Proposing a stratigraphy based on the CO2–CH4 relationship

Figure 2 shows the concentration of CO₂ vs the CH₄ concentration for the samples from the silty ice of the Dye 3 core. on this diagram, four different relationships are displayed. By contrast, a unique relationship was obtained by Reference Souchez, Lemmens and ChappellazSouchez and others (1995) in the silty ice of the GRIP ice core (crosses and dotted line in Fig. 2). the measured values fall along distinct straight lines with striking differences in slopes and concentration ranges, depending on the sample’s location in the core. Transitions from samples on one trend line to those on another coincide with clear-cut visual contrasts in debris load along the core (Fig. 3). These relationships lead us to define four corresponding units.

Fig. 2

The CO₂–CH₄ relationships in the basal part of the Dye 3 core. Meteoric-ice samples (unit 1: open triangles; sample Nos. 1–3), unit 2 samples (circles, 4–22), unit 3 samples (dotted circles, 23–25 and 32–35) and unit 4 samples (diamonds, 26–31) are represented with their respective regression lines. Crosses represent the basal ice from the GRIP ice core (from Reference Souchez, Lemmens and ChappellazSouchez and others, 1995). Inset: enlarged diagram of the lowest-concentration area.

In Figure 2, open triangles represent the three debris-free meteoric-ice samples forming unit 1 (samples 1–3 in Fig. 1a). This unit corresponds to a cluster of points. the CO₂ and CH₄ concentrations of ice samples from the upper part of the Dye 3 silty ice (unit 2: samples 4–22 in Fig. 1a; black circles in Fig. 2) show a linear relationship. the range of concentration values is small and the correlation strength moderate (r² =0.67 However, the regression line intersects the group of points representing the debris-free meteoric-ice samples (open triangles of unit 1in Fig.2), and the correlation strength increases to 0.83 when these samples are included. Another linear relationship (r² = 0.90) with smaller slope and larger concentration range, is inferred for the debris-richer ice samples from the lower part of the Dye 3 silty ice (unit 3: samples 23–25 and 32–35 in Fig. 1a; dotted circles in Fig. 2). Finally, yet another strong linear relationship (r² = 0.99)characterizes a 2 m thick layer (unit 4: samples 26–31 in Fig. 1a; diamonds in Fig. 2) with distinctively lower debris content, embedded in unit 3.

Two important features need to be pointed out from the trend lines in Figure 2, that will be useful in further interpreting the stratigraphy:

(a) the gross similarity between the slopes (s) of the regression lines obtained for the samples in units 2 and 4, and also for the GRIP core samples (s =15–21), is in strong contrast with the much lower value (s = 2.61) obtained for the samples from unit 3.

(b) At higher resolution (inset of Fig. 2), there is a clear contrast in behaviour between samples from units 2 and 4, the latter showing a distinctly stronger correlation (0.99 vs 0.67), a slightly higher slope of the trend line (20.27 vs 15.20) and a higher independent term (1453.9 vs 332.7 ppmv).

Coherence with the previous ``stable-isotope/gas’’ stratigraphy

Reference Souchez, Bouzette, Clausen, Johnsen and JouzelSouchez and others (1998) defined three major units from the variations of δ¹⁸O with depth (Fig. 1b): units 1 (2008.6– 2012.6 m) and 3 (2030–2037.8m) both showing surprisingly constant values, and unit 2 (2012.6–2030m) with δ values characterized by large and rapid variations. They showed that tight links do exist between the δ¹⁸O and the gas properties in the silty ice (Fig.1c). Their main statements are the following: unit 1 shows gas and isotopic properties of ``ice-sheet’’ ice. In unit 2 (Fig.1b and c), both δ¹⁸O and gas parameter (CO₂,O₂ / N₂ and total gas content) are correlated, which is a strong argument in favour of a mixing process in different proportions between two end-members of fixed isotopic and gas characteristics. This correlation between δ¹⁸O and gas parameters is negative or positive depending on the relative composition of each of the end-members involved in the mixing. Finally, in unit 3 (Fig. 1b and c), no correlation is found between δ¹⁸O and gas parameters, but a strong correlation exists between the latter, which does not preclude mixing, if the two end-members have similar δ¹⁸O values. the interpretation given by Reference Souchez, Bouzette, Clausen, Johnsen and JouzelSouchez and others (1998) is that unit 3 results from the mixing of two end-members with strongly contrasted gas properties and weakly contrasted isotopic signature, whilst unit 2 results from mixing between the bottom of unit 1 and the top of unit 3.

The increased resolution of the CO₂ measurements and the comparison with the CH₄ record allows interesting tuning of this initial stratigraphy. There are two significant discrepancies: a shift in the limit between units 2 and 3 downward by about 1.5 m, and the presence of a fourth unit (Fig. 1a and b). Samples 19–22 correspond to those considered by Reference Souchez, Bouzette, Clausen, Johnsen and JouzelSouchez and others (1998) as the top of unit 3 on the basis of the δ¹⁸O values. Since these samples show CH₄ and CO₂ levels and CO₂/CH₄ ratios typical of unit 2 (Fig.2) and very distinct from unit 3, they should be a part of unit 2. This does not contradict the data interpretation from 1998 since:

(a) High values of δ¹⁸O with low variability exist higher up in unit 2 (e.g. 2025.5–2027m; Fig. 1b), and

(b) Samples19–22 are at the hinge between unit 2 and unit 3 curves in figure 3b and c of Reference Souchez, Bouzette, Clausen, Johnsen and JouzelSouchez and others (1998), and can therefore be attributed to either of those two.

Samples 23–35 are all part of unit 3 as defined by Reference Souchez, Bouzette, Clausen, Johnsen and JouzelSouchez and others (1998). In this unit, characterized by a homogeneous isotopic profile, only the gas properties can be used as tracers of possible mixing processes. However, looking at the CO₂–CH₄ relationship (Fig. 2), unit 3 of Reference Souchez, Bouzette, Clausen, Johnsen and JouzelSouchez and others (1998) is clearly made up of two groups of samples:

Samples 23–25 and 32–35 showing a very strong CO₂– CH₄ correlation (0.90) with a slope of 2.61, and generally higher CH₄ and CO₂ concentrations, and

Samples 26–31 showing low CH₄ and CO₂ concentrations and a strong correlation (0.99) but with a much higher slope of 20.27, close to that of the best-fit line of the GRIP samples (20.96) and to that defined for unit 2 (15.20).

This has led us to distinguish unit 4, embedded into unit 3. Finally, it should be stressed that visual examination and discrete measurements (Reference ThorsteinssonThorsteinsson,1990) of the dirt content in the core (Fig. 3) clearly lend further support to this new stratigraphical approach.

Build-up of the basal-ice sequence

The contrasted linear relationships of CO₂ and CH₄ in Figure 2 are thought to be indicative of distinct mixing events between different end-members (i.e. between ice of local origin and ice-sheet ice flowing from upstream), leading to the progressive build-up of 25m thick basal ice at Dye 3, which has taken place along the basal flowline up-glacier from the drill site.

Any attempt to reconstruct this progressive build-up will have to integrate the following statements:

(a) Unit 1 represents debris-free meteoric glacier ice.

(b) Build-up of unit 2 must be the first step (and thus the oldest) of the sequence, since unit 1is one of its end-members. Indeed, the regression line of unit 2 is the only one which intersects the cluster of points representing the meteoric-ice samples (unit 1).

(c) Unit 2 shows the weakest regression coefficient along the trend line, indicative of dispersion of the samples around the mixing lines.

(d) the lower end-member at the origin of unit 2 shows a CO₂/CH₄ ratio that is similar to that of GRIP. Although the nature of the various chemical processes leading to this CO₂/CH₄ balance in the ice is still a major unknown, such values of the ratio could be indicative of similar sub-or pre-ice-sheet environmental conditions. In the case of GRIP, several arguments (see, e.g., Reference Souchez, Lemmens and ChappellazSouchez and others, 1995; Reference SouchezSouchez, 1996; Reference Tison, Souchez, Wolff, Moore, Legrand and de AngelisTison and others, 1998) are in favour of biologically mediated redox reactions with organic matter, drastically increasing the CO₂ concentration relative to the CH₄ in the local end-member. the observed methane concentration in this local end-member could result either from local bedrock gas diffusion or from biological processes.

(e) Unit 3, being the lowest, is likely to be the youngest, and built from interactions with the local end-member in the vicinity of the Dye 3 ice–bedrock interface. This local end-member obviously has a much lower CO₂ content for equivalent CH₄ concentration, because the slope of the unit 3 regression line is smaller than the others. This could be explained by a different balance between bedrock gas diffusion and biological processes in the local end-member.

(f) Samples from unit 4, although close to those of unit 2 (Fig. 2), show a much stronger CO₂/CH₄ correlation. They also systematically have higher CO₂ concentration than unit 2 and show a somewhat higher dirt content (0.05 weight per cent vs a mean value of 0.03 for unit 2 (Reference ThorsteinssonThorsteinsson, 1990)). the slope of the CO₂/CH₄ regression line is nearly identical to those at GRIP and in unit 2.

Bearing these characteristics in mind, we propose two plausible build-up scenarios, as schematically illustrated in Figure 4. A first step, common to both scenarios, occurs close to the central part of the ice sheet. Mixing interaction between the ice-sheet ice and a local ice component at the ice–bedrock interface builds a complete unit 2, resulting in a profile similar to that observed at GRIP, for example. Whilst the lower part of this unit remains in a local low-bedrock area, decoupling occurs and the upper part of unit 2 is entrained further downstream. Deformation around subsequent bumps is likely to result in the vertical thickening of unit 2 by further incorporation and mixing with higher levels from unit 1. Dilution of the signal (i.e. gas concentration) and hence dispersion of the sample points around the mixing trend line occurs along the flowline, eventually combined with small-scale diffusion and chemical reactions within the ice. Further downstream, in the next suitable area, the process is repeated twice to integrate either unit 4 and then unit 3 (Fig. 4A) or unit 3 and then unit 4 (Fig. 4B). In both alternatives, shearing and folding associated with an irregular bed are needed to explain the observed stratigraphy.

Fig. 4

Schematic sketch of the progressive build-up of basal ice through time and space, along a streamline that passes by the Dye 3 core. Depths of boundaries between units are arbitrary. Dashed line stands for a given tracer (e.g. gas concentration).

Our experience from the GRIP core shows that, in the course of a single flow-induced mixing event, the general trend is towards higher concentration of the gas tracers closer to the bedrock. This lends some support to scenario A, since in scenario B, unit 4, which is the youngest in the sequence and the closest to the bedrock, shows low concentration levels of the order of magnitude of those encountered in the upper part of unit 2, and in the top metre of the GRIP basal-ice sequence. Unfortunately, the mixing trend lines of Figure 2 do not bring any further decisive argument since the top end-member of unit 3 could undiscernibly be issued from bottom layers of unit 2 (B scenario) or unit 4 (A scenario). It is interesting to note, in this regard, that in both cases, the best candidates for the upper end-member of unit 3 (point A or B in Fig. 2 depending on the scenario) are not observed in the profile. Furthermore, our high-resolution record lacks sample points showing mixing signatures close to these end-members. This suggests important lamination of the bottom of units 2 and 4 after the building of unit 3. the same process probably explains that no sample with unit 4 characteristics is detected on the top of unit 3 (note, however, that sample 22, located at the hinge, is intermediate between units 2 and 4 in the insert of Figure 2).

In both scenarios, it is clear that similar heavier isotopic values, not related to climate, characterize the local end-member of all mixing processes. Finally, if scenario A applies, it appears (Fig. 4) that there might be a strong contrast in the environmental conditions characterizing the local end-members of the mixing processes between the central and the marginal parts of the ice sheet. These conditions result from the balance between gas diffusion from bedrock and the effective set of redox reactions occurring in the ice, which strongly depends on the local type and availability of the oxidizer (O₂, NO₃ ^–, H₂O, SO₄ ^2–, fermentation).