1. Introduction

The glaciers that drain the West Antarctic Ice Sheet into the Amundsen Sea are accelerating, presumably because of melting of the oceanic ice shelves (e.g., Pritchard et al. 2009; Shepherd et al. 2004; Thomas 2004; Jacobs et al. 1996; Rignot and Thomas 2002; Walker et al. 2007). The suspect heat source for such glacial melting (Walker et al. 2007; Thoma et al. 2008) is the relatively warm Circumpolar Deep Water (CDW) found in the adjacent oceanic domain, with temperature above 1.8°C near the southern boundary of the Antarctic Circumpolar Current (Orsi et al. 1995). Farther to the south, the Lower CDW available to the Amundsen Sea has neutral density (Jackett and McDougall 1997) between 28.03 and 28.27 kg m⁻³ (Whitworth et al. 1998). Even after mixing with colder surface water as it enters the continental shelf, Modified CDW (MCDW) is still warmer than 1°C in the Amundsen Sea. MCDW with temperatures approximately 3°C above freezing has been observed to penetrate across the east Amundsen shelf into the ice shelves through a deep channel there (Walker et al. 2007). The central Amundsen shelf is separated from the east Amundsen shelf by a 400–500-m-shallow bank (Fig. 1), which extends all the way to Bear Peninsula, where it also divides the deeper ice shelf basins into a western part and an eastern part (Nitsche et al. 2007).

The circulation of MCDW in the central Amundsen shelf is not known. The only existing estimate of MCDW inflow to the Amundsen shelf was done in the eastern channel (Fig. 1) leading to the east Amundsen shelf (Walker et al. 2007). Geostrophic calculations using a hydrographic transect across the channel showed an inflow of 0.25 Sv (1 Sv ≡ 10⁶ m³ s⁻¹). Here, new hydrographic data, complemented with lowered acoustic Doppler current profiler (LADCP) velocity measurements, across the western channel leading into the central Amundsen shelf are presented for the first time.

2. Methods

During December 2008, a survey was conducted in the central Amundsen shelf, including one section across the western channel and three “yo-yo” stations (Fig. 1). Hydrographic data were sampled with the conductivity–temperature–depth (CTD) system from Sea-Bird Electronics, Inc., (SBE)—the SBE 911+ CTD—with double sensor pack: SBE 3+ temperature, SBE 4C conductivity, and SBE 43 oxygen sensors. The sensors were calibrated before and after the cruise, and the data were compensated for the drift in salinity, temperature, and oxygen. Current shear was measured using a Teledyne RD Instruments, Inc., 300-kHz Workhorse Sentinel LADCP attached to the rosette sampler. The LADCP had bottom track everywhere where MCDW was detected. The velocity data were detided using the barotropic tide model (Circum-Antarctic Tidal Simulation Model CATS2008b) described in Padman et al. (2002), using all 10 available tidal components (further details of the model can be found online at http://www.esr.org/~padman/Antarctic_Tides/cats0201.html).

Historical data from the World Ocean Database (Boyer et al. 2006) have also been included. All stations located on the shelf between 71° and 75°S and between 130° and 100°W that were available in the database were used. The data collected by Walker et al. 2007 were not present in the database.

3. Results

Figure 2 shows across-channel sections (note that the easternmost part, stations 33 and 34, were occupied 3 days before the rest of the section) of temperature T, salinity S, dissolved oxygen O₂, and the detided along-channel velocity component. Thin white lines show contours of neutral density γ_N (Jackett and McDougall 1997), and the thick white line shows γ_N = 28.03 kg m⁻³, which marks the boundary between MCDW and overlying water masses (Whitworth et al. 1998).

Three main water masses can be seen in the across-channel sections. Extending from the surface down to 200–400-m depth is cold Antarctic Surface Water [often also referred to as Winter Water (WW); Jenkins 1999], which is defined as having γ_N < 8.03 kg m⁻³ and potential temperature θ < −1.6°C (Bindoff et al. 2000). It is characterized by oxygen values higher than 5 mL L⁻¹ and has blue color in the temperature section and green to blue colors in the salinity section. At the bottom, in the northeastern part of the channel, there is a layer of comparatively warm and salty MCDW with γ_N > 28.03 kg m⁻³, T ≈ 1°C, S ≈ 34.7 psu, and O₂ ≈ 4.3 mL L⁻¹. Below the WW and above the MCDW (where the MCDW layer is present), there is an approximately 150-m-thick layer of water with 27.75 < γ_N < 28.03 kg m⁻³, T ≈ 0°C, S ≈ 34.3 psu, and oxygen values below 4.8 mL L⁻¹.

The warmest and saltiest water is concentrated along the eastern flank of the channel and shoaling toward the northeast end, giving rise to a significant upward slope of the neutral density surfaces at depths below 250 m. The slope is consistent with a southward geostrophic flow along the channel as evident in the LADCP data. The along-channel velocities from LADCP measurements (Fig. 2d) show generally currents in the southerly direction from the bottom to 100–150 m above bottom. The southward current is strongest, with speeds up to 10 cm s⁻¹ at stations 56 and 57, at the deepest part of the channel, and at station 63, farther up the slope. Five of the stations, 58–62, were very closely spaced, and only the temporal average of the stations is plotted in Fig. 2.

During the cruise, three time series yo-yo stations were occupied (Fig. 1), during which the ship drifted with the ice and CTD casts were performed every 2 h. A comparison between the velocity measured by the LADCP (the vertical average in the water column) at the yo-yo stations and the on-shelf tidal current component, calculated by the tide model in Padman et al. (2002), is shown in Fig. 3. The tidal currents are generally less than 3 cm s⁻¹, and the observed velocities are consistent with the tide model results.

The cumulative along-channel transport of MCDW (i.e., water with γ_N > 28.03 kg m⁻³ bounded by the thick white line in Figs. 2a–c) is shown in the bottom panel of Fig. 3. The dashed line is the (detided) LADCP data, and the solid line is the geostrophic transport. The geostrophic transport was referenced to the surface; that is, the baroclinic pressure gradient obtained from the CTD section was used to calculate the baroclinic part of the geostrophic velocity, and the barotropic part was obtained from the (detided) LADCP measurements in the surface. Figure 3 shows the sum of the barotropic and the baroclinic part. Both LADCP and geostrophic calculations indicate that approximately 0.3–0.4 Sv of MCDW entered onto the central Amundsen shelf through the western channel, which can be compared to the geostrophic transport of 0.25 Sv measured by Walker et al. (2007) for the same water mass in the eastern channel.

Figure 4 shows T–S diagrams as well as oxygen concentration of the data from the present cruise and historical data from the World Ocean Database (Boyer et al. 2006). The historical data have been divided into those taken in the vicinity of ice shelves (green dots in Fig. 1) and others (pink dots in Fig. 1). Also shown in Fig. 4 is the surface freezing temperature (red line). The solid black line is the “Gade line” discussed below.

4. Meltwater mixture

When relatively warm ocean water comes into subsurface contact with ice, the ice will warm up until it begins to melt. Unless the ice is freshly calved and colder than approximately −30°C, the energy for warming the ice to freezing temperature (i.e., the sensible heat transfer from ocean to ice) is small relative to the energy for the melting of the ice (i.e., the latent heat transfer from ocean to ice). Assuming that the sensible heat transfer is negligible relative to the latent heat transfer and that the volume of meltwater is small relative to the volume of ocean water (and hence the energy for warming the meltwater is also negligible relative to the latent heat transfer), then the ocean–meltwater mixture that results from such a process will have temperature T_P and salinity S_P given by (e.g., Gade 1979; Jenkins 1999)

View Expanded

where T_OCEAN and S_OCEAN are, respectively, temperature and salinity of the ocean water prior to the melting; L_F = 334 kJ kg⁻¹ is the latent heat of fusion for ice; and C_P = 3.97 kJ kg⁻¹ K⁻¹ is the specific heat of water with salinity of 34.7 psu, temperature of 1°C, and pressure of 400 dbar. Equation (1) is hereafter referred to as the Gade line. It has been plotted in Fig. 4 (black solid line) using the characteristic T and S for MCDW: that is, T_OCEAN ≈ 1°C and S_OCEAN ≈ 34.7 psu. Water that is a mixture between the MCDW and the ice shelf meltwater, referred to here as MCDW–meltwater mixture, will fall close to the Gade line.

In Jenkins (1999), the hydrographic properties of water in the east Amundsen shelf and the effect of subsurface glacier melting (from the Pine Island Glacier) on the water masses were examined. Based on stations in the east Amundsen shelf, it was shown that the water on the shelf that is not affected by subsurface ice melt is a mixture of MCDW (with T ≈ 1°C, S ≈ 34.7 psu, O₂ ≈ 4.3 mL L⁻¹) and WW (with T ≈ −1.7°C, S ≈ 34 psu, O₂ ≈ 7 mL L⁻¹) that falls on a straight line in the S–O₂ plane and in the T–O₂ plane. Subsurface melting of the Pine Island Glacier lowers the salinity and the temperature but does not increase the oxygen content as much as mixing with WW does, and hence the MCDW–meltwater mixture deviates from the mixing line between MCDW and WW. In Fig. 5, the mixing line between MCDW and WW (Jenkins 1999; Jenkins et al. 1997) has been plotted in the S–O₂ plane and in the T–O₂ plane. Also shown are the present dataset (circles) and the historical data (plus signs). The color denotes R, the deviation from the Gade line in degrees Celsius, given by

View Expanded

where T_i and S_i are, respectively, the measured temperature and salinity and T_P(S_i) is the temperature calculated from Eq. (1).

The present data, unlike the historical data, show a strong signal of MCDW–meltwater mixture in the T–S, T–O₂, and S–O₂ planes. In the T–S plane, the strong signal is indicated by the fact that the data points follow the Gade line and have low oxygen values. In the T–O₂ and S–O₂ planes, it is indicated by the fact that the data have lower oxygen values than the mixing line between MCDW and WW and that they have low R values. The only historical stations that show the presence of the MCDW–meltwater mixture are three stations very close to Pine Island Glacier (Jenkins 1999; Jacobs et al. 1996). Data from these three stations are marked in Fig. 5.

To determine the spatial distribution of MCDW–meltwater mixture, an across-channel section plot of R [cf. Eq. (2)] has been constructed (Fig. 5, bottom). The MCDW–meltwater mixture is clearly identifiable and occupies a 100–150-m-thick layer between the incoming MCDW (bounded by the thick white line) and the WW (yellow and red colors). The MCDW–meltwater mixture is relatively warm (T ≈ 0°C), salty (S ≈ 34.3 psu), and oxygen poor (O₂ < 4.8 mL L⁻¹) relative to the WW but colder and fresher than the MCDW. Previous observations of MCDW–meltwater mixture (e.g., in the east Amundsen shelf; Jacobs et al. 1996; Jenkins 1999) show thin intrusions of MCDW–meltwater mixture interleaving into the ambient water. A distinct water layer consisting of MCDW–meltwater mixture has not been previously reported in the region.

The MCDW–meltwater mixture is approximately 1°C cooler than the MCDW. The heat transferred to the ice shelves or icebergs through subsurface melting is hence estimated to be 1.2–1.6 TW (using the measured transport of 0.3–0.4 Sv), equivalent to melting 110–140 km³ ice annually if the melting persists throughout the year. This value can be compared, for example, to the recent estimate of the dynamic thinning of the glaciers feeding the Dotson and Crosson Ice Shelves (Pritchard et al. 2009). Using a satellite-borne laser altimeter, the thinning in 2003–07 in the most rapidly moving parts was estimated to exceed 9 m yr⁻¹, which gives a net volume loss there of approximately 50 km³ yr⁻¹.

Regardless of whether the subsurface ice melt occurs underneath the ice shelves or by melting of icebergs, we expect the mean flow of the MCDW–meltwater mixture to be directed out from the shelf region. Because the cross-sectional area occupied by this water mass is larger than the one occupied by MCDW, the average velocity in the layer will be smaller. Using a cross-sectional area of approximately 25 km² and the measured inflow of 0.3–0.4 Sv gives an average outward velocity in the MCDW–meltwater mixture layer of less than 1.5 cm s⁻¹—probably too slow to detect.

5. Discussion

The present results show the existence of a large body of water produced by MCDW that has been cooled and freshened in a subsurface melting process of ice shelves and/or icebergs. This MCDW–meltwater mixture occupies a 100–150-m-thick intermediate layer between the MCDW and the WW. Previous observations of MCDW–meltwater mixture on the shelf (e.g., Jenkins 1999; Jacobs et al. 1996) has been in close proximity to the ice shelves (e.g., Jenkins and Jacobs 2008; Potter et al. 1984, 1988) where it has formed thin intrusions into the ambient water. However, only approximately 6–10 of the historical stations from the Amundsen shelf were located in the deep channels, and most of these stations were in the eastern channel (Fig. 1). Hence, it is possible that the MCDW–meltwater mixture has been present on the shelf previously but not detected. In March 2003 (Walker et al. 2007), the entrance of the eastern channel was transected with closely spaced CTD stations that show the presence and inflow of MCDW onto the shelf. However, the layer of WW extended all the way down to the MCDW; if there was any MCDW–meltwater mixture present, it occupied only a very thin layer. The MCDW–meltwater mixture may exit the shelf at another location: for example, flowing westward along the coast first, steered by the earth’s rotation. Planned future field studies will attempt to better resolve the spatial distribution and shorter-term time changes of the meltwater-enriched deep flow.