Drilling and Ice-Core Treatment

Ice cores were recovered using an adapted RAND corer, powered by a portable generator (Reference Tison, Bondesan, Ronveaux and MeneghelTison and others, 1997). Individual core segments, which were typically 75 mm in diameter and 0.25–0.4 m long, were logged, bagged and stored on site in a freezer maintained at ∼−30°C. The core remained in these freezers during transport from the Alps and during storage until the ice was analyzed in a cold laboratory at ∼−18°C. Once in the cold laboratory, ice was logged and sampled using a stainless-steel band-saw. Ice was handled at all times with non-contaminating plastic gloves.

Ice classification

Ice cores were visually logged in transmitted light and classified at a length resolution of 10 mm in two stages. First, bubble density and size were recorded on scales of 0–3 (bubble-free to bubble-rich) and 1–3 (small bubbles to large bubbles), respectively (Table 1). For example, ice containing a high density of small bubbles was logged as 3/1, and bubblefree ice as 0. Secondly, ice was classified into a variety of types defined on the basis of a progression of forms designed to reveal the generation and presence of clear-facies ice in the cores (Table 2). Thus, type 1 ice is characterised by a medium or high density of small bubbles or the fine-scale intercalation of such ice with thin (≤60 mm) layers or lenses of any other ice type. Such foliation is typical of englacial ice formed by firnification, where relatively bubble-poor, blue ice layers alternate at a scale of cm with relatively bubble-rich, white ice layers (e.g. Reference HambreyHambrey, 1975; Reference Hambrey and MilnesHambrey and Milnes, 1977). The origins of these and other bubble-poor ice types are investigated further below. Ice types 2–5 are characterised by progressively lower densities of larger bubbles, while ice type 6 is bubble-free and ice type 7 is both bubble-free and debris-rich. All 90.41 m of core recovered fell into one of these seven ice-type categories.

Table 1. Bubble-density and -size classification

Table 2. Ice-type classification adopted in the study

Ice characteristics

Ice cores and ice types

Five ice cores, each extending from the ice surface to the underlying bedrock, were recovered along an approximate flowline extending away from the terminus of the glacier (Fig. 1). Details of core length and stratigraphy are given in Table 3 and Figure 2. The most notable feature of the logs presented in Figure 2 is the presence of a relatively constant thickness of ice types 4–7 at the base of each of the cores. This layer is exposed at the glacier surface in cores 1 and 2 and underlies ice types 1–3 in the longer cores 3–5. On the basis of this division, we classify each core into an upper zone (UZ), a lower zone (LZ) and a basal zone (BZ). The boundary between the UZ and the LZ is defined as the lowest location of type 1 ice in each core, and the boundary between the LZ and the BZ is defined as the highest location of debris-rich basal ice (type 7) in each core. Our definition of the LZ, therefore, is that it is debris-poor and devoid of all remnants of bubbly white ice (or ice characterised by a high density of small bubbles). In the field, its blue colour contrasts strongly with the overlying, white UZ ice. In reality, the LZ is mainly composed of ice of types 4–6. Application of this definition to the core stratigraphies presented in Figure 2 reveals that LZ and BZ thicknesses are 8–14 and 0.3–0.95 m, respectively (Table 3). The BZ, type 7 ice corresponds visually with dispersed-facies basal ice identified by Reference Hubbard and SharpHubbard and Sharp (1995). Close examination of Figure 2 also reveals that individual lengths of type 1 englacial ice rarely exceed 1 m without being interrupted by some decimetres of types 3–6 ice, even in the upper sections of the longer cores, 4 and 5.

Table 3. UZ, LZ and BZ ice thickness in each core (zones are defined in the text)

Fig. 2. Ice-type stratigraphies fo each of the five ice cores. Stratigraphic logs represent the presence of ice types 1–7 as defined in Table 2.

Ice characteristics

Major-ion concentration, stable-isotope composition and gas content and composition were determined using lengths of core that were short enough to provide high-resolution records of variability within and between individual ice types and zones. Because economic constraints limited the total number of samples that could be analyzed, sampling focused on specific core sections (particularly from core 5) and was not, therefore, continuous over the entire length of any of the cores.

Major-ion concentration

Major-ion concentrations were determined for 253 samples. Mean concentrations for all samples range from 0.78 μeq L⁻¹ Mg²⁺ to 33.75 μeq L⁻¹ HCO₃ ⁻ (Table 4). These concentration ranges are consistent with those of Reference Fairchild, Bradby, Sharp and TisonFairchild and others (1994). Ionic concentration data were also subsampled by zone, and the significance of differences between these sample sets evaluated by t-test statistics (Table 5). With the exception of debris-rich, type 7 ice, there is no evidence for significant ionic differentiation between individual ice types. As a sample group, however, debris-free LZ ice is significantly enriched in Na⁺, Mg²⁺, Ca²⁺, NO₃ ⁻ and HCO₃ ⁻ relative to the UZ ice (given by bold numbers in Table 5). Down-core plots of individual species for core 5 (Fig. 3) indicate that this enrichment is principally confined to ice (mainly types 4–6) located within a few decimetres of the glacier bed, just above the type 7, BZ ice (Fig. 2). Significantly, concentrations of crustally derived species, in particular Ca²⁺ (Fig. 4), Mg²⁺ and HCO₃ ⁻, increase approximately logarithmically between depths of ∼43.80 m (where their concentration approximates the mean of all LZ samples; Table 4; Fig. 4) and 44.40 m, the top of the (debris-rich) type 7 ice layer. This BZ, type 7 ice is, in turn, significantly enriched in K⁺, Mg²⁺, Ca²⁺ and HCO₃ ⁻ relative both to all other core samples and to LZ samples (Table 5). In accordance with Reference Fairchild, Bradby, Sharp and TisonFairchild and others (1994), the ionic composition of this debris-rich ice is dominated by crustally derived species, in particular HCO₃ ⁻, Ca²⁺ and Mg²⁺. In contrast, concentrations of atmospherically derived ions (in particular Na⁺, Cl⁻ and NO₃ ⁻) are lower in BZ ice samples (significantly lower in the case of Cl⁻ and NO₃ ⁻) than in the overlying, debris-free ice column.

Table 4. Summary of ionic chemistry data, classified by zone. Standard deviations are given in parentheses

Table 5. One-tailed student’s t-test statistic at p = 0.05 for all ionic concentration measurements

Fig. 3. Major-ion concentrations measured in core 5. The horizontal lines at ∼30.8 and ∼44.4 m mark, respectively, the boundaries between the UZ and the LZ and the LZ and the BZ.

Fig. 4. Ca²⁺ concentrations measured at the base of core 5. Note enhanced concentrations in the clean ice located at 43.844.4 m, immediately above the type 7, dispersed-facies BZ ice (shaded) The dashed line gives the mean Ca²⁺ concentration for all LZ samples. Similar patterns are evident in plots of Mg²⁺ and HCO₃ ⁻.

Stable-isotope composition

δ ¹⁸O and δD were determined for 153 samples from core 5. Mean compositions for the entire core were −12.7‰ in δ ¹⁸O and −91‰ in δD (Table 6). However, down-core variability in stable-isotope composition is apparent (Fig. 5). In particular, these data reveal two core sections that are characterised by anomalous isotopic compositions: a section of heavy-isotope depletion between ∼32 and ∼33 m depth, and heavy-isotope enrichment in the debris-rich, basal ice layer. The δ ¹⁸O composition of ice in the former narrow band is depleted by ∼1‰ relative to the mean composition of samples from the rest of the core, with the lightest sample from 31.8–33.3 m being −14.6‰ in δ ¹⁸O. The mean δ ¹⁸O composition of the type 7, BZ ice is −11.1‰, contrasting strongly with a value of 12.8‰ for the rest of the core. Corresponding patterns in both of these anomalous zones are recorded in the δD signal. There is no evidence for significant isotopic variability between the ice of the UZ and the LZ (Table 6), or for heavy-isotope enrichment in types 4–6 (LZ) ice located close to the glacier bed (Fig. 5).

Table 6. Stable-isotope summary data for core 5

Fig. 5. Stable-isotope ratios measured in core 5. The horizontal lines at ∼30.8 and ∼44.4 m mark, respectively, the boundaries between the UZ and the LZ and the LZ and the BZ.

Least-squares linear regression of bivariate co-isotopic plots of these sample groups reveals some variability in slope (Fig. 6). Samples other than those located at 31.8–33.3 m and those comprising the BZ define a slope of 6.9 (± 0.30 at P = 0.05). This is in close agreement with the co-isotopic composition of “unaltered” surface-ice samples recovered from the glacier by Reference Hubbard and SharpHubbard and Sharp (1995). In contrast, isotopically light samples recovered from 31.8–33.3 m define a slope of 6.0 (±0.49), and isotopically heavy, BZ ice samples define a co-isotopic slope of 6.3 (±1.30). Gas content and composition Total gas content was determined for 44 ice samples, of which 28 were analyzed for CO₂, O₂ and N₂ concentrations. Since ice types 6 and 7 are visually bubble-free, only four samples of the former and none of the latter were analyzed. Total gas content for all 44 samples ranges from <0.005 to 0.035 mL g_ice ⁻¹. This range is consistent with, but extends lower than, that reported in surface ice close to the terminus of Griesgletscher, Switzerland, by Reference Berner, Bucher, Oeschger and StaufferBerner and others (1977). As expected on the basis of visual classification, total gas content varies inversely with ice type, ranging from 0.028 mL g_ice ⁻¹ for type 1 ice to 0.010 mL g_ice ⁻¹ for type 5 ice and <0.005 mL g_ice ⁻¹ for all four type 6 ice samples (Table 7). Plotting total gas content, CO₂ concentration and O₂/N₂ ratio against ice type (under the assumption that ice types 6 and 7 contain no gas) reveals a strong negative relation in the case of total gas content, but weaker positive and negative relationships, respectively, for CO₂ concentration and O₂/N₂ ratio (Fig. 7).

Table 7. Summary gas data, classified by ice type

Fig. 6. Bivariate co-isotopic plot of ice samples recovered from core 5. Least-squares linear regression statistics are presented for selected sample sub-groups (r² values are given in parentheses).

Fig. 7. Bivariate plot of mean total gas content, CO₂ concentration and O₂/N₂ against ice type.

Analysis of gas characteristics along core 5 (32 samples, of which 17 were analyzed for CO₂, O₂ and N₂ concentrations) reveals systematic trends in both total gas content and gas composition (Fig. 8). As expected on the basis of core stratigraphy, total included gas content is markedly lower in the LZ and BZ than in the UZ (Fig. 8a). Typical UZ gas concentrations exceed 0.02 mL g_ice ⁻¹ (mean = 0.025 mL g_ice ⁻¹), while LZ concentrations are generally well below this value (mean = 0.010 mL g_ice ⁻¹), commonly falling below the analytical detection limit. Initial examination of Figure 8a indicates that total gas content falls steadily (at a rate of about 0.005 mL g_ice ⁻¹ per 10 m depth) throughout the entire core length, rather than in a step-like manner at the UZ/LZ transition. This effect, however, may be an artefact of our sampling strategy, from which the analysis of bubble-free ice of types 6 and 7 was largely excluded (above). Since these ice types account for a markedly higher proportion of the LZ than of the UZ (Fig. 2), their under-representation results in a statistical over-calculation of the total gas content of the LZ. With this sampling bias in mind, we have constructed a proxy record of total gas content for core 5 by matching the core’s ice-type stratigraphy (Fig. 2) to the total gas content predicted (Table 7) on the basis of the bi-variate plot presented in Figure 7. While the accuracy of this technique is limited by variability in the relationship between ice type and total gas content, it does give a valid, alternative record of the pattern of change in total gas content along the core. In contrast to the raw data, this proxy record reveals a much stronger change in total gas content across the UZ/LZ boundary (Fig. 8a).

Fig. 8. Included gas characteristics measured in core 5: (a) total gas content, including that both measured (solid dots) and reconstructedfrom ice-type stratigraphy (solid line; see text for explanation); (b) CO₂ concentration; (c) O₂/N₂. The horizontal lines at ∼30.8 and ∼44.4 m mark, respectively, the boundaries between the UZ and the LZ and the LZ and the BZ.

In contrast to total gas content, a similar reconstruction of down-core patterns of CO₂ concentration and O₂/N₂ ratio from ice-type data cannot be justified, since no type 6 or 7 gas composition data were available to plot against ice type (Fig. 7). The raw data, however, reveal that CO₂ concentration rises from ∼100 ppmv at the top of the core to almost 600 ppmv at the base (Fig. 8b), while O₂/N₂ falls from a maximum value of 0.245 near the top of the core to a minimum value of 0.085 close to the glacier bed (Fig. 8c). Although O₂/N₂ appears to fall steadily along the entire core length, we have insufficient data to determine whether this decrease is characterised by a step at the UZ/LZ boundary. Sparse sampling also makes it difficult to identify any definite trend in CO₂ concentration through the UZ (mean = 126 ppmv), although there is evidence of a rapid downward increase through the LZ. The only samples with CO₂ concentrations that are greater than atmospheric (∼360 ppmv) are located towards the base of the LZ (the highest measured CO₂ concentration, 837 ppmv, was recorded within 1 m of the base of core 4).

Near-surface processes of bubble-poor ice formation

These include surface crevasse or crack fills, and the freezing of percolating meltwater within the surface snowpack or firn layer. The latter will tend to be thin, probably on the order of 10⁻² m, forming layers or lenses that are broadly conformable with primary stratification. Crevasse fills may be slightly thicker, perhaps up to 10⁰ m, may cut across dominant stratification and foliation at a high angle and are likely to be separated from each other by at least 10⁰ m of normal englacial ice (e.g. Reference PohjolaPohjola, 1994).

Englacial processes of bubble-poor ice formation

These include gas removal by meltwater generated by deformation-related grain-boundary melting and refreezing, and the local storage and/or refreezing of gas-poor meltwaters percolating through the ice. The former process would be expected to result in the gradual formation of a massive bubble-free ice layer that may be isotopically heavy relative to its initial composition (Reference Hubbard and SharpHubbard and Sharp, 1995). In contrast, local meltwater storage and/or refreezing would affect a thinner layer of ice that may be isotopically light relative to its initial composition.

Basal processes of bubble-poor ice formation

These involve either the formation of a basal ice layer by basal adfreezing or regelation, or the pressurised transfer of basal meltwater into and through the overlying ice. Ice layers formed by the former process are likely to contain more debris than overlying englacial ice, and are unlikely to be thicker than 10⁰ m at predominantly temperate-based glaciers (Reference Hubbard and SharpHubbard and Sharp, 1989). Basal ice layers affected by the latter process may be isotopically heavy, and are predicted to be no thicker than 10⁻¹ m (e.g. Reference LliboutryLliboutry, 1993).

The geometry, location and stratigraphy of the < 60 mm thick, bubble-poor ice layers contained within type 1 ice at Glacier de Tsanfleuron strongly indicate a near-surface origin as ice layers or lenses formed from the refreezing of per-colating meltwaters. Such processes universally accompany seasonally variable firnification, resulting in the formation of bubble-foliated englacial ice (e.g. Reference HambreyHambrey, 1975; Reference Hooke and HudlestonHooke and Hudleston, 1978). In contrast, the thicker (> 60 mm) layers of bubble-poor ice present throughout the UZ show some evidence of increasing in both thickness and frequency with depth (Fig. 2). These characteristics suggest an englacial mode of formation. However, the scale and frequently repeated occurrence of these layers reduces the likelihood of formation as local zones of meltwater storage or as healed crevasses. We therefore believe these layers are formed in response to locally enhanced ice deformation and associated internal meltwater generation. They may be similar to the layers observed by Reference Raymond and HarrisonRaymond and Harrison (1975), who identified metre-long, bubble-free sections of ice core at depths of 10, 20 and 23 m at Blue Glacier, U.S.A.

LZ clear-facies ice is massive, predominantly debris-free and ∼10 m thick, precluding formation as a result of nearsurface or basal processes. The facies therefore appears to form englacially: by deformation-related meltwater generation and removal, and/or by local storage or refreezing of percolating meltwaters. The former process is favoured, since, with the exception of the ice layer located at ∼33 m depth in core 5 (discussed below), there is no evidence that clear-facies ice is isotopically lighter than overlying englacial ice. Further, the general down-core decrease in total gas content and the ratio of (more soluble) O₂ to (less soluble) N₂ in core 5 (Fig. 8c) is broadly consistent with the removal of gas in solution along triple-grain veins during such a process. Our evidence is therefore consistent with deformation-related, bubble-free ice formation as distinct, 10⁻¹ m thick layers within the UZ (perhaps increasing in scale and frequency with depth) and as a massive 10¹ m thick layer within the LZ. This interpretation implies that deformation is similarly distributed, being focused along relatively discrete planes in the UZ and operating more or less pervasively through the LZ. In this case, the stratigraphic location of the transition between the UZ and the LZ may be related to a rapid increase in ice deformation as shear strain rate increases non-linearly with depth, perhaps supplemented by a local threshold in the response of ice metamorphism to deformation rate.

Bubble-poor, dispersed-facies basal ice is isotopically heavy and located within 10⁰ m of the glacier bed. The facies is also debris-rich, merging into the overlying clear-facies ice via an increase in the spacing between, and a decrease in the thickness of, individual debris-rich bands. These properties are consistent with formation by refreezing at the basal interface, perhaps supplemented by permeating basal water flow, and have been analyzed in detail elsewhere (Reference Tison and LorrainTison and Lorrain, 1987; Reference Hubbard and SharpHubbard and Sharp, 1995).

A ∼1 m thick layer of bubble-poor ice that is isotopically light and characterised by an anomalously low co-isotopic slope is located just below the top of the clear-facies ice layer in core 5, ∼12 m above the glacier bed. Statistical testing indicates that the isotopic composition of this layer is significantly (P = 0.05) different from that of the remainder of the ice core, suggesting the influence of a specific physical process, either during the formation of this layer or during its subsequent flow through the glacier. Since this layer is isotopically distinct from adjacent clear-facies ice, an alternative or supplementary mode of formation is sought. Two explanations for the bubble-poor and isotopically light character of the layer seem feasible: formation (1) as a crevasse fill near the ice surface, and (2) by the trapping of percolating water in an englacial location. In either case, the isotopic character of the layer is consistent with a water source supplied by melting–refreezing processes within overlying ice or snow. Reference Herrmann, Lehrer and StichlerHerrmann and others (1981) added water to the upper surface of a recrys-tallising, temperate snow column and sampled the outflow as it emerged from the column’s base. The first 50% of the outflow was isotopically lighter than the initial water and snow, and was aligned on a (relatively low) co-isotopic slope of 6.3. The generation of such water is possible both at the surface of Glacier de Tsanfleuron and within the body of the glacier, in the former case from the base of the seasonal snowpack and in the latter from melting and incomplete refreezing within overlying ice. For percolating meltwater to be preferentially accumulated within this layer, the layer must be more permeable than the surrounding ice, and/or overlie ice that is relatively impermeable, creating a perched water table. It may therefore be significant that the isotopically light layer is located near the top of the LZ in core 5, since ice character (and therefore, to some extent, its permeability) changes at this location. The possibility that the isotopically light layer may be (or have been) more permeable than the overlying englacial ice is not surprising given the sharp local reduction in gas-bubble concentration across the UZ/LZ boundary. Such bubbles serve to impede inter-granular water flow both by blocking veins directly (e.g. Reference LliboutryLliboutry, 1971; Reference Raymond and HarrisonRaymond and Harrison, 1975) and, where pressurised, by freezing vein waters (Reference RaymondRaymond, 1976). These processes, however, cannot easily be invoked to explain reduced permeability immediately beneath ∼33 m, although such an effect could be related to intense ice deformation and crystallographic changes in this zone (Reference LliboutryLliboutry, 1971). Such influences, however, are currently unconstrained, pending future detailed investigations of ice structure and crystallography over the UZ/LZ transition.

The percentage mass of influent, isotopically light vein water required to produce the measured isotopic shift at ∼33 m in core 5 may be approximated from simple mass-balance considerations. Here:

(1)

where δ _f and δ _i are, respectively, the final and initial isotopic compositions of the ice layer affected and V is the mass proportion of the final ice layer composed of influent, percolating meltwater of isotopic composition δ _p. For δ ¹⁸O, our measurements indicate that δ _f = −13.47‰ and that δ _i (for UZ ice) = −12.61‰ (Table 6). We assume that the composition of δ _p corresponds to the final 5% unfrozen fraction of a liquid with an initial isotopic composition similar to that of the initial ice, δ _i (i.e. −12.61‰). Following Reference Jouzel and SouchezJouzel and Sou-chez (1982), the isotopic composition of ice (δ _s) formed by closed system freezing is:

(2)

where δ ₀ is the initial composition of the parent water, K is the equilibrium fractionation coefficient between water and ice, and f is the proportion of the parent water source that is frozen. Substituting δ ₀ = −12.61‰, K = 1.00291 (Reference Lehmann and SiegenthalerLehmann and Siegenthaler, 1991) and f = 0.95 into Equation (2) gives a bulk newly formed ice composition of −12.21‰ in δ ¹⁸O. The corresponding composition of the unfrozen liquid remaining at f = 0.95 (and thus δ _p in Equation (1)) is −20.21‰. Substituting these isotopic values into Equation (1) and solving for V indicates that ∼11% of the 1 m thick ice layer located at ∼33 m depth in core 5 needs to be composed of isotopically light influent water. Since temperate ice typically contains < 2% liquid water (e.g. Reference Raymond and HarrisonRaymond and Harrison, 1975; Reference LliboutryLliboutry, 1976), much of the postulated storage within this layer must be in the form of refrozen ice.

Most of the clear-facies (LZ) ice layer is characterised by low ionic concentrations, as is the UZ ice (Table 4). These values conflict with those measured in clear-facies ice samples recovered from basal cavities in previous studies. For example, taking Ca²⁺ as a crustally derived indicator ion, Reference HubbardHubbard (1992) recorded a mean concentration of 115 μeq L⁻¹ in clear-facies basal ice at Glacier de Tsanfleuron. Similarly, Reference Jansson, Kohler and PohjolaJansson and others (1996) reported a mean Ca²⁺ concentration of 51 μeq L⁻¹ for a visually similar basal ice layer (their “Unit C”) characterised by a mean sediment concentration of ∼0.4 g L⁻¹ (0.4% wt.) at Engabreen, Norway, which flows over schists and gneisses. In contrast, the mean LZ Ca²⁺ concentration in the present study (excluding debris-rich, type 7 ice) is 14.9 μeq L⁻¹ (Table 4). The discrepancy between the low ionic concentrations measured in clear-facies ice in the present study and the relatively high ionic concentrations measured in similar ice by Reference HubbardHubbard (1992) and possibly Reference Jansson, Kohler and PohjolaJansson and others (1996) may reflect a variety of factors. These include sample contamination in difficult conditions at the glacier bed, particularly in the earlier Tsanfleuron study where ice was melted in situ for sampling; the presence of reactive debris entrained within samples of what appeared visually to be clean ice; and, perhaps most significantly, the transport of ionically enriched vein waters upward from the debris-rich basal ice where the solute was acquired. Measurements of enhanced concentrations of crustally derived ions in visibly debris-free, clear-facies ice above the debris-rich, dispersed-facies ice layer in the present study (Fig. 4) suggest that the latter process occurs over a scale of some decimetres. We therefore infer that, while ice metamorphism throughout the entire glacier thickness (but particularly within the LZ and BZ) results in the transfer of meltwaters towards the glacier bed, an active meltwater exchange may occur within a few decimetres of the glacier bed. Where such exchange occurs, concentration gradients in water-borne characteristics, such as ionic and dissolved-gas content, will tend to be reduced. It is interesting to note in this context that the concentration of atmospherically derived species (in particular, Cl⁻ and NO₃ ⁻) may be depleted within the type 7, BZ ice (Table 5). This scale of proposed vein-water exchange (10⁻² to 10⁻¹ m) agrees closely with Reference LliboutryLliboutry’s (1993) theoretical calculations of the thickness of the basal ice layer that experiences internal regelation cycles resulting from basally induced stresses.

Implications

Our data suggest that gas is expelled from englacial ice over much of the glacier thickness, and that the effectiveness of this metamorphism is markedly increased within 10–15 m of the glacier bed, where it results in the formation of a massive layer of clear-facies ice. The physical characteristics of this clear-facies ice are consistent with metamorphism through intense and variable ice deformation and associated internal pressure melting and refreezing. This process of ice metamorphism results in the expulsion of gas and meltwater to the basal interface in a manner that is probably both temporally continuous and spatially extensive. Such a supply of gas and water may have a significant impact on the geochemistry of subglacial water flow in a distributed drainage system, since these are not diluted by large and variable surface-generated meltwater fluxes. Indeed, meltwater samples collected from the bases of boreholes drilled in areas of distributed flow at Haut Glacier d’Arolla, Switzerland, are characterised by ratios of HCO₃ ⁻ to SO₄ ⁻² of 2.2:1 (Reference TranterTranter and others, 1997). Such values suggest the presence of a local subglacial CO₂ supply, since normal coupled sulphide oxidation–carbonation reactions result in a ratio of HCO₃ ⁻ to SO₄ ²⁻ of 1:1. Gas expulsion during basal ice metamorphism may well contribute to this subglacial CO₂ supply, the magnitude of which may be approximated in a general manner from our ice-core data. Since clear-facies ice is effectively devoid of gas at its contact with dispersed-facies basal ice, the quantity of CO₂ delivered per square metre of bed area per year to the basal interface by clear-facies ice formation (ξ) is equal to the annual thickness of englacial ice that is metamorphosed to clear-facies ice at the UZ/LZ boundary (Q _ec) multiplied by the initial CO₂ concentration of that englacial ice (CO_2e):

(3)

Here, CO_2e is given by the product of the initial total gas content (standard temperature and pressure) of the englacial ice (∼0.03 ML g_ice ⁻¹ for core 5 (or, assuming an ice density of 900 kg m⁻³, 27 L m_ice ⁻³); Table 7; Fig. 8a) and the concentration of CO₂ in that gas (∼0.015% for core 5; Table 7; Fig. 8b). At any given time, Q _ec may be considered in terms of (i) the rate of increase of clear-facies ice-layer thickness as it progressively forms from the head of the glacier to the terminus (Q _ec(f)), and (ii) the maintenance of that thickness as the layer is progressively melted from its base (Q _ec(m)) (illustrated conceptually in Fig. 9). In approximating Q _ec(f), we assume that the thickness of the clear-facies ice layer increases from zero at the head of the glacier to its maximum thickness at the terminus, such that:

(4)

Fig. 9. Conceptual model of clear-facies ice formation at Glacier de Tsanfleuron used for CO₂ flux calculations facies ice growth by englacial ice metamorphism,and loss by basal melting are described in the text

where T _c is the maximum thickness of the clear-facies ice layer at the glacier terminus (∼10 m; Table 3), L is the glacier length (∼2000 m; Fig. 1) and U _b is the spatially averaged basal sliding velocity of the glacier (m a⁻¹). We take U _b to be 5 m a⁻¹ on the basis of the recorded motion of a marker placed within a cavity beneath the glacier (Reference Hubbard and HubbardHubbard and Hubbard, 1998). Our assumption that the formation of clear-facies ice begins at the head of the glacier is supported by the ice-core records of Reference Vallon, Petit and FabreVallon and others (1976) reported above. Q _ec(m) may be approximated from:

(5)

where the righthand terms are annually and areally integrated basal melt rates due to geothermal heating (G _b), frictional heating (F _b) and frictional dissipation of heat in subglacial channels (W _b). Reference PatersonPaterson (1994, p. 112) gives typical values of G _b and F _b of 6 mm a⁻¹ each, and we make the assumption that W _b above a linked-cavity-type drainage system (Reference Sharp, Tison and FierensSharp and others, 1990) brings the total to 20 mm a⁻¹ Thus Q _ec = 0.045 m a⁻¹, and Equation (3) gives a specific CO₂ flux due to clear-ice facies formation (ξ) of 0.012 L m⁻² a⁻¹ at Glacier de Tsanfleuron. The precise distribution of this supply, however, will depend on spatial and temporal variability in rates of clear-facies ice formation across the glacier bed, which is currently unconstrained. Further, this rate of CO₂ delivery may be supplemented by that generated by bacterially mediated oxidation of organic carbon within glacier ice, and in particular the debris-rich BZ (Reference Sharp, Parkes, Cragg, Fairchild, Lamb and TranterSharp and others, 1999). This process may be at least partly responsible for the very high concentrations of CO₂ recorded close to the bases of our cores (e.g. Fig. 8b), although other processes such as CO₂ release during CaCO₃ precipitation and pyrite oxidation (Reference Fairchild, Bradby, Sharp and TisonFairchild and others, 1994b) may also be significant.