Introduction

In 1952, John Glen published results of his uniaxial-compression experiments on ice. The resultant flow rule (Glen, 1952, 1955) provided the means to calculate the velocity of a glacier for the first time (Nye, 1952). His experiments helped establish modern glaciology (Clarke, 1987c), and although his flow rule continues to be revised and its limitations better understood, it remains a cornerstone of the discipline.

His results and their lasting impact illustrate the potential utility of laboratory experiments in glaciology. Although experiments are usually viewed as the gold standard for hypothesis testing in physics and chemistry, their perceived importance varies greatly across geoscience disciplines. Critics cite the scale and complexity of open natural systems and conclude, as a result, that Earth processes cannot be fully simulated in the laboratory. These critics are correct but miss the point. Well-designed experiments do not attempt to fully simulate natural conditions. Rather they attempt to isolate and thereby explain important phenomena that cannot be isolated in the field. Through known boundary and initial conditions, complete knowledge of parameter values, control of independent variables and the capacity for true reproducibility, laboratory experiments offer clear advantages that complement full-scale but less controlled field studies.

This review describes laboratory experiments in glaciology, with an emphasis on recent work. It is intended to highlight the contributions of experiments, address their limitations and point to gaps in knowledge that potentially could be filled experimentally. The discussion will be limited to work on glacial processes; for example, experiments that address periglacial processes and sea-ice mechanics will not be discussed.

91.2 Laboratory experiments in glaciology

91.2.1 Ice deformation and structure

The impact of laboratory experiments in glaciology has been greatest in the study of ice rheology and structure. Experiments provide the surest means of determining constitutive relations for ice deformation that when combined with classic conservation rules provide the basis for modelling glacier flow. Relative to rocks in Earth's crust, glaciers are homogeneous and deform at high rates. Laboratory studies are, therefore, more readily applied to glacier flow than to crustal deformation. However, despite the relative homogeneity of glacier ice and its rapid deformation, laboratory studies continue to demonstrate the complexity of ice deformation and to be seriously limited by time-scales for deformation that are too long to be explored in the laboratory. Thus, as was noted by Kamb (1972), laboratory studies of ice deformation, in addition to their relevance to glacier flow, serve as a warning to overly simplistic interpretations of experimental creep results for more complicated geological materials.

Field measurements have also been influential in the study of ice rheology and structure. In such studies, however, the state of stress is usually more complicated than is desirable and usually cannot be measured, requiring estimation of stresses with simplified models. Moreover, because more than one ice property (e.g. temperature, crystal fabric, impurities) commonly changes with depth in glaciers and ice sheets, isolating the effect of a single variable on ice flow is difficult.

Most experiments with polycrystalline ice have been simple in design. Either synthetic ice, usually with initially random c-axis orientations, or glacier ice, usually with some anisotropy, is squeezed. Experiments are normally carried out under a constant stress and under temperatures regulated to 0.1-1.0°C in a cold room or fluid bath. Stress is most commonly applied in uniaxial compression but also in shear and in combined compression and shear, depending upon the objectives of the study. Experiments are conducted over periods ranging from weeks to over 2yr (e.g. Jacka, 1984b) with the goal of reaching steady-state deformation or at least the minimum strain rate that marks the beginning of tertiary creep.

One of the most important realizations of the last few decades is that no simple flow rule can adequately characterize deformation of ice in glaciers (Lliboutry & Duval, 1985; Alley, 1992). Experiments indicate that the original rule of Glen (1952, 1955) really represents part of an amalgam of flow rules that apply over different ranges of stress and strain rate. A generalization of Glen's flow rule can be written as e = EA exp(-Qyn (1)

where e and t are the strain rate and differential stress, respectively, A is constant for clean, isotropic ice, E is an enhancement factor that accounts for ice softening due to crystal fabric, impurities and other factors, Q is the activation energy for creep, T is the homologous temperature, R is the universal gas constant and n describes the sensitivity of the strain rate to the stress.

The value of n in Equation (1) is traditionally taken to be a constant equal to 3, based on assessments of both laboratory tests performed primarily at high stresses (>0.1 MPa) and field data (Paterson, 1994). Early laboratory studies conducted at low stresses indicated n < 2 (e.g. Mellor & Testa, 1969), but these studies may not have been carried out to sufficient strains to be indicative of steady-state creep, rather than transient creep (e.g. Weertman, 1983). However, strong laboratory evidence has accumulated more recently that, at stresses less than about 0.2MPa, n is somewhat less than 2 (Pimienta & Duval, 1987; Goldsby & Kohlstedt, 1997, 2001, 2002; De La Chapelle et al., 1999; Duval et al., 2000; Duval & Montagnat, 2002)—a result also supported by some borehole-deformation measurements (Dahl-Jensen & Gundestrup, 1987). This result is important because deviatoric stresses less than 0.1 MPa are typical of ice sheets. Experiments indicate that n = 4 at stresses greater than about 0.5 MPa (e.g. Barnes et al., 1971), but such high deviatoric stresses are not usually relevant to glaciers (Paterson, 1994).

Although the lower value of n at low stresses is relatively well accepted, deformation mechanisms responsible for the low value are controversial. Goldsby & Kohlstedt (1997, 2001), aware of the difficulty of reaching steady deformation rates at low stresses, used an innovative sample-preparation procedure to produce very fine-grained ice (3-200mm). Deformation mechanisms at low stresses commonly result in strain rates that depend inversely on grain size, so reducing grain size better assured that steady-state deformation would be achieved. Goldsby & Kohlstedt found in experiments conducted over a wide range of stress that the low value of n (1.8) was due to basal slip rate-limited by grain-boundary sliding—the deformation mechanism associated with so-called 'superplastic flow'. This differs from the traditional view that dislocation creep is the rate-limiting deformation mechanism. Features indicative of grain-boundary sliding, such as straight grain boundaries, equant grains and four-grain intersections, developed during deformation. High-stress tests yielded results consistent with n = 4, as expected for deformation purely by dislocation creep. These data indicate that Glen's rule with n = 3 may reflect experiments conducted at stresses near the transition between the grain-boundary-sliding and dislocation-slip regimes (Fig. 91.1). Peltier et al. (2000) provocatively assert that an implication of these results is that rates of glacier flow may be underestimated in models by as much as one to two orders of magnitude.

Duval and colleagues (Duval et al., 2000; Montagnat & Duval, 2000; Duval & Montagnat, 2002), however, argue that the low value of n for the coarser-grained ice of glaciers is the result of dislocation creep accommodated by grain-boundary migration. By absorbing dislocations, grain-boundary migration may reduce work hardening that results from local accumulations of dislocations that develop during slip on basal planes. Evidence in support of this viewpoint is that traditional deformation mechanisms, which involve crystal lattice rotation by dislocation slip, have been

Differential stress, MPa

Figure 91.1 Reinterpretation of data from creep experiments on coarse-grained ice at temperatures near the melting point, based on the laboratory results of Goldsby & Kohlstedt (1997, 2001). (Modified from Peltier et al., 2000.)

used to successfully model the development of fabric in ice sheets (e.g. Duval et al., 2000; Montagnat & Duval, 2000). Deformation involving grain-boundary sliding, however, can also cause fabric development because strain that is rate-limited by grain-boundary sliding can accrue primarily from dislocation slip along basal planes (Goldsby & Kohlstedt, 2002). Studies of calcite rocks that develop fabric during superplastic deformation bear this out (e.g. Rutter et al., 1994). However, empirical studies of ice fabrics that result from deformation that is rate-limited by grain-boundary sliding have not been conducted, and, unlike the case for traditional dislocation creep, observed ice-sheet fabrics have not yet been modelled successfully with grain-boundary sliding. Until such studies are conducted, the Goldsby-Kohlstedt hypothesis will probably remain controversial.

This controversy illustrates both the value and limitations of experimental approaches. Use of synthetic fine-grained ice in the experiments of Goldsby & Kohlstedt (1997,2001) allowed steady-state deformation to be achieved unequivocally, revealing a potentially important deformation mechanism for ice in glaciers. On the other hand, application of the results to ice sheets required an uncertain extrapolation to large grain sizes. Despite this uncertainty, the overall result is positive. Old concepts thought to be quite sturdy are being re-evaluated. For example, there was generally thought to be no influence of ice grain size on deformation rate (Jacka, 1984b; Budd & Jacka, 1989). If grain-boundary sliding, however, is the rate-limiting process, deformation rate should be inversely related to grain size (Goldsby & Kohlstedt, 2001). Thus, this hypothesis provides new context for field studies that demonstrate grain-size dependence (Cuffey et al., 2000c)

and/or that seek to infer deformation mechanisms from microstructures.

Other recent laboratory experiments have been less controversial but also innovative and important. De la Chapelle et al. (1999) made synthetic ices of different salinities to study the effect of water in ice on its creep rate, extending the work of Duval (1977). Ice consisting of 7% water increased strain rates by more than an order of magnitude relative to pure ice, over stresses ranging from 0.02 to nearly 1MPa. Softening was attributed to reduction of stress concentrations at grain boundaries that allowed more deformation to occur by easy dislocation slip along basal planes. Thus, by increasing the volume of water at grain boundaries, impurities that are ubiquitous and highly variable in ice cores may help soften ice. These results also may bear on the low effective viscosity of ice near the beds of temperature glaciers, where water contents may exceed 2% (Cohen, 2000).

In experiments conducted in combined compression and shear on initially isotropic ice, Li et al. (1996) demonstrated that minimum creep rates prior to the development of flow-induced anisotropy are independent of stress configuration. This result supports the fundamental assumption that deformation of isotropic ice is dependent only on the second invariant of the stress tensor (e.g. Nye, 1953). Not surprisingly, steady-state tertiary creep rates that were attained as ice acquired a fabric depended strongly on stress configuration. For simple shear without compression, enhancement factors of10 were indicated, consistent with previous experiments (Shoji & Langway, 1988). Subsequent experiments conducted to strains greater than 100% produced single-maximum fabrics with c-axes concentrated perpendicular to the shear plane, similar to fabrics observed deep in polar ice sheets (Li et al., 2000). Most previous simple-shear experiments had produced double-maximum fabrics (Kamb, 1972; Gao et al., 1989), presumably because they were terminated at too low a strain.

91.2.2 Till deformation

Widespread recognition in the 1980s that fast glacier flow might depend on shear deformation of till beneath glaciers provided initial stimulus for experimental work on the mechanical properties of till. At that time, various till rheological rules were beginning to be used in bed-deformation models of glacier flow (e.g. Alley, 1989), usually with the assumption that till obeyed a fluidlike viscous or viscoplastic rheology. Apparent support for this assumption came from deformation profiles measured in subglacial till by Boulton & Hindmarsh (1987) at BreiSamerkur-jokull. Since that time, however, deformation profiles measured in that study have been shown to not be uniquely indicative of a particular till rheology (Tulaczyk et al., 2000a; Iverson & Iverson, 2001). This non-uniqueness highlights the difficulty of making definitive interpretations regarding till mechanical properties from measurements of till strain made subglacially, where the state of stress varies spatially and temporally and is difficult to either measure or estimate.

A wide variety of equipment and procedures have been used, to date, in experiments aimed at assessing the rheology of water-saturated till. Kamb (1991) conducted direct-shear tests on finegrained till collected from the bed of Whillans Ice Stream.

200 400 600

Shearing rate, m yr-

200 400 600

Shearing rate, m yr-

1000

ra w ee

ra s cc ra

b

-

■ft ■

i}

-

Whillans Ice Stream till

1000

Shear strain rate, yr1

Figure 91.2 (a) Steady-state ratio of shear stress to effective normal stress as a function of shearing rate, as measured in ring-shear experiments on remoulded basal till of Storglaciaren (4% clay, 21% silt, 75% sand and gravel), the Two Rivers till of the Lake Michigan lobe (32% clay, 30% silt, 38% sand and gravel), and the basal till of the Des Moines Lobe (16% clay, 36% silt, 48% sand and gravel). Shear rate is the rate of displacement across a shear zone 10-30mm thick. Effective normal stresses were 20-150kPa. (Modified from Iverson et al., 1998.) (b) Ratio of shear stress to effective normal stress as a function of shear strain rate, as measured in triaxial experiments on the basal till of the Whillans Ice Stream (35% clay, 23% silt, 42% gravel). Different symbols indicate different effective normal stresses (25-320kPa) (modified from Tulaczyk et al., 2000a). In both (a) and (b) there is no increase in the ratio of shear stress to effective stress with strain rate that would indicate viscous deformation resistance.

Direct-shear experiments have the great advantage of being standard and easy to perform. Iverson et al. (1997, 1998) constructed a large ring-shear device (specimen chamber, 0.6m o.d., 0.125m width) and used it to study the mechanical behaviour of basal tills with different grain-size distributions. This device allowed very high shear strains, so that steady-state deformation was ensured, and allowed measurement of all boundary stresses. Tulaczyk et al. (2000a) constructed a smaller ring-shear device and also used triaxial and uniaxial testing equipment to study more thoroughly the till tested by Kamb (1991).

Despite the different tills and equipment used in these studies, all results indicate that the steady-state shearing resistance of till is extremely insensitive to strain rate (Fig. 91.2a & b) and that shearing resistance varies linearly with effective normal stress. These are the properties of a Coulomb (frictional) plastic material. Till, therefore, like granular materials in general, does not exhibit intrinsically viscous or Bingham-viscous behaviour, in which shearing resistance increases with strain rate. In contrast with these results, Ho et al. (1996) inferred mildly non-linear viscous behaviour from results of stress-controlled direct-shear experiments. Inspection of data from these tests (Vela, 1994), however, indicates that steady strain rates were not attained at a given stress; strain rates were still decreasing when a new larger stress was applied. Use of these data in models of glacier motion on soft beds is dubious (e.g. Licciardi et al., 1998).

Although laboratory results indicate that Coulomb models for till are appropriate for steady deformation (so-called critical-state deformation in which porosity, shearing resistance, and strain rate remain constant), laboratory experiments indicate that transient shearing resistance of till can depend strongly on deformation rate. Moore & Iverson (2002) conducted ring-shear experiments on overconsolidated tills that dilated during the initial stages of shear. Shear stress was held constant and normal stress was reduced until friction within the till was decreased sufficiently to initiate shear deformation. Rapid dilation, as shear rate increased, reduced pore-water pressure, thereby strengthening the till. This apparent viscous response, called dilatant strengthening, occurs when pore-pressure diffusion toward opening voids cannot keep pace with the rate at which porosity increases during shear. Dilatant strengthening eventually slowed deformation, and the consequent decrease in dilation rate allowed pore-water pressure to slowly increase. Pore pressure eventually increased sufficiently to trigger another shearing episode that was again arrested by pore-pressure decline—a cycle that repeated up to 10 times until the to

"O

1000

Des Moines lobe basal

0.00

0.05

0.10 0.15 Dilatancy

Figure 91.3 Transient shearing velocity as a function of dilatancy averaged for two tills (see Figure 91.2a) over the durations of creep episodes in stress-controlled ring-shear experiments. Dilatancy is the ratio of shear-zone thickening to shearing displacement; this ratio depends inversely on porosity. D is hydraulic diffusivity under an effective normal stress of 25kPa. (From Moore & Iverson, 2002.)

In addition to hydraulic effects, alignment of till particles during deformation can be used to infer the extent to which a basal till of the geological record has been sheared. Hooyer & Iverson (2000a) used a ring-shear device to study the fabric formed by rigid, elongate particles in both till and viscous putty at various total shear strains up to 475. The results with putty agreed with the theory of Jeffery (1922), who predicted that elongate particles in a shearing fluid orbit periodically but reduce their rotation rates when oriented near the shear plane. In tests with till, however, particle behaviour was strikingly different. Rather than rotating continuously, clasts rotated into the shear plane at shear strains of approximately 2 and remained there, resulting in a strong fabric (Sj eigenvalue of ca. 0.8) that persisted to high strains. This behaviour was caused by slip of the till matrix along clast surfaces, an effect not considered in the theory of Jeffery (1922). These results indicate that if total bed shear strain is high, strong particle fabrics result, contrary to some field inferences (e.g. Hart, 1994).

till reached its critical-state porosity and could not dilate further, resulting in catastrophic failure. Time averaged shear velocities prior to catastrophic failure depended inversely on the magnitude of pore dilation with shear and were significantly lower for finegrained till than for coarse-grained till, owing to different rates of pore-pressure diffusion in the two materials (Fig. 91.3).

These results illustrate that the apparent rheological behaviour of till beneath glaciers will be sensitive to till porosity, hydraulic diffusivity and the rate and magnitude of stress changes. Porosity less than the steady-state value for a given effective normal stress, low hydraulic diffusivity and rapid but small stress changes (limiting frictional equilibrium cannot be sustained by dilatant strengthening if stress changes are too large) favour dilatant strengthening and apparent viscous response. Given the likely variability of these factors beneath glaciers, a reasonable expectation is that the apparent rheological response of till can be highly variable, both temporally and spatially. This conclusion is consistent with recent measurements on Siple Coast ice streams, interpreted as indicating radically different rheological behaviour of tills beneath the Whillans Ice Stream (Coulomb behaviour) and ice streams C and D (apparent viscous behaviour) (Bindschadler et al., 2003).

Till, like ice, develops anisotropy when it deforms. Grains of all sizes tend to become oriented with their long axes parallel to the direction of the largest principal strain. Murray & Dowdeswell (1992) conducted direct shear and triaxial experiments to low strains (<0.12) on various tills and used scanning electron microscopy to evaluate alignment of fine particles. Despite the small strain, alignment of particles was both visually and statistically discernable. A fourfold difference in hydraulic conductivity was estimated parallel and normal to the direction of particle alignment. Shear deformation of a till bed, in the absence of macroscale voids that might collapse during shear (Clarke et al., 1984), should therefore enhance bed-parallel permeability, with likely pore-pressure feed-backs that would affect till mechanical behaviour.

91.2.3 Sliding

The goal of focusing on recent experimental work cannot be met when considering sliding of glaciers over either a rigid or deformable substrate because there has been no recent experimental work. Indeed, there appear to be no published laboratory studies of glacier sliding since the mid-1980s. Oddly, this situation persists despite growing awareness that accurately modelling processes such as rapid decay of past ice sheets, motion and shutdown of ice streams, Heinrich events and surging of valley glaciers depends on parameterizing basal motion successfully (Clark et al., 1999; Marshall et al., 2000).

Various efforts have been made in the laboratory to study sliding of synthetic ice over rigid roughness elements of very small scale (<0.01 m) (e.g. Barnes et al., 1971) but the study of Budd et al. (1979) has been the most influential. Ice, kept at the melting temperature in ice-water baths, was slid across rigid substrates with a wide range of roughnesses. Data collected over the most glacially relevant ranges of shear stress, Tb, normal stress, N, and sliding speed, v, indicated a relationship of the form, v = k Tbp N~q, where k is a constant dependent on bed roughness, p = 3 and q = 1. Many field studies of sliding have indicated relationships of similar form, if N is replaced by effective normal stress (Paterson, 1994), although values of p and q fitted to field data vary widely. Budd et al. (1979) speculated that the value of p from their study reflected the dominance of ice deformation over regelation during sliding because regelation would depend linearly on shear stress. This inference is supported by the likely loss of water from the film at the ice-rock interface to the surrounding ice-water bath (Hooke, 1998). The resultant loss of latent heat would have reduced the heat available for melting on stoss surfaces. Thus, the value of p was undesirably dependent on the proximity of the water film to the bath at atmospheric pressure. Similarly, the value of q may reflect this proximity, which would have controlled the thickness of the water film for a particular value of N. Sliding speed may have been especially sensitive to the thickness of the water film because the amplitudes of roughness elements were small (ca. 0.1-10mm).

In other experiments, synthetic temperate ice has been slid across larger bed obstacles to study velocity fields and ice rheol-ogy. The most ambitious of these studies involved use of a large Couette-type viscometer to slide an annulus of ice at a controlled velocity across two sinusoidal bumps (wavelength, 0.53 m) (Brepson, 1979; Meyssonnier, 1982). The bumps were of low thermal conductivity (epoxy), thereby suppressing regelation. Leeward cavities formed that were comparable in size to the bumps. Using a finite-element model and measured cavity ceilings as a free-surface boundary of the modelling domain, Meyssonnier (1982) calculated steady-state velocity fields in good agreement with measured values using n = 3 in the flow law. Hooke & Iverson (1985) pushed streamlined bumps (wavelength, 0.16m) beneath temperate ice under controlled shear stresses. Motion was accommodated primarily by ice deformation. Comparison of ice deformation measured around bumps with the results of a theoretical model for ice flow past a hemisphere (Lli-boutry & Ritz, 1978) indicated n > 1.5.

The simplest and possibly most illuminating experiments motivated by the sliding problem have focused on regelation of round wires through ice. Such experiments with temperate ice by Drake & Shreve (1973) indicated that at driving stresses <0.1MPa, wire speeds were one to two orders of magnitude lower than predicted by simple regelation theory (Nye, 1967). This result agreed in general with previous but less comprehensive laboratory studies (e.g. Townsend & Vickery, 1967). Drake & Shreve's study isolated several seemingly insignificant processes that slowed regelation dramatically. The most relevant of these for glacier sliding was the accumulation of solutes in the rear of wires. This reduced the freezing temperature there and thereby reduced heat flow to the front of the wire. This effect was subsequently incorporated in glacier sliding theories (Hallet, 1976a) and was also used to help interpret the origin of chemical precipitants on deglaciated bedrock surfaces (Hallet, 1976b). Wire-regelation experiments have also been conducted at subfreezing temperatures (Gilpin, 1980) and provide fundamental empirical support for the theory of cold-based glacier sliding (Shreve, 1984). Wire-regelation studies remind us that simple experiments incorporating highly idealized materials and geometries can yield insights that are unlikely to be brought to light by either experiments that are more complex or field studies.

91.2.4 Glacial erosion and sediment transport

Glacial erosion and sediment transport may have had major impacts on uplift in orogenic belts, weathering rates and atmospheric CO2 (Hallet et al., 1996; Jaeger et al., 2001). This realization has spawned numerous large-scale field and modelling studies of glacial erosion and sediment yields. In comparison, experimental approaches have been unpopular.

Glacial erosion of bedrock occurs principally by abrasion and quarrying. Early experimental efforts to study abrasion of rock by debris-laden sliding ice were performed at temperatures less than —16°C (Lister et al., 1968; Mathews, 1979). Significant abrasion occurred in these experiments, but sliding of cold-based glaciers is too slow to cause much abrasion. Experiments in which a flat rock bed was pushed beneath isolated, gravel-sized particles in temperate ice (Iverson, 1990) have allowed some elements of abrasion theories (Boulton, 1974; Hallet, 1979) to be tested. The glaciological variables in these models responsible for the normal stress in excess of hydrostatic that rock particles exert on the bed have been controversial. These experiments indicated that such stresses depend on the rate of ice movement toward the bed, consistent with the model of Hallet (1979). Stresses associated with ice movement differed from that predicted by regelation and creep theory for an isolated sphere (Watts, 1974) by less than a factor of two, despite the bed as a heat source and rigid boundary. The detachment of rock fragments from the bed by quarrying is probably volumetrically more important than abrasion (Drewry, 1986). However, there have been no laboratory efforts to study this process, and as a result quarrying theories (Iverson, 1991; Hallet, 1996) have not been tested.

Entrainment of sediment from an unlithified bed has been studied in a few experiments. Regelation of temperate ice under pressure through dense arrays of idealized and natural particles has been studied in experiments with and without sliding (Iverson, 1993; Iverson & Semmens, 1995). These experiments indicated that rates of ice motion toward the bed through particles were within a factor of five of that predicted by regelation theory for dense arrays of particles (Philip, 1980). Ice may intrude the glacier substrate by this kind of regelation and entrain particles (Clarke et al., 1999). Particles in experiments ranged from gravel-sized clasts to coarse-grained till (primarily sand and coarse silt). Tightly bound pore water in finer sediments probably impedes or prevents such regelation (Alley et al., 1997c). Knight & Knight (1999) pressed ice at —1°C against a wet sediment bed and observed entrainment of fine sediment (primarily coarse silt) in water at crystal boundaries. In other experiments, they imposed a cold front across the ice-sediment boundary and noted similar movement of fine sediment into ice.

Interactions between grains during basal transport have been studied experimentally to quantify debris communition. In ball-mill experiments, in which rock particles and steel balls were tumbled in a drum (Haldorsen, 1981), two size ranges of particles were produced: a coarse range (0.016-2 mm) due to grain crushing, and a fine range (0.002-0.063 mm) due to abrasion of grains sliding past one another. Momentum exchange between particles in these experiments differed from non-inertial grain interactions expected subglacially. However, ring-shear experiments conducted at glacial rates have revealed similar particle modes (Iverson et al., 1996). Stress concentrations, measured normal to the shearing direction in initially equigranular mud-stone, were large at low strains, causing primarily crushing. Consequent production of finer grains reduced stress concentrations by cushioning larger particles, so that abrasion dominated comminution at larger strains.

Mixing between sediment units is another consequence of debris transport in a deforming bed. In ring-shear experiments, Hooyer & Iverson (2000b) studied mixing between equigranular beads of different colours and between lithologically distinct tills. They found that random vertical motions of particles induced by shearing caused linearly diffusive mixing. Mixing coefficients were determined from laboratory measurements. These coefficients were applied to the contact between basal tills of the Des Moines and Superior lobes of the Laurentide Ice Sheet to place an upper bound on the total bed shear strain.

91.2.5 Water in glaciers

The laboratory study of water movement in glaciers is limited by the same scaling considerations that guide river studies (e.g. Peakall et al., 1996), except at the smallest scale in which water moves through microscopic veins at crystal edges. Mader (1992) developed novel methods for making synthetic polycrystalline ice and determining the geometry of such veins photographically. Equilibrium veins—those with no temperature or impurity gradients—were similar in geometry to those predicted theoretically by Nye (1989). Pinched-off veins that help limit ice permeability constituted less than 5% of veins observed.

The flume approaches that have advanced fluvial geomorphol-ogy are difficult to extend to englacial and subglacial channels because of the difficulty of incorporating ice melting and deformation at appropriately large scales. Catania & Paola (2001), however, studied braiding of channels at glacier beds in the absence of such melting and deformation to determine how pressure-driven braiding patterns differ from those of subaerial rivers. A stream table was filled with non-cohesive sediment and fitted with piezometers and a transparent, rigid lid to allow direct observation of braiding patterns. The rigid lid, by hindering flow of water over banks and bars, caused lateral gradients in pressure head that were much larger and more variable than elevation-head gradients in rivers. These large, variable pressure gradients resulted in greater channel density and variability of flow direction than in braided rivers. A lid that could both melt and deform like temperate ice would have likely affected this result, but large lateral pressure gradients are certainly characteristic of subglacial hydraulic systems (e.g. Engelhardt & Kamb, 1997).

91.2.6 Field experiments

The manipulation of variables that characterizes experimental work is also clearly a part of some field studies on modern glaciers. For example, borehole water pressure is commonly perturbed to study the nature of subglacial hydraulic systems (Stone & Clarke, 1993; Engelhardt & Kamb, 1997; Stone et al., 1997; Kamb, 2001). Human access to the beds of some glaciers provides opportunities to manipulate the basal environment further. At the Svartisen Subglacial Laboratory in Norway, a conical concrete obstacle (0.2 m high) was installed at the bed of the temperate glacier Engabreen (Cohen et al., 2000). Stresses on the obstacle and sliding speed were measured and used in conjunction with a numerical model of ice flow to estimate A in Glen's flow rule (Equation 1) for basal ice (Cohen, 2000); the mean value of A determined was 5.4 times larger than for 'normal' temperate ice (Paterson, 1994, p. 97). In subsequent experiments beneath Engabreen, a trough (2.5m long, 2.0m wide, 0.5 m deep) was blasted in the subglacial bedrock and filled with simulated till (Iverson et al., 2003). After ice closed on the till, water was pumped to it to increase its pore-water pressure. Till sheared beneath the ice only at intermediate values of pore pressure: when pore pressure was sufficiently low or high, ice moved over the till by sliding and ploughing without pervasive shear of the bed, as predicted by some models (Brown et al., 1987; Iverson, 1999).

91.3 Conclusion: future directions

A survey of the Journal of Glaciology from 1997 to 2002 indicates that only 4% of papers published during that period involved laboratory experiments. Thus, although such experiments have played an important role in the discipline historically, they are the basis for only a minor fraction of glaciological research. The reasons for this may be more cultural than scientific. The aesthetic appeal of field studies and the increased ease of theoretical approaches made possible by inexpensive and potent computer technology probably contribute to the relative unpopularity of experimental approaches. If so, this is unfortunate; major progress solving longstanding problems in glaciology could be made experimentally.

The most glaring of these problems involves glacier sliding. At present, fundamental aspects of sliding theories that help guide large-scale model parameterizations (e.g. Marshall & Clarke, 1997b) are untested. For example, there are no direct data that unequivocally link basal water pressure to the sizes of basal cavities and cavity size to sliding speed. There are no direct data that link a known basal shear stress, bed roughness and ice-overburden pressure to the different water pressures required for ice-bed separation and unstable sliding. There are no direct data on rates of growth or shrinkage of cavities during their adjustment to variable subglacial water pressure. There are no data to directly compare the contrasting responses of soft- and hard-bedded glaciers to changing basal water pressure. All of these issues could be investigated with an appropriate sliding apparatus; temperatures, boundary stresses and basal water pressure could be controlled, with direct observation of ice-bed separation. The resources required for such experiments would be substantial but no larger than those commonly marshalled for experimental work in disciplines with stronger experimental traditions (e.g. geochemistry).

This review also points to other experimental directions, only a few of which can be mentioned herein. Studies of fabric development in ice deformed by grain-boundary sliding would help clarify key aspects of the debate regarding ice-deformation mechanisms (e.g. Goldsby & Kohlstedt, 2001, 2002; Duval & Montagnat, 2002). The possible dependence of ice effective viscosity on grain size should be revisited experimentally, given the potential importance of grain-boundary sliding and recent field measurements that illustrate grain-size dependence (Cuffey et al., 2000c). Laboratory studies of till deformation should further explore how variables such as till porosity, hydraulic diffusivity and loading rate affect the apparent rheological response of till. Laboratory studies of bedrock quarrying would provide a test of quarrying models and thereby help guide large-scale erosion models (e.g. MacGregor et al., 2000). Laboratory studies of abrasion and associated debris-bed friction would allow subglacial measurements of unexpectedly high debris-bed friction (Iverson et al., 2003) to be critically assessed.

As an example of the potential impact of laboratory experiments, consider progress in understanding a process that is similar in some ways to glacier sliding: slip along crustal faults. Over the past several decades, a rate- and state-dependent friction rule has become widely accepted in the fault-mechanics community (Scholz, 2002) and is used routinely in efforts to model fault-slip dynamics (e.g. Segall & Rice, 1995). This rule did not result primarily from theoretical studies or field monitoring of fault slip. Rather, this rule emerged from multiple long-term efforts to study fault slip in well-controlled laboratory experi ments. The experimental ethic responsible for this success story is similar to that which inspired initial laboratory studies of ice rheology (Glen, 1952,1955). Expansion of this ethic would hasten progress in glaciology.

Was this article helpful?

0 0

Post a comment