Introduction

It is now accepted that tributaries of enhanced flow are widespread throughout both East and West Antarctica (Reference JoughinJoughin and others, 1999; Reference Bamber, Vaughan and JoughinBamber and others, 2000). Rather than there being an abrupt transition between streaming and inland flow (Reference Alley and WhillansAlley and Whillans, 1991), tributaries represent a zone of gradual transition between these two flow regimes, in which flow occurs by some intermediate combination of internal deformation and basal motion (Reference JoughinJoughin and others, 1999). The widespread existence of these tributaries within ice-sheet interiors has farreaching implications for understanding ice-sheet flow and stability, which in turn is crucial for improving estimates of present-day and possible future changes in mass balance. As a consequence, it is important to better understand the mechanisms which control tributary flow.

An airborne radio-echo sounding (RES) campaign carried out in the upper reaches of the Slessor Glacier complex, in Coats Land, East Antarctica, has revealed that tributaries of enhanced flow lie within well-defined basal troughs, and are separated from each other by bed highs. As a consequence of these confining troughs, the large-scale location of these tributaries appears to be relatively stable. A numerical-modelling study indicates, however, that the contribution of basal motion to tributary flow can vary between tributaries, i.e. flow in the central and most southerly of the tributaries of Slessor Glacier can virtually all be explained by ice deformation (with basal motion, within the range of uncertainties, playing a very minor role), while flow in the northernmost tributary, under a similar glaciological setting, is largely due to basal motion (Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a).

We propose that variations in roughness may be responsible for this difference, because of the fundamental importance of the degree of coupling between ice and bed in controlling the amount of basal motion. This coupling is controlled by bed roughness and subglacial water pressure; however, bed roughness is often completely ignored, usually because of difficulties in measuring this quantity (Reference BennettBennett, 2003; Reference Taylor, Siegert, Payne and HubbardTaylor and others, 2004). In this paper, we investigate bed roughness as a potential control on basal motion over the Slessor Glacier region.

Methodology

Radio-echo sounding data collection

During the austral summer of 2001/02, an airborne RES campaign was carried out in the upper reaches of Slessor Glacier. From a field camp and fuel depot located at 78°58.60′S, 7°24.97′ W, five airborne RES sorties, covering a total of ≈5000km, were flown, over a grid measuring 200 × 280 km (Figs 1 and 2).

Fig. 1.

Part of the RADARSAT Antarctic Mapping Project (RAMP) mosaic of Antarctica (Reference JezekJezek and RAMP Production Team, 2002), showing the tributaries of Slessor Glacier, with flight-lines and the field camp/fuel depot marked. Inset (a) shows the location of the base map in Antarctica. Inset (b) shows the bed topography for the surrounding region, covering an area of 3000 × 2000 km. The thicker contour marks zero elevation, while the black rectangle delineates the 200 × 280 km region covered by the flight-lines, as shown in the main figure.

Fig. 2.

Contoured bed topography (20 km gridcells) in the survey area, with flight-lines superimposed (in black) to aid orientation in Figure 1. Thicker contours represent elevations (m) above ellipsoid (WGS 84), and thinner contours represent elevations (m) below ellipsoid (from Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a). The three troughs containing the tributaries of enhanced flow under investigation – Slessor tributary north (STN), Slessor tributary central (STC) and Slessor tributary south (STS) – and the deep southeastern trough (DSET) are marked. The thicker flight-line is that studied in more detail in Figure 3.

The British Antarctic Survey (BAS) airborne radar is mounted on a de Havilland DHC-6 Twin Otter aircraft. It is a coherent system, transmitting a modulated pulse with a peak power of 1200 W, at a centre frequency of 150 MHz. Pulses are transmitted through a four-element dipole antenna array mounted under the port wing, and return echoes are received by an identical array mounted under the starboard wing. A constant aircraft speed of 60 ms^–1 was maintained, giving an effective sampling interval of ~25 m. Two different pulses were transmitted alternately: a standard short pulse of 0.25 ms and a 4 μs chirp pulse that gave greater penetration of thick ice and a higher resolution. Consequently, all subsequent analysis was carried out on the 4ms pulse, giving a true along-track sampling resolution of ~60–75 m, while the across-track spacing of the radar lines was 40 km. The aircraft’s on-board avionics radar system was used to record terrain clearance, which gives a more accurate measure of surface elevation than the RES system itself. Flying was carried out between predefined waypoints (Figs 1 and 2), and coordinates were recorded using an Ashtech GPS (global positioning system) receiver on board the Twin Otter, with reference to a second receiver at the field camp. RES processing was carried out using the seismic processing package ProMAX v2003.0. Data were not migrated, but an automatic gain control (AGC) gain was applied before the locations of the surface and bed were identified using a semi-automatic picking process, and ice thickness was determined as the difference between surface and bed elevation. GPS data were processed using the Trimble package GPSurvey. These procedures are explained in more detail by Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others (2003a).

Results and Analysis

The area of investigation is divided into a series of subglacial troughs associated with the enhanced-flow tributaries. These are divided from each other by two areas of high bed, which separate the three tributaries of Slessor Glacier. It is important to point out that in our original analyses we referred to the northernmost tributary of Slessor Glacier incorrectly as a tributary of the nearby Bailey Ice Stream (e.g. Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a, 2004). More recent inspection of interferometric synthetic aperture radar (InSAR) velocity data clearly indicates that this is, in fact, a tributary of Slessor Glacier (e.g. Reference BamberBamber and others, 2006). We refer to the three tributaries as: Slessor tributary north (STN), Slessor tributary central (STC) and Slessor tributary south (STS) (Fig. 2).

In addition, there is another large, deep area towards the southeast of the study area. This trough lies upstream of the head of STC and STS, and we refer to this trough as the deep southeastern trough (DSET). We refer to the inter-tributary areas as the Slessor-north/Slessor-central (STN/STC) intertributary area and the Slessor-central/Slessor-south (STC/STS) inter-tributary area (Fig. 2).

Prior to our surveys, the data coverage in this region was very limited (note the sparse existing BEDMAP tracks on Fig. 1). There are, however, some existing data from the surrounding wider vicinity, and the bed topography derived from this portion of the BEDMAP dataset is displayed in Figure 1b. This more extensive coverage reveals that the entire Slessor region lies in a major basin that is largely grounded below sea level under the present ice load, and this submarine grounding persists for some distance (~200–400 km) upstream of the study area.

Figure 3 is an example of how the three indicators of roughness vary along part of one flight-line. The data are from a line which runs approximately east-west across the region, crossing part of the STC trough in the west, an intertributary area of high ground in the centre and part of the DSETin the east (see Fig. 2 for location of this line). All three indicators of bed roughness suggest that the DSET is markedly smoother (lower SDs and shorter pulse length) than along the rest of the flight-line. The inter-tributary area is the roughest area according to all three indicators, with STC being marginally smoother. Additionally, the variability of these roughness measures is smallest in the DSET, then STC, and greatest in the inter-tributary area.

Fig. 3.

Example of how the three indicators used as proxies for bed roughness vary along part of one flight-line, which crosses the whole survey area from east to west (approximate y coordinates = 1145 km; Fig. 2). (a) Bed topography; (b) SD of bed elevation (≈ 60–75 m resolution, averaged over 5 x 5 km gridcells); (c) SD of BRP_corr (≈ 60–75 m resolution, averaged over 5 x 5 km gridcells); and (d) pulse length (ms) (≈ 1.12m resolution, no averaging).

These roughness indicators for the whole survey area are shown in Figure 4. We have interpreted our roughness measures across the whole survey area by contouring data using a simple (smoothed) inverse distance interpolation routine (within the data analysis/visualization package IDL v6.3), in which closer raw data points have a greater effect on interpolated points than those further away. The SD of bed elevation indicates that at the scale of the true sampling interval (~60–75 m), there are large areas of rough bed, associated with the areas of high ground (Fig. 4a). These areas constitute slow-moving inter-tributary areas (cf. Fig. 2). There are also substantial areas of bed with much reduced roughness. These areas of reduced roughness are concentrated in STN and the DSET, although there are also areas of reduced roughness along much of STS and in parts of STC, particularly at the down-tributary end (west) (Fig. 4a; cf. Fig. 2). Table 1 quantifies the roughness characteristics of each area by expressing the roughness indicators in each area as a percentage of the mean roughness over the whole region. This table shows that according to the SD of bed elevation indicator, the two inter-tributary areas are significantly rougher than the mean over the whole area (35% and 41% rougher), while the four troughs are relatively smoother (19–68% smoother than the overall mean).

Fig. 4.

Three different indicators of roughness over the Slessor region. (a) SD of variation in bed elevation (~60–75 m resolution, averaged over 5 x 5 km gridcells). (b) SD of variation in BRP_corr (~60–75 m resolution, averaged over 5 x 5 km gridcells). (c) Pulse length (~1.12 m resolution, no averaging). In all three images, the relevant measure of roughness is contoured and displayed with greyscale shading (darker areas are smoother and lighter areas rougher). In addition, contoured bed topography (m) is also displayed with a 20 x 20 km grid size. As in Figure 2, thicker contour lines indicate elevations above ellipsoid (WGS 84), while thinner contours indicate elevations below ellipsoid. In (a), the star at approximate coordinates (-260,1185) indicates an example of a region of reduced roughness within a slow-moving intertributary area, as discussed in the text.

Table 1.

Roughness characteristics in the four troughs and two inter-tributary areas. In columns two and three, standard deviation (SD) of bed elevation and BRP_corr, respectively, are used as proxies for bed roughness, while in column four pulse length is a proxy for roughness. In each case, roughness is expressed as a percentage less than or more than the mean SD, or pulse length, over the whole area. Positive values indicate that the bed is rougher, while negative values indicate that the bed is smoother than the overall mean. Thus, these percentage measurements are relative indicators of roughness

The second indicator of roughness (the SD of BRP_corr) suggests that the inter-tributary areas of higher ground are rough (46% and 56% rougher than the overall mean), and that much of STC and STS, although comparatively smoother (4% rougher and 20% smoother than the overall mean, respectively), is still relatively rough compared with the much smoother STN and the DSET (58% and 55% smoother than the overall mean, respectively). The area of reduced roughness in the DSET also extends a short distance into the upper parts of STC and STS, particularly STS (Fig. 4b; Table 1; cf. Fig. 2). These areas of reduced roughness in the upper parts of STC and STS go some way towards accounting for the surprisingly low percentage roughness values (especially in STS) for the troughs as a whole, given that Figure 4b indicates that large parts of these troughs appear significantly rougher.

The roughness revealed by the third indicator (the mean of pulse length) shows a similar pattern (Fig. 4c). The intertributary regions are dominated by high roughness measurements, confirmed by the data in Table 1 (64% and 27% rougher than the overall mean). By contrast, the deeper trough of STN and the DSET are significantly smoother than elsewhere (61% and 62% smoother than the overall mean respectively; Table 1). STC and STS display an intermediate roughness (3% rougher and 28% smoother than the overall mean; Fig. 4c; Table 1), although, as before, there is an area of reduced roughness that extends from the DSET down the upper part of STS.

We attempted to investigate whether roughness varied both along-and across-flow, in all tributaries, where the orientation of flowlines allowed. These analyses revealed that, using the methods presented here, there was no significant difference in roughness between these two flow directions.

Discussion

STN contains the region’s thickest ice (>2800 m; Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a) and is characterized by very low roughness according to all three roughness indicators; the magnitude of roughness is similar using all three methods (Fig. 4; Table 1). A smooth bed implies fewer obstacles to sliding, and this may be a result of: (i) erosion of bed obstacles, (ii) infilling of gaps between obstacles by sediment and/or (iii) the presence of basal water which drowns bed obstacles. There are no InSAR velocities available for much of STN, but balance velocities for the region indicate that velocities of >60 m a^–1 are reached here (Reference BamberBamber and others, 2006).

Balance velocities were calculated, where the ice sheet is grounded, from surface slope, ice thickness and the mean net surface mass balance, using a two-dimensional finite-difference scheme (Reference Budd and WarnerBudd and Warner, 1996; Reference Bamber, Vaughan and JoughinBamber and others, 2000). We are confident of the accuracy of the balance velocities, particularly the location of these tributaries, since errors are relatively small. Surface slope is the most important factor controlling balance velocities, and this is scaled by mass balance (estimated 10% error) and ice thickness (estimated 20% error, although in some locations this is greater because of significant regional variations) (Reference Bamber and HuybrechtsBamber and Huybrechts, 1996; Reference Vaughan, Bamber, Giovinetto, Russell and CooperVaughan and others, 1999; Reference Bamber, Vaughan and JoughinBamber and others, 2000). Patterns of tributary flow are not affected by these errors because areas of enhanced flow move at least 100–200% faster than the slower-moving surrounding ice sheet, and errors of a similar order are necessary if thickness is to impact on tributary location (Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003c). Large-scale modelling experiments carried out by Reference Takeda, Cox and PayneTakeda and others (2002) have indicated that in this region of East Antarctica large parts of the bed reach the pressure-melting point. Additionally, a numerical modelling study has shown that basal motion contributes to a large proportion of flow here (Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a). The smooth bed revealed here supports the suggestion that basal motion is important.

Like STN, the DSET contains thick ice (~2800m; Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a) and is also characterized by low roughness according to all three indicators (Fig. 4; Table 1). The roughness magnitudes are very similar to those in STN, but, unlike STN, combined InSAR and balance velocities in much of the DSET are only ~10ma^–1 (Reference BamberBamber and others, 2006), and ice deformation is considered to explain all flow here (Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a).

Ice in STC and STS is ~800 m thinner than in STN (Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a). InSAR velocities are available (personal communication from I. Joughin, 2002), and indicate that while the eastern, upstream ends of the tributaries reach velocities of ~20 m a^–1, at the downstream western end they are >60 m a^–1 (Reference BamberBamber and others, 2006). A numerical modelling investigation has indicated that, within the range of uncertainties, virtually all flow within STC and STS may be explained by ice deformation alone, without the need for basal motion (Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a). The results presented here generally agree with this observation, although STS is surprisingly smooth (albeit rougher than STN or the DSET). Although ice deformation explains all flow here, perhaps the trough does bear the signature of recent basal motion. Additionally, there is some disagreement as to the roughness of STC at the ~60–75 m scale, derived from roughness methods 1 and 2 (Table 1). This is not of particular concern since, despite the suggestion that these two methods indicate roughness at the same range of scales, we do not expect that identical results would be achieved using both methods, particularly as method 2 may also reflect variations in the overlying ice column.

The two inter-tributary areas between STN and STC, and STC and STS are characterized by slow flow velocities of the order of 5 ma^–1 and are dominated by internal ice deformation (Reference BennettBennett, 2003; Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a). These intertributary areas have rough beds according to all three indicators (Fig. 4; Table 1). There are, however, a small number of regions within the Slessor-north/Slessor-central inter-tributary area which display reduced roughness at the ~60–75 m scale (e.g. Fig. 4a), despite velocities being low. Close inspection of raw track data indicates that these regions consist of small subglacial valleys (not visible in the gridded bed topography because of the small size of these troughs and the 20 × 20 km grid resolution). It has been shown that relict subglacial valleys may be preserved beneath the East Antarctic ice sheet, where ice velocities are low, as a consequence of cold basal ice (Reference NäslundNäslund, 1997). We suggest that in the early stages of ice-sheet growth, when ice cover was much thinner, small valley glaciers may have occupied these troughs, and due to low flow velocities have been preserved. Small regions of smooth bed at the ~60–75 m scale may be a product of these palaeoglaciers.

It should also be noted that in significant areas of STN, there is a loss of radar signal. It is possible that this may be attributed to drastically increased radar absorption, if STN is underlain by temperate ice with a liquid-water layer (Reference Bogorodsky, Bentley and GudmandsenBogorodsky and others, 1985).

Explaining roughness patterns

As stated previously, a smooth bed means fewer obstacles to sliding, either as a consequence of erosion of bed obstacles or the infilling of gaps in a rough bed by water and/or sediment. In an attempt to investigate the feasibility of these proposals and explain the observed patterns of roughness, we investigate two possibilities.

1. Basal water

If basal ice is at the pressure-melting point, water may be present. In order to investigate this, we determined theoretical patterns of glacier-wide subglacial drainage. It is acknowledged that, in general, the detailed structure of the drainage system beneath ice streams and their tributaries is uncertain. While borehole investigations in West Antarctica have indicated that a sheet of water ~2.5mm thick may be present beneath Kamb Ice Stream, it is uncertain whether this distributed drainage system also links with larger-scale channelized drainage (Reference Kamb, Alley and BindschadlerKamb, 2001). Regardless of the precise drainage structure, borehole investigations beneath Whillans, Kamb and Bindschadler Ice Streams, Siple Coast, West Antarctica, have indicated that the subglacial drainage system carries substantial amounts of water at a pressure which is approximately equal to the ice overburden pressure (Reference Engelhardt and KambEngelhardt and Kamb, 1997; Reference Kamb, Alley and BindschadlerKamb, 2001).

We calculate the steady-state subglacial hydraulic potential when subglacial water pressure (P_w) equals ice overburden pressure (P_i). Under such conditions, drainage elements are completely filled with water, and they are orientated at right angles to contours of hydraulic equipotential (e.g. Reference ShreveShreve, 1972). Other workers have investigated the impact of other steady-state drainage scenarios on drainage routing (e.g. when P _w = 0.5P _i or when P_w = P_a (atmospheric pressure);Reference Flowers and ClarkeFlowers and Clarke, 1999; Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003b), but in light of the discussion above, we do not consider these scenarios realistic here.

We have no knowledge regarding flow volumes or specific flow paths, and so the orientation of equipotential contours is used as guidance. Total hydraulic potential (F) is the sum of the elevation and pressure potentials and can be expressed as follows (adapted from Reference ShreveShreve, 1972; cf. Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003b):

where ρ _w is the density of water (1000 kg m^–3), ρ _i is the density of ice (917 kg m^–3), g is acceleration due to gravity (9.81 m s^–2) and H and B are the elevations of the ice surface and bed respectively (m). Theoretical drainage flows perpendicular to equipotential surfaces, from regions of high to low equipotential (cf. Reference SharpSharp and others, 1993).

Figure 5 shows contours of hydraulic equipotential assuming P _w = P _i. These bear some resemblance to contours of surface elevation (Rippin, unpublished data), but the role of subglacial topography is clear to see, with greater potential in the troughs than the inter-tributary regions. Interestingly, potential is greatest in the troughs of STC and STS, but it is clear that any basal water present (assuming everywhere is below the pressure-melting point (Reference Takeda, Cox and PayneTakeda and others, 2002) and water is generated at the bed everywhere across the study region) would be routed at the bed, from the inter-tributary areas, along all three Slessor troughs (perpendicular to equipotential contours) (Fig. 5; cf. Fig. 2). Additionally, there appears to be significant potential drainage from the DSET, westwards into the STC and STS troughs, while the STN trough is comparatively isolated.

Fig. 5.

Contours of hydraulic equipotential (kPa) when P_w = p. Contours of bed elevation (m) are overlain in black, with thicker contours representing elevations above ellipsoid (WGS 84) and thinner contours representing elevations above ellipsoid, as Figure 2.

This result indicates that any water generated at the bed is likely to be routed to the troughs of STN, STC and STS, and thus we cannot easily invoke the presence of large amounts of liquid water as the cause of the difference in bed roughness between these troughs.

2. Marine sediments

In addition to possible drainage routing, we investigated the potential distribution of marine sediments to explain patterns of roughness. Previous maps of isostatic rebound indicate that the Slessor region was not grounded below sea level in preglacial times, and thus marine sediments, as in most of East Antarctica, cannot be present at the bed (Reference DrewryDrewry, 1983). However, previously available ice-thickness data in the region were of very limited extent (Reference Steinhage, Nixdorf, Meyer and MillerSteinhage and others, 1999; Reference Lythe and VaughanLythe and others, 2001), and our new maps of bed topography and ice thickness provide data where previous coverage was extremely poor or non-existent. As a result, we use our new data to recalculate isostatic recovery, using a simple Airy compensation which ignores crustal flexure and assumes that all loading of the crust is entirely supported by the asthenosphere. In this model, the unloaded bed elevation following removal of the ice sheet is determined simply from: uplift = ice thickness × [density of ice (917 kgm^–3)/density of the asthenosphere (3300 kgm^–3)] (Reference Brotchie and SilvesterBrotchie and Silvester, 1969; Reference Bamber, Layberry and GogineniBamber and others, 2001). This represents a conservative approach to calculating uplift, since it ignores crustal flexure.

Figure 6 shows a new map of isostatically rebounded topography in the region, following the removal of the Antarctic ice load, taking into consideration an estimated 70m sea-level rise as a consequence of the melting of this ice load (Reference HughesHughes, 1998). This figure illustrates that while most of the area would have been above sea level in preglacial times, there is a substantial region within STN which may have been below sea level (cf. Fig. 2). The possibility that, contrary to generally held opinion, a substantial thickness of marine sediments could have accumulated in STN is of great significance and could be responsible for the proposed basal motion here.

Fig. 6.

Subglacial topographic map of the Slessor region, following isostatic recovery as a consequence of the removal of the Antarctic ice load. Contours are in metres above sea level (m a.s.l.), and take account of a 70m sea-level rise (Reference HughesHughes, 1998). The shaded area indicates where the bed would have been below sea level in preglacial times.

Measurements of total magnetic intensity (TMI) were collected as part of the geophysical survey, and these data provide information about shallow and deep magnetic sources (Reference BamberBamber and others, 2006). A region of anomalously low value is revealed, coincident with STN, and forward and inverse modelling of these data reveals a non-magnetic sedimentary layer ~3 ± 1 km thick. This sedimentary basin, along with the geographic setting, provides strong evidence that these sediments are of marine origin (cf. Reference BamberBamber and others, 2006). Additionally, we have calculated driving stresses to be ~50kPa in STN, which is lower than STC and STS, but very similar to driving stresses calculated for the Siple Coast ice streams which have been shown to lie on marine sediments (Reference BellBell and others, 1998; Reference Bindschadler, Bamber, Anandakrishnan, Alley and BindschadlerBindschadler and others, 2001; Reference Blankenship, Alley and BindschadlerBlankenship and others, 2001; Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others 2003a; Reference BamberBamber and others, 2006).

Coupled with the routing of subglacial water beneath STN, the presence of such marine sediments may provide the mechanism by which basal motion might occur, and thus explain a smooth bed. This is because it is now widely accepted that (in West Antarctica) a basal layer of water-saturated dilatant till is required for basal motion to operate (Reference Alley, Blankenship, Bentley and RooneyAlley and others, 1986; Reference Blankenship, Bentley, Rooney and AlleyBlankenship and others, 1986; Reference Studinger, Bell, Blankenship, Finn, Arko and MorseStudinger and others, 2001).

In the DSET, subglacial water might also exist (Fig. 5), but the bed would not have been below sea level in preglacial times, so marine sediments could not have accumulated here (Fig. 6). The presence of water and low roughness means that the DSET shares a lot of features with STN, implying that it too may have undergone significant basal motion in the past. However, ice deformation rates alone are sufficient to explain current flow rates. We propose that basal motion may have been active here in the past when the East Antarctic interior was thicker and flux rates different. If this is accurate, then any sediment required to form a deformable till layer must have: (1) come from other non-marine sources (e.g. fluvial, lacustrine), (2) been formed in situ by erosion or (3) been transported into the region from elsewhere. This is an important consideration (in all regions, even where we infer the presence of marine sediments) since if East Antarctica has been buried beneath ice for ~36 × 10⁶years (Reference Ehrmann and MakensenEhrmann and Makensen, 1992), then there would have been considerable erosion both here, and upstream in the large subglacial basin in which the Slessor region appears to sit (cf. Fig. 1b). Reference Hallet, Hunter and BogenHallet and others (1996) suggest that 0.1mm a^–1 is the minimum erosion rate under temperate glaciers, and 0.01mm a^–1 is the minimum for cold-based glaciers (equivalent to total erosion of 3600m and 360m, respectively). We are sceptical as to whether such high erosion rates, derived from valley glaciers, are applicable to ice sheets. However, it is possible that there is a deformable till layer present that is a consequence of both in situ erosion and transport of eroded material from upstream.

The rougher bed and lack of basal motion in STC and STS is consistent with them not having an accumulation of sediments (either marine or as a consequence of in situ erosion), even though there are significant drainage route-ways along them (Fig. 5a). This, perhaps, serves to indicate the significance of both sediments and water for basal motion to occur.

Conclusions

In this paper, we have used a new airborne RES dataset to assess bed roughness – an often neglected, but important control on basal motion – in the enhanced flow tributaries of Slessor Glacier. We have also carried out simple modelling of the distribution of subglacial water and possible marine sediments in an attempt to explain the roughness patterns observed. Our analyses enabled us to identify a number of distinct regions:

1. 1. Inter-tributary areas where the bed is shallow, velocities low and no basal motion occurs. These are characterized by rough beds, a lack of drainage and any possibility of (marine) sediment accumulation.

2. 2. The fast-flowing STN, where basal motion dominates (Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a) and which lies in a deep trough characterized by a smooth bed. Here, marine sediments may have accumulated and subglacial drainage may be present.

3. 3. The moderately fast-moving STC and STS, where all flow can be explained by ice deformation alone (Reference Rippin, Bamber, Siegert, Vaughan and CorrRippin and others, 2003a). High roughness is evidence of a lack of past basal motion and evidence also of obstacles to current or future basal motion.

4. 4. The DSET, which has a smooth bed but is currently slow-moving. There is no basal motion, despite the necessary conditions for this to occur. We propose that basal motion may have occurred here in the past.

In determining why STN displays such different flow characteristics to the adjacent STC and STS, the degree of bed roughness may be of fundamental importance. It is, however, not clear whether STN has a smooth bed because sediment and water drown bed obstacles, thus enabling basal motion, or if basal motion results in a smoothing of bed obstacles. It is likely that there are feedbacks between bed roughness and the effect on basal motion. It is also important to stress that although we have focused on the presence or absence of marine sediments, it is possible that sediments produced by in situ erosion, as a consequence of basal motion, exist; that sediments exist as a consequence of transport from upstream areas; or that they were deposited under different preglacial scenarios (e.g. lacustrine or fluvial settings).

In assessing the likelihood of areas of bed being above or below sea level, it is important to point out that our model of isostatic recovery is simplistic, and that we have not considered the effects of tectonic elevation change, or the implications of subglacial erosion on bed elevation over the ~36 × 10⁶ years during which East Antarctica has been buried beneath ice (Reference Ehrmann and MakensenEhrmann and Makensen, 1992). However, we believe that identification of part of STN having been below sea level in preglacial times is a significant discovery, and this is backed up by the TMI analyses (Reference BamberBamber and others, 2006).

Additionally, the role of large-scale bed topography on flow dynamics should not be overlooked, since the deep subglacial valleys where these enhanced-flow tributaries exist are locations where water and sediments collect. Finally, the recognition of adjacent areas of enhanced flow in East Antarctica, with significantly different flow mechanisms, is significant, as is the identification of potential areas of relict enhanced flow, indicating perhaps the ability of these tributaries to ‘switch’ on and off.