The bathymetric data from Expedition PS128 were acquired with the multibeam echosounder (MBES) Teledyne Reson Hydrosweep DS3 with a frequency of ∼14 kHz in chirped pulses, arranged in a Mills Cross configuration of 3×3 m. The bathymetric data from NBP0101 were acquired with the SEABEAM system (Leventer et al., 2006). Both the PS128 and NBP0101 data were processed with the MB-system and spatially gridded with 25×25 m and 50×50 m, respectively, for further analysis with QGIS (QGIS.org, 2026).

Sub-bottom profiles were acquired with the hull-mounted Deep-Sea Sediment Echo Sounder System PARASOUND P70 (Teledyne Reson, Bremen, Germany). For the expedition setting, two high-frequency acoustic signals of 18 and 22 kHz were transmitted, generating a low-frequency signal of 4 kHz in the water column. In our study, this low-frequency signal penetrated the sediment column to a depth of about 50 m and with a vertical resolution of about 30 cm to image the upper tens of meters of the subsurface on formerly glaciated shelves. This largely depends on the sediment's degree of stiffness (e.g., Klages et al., 2017; Arndt et al., 2017). The combined acoustic signal has a beam width of ∼4°. The resulting beam's footprint and thus the horizontal resolution on the seabed is about 7 % of the water depth.

Before sampling the cores, all sections were X-rayed either with computed tomography (CT) or by scanning 1 cm-thick sediment slabs at the AWI in Bremerhaven. The standard sampling interval for shear strength measurements and further sample analyses was usually set to 10 cm, but guided by lithological or structural changes. Smear slides were taken from each identified lithological unit to support visual core descriptions and to further define microfossil content and composition. Magnetic susceptibility has already been determined onboard R/V Polarstern with a GEOTEK multi-sensor core logger. Syringe samples (10 cubic centimeters; cc) were removed, wet-weighed, freeze-dried, ground, and dry-weighed to determine water content and geochemical parameters such as total organic carbon (TOC) and nitrogen (N) content. For TOC, inorganic carbon – including carbonates – was removed from 0.1 g of sediment with hydrochloric acid (20 % HCl) before the samples were measured with an ELTRA CS800 carbon-sulfur determinator (error ±0.02 %). Total C and N contents of the bulk sediment were obtained with a Vario cube organic elemental analyzer by high-temperature combustion (error C, N ±0.02 %). The ratio of carbon to nitrogen (C:N) was calculated from TOC / N_tot(%).

Spatula samples (∼30 cc) from identical depths were sieved for grain size classification after freeze-drying with >2 mm and >63 µm-meshed sieves, respectively. The weight percentages were calculated from the total dry weight of the sample and the resulting weight of the dried grain size fractions, i.e., gravel (>2 mm), sand (>63µm), and mud (<63µm), with water content calculated by subtracting the total dry weight from the wet weight. Sufficient amounts of calcareous microfossils were extracted for AMS ¹⁴C dating from outer (PS128_44-1, 45-1, PS128_46-1, and PS128_47-1) and mid-shelf core PS128_39-1 (Tables 3 and S1 in the Supplement). Before microfossil extraction, each depth was closely investigated on X-radiographs to exclude iceberg and/or bioturbation, potentially leading to unreliable measurements. Extracted foraminifer assemblages mainly included planktic species in core PS128_45-1 and benthic species in cores PS128_39-1, PS128_44-1, PS128_46-1, and PS128_47-1. Even though calcareous microfossils are poorly preserved in continental shelf sediments (Hauck et al., 2012) as they are often bathed in corrosive Southern Ocean deep waters that currently spill onto shelves in many regions around the continent (e.g., Lanham et al., 2025), they provide the most reliable ¹⁴C age control for Antarctic shelf sediments deposited during the last ∼40 ka (e.g., Klages et al., 2017; Mawbey et al., 2026). For the present work, AMS ¹⁴C dating was carried out with the MIni CArbon Dating System (MICADAS) on planktic and benthic calcareous microfossils at the Alfred Wegener Institute, Bremerhaven, Germany, following established workflows (Mollenhauer et al., 2021).

For core PS128_39-1, we additionally applied Ramped-Pyrolysis-Oxidation (RPO) ¹⁴C dating to improve the age constraint for the alternating diatomaceous oozes in the upper part of the core. For that, we used a modified SoliTOC Cube Carbon Analyzer (Elementar Analysensysteme GmbH, Germany), also located at the MICADAS facility. Based on the shape of the thermograms (evolved CO₂ over temperature), the temperature limit for an appropriate fraction was determined such that the upper temperature limit was as low as possible while including a total amount of 80–100 µgC (Mollenhauer et al., 2021; Kattein et al., 2026; cf. Table S1). To ensure comparability to other Antarctic shelf sediment records, we corrected all ¹⁴C ages by subtracting the marine reservoir effect for Antarctic shelf sediments (1300±100 ¹⁴C years; Berkman and Forman, 1996; Ingolfsson et al., 1998; Anderson et al., 2002). We then calibrated the ages with the CALIB 8.2 software (Stuiver and Reimer, 1993), including the Marine20 calibration curve (Heaton et al., 2020), also considering their approach for Antarctic Holocene samples younger and older than 11.5 cal. ka BP (Heaton et al., 2023). To ensure comparability to previously dated sediment cores in the region, we recalibrated published ¹⁴C ages from JPC40 and JPC43B (Leventer et al., 2006) and PS69/849-2 (Borchers et al., 2016).

Identifying submarine glacial landforms that can be clearly linked to the former presence of ice gives us information about the past ice sheet's grounding line position, dynamics, and retreat styles, which is possible because many of these landforms are well preserved on the seafloor (Wellner et al., 2006; Dowdeswell et al., 2016).

Submarine glacial landform assemblages in the study area can be grouped into three classes: (1) Elongated, streamlined, and/or (sub)parallel, (2) wedge-like, and (3) randomly oriented curvilinear features. In the Nielsen Basin, we defined a distinct bedrock sill (elevated by ∼150 m relative to the surrounding topography) that separates the deeper inner-shelf part of the trough from the mid-shelf. Adjacent to the sill, deeper basins characterized by clusters of (sub)parallel and elongated ridges with elongation ratios >10:1 are present. Towards the sill's top, those ridges transition into teardrop-shaped features before becoming more elongated again in seaward direction (elongation ratios >20:1; Fig. 3b). Towards the top of the sill, sediment cover thins out, whereas it significantly thickens towards the center of adjacent basins (Fig. 3; profile A–A'). In the mid-shelf Nielsen Basin, several wedge-shaped (steep distal and gentle proximal slope) ∼1 km long and ∼8 m high bathymetric features are located perpendicular to the trough axis. Their internal composition appears largely transparent to chaotic on sub-bottom profiles (Fig. 3; profile X–X'). At the mid–outer shelf transition, the seafloor shallows significantly from ∼500 to ∼300 m b.s.l. (Fig. 2a; vertical cross-section c). Here, the sets of elongated, streamlined features become increasingly overprinted by randomly oriented curvilinear structures, which are characterized by a general NNW orientation.

On the outer shelf, randomly oriented curvilinear features comprise the entire area with highly variable orientations and depths. The presumed LGM GZW-II (cf. Mackintosh et al., 2011; Fig. 2a and cross-section panel c) lies in water depths of ∼ 390 mbsl with a crest height of ∼10 m and an along-flow length of ∼6 km. Seaward from GZW-II, several multidirectional curvilinear features with ∼5 to 10 m depths and with up to 200 m widths are preserved.

The Iceberg Alley trough is characterized by elongated, streamlined, (sub)parallel features on its mid-shelf and changes towards more randomly oriented curvilinear features on the outer shelf (cf., Leventer et al., 2006; Fig. 3c), including a wedge-like feature previously interpreted as a LGM GZW (Mackintosh et al., 2011; Fig. 3c; i.e., GZW-I). The outer shelf assemblages resemble those identified on the outer Nielsen Basin, including GZW-I (Fig. 2; panel a and cross-section panel b; cf. Mackintosh et al., 2011) and NNE-oriented curvilinear and linear features (Fig. 3). It has a height of ∼8 m and an along-flow length of 2 km and is superimposed on elongated, streamlined, parallel features (Fig. 3c). The water depths, crest height, and location on the outer shelf of GZW-I are comparable to GZW-II.

Based on numerous previous Antarctic shelf studies (e.g., Klages et al., 2014, 2015; Livingstone et al., 2016a; Halberstadt et al., 2016), we interpret the elongated, streamlined, and parallel bedform assemblages identified on the Nielsen Basin mid-shelf as mega-scale glacial lineations (MSGL). These features usually form under wet-based, fast-flowing ice streams on soft, deformable substrates such as glacial tills as a result of till deformation and basal sliding, often underlain by a consolidated, stiff till (e.g., Ó Cofaigh et al., 2007; Reinardy et al., 2011). Differently oriented sets of N-NNW-oriented MSGLs within the curved trough (Fig. 3a), disrupted by rugged bedrock topography, may imply their multi-temporal formation (e.g., Graham et al., 2009). They show an increase in elongation on the lee side of the bathymetric sill (Fig. 3a). Usually, this is associated with flow acceleration at substrate boundaries and deforming bed conditions (Ó Cofaigh et al., 2007; Graham et al., 2009, Livingstone et al., 2016b; Jamieson et al., 2016). In the absence of sufficiently penetrating sub-bottom profiles, we generalize the teardrop-shaped features on top of the Nielsen Basin bedrock sill to “drumlinoid features” (Fig. 3a), usually associated with ice-flow velocity changes (e.g., Graham et al., 2009). Varying sizes and shapes of MSGLs and drumlinoids likely result from a complex interplay between differences in local substrates, changing ice flow conditions, or small-scale variations in basal conditions (Graham et al., 2009; Greenwood et al., 2021).

The wedge-like features on the Nielsen Basin's mid to outer shelf transition are identified as grounding-zone wedges (GZWs; cf. Figs. 2a, 3c and profile X–X'), closely resembling numerous features previously reported from Antarctic continental shelves (e.g., Jakobsson et al., 2012; Klages et al., 2014, 2015, 2017; Batchelor and Dowdeswell, 2015). GZW dimensions are controlled by sediment delivery, duration of stillstand, cavity shape, and the width of the streaming ice margin (e.g., Dowdeswell and Fugelli, 2012; Simkins et al., 2018). In our study area, they co-occur with linear to curvilinear iceberg scours formed by the seafloor-scouring keels of grounded icebergs (Fig. 3b).

In Iceberg Alley, the streamlined, subparallel features from the inner to mid shelf were equally classified as “drumlinoid features”. They likely formed similarly to the ones in Nielsen Basin, i.e., from ice flow velocity changes due to varying bedrock composition underneath the streaming ice. On the outer shelf, we identified the clustered parallel ridges as MSGL (Fig. 3c) and the deeper ridges with low parallelism as iceberg scours. Some of the scours are cross-cutting, indicating the presence of large, multidirectional icebergs that were calving close to the ice sheet margin at different timescales. GZW-I on the outer shelf is superimposed on the MSGLs and, therefore, suggests an earlier formation of these MSGLs (Fig. 3c; dotted blue lines).

For this study, eight gravity cores were analyzed from the central mid-outer Nielsen Basin and outer Iceberg Alley (Fig. 2) to explore past grounding line extent and subsequent retreat dynamics since the LGM (Figs. 4 and 5; Tables 2 and 3).

The outer shelf gravity cores PS128_45-1 and 44-1 were retrieved near the continental shelf break of Nielsen Basin, and just seaward of GZW-II (cf. Mackintosh et al., 2011; Fig. 2). Mid-shelf cores PS128_42-1 and 41-1 were recovered from the top sets of two GZWs, PS128_39-1 from a sedimentary basin, and PS128_38-2 from a bedrock sill. Gravity cores PS128_46-1 and PS128_47-1 were recovered from outer Iceberg Alley (Fig. 2), with PS128_46-1 from atop GZW-I (cf. Mackintosh et al., 2011; Fig. 2) and PS128_47-1 located just seaward from GZW-I (Fig. 3c).

The geological record of grounding line retreat in gravity cores from the Antarctic continental shelf is usually evident from three to four sedimentary facies reflecting the changing conditions at a particular site from subglacial over grounding line-proximal and sub-ice shelf to seasonal open-marine deposits (e.g., Hillenbrand et al., 2014; Klages et al., 2013). Ideally, such facies successions can be put into chronological context by dating in-situ calcareous microfossils or, if the sedimentary environment allows (Hillenbrand et al., 2010), bulk organic carbon. We assigned depositional settings for the lithological units based on our multi-proxy approach, summarized in Table 2.

Previous studies from around the Antarctic continental shelf suggest a grounded ice sheet at or close to the continental shelf break at the LGM (e.g., Wellner et. al., 2006; Anderson et al., 2014; Hillenbrand et al.,2014; Hodgson et al., 2014; Larter et al., 2014; Mackintosh et al., 2014; Ó Cofaigh et al., 2014; Klages et al., 2014, 2015, 2017), with only a few studies around the East Antarctic margin focusing on grounding line retreat since the Last Glacial Maximum. For the Mac. Robertson continental shelf, previous studies revealed characteristic glacial sedimentary deposits in the outer Nielsen Basin, as well as prominent deglacial deposits of alternating laminations of iron-rich sediments and diatomaceous oozes in cores from both inner shelves (Harris and O'Brien, 1998; Leventer et al., 2006). Previously, only bulk AIO radiocarbon ages were available, and carbonate ages from foraminifera are a big improvement on those age constraints, because the carbonate retrieved from calcareous microfossils formed in situ. Our multi-proxy dataset, including foraminifera-derived ¹⁴C age constraints, indicates that grounding lines of fast-flowing ice streams within Nielsen Basin and Iceberg Alley reached the shelf break prior to ∼12.7 cal. ka BP (Table 3; Figs. 6a and 3). The multiple generations of keel scours documented on the outer shelf indicate several calving events of different stages of ice retreat in varying water depths and different constraints on drift directions. The scours are interpreted to be of Holocene age, due to their westward orientation, implicating a drift caused by the current-day configuration of outer shelf currents without lateral constraints from ice bodies (Harris and O'Brien, 1998; Leventer et al., 2006). Our foraminifera-derived ¹⁴C age in cores PS128_45-1, 46-1, and 47 from the GL-proximal facies indicates a minimum retreat age of the GL around ∼12.2 to 12.7 cal. ka BP from the outer shelf towards the mid-shelf in both basins. This implies that GZW-I and GZW-II (Fig. 2) did not form during the LGM and are post-LGM features. We assume a formation of these GZWs shortly after ∼12.7 cal. ka BP. The overprinted MSGLs (Fig. 3c, blue dotted lines) indicate fast-streaming wet-based ice before the formation of GZW-I. Thus, we propose a nearly synchronous initial retreat of the GL from the outer shelf in both glacial troughs (Fig. 6a). A lateral narrowing of the trough might have slowed down ice retreat due to lateral pinning on the shallower banks on each side of the trough and after the lateral pinning of the ice sheet, conditions favored GZW formation (Fig. 6b), because lateral and vertical bathymetric pinning points are important factors slowing down ice retreat (Batchelor and Dowdeswell, 2015) and is in accordance with the finding that commonly during the last deglaciation, grounded ice remained on shallow marine areas and adjacent islands, while faster retreat occurred in deeper marine troughs (Dyke et al., 2003). Both troughs significantly deepen into retrograde slopes at the transition from the outer to the mid-shelf. Here, MSGLs are abundant, indicating a fast-flowing ice stream at the time of formation. We inferred an acceleration of GL from the trough topography and the narrow time difference between the ¹⁴C ages retrieved from the transitional facies in core PS128_45-1 and PS128_39-1. In Nielsen Basin, we identified several stabilization events on the mid-shelf during GL retreat. The small size and close spacing of the identified GZWs (Fig. 3; Profile X–X') point toward only a short ice sheet stabilization at the transition from the outer shelf to the mid-shelf, and the mid-shelf GZWs are likely deposited after ice sheet pinning to outcropping bedrock (Fig. 6c).

Earlier work by Leventer et al. (2006) reveals an AIO age of ∼12.4 cal. ka BP for the onset of the deglaciation at the outer-mid-shelf boundary in Iceberg Alley, and Borchers et al. (2016) propose a similar AIO deglaciation age (∼12.5 cal. ka BP) for the mid-shelf region in Burton Basin. These ages are almost identical to ours (∼12.7 cal. ka BP), and this strengthens the assumption of a nearly synchronous grounding line retreat across the Mac. Robertson continental shelf. Yet, comparisons between AIO and foraminifera-derived ¹⁴C ages need to be treated tentatively, because AIO dates are often affected by contamination from reworked fossil carbonate, resulting in older AMS ages with a drastic down-core increase in ¹⁴C ages (Andrews et al., 1999). We attribute the older ¹⁴C ages from core JPC40 to entrained carbon in the bulk samples, leading to a modified deglaciation age. However, we cannot completely rule out the possibility that the outer shelf deglaciation started earlier, between 13 and 14 ka BP. Our additional RPO ages in core PS128_39-1 (blue ages in Fig. 4) help to better constrain the onset of deglaciation for the Nielsen Basin's mid-shelf. Here, the reworking of our foraminifera-derived carbonate material proximal to the retreating GL complicates the constraint on the exact timing. The material in the transitional/GL-proximal facies in PS128_39-1 is composed of glacially-transported sediments with unsorted grain sizes, transitioning into a slightly stratified sediment with big dropstones and a prominent sand layer. From the depositional environment and the regional topography, we inferred a varying GL during this time, depositing subglacial sediments downhill in front of the sill. Bioturbation marks at the depths of ∼350 to 330 cm b.s.f. can be explained by increased biological activity after the GL retreated farther from the sill. The RPO ages above 330 cm b.s.f. of 11.0 cal. ka BP indicate the onset of the Holocene. This strengthens the foraminifera-derived ages of 10.8 cal. ka BP retrieved at this interval. The ¹⁴C ages retrieved from some sparse specimens (Globocassidulina sp.) in the GL-proximal of PS128_47-1 indicate an age of ∼28.9 cal. ka BP (Iceberg Alley) and the subglacial till in PS128_39-1 of ∼34.6 cal. ka BP (Nielsen Basin). These specimens are likely reworked but still radiocarbon active. We cautiously propose this as a maximum glaciation age from older entrained microfossils in the subglacial till.

Reliable deglaciation ages are sparse for East Antarctic continental shelves, and because GL retreat around Antarctica results from several driving forces, such as oceanic conditions, bathymetry, prior ice extent, and ice sheet thickness, as well as bottom substrate composition, the overall ice sheet retreat varies widely between different regions. In comparison to the deglaciation history of Prydz Bay, where the continental shelf break of the Prydz Channel in front of the Amery Ice Shelf was already ice-free at the LGM (Domack et al., 1998; Hemer and Harris, 2003; Guitard et al., 2016), the Mac. Robertson continental shelf appears to have remained glaciated for a substantially longer period. Regional differences in bathymetric configuration likely contributed to these contrasting retreat histories around Antarctica. For example, GL retreat is facilitated and accelerated on retrograde sloping beds and topography below sea level. Such retrograde slopes are less pronounced across much of the East Antarctic continental shelf, except within cross-shelf glacial troughs (Morlighem et al., 2020). In addition to bathymetry, the intensity of mCDW intrusions onto the continental shelf controls the timing and intensity of the retreat. This is influenced by the stratification and composition of the water column, as well as by prevailing currents at the continental shelf break, and facilitated by oceanic gateways (Nicola et al., 2025). Isopycnals, stratified density layers of the ocean, at the shelf break influence which water masses can access or ventilate the shelf. Thompson et al. (2018) classified the East Antarctic area around the Mac. Robertson continental shelf as a “dense shelf” regime. This oceanic regime is characterized by persistent strong easterly winds and a westward-flowing ASC. Here, strong fronts separate cold and fresh shelf water from warm and salty CDW offshore, leading to a weak cross-slope exchange of water masses. In this regime, it is much harder for mCDW to intrude onto the shelf compared to a “warm shelf” regime, which prevails around most of, e.g., West Antarctic continental shelves (Thompson et al., 2018). This may likewise have limited the intrusion of mCDW onto the shelf in East Antarctica in the recent past. Our deglaciation age of the Mac. Robertson continental shelf of ∼12.7 cal. ka BP coincides with the end of the Antarctic Cold Reversal (ACR), a temporary atmospheric and oceanic cooling in the Southwest Pacific due to deglacial changes in the Atlantic Meridional overturning circulation, while a warming occurred in the Northern Hemisphere (Eaves et al., 2024), and it lasted from ∼14.5 to ∼13.0 ka BP (Pedro et al., 2016). Our deglaciation ages reveal an onset at the end or shortly after this short-term cooling interval. The initial retreat age of ∼12.7 cal. ka BP is too late (after 14.2 ka BP) to be assigned to the rising sea level from Meltwater Pulse 1a (MWP-1a). Yet, the sea level rise during Meltwater Pulse 1b (MWP-1b), triggered by elevated meltwater discharge and the disintegration of large parts of the ice sheet between 11.5 and 11.2 ka BP (Peltier and Fairbanks, 2006), falls into the same time interval as our deglaciation ages. This additional sea level rise may have lifted the depth interval in which the CDW resides. The complex topography of the Mac. Robertson continental shelf, with Nielsen Basin and Iceberg Alley as deeper cross-shelf troughs, might have provided geostrophic pathways for CDW to enter the continental shelf (cf. Fig. 6b). Additionally, southward migration of westerly winds and stronger coastal easterly winds might have enhanced upwelling (Fripiat et al., 2026) and the westward-flowing coastal current (Kusahara et al., 2017). All of the above may have delayed mCDW intrusion onto the continental shelf, and in combination with the regional topography, led to a more resilient ice sheet, delaying GL retreat on the Mac. Robertson continental shelf. Based on our multi-proxy analyses, we therefore propose a fully glaciated continental shelf during the LGM, which would imply absent DSW production on the shelf. This implies a different formation area of AABW, e.g., polynyas beyond the continental shelf break, farther offshore.

To get a better understanding of the forcing mechanisms driving ice sheet retreat, future studies might investigate grounding line retreat and advance of outer shelf deglaciation from adjacent areas along the East Antarctic Indian and Atlantic margin to connect grounding line retreat patterns. Volume estimations of past ice sheets on the East Antarctic shelves by combining terrestrial cosmogenic exposure dates with marine radiocarbon measurements might complete the overall picture of past ice sheet dynamics.