Ice-thickness measurements

Radar surveys were conducted during the 1995 and 1997 Antarctic seasons (Fig. 2). The radar equipment used in 1995 (Reference Tabacco, Passerini, Corbelli and GormanTabacco and others, 1998) includes a digital pulse generator that creates a 60 MHz pulse of controlled duration (0.3 or 1.0 μs), at about 1 mW peak power. A solid-state power amplifier raises that to > 2 kW peak power. The receiver consists of a solid-state low-noise pre-amplifier and a controlled bandwidth successive detection logarithmic amplifier. The dynamic range of the amplifier is 80 dB, and the output is digitized (by means of a 10 MHz digital oscilloscope) to eight-bit resolution, acquiring 512 samples every 100 ns (±8 m accuracy in ice-depth determination) for a total of 51.2 μs. The system allows an acquisition rate of one record every 3 s. The calibration test of the overall system shows a system performance of 150–160 dB (not including antenna gain), depending upon pulse length and bandwidth settings. The radar equipment was deployed by a Twin Otter aircraft, with folded dipole radar antennas under each wing, one for transmittal, the other for reception. Antenna gain is estimated to be 8.2 dB. A global positioning system (GPS) Trimble 4000SSE system (L1 and L2 frequency), with a geodetic antenna mounted on the fuselage, was installed and linked to the radar. A master GPS station, synchronized with the rover, was installed at Terra Nova station to enable differential correction of data. In 1997, improvements were made to the system’s hardware and software. Relevant changes were made to the acquisition unit, the radio-frequency and control unit and the software. A new digitizing board, based on an eight-bit 20 MHz analog/digital flash converter and Fast integrated-circuit components, was developed, along with related software. The sampling rate of the analog signal was doubled to 1024 samples every 50 ns, so as to increase the accuracy of ice-depth determination (±4 m). The new system, with a 166 MHz Pentium-based personal computer and a TX trigger pulse repetition rate of 50 Hz, allows an acquisition rate of ten records every second.

Fig. 2. Landsat TM satellite image map with positions and name of RES profiles, velocity data points determined by tracking technique and GPS, location of accumulation/ablation stakes, position and name of gate and location of Figures 7 and 8.

The survey was completed in four flights (two in 1995 and two in 1997), with a cruise speed of 100–120 kt (185–220 km h⁻¹) and an average altitude of 1000 ft (300 m; controlled by radar altimeter). In 1995 the data-acquisition rate was every 150–185 m; in 1997 the acquisition rate was every 5–6 m. A synchronized GPS with ±20 m precision in x, y coordinates (only pseudo-range differential corrections were made) was used to geo-reference the radar data. Priestley, Reeves, David and Mawson Glaciers were surveyed to determine the grounding line of each (by means of longitudinal profiles) and to evaluate ice thickness along floating portions of the ice tongues (by means of transverse profiles). A total of 35 radar profiles were made: 3 longitudinal and 4 transverse at Priestley Glacier, 3 longitudinal and 2 transverse at Reeves Glacier, 4 longitudinal and 11 transverse at David Glacier–Drygalski Ice Tongue, and 2 longitudinal and 6 transverse at Mawson Glacier. All radar profiles were processed with a software package specially designed for the new radio-echo sounder. A constant electromagnetic wave propagation velocity of 168 m μs⁻¹ was assumed. Thickness was determined to the nearest digital sample on the record with a digitizing error of ±100 ns (1995) and ±50 ns (1997) for both surface and bed reflections. This corresponds to a rms uncertainty in thickness of about ±16 m (1995) or ±8 m (1997). Ice-thickness data at each intersection of longitudinal and transverse profiles were compared; measurements differed by <10 m, i.e. less than the estimated accuracy in thickness determination. By combining 1995 and 1997 data, it was possible to determine the ice thickness and to locate the grounding line of the surveyed glaciers.

Surface velocity measurements

Several investigators have shown that ice-flow velocities can be determined by measuring the displacement of features observed in pairs of visible or synthetic-aperture radar images (e.g. Reference MorganMorgan, 1973; Reference Lucchitta and FergusonLucchitta and Ferguson, 1986; Reference Bindschadler and ScambosBindschadler and Scambos, 1991; Reference Lucchitta, Mullins, Allison and FerrignoLucchitta and others, 1993, Reference Lucchitta, Rosanova and Mullins1995; Reference Bindschadler, Vornberger, Blankenship, Scambos and JacobelBindschadler and others, 1996; Reference Frezzotti, Capra and VittuariFrezzotti and others, 1998a). Landsat TM 4 and 5 (1990–92) images of Priestley, Reeves, David and Mawson Glaciers, Drygalski Ice Tongue and the Nansen Ice Sheet, and SPOT 1 XS (1988) images of Reeves and Priestley Glaciers and the Nansen Ice Sheet were acquired. All images were provided in digital format on computer tape and were processed to eliminate scan-lines in each spectral band. A first-principal-component image was generated to enhance the surface features of glaciers and to reduce image noise (Reference Bindschadler and ScambosBindschadler and Scambos, 1991). New displacement velocities were measured by comparing the 1990 Landsat TM reference image with the 1988 SPOT XS and the 1992 Landsat TM images. Images were co-registered using 19 tie points for the 1990 and 1992 Landsat TM data of David–Drygalski, Priestley and Mawson Glaciers, and 20 points for the 1992 Landsat TM and 1988 SPOT XS imagery of Reeves Glacier. Small rock outcrops (a few pixels in size) near the glaciers, at low elevation and widely scattered across the velocity-measurement area, were chosen as image tie points. Images were co-registered using a first-order polynomial, and the data were resampled using a cubic convolution algorithm to a common pixel size of 28.5 m (that of the reference image). Residual errors from the co-registration process range from the two-pixels to sub-pixel level. Because Drygalski Ice Tongue extends some 90 km out to sea, co-registration errors may increase to the east along the ice tongue where there are no fixed features to be used as tie points. The displacement of surface features (crevasses, ice fronts, snowdrifts, drift plumes, etc.) in sequential images was determined using a semi-automatic procedure. Distinct features that occur in both images were identified first, then a 16 pixel by 16 pixel image chip containing a feature was extracted from each image. A comparator then determined the relative location of the feature in the two chips to a 1/4 pixel accuracy. Based on the measured feature displacement, the known time gap between images and image pixel size (28.5 m), the average velocity of each feature point was calculated in image coordinates. The time gaps for the different image pairs vary from 2 to 4 years. A total of 750 points were measured on the 1990 and 1992 Landsat TM images of David Glacier–Drygalski Ice Tongue; 650 points in the 1988 SPOT XS and 1992 Landsat TM images of Reeves Glacier–Nansen Ice Sheet; and 430 and 100 points in the 1990 and 1992 Landsat TM images of Priestley and Mawson Glaciers, respectively.

Satellite image maps, with 30 m pixel spatial resolution, were created using the Landsat TM reference image (Fig. 2). The image was geo-referenced by identifying image points corresponding to 15 ground-control points established by the Italian Antarctic Research programme with GPS surveys. The image was rectified to a Lambert Conformal Conic cartographic projection by using a linear conversion matrix, with a rms error of about two pixels. The same conversion matrix was applied to the satellite-image velocity points, which were rectified to the same projection. Reference Frezzotti, Capra and VittuariFrezzotti and others (1998a) have shown that flow rates derived from the analysis of satellite imagery compare well with those obtained from GPS measurements of Priestley, Reeves and David Glaciers. When compared to GPS measurements in areas close to image tie points, the error in average velocity measurements using feature tracking may be as little as ±15−20 m a⁻¹ (e.g. grounding-line area). In areas far from tie points, such as the outer part of Drygalski Ice Tongue, the two types of measurements differ by about ±70 m a⁻¹.

Ice flux

The co-registration of airborne radar data and satellite imagery was based on geographic coordinates. In regions with gaps in velocity (e.g. ice margins) or ice-thickness data (e.g. close to valley walls), dummy values were kept consistent with other data (velocity or ice thickness) and with our reconstruction of the morphology of the glacier bed and margin. To calculate the ice-equivalent volume flux (F) at the grounding line and along flow transects, Equation (2) (modified) of Reference Bindschadler, Vornberger, Blankenship, Scambos and JacobelBindschadler and others (1996) was used:

where w is the width, V _t is the velocity component perpendicular to the transect, H is the ice thickness and Δw is the increment in width (Table 1).

Since we do not know how the ice velocity varies with depth inland, ice flux was measured only close to the grounding line and along the floating part of the glacier, where it can safely be assumed that the velocity is constant with depth. Where radar surveys were not perpendicular to the flow across a gate, a projection of the nearest ice-thick-ness measurement was used. The grounding lines of outlet glaciers were surveyed using satellite imagery and radar data by identifying linear transverse depressions and change in surface slope that cut across glaciers in the satellite images and bottom crevasses in the radar profiles. Figure 3a shows the Drygalski Ice Tongue longitudinal profile AB surveyed by radar, with features characteristic of bottom crevasses observed about 130 km from the ice front. Ice thickness ranges from 200 to >1500 m.

Fig. 3. Radar profiles of David Glacier and Drygalski Ice Tongue. (a) Longitudinal profile BD; arrows indicate the position of transversal. (b, c) Transversal radar profiles EF and CD; vertical lines indicates the position of the “gate-limited” width. Profile positions in Figures 2 and 7.

Annual surface-accumulation/ablation rates along Drygalski Ice Tongue and the Nansen Ice Sheet were collected between 1991 and 1995 using a stake network (Fig. 2). Those data show spatial variation due to aeolian processes induced by katabatic winds descending from the plateau to Terra Nova Bay (Reference FrezzottiFrezzotti, 1993, Reference Frezzotti1998). Accumulation/ablation data from the stake network were extrapolated over flux gate; using Landsat TM image observations of accumulation, ablation-rate indicators (blue ice, surface morphology, etc).

The flow margins of the Nansen Ice Sheet and Davie Glacier–Drygalski Ice Tongue were surveyed through satellite image analysis. Discharge fluxes across flow band; (termed “gates” in this paper) along the Nansen Ice Shee and Drygalski Ice Tongue were calculated to determine basal melting and freezing rates; rates for regions bounded by two gates and a flow margin were calculated by summing the upstream gate flux with the surface accumulation/sublimation flux, and subtracting the downstream gate flux using equation (3) (modified) of Reference Bindschadler, Vornberger, Blankenship, Scambos and JacobelBindschadler and other; (1996):

where m is melt rate, δH/δt is the rate of thickness change, F and F′ are the up- and downstream fluxes, A is the surface flux and S is the flow-band surface area.

Drainage area

Drainage was mapped using a DEM of Antarctica provided by Reference Rémy, Shaeffer and LegrésyRémy and others (1999). The DEM was derived from the two geodetic 168 day repeat orbits of the European remote-sensing satellite (ERS-1) radar altimeter. Reference Rémy, Shaeffer and LegrésyRémy and others (1999) have analyzed the waveform altimetry data from geodetic orbits to create maps with a 1/30° grid size. The DEM has better than 1 m accuracy over the parts of East Antarctica where slope gradients are <0.5%. Around Victoria Land, where slopes are steeper, tracking was not continuous due to loss of tracking, so altimetry data were integrated with data from digitized U.S. Geological Survey (USGS) reconnaissance maps of Antarctica at a 1:250 000 scale. By integration of the DEM with digitized USGS data, a new DEM of the study area, with a Lambert Equal Area projection and a grid size of 1 km, was created. The catchment basin was delineated by applying standard hydrological modelling tools included in the proprietary GIS ARC/info (version 7.2) to the new DEM; results are similar to those of Reference Vaughan, Bamber, Giovinetto, Russell and CooperVaughan and others (1999). The grounded basin area was calculated taking into account the new grounding line inferred from the analysis of satellite imagery and radio-echo sounding (RES) data.

Drainage area

Priestley and Reeves Glaciers (Figs 2 and 4) drain the southeastern portion of the Talos Dome area and part of the northern Victoria Land névé (southern part of the Deep Freeze and Eisenhower Ranges). David Glacier, the largest outlet glacier in Victoria Land, drains the inner part of the plateau flowing from eastern Dome C, southern Talos Dome and northern Taylor Dome. Mawson and Mackay Glaciers flow through the Scott Coast and drain the northern peripheral portions of Taylor Dome and local ice originating within a few tens of kilometres of the Transantarctic Mountains.

Fig. 4. Delineation of drainage area of Priestley, Reeves, David, Mawson and Mackay Glaciers. Dotted lines indicate drainage system (e-d-d′-D″-E) drawn by Reference Giovinetto and BentleyGiovinetto and Bentley (1985).

On the DEM constructed, the estimated size of the drainage basin of these outlet glaciers is 235 000 km². Figure 4 shows a direct comparison between drainage basins derived by Reference Giovinetto and BentleyGiovinetto and Bentley (1985) and those derived as above. Our drainage results are similar to those of Reference Vaughan, Bamber, Giovinetto, Russell and CooperVaughan and others (1999). The drainage basin derived from the DEM is different close to Dome C from that derived by Reference Giovinetto and BentleyGiovinetto and Bentley (1985) on the basis of historical data (Reference DrewryDrewry, 1983); the catchment area ends 300 km short of the Dome C culmination. The first 300 km of the eastern Dome C drainage area flows into the Ross Ice Shelf and not into the Scott Coast, as suggested by Reference DrewryDrewry (1983) and Reference Giovinetto and BentleyGiovinetto and Bentley (1985). Drainage system e-d-d′-D″-E (Fig. 4) of Reference Giovinetto and BentleyGiovinetto and Bentley (1985) and Reference Vaughan, Bamber, Giovinetto, Russell and CooperVaughan and others (1999) includes the McMurdo Dry Valleys region, the largest ice-free area in Antarctica, which is not considered in this paper. Along the Scott Coast there are other glaciers (Marin, Harbord, Davis, Clarke and Larsen; Fig. 2) which drain a small area of local nunataks of about 5250 km². The ice thickness of these glaciers is not available at the grounding line and they are not considered in this paper.

Outlet-glaciers discharge

Priestley and Reeves outlet glaciers occupy deep troughs (Fig. 2) that stem from cirques cut into the bedrock beneath inland ice (Reference Orombelli, Baroni and DentonOrombelli and others, 1991). Reeves and Priestley Glaciers merge and form the Nansen Ice Sheet (the name given by the first explorers, although it is technically an ice shelf). The grounding lines of these glaciers were mapped by Reference Frezzotti, Baroni and BondesanFrezzotti and others (1996) using satellite imagery; the grounding-line locations have been confirmed by the new radar data. Reference Frezzotti, Baroni and BondesanFrezzotti and others (1996) place the grounding line of Priestley Glacier close to Black Ridge, about 62 km from the ice front of the Nansen Ice Sheet. An ice-thickness transect of Priestley Glacier was made 15 km upstream of the grounding line. At this transect (Fig. 5a) the ice is > 1000 m thick and ice velocity reaches 130 m a⁻¹. The ice velocity was calculated using new satellite-image tracking and was confirmed by GPS measurements (Reference Frezzotti, Capra and VittuariFrezzotti and others, 1998a). Based on these data, the estimated ice discharge of Priestley Glacier is 0.77 ± 0.13 km³ a⁻¹ (Fig. 6; Table 1).

Fig. 5. Radar profiles of Priestley (a), Reeves (b) and Mawson (c) transects; profile positions in Figures 2 and 8.

Fig. 6. Four transects of ice thickness and surface velocity of Priestley, Reeves, David and Mawson Glaciers.

Where Reeves Glacier flows into the Nansen Ice Sheet it separates into two streams because of interaction with Teall Nunatak (shown later, Fig. 8) and with buried topography (Reference Baroni, Frezzotti, Giraudi and OrombelliBaroni and others, 1991). Reference Frezzotti, Baroni and BondesanFrezzotti and others (1996) located the grounding line just at Teall Nunatak, about 40 km upstream of the front of the Nansen Ice Sheet. Figure 5b illustrates the radar cross-section of Reeves Glacier close to the grounding line, acquired during the 1995 expedition. In spite of the low horizontal resolution (1 record/170 m), ice thickness can be measured along the whole section. The feature at 10 km is Teall Nunatak which splits the ice-flow discharge. At the grounding line the ice is up to 700 m thick and ice velocity reaches 200 m a⁻¹ (Fig. 6; Table 1). Ice-thickness and velocity data (satellite tracking confirmed by GPS) are known across the grounding line, and ice discharge is calculated to be 0.52 ± 0.06 km³ a⁻¹.

David Glacier is the main outlet glacier of Victoria Land, coinciding with a subglacial trench which crosses the Transantarctic Mountains, and linking the Ross Sea basin with the internal Wilkes subglacial basin (Reference Steed, Drewry and CraddockSteed and Drewry, 1982). As most of David Glacier basin is below sea level, it is considered to be a marine-based glacier (Reference OrombelliOrombelli, 1991). David Glacier is fed by two main flows, a northern one from Talos Dome and the major southern flow from Dome C (Reference FrezzottiFrezzotti, 1993). Satellite images and GPS measurements show that the southern flow is formed by the convergence of three ice streams under an icefall, the David Cauldron (Fig. 7). Satellite images display three ice streams separated by heavily crevassed shear zones from ridges of slow-moving ice on either side. These ice streams have widths of > 10 km within 150 km of David Cauldron. Based on GPS measurements, ice velocities within these flows are 50–100 m a⁻¹, whereas ice velocities outside the flows are 20–27 m a⁻¹ (personal communication from L. Vittuari, 1998). At David Cauldron the streams enter a deep fjord-like valley (Reference McIntyreMcIntyre, 1985). Reference FrezzottiFrezzotti (1993) found that the grounding line is about 50 km inland from the coast, where transverse depressions, characteristic of a rapidly spreading flow in the floating glacier, can be seen in satellite imagery (Fig. 2). These observations are confirmed by radar surveys and by kinematic GPS measurements of tidal motion for the southern flow (Reference Frezzotti, Capra and VittuariFrezzotti and others, 1998a, b). Ice thickness in section CD, close to the grounding line, is 1100–1700 m (Fig. 3c). Radar cross-section CD runs close to the grounding line of the southern flow but, unfortunately, does not cover the entire southern ice flow. Cross-section EF (Fig. 3b) shows the thickness of the northern flow (about 700–900 m) compared to that of the southern flow (1000–1300 m). Radar survey EF runs 10 km downstream of the grounding line of the southern flow and 5 km downstream of the grounding line of the northern flow. The ice velocity close to the grounding line of the southern flow is up to 560 m a⁻¹, whereas the ice velocity of the northern flow is up to 120 m a⁻¹ (Fig. 6; Table 1). Ice thickness at the grounding line reaches 1740 m for the southern and 920 m for the northern flow. Ice discharge along transect EF is 5.0 ± 0.2 km³ a⁻¹ for the southern and 0.9 ± 0.1 km³ a⁻¹ for the northern flow. Between radar surveys CD and EF, ice thickness decreases at a rate of 16.4 ± 5.8 m a⁻¹ due to basal melting (see next subsection). By summing of basal melting rates, ice-discharge rates were calculated: 6.9 ± 0.6 km³ a⁻¹ for the southern flow (5.0 ± 0.2 km³ a⁻¹ at section EF plus 1.9 ± 0.4 km³ a⁻¹ of basal melting between section EF and the grounding line) and 7.8 ± 0.7 km³ a⁻¹ for both southern and northern flow of David Glacier.

Fig. 7. Landsat TM image of David Glacier with positions and name of RES profiles, velocity data points determined by tracking technique and GPS, location of accumulation/ablation stakes, position and name of gates and flow limits. See Figure 2 for location of this region.

Satellite imagery and radar data show that the grounding line of Mawson Glacier is located at the slope of Mount Murray, about 15 km upstream of the coast. Close to the grounding line, ice thickness (Fig. 5c) reaches 620 m and ice velocity reaches 190 m a⁻¹. Ice discharge is estimated to be 0.52 ± 0.07 km a⁻¹ (Fig. 6; Table 1). Reference FrezzottiFrezzotti (1997), based on ice-velocity data derived from satellite tracking and on ice-thickness data reported by Reference Keys and FowlerKeys and Fowler (1989), calculated an ice discharge of 0.33 ± 0.06 km³ a⁻¹ for Mackay Glacier.

Mass balance

Based on these data, the ice-discharge rate of outlet glaciers is 10.0 ± 1.0 km³ a⁻¹ (Table 1), and, assuming an ice density of 917 kg m ⁻³, we have a value of 9.1 ± 0.9 Gt a⁻¹. Based on ice- discharge rates (9.1 ± 0.9 Gt a⁻¹) for the new catchment area (235 000 km²) and assuming steady state, a surface-snow accumulation balance of 39 ± 4 kg m a⁻¹ is estimated. For this area (e-d-d′-D″-E drainage system), Reference Giovinetto and BentleyGiovinetto and Bentley (1985) and Reference Vaughan, Bamber, Giovinetto, Russell and CooperVaughan and others (1999) reported a net surface accumulation rate on grounded ice of 74 and 97.3 kg m⁻² a⁻¹, respectively, and a net surface balance of 19.6 and 25.5 Gt a using snow-accumulation rates derived from historical snow-pit stratigraphy (Table 2). The ice discharge of outlet glaciers that flow along the Scott Coast is less than half that required for a zero net surface mass balance according to the inputs given by Reference Giovinetto and BentleyGiovinetto and Bentley (1985) and Reference Vaughan, Bamber, Giovinetto, Russell and CooperVaughan and others (1999). For David Glacier, McIntyre (in Reference SwithinbankSwithinbank, 1988) assessed a balanced discharge of 14 km³ a⁻¹ for a drainage area of 224 000 km². Reference FrezzottiFrezzotti (1997) has estimated iceberg calving (19.4 ± 20% Gt a⁻¹) and floating-glacier melting (20 ± 50% Gt a⁻¹) for the glaciers that fringe the entire Victoria Land coast. Comparison with Reference Giovinetto and BentleyGiovinetto and Bentley’s (1985) accumulation value (64.3 ± 20% Gt a⁻¹) suggests a combined positive mass balance of these glaciers; the imbalance relative to the output flux that Reference FrezzottiFrezzotti (1997) measured represents 37% snow accumulation.

As there is evidence from radar altimetry that the drainage system is not grossly out of steady state (Reference Lingle and CoveyLingle and Covey, 1998; Reference Rémy and LegrésyRémy and Legrésy, 1998), the explanation for this large apparent imbalance is probably an over estimation of snow accumulation using historical data from snow-pit stratigraphy (1959–60 U.S. Victoria Land Traverse; Reference Stuart and HeineStuart and Heine, 1961).

Recent studies of the International Trans-Antarctic Scientific Expedition (ITASE) Italian programme (unpublished information from B. Stenni and others) have revealed an overestimation of snow accumulation in the Talos Dome area. For example, chemical and isotopic analyses of snow–firn cores show accumulation rates up to 50% less than previously reported (Reference Stuart and HeineStuart and Heine, 1961); moreover, Reference Cullather, Bromwich and van WoertCullather and others (1998), using atmospheric moisture budget, have identified low net precipitation along the David, Mackay and Mawson catchment area. The ITASE Italian programme will continue to study the area extending from the coast to Dome C in order to refine the surface mass balance over the entire basin that drains the ice sheet across the Scott Coast.

Floating-glacier discharge flux

Ice-thickness, ice-velocity and accumulation/ablation data from airborne surveys along the floating extensions of the outlet glaciers were collected to study the pattern of basal melting/freezing processes (the glacier is assumed to be in a steady state). The northern flow band of Reeves Glacier within the Nansen Ice Sheet, and the central flow band of David Glacier and Drygalski Ice Tongue were investigated.

The discharge flux of the Reeves flow band was investigated at four gates (Fig. 8) located within 30 km of the grounding line. The flow band is 4.5–6.0 km wide, with average ice thicknesses of 661–271 m and average ice velocities of 113–168 m a⁻¹ (Table 3). Blue ice at the surface of the flow band is exposed to katabatic winds. Stake measurements indicate ablation rates ranging from 500 ± 100 kg m⁻² a⁻¹ (close to the grounding line) to 400 ± 80 kg m⁻² a⁻¹. The rate of change in thickness indicates basal melting values of 0.26 ± 0.92 m a⁻¹ from the grounding line (gate R1–R2) to gate R3–R4, 16 km away, and 0.55 ± 0.2 m a⁻¹ from gate R3–R4 to gate R5–R6 (Table 3). A basal freezing rate of 0.75 ± 1.83 m a⁻¹ was calculated between gate R5–R6 and gate R7–R8. These melting/freezing rates agree with those provided by Reference FrezzottiFrezzotti (1997) for Carnein Glacier (basal melting rate 650–750 kg m⁻² a⁻¹) and by Reference BaroniBaroni (1990) and Reference Tison, Lorrain, Bouzette, Dini, Bondesan, Stiévenard and JeffriesTison and others (1998) for Hells Gate ice shelf (basal freezing rates of 0.10–0.58 m a⁻¹; Figs 2 and 8)

Fig. 8. Landsat TM image of the Nansen Ice Sheet (Reeves branch) with positions and name of RES profiles, velocity data points determined by tracking technique and GPS, location of accumulation/ablation stakes, position and name of gates and flow limits. See Figure 2 for location of this region.

The discharge flux of David–Drygalski Glacier was measured at six gates (Figs 2 and 7) located up to 120 km from the grounding line. The flow band is 3.5–10 km wide, with average ice thicknesses of 1409–197 m and ice velocities of 529–714 m a⁻¹ (Table 4). Snow accumulation (measured at stakes) ranges from 50 ± 10 to 185 ± 37 kg m a^{−2 −1}. An area with a surface ablation rate of 80 ± 16 kg m⁻² a⁻¹ lies between gates D9–D10 and D11–D12 (Fig. 9). The rate of ice-thickness change (m − δH/δt) ranges from −16.4 ± 5.8 m a⁻¹ near the grounding line to +8.6 ± 12.0 m a⁻¹ between gates D7–D8 and D9–D10, where the ice tongue leaves the coast (Table 4).

Fig. 9. Schematic sketch of the David Glacier–Drygalski Ice Tongue ice/ocean interaction.

Ice/ocean interaction

Basal melting/freezing processes below floating glaciers of the Terra Nova Bay area have been discussed by various authors (Reference BaroniBaroni, 1990; Reference SouchezSouchez and others, 1991, Reference Souchez1995; Reference FrezzottiFrezzotti, 1993; Reference Tison, Lorrain, Bouzette, Dini, Bondesan, Stiévenard and JeffriesTison and others, 1998). These processes are driven by deep thermohaline circulation triggered by high-salinity shelf water (HSSW) formed as a result of the expulsion of salt during sea-ice formation. Terra Nova polynya is one of the Ross Sea’s most important “sea-ice factories”, estimated to produce 5–15% of the sea ice generated in the entire Ross Sea shelf (Reference Kurtz, Bromwich and JacobsKurtz and Bromwich, 1985; Reference Frezzotti and MabinFrezzotti and Mabin, 1994).

Analyses of thickness variations along the flow bands of David–Drygalski (Fig. 9) and the Nansen Ice Sheet show large uncertainties, but lead to several conclusions about ice/ocean interaction in this region:

1. (1) David Glacier has a high basal melting rate (16.4 ± 5.8 m a⁻¹) at 1200 m below sea level where ice starts to float, and decreases downstream. The melting rate at the grounding line is higher than previously hypothesized (e.g. Reference Jenkins and DoakeJenkins and Doake, 1991; Reference Jacobs, Hellmer, Doake, Jenkins and FrolichJacobs and others, 1992; Reference FrezzottiFrezzotti, 1993) for thermohaline circulation led by HSSW. This high basal melting rate is close to the predicted melting rate for Circumpolar Deep Water intrusion beneath Pine Island Glacier evaluated by Reference Jacobs, Hellmer and JenkinsJacobs and others (1996), but far from the basal melting of 50 ± 10 m a⁻¹ calculated by Reference RignotRignot (1998). A high local basal melting rate for David Glacier could be related to over-production of HSSW in the Terra Nova polynya.

2. (2) Basal melting at the grounding line of the Reeves flow band is about 10% of that at David Glacier. The grounding line of Reeves Glacier is about 550 m below sea level, much shallower than the deep troughs of David Glacier’s southern flow (1200 m below sea level) and Priestley Glacier (about 1000 m below sea level; Fig. 10). The ocean bottom in Terra Nova Bay deepens from the Nansen ice front (700 m below sea level) to the east (about 1000 m below sea level, 5 km from the ice front). The different depths of the grounding lines indicate that David and Priestley Glaciers have deeper sub-ice-shelf cavities which can concentrate HSSW formed in the Terra Nova polynya. This deep-water convection melts large volumes of meteoric ice. The small melting rate at the Reeves grounding line could be linked to intermediate-depth water convection drawn from ice-shelf water (ISW) formed at deeper grounding lines (Priestley and David). The only oceanic profile presently available in Terra Nova Bay (Reference Fabiano, Povero, Catalano and BenedettiFabiano and others, 1991) shows ISW from 100 m deep to 600 m above sea floor (700 m deep). This intermediate-depth convection does not melt large volumes of ice (Reference Jacobs, Hellmer, Doake, Jenkins and FrolichJacobs and others, 1992).

3. (3) Basal accretion is present along the David–Drygalski flow band where the bottom ice slope steepens and the floating glacier leaves the valley, in agreement with the previous hypothesis of Reference FrezzottiFrezzotti (1993). Basal accretion is also active along the Reeves Glacier–Nansen Ice Sheet flow band between gates R5–R6 and R7–R8, where the Reeves flow meets the deeper and thicker Priestley flow (600–300 m; Fig. 10). The location of basal freezing agrees with modelling of Deep Thermohaline Convection, which suggests that HSSW evolves into very cold but relatively fresh and light ISW that returns to the ice front. Where the basal slope is steep (where Drygalski Ice Tongue leaves the coast, and the merging area along the Nansen Ice Sheet between the Reeves and Priestley subglacial cave), supercooling of the water mass produces frazil-ice crystals that accrete at the bottom of floating glaciers and slowly consolidate to form marine ice.

Fig. 10. Radar profiles LG of Nansen Ice Sheet (Priestley branch). Location of profiles in Figure 2.

The net ablation rates of meteoric ice were estimated from the change in ice thickness along flow channels. About 50% and 25% of the ablation of David and Reeves Glaciers, respectively, is due to basal melting, 50% and 45% to iceberg calving and, for Reeves Glacier, 30% to surface ablation induced by katabatic winds. The 50% melting rate for David Glacier is higher than has been reported by other authors (Reference Jacobs, Hellmer, Doake, Jenkins and FrolichJacobs and others, 1992), for whom iceberg calving represents 100–70% of the negative term of the Antarctic mass balance. Estimated mass-loss rates agree with those of Reference FrezzottiFrezzotti (1997), who estimated the iceberg calving flux during the last 50 years for 50 Victoria Land glaciers along 3600 km of coast; he found that mass loss is due to an iceberg calving rate of 19.4 ± 20% Gt a⁻¹ and a floating-glacier melting rate of 20 ± 50% Gt a⁻¹.