-30
-20
Fig. 3. COD evolution versus accumulation rate and temperature calculated by the steady-state firnification model of Arnaud et al. (2000). Firn characteristics of current polar sites are displayed.
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
56
compiled in Table 2 pointing to the existence of a larger non-diffusive zone for sites with higher accumulation rate. However, the non-diffusive zone does not exceed 13 m and cannot explain the large discrepancy (up to 40 m) between modelled COD and d15N inferred LID during glacial periods in low accumulation rate sites (Vostok, Dome C, Dome F). Parallel to this relatively shallow non-diffusive zone, the modern convective zone seems to be restrained to a maximum of 12 m (Table 2). Notice, however, that we did not include a recent study by Severinghaus et al. (2004) describing a 20 m deep convective zone in a very particular polar site in the Megadunes near Vostok with an accumulation rate close to zero. Finally, this review of modern firn studies shows that the current firn densification models are able, within an uncertainty range around 10%, to predict the LID, with the uncertainty resulting at least partly from the existence of a non-diffusive zone. Modern firn showing significant non-diffusive zones are systematically observed without any convective zone. This means that the depth of the COD modelled by Arnaud et al. (2000) does not differ by much than 10% from the depth of the diffusive zone (Table 2). Such a model should therefore be valid to describe the d15N evolution during the deglaciation over most polar sites whose LGM characteristics are in the accumulation rate and temperature ranges depicted in present-day polar sites.
3. Analysis of two d15N profiles over the deglaciation in Antarctica We now present new d15N measurements from two coastal cores, DSS (661440 S, 1121500 E) and EDML (751000 S, 01040 E), where estimated glacial surface conditions fall within the model empirical validity range. d15N measurements were performed on the DSS ice core in 1997–1998 (Fig. 4) between 1100 and 1140 m, corresponding to the last deglaciation (Morgan et al., 2002). The d15N profile shows an increase from 0.19 % during the LGM to 0.27 % during the Early Holocene (EH). Due to a relatively large analytical uncertainty on d15N measurements at LSCE in the late 1990s (0.04%), the validity of the unexpected results was initially questioned. The EDML ice core has been recently retrieved (2002–2003) from Kohnen Station; note that this ice core drilling site is located 362 km away from the BAS depot firn drilling site on the Dronning Maud Land plateau presented previously. d15N measurements were performed with an improved analytical precision (the pooled standard deviation with two replicates at each depth is now 0.006%). Again, this precision benefits from the standard corrections to take into account the formation of CO+ (mass 28) from CO2 and the sensitivity of d15N measurements to the ratio O2/N2 (Severinghaus et al., 2001). In order to capture the last
0.45
0.40
δ15 N (‰)
0.35
0.30
0.25
0.20
0.15
measurements model (spatial slope)
0.10 1100
model (temporal slope)
1110
1120 depth (m)
1130
1140
Fig. 4. d15N measurements on the Law Dome DSS ice core (grey circles). d15N evolution from the Arnaud et al. (2000) model forced by surface conditions (accumulation rate and temperature) assumed to be modern levels for EH and deduced from the water isotopes for LGM (the black continuous line is inferred when using the spatial slope, 0.67% 1C1, and the black dotted line when using the temporal seasonal slope, 0.44% 1C1, see text). As we used the steady-state model by Arnaud et al. (2000), we simulate only one d15N mean value for the LGM and one for the EH. The modelled d15N signal between LGM and EH should therefore not be considered with high confidence.
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
deglaciation, measurements were performed from 548 to 1172 m depth (Fig. 5). Additional d18O of O2 measurements (not shown here) used for dating purposes (i.e. Sowers et al., 1991, 1993) confirm that the time period explored encompasses the deglaciation. As in DSS, the profile depicts a clear d15N increase from 0.41% during the LGM to 0.46% during the EH. A comparison of measured and modelled d15N was undertaken using the steady-state firnification model by Arnaud et al. (2000) which describes the dependence of the COD on temperature and accumulation rate (Figs. 3 and 6). In a first approximation, the EH surface conditions were assumed to be 1 1C warmer (Masson et al., 2000) than today both at DSS and at Kohnen Station. At DSS, Morgan et al. (2002) gave two estimates of the LGM temperature using the water isotopes: either through the commonly used spatial slope (Dd18Oice ¼ 0.67DT (Jouzel and Merlivat, 1984)) or through a seasonal temporal slope (Dd18Oice ¼ 0.44DT (van Ommen and Morgan, 1997)). The applicability of both the spatially and the seasonally derived calibrations at Law Dome for LGM-to-Holocene shifts is open to question, and the temperature uncertainty proposed by Morgan et al. (2002) (between 32 and 37 1C) may be underestimated. Temporal calibrations of the non-seasonal signal (annual to multi-annual averages) at Law Dome give slopes down to 0.3% 1C1 (Masson-Delmotte et al., 2003) resulting in a larger uncertainty (Fig. 6). Following Lorius et al., (1985) and Jouzel et al. (1987), we derived the LGM accumulation rate from past temperature estimates
57
through the relationship between water vapour pressure and temperature. The resulting range of LGM conditions varies between [42 1C, 20 cm eq ice yr1] and [32 1C, 40 cm eq ice yr1]. For EDML, the use of the spatial slope suggests that the glacial conditions are between [54.5 1C, 3.7 cm ice eq yr1] and [51.5 1C, 4.5 cm ice eq yr1] (see caption of Fig. 6). These LGM surface conditions enable us to model the d15NEH and the d15NLGM assuming that the COD is close to the diffusive column depth (i.e. no strong nondiffusive and convective zones). The simulated DSS d15N evolution shows a decrease from 0.38-0.52% at LGM (depending on the temperature scenario for the last glacial maximum) to 0.31% at EH (Fig. 4). The simulated EDML d15N evolution is similar with a LGM to EH decrease from 0.5–0.55 % to 0.42–0.43 % (Fig. 5). Both for DSS and EDML, the simulations show opposite LGM to EH trends to the measured d15N increases. This model-data disagreement cannot be solved by invoking only the temperature uncertainty for the LGM (Fig. 6). These two new records therefore extend the formerly presented list of East Antarctica sites where the d15N measurements contradict the firn modelling results. Such a result is very surprising since both EDML and DSS are far from being extreme sites like Vostok, Dome F and Dome C: the current temperature at Kohnen Station (EDML) is 18 1C higher, and the accumulation rate 3 times larger than at Vostok. At DSS and Kohnen Station, LGM surface characteristics are in the range covered by the current polar sites depicted in Table 2
0.50
0.48
δ15 N (‰)
0.46
0.44
0.42
0.40 measurements model
600
700
800
900
1000
1100
depth (m) Fig. 5. d15N measurements on the EDML ice core (grey circles). d15N evolution from the Arnaud et al. (2000) model forced by surface conditions (accumulation rate and temperature) assumed to be modern levels for EH and deduced from the water isotopes for LGM (black dotted line). As we used the steady-state model by Arnaud et al. (2000) in the absence of any d18Oice data available, we simulate only one mean value for d15N for the LGM and one for the EH. The modelled d15N signal between LGM and EH should therefore not be considered with high confidence.
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
58
accumulation rate (cm ice. eq. yr-1)
40
0.75
0.5
today acc. 70 cm ice eq / yr
0.65
0.95
0.7
0.9 0.85 0.8
50
0.8
acc. van Ommen
DSS LGM acc. derived from temperature
30
DSS EH 20
0.55
0.45
0.4 0.3
EDML EH acc. derived from layer counting 10
acc. derived from temperature from Dome C EDML synchronization -60
DSS LGM 0.25
from temperature
0.2
acc. van Ommen
EDML LGM
0.15 -50
-40
-30
-20
temperature (deg) Fig. 6. Modeled d15N evolution assuming that the thickness of the diffusive zone is equal to the COD modeled by Arnaud et al. (2000). We compare the modelling with our data sets for EDML and DSS over the deglaciation.
The darker grey zones indicate the d15N data range for different sites and periods assuming that the range of temperature is known. For the early Holocene (EH) on both EDML and DSS, we assume that the temperature was equal to the current surface temperature or superior by 11. For the LGM at DSS, we took a d18Oice/temperature slope between 0.3 and 0.67% 1C1 (see text) and for EDML, the LGM d18Oice/temperature was taken equal to the current spatial slope (0.77% 1C1) and a correction of d18Oice for the oceanic isotopic composition was included. We attributed a 20% uncertainty for the last temperature estimate by comparison to Vostok (Jouzel et al., 2003). The dashed zones indicate the d15N modelling for DSS and EDML during the EH and the LGM. The temperature input of the modelling is the same as for d15N data. The accumulation rate inputs were estimated through different methods: from water isotopes or from independent calculations (see text). The light grey zone indicates the d15N data corrected for a 10 m convective zone at EDML during the LGM.
and Fig. 3. As the firnification models have been empirically validated over this range of surface conditions, we argue that the processes embedded in models themselves are not questionable (hypothesis 2). Firn temperature profiles have revealed that, for very low accumulation rate sites, the geothermal flux maintains a thermal gradient in the firn inducing a thermal signal superimposed onto the gravitational signal on the d15N value (hypothesis 3). Goujon et al. (2003) quantified a 2–3 1C thermal gradient in Vostok firn during the LGM. Such a thermal effect is, however, expected to decrease the modelled d15N by less than 0.05%. For EDML, the thermal gradient in the firn is expected to be 3 times weaker than in Vostok because the accumulation rate and thus the vertical advection are 3 times larger. As a consequence, such an argument is unable to bring modelled and measured d15N into
agreement. The same argument holds true for DSS and therefore hypothesis 3 should be rejected. The 20 m difference between modelled COD and d15N inferred LID for EDML and 30 m difference for DSS still remains to be explained. Given the previous firn studies, it is almost possible to explain the EDML difference by the uncertainty range in the modelled COD and, following hypothesis 1, by the existence of large convective and non-diffusive zones (10 m+10 m). However, the estimated LGM accumulation rate for EDML is relatively small (4.170.4 cm eq ice yr1), and modern low accumulation rates are associated with a systematically small non-diffusive zone (probably o3 m due to a lack of seasonal layering at sites with low accumulation rate). We therefore rule out the existence of a 10 m LGM non-diffusive zone for EDML. A 10 m convective zone could be possible by analogy with
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
Vostok and Dome F, though the lack of any convective zone at Dome C suggests that there is no simple relationship between accumulation rate and convective zone depth. For DSS, combining the uncertainty in the modelled COD and the maximum difference observed between LID and COD on firn is insufficient to reconcile model and data. An alternative solution (hypothesis 4) for the DSS mismatch was inspired by the new accumulation rate estimates derived by van Ommen et al. (2004) combining dating constraints and an ice-flow model. The Holocene accumulation rate variability at DSS does not seem related to the fluctuations of water isotopes probably because this coastal site is affected by strong cyclonic activity, in contrast to the East Antarctic plateau, where the atmospheric moisture content is controlled by local temperature. Van Ommen et al. (2004) estimate that the accumulation rate during the EH was 40720 cm ice eq yr1 (compared to 70 cm today) and 572.5 cm ice eq yr1 during the LGM (i.e. at least 5 times less than inferred from the water isotopes). As can be seen from Fig. 6, when firn models are forced with such smaller accumulation rates, they simulate a smaller LGM d15N. Modelled and measured d15N can then be reconciled by assuming a LGM temperature range between 37 and 32 1C (Fig. 6). Our approach demonstrates that LGM d15N changes at DSS can be simulated by firn models only when low glacial accumulation rates are used, in agreement with the dating constraints on accumulation rates proposed by van Ommen et al. (2004). Given the d15N uncertainties, we conclude that LGM conditions in DSS should be associated with temperatures between 37 and 32 1C and accumulation rate between 2.5 and 6 cm eq ice yr1. EDML drilling site is located further inland than DSS and therefore should be less influenced by cyclonic activity changes at LGM and EH. Preliminary layer counting has recently been performed (A. Lambrecht, pers. comm.) suggesting that the accumulation rate inferred from the water isotopes is underestimated by 30% during the EH. Interestingly, the simulated d15NEH for EDML (0.41–0.42%) is indeed significantly below the observed EH levels (0.43–0.5%), when the model is forced by the water isotopes derived accumulation rate (Fig. 5). Using the layer counting derived accumulation rate to infer the COD results in a simulated d15NEH of 0.45% for EDML, bringing model and measurements into agreement (Fig. 6). For the LGM, two ongoing efforts from Dome C—EDML dust and volcanic peaks synchronisation suggests a 3.2–3.5 cm ice eq yr1 accumulation rate to be compared to 3.7–4.5 cm ice eq yr1 from the water isotopes (H. Fischer, pers. comm. and additional unpublished work). These lower accumulation rates reduce only slightly the disagreement between measured and modelled d15N. If
59
accumulation rate changes would be to reconcile fully modelled and measured d15NLGM, the accumulation rate should be decreased below 2.6 cm eq ice yr1 which is incompatible with other estimates. Finally, the proximity of estimated LGM surface conditions at EDML with those prevailing today at Vostok and Dome F makes it possible to propose a convective zone of 10 m at that time. Assuming such a convective zone, it is now possible to reconcile model and data with a EDML LGM temperature between 53 and 52 1C and an accumulation rate between 3.2 and 3.8 cm eq ice yr1 (Fig. 6). Revised estimates of LGM in accumulation rate do, however, not only affect modelled d15N but also affect the ice core chronology. For DSS, van Ommen et al. (2004) have already shown that the LGM accumulation rate proposed here best fits age dating constraints from a flow model and age-ties to the GRIP ice core. Still, the decrease in accumulation rate increases the Dage and then influences the DSS gas dating toward younger ages when using the gas tie points relative to GRIP. Such modifications of the DSS dating were already discussed by Morgan et al. (2002). They considered alternative high and low accumulation scenarios, obtaining Dage estimates that, while varying by up to several centuries, were still too small to modify the dating and the sequence of the millennial variability. For EDML, no complete age model has been developed until now but the current uncertainty of the LGM thinning function is of 20% (O. Rybak, personal communication). The proposed decrease of the accumulation rate from 4.1 to 3.2 cm eq ice yr1 is therefore compatible with current dating uncertainties. Moreover, while EDML modelled Dages are large during the LGM (2000 years), our proposed decrease of accumulation rate increases Dage by only 300 years. Again, detailed gas measurements (CH4 or d18O of O2) are not available in EDML to provide a chronology by gas age-ties but such a 300 years difference will have a negligible influence on the final dating and on the sequence of the millennial variability (even less than for DSS). As our hypothesis for revised LGM accumulation rate on the basis of d15N measurements is solid for DSS and EDML, it is very tempting to systematically extend it to the various East Antarctic sites where model and data depict huge disagreements. In low-temperature sites (o50 1C), the dependence of COD and d15N on accumulation rate and temperature is very important (Figs. 3 and 6) thus suggesting that a small change in accumulation rate can strongly improve the model-data agreement. For example, the recently revised Vostok timescale (Parrenin et al., 2004) now attributes an accumulation rate of 1.1 cm eq ice yr1 to the LGM, about 20% less than used for the previous GT4 agescale (Petit et al., 1999). Such a modification of the past
ARTICLE IN PRESS 60
A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
accumulation rate, in agreement with dating of Beryllium 10 profile (Jouzel et al., 1989), helps to bring model and measurements into better agreement by decreasing the modelled COD by 8 m (Fig. 3) and hence d15N by 0.05 % (i.e. one-fourth of the model-data disagreement). Further decreasing the accumulation rate in order to improve this agreement raises the problem of the chronology. In Vostok, we have only two estimates of the depth difference between contemporaneous air and ice (to be directly related to Dage providing that the age model is correct): combined to thinning function calculations, they support the water isotope derived accumulation rate (Caillon et al., 2001; Caillon et al., 2003). Such constraints limit minimum hypothetical accumulation rates for cold and low accumulation rate East Antarctica sites during glacial period unless the thinning function from glaciological models is revised accordingly (Huber et al., in revision b). Summarizing, large uncertainties still remain in the understanding of firn evolution between glacial and interglacial periods, especially for sites with low accumulation rates and temperatures. First, the size of the convective zone has been shown for one very extreme site in the Megadunes to be up to 20 m nowadays. Second, uncertainties in the thinning function can be up to 20% (Parrenin et al., 2004). Third, the semi-empirical description of the firn models still has to be improved for those sites under glacial conditions. Few constraints are available: we have very few estimates of Dage in Antarctica and d15N is probably not a faithful tracer of the past COD because of the possible existence of a convective zone on those sites. However, these firn data/model comparisons show here that the traditional way to infer past accumulation rates from water isotopes should be regarded with caution, as already shown by other indicators in different Antarctic sites, e.g. Taylor Dome (Monnin et al., 2004), Siple Dome (Taylor et al., 2004; Severinghaus et al., 2003), Vostok (Parrenin et al., 2004) and Dome C (Udisti et al., 2004). Finally, if we exclude the extreme low accumulation sites until new field and modelling studies enable us to understand the development of deep convective zones, we have shown that using d15N over LGM to EH and firnification models can usefully constrain the couple accumulation rate and temperature; it should hence be included in the ice core dating strategy. This could be achieved more continuously with improved analytical precision and automation of the measurements (Huber et al., 2003).
4. Conclusion Using new air isotopic data from the firn in a range of polar sites and from two ice cores over the deglaciation,
we reviewed the different processes that could explain the observed mismatch between modelled and measured air d15N between LGM and Holocene in East Antarctica. The understanding of firn processes is of crucial importance when studying the temporal relationship between atmospheric concentration and surface temperature. The firn review presented here first confirms that firnification models developed over the past 20 years correctly predict the different COD over the wide range of current polar sites (accumulation rates from 2.4 to 150 cm ice eq yr1 and temperature ranging from 58 to 20 1C). From a review of firn studies, we also suggest that the convective and non-diffusive zones are not expected to exceed 10 m, except for extreme glacial conditions (i.e. outside the accumulation rate and temperature ranges covered by current studies). Analysis of d15N over the deglaciation in air trapped in two Antarctic ice cores, whose glacial characteristics are within the empirically validated model range, depicts an evolution opposite to that calculated by the models when considering temperature and accumulation rate derived from the water isotopes. We show that, for these sites, a revision of the past accumulation rate, in accordance with independent estimates, allows the modelled and measured d15N to be brought into agreement. Part of the observed disagreement between firn model and d15N data could be the result of an overestimation of past accumulation rate from water isotopes. We note, however, that the disagreement between models and data at East Antarctic sites, which lack modern analogues for their glacial conditions, is too large to be resolved entirely by revising the surface parameters used to force the models because of dating constraints. Finally, our studies of Kohnen Station and Law Dome ice suggest that the air isotopic composition could be used to better constrain past ice core accumulation rate and temperature to improve the dating.
Acknowledgements We warmly thank F. Parrenin, H. Oerter, H. Fischer, J.P. Severinghaus and two anonymous reviewers for their critical readings of the paper. The firn samplings and analyses were realised in the frame of the European projects FIRETRACC and CRYOSTAT. The EDML ice core is drilled in the frame of the EPICA programme. The DSS part was supported by the Australian Government’s Cooperative Research Centres Programme through the Antarctic Climate and Ecosystems Cooperative Research Centre. We also wish to thank the CEA, the CNRS in the frame of the PNEDC and the Balzan Foundation. AL is currently funded by a Lady Davis Fellowship and the Israel Science Foundation.
ARTICLE IN PRESS A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
References Albert, M.R., Shultz, E.F., 2002. Snow and firn properties and airsnow transport processes at Summit, Greenland. Atmospheric Environment 36, 2789–2797. Arnaud, L., Barnola, J.-M., Duval, P., 2000. Physical modeling of the densification of snow/firn and ice in the upper part of polar ice sheets. In: Hondoh, T. (Ed.), Physics of Ice Core Records. Sapporo, pp. 285–305. Barnola, J.-M., Pimienta, P., Raynaud, D., Korotkevich, Y.S., 1991. CO2 climate relationship as deduced from the Vostok ice core: a reexamination based on new measurements and on a re-evaluation of the air dating. Tellus 43B (2), 83–91. Battle, M., Bender, M.L., Sowers, T., Tans, P.P., Butler, J.H., Elkins, J.W., Ellis, J.T., Conway, T., Zhang, N., Lang, P., Clarke, A.D., 1996. Atmospheric gas concentration over the past century measured in air from firn at the South Pole. Nature 383, 231–235. Bender, M.L., Sowers, T., Barnola, J.-M., Chappellaz, J., 1994a. Changes in the O2/N2 ratio of the atmosphere during recent decades reflected in the composition of air in the firn at Vostok station, Antarctica. Geophysical Research Letters 21 (3), 189–192. Bender, M.L., Sowers, T., Dickson, M.L., Orchado, J., Grootes, P., Mayewski, P.A., Meese, D.A., 1994b. Climate correlations between Greenland and Antarctica during the last 100,000 years. Nature 372, 663–666. Caillon, N., Severinghaus, J.P., Barnola, J.-M., Chappellaz, J., Jouzel, J., Parrenin, F., 2001. Estimation of temperature change and gas age–ice age difference, 108 Kyr BP, at Vostok, Antarctica. Journal of Geophysical Research 106 (D23), 31,893–31,901. Caillon, N., Severinghaus, J.P., Jouzel, J., Barnola, J.-M., Kang, J., Lipenkov, V.Y., 2003. Timing of atmospheric CO2 and Antarctic temperature changes across Termination III. Science 299, 1728–1731. Colbeck, S.C., 1989. Air movement in snow due to windpumping. Journal of Glaciology 35, 209–213. Delmotte, M., Raynaud, D., Morgan, V., Jouzel, J., 1999. Climatic and glaciological information inferred from air-content measurements of a Law Dome (East Antarctica) ice core. Journal of Glaciology 45, 255–263. Dreyfus, G., 2003. Analyse de la composition isotopique de l’air pie´ge´ dans la glace sur la dernie`re de´glaciation a` EPICA Doˆme C, Antarctique, rapport de DEA. University Pierre et Marie Curie, Paris. Goujon, C., Barnola, J.-M., Ritz, C., 2003. Modeling the densification of polar firn including heat diffusion: application to close-off characteristics and gas isotopic fractionation for Antarctica and Greenland sites. Journal of Geophysical Research 108 (D24). Goosse, H., Masson-Delmotte, V., Renssen, H., Delmotte, M., Fichetet, T., Morgan, V., van Ommen, T., Khim, B.K., Stenni, B., 2004. A late medieval warm period in the Southern Ocean as a delayed response to external forcing? Geophysical Research Letters 31. Graf, W., Oerter, H., Reinwarth, O., Stichler, W., Wilhelms, F., Miller, H., Mulvaney, R., 2002. Stable-isotope records from Dronning Maud Land, Antarctica. Annals of Glaciology 35, 195–201. Huber, C., Leuenberger, M., Zumbrunnen, O., 2003. Continuous extraction of trapped air from bubble ice or water on-line determination of isotope ratios. Analytical Chemistry 75, 2324–2332. Huber, C., Beyerle, U., Leuenberger, M., Schwander, J., Kipfer, R., Spahni, R., Severinghaus, J.P., Weiler, K., in revision a. Evidence for size dependent gas fractionation in firn air derived from noble gases, oxygen and nitrogen measurements. Earth and Planetary Sciences Letters.
61
Huber, C., Leuenberger, M., Spahni, R., Flueckiger, J., Schwander, J., Stocker, T. F., Johnsen, S., Landais, A., Jouzel, J., in revision b. Isotope calibrated Greenland temperature record over marine isotope stage 3 and its relation to CH4. Earth and Planetary Sciences Letters. Jouzel, J., Merlivat, L., 1984. Deuterium and oxygen 18 in precipitation: modeling of the isotopic effects during snow formation. Journal of Geophysical Research 89 (D7), 11,749–11,757. Jouzel, J., Lorius, C., Petit, J.-R., Genthon, C., Barkov, N.I., Kotlyakov, V.M., Petrov, V.M., 1987. Vostok ice core: a continuous isotope temperature record over the last climatic cycle (160,000 years). Nature 329 (6138), 403–408. Jouzel, J., Raisbeck, G.M., Benoist, J.P., Yiou, F., Lorius, C., Raynaud, D., Petit, J.-R., Barkov, N.I., Korotkevitch, Y.S., Kotlyakov, V.M., 1989. A comparison of deep Antarctic ice cores and their implications for climate between 65,000 and 15,000 years ago. Quaternary Research 31, 135–150. Jouzel, J., Vimeux, F., Caillon, C., Delaygue, G., Hoffmann, G., Masson, V., Parrenin, F., 2003. Temperature reconstruction from Antarctic ice cores. Journal of Geophysical Research 108 (D12). Kawamura, K., 2000. Variations of atmospheric components over the past 340 000 years from Dome Fuji deep ice core, Antarctica. Ph.D. Thesis, Tohoku University. Lorius, C., Jouzel, J., Ritz, C., Merlivat, L., Barkov, N.I., Korotkevitch, Y.S., Kotlyakov, V.M., 1985. A 150,000-year climatic record from Antarctic ice. Nature 316, 591–596. Martinerie, P., Lipenkov, V.Y., Raynaud, D., Chappellaz, J., Barkov, N.I., Lorius, C., 1994. Air content paleo record in the Vostok ice core (Antarctica): a mixed record of climatic and glaciological parameters. Journal of Geophysical Research 99, 10,565–10,576. Masson, V., Vimeux, F., Jouzel, J., Morgan, V., Delmotte, M., Ciais, P., Hammer, C., Johnsen, S., Lipenkov, V.Y., Mosley-Thompson, E., Petit, J.-R., Steig, E.J., Stievenard, M., Vaikmae, R., 2000. Holocene climate variability in Antarctica based on 11 ice-core isotopic records. Quaternary Research 54, 348–358. Masson-Delmotte, V., Delmotte, M., Morgan, V.I., Etheridge, D., van Ommen, V., Tartarin, T., Hoffmann, G., 2003. Recent Southern Indian Ocean climate variability inferred from a Law Dome ice core: new insights for the interpretation of coastal Antarctic isotopic records. Climate Dynamics 21, 153–166. Monnin, E., Steig, E.J., Siegenthaler, U., Kawamura, K., Schwander, J., Stauffer, B., Stocker, T.F., Morse, D., Barnola, J.-M., Bellier, B., Raynaud, D., Fischer, H., 2004. Evidence for substantial accumulation rate variability in Antarctica during the Holocene, through synchronization of CO2 in the Taylor Dome, Dome C and DML ice cores. Earth and Planetary Sciences Letters 224, 45–54. Morgan, V.I., Wookey, C.W., Li, J., van Ommen, T.D., Skinner, W., Fitzpatrick, M.F., 1997. Site information and initial results from deep ice drilling on Law Dome, Antarctica. Journal of Glaciology 43, 3–10. Morgan, V., Delmotte, M., van Ommen, T., Jouzel, J., Chappellaz, J., Woon, S., Masson-Delmotte, V., Raynaud, D., 2002. Relative timing of deglacial events in Antarctica and Greenland. Science 297, 1862–1864. Mosley-Thompson, E., Thompson, L.G., Paskievitch, J.F., Pourchet, M., Gow, A.J., Davis, M.E., Kleinman, J., 1995. South Pole snow accumulation has increased in recent decades. Annals of Glaciology 21, 182–188. Nishio, F., Furukawa, T., Hashida, G., Igarashi, M., Kameda, T., Kohno, M., Motoyama, H., Naoki, K., Satow, K., Suzuki, K., Takata, M., Toyama, Y., Yamada, T., Watanabe, O., 2002. Annual-layer determinations and 167 year records of past climate of H72 ice core in east Dronning Maud Land, Antarctica. Annals of Glaciology 35, 471–479.
ARTICLE IN PRESS 62
A. Landais et al. / Quaternary Science Reviews 25 (2006) 49–62
Oerter, H., Graf, W., Meyer, H., Wilhelms, F., 2003. The EPICA ice core from Dronning Maud Land: first results from stable isotope measurements. Annals of Glaciology 39, 307–312. Parrenin, F., Re´my, F., Ritz, C., Siegert, M.J., Jouzel, J., 2004. New modelling of the Vostok ice flow line and implication for the glaciological chronology of the Vostok ice core. Journal of Geophysical Research 109. Petit, J.-R., Jouzel, J., Raynaud, D., Barkov, N.I., Barnola, J.-M., Basile, I., Bender, M.L., Chappellaz, J., Davis, M., Delaygue, G., Delmotte, M., Kotlyakov, V.M., Lorius, C., Pepin, L., Ritz, C., Saltzman, E., Stievenard, M., 1999. Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica. Nature 399, 429–436. Schwander, J., Stauffer, B., 1984. Age difference between polar ice and the air trapped in its bubbles. Nature 311, 45–47. Schwander, J., Barnola, J.M., Andrie´, C., Leuenberger, M., Ludin, A., Raynaud, D., Stauffer, B., 1993. Age scale of the air in the summit ice: Implication for glacial–interglacial temperature change. Journal of Geophysical Research 98, 2831–2838. Schwander, J., Sowers, T., Barnola, J.-M., Blunier, T., Fuchs, A., Malaize´, B., 1997. Age scale of the air in the summit ice: implication for glacial–interglacial temperature change. Journal of Geophysical Research 102 (D16), 19,483–19,493. Severinghaus, J.P., Brook, J., 1999. Abrupt climate change at the end of the last glacial period inferred from trapped air in polar ice. Science 286 (5441), 930–934. Severinghaus, J.P., Grachev, A., Battle, M., 2001. Thermal fractionation of air in polar firn by seasonal temperature gradients. Geochemistry, Geophysics, Geosystems, vol. 2, Paper No. 2000GC000146. Severinghaus, J.P., Grachev, A., Luz, B., Caillon, N., 2003. A method for precise measurement of argon 40/36 and krypton/argon ratios in trapped air in polar ice with applications to past firn thickness and abrupt climate change in Greenland and at Siple Dome, Antarctica. Geochemica et Cosmochemica Acta 67, 325–343. Severinghaus, J.P., Fahnestock, M., Albert, M., Scambos, T., Shuman, C., 2004. Do deep convective zones occur in low accumulation firn? AGU Fall Meeting, C31C-07. Sowers, T.A., 1991. The isotopic composition of trapped O2 and N2 in ice cores: Biogeochemical and stratigraphic implications. Ph.D. Thesis, University of Rhode Island.
Sowers, T., Bender, M.L., Raynaud, D., Korotkevich, Y.S., Orchado, J., 1991. The d18O of atmospheric O2 from air inclusions in the Vostok ice core: timing of CO2 and ice volume change during the penultimate deglaciation. Paleoceanography 6 (6), 669–696. Sowers, T., Bender, M.L., Raynaud, D., Korotkevich, Y.S., 1992. 15N of N2 in air trapped in polar ice: a tracer of gas transport in the firn and a possible constraint on ice age–gas age differences. Journal of Geophysical Research 97, 15,683–15,697. Sowers, T., Bender, M.L., Labeyrie, L., Martinson, D.G., Jouzel, J., Raynaud, D., Korotkevich, Y.S., 1993. A 135,000 year VostokSPECMAP common temporal framework. Paleoceanography 8, 737–766. Taylor, K.C., White, J.W., Severinghaus, J.P., Brook, E.G., Mayewski, P.A., Alley, R.G., Steig, E.J., Spencer, M.K., Meyerson, E., Meese, D.A., Lamorey, G.W., Grachev, A., Gow, A.J., Barnett, B.A., 2004. Abrupt climate change around 22 ka on the Siple Coast of Antarctica. Quaternary Science Reviews 23, 7–15. Trudinger, C.M., Enting, I.G., Etheridge, D.M., Francey, R.J., Levchenko, V.A., Steele, L.P., Raynaud, D., Arnaud, L., 1997. Modeling air movement and bubble trapping in firn. Journal of Geophysical Research 102 (D6), 6747–6763. Trudinger, C.M., Etheridge, D.M., Rayner, P.J., Enting, I.G., Sturrock, G.A., Langenfelds, R.L., 2002. Reconstructing atmospheric histories from measurements of air composition in firn. Journal of Geophysical Research 107 (D24), 4780–4793. Udisti, R., Becagli, S., Castellano, E., Mulvaney, R., Schwander, J., Torcini, S., Wolff, E., 2004. Holocene electrical and chemical measurements from the EPICA-Dome C ice core. Annals of Glaciology 30, 20–26. van Ommen, T.D., Morgan, V.I., 1997. Calibrating the ice core paleothermometer using seasonality. Journal of Geophysical Research 102 (D8), 9351–9357. van Ommen, T.V., Morgan, V.I., Curran, M.A.J., 2004. Deglacial and Holocene changes in accumulation at Law Dome. Annals of Glaciology 39, 39A034. Watanabe, O., Kamiyama, K., Motoyama, H., Fujii, Y., Shoji, H., Satow, K., 1999. The paleoclimate record in the ice core at Dome Fuji station, East Antarctica. Annals of Glaciology 29, 176–178.
