Surface motion and deformation of the ice

The glacier’s surface motion at the drill site was measured over various time-scales. The annual displacement of the borehole marker was measured by resection with a theodolite. In June 1997 a global positioning system (GPS) receiver recorded the position of a nearby stake four times a day for 3 weeks. Upstream of the drill site a camera has taken one picture a day since 1986 to record the motion of a pole.

Internal ice deformation in borehole N1 (located 10 m to the north of N1A) was measured with a retrievable inclinometer equipped with two mutually perpendicular tilt sensors and a compass. The inclinometry was performed on 16 May, 1 June and 28 June 1997.

Measurements within the till

Three tiltmeters were installed near the bed and in the till: the uppermost at the ice–till interface, and the others at 1 and 2 m below the interface (Fig. 2). This distance was measured from the top of the borehole down, as well as from the bottom up. The total length of the borehole was well known from the wireline drilling. The two methods of measurement were in excellent agreement. The elastic properties of the instrument cable were measured prior to installation, and the 1.5 m stretch of the cable due to a 20 kg load at its end was taken into account. The buoyant weight of the cable was 20 kg. Potential errors result from additional stretching of the cable due to its own weight, the buoyancy of the steel weight in a drill mud of a density unknown to us, and the possibility of caving-in of parts of the borehole before instrument insertion. These errors add up to a maximum of 0.5 m in the position of the instruments.

Fig. 2. Schematic drawing of the borehole instruments and their position in the borehole, as discussed in the text.

The instruments were mounted on a cable of 1.3 cm diameter (Cortland Cable Company) that was designed to stretch up to 30% before breaking. We used dual-axis electrolytic tiltmeters (Fredericks) with a measuring range of 0–30°. A data logger (Campbell 21X) recorded tilt along both axes, which was converted to total tilt and azimuth (Fig. 3). Changes in azimuth cannot be distinguished from rotation of the instrument around its own axis since no compass was used. We assume that such rotation is small, because it would induce torsion in the short section of cable between the individual instruments. The tiltmeters were mounted in 16 cm long cylindrical pressure cases, 7 cm in diameter.

Fig. 3. Tilt is measured along two axes (dotted lines), and then converted into total tilt and azimuth (bold lines).

Pore-water pressure was measured with a vibrating wire piezometer (Geokon) 0.5 m below the ice–till interface. The instrument had a range of 7 MPa and a resolution of 3 kPa. Borehole water levels were also recorded with a pressure transducer at 150 m below the glacier surface. This transducer was less accurate, and subject to systematic errors.

Surface displacement and ice deformation

The average annual surface velocity was about 60 m a⁻¹ at the borehole location. This implies a surface displacement of almost 70 m for the 410 days of borehole data collection. The position of a marker 300 m to the south of the drilling site was measured with a theodolite and an electronic distance meter (EDM) until Julian day (JD) 155, and then four times a day using GPS methods. Figure 4a shows the horizontal velocity. The spring speed-up on 7 June and several other events are prominent features of the record. The event on 21 June was a result of the drainage of a marginal lake upstream of the drill site, observed by an automatic camera. These events are very well resolved, as the GPS-derived velocities have an error of <6 cm d⁻¹ (see error bar in Fig. 4a).

Fig. 4. (a) Velocity measured with theodolite and EDM (before JD155) and GPS methods (thereafter). (b) Water pressure measured 0.5 m below the ice-till interface. Total tilt of the tiltmeters at the interface (c)1 m below it (d) and 2 m below it (e). Note the different scales on the ordinate axes.

Only the lowermost 100 m of the inclinometry record could be interpreted because, apparently, the borehole was too wide in its upper part to yield meaningful results. Assuming that the upper part of the borehole does not contribute significantly to the total deformation, we derived deformational speeds of 19 m a⁻¹ for the period 16 May–1 June, and of 23 m a⁻¹ for the period 1–28 June. On 7 June 1997 the annual spring speed-up occurred and the surface velocity of the glacier rose by a factor of three (Fig. 4a). This did not significantly affect the results of the inclinometry. A second estimate of deformational speed can be obtained from winter velocities. Minimum observed velocities amount to 35 m a⁻¹. This is an upper estimate for the deformational speed, because there is reason to believe that basal motion is occurring year-round (Reference Heinrichs, Mayo, Echelmeyer and HarrisonHeinrichs and others, 1996). A third estimate, of about 30 m a⁻¹, was obtained by using a model based on Reference Kamb and EchelmeyerKamb and Echelmeyer’s (1986) longitudinal averaging method. For the remainder of the paper we will assume that 20–30 m a⁻¹ is due to deformation of the ice. This implies basal motion of 30–40 m a⁻¹.

Measurements in the till

Tiltmeter and piezometer data were recorded from 8 May 1997 until 22 June 1998, when water flooded the data-logger box. Data gaps resulted from using up the storage module’s memory in summer 1997, and loss of battery power in winter 1998 (Fig. 5). We had expected the cable to break during one of the early speed-up events in spring 1997. The fact that data were still being recorded well over a year after initiation of the experiment came as a major surprise.

Fig. 5. Total tilt at the interface (a), 1 m below it (b) and 2 m below it (c). Azimuth is not shown in this graph. (d) Water pressures measured at 0.5 m below the interface (solid line), and at 150 m below the glacier surface (dotted line). The borehole water-level record was converted to pressure at the position of the piezometer in the till. They overlap from JD 180 to JD 195, when a shift in the water-level record occurs.

Figure 6 shows the tiltmeter records in a projection of the tiltmeter axis on a horizontal plane. The total change in tilt between the first and the last measurement is 3.4° for the uppermost, 4.9° for the second and 1.8° for the lowest tiltmeter, implying 13 cm of basal motion over the uppermost 2 m of till.

Fig. 6. The total path of the tiltmeter axis projected on a horizontal plane (i.e. the x–y plane in Fig 3a) Tiltmeter at the interface (a), 1 m below it (b) and 2 m below it (c). The open circles mark the start, and the open squares the end, of the record. The circles are contour lines of total tilt.

The small values of total tilt, the fact that the cable was still intact after 410 days, and the estimate of 35–45 m of basal motion during that time interval lead us to conclude that a major part (50–70%) of glacier motion occurred below the lowermost tiltmeter, i.e. below 2 m in the till (Fig. 7).

Fig. 7. Diagram showing the inferred distribution of glacier motion over the 410 day measurement period. Fifty to seventy per cent of the observed motion occurs below 2 m in the till, either on a series of shear bands or at the till–bedrock interface.

At times there appears to be larger tilt, but this tilting seems to be a reversible phenomenon. The tiltmeter at 1 m below the interface shows one jump of about 15° on JD 207, and then several excursions later on, all in the same direction (Figs 5 and 6). All of the tiltmeters show diurnal variations at times. No permanent tilting is associated with speed-up events (Fig. 4), although the lowermost tiltmeter seems to react to the spring speed-up, and the uppermost and lowermost instruments show strong but reversible spikes prior to the large speed-up event on 21 June (Fig. 4).

Figure 5d shows the pore-water pressure record. For about 2 months, water pressures were also simultaneously measured in the borehole at 150 m below the glacier surface. These measurements show that the piezometer in the till stayed hydraulically connected to the borehole for at least that long, as no phase lags and no attenuation of high-frequency variations were observed. The transducer at 150 m below the surface is subject to systematic errors, which explains the shift after JD 195.

Engineering properties of a till sample

A cylindrical till matrix sample 11 cm long and 5 cm in diameter was recovered from the top of the till layer in hole N1A. The sample contained some ice, and a water content of 29% was measured after it was allowed to melt. Sieving showed that only 6% of the sample’s weight is in the silt-and-clay fraction (Reference Truffer, Motyka, Harrison, Echelmeyer, Fisk and TulaczykTruffer and others, 1999). The remolded till was tested in a triaxial apparatus and an oedometer. In addition, an Atterberg liquid limit of 10% was determined in a liquid limit testing device (Reference BowlesBowles, 1992). Table 1 summarizes the results. The friction angle ϕ and the apparent cohesion c _a define a linear relationship (Coulomb friction law) between the sample strength τ and the effective pressure σ′:

(e.g. Reference Lambe and WhitmanLambe and Whitman, 1979). The low apparent cohesion is expected for a granular material, but previously measured friction angles on glacial till were typically <30° (Reference Iverson, Hooyer and BakerIverson and others, 1998). Our high value reflects the low silt-and-clay content (Reference Lambe and WhitmanLambe and Whitman, 1979) and the high angularity (data by D. B. Simons shown in Reference JulienJulien (1995, fig. 7.2)), although it might have been caused by a few large particles that were not removed before the test. The in situ diffusivity of the till is likely to be lower than that measured in the oedometer due to the inferred presence of ice in the uppermost part of the till (Reference Truffer, Motyka, Harrison, Echelmeyer, Fisk and TulaczykTruffer and others, 1999).

Table 1. Engineering properties of a till sample from borehole N1A

Ring-shear testing was done by S. Tulaczyk at the California Institute of Technology, Pasadena, CA, to test the strain-rate dependence of the shear strength. The shear strength increased by only 2 kPa (from 21 to 23 kPa) as the strain rates were varied over two orders of magnitude, reflecting a highly non-linear nature of the till. The slight increase in shear stress may reflect strain-rate-induced variations of effective pressure, and not a viscous effect (Reference KambKamb, 1991; Reference Tulaczyk, Kamb and EngelhardtTulaczyk and others, 2000). All ring-shear tests were carried out at the same low normal load of 20 kPa. They give a second estimate of the friction angle equal to 45°. This somewhat higher value is caused by the walls of the ring-shear device (Reference Iverson, Baker and HooyerIverson and others, 1997).

Alternative interpretations of the data

In the above section we have concluded that most of the observed surface motion originates from >2 m below the ice–till interface. We have considered several other possibilities, such as failure of the instruments, pulling of the instruments out of the till and misinterpretation of a dirty basal ice layer as water-saturated basal till. None of these possibilities can satisfactorily explain the results. If, for some reason, the instruments were placed in the basal ice rather than in the till, a steady tilting of much higher magnitude would be expected. The most likely alternative interpretation is that a substantial amount of basal motion occurs as sliding at the ice–till interface (Fig. 8). This could be the case under three conditions. First, the uppermost tiltmeter would have to be placed far enough from the ice–till interface not to be affected by the sliding motion. This is a possibility due to the uncertainties in the emplacement (see Methods). Second, the sliding motion would not have destroyed the cable. In principle, this is a possibility, but it would require that the induced stretch be distributed over the entire length of the cable. It could then conceivably stretch up to 150 m, substantially more than the 35–45 m required from basal motion. It seems rather unlikely, however, that the cable survived such an intense shearing environment for well over a year, a view shared by the manufacturer (personal communication from J. Dower, Cortland Cable Company, 1999). And, third, the instruments would have to be firmly emplaced in a very strong till. If they were not firmly emplaced, they would be pulled out of the till and into the ice. Although ice deformation accounts for only half or less of the total surface motion, calculated tilting rates at the bottom of the glacier would be significantly higher than those recorded by the tiltmeters. If the instruments had been pulled out of the till, one after the other would have recorded higher tilting rates. If the till was not strong, the instruments would be dragged through the deformable till, again causing appreciable amounts of tilting in one instrument after the other. This is not observed. On the other hand, the tiltmeter at 1 m below the interface exhibits several excursions of up to 15° of tilt. Also, the lowermost tiltmeter jumps by about 5° on 25 May 1998 (Fig. 5). At about this time the spring speed-up is expected to occur (Fig. 4), although we have no direct observations of this from 1998. The other tiltmeters do not record this event, or at least not as clearly. This observation fits very well with the idea of shear layers below the instruments, but not so well with the idea of a strong till and a sliding interface between the ice and the till.

Fig. 8. Position of the instruments if most of the basal motion is occurring on a discrete sliding plane or in a thin layer of sediments.

Short-term fluctuations

We tested the idea that a diffusing water-pressure wave affected the till strength (Reference Fischer, Iverson, Hanson, Hooke and JanssonFischer and others, 1998) and thus some of the structure in the tiltmeter record (Reference Tulaczyk, Mickelson and AttigTulaczyk, 1999; Reference Tulaczyk, Kamb and EngelhardtTulaczyk and others, 2000). We assumed that the pore-water pressures in the till are driven by a fluctuating water pressure at the ice–till interface. The water pressure beneath a valley glacier normally drops to a minimum in summer, once a good drainage system is established (Fig. 5). This increases the effective pressure in the upper part of the till, making it stronger. Underlying till will still be at higher pore-water pressures and therefore lower effective pressures. Thus, it will deform more readily. We modeled pore-water pressures in the till by numerically solving the one-dimensional diffusion equation

(1)

(Reference De MarsilyDe Marsily, 1986), where p is the pore-water pressure and C _v is the hydraulic diffusivity of the till. The measured water pressures were prescribed at the top of the till layer, interpolating through the data gap and assuming the pressures have a periodicity of 1 year. A zero gradient of hydraulic head was assumed at the bottom. We then compared the modeled results with a smoothed record of measured tilts (Fig. 9). The largest changes in tilt are observed at the time when the low-water-pressure wave reaches the corresponding tiltmeter, but otherwise the correlation between the two curves is not strong. The best agreement between the curves was obtained when a hydraulic diffusivity of 3 × 10⁻¹⁷ m² s⁻¹ was used, and 5 day averages of tilts were plotted. The diffusivity is somewhat smaller than that measured on a sample (10⁻⁵ m² s⁻¹), but due to the inferred presence of ice in the uppermost part of the till layer, we expect that the laboratory test overestimated the hydraulic diffusivity (Reference Truffer, Motyka, Harrison, Echelmeyer, Fisk and TulaczykTruffer and others, 1999).

Fig. 9. (a) Observed water pressure used as input for the model. (b, c) Modeled pore-water pressures and averaged total tilt (step plot) 1 m (b) and 2 m (c) below the interface.

Although the contribution of the overall measured tilt rate to the surface motion is negligible, each of the tiltmeters indicates periods of high strain rates, both positive and negative. Such variations have been observed before (Reference BlakeBlake, 1992), and have been interpreted as elastic effects, due to changing local basal shear stress (Reference Iverson, Baker, Hooke, Hanson and JanssonIverson and others, 1999) or effective stress (Reference Tulaczyk, Kamb and EngelhardtTulaczyk and others, 2000). Such elastic effects are predicted by an elastic–plastic soil model if stress changes are not large enough to exceed a previously reached maximum (e.g. Reference WoodWood, 1990). Reference Nolan and EchelmeyerNolan and Echelmeyer (1999b) concluded that lake-drainage events cause hydraulic jacking and temporary changes in local ice-overburden pressure. Together with changes in the pore-water pressure and local shear stresses, rather dramatic switches in the stress state should be expected.

It should be kept in mind that the small-scale features in the tiltmeter record are much harder to interpret because the tiltmeters are connected with a cable. The cable allows for a certain amount of interaction between the instruments, and some of the tilting could be due to pulling on the cable.

The tiltmeter record indicates that basal motion due to shearing in the upper 2 m of the till amounts to only 13 cm in 410 days. Fifty to seventy per cent of the glacier’s surface motion occurs more than 2 m beneath the ice–till interface. The total tilts reported above are fortuitously small, since tilt events as large as 20° are observed (Fig. 5). A uniform tilt of 20° would still only amount to 0.7 m of basal motion, a negligible amount.

Motion deep in a till layer: other observations

Direct observations of the distribution of motion in till layers have been made in tunnels near the glacier terminus (Reference Boulton and HindmarshBoulton and Hindmarsh, 1987; Reference Echelmeyer and ZhongxiangEchelmeyer and Wang, 1987). Echelmeyer and Wang observed shear bands in the frozen subglacial till beneath Ürümqi Glacier No. 1, and concluded that 10–25% of the total surface motion occurred on such bands. The present study concludes that an even larger amount of motion can occur deep in the till layer of a relatively fast-moving temperate glacier. In retrospect, several other observations also fit into this picture. Reference Harrison, Kamb and EngelhardtHarrison and others (1986) used borehole television on surge-type Variegated Glacier, Alaska, and were surprised that they did not manage to measure significant motion across the ice–till interface. Reference Iverson, Baker, Hooke, Hanson and JanssonIverson and others (1999) attributed the lack of significant till deformation in the upper 20 cm of the till layer to sliding of the ice over the underlying till, but did allow that it could also be due to motion deeper in the till.

In a study of the Puget glacial lobe, Washington, U.S.A., Reference Brown, Hallet and BoothBrown and others (1987) concluded that “basal motion was confined to the ice–bed interface or to distinct faults within the substrate”, and that “shearing along discrete shear zones is evident and may amount to significant displacement” (Reference Brown, Hallet and BoothBrown and others, 1987, p. 8994). Reference Hiemstra and van der MeerHiemstra and Van der Meer (1997) also concluded that glacial deposits from Wijnjewoude in The Netherlands contained discrete horizontal to sub-horizontal shear zones, along which strain was focused. Reference Astakhov, Kaplyanskaya and TarnogradskiyAstakhov and others (1996) reported observations of shear bands in West Siberian permafrost that was formerly overlain by glaciers. The occurrence of such discrete shear zones in glacial deposits has thus become textbook knowledge in glacial geology (e.g. Reference Menzies, Shilts and MenziesMenzies and Shilts, 1996, p. 48), although it has not been possible to show the amount of displacement that occurs in such zones.

Reference AlleyAlley (1991) presented arguments for widespread till deformation beneath the Laurentide ice sheet. This contrasts with the view of Reference Clayton, Mickelson and AttigClayton and others (1989) who argue against pervasive till deformation on the basis of till stratigraphy. It is also often assumed that coherent thrust sheets imply a frozen bed (e.g. Reference Mickelson, Clayton, Fullerton, Borns and WrightMickelson and others, 1983). The results presented in this paper might help resolve these paradoxes by allowing slip across discrete planes deep within unfrozen till layers. Such deep-seated slip may also help explain observations of large till sheets (up to 1000 km²) which have been displaced up to 250 km with little apparent internal deformation (references in Reference Van der Wateren and MenziesVan der Wateren, 1995, p. 328).

Discrete shear zones are also observed in shear tests on granular materials (e.g. Reference Mandl, dejong and MalthaMandl and others, 1977).

Is till deformation at > 2 m depth physically plausible?

The question of till deformation is intimately linked with that of the strength of the ice–till interface (Alley, 1983; Reference Brown, Hallet and BoothBrown and others, 1987; Iverson, 1999; Reference Tulaczyk, Mickelson and AttigTulaczyk, 1999). These studies show that this strength depends on the size of clasts and the grain-sizes in till, and on the effective pressure at the interface. An interface between ice and coarse clast-rich till, such as the one encountered beneath Black Rapids Glacier, is expected to have a high strength, thus making till deformation more probable. The inferred presence of regelation ice in the upper part of the till (Reference Truffer, Motyka, Harrison, Echelmeyer, Fisk and TulaczykTruffer and others, 1999) only serves to make this coupling stronger.

Reference Tulaczyk, Mickelson and AttigTulaczyk (1999) described three different models of basal motion over a water-saturated till: sliding over the till, shear deformation at a depth that is determined by the minimum effective pressure (overburden minus pore-water pressure) and shear deformation at a depth determined by the minimum of the past maximum effective pressure. The last model, “the perfectly overconsolidated till”, takes into account the memory of a granular material. When subjected to a shear load, a granular material will exhibit a peak strength that is determined by the maximum effective pressure it has experienced since it was last disturbed. If this maximum is higher than the current effective stress, the till is called overconsolidated (Reference ClarkeClarke, 1987). The maximum effective pressure is the difference between the overburden pressure, which increases linearly with depth in the till, and the minimum in water pressure over a given time period at a certain depth. This minimum water pressure increases with depth due to diffusion. To model the depth dependence of the maximum effective pressure throughout the year, the water pressure was calculated by solving Equation (1) for a 7 m thick till of hydraulic diffusivity C _v = 3 × 10⁻⁷ m² s⁻¹ and density ρ _till = 2700 kg m⁻³. Figure 10 shows the resulting annual maximum effective pressure as a function of depth. A strong decrease in the first 2 m is followed by a broad minimum at around 3.5 m and an asymptotic decrease that reflects the increasing overburden pressure. The position of this minimum in the annual maximum effective pressure reflects the position of the smallest value of peak strength, and marks a likely place for a shear layer. Using a hydraulic diffusivity of C _v = 10⁻⁵ m² s⁻¹ (Table 1) places such a layer even deeper in the basal zone, at 4.6 m below the ice–till interface.

Fig. 10. Maximum effective pressure throughout the year as a function of depth. Overburden pressure increases linearly with depth. The annual minimum of the water pressure was calculated by solving a diffusion equation.

In a proper model of basal motion, the strength of the till–bedrock interface has to be considered as well. It is possible that this interface is rather smooth, and that it is the weakest point in the basal ice–till–bedrock system. Sliding of till over bedrock has been discussed by Reference Cuffey and AlleyCuffey and Alley (1996) and Reference HindmarshHindmarsh (1996).

Erosion and till mass balance

The glacier’s main accumulation area lies to the south of the Denali Fault from where it crosses into the main fault valley about 3 km upstream of the drill site. In Reference Truffer, Motyka, Harrison, Echelmeyer, Fisk and TulaczykTruffer and others (1999) we concluded that about 85% of the till sampled in the boreholes was derived from the northern side of the fault. On the northern side or along this fault the glacier extends between 7 km (ice divide) and 12 km (head of the westernmost tributary) upstream from the drill site (Fig. 1). If 2–7 m of till were to be carried along at the rate of the basal motion, the average upstream erosion rate would be 4– 30 mm a⁻¹, ignoring cross-sectional variations and till focusing towards the center of the channel. This is somewhat higher than is thought to be typical of alpine glaciers (e.g. Reference Bennett and GlasserBennett and Glasser, 1996), but agrees well with the high rates found in southeast Alaska (Reference Hallet, Hunter and BogenHallet and others, 1996). Reference Humphrey and RaymondHumphrey and Raymond (1994) found that the erosion rate on Variegated Glacier was directly proportional to the ice velocity. They calculated a dimensionless erosion rate — the ratio of erosion rate to ice velocity — on the order of 10⁻⁴ for Variegated Glacier, and they showed that a similar rate was found for Glacier d’Argentière, France, and Breiðamerkurjökull. If this ratio applied at Black Rapids Glacier, an erosion rate of 6 mm a⁻¹ would result, just at the lower limit of the above estimate. It is possible that sliding of till over the underlying bedrock is the erosional mechanism for supplying this substantial amount of till (Reference Cuffey and AlleyCuffey and Alley, 1996; Reference HindmarshHindmarsh, 1996). The presence of large clasts and boulders in the till (Reference Truffer, Motyka, Harrison, Echelmeyer, Fisk and TulaczykTruffer and others, 1999) shows that abrasion cannot be the only erosional mechanism, however, and plucking has to occur as well.

Implications for surge behavior

Black Rapids is a surge-type glacier, and naturally the question arises if our observations can bring us a step closer to understanding surge behavior and initiation. The fast motion during the 1982/83 surge of Variegated Glacier was accompanied by water pressures close to overburden (Reference KambKamb and others, 1985). Dye-tracing experiments indicated a complete switch from a fast to a slow drainage system between non-surge and surge conditions (Reference BrugmanBrugman, 1986), and large volumes of water were released at times of decreasing velocities (Reference Humphrey and RaymondHumphrey and Raymond, 1994). This has also been observed at the surge termination of Kolka Glacier, Caucasus (Reference Krenke and RototayevKrenke and Rototayev, 1973) and West Fork Glacier, Alaska (Reference Harrison, Echelmeyer, Chacho, Raymond and BenedictHarrison and others, 1994). Reference KambKamb’s (1987) model of a switch from a drainage system consisting of Röthlisberger (R) channels to a linked cavity system can account for these observations. It assumes that the glacier is underlain mainly by clean bedrock, however, most likely a significant oversimplification (Reference Harrison, Kamb and EngelhardtHarrison and others, 1986; Reference Humphrey and RaymondHumphrey and Raymond, 1994; Reference Truffer, Motyka, Harrison, Echelmeyer, Fisk and TulaczykTruffer and others, 1999).

We suggest that a switch in the drainage system is not the cause but rather an effect of fast glacier motion. Surges could be initiated by the large-scale mobilization of subglacial till. This would happen as basal shear stresses exceed a critical value (Reference Boulton and JonesBoulton and Jones, 1979; Reference Clarke, Collins and ThompsonClarke and others, 1984). Increasing shear stresses were observed on Variegated Glacier prior to the 1982/83 surge (Reference Raymond and HarrisonRaymond and Harrison, 1988). An efficient drainage system, such as one consisting of R channels, will presumably be destroyed by high basal motion, causing a switch to a slowly draining system, such as the distributed “canals” suggested by Reference Walder and FowlerWalder and Fowler (1994), with a substantial amount of water storage and high basal water pressures. This makes sustained fast motion possible. The sudden release of the stored water will cause an abrupt slowdown or an end to the surge. The interaction between the subglacial hydraulic system and basal motion in the context of till mechanics is not understood well enough to elaborate further.

The mechanism suggested here is in accord with observations made on the surging Variegated Glacier, and it emphasizes the role of a subglacial till layer. It establishes the presence of a highly erodible glacier bed as a necessary but not sufficient condition for glacier surging, as has been proposed before (Reference WilburWilbur, 1988; Reference Hamilton and DowdeswellHamilton and Dowdeswell, 1996). A glacier geometry that allows the build-up of high basal shear stresses is possibly a second necessary condition.