Detailed Seismic Bathymetry Beneath Ekström Ice Shelf, Antarctica: Implications for Glacial History and Ice‐Ocean Interaction

The shape of ice shelf cavities are a major source of uncertainty in understanding ice‐ocean interactions. This limits assessments of the response of the Antarctic ice sheets to climate change. Here we use vibroseis seismic reflection surveys to map the bathymetry beneath the Ekström Ice Shelf, Dronning Maud Land. The new bathymetry reveals an inland‐sloping trough, reaching depths of 1,100 m below sea level, near the current grounding line, which we attribute to erosion by palaeo‐ice streams. The trough does not cross‐cut the outer parts of the continental shelf. Conductivity‐temperature‐depth profiles within the ice shelf cavity reveal the presence of cold water at shallower depths and tidal mixing at the ice shelf margins. It is unknown if warm water can access the trough. The new bathymetry is thought to be representative of many ice shelves in Dronning Maud Land, which together regulate the ice loss from a substantial area of East Antarctica.


Introduction
Ice shelves surrounding Antarctica act as buttresses, restraining ice discharge from the continent into the oceans, and therefore regulating Antarctic contributions to sea level rise (Dupont & Alley, 2005). Mass loss from Antarctica has been accelerating over the past 20 years (IPCC, 2019), driven by increased basal melting of ice shelves (Paolo et al., 2015;Pritchard et al., 2012). Accurate knowledge of the geometry of the ice shelf cavities and the properties of the water column beneath them are essential for understanding processes active at the ice shelf ocean interface. Recent studies have highlighted a lack of sub-ice-shelf bathymetry as a "major limitation" (Pattyn et al., 2017) for future projections of Antarctic mass balance. Improved bathymetric mapping allows determination of water access pathways and calculation of spatially and temporally variable melt rates (e.g., Cochran et al., 2014;Goldberg et al., 2019;Milillo et al., 2019;Morlighem et al., 2020;Pattyn et al., 2017;Tinto et al., 2019). In addition, sub-ice-shelf bathymetry also provides information about ice dynamic history. Understanding past ice dynamics and implementing an accurate bathymetry in ice-flow and oceanographic models, is a critical step to improve projections of the evolution of the ice sheets.
The coast of Dronning Maud Land (DML), East Antarctica ( Figure 1) is fringed by numerous small ice shelves. In this area, satellite-derived melt rates are typically low (Rignot et al., 2013). However, the continental shelf is narrow (Figure 1b), meaning the ice shelves of DML are in close proximity to Warm Deep Water (WDW) masses which flow along the continental slope, making this a potentially sensitive region to future change (Hattermann, 2018;Heywood et al., 2013;Thompson et al., 2018). It has also been highlighted as susceptible to marine ice sheet instability (Morlighem et al., 2020;Ritz et al., 2015). In addition, the ice shelf-ocean interactions along the DML coast play an important role in preconditioning the structure and water mass properties of the westward flowing boundary current (Fahrbach et al., 1994;Hattermann, 2018). This current is a key control on warm-water inflow toward the Filchner-Ronne Ice Shelf (Hellmer et al., 2017;Timmermann & Hellmer, 2013) and bottom water formation in the southern Weddell Sea (Meijers et al., 2016;Meredith et al., 2011). The few bathymetric measurements that exist under DML ice shelves have revealed cavities that are much deeper than those included in current gridded data sets of Antarctica (e.g., Fretwell et al., 2013;Morlighem et al., 2020). Under the Fimbul Ice Shelf (Figure 1b), a deep trough within the subshelf cavity was discovered (Nøst, 2004) and confirmed to contain modified WDW (Hattermann et al., 2012. At the front of the Roi Baudouin Ice Shelf, in eastern DML, an 850-m-deep trough is present, which has major consequences when simulating ice sheet advance and retreat (Berger, 2017;Favier et al., 2016). This emerging picture highlights the need for more accurate bathymetry measurements in this region.
The general lack of bathymetry measurements beneath ice shelves is the result of difficulties in access: radar systems do not penetrate through the water column; gravity inversions are possible, but without control points are sensitive to assumptions about the underlying geology (Eisermann et al., 2020); drilling through the ice provides limited spatial range and is logistically challenging; and the use of autonomous underwater vehicles is rare, as they require ship support (e.g., Jenkins et al., 2010;Nicholls et al., 2006). Seismic surveys currently provide the most reliable method to map water column thickness, as well as to image the sea bed below.
Here we address this problem using a specialized vibroseis seismic source and snow streamer system (Eisen et al., 2015) to create a new map of the ice shelf cavity and sea floor bathymetry beneath the Ekström Ice Shelf, DML. Ekström Ice Shelf ( Figure 1) covers an area of approximately 6,800 km 2 with basal melt rates of up to 1.1 ma −1 (Neckel et al., 2012). It is laterally constrained by the grounded ice rises of Søråsen to the west and Halvfarryggen in the east (Figure 1c). The present ice shelf front is less than 20 km from the continental shelf break. Until now, little was known about the bottom topography of the ice-shelf cavity. A number of seismic reflection measurements by Kobarg (1988) suggest a southward deepening of the sea floor with a maximum water column thickness of 500 m. However, a map of the cavity is currently lacking. We integrate our bathymetric mapping with conductivity-temperature-depth (CTD) data acquired through hot water drilled access holes in the ice shelf and under sea ice in Atka Bay ( Figure 1c). The combined observations are used to identify the primary implications of this new bathymetry for ice-ocean interaction and ice dynamic history in the region. They show the urgent need for more measurements of sub-ice-shelf bathymetry along the coast of this sector of Antarctica and for many other ice shelves, where the bathymetry is equally poorly constrained.

Seismic Data Acquisition
Seismic data were collected between 2010 and 2018, using two different vibroseis seismic sources. The snow streamer, used for all data acquisition, was 1,500 m long, containing 60 channels, with a 25-m group spacing. Each group contained eight gimballed P-wave SM-4, 14-Hz geophones. For all data collection the vibroseis source was towed behind a snow tractor with the snow streamer towed behind that. This method of operation allowed for data acquisition rates of up to 20 km per day for ten fold data. Fold refers to the amount of times a subsurface point (referred to as a common midpoint, CMP) is sampled by different source/receiver pairs. A more detailed explanation of both these seismic sources, the snow streamer, and the operational method is given in Eisen et al. (2015).
The main grid of data, at the ice shelf front, was collected during the 2016/2017 and 2017/2018 austral summers (Figure 1c, green) as part of the Sub-EIS-Obs project (Kuhn & Gaedicke, 2015). The seismic source was the AWI IVI EnviroVibe, producing a 10-s linear sweep from 10 to 220 Hz. These data are relatively high fold (6-15), with the exception of the two lines in the north-western corner of the grid, which are single  (Arndt et al., 2013), with the 1,000 m below sea level (BSL) contour shown as a dashed line, indicating the location of the continental slope. The CTD location labels "4" to "8" refer to hot water drilled access holes "EIS-4" to "EIS-8" and "AB" to the location of a sea ice lead in Atka Bay. The location of the seismic line shown in Figure 2 is X-X'.  Figure 1c, yellow) and were acquired using the University of Bergen Failing Y-1100 vibroseis source (Kristoffersen et al., 2014), with a 10-s sweep from 10 to 100 Hz; fold varies between one and eight.
All seismic data were processed specifically for this study to ensure consistent treatment of data from different surveys (supporting information, Text S1). The resulting seismic time-stacked sections all have clear ice base and the sea floor reflections ( Figure 2). The two-way traveltimes (TWTs) of these reflections were handpicked on each section.

Depth Conversion and Gridding of Seismic Measurements
The handpicked TWTs from all seismic lines were used to create grids of the TWT to the ice base and sea floor, using a kriging algorithm. Any misties between picks, in areas where seismic lines overlap, were handled by assigning priority to the higher resolution surveys. Each grid was then depth converted using a seismic velocity of 3,601 ms −1 for the ice shelf and 1,451 ms −1 for the water column (supporting information, Text S2). The final step was to correct the depth of each grid for the ice surface elevation, using the REMA digital elevation model v1.1 (Howat et al., 2019), which was rereferenced to the GL04C geoid (Förste et al., 2008).

Uncertainties in Seismic Depths
Uncertainties in the sea floor depth, from seismic measurements, come from four main sources: (i) accuracy of the horizon picking, (ii) velocities used for depth conversion of these horizons, (iii) errors in the REMA DEM used for surface elevation corrections, and (iv) depth errors from unmigrated data. A detailed analysis of these individual error sources was made (supporting information, Text S3), resulting in cumulative error at the sea floor of ±14.8 m in the area of the main data grid and ±34.4 m in the areas of the 2014 seismic lines, which extend from the main grid toward the grounding lines ( Figure 1). The gridded bathymetry may have larger errors away from the seismic lines.
It was possible to measure the ice thickness and sea floor depth at the five hot water drilled access hole locations (4-8; Figure 1) on the ice shelf, both during drilling and with camera equipment and coring devices deployed through the holes to the sea floor. These measurements confirmed that the seismic determined ice thickness was within ±5 m of the measured thickness and the sea floor within ±10-20 m. This validates our error estimates, as the latter range also includes horizontal displacement of the sampling rope by ocean currents.

CTD Data
During the 2018/2019 austral summer, hot water drilled access holes were made at five locations on the ice shelf (4-8; Figure 1). An RBR Concerto CTD sensor was repeatedly lowered through each hole and additionally through a sea ice lead in Atka Bay (AB; Figure 1), giving a total of six CTD measurement sites. The CTD sensor recorded water mass properties at a frequency of 1 Hz (vertical resolution of 0.5-2 m), from which pressure, in situ temperature, and practical salinity data were extracted (supporting information, Text S4). The uncertainty for in situ temperature is 0.02 • C and for practical salinity 0.03. These data were used to 10.1029/2019GL086187 calculate seismic velocities for depth conversion of the sea floor seismic reflection (supporting information, Text S2) and to investigate water masses present beneath the ice shelf and sea ice.

Results and Discussion
The new sub-ice-shelf bathymetry (Figure 3a) is independent of any previously available products of ice thickness and water depth. Here, we compare the new bathymetry to the Bedmap2 (Fretwell et al., 2013) Antarctic bed topography (Figure 3b), which has been the baseline Antarctic dataset for the large majority of modeling studies. We highlight the differences to emphasize the need for dedicated measurements of sub-ice-shelf bathymetry.
Given the lack of previously available data, the Bedmap2 (Fretwell et al., 2013) sea floor bathymetry beneath Ekström Ice Shelf closely follows the ice base topography, deepening toward the grounding line. This suggesting a relatively thin, uniform water column height of the ice shelf cavity (Figure 3b). The new seismic bathymetry, in contrast, reveals a much more distinct geometry of the ice shelf cavity (Figure 3c). Similar mismatches have been documented for other ice shelves in this sector (e.g., Nøst, 2004).
Beneath the main grid of our data (Figure 3a) we find a bathymetric trough under the central part of the ice shelf, which appears to be aligned with the current ice flow direction. In this area the trough is 30 km wide and reaches depths of up to 800 m below sea level (Figures 3a and 3c, C-C'). The trough flanks have depths around 450-500 m (Figures 3a and 3c, C-C' and D-D'), shallowing to around 300 m depth at the marginal grounding line joining the ice shelf to the ice rise of Halvfarryggen (Figures 3a and 3c, B-B'). Shallowing topography is also seen toward the western grounding line at Søråsen, and it is likely this mirrors the cavity shape to the east. The sea floor directly in front of the ice shelf is 450 m deep, similar to the flanks of the trough. A basin-like depression, around 570 m deep, is seen on the eastern plateau to the south of Neumayer Station III. The profile from the front of the ice shelf edge to the current grounding line (Figures 3a and 3c, A-A') follows the axis of the trough under the main grid and the single seismic line connecting the south of the grid to the grounding line. This indicates an inland-sloping sea floor (Figures 3a and 3c, A-A'), reaching a maximum depth of 1,100 m around 10-km seaward of the current grounding line.

Ice Dynamic History
The sea-floor bathymetry under Ekström Ice Shelf allows us to interpret features associated with past ice sheet configurations in this region. The deepened trough under the center of the ice shelf (Figure 3a) is interpreted as a relict landscape, formed through erosion by former ice streams. The flanks of the trough and the sea floor at the ice-shelf front are around 300-400 m shallower than the trough. The trough does not cross-cut the outer parts of continental shelf, and it is therefore likely that the seaward rising continental shelf was a previous grounding line.
Sediment core records in front of Ekström Ice Shelf are sparse (Grobe & Mackensen, 1992;Hillenbrand et al., 2014) and do not provide conclusive evidence as to whether grounded ice covered the entire continental shelf, as far as the continental slope (approximately the 1,000-m contour in Figure 1), during the Last Glacial Maximum (LGM: 23-19 ka BP). Sediment cores from the continental shelf, in front of Ekström Ice Shelf, were not deep enough to penetrate into material from the LGM. The recovered Holocene aged sediments (11.7 ka BP to present), were likely deposited beneath an ice shelf (Grobe & Mackensen, 1992), close to the grounding line. Our finding that the trough does not cross-cut the outer parts of continental shelf suggests that grounded ice likely reached to the inner parts of the continental shelf (around the current ice-shelf front) in the past, possibly at the LGM, but the precise grounding line position remains unresolved.
The seismic line extending from the main grid to the south (Figure 3c, A-A') shows a sea floor that deepens toward the current grounding line. The general trend of a retrograde slope within the trough would likely have put this area at risk of rapid ice retreat after the LGM, until a stable grounding line position (e.g., a topographic high) was reached. There are a number of topographic highs along the ice flow direction. Particularly prominent is the topographic high at 100 km along profile A-A' (I; Figure 3c), which is around 200 m above the surrounding sea floor. We suggest that this is a former grounding line position, given its significant elevation. However, current bathymetry measurements in this region do not extend laterally to confirm the extent of this feature. There is a significant overdeepening upstream of this high (II; Figure 3c), reaching 1,100 m depth around 10-km seaward of the current grounding line. Overdeepenings are commonly formed in areas of convergent ice stream flow, where ice velocities and erosional potential are high   (Gade, 1979) when assuming zero conductive heat flux into the ice. Black curves indicate horizons of constant density, the thick near-horizontal black line indicates the melting point temperature at atmospheric pressure. (Patton et al., 2016). The location of this overdeepening is at the convergence of two tributary glaciers, with higher modern-day ice flow velocities than the surrounding ice (Neckel et al., 2012). When ice was thicker and grounded further seaward, the overdeepened area would have been at the junction between these two tributaries, eroding the deep basin. Overdeepenings typically terminate in sills (Benn & Evans, 2014), where ice flow becomes less constrained, which could explain the origin of the topographic high we observe at 100 km along profile A-A'. Two smaller topographic highs at 35 km (III; Figure 3c) and 45 km (IV; Figure 3c) along profile A-A' are around 50 m in height, each separated by deeper basin areas, indicative of ice having been grounded at these points for some time.
A deep central trough punctuated with transverse topographic highs has also been observed under the neighboring Fimbul Ice Shelf (Nøst, 2004) and along the adjacent Coats Land ice margin (Hodgson et al., 2018(Hodgson et al., , 2019. Under the front of Roi Baudouin Ice Shelf, eastern DML, an 850-m-deep trough (Berger, 2017;Favier et al., 2016) is present, hinting that this may also extend under the ice shelf in a similar way. This emerging picture suggests such deep troughs are ubiquitous in this sector of East Antarctica, indicating ice streams were a prevalent feature in the past and supporting the need for more widespread sub-ice-shelf bathymetry measurements.

Ice-Ocean Interaction
The implications for ice-ocean interactions in the Ekström region, from the newly mapped ice shelf cavity, refine our understanding of the role of DML ice shelves in the mass balance of the Antarctic ice sheets. Along this sector of the Antarctic coast, WDW masses are suppressed below the continental shelf break (Heywood et al., 2013) by the prevailing easterly winds in the "fresh shelf" regime (Thompson et al., 2018) that extends from about 20 W to 30 E. The sub-ice-shelf CTD profiles (Figure 4) confirm this, underpinning the relatively low basal melt rate estimates (Neckel et al., 2012) for the Ekström Ice Shelf. Most of the cavity is filled with relatively cold Eastern Shelf Water (ESW) that resides above the WDW along the DML coast (Nøst et al., 2011). In situ temperatures are close to the surface freezing point around −1.9 • C and practical salinity around 34.4. Only the upper tens of meters of the individual profiles show colder (Figure 4a), less saline water masses, indicating buoyant outflows of Ice Shelf Water in a surface layer near the ice shelf base that is aligned along common melt water mixing line (Figure 4b, Gade, 1979) originating from the ESW.
Although the deeper trough beneath the central part of the ice shelf does not cross-cut the continental shelf break, it provides a possible conduit for warmer inflows reaching onto the continental shelf to enter the ice-shelf cavity (Figure 3c, A-A'). Intermittent, warmer near-bottom inflows have been observed across a slightly deeper sill beneath the neighboring Fimbul Ice Shelf (Hattermann et al., 2012) and a rise of WDW along the DML coast has been suggested as a possible response to future climate change (Hattermann, 2018;Hellmer et al., 2017). In these scenarios, the newly revealed several hundred meters of water depth of the Ekström cavity would support a cavity overturning circulation, where warm near-bottom inflows can propagate undiluted toward the grounding line, rendering Ekström Ice Shelf more vulnerable to warm inflows than implied by previous bathymetric datasets. Our example is also likely to be instructive for many smaller, unmapped ice shelves that have a similar configuration and glacial history along the DML coast and elsewhere.
The presence of extensive ocean cavities beneath Ekström and potentially other DML ice shelves has implications for the susceptibility to local marine ice sheet instability (Morlighem et al., 2020;Ritz et al., 2015), as well as for far field effects through melt induced altering water mass properties of the westward boundary current. While ice shelf melt water presently contributes little to the freshening of the ESW (Zhou et al., 2014), there is growing evidence that basal melting feedbacks play an important role for the slope front overturning that controls the depth of the thermocline at the shelf break (e.g., Hattermann, 2018). This regulates the access of warm water to the vast Filchner-Ronne Ice Shelf in the southern Weddell Sea (e.g., Hellmer et al., 2017). A correct representation of warm water pathways to the ice shelves upstream of the Filchner Trough will hence be crucial to quantifying these effects and assessing the uncertainty of model projections.
Locally, our observed seafloor bathymetry suggests a separation of different circulation regimes beneath the ice shelf. While the central part of the cavity seems to be dominated by the trough system; the eastern portion of the ice shelf, adjacent to Atka Bay, comprises a relatively shallow water column (Figures 3a and  3c, C-C') that is likely subject to intense tidal mixing and may be responsible for the vigorous accretion of marine ice in the area (Arndt et al., 2020). The two successive temperature profiles from the EIS-4 site confirm the existence of a tidally mixed zone in this region. While the EIS-4b profile shows a similar vertical gradient in temperature to the profiles obtained in the deeper part of the cavity, the EIS-4i profile has a more homogeneous vertical temperature structure. The EIS-4i profile was taken about 11 hours later than EIS-4b, and according to the CATS regional tidal model (Padman et al., 2008), a relatively large amplitude tidal wave passed the EIS-4 hole location between the two measurements. Hoppmann et al. (2015) found that platelet ice crystals leave the ice-shelf cavity in intermittent pulses at this location, and similar tidal flushing events may be responsible for mixing larger volumes of potentially supercooled Ice Shelf Water in shallower depths, contributing to the platelet ice formation under the fast ice in Atka Bay that has been identified as a prominent habitat in the sea ice ecosystem (e.g., Smetacek et al., 1992). A recent study also shows that propagating tidal waves may play an important role in modulating basal melting near the ice shelf grounding lines (Sun et al., 2019), further emphasizing the need for knowledge of the local cavity shape to assess and interpret the melt rate variability of a given region.

Conclusions
We have presented new bathymetry data from under the Ekström Ice Shelf, Dronning Maud Land, Antarctica. The use of vibroseis seismic reflection surveys proves an effective method for collecting high-resolution data across large areas of the ice shelf. As a result, the Ekström Ice Shelf cavity is currently one of the best mapped in Antarctica. The discovery of a deep trough with transverse sills under Ekström Ice Shelf is the second example of such a feature under a DML ice shelf, after the neighboring Fimbul Ice Shelf, with similar features also found along the adjacent Coats Land margin (Hodgson et al., 2018(Hodgson et al., , 2019 and at the ice shelf front of Roi Baudouin (Berger, 2017;Favier et al., 2016) . This growing list of evidence suggests that the bathymetry we see at Ekström, Fimbul, and along the Coats Land coast is likely characteristic of other ice shelves in the DML and neighboring regions. While these ice shelves are small, they are numerous and very little is known about the cavity geometry, which is a fundamental gap in our ability to understand past ice dynamics and future stability of this region. The ice shelves of DML are known to play a key role in preconditioning the water mass properties of the westward flowing boundary current, which affects the much larger Filchner-Ronne Ice Shelf and thus large portions of the West Antarctica Ice Sheet. Improved knowledge of ice shelf cavities is a key required step toward better understanding and projections of the fate of marine ice sheets in a warming climate.

Author Contributions
All authors contributed to seismic target and profiling design, discussed the manuscript, and contributed comments toward the final version. ECS designed and wrote the paper, performed fieldwork, and led one field season, analyzed, and interpreted seismic data; TH analyzed CTD data and provided the oceanographic component to the paper; DF and A Lambrecht performed seismic data acquisition and analysis; CM performed seismic data acquisition and led one field season; SB acquired and analyzed CTD data and provided glaciological constrains for ice shelf processes; RD, TAE, and CM contributed ice flow modeling results and glaciological/geological constraints for data interpretation; CH was in charge of the seismic equipment and processed seismic data; GK, CG, A Läufer, and RT provided geological constraints for data interpretation; FW prepared and coordinated the hot water drilling system; RG led one field season and performed cavity sampling; OE coordinated and implemented the seismic field work, led three field seasons, performed seismic and CTD data acquisition. GK, CG, A Läufer, RT, FW, and OE are project Co-PIs, designed financial support and project implementation; CG and OE administratively coordinated the project. The transdisciplinary science component is based on Kuhn and Gaedicke (2015).