2.1. Photoblock data

Repeat aerial photography was used to obtain closely spaced velocity measurements along four sectors (herein referred to as ‘photoblocks’) of WIS and VIS (Reference Whillans and van der VeenWhillans and others, 1993). Each photoblock consists of several aerial photographs linked together to produce a stereo model. The photo mosaic is constrained by Transit satellite ground- control stations and image overlap. The locations of velocities derived within each photoblock are shown in Figure 1, overlain on a RADARSAT image.

Fig. 1. The locations of photogrammetric velocity data points (black dots) overlain on a portion of the RAMP mosaic. The transects used in this study are shown in white.

The aerial photographs are tied to ground-control stations on ridge ice and on the fast-moving ice stream. Network adjustment techniques were applied to improve the precision of fast-moving stations, the details of which are summarized in Reference McDonald and WhillansMcDonald and Whillans (1988, Reference McDonald and Whillans1992) and Reference Whillans and van der VeenWhillans and others (1993).

To derive velocities from time-sequential photo-mosaics, the displacement of surface features (crevasses and drift mounds) observed in both mosaics was measured (Reference JacksonJackson, 1991; Reference Whillans and van der VeenWhillans and others, 1993). As a result, the distribution of velocity data is irregular throughout the block and absent in the crevasse-free center of the ice stream. In order to obtain a more uniform distribution of velocities, a cubic spline with an r ² of 0.954 was fit to a cluster of ten transects.

Errors associated with the photoblock velocities can originate from the photogrammetric measurements (1.7 m a^-1), ground-control points (0.4 m a^-1), or coordinate transformation (Reference Whillans and van der VeenWhillans and others, 1993). Following Reference Whillans and van der VeenWhillans and others’ (1993) approximation, a total velocity uncertainty of 4.0m a^-1 is used in this study.

2.2. InSAR data

InSAR is a widely used technique for measuring ice-sheet surface velocities. As part of RAMP (Reference JezekJezek, 1999), interferometric data were acquired for selected areas, including portions of the Siple Coast ice streams. Surface motion can be measured by analyzing the differences between coherently interfered pairs of complex synthetic aperture radar images, with the aid of a DEM. InSAR analysis, using a combination of interferometric and speckle matching methods (Reference JoughinJoughin, 2002), provides closely spaced (500 m) velocity measurements across much of WIS and VIS (Fig. 2). The final dataset has a resolution range of 0.5-3.0 km, due to smoothing of the speckle-tracked data to reduce noise (Reference JoughinJoughin, 2002).

Fig. 2. InSAR velocities for WIS and VIS. Velocities <50ma^_1 are white, and areas with no data are speckled. Thick black lines spanning the ice stream represent transects used in this study (labeled in Fig. 1). Thin black lines trending down the ice stream are flowlines traced from the RAMP mosaic.

There is a large degree of spatial variability in the InSAR errors across WIS. As Reference JoughinJoughin (2002) points out, errors around the shear margins are inherently larger than in the interior of the ice stream (1ct errors are 1-4 m a^-1). For this study, the velocity and associated error (1-25 m a^-1) at each gridpoint along the specified transects was used in the mass- flux and force-budget calculations (see Reference JoughinJoughin, 2002, for a more complete description of the InSAR errors).

Absolute control for InSAR data in this region was done using a select set of Transit satellite measurements collected by Reference Whillans and van der VeenWhillans and others (1993). Because the photoblock velocities were tied to similar ground-control points, it is possible that the InSAR velocities and the photoblock velocities are no longer independent datasets. However, only control points in slow-flow zones on interstream ridges were used for the InSAR analysis (Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002), so we assume minimal coupling between the photoblock and InSAR velocities compared in this paper.

2.3. Ancillary data

To complete mass-balance and force-budget calculations, we require estimates of accumulation rate, ice thickness and bed topography. Accumulation rates are derived from a gridded compilation of isolated point measurements, with errors of ~15% or 0.016 m a^-1 ice equivalent (Reference Venteris and WhillansVenteris and Whillans, 1998). Based on these data, we assign a range of accumulation rates from 0.13 ma^-1 at the upstream end (gates G0), to 0.11 m a^-1 at the terminus (gates G3 and G4) (Fig. 1).

Reference Shabtaie, Whillans and BentleyShabtaie and others (1987) conducted airborne radar soundings of the ice sheet along a dense grid (nominal flightline spacing in both horizontal directions was roughly 50 km) to measure surface topography and ice thickness (from which basal topography is inferred) over the Ross ice streams (Reference Shabtaie and BentleyShabtaie and Bentley, 1988). These data were combined with isolated surveys and then gridded to form the BEDMAP compilation (Reference Lythe and VaughanLythe and Vaughan, 2000). Errors of 25 m for bed topography and 18 m for ice thickness are used (Reference Lythe and VaughanLythe and Vaughan, 2000). These datasets were only used to calculate the driving stress. To reduce errors, the driving stress was calculated at least ten times across each transect, the results were averaged and the standard deviation calculated.

The OSU DEM was used as an estimate of surface topography. It is an integration of different topographical datasets of Antarctica combined into a single high-resolution DEM (Reference JezekJezek and others, 1999; Reference Liu, Jezek and LiLiu and others, 1999). The vertical accuracy for the Ross ice streams is 15 m (Reference Liu, Jezek and LiLiu and others, 1999).

3.1. Spatial and temporal variations in velocity

We compare velocities along transects perpendicular to flow (see Figs 1 and 2). The 1987 and 1997 velocities (both measured and interpolated) along these transects are shown in Figure 3. A shift in the limbs of the parabolic velocity profile is interpreted as a migration of the shear margin.

Fig. 3. Spatial and temporal changes in velocity (m a^_1) along WIS. The dotted line is the difference between the 1997 InSAR velocities and the 1987 photoblock velocities, plotted at the same scale. There are no photoblock data across transects G0n and G0s.

We begin with the upstream transects G0n and G0s. Ice velocities across the farthest upstream profiles (G0n and G0s) are characterized by the confluence of several smaller tributaries, which develop up-glacier of the main trunks (WIS and VIS). Both the velocity profiles and the RADARSAT mosaic show at least five tributaries dissected by the G0 transects. The southern tributaries are distinctly revealed in the velocity profiles as local velocity maxima and minima, while the northern tributaries are only weakly detected. These tributaries follow subglacial valleys, and therefore the processes that channel ice-stream flow are already well developed, especially across the northern transect (Studinger and others, 2001;Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002). There are no velocities derived from the photoblocks in this region.

The 1987 and 1997 velocity profiles (transects G1n and G2n) across the main trunk of WIS are parabolic and symmetric about the center line. Over ten years, the trunk of WIS has decelerated an average of 2.9-4.1ma^-2. This estimate is slightly higher than the average deceleration of 2.1-2.6ma^-2 calculated in other studies (Reference Hulbe and WhillansHulbe and Whillans, 1997; Reference Echelmeyer and HarrisonEchelmeyer and Harrison, 1999; Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others 2002).

Along VIS, the three original tributaries have reduced to two and are in the process of merging together, as can be seen by comparing transects G1s and G2s. There is a slight deceleration (0.6-1.7 m a^-2) with no associated shift in the margin. The velocity across transect G1s is faster than the downstream velocity at G2s. We speculate that this is due to the oblique intersection of a smaller channel (seen at the left edge of the G2s profile), resulting in an increase in longitudinal compressive stress.

At the distal end of the ice stream (profile G3), the branching network of ice-stream flow is reduced to one main trunk as WIS and VIS merge. The profile is roughly parabolic, with a velocity maximum of 688 m a^-1 biased towards the north end of the channel. We interpret this asymmetry as being due to the convergence of the two main tributaries and the effective narrowing of the northern branch by approximately 13 km (VIS has only narrowed 2 km at G3). At this location, ice deriving from the two tributaries (WIS and VIS) is separated by a clearly visible suture zone coinciding with the asymmetry of the velocity profile. From 1987 to 1997, there was a deceleration along the northern portion of the ice stream of approximately 4.5 m a^-2. Possible reasons for the deceleration are addressed in section 6.2.

The velocity profile becomes more symmetric downstream at G4 where there is only slight evidence of the upstream tributaries in the velocity profile. That is, the velocity maximum remains displaced towards the northern margin, and velocities are somewhat smaller across the southern half of the transect. This likely reflects the generally lower speeds on VIS compared to WIS. More significantly, along this transect we observe a dramatic deceleration of 7.2 m a^-2 over 10years, with a 1.7 km shift inwards of the northern margin. This value compares well to a mean deceleration of 6.5 m a^-2 calculated by Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others (2002) over the same area.

3.2. Margin migration

The margins of an ice stream play an important role in the down-ice dynamics. Many modeling and field-based studies agree that the driving stress, τ _dx, along WIS is almost entirely supported by lateral drag (e.g. Reference Echelmeyer, Harrison, Larsen and MitchellEchelmeyer and others, 1994; Reference Jackson and KambJackson and Kamb, 1997; Reference Whillans and van der VeenWhillans and Van der Veen, 1997; Reference Harrison, Echelmeyer and LarsenHarrison and others, 1998; Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002). If this is true, then the velocity at the center of the ice-stream center line is proportional to the fourth power of the width of the ice stream, W, (Reference Van der VeenVan der Veen, 1999) according to 2

In this expression, B is a viscosity parameter that increases with depth; in the following calculations, a depth-weighted value (540kPaa^1/3) is used. The flow-law exponent, n, is taken to equal 3.

Even a slight increase in ice-stream width could lead to a large increase in glacier speed. The deceleration along WIS occurs with both outward (G1n, G2n) and inward (G4) margin migrations, giving clues to different processes controlling the migrations.

To determine changes in the shear margin positions, the photoblock velocity profiles were compared to the RADAR- SAT mosaic. Since the InSAR data are smoothed by 1-3 km, a direct comparison of the two velocity profiles proved tenuous (Reference JoughinJoughin, 2002). For the photoblock data, the shear margin was defined as the start of a sharp increase in velocity: a narrow and distinct zone (Reference Raymond, Echelmeyer, Whillans, Doake, Alley and Bind-schadlerRaymond and others, 2001). For the RADARSAT mosaic, the shear margin was determined as the outer extent of the arcuate crevasses. Reference Echelmeyer and HarrisonEchelmeyer and Harrison (1999) noted that these crevasses coincided with the greatest jump in along-flow velocity.

Using this technique, we found significant margin migration along the northern flank of the ice stream. Transect G1n has migrated outwards (north) 2.8 km; transect G2n has migrated outwards 1.7km; transect G4 has migrated inwards 1.6-2.2km. Along the southern side of G1n and G2n, there is up to 250 m outward migration. There was no noticeable shift in the margins of VIS; however, this could be due to poor visibility of arcuate crevasses in the RADARSAT imagery.

The shear margins of WIS have been studied extensively using various theoretical and field measurements. Velocity profiles (Echelmeyer and Harrison, 1998), crevasse curvature methods (Reference Vornberger and WhillansVornberger and Whillans, 1990), temperature profiles (Reference Harrison, Echelmeyer and LarsenHarrison and others, 1998) and theoretical studies (Reference Jacobson and RaymondJacobson and Raymond, 1998) all suggest a 7-30 ma^-1 outward migration of the southern margin of WIS, comparable to our findings. We find substantially higher migration rates for the northern margin of WIS: outwards 170-280 m a^-1 at G1n and G2n; inwards 170 m a^-1 at G4. The magnitudes of these estimates are comparable to those found by Reference Bindschadler and SeiderBind-schadler and Vornberger (1998). Rapid outward migration rates are mechanically possible if the bed under the ridge ice is still warm and wet from previous streaming flow (Reference Bindschadler and SeiderBind-schadler and Vornberger, 1998; Reference Jacobson and RaymondJacobson and Raymond, 1998). Fast inward margin migration (as seen at G4) is best explained by the advection of colder ice from adjacent ridges into the ice stream and is not unprecedented (Reference Van der Veen and WhillansVan der Veen and Whillans, 1996; Reference Jacobson and RaymondJacobson and Raymond, 1998; Reference Raymond, Echelmeyer, Whillans, Doake, Alley and Bind-schadlerRaymond and others, 2001). Reference Clarke and BentleyClarke and Bentley (2000) estimated an inward migration rate of 100 m a^-1 for the south margin of WIS (near G1n/G2n) as it migrated from the ‘Fishhook’ starting ~200years ago. Similarly, Reference Jacobel, Scambos, Nereson and RaymondJacobel and others (2000) suggested that the multiple scars on the north margin of Kamb Ice Stream indicate that the margin moved inward in rapid jumps.

The migration of a shear margin is dictated by the balance between inward flow of cold ice from adjacent ridges (which narrows the ice stream) and shear heating of the margins (which widens the ice stream as the ridge ice is warmed) (Reference Van der Veen and WhillansVan der Veen and Whillans, 1996; Reference Bindschadler and SeiderBindschadler and Vornberger, 1998; Reference Jacobson and RaymondJacobson and Raymond, 1998; Reference Raymond, Echelmeyer, Whillans, Doake, Alley and Bind-schadlerRaymond and others, 2001). Spatially stable ice-stream margins are unlikely, due to the delicate balance described above and the dynamic nature of ice streams. However, Reference Jacobson and RaymondJacobson and Raymond (1998) suggest that a stable width for an ice stream could be maintained if there is a geologically determined boundary that precludes further outward expansion. Basal troughs and minimal margin migration support findings that the margins of VIS are geologically bounded.

3.3. Spatial and temporal variations in lateral shear stress

Prior studies have revealed the dominance of side drag along the lateral margins in opposing the gravitational driving stress and, therefore, in controlling the flow dynamics of WIS (Reference Bindschadler, Stephenson, MacAyeal and ShabtaieBindschadler and others, 1987; Reference Echelmeyer, Harrison, Larsen and MitchellEchelmeyer and others, 1994; Reference Jackson and KambJackson and Kamb, 1997; Reference Whillans and van der VeenWhillans and Van der Veen, 1997; Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002). Using the two sets of velocity data, we investigate spatial and temporal changes in the role of lateral drag following the procedure outlined by Reference Whillans and van der VeenWhillans and Van der Veen (1997) (cf. Reference Van der VeenVan der Veen, 1999, section 5.5). That is, shear stresses, R _xy, are calculated solely from the shear strain rate (equal to one-half the transverse gradient in the downstream velocity, U) and Glen’s flow law, so that: 1

The variables B and n are the same as used in Equation (1). Because vertical shear is negligible on fast-moving ice streams where most of the motion derives from basal sliding or deformation of a subglacial layer of weak sediment, the depth-averaged velocity as called for in Equation (2) may be taken as the measured surface velocity. Finally, in this calculation longitudinal stresses are assumed negligible based on previous research (Reference Whillans and van der VeenWhillans and Van der Veen, 1993) so that only the horizontal shear strain rate contributes to the effective strain rate. Calculated shear stress variations are shown in Figure 4.

Fig. 4. Shear stress variations (kPa) across WIS, calculated from Equation (1). The 1987 shear stress is derived from the photoblock velocities, and the 1997 shear stress from InSAR velocities (as shown in Fig. 3).

Similar to the velocity field, the shear stress field simplifies as the effects of upstream tributaries attenuate and the ice stream approaches the Ross Ice Shelf. Across G0n/s, the shear stress fluctuates ~80 kPa above and below zero in response to merging tributaries of varying velocity. By G1n and G2n, the complex variations in shear stress observed at G0n/s have reduced to simple, near-linear variations in stress across the channel, characteristic of an area of the ice stream that is bounded by well-developed shear margins. A linear slope in shear stress is observed for both the InSAR and photoblock velocities. Because photoblock velocities are sparse away from the margins and because a cubic spline was fit to the photoblock velocity data points, the linearity of the shear stress slope could be artificial.

The situation at G2s and G3 is more complicated due to the confluence of nearby tributaries. For each of these profiles, the shear stress consists of a segment that is nearly constant with range and then abruptly changes to a segment that varies linearly. For both of these profiles the break point in slope is distinctly off-center. By G4, the pattern is again simplified to a near-linear variation in shear stress.

Temporal variations in shear stress are only noticeable in profiles G1n, G2n and G3 (Fig. 4). Along profiles G1n and G3 the variations appear to be spatially random. In G2n, there appears to be a systematic bias resulting from a decrease in shear stress along the northern portion of the profile.

4.1. Spatial and temporal changes in the stresses

With the above assumptions, the force-balance equation is simplified to (Reference Van der VeenVan der Veen, 1999, section 5.5):

Table 1 lists the values for each term in Equation (5) along sections of WIS and VIS bounded by transects calculated for 1987 and 1997. Before discussing the table, we note that there are two types of errors associated with the force- budget technique. First, there are errors that have propagated from the InSAR and photoblock velocities. Second, there are errors that are inherent in each dataset used: BEDMAP, the OSU DEM, and accumulation rates. The variance is computed by taking the partial derivative of each of the relevant equations discussed above. Errors are assumed to have a Gaussian distribution, and errors between variables are taken to be independent. We do not include an additional source of non-random error associated with assumptions in our model (e.g. neglecting the longitudinal stress gradients) and our averaging procedures.

The driving stress varies along the ice stream, as shown in Figure 5. To eliminate errors in deriving the driving stress, at least ten profiles (adjacent to each specified transect) were averaged for surface slope (RAMP DEM) and ice thickness (gridded BEDMAP) (Reference JezekJezek and others, 1999; Reference Liu, Jezek and LiLiu and others, 1999; Reference Lythe and VaughanLythe and Vaughan, 2000). For the surface slope, profiles greater than ten times the ice thickness were averaged and the standard deviation recorded (after Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002). VIS has thicker ice and a steeper surface slope (see Table 1) and therefore supports a driving stress that is larger than the adjacent branch of WIS.

Fig. 5. The distribution of driving stress along WIS and VIS based on the OSU DEM and BEDMAP. The value above represents the mean driving stress for each area with a 2σ error, at an average date of 1999.

A change in the driving stress from 1987 to 1997 could explain some of the observed velocity and force-budget variations. As Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others (2002) note, the surface velocity is very sensitive to changes in driving stress. Their calculated change in driving stress does explain their observed deceleration accounting for no, or minimal (up to 34 m a^-1), widening. Our estimates for widening are significantly larger: up to 300 m a^-1. In calculating the effect that thickening or thinning rates would have on the surface slope, we found that the prescribed elevation changes altered the driving stress by <1% of the total driving stress.

Figure 6 illustrates the respective distribution of basal shear stress and weighted lateral shear stress, as calculated from Equation (4). The weighted shear stress, τ_w, is relatively low (6.8-8.7kPa) in the upper reaches of the ice stream, increases by 30-45% in the middle reach (to values of 9.814.7 kPa) and decreases back (6.9-10.4 kPa) near G4. There is a noticeable, but not statistically significant, temporal change in τ_w along the northern margin of WIS (near G2n and G4). In both of these locations, τ _w is reduced by ~2 kPa from 1987 to 1997. These are also the only two locations where there is a large shift in northern margin location, albeit in opposite directions. At G1n and G2n, the north margin migrates outward 1.7-2.8km; at G4 the north margin migrates inward 1.6-2.2 km. This observation suggests that G4 is narrowing due to deceleration, while G1n and G2n are decelerating for another reason, possibly a change in basal conditions.

Fig. 6. Basal drag (italic labels in the center of the ice stream) and weighted shear stress values for 1987 (regular type) and 1997 (bold type).

Values of calculated basal drag are not uniform throughout the ice stream. In the upstream portions of the ice stream, the driving stress is high (especially along VIS) but the calculated weighted shear stress is lower than the downstream values. This results in high values for calculated basal drag, based on Equation (5). This equation assumes that longitudinal components of stress are negligible, an invalid assumption at the tributaries (Reference Price, Bindschadler, Hulbe and BlankenshipPrice and others, 2002). Once the ice-stream shear margins become well developed in the middle section of the northern tributary, basal drag resists less of the driving stress, as anticipated (Reference Echelmeyer, Harrison, Larsen and MitchellEchelmeyer and others, 1994; Reference Jackson and KambJackson and Kamb, 1997; Reference Whillans and van der VeenWhillans and Van der Veen, 1997; Reference Harrison, Echelmeyer and LarsenHarrison and others, 1998; Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002). However, for VIS, basal drag is still an important component, accounting for 34% of the driving stress. The high basal drag estimate, however, could be a function of large errors in the RAMP DEM in this area (Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002).

Along the main part of the WIS, basal drag is approximately 1-3 kPa, which agrees with estimates by Reference KambKamb (1991), Reference Echelmeyer, Harrison, Larsen and MitchellEchelmeyer and others (1994) and Reference Whillans and van der VeenWhillans and Van der Veen (1997). In these areas, basal drag opposes 15-30% of the driving stress, comparable to estimates by Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others (2002). The remaining 70-85% of flow resistance is provided by lateral drag, highlighting the importance of the margins to ice-stream flow.

6.1. Summary of observations

6.2. Potential cause

Evidence of both past (Reference Clarke and BentleyClarke and Bentley, 1995; Jacobel and others, 1996;Reference Nereson and RaymondNereson and Raymond, 2001) and current (Reference Bindschadler and SeiderBindschadler and Vornberger, 1998; Echelmeyer and Harrison, 1998; Reference Hamilton, Whillans and MorganHamilton and others, 1998; Reference Harrison, Echelmeyer and LarsenHarrison and others, 1998; Reference Jacobson and RaymondJacobson and Raymond, 1998) margin migrations gives clues about the changing behavior of WIS. Ice-stream width is dictated by a balance between the entrainment of ridge ice and the advection of streaming ice. As Reference Raymond, Echelmeyer, Whillans, Doake, Alley and Bind-schadlerRaymond and others (2001) posit, a decrease in basal velocity could cause a subsequent decrease in basal heat production inside the margin. This could cause streaming ice to be incorporated into the ridge, effectively narrowing the ice stream. This process explains the narrowing and deceleration at G3 and G4.

When the ice stream is widening and decelerating, as in G1n and G2n, the above shear-margin adjustment process does not apply. Here, the balance between entrainment and advection is perturbed by an external forcing, possibly a strengthening of the bed (Reference Tulaczyk, Kamb and EngelhardtTulaczyk and others, 2000; Reference Raymond, Echelmeyer, Whillans, Doake, Alley and Bind-schadlerRaymond and others, 2001; Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002).

Several studies find that the majority of melt beneath the Siple Coast ice stream is produced near the onset of streaming (Reference RaymondRaymond, 2000; Reference Whillans, Bentley, van der Veen, Alley and BindschadlerWhillans and others, 2001) underneath the tributaries (Reference JoughinJoughin and others, 1999, 2003; Reference Price, Bindschadler, Hulbe and BlankenshipPrice and others, 2002), with a possible contribution from subglacial lakes (Reference Siegert and BamberSiegert and Bamber, 2000). Meanwhile, minimal melt (and potentially freeze-on) is found underneath the most active part of WIS, in modeled (Reference Joughin, Tulaczyk and EngelhardtJoughin and others, 2003; Reference Vogel, Tulaczyk and JoughinVogel and others, 2003; this paper), theoretical (Reference RaymondRaymond, 2000), and observed (Reference Engelhardt and KambEngelhardt and Kamb, 1993) accounts. Using a numerical model of basal melt patterns under the Siple Coast tributaries, Reference Vogel, Tulaczyk and JoughinVogel and others (2003) infer a depletion of basal melt since the early Holocene. If this is the case, then depletion in basal meltwater should affect the terminus of the ice stream first. Less subglacial water could increase basal resistance (Reference ClarkeClarke, 1987; Reference Tulaczyk, Kamb and EngelhardtTulaczyk and others, 2000), and therefore could account for the observed deceleration (Reference Stephenson and BindschadlerStephenson and Bindschadler, 1988; Reference Hulbe and WhillansHulbe and Whillans, 1997;Reference Bindschadler and SeiderBind-schadler and Vornberger, 1998; Reference Echelmeyer and HarrisonEchelmeyer and Harrison, 1999; Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and others, 2002; Reference Vogel, Tulaczyk and JoughinVogel and others, 2003; this paper).

We speculate that the depletion of meltwater at the base of the ice stream is most likely responsible for the observed changes. Ice streams may stop flowing fast if the lubricating conditions are lost (Reference Tulaczyk, Kamb and EngelhardtTulaczyk and others, 2000), a hypothesized precursor to the shut-down of Kamb Ice Stream (Reference Price and WhillansPrice and Whillans, 2001; Reference AlleyAlley, 2002; Reference Joughin, Tulaczyk, Bindschadler and PriceJoughin and Tulaczyk, 2002). Not enough research has been done to say whether WIS is headed towards stagnation or not. However, there is mounting evidence of its current deceleration and a change in basal conditions.