The Antarctic Coastal Current in the Bellingshausen Sea

Abstract. The ice shelves of the West Antarctic Ice Sheet experience basal melting
induced by underlying warm, salty Circumpolar Deep Water. Basal meltwater,
along with runoff from ice sheets, supplies fresh buoyant water to a
circulation feature near the coast, the Antarctic Coastal Current (AACC). The formation, structure, and coherence of the AACC has been well documented along the West Antarctic Peninsula (WAP). Observations from instrumented seals collected in the Bellingshausen Sea offer extensive hydrographic coverage throughout the year, providing evidence of the continuation of the westward flowing AACC from the WAP towards the Amundsen Sea. The observations reported here demonstrate that the coastal boundary current enters the eastern Bellingshausen Sea from the WAP and flows westward along the face of multiple ice shelves, including the westernmost Abbot Ice Shelf. The presence of the AACC in the western Bellingshausen Sea has implications for the export of water properties into the eastern Amundsen Sea, which we suggest may occur through multiple pathways, either along the coast or along the continental shelf break. The temperature, salinity, and density structure of the current indicates an increase in baroclinic transport as the AACC flows from the east to the west, and as it entrains meltwater from the ice shelves in the Bellingshausen Sea. The AACC acts as a mechanism to transport meltwater out of the Bellingshausen Sea and into the Amundsen and Ross seas, with the potential to impact, respectively, basal melt rates and bottom water formation in these regions.


. Length of section (degrees latitude, km), number of winter (April-September), summer (October-March) and total profiles, and the averaging length (degrees latitude, km) for each of the seven hydrographic sections shown in Fig. 5.

Section
Length of Section Length of Section

Horizontal Maps
To assess horizontal variability of physical properties in the Bellingshausen Sea shelf region, the seal data were mapped onto a 1°longitude by 0.5°latitude grid (Fig. 3a). This grid size was chosen to provide the highest resolution on the shelf, while 125 maintaining an adequate amount of grid cells that contain at least one data point. In each grid cell, and for each depth bin, the median values of temperature and salinity were calculated from the seal dives in that cell, as well as the variance. Separate calculations for summer and winter months were also completed. The dynamic height, referenced to 400 m, was calculated from the median values.

Hydrographic Sections 130
Seven composite hydrographic sections, spanning the continental shelf break to the coast, were created to examine how the vertical structure of physical properties in the Bellingshausen Sea changes from east to west. Two of the sections are located in the WAP to compare the seal data with the APCC observations reported in Moffat et al. (2008), which we indicate with the red dot and 'M' in Fig. 9a. The other five sections are located in the Bellingshausen Sea (Fig. 5). A moving median of the seal profiles was taken using a different length scale for each section, based on the available data. For example, in Section 1, the 135 length of the section is 0.75°of latitude and the medians of the properties were taken every 0.075°of latitude. The length scale over which the median was calculated was similar across various sections, but was allowed to vary to ensure that each section avoided gaps with unavailable data. Table 1 provides details about the number of profiles, length, and averaging length for each of the seven sections.

140
To characterize the strength of the AACC, geostrophic velocities and transports perpendicular to each section were calculated based on the density structure. In order to calculate the total geostrophic velocity and transport, a reference level, or level of no  (2008). The 0% meltwater contour (see discussion of our meltwater index below) was used as a rough estimate 150 of the offshore limit for the transport estimates in each section, a similar approach as in Jenkins and Jacobs (2008). For the various hydrographic sections, the offshore extent of the AACC is marked by a dashed line; this location was used to estimate the transport associated with the AACC. We define westward velocities and transports as negative.
To provide an estimate of the error in the velocity/transport calculations, we applied a bootstrapping approach (Efron and 155 Tibshirani, 1994). Along each section, 1000 different composite hydrographic sections were created by randomly selecting only 40% of the profiles in each cross-shelf bin. Geostrophic velocities and geostrophic transports were calculated for each of these 1000 sections and error bars are reports as the root mean square (rms) of these values. The rms values are taken as the difference from the mean composite section using all the data.

160
For each section, meltwater was calculated using the composite tracer method outlined in Jenkins (1999), where definitions of the three water masses that dominate the properties on the shelf are required. These three water masses are CDW, Winter Water (WW), and glacial meltwater. When calculating the meltwater fraction, the same endpoints were used for every profile, as opposed to using individual end members for each profile. Using constant endpoints introduced negative values for meltwater fraction, due to some data points lying to the right of the mixing line for CDW and WW. For this reason, the values we report 165 in this study should be considered a meltwater "index," rather than a quantitative estimate of the meltwater concentration. We emphasize that we only use the meltwater index to define an offshore boundary for our AACC transport estimates. The use of constant end members will locally impact the quantitative estimate of meltwater, but we are concerned with large-scale geographic changes in meltwater concentration (Biddle et al., 2019). The endpoints used for meltwater were the same as those  To understand the effect that different end members have on the meltwater index, a variety of different end members were tested. When shifting the WW end members by decreasing temperature and increasing salinity, the meltwater index increases throughout the entire water column, although not in a uniform manner. The surface layer above 100 m, specifically north of 68.7°S, shows greater increases in meltwater index than the water below it. When shifting the CDW end members to saltier values, the meltwater index increases in some areas of the water column and decreases in others. However, when changing both 180 end members so as to preserve the slope of the WW and CDW mixing line, the meltwater index changes uniformly throughout the water column. We use constant end members to simplify the presentation and preserve quantitative comparisons across the Bellingshausen shelf. AASW, which represents the surface mixed layer, has potential temperatures ranging from -1.8°C to 1°C and salinity ranges from 33 to 34 psu. During austral winter, surface heat loss leads to a deeper mixed layer. In summer, surface heating and

200
Because of the broad coverage of the seal profiles, this data set offers a unique opportunity to construct horizontal mean fields that may be constructed either along isobars or isopycnals. We focus on the latter in the following subsections. These maps provide a more complete picture of hydrographic variations than is typically permitted from discrete hydrographic sections, e.g. (Castro-Morales et al., 2013). Summer melting of sea-ice freshens and cools the surface layers, which combined with heating later in the summer forms the fresher and warmer seasonal thermocline. In the winter, sea-ice formation increases 205 salinity through brine rejection, destroying the surface layer and imprinting hydrographic properties on the deeper WW layer. Thus, AASW shows the most variability in properties due to seasonal surface forcing variations (Whitworth et al., 1998). Our focus in the following is on layers below the surface.

Winter Water and Transition Layer
We illustrate the properties of the WW layer on the σ 0 = 27.4 kg/m 3 surface ( Fig. 3b-d), with median values of isopycnal layer 210 depth, temperature, and salinity taken from all available seal data. Properties off the continental shelf have been removed in panels (b-d) to better highlight variations over the continental shelf. Seasonal variations in the WW properties (divided into six month periods) are provided in the appendix (Fig. A2) A key feature of the WW layer is its downward slope along the entire coast of the Bellingshausen Sea, indicating a baroclinic, westward geostrophic current, under the assumption that the velocity decays with depth (Fig. 3b). The shape of the 27.4 kg/m 3 surface highlights the boundary-trapped nature of the AACC up to shausen Sea as compared to the west, with a difference of roughly 1.15°C. Close to the coast the temperature on this density surface is warmer than what is typically associated with WW, which may be due to upward mixing of warm CDW as these regions are associated with large polynyas (Tamura et al., 2008). Lateral changes in temperature within the coastal boundary current are smaller, but the trend shows a consistent cooling from east to west. Similarly, the salinity of the WW layer varies, freshening from east to west, both broadly over the continental shelf and in the boundary current (Fig. 3d). The difference of 225 the salinity from east to west has a magnitude of roughly 0.055. This cooling and freshening signal in the boundary current could be related to an introduction of meltwater from the ice shelves in the Bellingshausen Sea.
Comparing summer and winter properties of the WW layer reveals that there are larger horizontal gradients in summer as compared to winter months (Fig. A2). Temperature, salinity, and isopycnal layer depth all have larger lateral gradients in

Transport
We next use the composite hydrographic sections to construct geostrophic velocities perpendicular to the section and therefore largely oriented parallel to the shelf break and the coastline. Figs. 6c and 6d show an example of geostrophic velocity and cumulative volume transport for the reference hydrographic section, Section 3. Each of the remaining sections was analyzed similarly and are presented in the Appendix (Fig. A6). Our notation is that negative velocities and transports are directed west-340 ward; the transport is calculated by integrating the velocities with respect to depth and distance from the coast, such that the transport at the shelf break is equivalent to the net along-shelf volume transport in Sverdrups (1 Sv = 10 6 m 3 s −1 ).
Throughout the Bellingshausen Sea continental shelf, the density structure near the coast is consistent with a baroclinic, vertically-sheared flow that is westward near the surface. Additionally, the lateral density gradients intensify and extend further 345 away from the coast as the AACC moves towards the west, suggesting a strengthening of the AACC. Thus the evolution of the AACC, inferred from hydrographic properties through dynamic height, geostrophic velocity, and transport estimates, shows a consistent picture of a connected circulation feature that extends from the WAP through the western Bellingshausen Sea. Geostrophic velocities offer additional information about the structure of the flow and the processes influencing the flow field. Figure 6 shows the geostrophic velocity (panel c) and transport (panel d) for Section 3 (see Fig. A6 for Sections 1 through 7).

365
In order to arrive at an absolute geostrophic velocity, a reference level of no motion (or barotropic velocity) must be selected. reference level gives us confidence that our velocity and transport estimates are reasonable. For the remainder of the section, we will report geostrophic velocities and transports using a 400 m reference level for reasons discussed in Sect. 2.3. The choice of reference level quantitatively changes the transport magnitudes, but it does not impact the spatial variations in the transport, which is the primary focus of this study.

380
Throughout the WAP and Bellingshausen Sea there is westward flow along the coast. The extent of the APCC is defined as the region between the coast and the location where the 0 meltwater index outcrops at the surface, indicated by the dashed line in Fig. 6b. In Section 1, the geostrophic velocity in the AACC has a peak value of -0.21 m s −1 . For this particular section, the meltwater index does not coincide with the boundary current, but the alongshore transport is dominated by flow near the coast (Fig. A6a). Section 2 similarly shows the AACC tightly confined to the coast with a similar peak westward velocity of -0.20 385 m s −1 . Across Section 3, the first section in the Bellingshausen Sea, the velocity of the AACC has an average value of -0.16 m s −1 with a maximum value of -0.26 m s −1 . Here, the AACC occupies a much larger area than in the previous sections (Fig.   6). In Section 4, the average velocity decreases by more than half to -0.07 m s −1 and the maximum velocity is -0.26 m s −1 . The decrease in velocity in Section 4 is associated with our meltwater index extending over most of the shelf although as shown in Sea, has an average velocity of -0.06 m s −1 and a maximum of -0.12 m s −1 . As we discuss below, part of this weakening in the geostrophic velocity is tied to a broadening of the AACC that is better captured by the changing geostrophic transport.

395
The magnitude of the geostrophic velocity is variable across the various composite sections, which could result from a number of factors, including surface forcing effects (modifying the sea surface height) and width of the AACC. The along-coast transport, on the other hand, provides a clearer picture of the evolution of the AACC. The striking feature is a nearly linear trend in volume transport extending from the WAP through the western Bellingshausen Sea. Values along the WAP show that extracted from the seal data. Finally, A6 displays an index of meltwater for each of the sections, indicating roughly the extent of meltwater influence. We present these figures in this Appendix since they help to provide the broader context for the key results presented in the main text. Only the mean WW properties are presented in the main text because this is where the largest along-isopycnal gradients are found in hydrographic properties. Therefore Figs. A3 and A4 include panels to show the median values as described previously.

605
Figures A5 and A6 are provided to complement the information in the reference section (Section 3, Fig. 6). In the salinity panels of Fig. A5, the red dashed line shows the position of the pycnocline (halocline). Note that although the panels are all the same size, they cover different distances; along-section distance is given along the top of the potential temperature panels.
The red dashed line is also included in the geostrophic velocity panels in Fig. A6. The green dashed line shows the position of the 0 meltwater index contour (see discussion in Sect. 2.4). The location where this contour outcrops at the surface is taken 610 as the offshore boundary of the AACC. This is a somewhat arbitrary definition and in some sections it is a poor indicator of the boundary current (e.g. Section 4 and 6). Nevertheless, it is clear in all the transport panels, that most of the along-shore transport is concentrated close to the coast. The contributions to the AACC occur over a broader distance from the coast in the western part of the Bellingshausen Sea. Finally, the distribution of the meltwater index for all seven sections is presented in Fig. A7. Negative values are possible for this diagnostic because we chose to use constant end members for the entire shelf 615 region as discussed in Sect. 2.4.