Deformation, Strength and Tectonic Evolution of Basal Ice in Taylor Glacier, Antarctica

Observation and measurements of ice structure and deformation made in tunnels excavated into the margin of Taylor Glacier, a polythermal glacier in the McMurdo Dry Valleys of Antarctica, reveal a complex, rapidly deforming basal ice sequence. Displacement measurements in the basal ice, which is at a temperature of −18°C, together with the occurrence of cavities and slickenslides, suggest that sliding or rapid deformation in thin zones of high shear occurs at structural discontinuities within the basal zone. Strain measurements show that the highest strain rates occur in ice with average debris concentrations of 26% followed by ice with debris concentrations of around 12%. The lowest strain rates occur in clean bubbly ice that has very low debris concentrations (<0.02%). Deformation within the basal ice sequence is dominated by simple shear but disrupted by folding which results in shortening of the debris‐bearing ice followed by attenuation of the folds due to progressive simple shear which generates predominantly laminar basal ice structures. About 60% of glacier surface velocity can be attributed to deformation within the 4.5 m thick sequence of basal ice that was monitored for this study, and 15% of motion can be attributed to sliding or very localized shear. The combination of high debris concentrations and high strain rates in the debris‐bearing ice results in high rates of abrasion and the production of striated and facetted clasts typical of temperate glaciers, even though the basal ice is at a temperature of −18°C.


Introduction
The processes that occur at the beds of glaciers within a few meters of the contact between ice and the glacier substrate are fundamental controls of glacier behavior.Deformation of ice and debris close to this zone regulates the nature of motion at the bed, the rate of deformation experienced by ice and the deformation and entrainment of subglacial sediment (Clarke, 2005;Hubbard, 2006;Iverson, 2010).One of the products of glacier-substrate interactions is basal ice, which can be defined as ice close to the bed of glaciers that has distinctive physical, chemical and mechanical properties (Hubbard & Sharp, 1989;Hubbard et al., 2009).These properties are imparted by interactions between glacier ice and the substrate.The processes that occur within basal ice and at glacier beds underpin the large-scale behavior of glaciers and control the geological processes at the ice-substrate interface.These processes also determine the geological imprint of glaciation on the Earth's surface (Alley et al., 1997;Boulton, 2006;Iverson, 2010).
Debris and solutes within basal ice alter the rheology of materials close to the bed, which has direct implications for glacier behavior and glacial geological processes.For example, a cold-based glacier with easily deformable basal ice will have higher surface velocities and a flatter surface profile than a cold-based glacier without easily deformable ice (Pettit et al., 2014).Both experimental and field-based research on the behavior of ice-debris mixtures suggests that we do not yet understand the constitutive properties of debris-ice mixtures (Moore, 2014;Thompson et al., 2020).There is a substantial body of experimental data that suggests the addition of solid debris increases ice strength (Fitzsimons, 2006;Goughnour & Andersland, 1968;Nickling & Bennett, 1984;Waller, 2001;Waller et al., 2012).However, field observations remain equivocal (Moore, 2014).Some studies have determined that ice containing debris is stronger than adjacent pure ice (Fitzsimons & Howarth, 2020;Fitzsimons et al., 1999Fitzsimons et al., , 2000Fitzsimons et al., , 2008Fitzsimons et al., , 2023;;Rabus & Echelmeyer, 1997), whereas others have concluded that ice containing debris is weakened relative to adjacent ice (Cohen, 2000;Echelmeyer & Zhongxiang, 1987).Using compressive strength tests on ice from Taylor Glacier Lawson (1996) showed that the addition of 10% debris by volume resulted in increases and decreases in ice viscosity.Although some studies have concluded that the addition of debris has no impact on the strength of the ice (Cuffey et al., 2000;Jacka et al., 2003;Moore, 2014), the general view is that ice containing sediment is expected to be stronger than adjacent ice and less susceptible to ductile deformation well below the freezing point (Moore, 2014;Warbritton et al., 2020).In an extensive review of the behavior of frozen debris Moore (2014) concluded that ice-debris mixtures are more resistant to deformation at low temperatures than pure ice but that at temperatures closer to melting the growth of an interfacial water film can lead to profound weakening.Collectively, this work has demonstrated that there are several key gaps in our understanding of basal processes and a full understanding of the complexity of linkages between ice, debris, and glacier behavior remains elusive.
Taylor Glacier is an outlet glacier that flows from Taylor Dome, which is part of the East Antarctic Ice Sheet (Figure 1).This polythermal glacier terminates in the McMurdo Dry Valleys at the permanently ice-covered Lake Bonney, where the ice margin is characterized by 18-20 m-high cliffs.At the glacier terminus Robinson (1984) measured a mean ice surface temperature of 17°C and concluded that the glacier was at the pressure melting point some distance upstream of the terminus.Using an ice penetrating radar Hubbard et al. (2004) concluded that upstream of the glacier terminus the bed consisted of saturated sediment or ponded liquid.Later work has mapped widespread hypersaline groundwater that is connected to Blood Falls, a hypersaline water discharge at the glacier terminus (Badgeley et al., 2017;Foley et al., 2016;Mikucki et al., 2015).A study of ice-crystal textures and gas content of the basal ice from Taylor Glacier shows that significant shearing has occurred within the basal ice and concludes that variations in resistance to deformation occur at the cm to mm-scale within the basal ice (Samyn, Fitzsimons, & Lorrain, 2005).Pettit et al. (2014) used a two-layer flowband model to investigate glacier behavior and concluded that the basal ice contributes 85%-90% of glacier surface velocity and that the basal ice is likely 10-15 m-thick and 20-40 times softer that Holocene age glacier ice.These findings suggest that at least in some circumstances, cold-based ice can be very sensitive to subglacial deformation processes.
Here we hypothesize that strain in the basal zone of Taylor Glacier is localized to thin bands of intense shear or sliding and that deformation of the basal ice is the major component of glacier surface velocity.We further hypothesize that despite the low temperature at the margin of this glacier, localized shear and potentially sliding results in a structural evolution of the basal zone and that these processes impart a distinctive glacial geological signature.This paper describes the geomorphological setting and the methods used in the study (Section 2), and presents observations of the composition and deformation of the basal zone from a combination of field and laboratory analyses (Section 3).Interpretation of the data focuses on the patterns of ice deformation, the rheological controls of deformation in the basal zone of the glacier, and the geological imprint of these processes (Section 4).

Excavation
In the austral summer of 1998-1999 a 20 m-long tunnel was excavated approximately 1.1 km upstream of the terminus at 77.71994°S 162.21939°E (T1 in Figure 1) using electric chainsaws with tungsten carbide cutters on the chain to make relief cuts in the ice, a demolition hammer to break up the ice and a sled to remove the ice from the tunnel.This tunnel was studied for a 2 week period but was destroyed by a combination of melting and flooding before it could be resurveyed in the following summer.A second tunnel was excavated and instrumented in December 1999 approximately 200 m downstream of the first tunnel at 77.72068°S 162.22854°E (T2 in Figure 1).This tunnel was resurveyed in January 2000 and again in January 2001.Since the excavation, the margin of the glacier has remained largely unchanged.The tunneling process is documented in a series of short videos stored in Zenodo (Fitzsimons, 2023).The temperature of the ice measured along the tunnel was 18 ± 1°C beyond 8 m of the tunnel entrance (Figure 2).The air temperature at the back of the tunnel remained around 15°C during excavation.Both tunnels were approximately 20 m-long and curved to avoid direct sunlight getting into the tunnels.The walls were oriented parallel to the ice flow at the end of each tunnel.A 4 m-deep shaft was cut at the end of the second tunnel to expose the debris-bearing basal ice (Figure 3).Most of the observations and data reported in this paper come from the second tunnel.

Debris Characteristics
The particle size distribution of sediments was determined by treating subsamples with 5 ml of 30% H 2 O 2 to digest organic matter and rinsed in de-ionized water.Biogenic silica was removed using 10 ml of 1 M NaOH and rinsed with de-ionized water.The samples were then treated with 5 ml of 5% sodium hexametaphosphate solution and shaken gently for 2 hr to deflocculate clay-sized particles.The treated samples were analyzed using a Malvern Mastersizer 2000 laser particle size analyzer or sieved if the samples contained particles >2 mm in diameter.
Debris concentration measurements were made by displacing the samples in water to determine volume, then the sample was melted, the water evaporated and the remaining sediment weighed.Debris concentrations are expressed as percentages of volume calculated using a particle density of 2,700 kg.m 3 .Three representative samples of 30 particles >40 mm in diameter were retrieved by melting blocks of the massive ice facies (Section 4.1).These particles were examined for striations and their shape was classified using Powers roundness classes (Powers, 1953).

Ice Motion Measurements
Measurements of ice movement were obtained using plumblines and engineer dial gauges that were rock-bolted to boulders in the basal ice.The dial gauges used were Mitutoyo series 2 gauges that have a precision of ±0.02 mm and an accuracy of ±20 μm.The plumblines consisted of nylon lines attached to a peg drilled and frozen into the top of the tunnel wall (Figure 3a).The line was held taut with a plumb bob which was mounted above a brass target drilled and frozen into the tunnel floor.Displacements were measured from the plumbline and wooden markers drilled and frozen into the ice at 0.1 m intervals using a digital caliper for displacements up to 0.2 mm and a steel tape for displacements >0.2 m.The plumblines were resurveyed at intervals of 27, 42, and 403 days between December 1999 and January 2001.Although the motion of brass targets on the floor was measured and appeared to be stationary, we cannot entirely discount the possibility that they are moving.

Strain Measurements
Ice strain was measured using strain arrays that consisted of square arrays of stainless coach screws that were drilled and frozen into the tunnel walls parallel to the ice flow (Figure 3b).Each stainless steel coach screw had a cone milled into its head in order to accurately position a digital caliper that had locating cones fixed to the jaws.Measurements were made using a Mitutoyo digital caliper with a precision of 0.001 mm.The caliper was calibrated by repeat measurements of cones milled into a block of stainless steel.Repeated measurements of the calibration block resulted in an operator error of ±0.0096 mm and an accuracy of ±0.02 (n = 60).All the strain rates reported in this paper are based on measurements between 42 and 403 days.During this time, some bulging of the tunnel walls occurred, which could introduce errors related to differences in closure rates in different basal ice facies.This problem is addressed in Section 2.7 of this paper and bulging the tunnel walls was monitored by surveying the shape of the tunnel over the duration of the experiment.
Strain rates were calculated from the corners of triangles defined from the strain arrays using a numerical calculation of a Mohr circle based on the method of Ramsay (1967) and used in glaciology by Hambrey and Müller (1978) and Hambrey et al. (1980).We have followed the convention for principal strain rates that ε1 and ε2 are horizontal, that ε1 ≥ ε2 , that extension is positive, and that ε3 is always vertical (Sharp et al., 1988).Shear strain rate was calculated as: γ = Ė1 Ė2

2
. The analytical techniques and errors associated with the strain calculations derived from field measurements made in tunnels are described in Fitzsimons et al. (2008).If the walls of the tunnel did not bulge ε3 would equal zero, which is not the case with any of the measurements.However, ε3 is an order of magnitude less than ε1 and both ε1 and γ increase into the tunnel so we conclude that tunnel closure has not introduced major errors into the calculation of strain rates.

Ice Sampling
Samples of basal ice were collected for chemical and structural analysis using a chainsaw in ice with relatively low debris concentrations and a 59 mm diameter diamond corer where the ice contained high debris concentrations or larger particles.Following collection, the frozen samples were returned to the laboratory where they were subsampled in a freezer using a band saw to produce samples of about 50 ml.After subsampling, the ice was melted at room temperature within 10 min of being cut and immediately filtered with 0.45 μm cellulose nitrate filter paper.The laboratory procedure is described in Fitzsimons et al. (2008).

Direct Shear Tests
A series of direct shear tests were undertaken in the tunnel using a modified laboratory direct shear device (Figure 3c).The shear box was constructed on stainless steel plates that were drilled to accommodate a 0.059 m diameter cylindrical sample (Figure 3d) and driven by a stepper motor through a load cell to move the upper plate (Figure 3c).The device was capable of displacing the sample by about 5 mm, which is a strain of approximately 8%.Shear test samples were cut from the basal ice using a diamond corer driven by an electric drill.The experiments were conducted in the tunnel which had an air temperature of 15°C, with displacement rates of 0.85 mm hr 1 and a normal load of 200 kPa.The normal load was applied by a pneumatic truck shock absorber attached to the tunnel roof and a load cell above the sample (Figure 3c).The tests were run for 3-4 hr to reach peak shear strength.

Limitations
Making ice deformation measurements in tunnels is problematic because any instruments introduced into the glacier bed run the risk of changing the boundary conditions that they are designed to measure.This is particularly the case with tunnels cut into glacier margins because the ice will tend to creep into the tunnel.In this study, we attempted to mitigate the effects of creep of the tunnel walls from the strain measurements by a combination of orienting the planes in which measurements were made parallel to ice flow, and by simultaneously recording deformation of the tunnel walls.Although the presence of the tunnel has changed the strain behavior of the ice, the measurements we made are consistent with the results of the direct shear tests, which leads us to conclude that the measurements we made capture the relative differences between the three main ice facies and sub-facies that were present at Taylor Glacier described in Section 4 below.

Physical Characteristics of the Basal Ice
Three distinctly different facies can be identified in the basal zone; clean englacial ice, stratified ice that consists of two subfacies, and a coarse pebble-cobble gravel at the bottom of the excavation (Table 1).The stratified facies consists of two subfacies: a laminated subfacies composed of alternating laminae of debris-bearing and clear clean ice layers 1-5 mm thick and a massive subfacies that has an unstructured appearance.
The clean englacial ice consists of white bubbly ice with very low debris concentrations that average 0.19% (Figure 3b, Table 1).The debris consists of dispersed silt to sand-sized particles and most of the stretched bubbles have axial ratios between 1:3 and 1:5.The laminated subfacies consists of finely laminated debris-rich ice and clean, clear ice layers (Figures 4a and 4b).Individual laminae range from 5 mm to less than 1 mm-thick and the debris ranges from abundant silt-sized particles up to boulder-sized particles held in the ice (Figure 2).The massive subfacies consists of densely packed debris in a matrix of clear clean ice.Although this ice has the appearance of being massive, sections cut from the ice reveal the presence of abundant clear, clean ice lenses and laminae (Figures 4c and 4d).Average debris concentrations in this subfacies are 33.2%vol.The very low standard deviation Section 2.6 reflects remarkably uniform debris concentrations in the central part of this subfacies (Table 1).
The bottom of the shaft consisted of a 0.3 m-thick coarse cobble-pebble gravel that stopped the excavation.Attempts to drill through the gravel with an ice auger equipped with tungsten carbide cutters and an attempt to map the geometry of the unit with a ground penetrating radar were unsuccessful.Solute concentrations are highly variable in the basal ice (Table 2).The englacial ice is characterized by low concentrations in Cl and the major cations, and the values have a low standard deviation.The laminated subfacies is characterized by moderate concentrations of Cl and the major cations with a relatively high standard deviation, likely reflecting two subpopulations of the clean clear laminae and the debris-bearing laminae.The highest solute concentrations were found in the massive subfacies where Cl is an order of magnitude greater than the laminated subfacies and the major cations are all at least three times the values of the laminated subfacies with the exception of Na + which is lower than the value for the laminated subfacies (Table 2).The high standard deviation of the partial TDS measure shows that there is very high variability in the values in both stratified subfacies.

Sedimentology and Structure
Both the laminated and massive subfacies contain dolerite, gneiss and granite boulders up to 0.6 m in diameter (Figure 2).A selection of pebble to cobble-sized particles was recovered from the ice by allowing the ice matrix of the samples to sublimate.These particles were frequently striated in cross cutting patterns with fine rock debris preserved adjacent to the striae (Figures 5a and 5b).Although not as well developed or abundant as striae on  dolerite particles, gneiss and granite clasts also exhibited striae in cross-cutting directions, primarily on protuberances on the particles (Figure 5b).The cross-cutting striae suggest substantial rotation of the particles within the basal ice during abrasion.A few striated ventifacts were recovered from the laminated subfacies.(g) Roundness of pebbles from the main debris band classified using Powers roundness classes (Powers, 1953).(h) Predominantly subangular pebble-sized particles retrieved from melting 10 kg of ice from the massive subfacies 3 m above the datum.
Particle size distributions of the sediment fractions less that 3.5 ϕ (11 mm) from the massive and laminated subfaces show that all the samples were very poorly sorted and characterized by multi modal distributions (Figures 5c-5e).The largest peak in the distributions occurs between 3.5 and 4.0 (0.088-0.062 mm), which encompasses fine sand and coarse silt particles.A persistent secondary peak occurs at 2 ϕ (4 mm), which is composed of fine gravel particles.The modal particle shape is subangular, but there are large numbers of subrounded and angular particles (Figures 5g and 5h).
Although the laminated and massive subfacies are dominated by planar structures (Figure 6), numerous ductile deformation structures were observed, particularly in the laminated subfacies.Figure 6a shows a thin section of a layer of the laminated subfacies between two layers of englacial ice and there is a strong contrast in the diameter of ice crystals in the two facies.Two 10-15 mm-diameter pebbles show that the ice layers are warped over the particles and that there are dark shadow zones on the stoss and lee sides of the pebbles, some of which are airfilled cavities.

Strength of the Basal Ice
The direct shear tests show that the clean englacial ice has the highest shear strength, followed by the laminated subfacies and the massive subfacies (Table 3).The massive subfacies has half the average shear strength of the englacial samples.The average peak shear strength of the samples from the different facies are significantly different from each other (t-tests with p values between 0.000 and 0.008) and the tests have low standard deviations.The shear strain rate calculated from the strain arrays shows that the materials with the highest strength, the englacial facies, have the lowest strain rates and the lowest strength materials, the massive subfacies, have the highest shear strain rates (Table 3).A few triaxial tests using samples cut so that the layers are 45°to the long axis of the samples were undertaken in laboratory conditions.Although fewer in number, the results of these tests (Table S1) are in the same order as the direct shear tests conducted in the field shown in Table 3.While some of these tests resulted in barrel-shaped failures of the samples, many failed along ice layers within the laminated facies (Figure 7).

Velocity of Ice in the Tunnel
A plumbline was located on the right wall in the shaft excavated at the end of the tunnel (Figure 1), which spanned the 4.5 m section of the basal zone (Figure 8).Prominent offsets occur in the displacement profile at 1.7 and 2.7 m above the measurement datum.The offset 1.7 m above the datum is 0.074 m a 1 and the one located at 2.7 m is 0.13 m a 1 .Both the offsets occur between pegs that were 0.1 m apart at the boundaries of the laminated and massive subfacies.Displacements at the 1.7 m interface were also measured directly using engineers dial gauges rock-bolted onto cobbles and boulders embedded massive subfacies but protruding into the laminated subfacies.
At the four locations measurements made over a period 4-31 days the displacements ranged between 0.046 and 0.167 m a 1 .

Velocity of Ice at the Glacier Surface
Optically surveyed velocity measurements between 2.1 m a 1 close to the ice margin to 5.5 m a 1 in the glacier center line over 376 and 769 days (Figure 1).The surface velocity close to the tunnel was 2.1 m a 1 , which is within the error of a 2.5 m a 1 , velocity determined by optical surveying of pegs placed in the cliff (B.Hubbard pers.comm.2007).These velocity measurements are very similar to the magnitude and pattern of surface velocities determined by GPS surveying of stakes on the glacier surface (Pettit et al., 2014).

Strain of Ice in the Basal Ice Shaft
The highest shear strain rates are in the massive subfacies followed by the laminated subfacies and the englacial ice (Figure 6d).These differences align with observations made of tunnel closure and deformation of boreholes and strain markers cut into the tunnel walls.After 403 days, the greatest deformation of the tunnel occurred in the shaft excavated at the end of the tunnel where the rectangular cross-section of the tunnel was observed to deform into a keyhole shape (Figure 9a).Cylindrical holes cut in the tunnel walls with a diamond corer deformed rapidly in the laminated subfacies during the experiment (Figure 9b).Within 4 days, the holes deformed into ellipses with axial ratios between 1.38 and 1.53, whereas the corresponding holes in the low debris concentration laminated subfacies had average axial ratios of 1.2.There was no detectable deformation of the holes in clean ice.Closure occurred fastest at the interface between the laminated subfacies above and below the massive subfacies layer (Figure 9c).Bulging also occurred in the back wall of the tunnel, which was oriented transverse to ice flow (Figure 1).Here, the maximum rates of deformation occurred in the laminated subfacies and the bulging resembled the shape of a plug (Figure 9d).

Motion in Basal Ice Layer
The measurements made in the tunnel show that deformation is dominated by progressive simple shear within the debris bearing ice, with strong differences in the rate of deformation in different ice facies.The highest rates of strain occur in the massive subfacies followed by the laminated subfacies and the englacial facies, a pattern that is the same order as the direct shear tests.Two exceptions to the pattern of progressive simple shear are the presence of offsets in the plumbline profile which may be caused by thin shear zones or sliding interfaces and the presence of polyharmonic recumbent folds which point to flow perturbations within the basal ice.
The offsets in the velocity profile recorded at 1.7 and 2.7 m above the datum occur at interfaces between the massive and laminated subfacies (Figure 8).Measurements of displacements on boulders at the upper interface, together with the development of air-filled cavities and slickenslides, suggest that there may be sliding or intense local shearing at these interfaces.However, the spacing of the measurement pegs used in the plumbline (0.1 m) precludes a definitive interpretation.The offsets could be produced by sliding or by narrow zones of higher rates of shear.Shear localization is also a feature of some triaxial tests (Figure 7).The possibility of sliding is discussed in Section 4.2.
The deformation structures observed around particles in the laminated subfacies (Figure 6a) suggest a sense of vorticity (Samyn et al., 2009), and the pebbles appear very similar to sigma grains observed in deformed geological materials (Passchier & Simpson, 1986).Tight, recumbent, polyharmonic folds are common, particularly at the interfaces between the laminated facies and the massive facies (Figure 6c).Short wavelength parasitic folds that occur at the axes of the large folds are sheared in the direction of flow (Figure 6c).Such recumbent polyharmonic folds typically form in shear zones where instabilities develop at the boundaries of adjacent layers with different viscosities (Ramsay & Huber, 1987).This interpretation is consistent with our observations of folding at the interfaces between the laminated subfacies and the englacial ice and within the laminated subfacies where clean clear ice is interbedded with debris-bearing ice (Figure 4).The recumbent geometry of the folds, together with the presence of numerous parasitic folds that appear to be sheared at the fold hinges (Figure 6c), suggest a sinistral rotation of objects in the basal ice (i.e., anticlockwise) and progressively attenuation that results in the pervasive laminar structure of the basal zone.In the predominantly compressive flow regime combined with pervasive shearing that characterizes the glacier margin, the formation of folds leads to shortening and thickening of the basal layer.
Another source of flow instabilities in the basal zone consists of perturbations associated with pebble to bouldersized particles that occur throughout the basal ice.Since these particles rest in a deforming medium in which deformation increases from bottom to top, the particles bridge a velocity gradient which results in sinistral rotation.In shear zone, rotation occurs at a rate that can be significantly lower than the instantaneous shear strain rate (Simpson & De Paor, 1993).The combination of rotation of the particles and ice flowing over their upper surfaces has led to the development of morphologies similar to sigma grains in other geological materials.These particles are characterized by coatings of fine particles, striae in multiple directions, and pressure shadows at the stoss and lee sides of the grains.In the case of particles >10 mm in diameter, air-filled cavities have developed on both the stoss and lee sides of the clasts (Figure 6b).Cavities on the margins of boulder-sized particles were characterized by grooved linear markings on the cavity roofs, all aligned with the ice flow direction.These features resemble slickenslides in deformed rocks where polished and striated surfaces result from friction produced at fault surfaces.The slickenslides are a series of grooves which are impressions of the asperities of the particles that the ice has flowed over.Some of these cavities degassed when punctured, that is, the gas was above atmospheric pressure.Such cavities have also been observed at the base of Suess Glacier, a small cold-based alpine glacier located about 12 km east of Taylor Glacier (Fitzsimons et al., 2000).We did not observe any evidence to suggest that clasts that protruded through the boundaries were plowing.
Thin sections of small clasts (<10 mm) from the laminated subfacies show that the pressure shadows consist of clear ice, which may be the product of recrystallization (Samyn, Fitzsimons, & Lorrain, 2005;Samyn et al., 2009).Ribbons of fine, elongated ice crystals were observed at the boundaries of debris-bearing and clear laminae identified in microstructural analysis (Samyn et al., 2008).These ice crystals show clear signs of differential strain and dynamic recrystallization (Samyn et al., 2008), which might have been favored by the occurrence of low viscosity interfacial films.These patterns are consistent with the high rates of distributed shear and relatively low strength of the debris-bearing facies identified in Section 3.3.

Is Sliding at −18°C Plausible?
The two offsets in the velocity profile amount to 0.204 m.a 1 (0.130 m at 2.7 m and 0.074 m at 1.7 m above datum, Figure 8), which corresponds to 14% of motion within the basal zone and about 8% of glacier motion at the ice edge (using 2.5 m a 1 as the ice edge surface velocity).Taken together with the short-term point measurements of displacement above boulders and the presence of cavities and slickenslides, these observations suggest two possibilities: either sliding or rapid deformation in thin (<0.1 m) zones.
There is both theoretical and field evidence for sliding in the basal zone at subfreezing temperatures.Gilpin (1979) argued for the existence of liquid-like layers between the ice and the substrate, which led Shreve (1984) and Fowler (1986) to suggest that sliding is possible at subfreezing temperatures.Subsequent sliding at has been observed or inferred at temperatures ranging between 1 and 17°C in several glaciers (Cuffey et al., 1999;Echelmeyer & Zhongxiang, 1987;Fitzsimons et al., 2000).In Meserve Glacier, a small alpine glacier in the Dry Valleys with a basal temperature of 17°C, displacement measurements at the ice boundary layer over boulders led Cuffey et al. (1999) to conclude that ice was sliding over the boulders at velocities between 0.002 and 0.008 m a 1 .The sliding was associated with lee cavities on boulders which carried slickenslides like the ones described in this study.Cuffey et al. attributed the sliding at these low temperatures to the presence of interfacial films.Our point measurements on boulders range from 0.046 to 0.167 m a 1 and the two offsets in the vertical profile at the upper and lower boundaries of the massive subfacies are 0.130 and 0.074 m a 1 respectively.If these displacements are from sliding, as seems plausible, the velocities are more than an order of magnitude greater than those observed at Merserve Glacier at the same temperature.An important difference between the two sites is that the suspended sediment and solute concentrations in the basal ice at Taylor Glacier are considerably higher than in Meserve Glacier (Holdsworth, 1974;Samyn, Svensson, et al., 2005).The ice is around 20 m-thick at both locations.However, interpretation of sliding behavior remains uncertain and the observed behavior might be explained by localized strain within a shear band enhanced by the presence of liquid water.
The theory of premelting describes the formation and existence of liquid at temperatures below the solid region in a bulk phase diagram (Wettlaufer, 1999) and provides a useful theoretical context for field observations of subfreezing sliding.Premelting refers to the process in which a liquid or a liquid-like film is present at the surface of a crystal in contact with its vapor phase below the bulk freezing point.At the boundary of rock particles and ice crystals, it is known as interfacial premelting (Dash et al., 1995).The thickness of liquid films is extremely sensitive to the concentrations of solutes (Wettlaufer, 1999), and potentially to the nature of debris entrained within the ice (Dash et al., 1995).In Taylor Glacier, the average total dissolved solids in the laminated and massive subfacies are two and three orders of magnitude greater respectively than that of the englacial ice.It is likely that the solute values are conservative because our samples were 0.010 m wide and 0.1 m long and provide average values through multiple laminae that contain clear ice layers.
The high sediment concentrations in the massive subfacies together with the large component of fine sand and silt may also be conducive to sliding because finer-grained sediments are associated with depressed freezing points due to greater interfacial water curvature.An experimental study of sliding behavior by Emerson and Rempel (2007) pointed to particle size and concentration controls on temperate sliding behavior.They distinguish a "sandy" regime in which there is relatively high resistance to sliding associated with higher debris concentration and larger particle sizes.By contrast, when debris concentrations are lower and particle sizes are smaller, there is an abrupt transition into a "slippery" regime in which shear resistance is no longer dependent on normal load.The physical characteristics of the debris-rich basal ice in Taylor Glacier, together with the high solute concentrations (Table 2) and presence of fine-grained particles (Figure 8), seems to be compatible with larger volumes of premelted liquid and lower friction.Further evidence for the presence of premelted liquid in the basal ice of Taylor Glacier was provided by Samyn, Fitzsimons, and Lorrain (2005) who suggested that subtle changes in the gas composition were typical of phase changes involving minute quantities of water.These authors also provided crystallographic evidence of lattice loosening, indicating small-scale strain variations that are compatible with slight changes in interfacial water content.Finally, Souchez et al. (2004) proposed a co-isotopic model where the basal zone formation can be explained by interfacial premelting without apparent fractionation.

Origin and Evolution of the Basal Zone
The basal ice facies observed in tunnels in Taylor Glacier bear some striking similarities to the model of basal ice formation under freezing conditions proposed by Christoffersen et al. (2006) and to a thermomechanical treatment of periodic formation of ice lenses proposed by Meyer et al. (2023).Christoffersen et al. (2006) developed a numerical model for basal ice formation based on the ratio of the supply of subglacial water to the freezing rate in which the following types of ice are produced depending on the subglacial hydrology: clear ice is produced when water is freely available; laminated ice if the supply of water is constrained; massive ice if the water supply is further constrained; and a solid facies if meltwater is depleted and there is rapid freezing.Meyer et al. (2023) describe the thermomechanics of liquid water flow and freezing in ice-saturated sediments that account for the development of pure ice lenses that can develop in subglacial environments.The massive subfacies observed beneath Taylor Glacier is structurally very similar to these models because it consists of finely stacked laminae and lenses that appear to be massive from visual observation (Figures 4c and 4d).A regelation origin for the massive and laminated subfacies is consistent with many lines of evidence for the presence of liquid water beneath Taylor Glacier (Badgeley et al., 2017;Hubbard et al., 2004;Lyons et al., 2019;Mikucki et al., 2015) and with cooling of the base of the glacier as it thins toward the margin.Laminated basal ice has been described in a number of field settings including at Variegated Glacier (Sharp et al., 1994), Russell Glacier, Greenland (Waller et al., 2000), and borehole video from Antarctic ice streams (Carsey et al., 2002) and is attributed to a range of thermomechanical phenomena involving ice-rock-water interactions.
A significant difference with these regelation models, proposed by Christoffersen et al. (2006) and Meyer et al. (2023), is that the Taylor Glacier basal ice layer consists of the massive (debris-rich) ice interbedded with clean bubbly englacial ice (Figure 8).This interbedding, together with abundant evidence of sinistral rotation described above, the presence of recumbent folds and the development of parasitic folds suggests that tectonic processes play a role in basal ice formation.The occurrence of tight recumbent folds in the laminated subfacies (Figure 6c) and the interbedded nature of the laminated subfacies with the englacial facies suggests that debris bearing-ice is being mixed with adjacent relatively clean ice.Such mixing is expected at structural boundaries where there are rheological contrasts, an association that is well known to lead to flow perturbations in shear zones (Passchier & Simpson, 1986;Ramsay & Huber, 1987).These perturbations first lead to basal zone thickening due to the compressive component of flow and then to stretching and thinning as a result of the shearing component.Because of the pervasive simple shear in the basal zone, the folds are prone to be rapidly attenuated into the planar structures that characterize the bulk of ice in the basal zone.From the observed deformation structures, together with the patterns of deformation, we conclude that there is tectonic evolution of the basal ice along the flow path toward the glacier margin.We suggest that the laminated subfacies is at least partly a tectonic facies that is derived from mechanical deformation at the upper and lower boundaries of the massive subfacies.This tectonic origin is further supported by gas chromatography (Samyn, Fitzsimons, & Lorrain, 2005) and water co-isotopic measurements (Souchez et al., 2004), both conducted at the cm-scale within the basal ice zone, showing that the clean and debris-bearing ice layers from the laminated subfacies present a "meteoric" signature, thereby precluding their origin as resulting from the macro-scale regelation process.The clearly documented deformation structures described in Section 3.2 show that there is an unambiguous tectonic overprinting of the accreted ice.
The tunnel walls show that the basal ice layers rise toward the ice margin and that at some locations there is an abrupt termination of the basal debris layers before reaching the ice cliff (Figure 2).This termination appears to be the product of partial overriding and entrainment of the ice apron that forms at the foot of the terminal cliff in places.The process of apron entrainment has been described by Shaw (1977) and Fitzsimons et al. (2008).

Implications for Glacier Behavior
Displacements measured in the tunnel show that the basal zone contributes about 1.5 m of motion per year, which is about 60% of the 2.5 m a 1 surface velocity at the ice edge.If the targets drilled into the bed that have been used as measurement datums are moving, a possibility described in Section 2.3, these estimates are minimum values.The same value of basal motion was observed in "ice-laden drift" with a temperature of 4°C in Urumqi No 1 Glacier (Echelmeyer & Zhongxiang, 1987).Similar results have also been reported from Russell Glacier, a polythermal outlet glacier from Greenland where 16% of the motion was attributed to deformation of basal ice (Waller & Hart, 1999).At Taylor Glacier, a previous study (Pettit et al., 2014) used a two-layer flow model to predict the basal properties of and concluded that deformation of the basal ice layer accounted for 85%-98% of glacier motion, that the basal ice layer was likely 10-15 m thick, and that the basal ice was 20-40 times softer than clean englacial ice.While the measurements made in the tunnel are broadly consistent with the argument that the basal ice layer plays a significant role in the behavior of Taylor Glacier, the basal ice layer that we have characterized is thinner, does not account for such a large proportion of surface motion and is only 2 times softer than the values for the basal layer inferred by Pettit et al. (2014).These differences highlight the need for a thorough characterization of basal ice facies in ice flow models.
One of the issues in examining the behavior of ice and debris at the base of glaciers is that there is a continuum between debris-rich basal ice and ice-rich debris that might form the bed of a glacier.What one researcher might describe as basal ice another might describe as ice-rich sediment or frozen till.This problem is exacerbated when observations are made using bore holes because of uncertainties concerning the location of the glacier bed and because the holes cannot penetrate coarse sediment, unless local conditions allow diamond drilling equipment to be deployed (e.g., Truffer et al., 2000).For example, work by Echelmeyer and Zhongxiang (1987) is widely cited as evidence for subglacial deformation under subfreezing conditions.However, their description of the contact between the glacier bed and the substrate is a boundary between clear glacier ice with a low debris content and the bed, which contains 21%-39% debris.In the Taylor Glacier, such a boundary occurs between the englacial ice and the laminated subfacies, which is contained entirely within the debris-bearing basal ice.Although the tunnelbased observations at Taylor Glacier leave little doubt that the measurements have been made within the basal ice layer (Figure 4), it is clear that the displacement profile resembles that of a subglacial sediment deformation profile (e.g., Boulton, 2006;Boulton & Hindmarsh, 1987) because the strain is heterogenous and the velocity profiles are stepped owing to the presence of offsets due to sliding or localized shear.The observations made in the tunnels support the view that the base of glaciers should be defined as a zone and not a single zero-velocity boundary at a simple ice-substrate interface (Fitzsimons, 2006;Pettit et al., 2014;Waller, 2001).

Implications for Glacial Geology
Our measurements show that the basal zone has an emergence velocity (i.e., the velocity at which it the debris reaches the ice edge) of around 1.5 m a 1 .The average debris concentration through massive subfacies is 26% by volume for a 1.2 m-thick layer and for the laminated subfacies is 12% vol.for 1.6 m-thick layer, which together yields 0.5 m 3 a 1 of debris per 1 m of ice margin per year.These estimates are half of the estimate made by Pettit et al. (2014).The sources of the differences in our calculations are that the emergence velocity is lower and the thickness of the basal layers substantially smaller (4.5 vs. 10 m), although the measured debris concentrations in the basal ice are much greater than those suggested by Pettit et al. (2014).The volume of debris that is discharged to the ice margin is not consistent with the modest volumes of material that have accumulated along the ice edge and the absence of moraines (Figure 1).There are no substantial accumulations of material at the ice margin because fluvial processes rapidly remove most of the debris that is discharged from the ice cliff.A steep ephemeral marginal meltwater stream, Santa Fe Creek, is pinned against the ice margin by the adverse slope adjacent to the ice edge (Figure 1).Consequently, all the fine-grained sediment excavated from the basal zone was removed and transported into Lake Bonney within 1 year of the excavations.A boulder and cobble lag were all that remained.An exception to this process is the small ice cored moraine that rests on the delta of Santa Fe Creek adjacent to the glacier terminus.Basal ice is preserved in the core of this moraine, and ablation of the ice has produced a silty till with numerous striated clasts.This material is indistinguishable from the basal ice observed in the tunnels.
A combination of high debris concentrations in the debris-bearing ice and the high shear experienced within the ice has produced facetted and abraded clasts and a high proportion of silt-sized particles in particle size distributions.These characteristics are similar to the sedimentary signature of polythermal glacial environments which are dominated by high rates of abrasion and contrast strongly with the sedimentary signature of cold-based glaciers which are dominated by sandy gravel, glaciotectonically deformed permafrost and aeolian deposits (Fitzsimons & Howarth, 2020;Hambrey & Fitzsimons, 2010;Hambrey & Glasser, 2012).These observations highlight the complexity of basal ice processes at the margin of Taylor Glacier and that simple sedimentological criteria may not always be a sound basis for reconstructing the thermal regimes of glaciers.

Conclusions
The basal zone of Taylor Glacier is characterized by a complex strain distribution that results from heterogeneous deformation within the basal zone.The strongest materials and lowest rates of deformation occur in ice layers with low debris concentrations and the weakest materials and highest rates of deformation occur in layers of ice with the highest debris concentrations and high solute loads.Ice with a laminated appearance has an intermediate strength.
Deformation of the whole basal zone accounts for 60% of glacier motion at this ice marginal location.Velocity profiles determined using plumblines suggest that either the ice is sliding at structural interfaces or that high rates of shear are localized to narrow (<0.1 m-thick) zones accounting for 14% of glacier motion.Short term measurement of displacements using engineer dial gauges (described in Section 3.4.1),together with the development of cavities and slickenslides in the ice support the interpretation of sliding of up to 0.167 m a 1 or very localized interfacial flow despite the low basal temperature ( 18°C).These observations are consistent with studies of ice crystallography, which show evidence of pervasive shearing and strain localization at sedimentological and structural interfaces.
Heterogeneous deformation results in flow perturbations that cause folding, which drives mixing within the basal zone and results in blending of debris-bearing and clean ice facies.Mixing of adjacent ice facies results in tectonic evolution of the basal ice along the flow line.Basal zone thickening toward the of the glacier occurs initially as a result of folding, which is then attenuated by horizontal stretching as a result of simple shear.Deformation measurements together with observation of folding and shearing suggest that a significant part of the laminated subfacies has been produced by mechanical deformation within the basal zone.
High rates of strain combined with high debris concentrations in the debris-bearing ice produce a high abrasion environment, which is consistent with abundant heavily facetted and striated clasts and a strong silt mode in particle size distributions.The combination of the sedimentary signature of the abraded and striated particles, the production of high volumes of silt-sized particles and the relatively high sediment flux of the basal zone are characteristics normally associated with temperate glaciers.However, the structural and geochemical signature of the basal profile reflects the cold nature of the marginal ice.
Finally, our observation of the structure, composition and deformation of basal ice in Taylor Glacier support the view that glacier beds are zones in which deformation is spatially and temporally variable.Our observations and measurements show that the glacier bed is not a single zero-velocity boundary and suggest that glacier flow models need to incorporate rheologically and structurally distinct layers to capture spatially variable behavior of debris bearing basal ice.

Figure 1 .
Figure 1.(a) Location map of Taylor Glacier showing the location of the tunnels (T1 and T2) and the velocity distribution over the terminus area.Base image: Google Earth, CNES/Airbus, image date: 22 November 2016.The surface velocity measurements are annual displacements acquired 4 years before the tunnel excavation.(b) Elevation of the profile of the second tunnel excavated in 1999 showing the geometry of the debris bands in the light gray tone within the tunnel profile.The dashed line is the approximate location of the glacier bed.(c) Plan of the second tunnel showing the local direction of ice flow.Santa Fe Creek is a stream that is parallel to the ice margin.

Figure 2 .
Figure 2. (a) Entrance to the second tunnel after excavation showing basal debris layers rising toward the ice edge.The abrupt termination in the debris layers is the contact between the basal ice and the marginal ice apron.A 0.5 m-diameter boulder rests on the top of the basal debris layer in the upper left corner of the photograph.(b) Temperature measured using thermocouples placed in holes drilled 0.2 m into the tunnel wall from the entrance to the end of the tunnel.

Figure 3 .
Figure 3. Monitoring ice deformation.(a) Plumbline and wooden pegs in the shaft excavated at the end of the tunnel.(b) Strain arrays of four stainless steel coach screws in the upper left and lower center located at the interface between debris bearing and clean ice.(c) Direct shear device used for in situ strength tests.At left is a Wykeham Farrance direct shear machine with a 2.5 tonne load cell between the stepper motor and the moving stainless steel plate.The vertical load is applied by a truck pneumatic shock absorber (out of view on the top of the vertical pole) separated from the top of the sample with a 1.5 tonne load cell.(d) Stratified debris-bearing ice sample tested in the direct shear device.

Figure 4 .
Figure 4. Basal ice facies exposed in the tunnel walls.(a) Laminated subfacies interbedded with the englacial facies.The two bolts shown in the upper part of the photograph are part of a strain array.(b) Laminated subfacies with folded and sheared layers of clean bubble-free ice.(c) Massive facies with clear ice lenses and small pebbles with stoss and lee-side cavities.(d) Massive facies with small pebbles.Ice flow is from right to left in all photographs.

Figure 5 .
Figure 5. Characteristics of debris in the basal ice and adjacent surficial sediments.(a) Randomly oriented striae on a polished dolerite clast (b) Randomly oriented striae on a gneiss clast.The striae appear "fresh" judged from the presence of powdered rock debris adjacent to the striations.Both cobbles are from the top of the main debris band 3-3.5 m above the datum.(c) Particle size distribution from the center of the massive subfacies debris band 2.8 m above the shaft datum.(d) Particle size distribution from the top of the main debris band 3.4 m above the datum (e) Particle size distribution from the laminated subfacies facies 3.6 m above the datum.(f) Particle size distribution from the stream channel adjacent to the ice margin.(g)Roundness of pebbles from the main debris band classified using Powers roundness classes(Powers, 1953).(h) Predominantly subangular pebble-sized particles retrieved from melting 10 kg of ice from the massive subfacies 3 m above the datum.

Figure 6 .
Figure 6.Deformation structures in the laminated subfacies.(a) Thin section under polarized light showing clean bubbly ice above and below a debris-rich layer of the laminated subfacies.Ice crystals in the debris-bearing ice are less than 1 mm in diameter and the clean ice has crystals 5-10 mm in diameter.(b) Prolate-shaped particles in the debris-bearing ice with characteristic stoss and lee-side air filled cavities.(c) A tight polyharmonic recumbent fold with multiple shear bands extending into the overlying ice.(d) Box and whisker plot of the shear strain rate showing median and upper quartile ranges and the maximum and minimum values.The ice flow is from right to left in all photographs.

Figure 7 .
Figure 7. Result of a triaxial test of a sample of the laminated subfacies showing failure localized to an ice-rich layer in the laminated subfacies.This test was undertaken with the sample in an air-filled pressure vessel maintained at 200 kPa and 18°C for the duration of the 12 hr test.

Figure 8 .
Figure 8.(a) Stratigraphic log of the shaft excavated at the end of the tunnel, showing clean glacier ice underlain by and interbedded with stratified ice consisting of laminated and massive subfacies.Granite boulders in the ice have lee and stoss-side air filled cavities shown as black zones.Within the upper two units of laminated ice between 3 and 4 m there are approximately 25 debris-bearing layers separated by clear ice layers.The laminated ice between 1.4 and 1.8 m contains about 15 layers of debris bearing ice separated by clear ice layers.(b) Velocity measured with a plumbline surveyed episodically up to 403 days after installation.The two steps in the profile at 1.7 and 2.7 m above the datum occur at the interfaces between the massive and laminated subfacies.

Figure 9 .
Figure 9. Evidence of the differential and localized deformation in the tunnel.(a) Photograph of the end of the tunnel 403 days after excavation.The walls of the shaft were originally planar and vertical.Most of the closure has occurred at the interface between the laminated and massive subfacies in the lower part of the photograph.The end of the ladder is 0.6 m wide.(b) Closure of 59 mm diameter core holes in the massive subfacies 4 days after the samples were cored.(c) Closure of the crosssection of the shaft excavated at the end of the tunnel 403 days after excavation.(d) Bulging of the rear wall of the tunnel surveyed 403 days after excavation.

Table 1
Basal Ice Facies Characteristics a The vertical arrangement of the facies is described in Figure8.
a FITZSIMONS ET AL.

Table 2
Average Solute Concentrations for the Ice Facies a a Values in parts per million.b Partial measure of Total Dissolved Solids.

Table 3
Peak Shear Strength Values for Ice From the Basal Zone of Taylor Glacier FITZSIMONS ET AL.