Englacial diffused facies (Ed)

This facies comprised the bulk of the glacier and consisted of strongly bubble-foliated, debris-poor ice, with occasional thin layers of clear, blue ice and fine-grained ice. Ice crystals reached 8 cm in diameter. Debris occurred as dispersed aggregates which were sometimes concentrated along poorly defined planes sub-parallel to the bubble foliation. Debris concentrations averaged 0.3% by volume. Electrical conductivities averaged 26.6 µS cm⁻¹, somewhat higher than values measured on overlying snow (mean 9.7µScm⁻¹) (Table 1). This difference can be explained in terms of the greater leaching by percolating meltwater from snow deposited within the ablation zone of the glacier than from snow deposited at higher elevations (the presumed origin of glacier ice). Debris consisted primarily of coarse to medium silt (5.81 ± 1.19Φ; 84%), with additional minor modes in the coarse (10%) and fine (6%) sand-size ranges (Fig. 3). The combined evidence of fine mean grain-size, high degree of sorting and the dispersed distribution of the debris in the ice suggests an aeolian origin for the primary mode. The coarse-sand mode included angular lithic fragments which were presumably derived from the valley walls. The absence of coarser material of similar provenance in samples from this facies can be explained in terms of the relatively small size of ice sample collected for analysis.

Fig. 3.

Results of Gaussian component analysis of particle-size distributions of samples of debris collected from ice facies at Variegated Glacier: a. Standard deviation versus mean size for each component recognized; b. Proportion of the sample falling into each component as a function of the mean size for that component.

Englacial banded facies (Eb)

This facies consisted of debris-rich layers of clear ice embedded within longitudinally foliated ice of the englacial diffused facies, which could be traced for long distances across the glacier surface. Debris was commonly derived from a single rock type and was coarse-grained and predominantly angular in shape. The incorporated rock types varied systematically across the glacier in a manner consistent with their pattern of outcrop on the valley walls. This suggests that the debris was initially deposited on the glacier surface in the accumulation area, buried by winter snowfall and transported down-glacier in an englacial position before being brought back to the surface by ablation. The debris bands may therefore represent accumulations of debris on summer melt surfaces or material deposited on to the glacier by individual mass-movement events.

Basal stratified facies — laminated sub-facies (Bs₁)

The most widespread component of the basal stratified facies consisted of repeated laminations of clear and debris-rich ice (Fig. 4). In general, the laminae were flat-lying and sub-parallel to the glacier bed (cf. foliation S_b of Reference PfefferPfeffer (1992)) but they were locally folded. The orientation of the laminae was discordant with that of foliation in overlying Ed-facies ice. The clear ice layers were up to 2 cm thick and were composed of crystals generally less than 0.3 cm in diameter. Locally, however, crystals up to 3 cm in diameter cross-cut the lamination. Such crystals were presumably a product of secondary recrystallization. The debris-rich laminae were up to 0.3 cm thick and were composed of ice crystals up to 0.1 cm in diameter, which were commonly elongated perpendicular to the lamination. Air bubbles were rare within this facies but they could be concentrated along planes parallel to the lamination which lay immediately above the upper surface of debris-rich layers. Debris concentrations, measured on samples composed of several clear and debris-rich laminae, averaged 10.9% by volume, while electrical conductivities averaged 44 µS cm⁻¹ (Table 1).

Fig. 4.

Basal ice layer at site C, showing pods of Bd-facies ice which occur as intrafolial fold hinges and boudinaged remnants of fold limbs within Bs₁ sub-facies ice. Ice axe (85 cm) provides scale.

Debris consisted of a sand-and silt-sized matrix, with coarser clasts occasionally protruding from the upper surface of the laminae into overlying clear ice. Such clasts also occurred dispersed within the clear-ice layers and were composed almost entirely of black ortho-amphibolite, which is the immediately underlying bedrock. Debris samples from the Bsi sub-facies were characteristically polymodal. Gaussian component analysis revealed that four discrete modes occurred repeatedly in samples from this sub-facies (although all four were not present in every sample) (Fig. 3). These modes were centred on 6.5Φ (fine silt), 3Φ (fine sand), −1.5Φ (fine gravel) and −3.5Φ (medium gravel). The fine-sand mode was commonly the dominant one (30–60%), although this was not always the case (Fig. 3b).

It seems probable that these particle-size modes can be interpreted in terms of the processes of comminution which affect debris in the subglacial environment (Reference HaldorsenHaldorsen, 1981). The fine-silt mode would therefore represent the products of glacial abrasion and be composed of debris particles finer than the mineral grains in the parent bedrock. The fine-sand mode would represent the products of crushing and would be composed of particles comparable in size to mineral grains in the bedrock. The two coarser modes would represent the products of bedrock fracture and would be composed of lithic fragments. If this interpretation is correct, then the consistent occurrence of three or four of these modes in ice of the laminated sub-facies suggests that it formed in areas of intimate ice–bed contact, where rock debris was being actively produced and comminuted.

Although four size modes were recognized in samples of debris from the laminated sub-facies, the majority of samples could be described adequately using only the three finer modes. It was therefore possible to recognize two sub-types of debris-size distribution. Low-level samples, collected from site B (where basal ice was less than 1.5 m thick) and from the lowest 3.9 m of site D (where basal ice was at least 13 m thick), displayed all four modes. High-level samples, collected from higher levels at site D, displayed only the three finer modes. For purposes of comparing the two sets of samples, we combined the two gravel-sized modes of samples with four modes and determined the mean proportions of each group of samples which fell into the gravel sand-and silt-sized modes, respectively. Low-level samples were composed, on average, of 50% gravel, 30% sand and 20% silt, while high-level samples were composed of 25% gravel, 45% sand and 30% silt. The low-level samples were therefore enriched in gravel and depleted in the finer grain-sizes relative to the higher-level samples. The enrichment in gravel resulted from addition of the fourth, very coarse mode, which was absent at higher levels in the sequence.

Whilst the addition of the fourth mode to the low-level samples is obviously a real effect, the depletion in silt-and sand-sized debris need not be so. It could arise simply because the addition of the fourth mode automatically reduced the fraction of the sample assigned to the other modes. To test this hypothesis, we calculated the fraction of the low-level samples which would fall into the three finer modes if the fourth mode were not present. In four out of five cases, the low-level samples still contained less silt than the high-level ones and in three out of five cases they also contained less sand. In no case did they contain less gravel. This suggests that the low-level samples were preferentially depleted in silt and, to a lesser extent sand, relative to the high-level samples. This implies either that the ice from which the low-level samples were derived formed at a time when, or in a place where, debris comminution by crushing and abrasion was less active than it was when the high-level ice was formed, or that the subglacial hydrological conditions when the low-level ice was formed favoured the flushing of comminution products from beneath the glacier. The occurrence of the fourth, coarse mode in the low-level samples suggests that bedrock fracture was particularly active when the ice from which they were collected formed. It is hard to envisage an environment in which bedrock fracture would be enhanced but comminution suppressed, so we favour the suggestion that there was efficient flushing of comminution products when the low-level ice formed.

Basal stratified facies — clear sub-facies (Bs_c)

At elevations of more than 3.9 m above the bed at site D, the basal stratified facies contained much thicker lenses of clear ice than occurred within the laminated sub-facies. These lenses could be decimetres thick and several metres wide and they tended to pinch out laterally within the more pervasive laminated ice. Ice crystals were up to 4 cm in diameter. The bubble content of the ice was low and bubbles were concentrated along high-angle planes, aligned parallel to the local ice-flow direction, which cross-cut the ice layers. These planes are interpreted as incipient fracture surfaces (cf. foliation S₂ of Reference PfefferPfeffer (1992)). Pfeffer argued that this foliation formed in response to transverse lateral extension or longitudinal compression within the propagating front of the 1982–83 surge. Thus, the distribution of these fracture surfaces implies that the Bs_c-facies ice, and therefore the basal-ice sequence at elevations greater than 3.9 m above the bed at site D, was entrained within the glacier prior to the 1982–83 surge. Debris concentrations in this facies averaged 0.4% by volume and electrical conductivities averaged 22.8 µS cm⁻¹ (Table 1). Included debris consisted of scattered coarse clasts and grit particles, dispersed randomly through the ice.

Basal stratified facies — solid sub-facies (Bs_s)

This sub-facies consisted of layers up to 80 cm thick of massive structureless debris with interstitial ice. Occas-ionally, there were more extensive lenses of clear, debris-free ice within the debris. The layers had undulating boundaries and could, in places, be traced laterally for tens of metres. They ablated more slowly than the laminated ice and, at site D, protruded from the ice surface, splitting the basal ice layer into a scries of wedges, each up to several metres thick (Fig. 5). Clasts up to 26 cm in length derived from a variety of lithologies (including the local ortho-amphibolite, greenschist, mica-schist and a quartzite) were recovered from this facies. This clearly differentiates the debris in this facies from that in the Bs₁ sub-facies, which was all locally derived. No obvious primary sedimentary structures were ob-served within the debris of this facies. However, the ice lenses were commonest in the more silt-rich layers, suggesting an origin as segregation ice lenses (cf. Reference Tison, Souchez and LorrainTison and others, 1989). Debris concentrations averaged 36% by volume and electrical conductivities averaged 60.7 µS cm⁻¹ (Table 1).

Fig. 5.

Basal stratified facies ice at site D, where the sequence is split into wedges by the outcrop of bands of Bs_s sub-facies ice.

Size distributions of Bs_s sub-facies debris were polymodal but the modes recognized by Gaussian component analysis were inconsistent between samples and frequently did not correspond with those found in the laminated sub-facies (Fig. 3). Although only four samples were analysed and three different types of composition were observed, these are believed to be representative of the range of debris types found within this facies. Type (a) sediment had three modes but 89% of the sample consisted of coarse-to-medium silt, similar to that found in the Ed facies. Type (b) had three modes and a size distribution very similar to that of high-level Bs₁ sub-facies debris. Type (c) had four modes, the most prominent of which consisted of fine-to-medium gravel and of coarse-to-medium sand. Together, these modes comprise over 70% of the size distribution.

Basal diffused facies (Bd)

Enclosed within the basal stratified facies were layers and pods of ice, up to 0.65 m thick and 1.1 m in lateral extent, with physical characteristics indistinguishable from those of Ed-facies ice. On account of its occurrence within the basal stratified ice, this ice is referred to as the basal diffused facies. It occurred in two main situations: at site D it cropped out in a well-defined zone immediately above the bed as a breccia, in which individual blocks showed different foliation orientations (Fig. 2); however, at sites B and C, and at higher levels of site D, isolated pods occurred throughout the thickness of the basal ice layer as boudinaged remnants of formerly more extensive layers and as rootless intrafolial fold hinges (Fig. 4). Debris concentrations and electrical conductivities of these pods were very similar to those of the Ed facies (Table 1).

Englacial diffused facies and snow

Snow, sampled in the ablation zone, and Ed-facies ice had a relatively wide range of isotopic compositions (Table 2). Values were somewhat heavier than others reported for the St Elias Range (Reference Epstein and SharpEpstein and Sharp, 1959; Reference MacPherson, Krouse and StoutMacPherson and Krouse, 1967; Reference West and KrouseWest and Krouse, 1972) but they were consistent with the lower elevations at which sampling took place on Variegated Glacier. Reference West and KrouseWest and Krouse (1972) suggested that δ¹⁸O decreased with altitude at a rate of 0.4‰ per 100 m on Rusty Glacier, Yukon Territory. They recorded a mean δ¹⁸O value of −22‰ at 2640 m on Hubbard Glacier, which lies immediately to the west of Variegated Glacier. Given these results and the altitude distribution of Variegated Glacier (0–1950 m), a range of composition of −11.40 to −19.30‰ in δ¹⁸O is predicted. This is very close to the observed range for snow and Ed-facies ice (−12.60–18.75‰).

Table 2.

Summary of the stable-isotope composition of ice facies sampled at Variegated Glacier. For basal ice, results are presented for the basal stratified facies as a whole, and also for the individual sub-facies. No data are available for the englacial banded facies

The δD versus δ¹⁸O relationship for samples of snow and Ed-facies ice is described by the line δD = 6.57(±0.27)δ¹⁸O – 14.97 (r² = 0.92, DF = 54, p < 0.001) Separate treatment of snow and Ed-facies samples reveals similar trends. For snow δD = 6.73(±0.33)δ¹⁸O) – 15 (r² = 0.98, DF = 9, p < 0.001), while for Ed-facies ice from the glacier centre line δD = 6.54(±0.73)δ¹⁸O–15.3 (r² = 0.8, DF = 21, p < 0.001). Allowing for the standard errors on the slope coefficients, the slopes of these lines are lower than the global average value of 8 normally found in samples of precipitation and unaltered glacier ice (Reference DansgaardDansgaard, 1964). They are, however, similar to the value of 6.86 determined from monthly mean values of precipitation at Adak (the only Alaskan site for which there are data from the IAEA network). Thus, although we cannot completely rule out the possibility that the empirically determined slopes correspond to the local precipitation slope, it seems likely that both snow and Ed-facies ice have undergone some isotopic fractionation relative to precipitation.

Since co-isotopic analysis suggestes that fractionation may affect both old snow and Ed-facies ice, it seems likely that this occurs from the earliest stages of firnification. At Variegated Glacier, firnification occurs entirely within the wetted zone, in which isotopic fractionation (and lowering of the co-isotopic slope relative to that of precipitation) might result from two main processes: (a) evaporation, which enriches the surface of the snowpack in D and especially ¹⁸O, and makes percolating melt-waters derived from the enriched surface layer isotopically heavier than the original snow (Reference Stichler and HerrmannStichler and Herrmann, 1978; Reference Moser, Stichler, Fritz and CMoser and Stichler, 1980; Reference Stichler, Baker, Oerter and TrimbornStichler and others, 1982); and (b) fractionation between snow and percolating meltwater during recrystallization of the snowpack, which results in the enrichment of the snowpack in heavy isotopes, particularly ¹⁸O (Reference ArnasonArnason, 1969, Reference Arnason and Gat1980; Reference Martinec, Moser, Quervain, Rauert and StichlerMartinec and others, 1977; Moser and Stichler, 1980; Reference Stichler, Baker, Oerter and TrimbornStichler and others, 1982). These processes may have had a particularly marked effect on the co-isotopic characteristics of snow and Ed-facies ice at Variegated Glacier because of the unusually low elevation of its accumulation area and because of the periodic occurrence of Chinook winds which induce strong evaporation from the snowpack. If individual samples of snow and glacier ice display varying degrees of isotopic fractionation relative to precipitation depending upon the environment in which firnification took place, then empirical co-isotopic slopes determined from groups of such samples may be purely statistical artefacts which fall somewhere between the precipitation slope and the theoretical freezing or evaporation slopes (see below).

Basal diffused facies ice

The isotopic composition of Bd-facies ice from sites B–D is slightly heavier than that of snow and Ed-facies ice from the glacier centre line (Table 2) but its co-isotopic characteristics are identical. Samples plot on the line δD = 6.62 (±0.59)δ¹⁸O−13.98 (r² = 0.92, DF = 11, p < 0.001). This result suggests that Ed-and Bd-facies ice have a similar origin, although the slightly heavier isotopic composition of Bd-facies ice may imply a lower mean elevation of formation than that of Ed-facies ice collected on the glacier centre line. Isotopically, Bd-facies ice is distinguishable from the Bs-facies ice by which it is surrounded on the basis of its steeper co-isotopic slope (see below).

Basal stratified facies ice

As a single population, basal stratified ice is isotopically heavier than snow and Ed-facies ice from the glacier centre line but similar in composition to Bd-facies ice found within it (Table 2). Its co-isotopic characteristics are, however, different from those of snow and Ed/Bd-facies ice, since samples plot on lines of lower slope. For the Bs facies as a whole, samples plot on the line δD = 5.81 (±0.32)δ¹⁸O−25.87 (r² = 0.88, DF = 46, p<0.001). When sub-facies are considered, the Bs_c facies is most similar to Bd-facies ice (δD = 6.09(±0.39)δ¹⁸O −20.9; r² = 0.97, DF = 9; p < 0.001), while Bs₁ and Bs_s are more dissimilar (Bs₁: δD = 5.77(±0.45)δ¹⁸Ο −26.7; r² = 0.84, DF = 31, p <0.001; Bs_s: δD = 5.86(± 1.07)δ¹⁸O −24.8; r² = 0.94, DF = 2; NS).

Englacial/basal diffused facies

The results presented above clearly establish the distinc-tion between Ed/Bd-and Bs-facies ice. Ed-facies ice comprises the bulk of the glacier and is produced by firnification processes in the accumulation area. On account of the low elevation of Variegated Glacier, these processes take place entirely within the wetted zone. The co-isotopic characteristics of Ed-facies ice thus appear to show evidence of fractionation between ice and percolating meltwater during recrystallization of the snowpack. Debris incorporated in Ed-facies ice is primarily aeolian in origin, though it also includes small quantities of coarse sand, which may be derived from the valley walls. Locally, higher concentrations of valley-wall-derived material occur within the Eb facies. Bd-facies ice is similar to Ed-facies ice in its physical appearance, debris content, electrical conductivity and co-isotopic characteristics, so the two facies probably have a similar origin. The isotopic composition of Bd-facies ice is heavier than that of the Ed facies sampled on the glacier centre line, suggesting formation at lower elevations. The outcrop geometry of Bd-facies ice suggests that it has become tectonically incorporated within the basal ice layer.

Basal stratified facies—laminated sub-facies

Bs₁ sub-facies ice is physically, structurally and co-isotopically distinct from overlying Ed-facies ice. The laminae within it are discordant with the foliation in Ed-facies ice and they are defined by variations in debris concentration rather than in bubble content. The overall debris content is much higher and the bubble content much lower than in the Ed facies. These characteristics suggest either that the BS₁ sub-facies ice has been accreted on to the base of the glacier by processes quite different to those involved in the formation of Ed-facies ice, or that it has suffered an extreme degree of metamorphism as a result of deformation in the shear zone at the base of the glacier. Whilst there is no doubt that the basal ice layer has been strongly deformed (see below), the extremely sharp nature of the boundary between Ed and Bs₁ sub-facies ice leads us to believe that basal accretion has taken place.1

The textural characteristics of Bs₁ sub-facies debris suggest that it was subglacially derived and entrained in an environment in which bedrock erosion and debris comminution were active. This probably implies a rigid substrate. The lithology of the debris suggests an origin no more than 3 km up-glacier from the sample sites. The contrast in debris character between high-and low-level samples suggests temporal variations in the intensity of bedrock fracture and in the efficiency of fluvial flushing of fine-grained debris within the environment in which this facies was formed. Observations made during the 1982–83 surge suggest that these variations may have been associated with the active and quiescent phases of the surge cycle.

The 1982–83 surge was apparently linked to a change from channelized subglacial drainage during quiescence to distributed subglacial drainage (Reference KambKamb and others, 1985; Reference KambKamb, 1987), which would have given meltwaters access to a large proportion of the glacier bed. Extremely large quantities of suspended sediment were removed from the glacier during and immediately after the surge (Reference HumphreyHumphrey, 1985; Reference SharpSharp, 1988), suggesting that fluvial flushing of comminution products was particularly effective under these conditions. The absence of comminution products from ice which was in close proximity to the bed immediately after the surge suggests that this ice formed and/or interacted with the glacier bed during the surge, when this flushing took place. The presence of comminution products in ice at higher levels above the bed suggests that this ice formed under conditions when debris flushing was less efficient (e.g. meltwaters were confined to a few large subglacial channels) and that it did not interact with the bed during the surge. The observation that Reference PfefferPfeffer’s (1992) S₂ surge-induced foliation is present in layers of Bs_c sub-facies ice contained within high-level Bs₁ ice confirms that this ice was in existence prior to the 1982–83 surge. Although this ice could have formed during a previous surge, the characteristics of the debris entrained within it suggest that it is more likely to have formed during the quiescent phase of a surge cycle.

The presence of relatively large quantities of medium gravel in the low-level ice suggests that bedrock fracture was an active process when this ice formed. This might be expected under surge conditions, when subglacial water pressures were high (Reference KambKamb and others, 1985) and subglacial cavitation was presumably extensive. Such conditions significantly weaken rock on the lee side of bedrock bumps (Reference Boulton and CoatesBoulton, 1974; Röthlisberger and Iken, 1981). Fracture and removal of eroded bedrock are facilitated by fluctuations in the water pressure in leeside cavities (Reference IversonIverson, 1991). High rates offrictional melting at the glacier bed (associated with high basal shear stresses and large sliding velocities) also favour fracture by inducing unusually high contact forces between entrained particles and the bed (Reference HalletHallet, 1979).

Laminated ice similar in appearance to the Bs₁ sub-facies ice at Variegated Glacier is widespread at the base of glaciers in the western European Alps (Reference BoultonHubbard, 1992), where its formation has been linked to regelation as ice flows over bedrock of irregular geometry (Reference Boulton and CoatesHubbard and Sharp, 1993). In this model, it is argued that laminated ice forms by refreezing of meltwaters on the downstream side of bumps. Gas and debris (particularly the finer grain-sizes) are rejected from the ice during the initial stages of refreezing (Reference Bindschadler, Harrison, Raymond and CrossonHubbard, 1991), accounting for the low bubble and debris content of the clear-ice layers. Both debris and air bubbles may, however, be entrapped within the ice during the final stages of freezing, when the residual water becomes super-saturated with respect to dissolved gases, bubble nucleation occurs and the removal of gas from solution raises the freezing point, so that the rate of freezing increases (cf. Reference Rowell and DillonRowell and Dillon, 1972). This mechanism can account for the observed association of debris laminae and bubble planes in this facies. Partial melting of debris laminae on the upstream side of bumps of lower amplitude located further downstream may concentrate debris at the ice/bed interface and increase the local debris concentration. Internal melting within the basal ice could have a similar effect (Reference LliboutryLliboutry, 1993).

Co-isotopic analyses of BS₁ sub-facies ice help to test this hypothesis, because it has been demonstrated that isotopic fractionation during the refreezing of meltwaters usually produces ice samples which plot on a δD versus δ¹⁸O diagram with a slope lower than that on which source waters plot (Reference Jouzel and SouchezJouzel and Souchez, 1982; Reference Souchez and JouzelSouchez and Jouzel, 1984). When freezing occurs under closed-system conditions (as in regelation, for instance), this slope (S) is given by:

where α and β are the equilibrium fractionation coefficients for D and ¹⁸O, respectively (1.0212 and 1.00291; Reference Lehmann and SiegenthalerLehmann and Siegenthaler, 1991), and δ_i refers to the isotopic composition of the initial (source) liquid (Reference Jouzel and SouchezJouzel and Souchez, 1982). For open-system freezing, in which the composition of the water input to the reservoir during freezing is not significantly different from that of the initial liquid, the slope is given by:

(Reference Souchez and JouzelSouchez and Jouzel, 1984). On these slopes, samples which plot at the heavy end of the isotopic spectrum form during the early stages of freezing, whereas those which plot at the light end form during the final stages of freezing. For the case of closed-system freezing, the mean isotopic composition of the solid formed at each stage of the freezing process is given by:

where δ_s is the isotopic composition of the solid formed and K is the fraction of the liquid reservoir that has frozen (Reference Jouzel and SouchezJouzel and Souchez, 1982).

In evaluating the origin of the Bs₁ sub-facies, it is useful to consider whether or not the empirically determined co-isotopic slope for this sub-facies coincides with that predicted using either of the above models. Some caution is, however, necessary in adopting this approach. In previous work, it has been argued that the isotopic composition of the initial liquid can be defined by the point of intersection of the local precipitation slope and the empirically determined slope for basal ice facies Qouzel and Souchez, 1982). This makes two assumptions which are not necessarily valid: (i) that all samples of basal ice analysed are derived from initial liquids of identical composition, and (ii) that these initial liquids lie on the local precipitation slope. The first assumption ignores the range of isotopic compositions which characterizes waters in the glacial environment and assumes that the empirically determined basal ice slope is equivalent to the physically meaningful theoretical freezing slope. If, in reality, basal ice forms from a range of initial liquids, the empirically determined slope may be no more than a statistical artifact. The second assumption ignores the possibility that melting of pre-existing basal ice can provide the initial liquid for later generations of such ice and that such initial liquids may not lie on the precipitation slope.

We therefore make the more realistic assumption that basal ice could potentially form from any initial liquid within the range of compositions measured at Variegated Glacier. Taking initial liquids at the light and heavy extremes of the waters sampled (−142.60‰δD and −18.75‰ δ¹⁸O, and −100.00‰ δD and −12.60‰ δ¹⁸O, respectively), the predicted freezing slopes are 6.37 and 6.64 (closed system; Equation (1)), and 6.48 and 6.76 (open system; Equation (2)), respectively. These are all steeper than the empirically determined slope for the Bs₁ sub-facies (5.77 ±0.45). Although the fact that the empirical slope is gentler than the local precipitation slope is consistent with the suggestion that the Bs₁ ice formed by the refreezing of meltwaters, the poor correspondence between the theoretical and empirical slopes suggests either that the latter is a statistical artefact (resulting from basal ice forming from a variety of initial liquids) or that the isotopic composition of the Bsi sub-facies ice has undergone modification subsequent to its formation.

Let us first consider the possibility that basal ice forms from a range of initial liquids. Taking, as examples, initial liquids at the light and heavy extremes of measured values of snow and Ed-facies ice, and plotting the theoretical freezing slopes derived from these on a co-isotopic (δD versus δ¹⁸O) diagram, it is possible to define an envelope within which samples of basal ice should plot (Fig. 6). For the case of closed-system freezing, it is possible to useEquations (1) and (3) to define a grid within this envelope. In this grid, the sub-horizontal lines represent freezing slopes derived from different initial liquids. The sub-vertical lines link points on different freezing slopes corresponding to equivalent frozen fractions of their respective initial liquids (Fig. 6). The position within the grid at which an individual sample plots is determined by (a) the composition of the initial liquid, and (b) the frozen fraction represented by the sample. Some of the samples of Bs₁ sub-facies ice plot outside and to the right of this grid, implying that they formed from initial liquids slightly enriched in ¹⁸O relative to those used to define the envelope. However, all of the samples plot close to the freezing slope for an initial liquid comparable in composition to or slightly ligher than the mean of Ed-facies ice on the glacier centre line (−122.1 and −16.34‰), and within the range of isotopic compositions which could be produced by freezing such a liquid (frozen fractions of 0.2–0.7). This suggests (i) that, if Bs₁ sub-facies ice was formed by refreezing of melt-waters, refreezing actually took place under open-system conditions (because K values are <1), and (ii) that the waters involved in refreezing were isotopically more homogeneous than the range of possible initial liquids at Variegated Glacier.

Fig. 6.

Co-isotopic diagram, showing the envelope within which should lie samples of basal ice formed by the freezing meltwaters within the range of isotopic compositions defined by samples of snow and Ed-facies ice. Sub-horizontal lines are theoretical freezing slopes for different initial liquids, and span the range of compositions produced by freezing of different initial liquids. Crosses on these lines define the predicted isotopic composition of different frozen fractions. Sub-vertical lines link points corresponding to equivalent frozen fractions. Freezing slopes are plotted for initial liquids at the light and heavy extremes of samples of snow and Ed-facies ice (see text), and with a composition equivalent to the mean of Ed-facies ice (−122.12‰δD, −16.34‰δ¹⁸O). Circles and triangles represent samples of Bs_s and Bs₁ sub-facies ice, respectively.

Open-system freezing would be expected if regelation was accompanied by net basal melting due to geothermal heating or sliding friction. Net melting would result in a continuous influx to the freezing environment of water derived from Ed-facies ice, ensuring that the isotopic composition of the initial liquid remained close to that of Ed-facies ice. Initial liquids lighter in composition than the mean of Ed-facies ice could be produced by isotopic fractionation accompanying recrystallization either in the basal shear zone of the glacier (see below) or during firnification.

One problem in applying this model to the basal ice sequences at Variegated Glacier lies in the difficulty of explaining the large thicknesses of Bs₁ ice observed. Theoretical considerations and simple modelling (Reference LliboutryLliboutry, 1993; Reference Boulton and CoatesHubbard and Sharp, 1993) suggest that ice layers produced by this mechanism should not exceed a few decimetres in thickness (or less if there is net basal melting). Whilst such thicknesses are consistent with those observed at site A and possibly site B, they are much less than observed at sites C and D. Two mechanisms might explain this discrepancy: (i) the basal ice was formed by net basal adfreezing rather than by regelation, or (ii) the basal ice layer was thickened significantly by tectonic processes. Net basal adfreezing seems unlikely in the case of Variegated Glacier, because it is known to have a temperate thermal regime, and because it probably experiences high rates of net basal melting during surges which would have the effect of thinning rather than accreting a basal ice layer. The role played by tectonic processes is discussed below.

As an alternative to the regelation model discussed above, it is possible that the co-isotopic characteristics of Bs₁ sub-facies ice result from recrystallization, partial melting and loss of isotopically light meltwater from basal ice resulting from high stresses and strain rates close to the glacier bed, particularly under surge conditions. However, three factors lead us to believe that this is probably not the case. First, Bd-facies ice enclosed within Bs₁ sub-facies ice does not appear to have had its co-isotopic slope modified by such processes. Secondly, such processes do not provide a means of entraining the high debris content of Bs₁ sub-facies ice. Thirdly, only a small number of samples of basal ice seem to have formed by refreezing of isotopically light meltwaters such as would have been squeezed from basal ice by these processes. We therefore favour the idea that the Bs₁ sub-facies ice formed by regelation in association with net basal melting, although we cannot totally rule out the possibility that the isotopic composition of the ice has been modified by fractionation accompanying recrystallization. The sub-horizontal and planar nature of the lamination within this facies may, however, be attributable to rotation of fabric elements into the plane of shear.

Basal stratified facies ’ clear sub-facies

The occurrence of S₂ foliation within Bs_c sub-facies ice suggests that it formed prior to the 1982–83 surge, and its outcrop within Bs₁ sub-facies ice which contains comminution products suggests that it may well have formed during the quiescent phase of a surge cycle. Again, the question arises as to whether it was accreted on to the glacier sole by the refreezing of meltwaters, or whether it represents a metamorphosed form of Ed-facies ice. The low debris content and electrical conductivity of the ice are comparable to those of the Ed and Bd facies but the bubble content is much lower. The co-isotopic slope of Bs_c sub-facies ice (6.09 ± 0.39) is intermediate between those of Ed and Bs₁ ice. Whilst it is difficult to draw unequivocal conclusions about the origin of the Bs_c sub-facies on the basis of these data, the similarities to Ed-facies ice and the intermediate isotopic characteristics suggest that an origin as metamorphosed Ed-or Bd-facies ice may be more likely than formation by basal freezing. If this view is correct, the low bubble content of the ice would be explained as a result of the removal of gases in solution by percolating meltwater (cf. Reference BernerBerner and others, 1978), while the intermediate co-isotopic slope would be linked to the processes of recrystallization, partial melting and melt-water squeezing described above. This then suggests that Bs_c sub-facies ice is identical in origin to Bd-facies ice but that its physical appearance and co-isotopic characteristics have been modified as a result of it experiencing a much higher degree of metamorphism.

Basal stratified solid sub-facies ice

The character of solid sub-facies debris suggests that it consists of sediments which were accreted en masse (rather than particle-by-particle) on to the base of the glacier (cf. Reference LawsonLawson, 1979). It is therefore likely that this sub-facies formed where the glacier bed consisted of unconsolidated sediments, while the Bs₁ sub-facies formed where it consisted of rigid bedrock. The co-isotopic characteristics of the Bs_s sub-facies are indistinguishable from those of the Bs₁ sub-facies, suggesting that the ice formed by open-system freezing, with metamorphic effects possibly superimposed. The size characteristics of the sediments suggest considerable variation in the character of the source sediments, reflecting the range of transport and depositional processes active in the subglacial environment. Type (b) debris, which contains a wide range of particle sizes, probably originated as subglacial (perhaps lodgement) till, which was observed in the proglacial area a short distance from site D. Types (a) and (c) show a greater degree of sorting and are probably fluvial or lacustrine in origin. Type (c) contains very little silt-and clay-sized material, and has a size distribution characteristic of glaciofluvial gravels. Type (a) consists primarily of silt and represents a lower-energy depos-itional environment. It is significant that types (a) and (c) were most commonly observed at site D₁ which is immediately adjacent to one of the main meltwater streams draining the glacier. This suggests that the sediments may have been transported and deposited by this stream prior to being entrained by the glacier. Type (c) sediments were probably deposited within the stream channel or on a bar within it, while type (a) sediments were deposited in a slack-water environment. Long-distance debris transport by glacial meltwaters can account for the occurrence of exotic lithologies in the debris of this facies.