We report on under-ice profiling float observations of cold, dense, and oxygenated bottom layers on the continental slope of Mac Robertson Land (60°–72°E) in East Antarctica. This bottom layer water mass, with potential temperature in the range −1.8°C < θ < −0.4°C, is identified as modified shelf water. It is a downslope variety of dense water formed on the Antarctic continental shelf in winter and plays an important role in ventilating the deep Southern Ocean. The seasonal evolution of its thickness and density follows the sea ice cycle of growth and decay, reaching a maximum in October–November. The characteristics and location of this modified shelf water are similar to Cape Darnley Bottom Water, thus suggesting the same primary source in the Cape Darnley polynya region. These float data support recent results that the continental shelf along Mac Robertson Land is a significant source of dense waters in East Antarctica.
 Dense waters on the continental shelf around Antarctica are formed in winter as a result of brine rejection and deep convection during active sea ice production. Their production is enhanced in areas where the presence of coastal polynyas leads to increased sea ice formation. These dense waters have near-freezing temperatures and higher oxygen contents than surrounding waters and persist year-round as a bottom layer in major depressions on the Antarctic continental shelf. With suitable bathymetry and current transport, they can flow off the shelf, mix with the surrounding modified Circumpolar Deep Water (mCDW), and sink along the continental slope as modified shelf water (MSW). Thus, dense shelf waters around Antarctica are an important component of the thermohaline circulation in the Southern Ocean, as their export and sinking along the continental slope contribute to the formation of Antarctic Bottom Water (AABW).
 AABW is the water mass that occupies the bottom layer of the Antarctic Circumpolar Current and plays an important role in ventilating the abyssal depths of the world ocean. Three primary source regions of AABW have been established around Antarctica: the Weddell Sea [Foster and Carmack, 1976; Fahrbach et al., 2001], the Ross Sea [Jacobs et al., 1970; Whitworth and Orsi, 2006], and the Adélie Land coast [Rintoul, 1998; Williams et al., 2010]. More recently, Ohshima et al.  identified the Cape Darnley polynya region as a fourth source of AABW. Cape Darnley is situated at the eastern end of Mac Robertson Land, which is a portion of East Antarctica between approximately 60° and 72°E. The Cape Darnley polynya is located at approximately 65°–69°E and covers an offshore area > 104 km2. It was estimated that this newly identified bottom water accounted for 6–13% of the circumpolar total of AABW.
 Year-round observations of dense waters on the continental shelf and slope around Antarctica are sparse because the oceans at those high latitudes are covered by sea ice for most of the year. Ohshima et al.  documented winter measurements collected by seal-mount instruments that showed high-salinity waters over the continental shelf of Cape Darnley and overflowing dense waters on the continental slope. In this paper we report on observations made by under-ice profiling floats during 2010 and 2011 that show the presence and characteristics of MSW on the continental slope north of Mac Robertson Land (Figure 1). These float-based observations show that local production of dense shelf waters is a perennial event and that their export onto the continental slope north of Mac Robertson Land extends across a wide region. More importantly, this independent confirmation of MSW supports the results from Ohshima et al.  that the Cape Darnley area is a significant dense water production region in East Antarctica.
 Observations of MSW along the Mac Robertson Land coast were made by four profiling floats equipped with the ice-avoidance algorithm discussed in Wong and Riser . These four profiling floats were part of a larger group of floats deployed in the East Antarctic sea ice zone around 117–128°E and 64–65°S during September–October 2007 [Williams et al., 2011]. Most of the floats from this deployment moved westward with the Antarctic Slope Current (ASC) but became inactive before reaching 70°E. Only four floats remained active after they migrated west of 70°E and subsequently recorded the presence of MSW north of Mac Robertson Land in 2010 and 2011 (Table 1).
Table 1. The Four Profiling Floats That Observed MSW North of Mac Robertson Landa
Date of MSW Observations
Estimated Longitude Range
aData are available from the Argo Global Data Centers (http://www.coriolis.eu.org; http://www.usgodae.org) by using WMO ID of the floats. Numerical designations of profiles within individual float data record that sampled MSW are listed under Cycle #.
14 Jun to 5 Sep 2011
193 to 205
9 Sep to 24 Nov 2011
209 to 220
5 Apr to 7 Jun 2010
131 to 136, 138 to 140
28 Feb to 20 Jun 2011
178, 180, 184, 190, 191, 193, 194
 The combination of the ice-avoidance algorithm and the use of the Iridium communication system for data telemetry allowed these floats to collect and store conductivity-temperature-depth (CTD) data beneath sea ice in winter and transmit the accumulated data when the sea surface is free of ice in summer. CTD data collected by this group of floats were 2-dbar bin-averaged and were collected during ascent from 2000 dbar every 7 days. Between each CTD profile the floats parked and moved with the currents at 1000 dbar. The four floats described in Table 1 remained active for more than 4 years, which included five austral winters. Their CTD data showed no sensor drift over time and were estimated to have an accuracy of 0.01 for salinity, 0.002°C for temperature, and 2.4 dbar for pressure. These float data accuracies are consistent with results reported in Riser et al. .
 Two of the four floats (World Meteorological Organization buoy identification number (WMO ID) 2900123 and 2900126) were equipped with Aanderaa 3830 Optode sensors that collected dissolved oxygen samples at discrete levels during ascent at the same time as the CTD profiles. The oxygen sampling levels were every 50 dbar from 2000 to 400 dbar, every 20 dbar from 400 to 360 dbar, then every 10 dbar from 360 dbar to the surface. The dissolved oxygen data were postprocessed by using a method similar to Takeshita et al. , giving an accuracy of 10 µmol kg−1 for the adjusted values, which is sufficient for illustrating the high-oxygen characteristic of MSW.
 MSW over the Antarctic continental slope is identifiable as a sharp turn toward cold, fresh values at the bottom of vertical CTD profiles. This bottom feature is ubiquitous in float data from the vicinity of the Mac Robertson Land coast. We focus our description on the most prominent variety of MSW from this area: the bottom layers that show potential temperature values lower than −0.4°C. Note that this is not used as a temperature boundary to identify the top of the bottom layers, but rather, it is used to identify CTD profiles whose bottom values are significantly colder than surrounding waters. As such, the choice of −0.4°C was determined by examining all float data from the East Antarctic region.
 MSW in the range −1.8°C < θ < −0.4°C was found in data from 41 float profiles between April 2010 and November 2011 (Table 1). Almost all of these profiles were collected when the ocean was under an ice cover and the floats were unable to surface to obtain position fixes from satellites. Positions for these under-ice profiles are not known accurately but can be estimated by interpolating between float positions that have satellite fixes from the ice-free months. Linearly interpolating between known positions is a simple way to estimate the mean direction of flow, but it does not account for deviation of the float trajectories from the mean caused by eddy variability and topographic steering. A previous study by Wong and Riser  indicated that along the East Antarctic coast, linearly interpolated float positions could differ from reported positions by an average of 20 km. Thus, by using interpolated positions, we estimated that MSW was observed in the region between 62–72°E and 66–67°S, with an estimated error of 0.5° in longitude and 0.2° in latitude. The mean direction of flow according to interpolated positions is westward. This agrees with the sea surface height field from Meijers et al. , which shows that the regional flow field is dominated by westward jets associated with the ASC and follows the contours of the shelf break. Maximum CTD pressures indicate that most of these MSW profiles were collected at locations with water depths between 1300 and 2000 m (Figure 1), thus confirming their positions near the continental slope.
 The cold, fresh, bottom signature of MSW is accompanied by elevated dissolved oxygen values in excess of 220 µmol kg−1 (Figure 2). Potential temperature within the observed MSW reached values as low as −1.77°C, with corresponding salinity at 34.48, recorded at 1280 dbar during June 2011. The near-freezing bottom temperature indicates a shelf origin, and the elevated dissolved oxygen values provide further evidence that the water mass has recently been near the surface. Salinity as high as 34.67 was observed within this variety of MSW, with θ = −0.75°C at 1930 dbar during October 2011.
θ-S curves from float data show that the observed MSW exists as a dense water mass underriding the salinity maximum of regional mCDW (Figure 3). Whitworth et al.  defined this as AABW on the continental slope and used the neutral density criterion of γn > 28.27 kg m−3 to delineate mCDW from AABW (or MSW). Using this definition, we estimated that the thickness of the observed MSW varied between 50 m and 970 m. This wide range of MSW thickness is due to seasonal variations in MSW properties. During the period from May to November (Figure 2, top row), the MSW layers are thicker and show more interleaving than the period from February to April (Figure 2, bottom row).
 The annual cycle of formation and modification of dense shelf water is related to the cycle of sea ice growth and decay. The months from May to November are the active winter sea ice growth period when brine rejection and enhanced vertical convection lead to the formation of thick layers of dense water on the continental shelf. December to March are the spring melt and summer ice-free period with no vertical convection and hence no production of new shelf water. The remaining shelf water from previous winter gradually loses its near-freezing characteristic through mixing and becomes warmer and thus lighter in density. In a case study of the Adélie Depression, Williams et al.  show that shelf water densities reach a peak during September–October and then gradually decrease from November. Since MSW is the downstream variation of dense waters from the continental shelf, the seasonal evolution of MSW properties on the continental slope naturally follows the processes on the shelf. Our MSW observations on the continental slope show that both the thickness and the density of MSW increase steadily from February to November, with density reaching a maximum in October and thickness reaching a maximum in November (Figure 4a). The increase in density from February to June is mainly due to the decrease in temperature, while from June to October, density increase is mainly due to the increase in salinity (Figure 4b).
 No MSW was observed during December and January. A single January profile showed a sharp hook at the bottom toward cold, fresh, oxygenated values, but its minimum potential temperature only reached −0.2°C. This is a curious fact as MSW of the θ < −0.4°C variety was observed during February and March, much later in the shelf water modification period. We suspect that this sampling irregularity is less a result of seasonal variability and more due to spatial variability of the water mass. Of the four floats that were profiling along the Mac Robertson Land coast, only two were in the estimated MSW region during the austral summer (WMO ID 2900123 and 2900126). Their known positions from those ice-free months indicate that they remained east of ~69°E and north of Prydz Bay during December and January (Figure 1). The lack of MSW observations during December and January could be an indication that the export of cold, dense shelf water from Prydz Bay is sporadic.
4 Discussion and Conclusions
 Our observations of MSW from profiling floats are consistent with the results reported by Ohshima et al. . They documented MSW on the continental slope between 67.5° and 71°E based on seal-mount CTD data, with θ-S characteristics similar to our observations and with a shift to denser bottom values west of ~68.5°E (see their Figure 3c). The key result from Ohshima et al.  comes from their moored time series data (M3 at water depth 2608 m), which show plumes of newly ventilated AABW cascading down the slope to the bottom of the canyons northwest of Cape Darnley from May to January. This Cape Darnley Bottom Water (CDBW) has −1°C < θ < −0.5°C and 34.6 < S < 34.67 (see their Figure 2) and is estimated to have a thickness of > 300 m. It is believed to originate primarily from the Cape Darnley polynya region, an area noted to have the second highest sea ice production of all regions around Antarctica. High sea ice production leads to formation of a relatively large volume of dense shelf water, which flows out of the shelf, mixes with the ambient mCDW, and ultimately gets channeled downslope via canyons to abyssal depths as new CDBW. Our observed MSW shares the same θ-S range and thickness as CDBW, with an estimated position in the same location as CDBW, albeit at shallower depths. These similarities with CDBW suggest that the MSW reported in this study has the same primary source as CDBW.
 Not all modified shelf waters on the Antarctic continental slope descend to abyssal depths as dense plumes via submarine canyons. Some of these waters flow westward along the slope as they descend in approximate geostrophic balance. These broad sheets of dense waters serve to ventilate the interior of the Southern Ocean at their adjusted density range [Baines and Condie, 1998]. Since there is some uncertainty in the positions of the float-based MSW profiles, we cannot determine the exact extent of the MSW westward spreading from its source. However, we note that the four floats listed in Table 1 sampled MSW over an average period of 81 days. During that time they moved in a general westward direction at an average speed of 0.03 ms−1. This equates to a longitudinal distance of about 210 km, or about 5° of longitude at 67°S, over which the bottom characteristics of MSW remain distinguishable. Such spatial coverage is comparable to the dense shelf water overflows on the continental slope between 144°E and 148°E, northwest of the Mertz Depression [Williams et al., 2010].
Williams et al.  pointed out that dense shelf water sources along the East Antarctic coast were found in discrete coastal polynya regions, with the most conspicuous location being the Adélie and George V Land at 140–149°E. Despite having collected nearly 3000 CTD profiles from autonomous floats along the East Antarctic coast between 50°E and 128°E, the Mac Robertson Land coast was the only area where MSW with θ < −0.4°C was observed. It is inconclusive whether this absence of MSW from other areas in the float data record implies a lack of other AABW sources in the polynya regions of East Antarctica. There could be other dense shelf waters that were not sampled by floats, due to the float trajectories being too far from the Antarctic continental margin. Nevertheless, our MSW observations north of Mac Robertson Land from 2010 to 2011, along with those of Ohshima et al.  from 2008 to 2009, highlight the Cape Darnley polynya region as a significant source of dense shelf water for the Weddell-Enderby Basin. This deep basin between the Weddell Sea and the Kerguelen Plateau is a climatically sensitive area, as its bottom waters serve to ventilate the Atlantic sector of the Southern Ocean. We therefore echo the conclusions from these previous authors that this region of East Antarctica should be studied in more detail in order to assess the contribution of its shelf circulation to deep and bottom water formation in the Weddell-Enderby Basin.
 The authors wish to thank Robert Drucker for providing the adjusted dissolved oxygen data from the two floats that carried the Aanderaa Optode sensors. Comments from an anonymous reviewer helped improve the manuscript. The profiling floats used in this study were fabricated and programmed at the University of Washington by Dana Swift, Dale Ripley, and Rick Rupan. The float data were made freely available by the International Argo Program (http://www.argo.net). This work was funded by National Oceanic and Atmospheric Administration (NOAA) grant NA17RJ1232 task 2 to the University of Washington in support of the Argo float project.
 The Editor thanks an anonymous reviewer for his assistance evaluating this manuscript.