Reconstruction of changes in the Weddell Sea sector of the Antarctic Ice Sheet since the Last Glacial Maximum q

changes in ice-sheet extent and ice-sheet thickness since the Last Glacial Maximum (LGM at ca 23 e 19 calibrated kiloyears before present, cal ka BP). Terrestrial geomorphological records and exposure ages obtained from rocks in the hinterland of the Weddell Sea, ice-sheet thickness constraints from ice cores and some radiocarbon dates on offshore sediments were interpreted to indicate no signi ﬁ cant ice thickening and locally restricted grounding-line advance at the LGM. Other marine geological and geophysical studies concluded that subglacial bed-forms mapped on the Weddell Sea continental shelf, subglacial deposits and sediments over-compacted by overriding ice recovered in cores, and the few available radiocarbon ages from marine sediments are consistent with major ice-sheet advance at the LGM. Re ﬂ ecting the geological interpretations, different ice-sheet models have reconstructed con ﬂ icting LGM ice-sheet con ﬁ gurations for the Weddell Sea sector. Consequently, the estimated contributions of ice-sheet build-up in the Weddell Sea sector to the LGM sea-level low-stand of w 130 m vary considerably. In this paper, we summarise and review the geological records of past ice-sheet margins and past ice-sheet elevations in the Weddell Sea sector. We compile marine and terrestrial chronological data constraining former ice-sheet size, thereby highlighting different levels of certainty, and present two alternative scenarios of the LGM ice-sheet con ﬁ guration, including time-slice reconstructions for post-LGM grounding-line retreat. Moreover, we discuss consistencies and possible reasons for in-consistencies between the various reconstructions and propose objectives for future research. The aim of our study is to provide two alternative interpretations of glacial e geological datasets on Antarctic Ice-Sheet History for the Weddell Sea sector, which can be utilised to test and improve numerical ice-sheet models.


Introduction
The Weddell Sea region in the Atlantic sector of Antarctica (Fig. 1) plays a key role for the global thermohaline circulation by ventilating the abyssal World Ocean in the Southern Hemisphere (Rahmstorf, 2002).Interaction between sea ice, ice shelves and seawater on the continental shelf of the Weddell Sea Embayment (WSE) produces dense cool precursor water masses for Antarctic Bottom Water (AABW) which fills the deep Southern Ocean and spreads equatorwards into the deep-sea basins of the Atlantic, Indian and Pacific oceans: in the Atlantic sector AABW sensu lato reaches as far as w5 S latitude (e.g.Orsi et al., 1999;Nicholls et al., 2009).At present, about 40e70% of AABW is formed in the Weddell Sea, which therefore represents an important 'AABW factory' (Naveira Garabato et al., 2002;Fukamachi et al., 2010;Meredith, 2013).Glaciers, ice streams and ice shelves flowing into the WSE drain more than 22% of the combined area of the West Antarctic Ice Sheet (WAIS), the East Antarctic Ice Sheet (EAIS) and the Antarctic Peninsula Ice Sheet (APIS) (e.g.Joughin et al., 2006).Thus, as in other sectors of Antarctica, dynamical changes in the ice drainage basins surrounding the WSE have the potential to make major contributions to future sea-level rise (IPCC, 2007).The southern part of the embayment is covered by the FilchnereRonne Ice Shelf, one of the two major ice shelves in Antarctica, which has been identified as potentially critical to future WAIS stability (Hellmer et al., 2012).
Recently published data on subglacial topography have revealed that in the hinterland of the WSE (i) the WAIS is grounded at about 1000e1200 m below sea level on a bed with locally reverse slopes, (ii) the WAIS has a thickness close to floatation, and (iii) a large subglacial basin is located immediately upstream of the grounding line (Ross et al., 2012).Such a configuration is thought to make the ice sheet prone to rapid grounding-line retreat and ice-sheet drawdown (e.g.Weertman, 1974;Schoof, 2007;Vaughan and Arthern, 2007;Katz and Worster, 2010;Joughin and Alley, 2011), which could be triggered by grounding-line destabilisation in response to increased oceanic melting during the latter half of the 20th century (Hellmer et al., 2012).The presence of a smooth, flat bed upstream of the grounding line has been cited as evidence of previous deglaciation (Ross et al., 2012).Whilst much recent work has focussed on the Amundsen Sea sector of the WAIS, the recent findings have drawn attention to the Weddell Sea sector as another potentially important unstable part of the Antarctic Ice Sheets.
Furthermore, East Antarctica, including the eastern WSE, has been identified as a key region for better understanding glacial-isostatic adjustment (GIA) following the LGM (King et al., 2012;Shepherd et al., 2012).Estimates of mass balance based on satellite gravimetry (and to a lesser extent satellite altimetry) require a correction for crustal and mantle movements following ice (un-) loading; the uncertainty in such mass balance estimates is now dominated by the relatively poor knowledge of East Antarctic GIA (King et al., 2012).Overview map over the Weddell Sea sector with shelf bathymetry and ice-sheet surface elevation (in metres above sea level) according to Bedmap2 (Fretwell et al., 2013) and the main physiographic and glaciological features.Ice shelves are displayed in light blue shading.Inset map shows the Weddell Sea sector outlined by the red line within the context of Antarctica, with ice shelves being displayed in grey shading (APIS: Antarctic Peninsula Ice Sheet, EAIS: East Antarctic Ice Sheet, WAIS: West Antarctic Ice Sheet).
Reconstructions of the dynamical changes affecting the Weddell Sea sector during the last glacial cycle may give important clues about the future fate of its drainage basins.Such palaeo-studies have the potential to answer three fundamental questions hampering our understanding of Antarctica's glacial history: 1) Did the grounding line in the WSE advance to the shelf break during the LGM at w23,000 to 19,000 cal yrs BP (e.g.Gersonde et al., 2005) and thereby shut down the modern type of AABW production in this sector?2) How much did ice-sheet build-up in this sector contribute to the LGM sea-level low-stand of w130 m below present, and how much did post-LGM ice-sheet draw-down contribute to global meltwater pulses at 19.1 cal ka BP (e.g.Clark et al., 2004) and 14.6 cal ka BP (e.g.Clark et al., 2002)? 3) What was the icesheet history in the WSE and especially in its eastern part that contributed to modern day glacial-isostatic adjustment?Unfortunately, the available geological data constraining the LGM and post-LGM history of the Weddell Sea sector are so sparse that it can arguably be considered as one of the least well-studied sectors of Antarctica (e.g.Sugden et al., 2006;Wright et al., 2008).The main reasons for this lack of data are (i) the logistically very challenging access to the remote outcrops of rocks and till in the WSE hinterland, which are far away from any research station, and (ii) the nearly perennial sea-ice coverage, which has significantly restricted the access of research vessels to the southern WSE shelf, especially since the calving of huge icebergs from the Filchner Ice Shelf in 1986 (Grosfeld et al., 2001), with one of these icebergs remaining grounded on the shelf even today.Thus, at the time of the last major review of Weddell Sea glaciation (Bentley and Anderson, 1998) there was only fragmentary marine and terrestrial geological evidence to draw upon, much of it undated.As a consequence of the scarcity of data, LGM ice-sheet configurations reconstructed from numerical models show major discrepancies in the WSE, with some models indicating a thick ice sheet covering the entire continental shelf (e.g.Huybrechts, 2002;Bassett et al., 2007;Pollard and DeConto, 2009;Golledge et al., 2012) and others suggesting a thin ice-sheet extending across only shallower parts of the shelf (Bentley et al., 2010;Le Brocq et al., 2011;Whitehouse et al., 2012).Consequently, the estimated sea-level equivalent volume of LGM ice-sheet build-up in the Weddell Sea sector varies between 1.4 to 3 m and 13.1e14.1 m (Bassett et al., 2007;Le Brocq et al., 2011).
Despite these challenges, significant progress has been made over the last decade (and especially during the last few years) in mapping terrestrial palaeo-ice sheet surfaces and collecting rock samples for exposure age dating by analysing cosmogenic nuclides (e.g.Fogwill et al., 2004;Bentley et al., 2010;Hein et al., 2011;Hodgson et al., 2012) and in mapping glacial bedforms on the continental shelf for reconstructing past ice-sheet extent (Larter et al., 2012;Stolldorf et al., 2012).Furthermore, compilations of older datasets together with new results from sedimentological and chronological analyses on marine sediment cores recovered in the late 1960s, early 1970s and 1980s have recently been published (Hillenbrand et al., 2012;Stolldorf et al., 2012).Additional important information about the LGM ice-sheet configuration was obtained from the Berkner Island ice core drilled from 2002 to 2005 (Mulvaney et al., 2007).
All these recent studies have substantially increased the available palaeo-dataset and stimulated this paper.The main aim of our reconstruction is to provide a timely summary of current knowledge about the LGMeHolocene glacial history of the Weddell Sea sector.Together with the reconstructions of the other Antarctic sectors synthesised in this special issue by the community of palaeo-researchers, our study will provide comprehensive and integrated glacial-geological datasets on Antarctic ice-sheet history.The aim is that the datasets can be used to test and refine numerical ice-sheet models and to improve their reliability in predicting future sea-level rise from ice-sheet melting in response to global warming.
In the WSE there is still an apparent discrepancy between different lines of evidence for the extent of the ice sheet at the LGM (e.g.Bentley et al., 2010;Hillenbrand et al., 2012).The discrepancy has not yet been resolved and so this paper presents two alternative reconstructions for the LGM ice-sheet configuration in the Weddell Sea sector.We go on to discuss how these two reconstructions might (at least partly) be reconciled, and suggest priorities for future field, analytical and modelling work.

Study area
The Weddell Sea sector as defined for this reconstruction extends from w60 W to 0 W and from the South Pole to the continental shelf edge offshore from the large Ronne and Filchner ice shelves and the relatively small Brunt, Stancombe-Wills, Riiser-Larsen, Quar, Ekstrøm, Jelbart and Fimbul ice shelves, respectively (Fig. 1).The Ronne and Filchner ice shelves are separated by Berkner Island and fed by ice streams draining the APIS and the WAIS into the Ronne Ice Shelf (from west to east: Evans Ice Stream, Carlson Inlet, Rutford, Institute, Möller and Foundation ice streams) and draining the EAIS into the Filchner Ice Shelf (Support Force, Recovery and Slessor glaciers, Bailey Ice Stream) (Fig. 1; Swithinbank et al., 1988;Vaughan et al., 1995;Joughin et al., 2006).Mountain outcrops extend all along the eastern Palmer Land coast (Antarctic Peninsula), but around the rest of the WSE are restricted to high elevation regions in the Ellsworth Mountains (SW-hinterland of the Ronne Ice Shelf), the Pensacola Mountains (S-hinterland of the Filchner Ice Shelf), the Shackleton Range and Theron Mountains in Coats Land (east of the Filchner Ice Shelf) and Maudheimvidda in western Dronning Maud Land (Fig. 1).
North of the Ronne and Filchner ice shelves the continental shelf is w450 km wide and on average w400e500 m deep (Schenke et al., 1998).The shallowest water depth ( 250 m) is recorded in the vicinity of Berkner Island (Haase, 1986), and the deepest part of the shelf edge lies at w600e630 m water depth between ca 32 W and 34 W (Gales et al., 2012).In the region from w25 W to 0 W the distance between ice-shelf front and shelf break varies between 0 km and 80 km, with the water depths predominantly ranging from 300 to 400 m.Filchner Trough (also called Crary Trough, with its subglacial landward continuation usually referred to as Thiel Trough), Hughes Trough and Ronne Trough are bathymetric depressions that extend across the continental shelf offshore from the Filchner and Ronne ice shelves (Fig. 1; Schenke et al., 1998;Stolldorf et al., 2012).All three troughs have pronounced landward dipping bathymetric profiles, which are typical for cross-shelf troughs eroded by Antarctic palaeo-ice streams, with the over-deepening of the inner shelf mainly resulting from subglacial erosion during repeated ice sheet advances over successive glacial cycles (e.g.Anderson, 1999;Livingstone et al., 2012).Filchner Trough is located offshore from the Filchner Ice Shelf, up to w1200 m deep near the ice front (Schenke et al.,1998;Larter et al., 2012) and associated with a troughmouth fan (Crary Fan) on the adjacent continental slope (e.g.Kuvaas and Kristoffersen, 1991).Hughes Trough extends north of the central Ronne Ice Shelf and has a more subtle bathymetric expression with its floor lying at water depths shallower than 500 m (Haase, 1986;Stolldorf et al., 2012).Ronne Trough, which is located offshore from the westernmost Ronne Ice Shelf, is up to w650 m deep (Fig. 1; Haase, 1986;Mackensen, 2001;Nicholls et al., 2003Nicholls et al., , 2009;;Hillenbrand et al., 2012).Data on subglacial topography indicate that all three palaeo-ice stream troughs are the submarine northward expressions of subglacial troughs which deepen further inshore beneath the WAIS and EAIS, respectively (see Fig. 11;Vaughan et al., 1995;Nicholls et al., 2009;Ross et al., 2012;Fretwell et al., 2013).

Marine studies
Ice-sheet extent on the Antarctic continental shelf is usually reconstructed from subglacial bedforms mapped by multi-beam swath bathymetry or sidescan sonar imaging, glacial erosional unconformities observed in (shallow) seismic or acoustic subbottom profiles, and occurrence of subglacial diamictons (i.e.tills) recovered in marine sediment cores (e.g.Domack et al., 1999;Shipp et al., 1999;Pudsey et al., 2001;Anderson et al., 2001Anderson et al., , 2002;;Heroy and Anderson, 2005;Ó Cofaigh et al., 2005a,b;Wellner et al., 2006;Graham et al., 2009;Hillenbrand et al., 2010;Mackintosh et al., 2011;Smith et al., 2011;Jakobsson et al., 2012;Kirshner et al., 2012;Livingstone et al., 2012).In the Weddell Sea sector, several seismic, 3.5 kHz, TOPAS, PARASOUND and sparker surveys were conducted but only a few narrow strips of the shelf were mapped with high-resolution bathymetry (Fig. 2).While the distribution and geometry of subglacial bedforms, such as moraines, glacial lineations and drumlins, give unequivocal evidence for former icesheet grounding and ice-flow directions on the shelf, their preservation allows only crude age estimations, unless chronological information from sediment cores is available.Likewise, any interpretations of prominent (sub-)seafloor reflectors visible in seismic profiles as glacial erosional unconformities or seabed outcrops of subglacial till still require confirmation by sediment coring, and such reflectors alone do not provide chronological information about past grounding events.
Marine sediment cores have been recovered mainly from the southern and eastern parts of the Weddell Sea sector, while only sparse sedimentological information from a few short cores is available for the rest of the study area (Supplementary Table 1, Fig. 3).A particular problem in identifying palaeo-grounding events in sediment cores is the clear distinction of subglacial and glaciomarine facies (e.g.Anderson et al., 1980;Elverhøi, 1984;Licht et al., 1996Licht et al., , 1999;;Domack et al., 1999;Evans and Pudsey, 2002;Hillenbrand et al., 2005).For example, new sedimentological and micropalaeontological data on diamictons recovered from the WSE shelf that had previously been classified as subglacial tills (Anderson et al., 1980(Anderson et al., , 1983)), led to a reinterpretation of some of the diamictons as glaciomarine sediments (Stolldorf et al., 2012).Another challenge for the sedimentological identification of past grounding events on the WSE shelf is that here, in contrast to other sectors from the Antarctic continental shelf (e.g.Licht et al., 1996Licht et al., , 1999;;Domack et al., 1999;Heroy and Anderson, 2005;Ó Cofaigh et al., 2005a,b;Mosola and Anderson, 2006;Hillenbrand et al., 2010;Kilfeather et al., 2011;Smith et al., 2011;Kirshner et al., 2012), several cores contain glaciomarine sediments with low water content, high shear strength and high density, which may indicate their post-depositional over-consolidation by a grounded ice sheet (e.g.Elverhøi, 1981Elverhøi, , 1984;;Elverhøi and Roaldset, 1983;Haase, 1986;Melles, 1987;Melles and Kuhn, 1993;Hillenbrand et al., 2012).
The main dating method applied to shelf sediments in the Weddell Sea sector is radiocarbon ( 14 C) dating of calcareous microfossils, including radiometric 14 C dating and since the mid 1980s the much more sensitive Accelerator Mass Spectrometry (AMS) 14 C dating, which requires only 10 mg of calcareous material.Radiocarbon dating of biogenic carbonate does not suffer from the large uncertainties affecting 14 C dating of particulate organic matter (e.g.Andrews et al., 1999;Licht and Andrews, 2002;Mosola and Anderson, 2006;Rosenheim et al., 2008).However, calcareous microfossils are very rare in Antarctic shelf sediments and, as a consequence, only a few of the cores recovered from the WSE shelf have been dated (Supplementary Table 2, Fig. 4).Where calcareous microfossils had been sampled from glaciomarine sediments above Fig. 3. Sites of marine sediment cores retrieved from the continental shelf and upper continental slope (above 1000 m water depth) in the Weddell Sea sector and distribution of normally consolidated glaciomarine sediments, over-compacted glaciomarine sediments and subglacial tills recovered in these cores (for details, see Supplementary Table 1).
Fig. 4. Sites of marine sediment cores retrieved from the continental shelf and upper continental slope (above 1000 m water depth) in the Weddell Sea sector, for which radiometric and AMS radiocarbon dates have been published (for details, see Supplementary Table 2).Note that core PS1418 provided a core-top age only.
subglacial till, their 14 C dates were usually interpreted as minimum ages for grounded ice-sheet retreat (e.g.Anderson and Andrews, 1999).Most of the dated cores have provided just a single 14 C age (e.g.Kristoffersen et al., 2000b) or 14 C ages for horizons significantly above the transition of subglacial to glaciomarine sediments (e.g.Elverhøi, 1981).Several cores are characterised by down-core reversals of 14 C dates that may result from post-depositional sediment reworking and disturbance caused by iceberg scouring, current winnowing or debris flow redeposition (e.g.Anderson and Andrews, 1999;Kristoffersen et al., 2000a).Gravitational mass wasting is widespread on the continental slope of the Weddell Sea (e.g.Michels et al., 2002;Gales et al., 2012).Cores from further down the slope and the continental rise frequently recovered debris flow deposits, turbidites and contourites, i.e. sediments largely consisting of reworked material (e.g.Grobe and Mackensen, 1992;Kuhn and Weber, 1993;Melles and Kuhn, 1993;Anderson and Andrews, 1999).Therefore, we exclusively consider 14 C ages of cores collected from the continental shelf and the uppermost slope (i.e.shallower than 1000 m water depth) in this study.
Taking into account the problems of down-core age reversals and possible presence of subglacially compacted, originally glaciomarine sediments on the WSE shelf, the interpretation of the oldest or even the youngest 14 C date in a sediment core as a minimum age for the last retreat of grounded ice is not straightforward.These limitations, together with uncertainties about the increase of the marine reservoir effect (MRE) in the Southern Ocean during the last glacial period (e.g.Sikes et al., 2000;Van Beek et al., 2002;Robinson and van de Flierdt, 2009;Skinner et al., 2010), make it particularly challenging to reconstruct the timing of the last icesheet advance and retreat in the Weddell Sea sector from shelf sediments.
The marine 14 C dates mentioned under 'Datasets' (Section 4) are reported as in the original references, but the 14 C ages used for the 'Time-slice reconstructions and recent ice-sheet changes' (Section 5) and referred to in the 'Discussion' (Section 6) were all calibrated with the CALIB Radiocarbon Calibration Program version 6.1.0.We used an MRE correction of 1300 AE 70 years (Berkman and Forman, 1996), the uncertainty range of which overlaps with that of the core-top age of 1215 AE 30 14 C yrs BP obtained from site PS1418 on the upper slope just to the west of Crary Fan (Fig. 4, Supplementary Table 2), and the Marine09 calibration dataset (Reimer et al., 2009).Average calibrated 14 C ages are given for samples with replicate 14 C dates (Stolldorf et al., 2012), and corrected 14 C ages are given for 14 C dates that could not be calibrated.Uncorrected and corrected radiocarbon dates are given in 14 C ka BP (or 14 C yrs BP) and calibrated 14 C dates are given in cal ka BP (or cal yrs BP).All conventional and calibrated 14 C dates are listed in Supplementary Table 2.

Terrestrial studies
At the time of the last major review of ice-sheet extent and chronology in the WSE during the last glacial cycle (Bentley and Anderson, 1998) the mapped evidence of the onshore ice-sheet configuration, which included features marking the altitudinal extent of the former ice-sheet surface (e.g.erosional trimlines, moraines) and former flow direction indicators (e.g.striations, roches moutonnees), was limited and the dating control of these features was poor.Since then there has been a substantial increase in onshore glacial geological investigations around the embayment.The majority of studies have applied geomorphological mapping and cosmogenic surface exposure dating to mountain groups and nunataks located around the rim of the WSE, notably in the SE Antarctic Peninsula, Ellsworth Mountains, Pensacola Mountains, and Shackleton Range.These studies have provided important geomorphological constraints on former ice thickness configurations, including evidence from trimlines, sediment drifts, striated bedrock, and deposition of erratic clasts on exposed nunatak flanks.The latter have been particularly important because they have formed the primary target for dating former changes in ice-sheet elevation: erratics at a range of altitudes have now been dated at several locations extending around much of the WSE (e.g.Fogwill et al., 2004;Bentley et al., 2006Bentley et al., , 2010;;Hein et al., 2011Hein et al., , 2013;;Hodgson et al., 2012).We report the exposure dates in ka, corresponding to cal ka BP of the marine radiocarbon ages.A compilation of all the exposure dates from the hinterland of the Weddell Sea sector is provided in Supplementary Table 3.
There have also been other approaches to reconstructing former ice thickness.Two deep ice cores have been drilled in the WSE, or close to it, namely the Berkner Island core (Mulvaney et al., 2007) and the EPICA-Dronning Maud Land (EDML) core (EPICA Community Members, 2006) (Fig. 1).As with other ice cores the isotopic proxy records and gas bubble proxies can potentially be used to infer former ice sheet surface elevations.
Biological indicators of former ice absence (deglaciation) include accumulations of snow petrel stomach oil.Petrels rapidly colonise newly deglaciated areas of rock in East Antarctica, driven by competition for nesting sites, even up to 440 km from the coast.At their nest sites the petrels regurgitate stomach oil as a defence mechanism; this accumulates as a waxy grey coating, termed 'mumiyo', on the rocks, 100e500 mm thick, with a stratified internal structure.Radiocarbon ages show an increase with depth (Ryan et al., 1992) confirming that it is deposited by progressive accumulation of regurgitated oil, at a rate of 9e100 mm/ka.Dating of the base of these deposits has been shown to provide a minimum age for local deglaciation, and has been used in combination with cosmogenic isotopes to determine ice sheet thickness changes (e.g. in the Framnes Mountains in East Antarctica, Mackintosh et al., 2011).By using a sequence of dates on a single mumiyo deposit it is also possible to demonstrate continuous petrel occupation (i.e.ice absence) over millennia, or identify significant hiatuses (indicating that ice thickening may have occurred).Such deposits have been dated at a number of sites, but from the hinterland of the Weddell Sea only 14 C dates on mumiyo deposits collected from the Shackleton Range (Hiller et al., 1988(Hiller et al., , 1995)), western Dronning Maud Land (Thor and Low, 2011) and central Dronning Maud Land (Steele and Hiller, 1997) have been published.Nevertheless, it seems breeding sites of petrels are a near-ubiquitous feature of nunataks within a suitable range (up to ca 440 km) of feeding grounds.In line with the marine 14 C ages, we report all terrestrial 14 C dates mentioned under 'Datasets' (Section 4) as in the original references.A compilation of the terrestrial 14 C dates from the hinterland of the Weddell Sea sector is provided in Supplementary Table 4.
In almost all cases the primary focus of onshore studies has been the maximum configuration of ice at the local LGM in the region.Less is known about the post-LGM ice-sheet history but in some studies the deglacial portion of the last glaciation has also been constrained by thinning histories derived from dating material on nunatak 'dipsticks' (e.g.Todd and Stone, 2004;Bentley et al., 2010).
Other terrestrial studies in the Weddell Sea sector, such as radar and seismic investigations of the ice sheet, have also contributed to palaeo-ice sheet reconstructions.These datasets have helped to identify past changes in ice-flow directions (Campbell et al., 2013), reconstruct former ice-divide migration (Ross et al., 2011) and calculate palaeo-accumulation rates (Huybrechts et al., 2009).

Datasets
In the following, we summarise the datasets, outputs and interpretations of the marine and terrestrial studies that are relevant to reconstruct the LGMeHolocene glacial history of the Weddell Sea sector, thereby identifying their key constraints.

U.S. expeditions
Piston and gravity cores were recovered from the continental shelf of the Weddell Sea sector during the 'International Weddell Sea Oceanographic Expeditions' (IWSOE) aboard the USCGC Glacier from 1968 to 1970 and during cruise IO1578 aboard the ARA Islas Orcadas in 1978 (Supplementary Table 1, Fig. 3).Glaciomarine and subglacial facies on several of these cores were analysed by Anderson et al. (1980Anderson et al. ( , 1982Anderson et al. ( , 1983Anderson et al. ( , 1991)), but the first AMS 14 C ages obtained from glaciomarine sediments in the cores were not published until the late 1990s (Bentley and Anderson, 1998;Anderson and Andrews, 1999;Smith and Licht, 2000).According to these early studies, glaciomarine muds and glaciomarine diamictons overly subglacial till in Filchner Trough and seaward from the Riiser-Larsen to Fimbul ice shelves.The seabed of the eastern flank of Filchner Trough and its western flank (inner to mid shelf) consists of coarse-grained residual glaciomarine sediments and exposed basement rocks, while the rest of the WSE shelf comprises glaciomarine muds and diamictons (Bentley and Anderson, 1998).On the basis of the few available 14 C dates, Anderson and Andrews (1999) concluded that the last grounding event of the EAIS on the Weddell Sea shelf must predate w26 14 C ka BP (cf.Anderson et al., 2002).
Recently, Stolldorf et al. (2012) carried out more detailed grainsize analyses on some of the IWSOE and IO1578 cores and obtained numerous AMS 14 C dates from glaciomarine sediments, predominantly in cores from the eastern flank of Filchner Trough and the seabed offshore from the Brunt, Riiser-Larsen and Quar ice shelves.The authors reinterpreted some of the diamictons previously classified as subglacial tills as glaciomarine sediments (Fig. 3).This conclusion is consistent with the observation that the benthic foraminifera assemblages in those diamictons are identical with foraminifera assemblages characterising various glaciomarine environments in the Weddell Sea today and show no sign of subglacial reworking (Anderson, 1972a(Anderson, , 1972b)).Stolldorf et al. (2012) concluded from the range of the AMS 14 C dates that the EAIS did not ground on the shelf to the east of Filchner Trough after 30,476 cal yrs BP (Fig. 5).A single AMS 14 C date from the western flank of the inner shelf part of Filchner Trough (core G10) yielded an age of 48,212 cal yrs BP, while the only date from within Filchner Trough (core G7) provided an age of 8521 cal yrs BP.The older of two dates in core 2-19-1, which is located on the outermost shelf just to the west of Filchner Trough, gave an age of 17,884 cal yrs BP (Figs. 4 and 5;Anderson and Andrews, 1999;Smith and Licht, 2000;Stolldorf et al., 2012).

Norwegian expeditions
During the 'Norwegian Antarctic Research Expedition' (NARE) cruises with R/V Polarsirkel from 1976 to 1979, seismic profiles and sediment cores were collected from Filchner Trough, its eastern flank and offshore from the ice shelves extending eastward to the Fimbul Ice Shelf (Supplementary Table 1 , Fig. 3;Elverhøi, 1981Elverhøi, , 1984;;Elverhøi and Maisey, 1983;Elverhøi and Roaldset, 1983;Haugland, 1982;Haugland et al., 1985).The seismic profiles revealed a thin sediment drape overlying an unconformity extending from the Brunt to the Riiser-Larsen ice shelves and were interpreted to indicate repeated advance and retreat of grounded ice across the continental shelf during the Late Pleistocene (Elverhøi and Maisey, 1983;Elverhøi, 1984;Haugland et al., 1985).Profiles from Filchner Trough were interpreted as showing outcrops of Proterozoic crystalline basement along its eastern flank (Elverhøi and Maisey, 1983;Haugland et al., 1985).Near the iceshelf front, westward dipping units of stratified to massive sedimentary rocks, which are separated by erosional unconformities and assumed to be of Jurassic to Cenozoic age, onlap the acoustic basement and form the trough floor (Elverhøi and Maisey, 1983;Haugland et al., 1985).Subsequent analysis of palynomorphs in subglacial and glaciomarine sediments recovered in IWSOE cores from this area suggested an Early-Late Cretaceous age for these westward dipping strata (Anderson et al., 1991).On the inner and mid-shelf part of Filchner Trough, an angular unconformity separates the dipping strata from a thin veneer of Late Pleistocene-Holocene sediments on the trough floor and thick semiconsolidated flat-lying glacigenic sediments on the western trough flank (Elverhøi and Maisey, 1983;Haugland et al., 1985).At the transition from the middle to the outer shelf, these flat-lying strata, which are assumed to be of late NeogeneeQuaternary age, extend onto the trough floor and are underlain by a second unit of flat-lying glacigenic sediments of assumed early Neogene age.The upper unit displays a wedge-shaped geometry on the outer shelf part of Filchner Trough (Elverhøi and Maisey, 1983).The shelf in the vicinity of the Filchner Trough mouth and Crary Fan is characterised by pronounced glacial progradation (Haugland, 1982;Haugland et al., 1985;Kuvaas and Kristoffersen, 1991;Bart et al., 1999).
According to the lithological analyses on the NARE sediment cores (Elverhøi, 1981(Elverhøi, , 1984;;Elverhøi and Maisey, 1983;Elverhøi and Roaldset, 1983), the seabed of the Weddell Sea sector is characterised by the presence of a stiff pebbly mud interpreted as subglacial till or glaciomarine sediment that was subsequently compacted by grounded ice.This over-consolidated pebbly mud is locally overlain by a soft pebbly mud interpreted as glaciomarine sediment (for locations of subglacial, over-consolidated and normally consolidated sediments, see Fig. 3).Two radiometric 14 C dates obtained from glaciomarine sediments in core 212 on the outermost shelf to the west of Filchner Trough and core 214 from the uppermost continental slope yielded uncorrected radiocarbon ages of 31,290 14 C yrs BP and >35,100 14 C yrs BP, respectively (Supplementary Table 2 , Figs. 4e6;Elverhøi, 1981).However, the sediments in core 212 were subsequently considered to be disturbed by iceberg scouring and those in core 214 to be affected by current winnowing, and therefore these 14 C ages may not constrain the time of the last ice-sheet retreat (Bentley and Anderson, 1998;Anderson and Andrews, 1999).Another single 14 C radiometric date obtained from a glaciomarine diamicton in core 206 offshore from the Fimbul Ice Shelf provided an uncorrected radiocarbon age of just 3950 14 C yrs BP, and three more 14 C dates from core 234 at the uppermost slope offshore from the Riiser-Larsen Ice shelf gave uncorrected ages ranging from 21,240 to 37,830 14 C yrs BP in normal stratigraphic order (Supplementary Table 2, Figs. 4e6;Elverhøi, 1981Elverhøi, , 1984;;Elverhøi and Roaldset, 1983).
During NARE 84/85 with K/V Andenes additional side-scan sonar and shallow seismic data as well as several gravity and vibro-cores were collected north of the Kvitkuven Ice Rise, Riiser-Larsen Ice Shelf (Orheim, 1985;Lien et al., 1989).The same area was targeted with a detailed seismic survey during the Nordic Antarctic Research Expedition 1995/1996 aboard the Finnish R/V Aranda (Kristoffersen et al., 2000b), during which a 14.05 m long core with a recovery of 18% was drilled (core KK9601; Kristoffersen et al., 2000a).The seismic profiles revealed not only significant shelf progradation caused by repeated advances of a grounded EAIS to the shelf break during the Plio-/Pleistocene, but also that the shelf progradation west of Kvitkuven Ice Rise started earlier than further east (Kristoffersen et al., 2000b).The side-scan sonar data showed iceberg scour marks (Lien et al., 1989), while the seismic survey mapped two submarine moraine ridge complexes on the shelf that  2).5a: All ages grouped for different regions (note: Brunt Ice Shelf dates are exclusively from core 3-7-1, Quar Ice Shelf dates are exclusively from core 3-17-1 and the Fimbul Ice Shelf date is from core 206).Minimum ages are marked with arrows, and dates from cores recovered on the continental slope are underscored.Light grey shading indicates the time span of a are orientated parallel to the shelf edge (Fig. 2; Kristoffersen et al., 2000b).The sediment cores recovered glaciomarine sediments, with only two cores retrieving over-consolidated diamictons at their bases (Fig. 3; Orheim, 1985;Lien et al., 1989).A single AMS 14 C date was obtained from a normally consolidated diamicton in core A85-10 that was recovered from between the two moraine ridges (Fig. 4).Its uncorrected radiocarbon age of 18,950 14 C yrs BP was interpreted to indicate that either grounded ice had retreated from an earlier outer shelf position to the core site by this time or that the inner moraine ridge marks the maximum ice-sheet extent at the LGM (Supplementary Table 2, Fig. 6; Kristoffersen et al., 2000b).Core KK9601 was drilled landward from the inner moraine ridge and recovered glaciomarine muds, sands and diamictons that overlie a subglacial diamicton at its base (Kristoffersen et al., 2000a).Two AMS 14 C dates obtained from glaciomarine diamicton just above the till provided uncorrected radiocarbon ages of 30,040 and 37,750 14 C yrs BP, respectively, while six more dates obtained from the overlying sediments range from 3870 to 11,440 14 C yrs BP but not in stratigraphic order (Supplementary Table 2; Figs.4e6).These ages were interpreted to indicate (i) an initial phase of EAIS advance and retreat before w38 14 C ka BP, (ii) a second phase of grounded EAIS advance after w30 14 C ka BP and retreat before w11 14 C ka BP, and (iii) a short phase of local ice advance or iceberg ploughing during the Holocene (Kristoffersen et al., 2000a).
High-resolution seismic profiles collected along the front of the FilchnereRonne Ice Shelf in the season 1994/1995 indicate a westward transition of the westward dipping Jurassic to Cenozoic sedimentary strata described by Elverhøi and Maisey (1983) and Haugland et al. (1985) into flat-lying strata north of the central Ronne Ice Shelf and into a folded sequence north of the western Ronne Ice Shelf (Jokat et al., 1997).Recently, Stolldorf et al. (2012) presented the first multi-beam data from the Weddell Sea sector, which had been collected just north of the Filchner-Ronne Ice Shelf on R/V Polarstern cruise ANT-XII/3 in 1995.The seabed images revealed mega-scale lineations (MSGLs) on the inner shelf parts of Ronne Trough and Hughes Trough and more subtle subglacial lineations on the inner shelf part of Filchner Trough (Fig. 2).Based on potential hiatus from w46.5 to w41.5 corrected 14 C ka BP that may have affected the entire Weddell Sea sector.However, 14 C dates obtained from calcareous (micro-)fossils exceeding ca 35 14 C ka BP may be unreliable, and the true ages may be older (e.g.Takada et al., 2003;Hughen, 2007).5b: Conventional radiocarbon dates versus calibrated (or corrected) radiocarbon ages (i) offshore from the Ronne Ice Shelf and from within Filchner Trough, (ii) from the eastern flank of Filchner Trough, and (iii) offshore from the Riiser-Laren Ice Shelf.Only dates from cores recovered on the continental shelf are shown.Grey shading indicates the time spans of potential hiatuses.Note that the radiocarbon dates exceeding ca 35 14 C ka BP and the corresponding hiatuses may be unreliable.The dark grey shading highlights the most extended hiatuses in the three areas.These apparent hiatuses overlap during the time interval from w31.0 to w21.5 cal ka BP. the pristine preservation of the MSGLs, the authors proposed an LGM age for the last grounding event offshore from Ronne Ice Shelf.
The sediments recovered along the Ronne Ice Shelf front consist mainly of glaciomarine deposits with subglacial till reported only from site PS1197 in Hughes Trough and site PS1423 at the western flank of Ronne Trough (Fig. 3; Haase, 1986;Wessels, 1989;Crawford et al., 1996).An acoustically transparent layer in a subbottom profile from the inner shelf part of Ronne Trough suggests the presence of a soft till layer (Hillenbrand et al., 2012), which is consistent with the recent discovery of MSGLs on the trough floor there (Stolldorf et al., 2012).Along the ice-shelf front acoustic profiles extending from Ronne Trough to Filchner Trough revealed few details (Haase, 1986;Fütterer and Melles, 1990), but several of the glaciomarine sequences recovered from Hughes Trough and its flanks are overcompacted, possibly as a result of ice-sheet loading at the LGM (Fig. 3; Haase, 1986;Wessels, 1989;Hillenbrand et al., 2012).Two AMS 14 C dates from a normally consolidated glaciomarine diamicton at site PS1423, which was interpreted as an iceberg-rafted sediment deposited at a former ice-shelf calving line, provide the only age constraints for cores collected along the Ronne Ice Shelf front and yielded uncorrected radiocarbon ages of 3250 and 5910 14 C yrs BP, respectively (Supplementary Table 2 , Figs. 4e6;Hedges et al., 1995;Crawford et al., 1996).
Cores from the deepest part of Filchner Trough often recovered tills, while cores collected from the outer shelf frequently recovered over-consolidated glaciomarine sediments (Fig. 3; Melles, 1987Melles, , 1991;;Fütterer and Melles, 1990;Melles and Kuhn, 1993;Hillenbrand et al., 2012).Although this over-compaction was attributed to iceberg ploughing at some core sites (Melles, 1991;Melles and Kuhn, 1993), an LGM advance of a grounded ice sheet through Filchner Trough to the shelf break was considered as the most likely explanation for the distribution of over-consolidated glaciomarine sediments and tills in this area (Melles, 1987(Melles, , 1991;;Fütterer and Melles, 1990;Melles and Kuhn, 1993;Hillenbrand et al., 2012).This suggestion is supported by sedimentary sequences recovered on the adjacent continental slope, which indicate that during the last glacial period (i) glaciogenic detritus originating from the continental shelf was transported down-slope by mass movements and bottom-water flow, and (ii) katabatic winds blowing off an expanded ice sheet formed a polynya above the uppermost slope (Melles, 1991;Ehrmann et al., 1992;Melles and Kuhn, 1993).The conclusion of LGM ice-sheet grounding seems also to be consistent with: (i) the observation of 'hard' reflectors in acoustic subbottom profiles from the outer shelf, which are high-amplitude reflectors without reflections beneath them, suggesting that they are the acoustic expressions of glacial unconformities and surfaces of tills, respectively (Melles and Kuhn, 1993), and (ii) the recent discovery of subglacial bedforms within Filchner Trough (Larter et al., 2012;Stolldorf et al., 2012).Only eight 14 C dates were obtained from glaciomarine sediments recovered by R/V Polarstern from the continental shelf and the uppermost slope in the vicinity of Filchner Trough.The corresponding ages range from 1215 to 8790 14 C yrs BP (Supplementary Table 2, Figs. 4e6;Hillenbrand et al., 2012).Down-core abundance of planktonic and benthic foraminifera was sufficient at three sites from the outer WSE shelf (PS1420, PS1609, PS1611) for analysing stable oxygen isotopes (d 18 O) (Melles, 1991).However, the suitability of these data for applying d 18 O stratigraphy by identifying d 18 O shifts related to glacialeinterglacial transitions remains uncertain (Hillenbrand et al., 2012).
A hard seabed reflector was recorded in subbottom profiles offshore from the Brunt and Riiser-Larsen ice shelves but it remained unclear if this acoustic character resulted from coarse grain-size, over-compaction or a combination of both (Kuhn and Weber, 1993;Michels et al., 2002).The shelf cores collected offshore from the eastern Riiser-Larsen Ice Shelf and the Ekstrøm Ice Shelf contain exclusively glaciomarine sediments, for which a Holocene age was assumed (Grobe and Mackensen, 1992;Michels et al., 2002).

British expeditions
Multibeam swath bathymetry data and acoustic subbottom profiles (TOPAS) were collected in the Filchner Trough area by the British Antarctic Survey (BAS) during RRS James Clark Ross cruises JR97 in 2005 and JR244 in 2011 (Gales et al., 2012;Larter et al., 2012).On the inner shelf, these data revealed the presence of subglacial lineations in the axis of the trough and of drumlins on the lower part of its eastern flank (Fig. 2).Subglacial lineations that are locally eroded into an acoustically transparent layer were mapped in the mid-shelf part of Filchner Trough, and a groundingzone wedge located landward of linear iceberg furrows was mapped on the outer shelf (Fig. 2).These bedform assemblages were interpreted as the results of a Late Pleistocene ice-sheet advance through Filchner Trough, and an LGM age was proposed for their formation (Larter et al., 2012).

Summary of marine studies
The available seismic, swath bathymetry and sediment core data indicate ice-sheet grounding on the continental shelf of the Weddell Sea sector during the past, with ice grounding even in the deepest parts of the palaeo-ice stream troughs (Stolldorf et al., 2012) and grounded ice in Filchner Trough advancing onto the outer shelf to within at least 40 km of the shelf break (Larter et al., 2012).The pristine preservation of the mapped subglacial bedforms (Fig. 2) suggests that the last ice-sheet grounding directly north of Ronne Ice Shelf and within Filchner Trough occurred during the Late Pleistocene.However, the few available 14 C dates poorly constrain the timing of this grounding event, and therefore it remains unclear whether it happened at the LGM.When only shelf sites are considered and the date from core 212 is ignored (because of possible disturbance of its stratigraphy), all but one of the oldest ages obtained from cores recovered north of the Ronne Ice Shelf and within Filchner Trough are consistent with LGM grounding (Figs. 5 and 6;Hillenbrand et al., 2012;Stolldorf et al., 2012).However, these few dates are all minimum limiting ages and so do not rule out the grounding event being older.In contrast, the oldest ages obtained from cores on the uppermost continental slope and on the shelf to the east of Filchner Trough can be interpreted to indicate grounded ice-sheet retreat before 34 cal ka BP or even before 50 14 C ka BP (Figs. 5 and 6).It has to be taken into account, however, that (i) the sediments from the flanks of Filchner Trough and the upper continental slope are prone to reworking by debris flows because of a steep seafloor gradient, and (ii) those from the eastern Filchner Trough flank are prone to iceberg scouring because the corresponding core sites are located at water depths shallower than 550 m and thus above the mean keel depth of icebergs calving from the Filchner Ice Shelf (Dowdeswell and Bamber, 2007).Therefore, the dates from all those cores may be dismissed as unreliable for constraining the age of the last grounding-line retreat, which may be supported by down-core age reversals observed in some of the cores (cf.Anderson and Andrews, 1999).In addition, very old 14 C ages of near-surface sediments in conjunction with down-core dates in normal stratigraphic order at sites to the east of Filchner Trough indicate that sediments younger than w30 cal ka BP are missing at these locations (e.g.core 2-20-1; Fig. 4, Supplementary Table 2), which might be explained by subglacial erosion at the LGM.
Five 14 C dates spanning 15,876 to 27,119 cal yrs BP in normal stratigraphic order in core 3-17-1 offshore from the Quar Ice Shelf strongly suggest that the EAIS had retreated before ca 27.3 cal ka BP in this part of the Weddell Sea sector (Supplementary Table 2, Figs. 4 and 5; cf.Anderson and Andrews, 1999;Stolldorf et al., 2012).This scenario would not necessarily contradict a later, limited readvance north of the Riiser-Larsen Ice Shelf (Kristoffersen et al., 2000a,b).On the WSE shelf west of Filchner Trough, the time of the last WAIS retreat is only constrained by 14 C dates from five cores on the outermost shelf and upper slope seaward of the eastern Ronne Ice Shelf and from core PS1423 on the inner shelf part of Ronne Trough (Supplementary Table 2, Figs. 4 and 6).Thus, the assumption of an LGM age for the last ice-sheet advance in this area is based on (i) very few dates (Hillenbrand et al., 2012) and (ii) analogy with the glacial history of other WAIS drainage sectors (Stolldorf et al., 2012).
All 14 C ages available from the Weddell Sea sector extend back to 54 14 C ka BP and seem to hint at a possible hiatus spanning the time interval from w46.5 to w41.5 cal ka BP (Supplementary Table 2, Fig. 5a).It has to be pointed out, however, that radiocarbon dates on calcareous (micro-)fossils exceeding ca 35 14 C ka BP are inherently unreliable because of diagenetic alteration, and that the true ages may be much older (Hughen, 2007).For example, electron spin resonance (ESR) dating of mollusc shells from raised beach deposits in Lützow-Holm Bay in East Antarctica, which had provided uncorrected AMS 14 C ages spanning from 34.7 to 42.8 14 C ka BP, demonstrated that these molluscs had been deposited between 50 and 228 ka (Takada et al., 2003).On the Weddell Sea shelf, marine radiocarbon dates younger than 35 14 C ka BP do show considerable regional variability, which could be significant.So far, none of the cores from the shelf north of the Ronne Ice Shelf, within Filchner Trough, on the eastern flank of Filchner Trough and offshore from the Riiser-Larsen Ice Shelf (Fig. 4) provided ages falling into the intervals from w34.0 to w18.5 cal ka BP, w31.0 to w14.0 cal ka BP and w33.0 to w21.5 cal ka BP, respectively (Fig. 5b).These apparent hiatuses, which are the most extended in the three areas and include the time span of the LGM from 23 to 19 cal ka BP, may be interpreted as evidence that those parts of the WSE shelf were affected by subglacial erosion or non-deposition at the LGM.In contrast, the dates obtained from cores located offshore from the Brunt and Quar ice shelves (sites 3-7-1 and 3-17-1; Fig. 4) do not indicate any pronounced hiatus after w33 and w28 cal ka BP, respectively (Fig. 5a).
At the moment, we cannot preclude the possibility that the apparent hiatus from w31.0 to w21.5 cal ka BP observed north of the Filchner-Ronne and Riiser-Larsen ice shelves is an artefact resulting from the low number of available 14 C ages.Even if this hiatus is real, however, it does not necessarily imply an advance of grounded ice across the Weddell Sea shelf during that time because coverage with an ice shelf or perennial sea ice alone may have prevented the deposition of microfossils.Moreover, even the studies favouring grounded WAIS and EAIS advance across the southern WSE shelf at the LGM argue that the grounded ice had a very low profile, i.e. the grounding event itself was merely a slight 'touchdown' of an advancing ice shelf, and that the grounding may have been brief and in the order of a few thousand years (Hillenbrand et al., 2012;Larter et al., 2012).

Weddell Sea terrestrial studies
The subglacial topography of the WSE has been partly mapped by airborne radio echo sounding and seismic profiles (the latter especially over the FilchnereRonne Ice Shelf) by several nations.These surveys have been compiled into the Bedmap2 dataset (Fretwell et al., 2013).Terrestrial studies have focussed on five nunatak groups around the WSE rim and we describe results from these areas in turn.The terrestrial data are consistent in suggesting that ice-sheet thickening around the WSE rim during the LGM was of the order of only a few hundred metres, and in some areas may have been zero.This view of minor LGM thickening is critical in determining the reconstruction of post-LGM ice in the WSE, and so we spend some time discussing the assumptions that underpin it.

SE Antarctic Peninsula
The western Weddell Sea is fed partly by ice from the SE Antarctic Peninsula.Early glacial geological work (Carrara, 1979(Carrara, , 1981;;Waitt, 1983) suggested that the area had been over-ridden by an expanded ice sheet but the timing remained unknown.Further mapping and reconnaissance-level dating of this expanded ice sheet by Bentley et al. (2006) suggested that during the LGM the ice sheet thickened by over 300e540 m in the southernmost part of the Antarctic Peninsula and by 500 m further north.Striation data in Palmer Land show that when the APIS thickened, it did not merge to form a single dome, but rather, two or more of the present-day ice domes expanded and became thicker, and drove ice-sheet flow oblique to present trends (Bentley et al., 2006).Thinning of this ice sheet on the east side of the peninsula was underway by the Early Holocene such that it was <300 m thicker than present in the Behrendt Mountains by 7.2 ka (Bentley et al., 2006).Other attempts to date deglacial thinning were confounded by very high proportions of reworked erratic clasts that yielded complex ages and which were in some cases as old as 1.2 Ma (Bentley et al., 2006).

Ellsworth Mountains
Evidence for a formerly thicker ice sheet was mapped in detail by Denton et al. (1992).Although their evidence of past thickening was undated they provided a detailed map of former erosional and glacial drift evidence of the upper limits of ice sheet glaciation.Much of this effort focussed on a high (800e1000 m above present ice) glacier trimline, which is especially well-preserved in the Sentinel Range and also observed elsewhere in the Ellsworth Mountains.The altitudinal relationship between this erosional trimline and the present ice sheet surface throughout the region may suggest long-term stability of ice divide location even during ice-sheet expansion (Denton et al., 1992).This conclusion would be consistent with the interpretation of radio-echo-sounding and GPS data collected between Pine Island Glacier and Institute Ice Stream that document a stable position of the ice divide between the Amundsen Sea and the Weddell Sea drainage sectors of the WAIS for at least the last 7 ka and possibly for the last 10 to 20 ka, or even longer (Ross et al., 2011).Bentley et al. (2010) subsequently mapped a second trimline, which is significantly below the trimline reported by Denton et al. (1992) and exposed as a drift limit in the Heritage Range, southern Ellsworth Mountains.This lower drift limit drapes nunatak flanks in Marble Hills, Patriot Hills and Independence Hills and was deposited by ice that reached 230e480 m above present-day levels.Material below the limit is relatively fresh and lithologically diverse whilst the sparse patches of drift above the limit are highly weathered and have a much more restricted range of lithologies.Based on weathering and dating of erratics this lower drift limit was interpreted by Bentley et al. (2010) as the LGM upper surface of the ice sheet in this region (cf.Fogwill et al., 2012).Cosmogenic surface exposure dating showed that a moraine ridge forming the upper edge of this lower drift was abandoned by the ice sheet at or around 15 ka and that ice thinning to present levels occurred progressively through the Holocene.The data were inadequate to determine whether thinning continued through to the present-day or whether present ice elevations were achieved sometime earlier in the Late Holocene: exposure ages of erratics at the present-day margin yield youngest ages of 2 ka (Marble Hills) or 490 yrs (Patriot Hills).Above the moraine delineating the top of the lower drift, erratics yielded only older ages, some of which appeared to imply continuous exposure for several 100 ka (Todd and Stone, 2004;Bentley et al., 2010;Fogwill et al., 2012).
The underpinning assumptions of the chronology were subsequently debated (Clark, 2011;Bentley et al., 2011a).Specifically Clark (2011) questioned whether there was a (short-lived) ice sheet thickening above the lower trimline during the LGM.Bentley et al. (2011a) argued that this would require that the ice sheet did so without leaving behind fresh erratics, and that it would require an explanation for the weathering contrast above and below the lower drift limit, and why the depositional regime shifted from almost no deposition to extensive supraglacial deposition below a critical altitude.So, although the problems of using negative evidence were acknowledged, and that such a scenario could not be conclusively ruled out, Bentley et al. (2011a) argued that the most parsimonious explanation is that the lower trimline is the LGM limit and not an intermediate limit.
One implication of this is that discovery of any young (e.g.Holocene) ages above the lower trimline would invalidate the Bentley et al. (2010) model.
In an attempt to constrain the postglacial crustal rebound in Antarctica, Argus et al. (2011) analysed GPS data from stations around the Antarctic coast, the WSE and in the Ellsworth Mountains that had been recorded between 1996 and 2011.The authors found that the Ellsworth Mountains are currently rising at a rate of ca 5 AE 4 mm/yr (95% confidence limits) and concluded that significant ice loss there must have ended by 4 ka.

Pensacola Mountains
Boyer (1979) made geomorphological observations in the Dufek Massif (northern Pensacola Mountains) that showed a complex glacial history of regional ice sheet over-riding and local outlet glacier advance.As with other early work the available techniques meant that the author was unable to date the evidence of glacial fluctuations.The first cosmogenic surface exposure dating of Dufek Massif was carried out by Hodgson et al. (2012).This study revealed evidence of a long glacial history, mostly prior to the timescale relevant to this paper.However, mapping of boulder ice-sheet moraines in Davis Valley and cosmogenic surface exposure dating of erratics on the moraines, along with radiocarbon ages around the margins of a pond in the adjacent Forlidas Valley suggest only moderate ice sheet thickening and advance of less than 2.5 km along-valley during the last glacial advance, assumed to be the LGM (Hodgson et al., 2012).The timing of this advance is not well constrained but radiocarbon dates on lacustrine algae show that the ice sheet had retreated from Forlidas Valley by 4300 cal yrs BP (Hodgson and Bentley, 2013).Hegland et al. (2012) and Bentley et al. (2012) reported preliminary results of fieldwork undertaken in the Williams, Thomas and Schmidt hills.They observed glacial scours on Mount Hobbs, Williams Hills, and striations on Martin Peak, Thomas Hills, suggesting a maximum ice thickness that was at least 562e675 m greater than today.However, no chronological constraints are currently available for this ice-sheet elevation highstand.The authors also observed moraines consisting of unweathered till at altitudes between 20 and 100 m above the present ice-sheet surface and assumed that these are likely to post-date the LGM.Using measurements of radar-detected stratigraphy, surface ice-flow velocities and accumulation rates Campbell et al. (2013) investigated the relationships between local valley-glacier and regional icesheet dynamics in and around the Schmidt Hills.The authors found evidence that ice-margin elevations in the Schmidt Hills have lowered by about 3 m over the last ca 1200 years without a concurrent change in the surface elevation of the neighbouring Foundation Ice Stream.

Shackleton Range
In the Shackleton Range the summits have been over-ridden by the ice sheet but cosmogenic isotope data suggest this happened over 1 Ma ago (Fogwill et al., 2004).Lateral moraines that lie w250 m above present day ice at the grounding line and w200 m above present ice further upstream on Slessor Glacier were originally suggested to most likely mark the upper limit of the LGM ice sheet but were not dated directly (Höfle and Buggisch, 1993;Kerr and Hermichen, 1999;Fogwill et al., 2004;Bentley et al., 2006).More recently, a comprehensive geomorphological and cosmogenic dating study of the lower flanks of the Shackleton Range showed that there was no direct evidence of any significant thickening during the LGM (Hein et al., 2011(Hein et al., , 2013)), and indeed the data are best explained by stability of the Slessor-Recovery ice stream system during the LGM.Dating of erratic boulders yielded a pattern of 'young' (<50 ka) ages that were confined, without exception, to the moraines forming at the present-day ice sheet margin.Above these moraines all exposure ages were >109 ka, and many of these showed a complex exposure history.
The simplest explanation of this pattern is that the LGM ice sheet did not thicken in the Shackleton Range e and may even have been thinner than present e and that the higher, older erratics all date to previous (pre-LGM) ice sheet expansions (Hein et al., 2011(Hein et al., , 2013)).As with the Ellsworth Mountains it is not possible to rule out the possibility of short-lived thickening events that spanned only several hundred to a few thousand years and left no erratics or other geological imprint, but after discussing such alternative explanations (cold-based ice leaving no erratics, or change in ice dynamics such that erratics were not brought to the margin along the ice streams), Hein et al. (2011) concluded that these would require conditions for which there was neither data nor observations, and hence they favoured the minimal LGM thickening model.
We note also that two dates on sub-samples of mumiyo from a site on Mt.Provender were reported by Hiller et al. (1988Hiller et al. ( , 1995) ) but the precise sample location was not reported, and so we cannot assess its relationship to present-day ice.The uncorrected ages were 8970 AE 250 and 9770 AE 200 14 C yrs BP (no laboratory codes given).

Western and central Dronning Maud Land
Constraints for ice thickness changes in western Dronning Maud Land since the LGM are restricted to the Heimefrontfjella region (the westernmost part of Maudheimvidda, see Fig. 1), where Hättestrand and Johansen (2005) carried out geomorphological mapping and Thor and Low (2011) collected mumiyo samples for radiocarbon dating.Hättestrand and Johansen (2005) mapped moraines in the vicinity of the Scharffenbergbotnen valley (centred at ca 74 35'S, 11 08'W and 1200e1600 m above sea level), which extend up to 200e250 m above the present ice surface on the surrounding valley slopes, and generally to less than 100 m above the present ice surface on slopes outside the valley.Although the authors did not obtain dates from the moraines, they tentatively inferred an LGM age for them.The radiocarbon dates from the basal layers in two mumiyo samples collected on the Haldorsentoppen nunatak in Sivorgfjella directly to the SW of the Scharffenbergbotnen valley (at ca 74 34'36 00 S, 11 13 0 24 00 W and 1245 m above sea level) yielded ages of 37,400 AE 1500 and 3120 AE 70 uncorrected 14 C yrs BP, respectively (Thor and Low, 2011).These dates indicate that Sivorgfjella may not have been over-ridden by ice since at least w37 14 C ka BP.Huybrechts et al. (2007) carried out modelling of stable hydrogen and oxygen isotopic data from the EDML ice core drilled in central Dronning Maud Land (75 00 0 S, 0 04 0 E; Fig. 1).The results suggest there was initial post-LGM thickening followed by thinning over the last 5 ka (Huybrechts et al., 2007).Accumulation rates in central Dronning Maud Land were shown to have been 1.5e2 times lower during the last glacial period than after ca 15 ka (Huybrechts et al., 2009).Steele and Hiller (1997) reported a large number of mumiyo ages from the near-coastal part of central Dronning Maud Land.These were from a variety of sites including close to present ice (nunatak foot), nunatak summits and intermediate sites.Dates from the nunatak foot locations show that ice was at present-day levels by 5590 corrected 14 C yrs BP ('Ice Axe Peak' locality at Robertskollen, 71 28 0 S, 3 15 0 W) and 6400 corrected 14 C yrs BP (Vesleskarvet, 71 40 0 S, 2 51 0 W).Minimum ages for clearance of summits are 7030 corrected 14 C yrs BP ('Tumble Ice' locality at Robertskollen, 40 m above present ice surface) and 6720 corrected 14 C yrs BP ('Nunatak V' locality at Johnsbrotet, 71 20 0 S, 4 10 0 W, 100 m above present ice surface).A further study at the same summit locality at Robertskollen yielded mumiyo showing continuous ice absence since 7000 cal yrs BP (Ryan et al., 1992).Based on their GPS data analysis, Argus et al. (2011) reported that the near-coastal part of Dronning Maud Land (Vesleskarvet) is currently rising at a rate of ca 4 AE 2 mm/yr in response to Holocene unloading of ice.
Although outside our sector it is relevant to note that samples from the Untersee Oasis (71 S, 13 E) show ice absence at nunatak foot locations as far back as w33 corrected 14 C ka BP (Hiller et al., 1988(Hiller et al., , 1995;;Steele and Hiller, 1997;Wand and Hermichen, 2005): these are at near-coastal locations landward of the narrow East Antarctic shelf and so may be indicative of ice-sheet history on the shelf immediately east of Filchner Trough.

Berkner Island
At the site of the Berkner Island ice core (79 34 0 S, 45 39 0 W; Fig. 1) the stable isotope data are consistent with continuous accumulation on a local ice dome, and appear to exclude the possibility that Berkner Island was over-ridden by interior ice during the LGM.For this reason they can be used to provide a maximum constraint for former ice sheet configurations in the embayment, namely that Berkner Island remained an independent ice dispersal centre throughout the LGMeHolocene (Mulvaney et al., 2007;Bentley et al., 2010).

Summary of terrestrial studies
The terrestrial data show that the WSE preserves a complex glacial history extending over millions of years but with only very minor thickening during the LGM.The available dating evidence suggests that maximum ice sheet expansion (to upper trimline in Ellsworth Mountains, over nunatak summits in Shackleton Range and Dufek Massif) occurred substantially prior to the last glacial cycle, and in some cases millions of years ago.Where dating evidence exists the LGM is represented by modest thickening (>340e540 m in SE Antarctic Peninsula, 230e480 m in Ellsworth Mountains, very minor in Dufek Massif, and near to zero in the Shackleton Range).Bentley et al. (2010), Le Brocq et al. (2011) and Whitehouse et al. (2012) have explored the use of the terrestrial constraints on former ice sheet thickness to delimit former ice sheet extent in the WSE, and specifically in Filchner Trough.The model results were consistent with very limited grounding-line advance in the Filchner and Ronne troughs.On the other hand, a recent modelling study on LGM ice-sheet thickness in Antarctica could reproduce successfully constraints on former ice-sheet elevations provided by terrestrial data and ice cores in most Antarctic sectors, but notably not in the eastern WSE, where the predicted ice sheet is thicker than indicated by the terrestrial data (Golledge et al., 2012).Fig. 7. Oldest calibrated radiocarbon ages obtained from cores offshore from the Brunt, Quar and Fimbul ice shelves (Fig. 5a) and calibrated radiocarbon ages constraining the termination of the most extended hiatuses observed north of the Filchner-Ronne and Riiser-Larsen ice shelves (see Fig. 5b).Only dates from cores collected from the continental shelf are displayed.These ages form the basis for the time-slice reconstructions according to Scenario B (see Figs. 12e16).

Time-slice reconstructions and recent ice-sheet changes
At present the terrestrial and marine data suggest two alternative reconstructions of the LGM ice-sheet extent in the Weddell Sea sector.Importantly both of these scenarios are consistent with low excess ice volumes during the LGM and deglacial period, which would imply only a minor contribution (i.e.just a few metres) to global meltwater pulses during the last deglaciation (Bentley et al., 2010;Hillenbrand et al., 2012).
Scenario A assumes that the LGM extent of ice in the Weddell Sea sector was largely as modelled using terrestrial data to constrain ice-sheet thickness by Bentley et al. (2010), Le Brocq et al. (2011) and Whitehouse et al. (2012).In this scenario even the oldest dates obtained from the marine sediment cores (Fig. 6) are minimum ages for grounded ice retreat from the continental shelf, and the grounding event recorded in subglacial bedforms and sediments was substantially pre-LGM.Scenario A implies that significant grounded ice-sheet advance during the LGM was restricted to the shelf offshore from the eastern and central Ronne Ice Shelf, whereas the grounding line remained in the vicinity of its modern position or showed only minor advance in most of the Weddell Sea sector and especially in the deep Filchner and Ronne troughs.This scenario was also the preferred explanation for the old marine radiocarbon ages obtained from the East Antarctic continental shelf of the Weddell Sea sector (Anderson and Andrews, 1999;Stolldorf et al., 2012).For the various time-slices of grounding line position in Scenario A we give those derived from modelling studies of Whitehouse et al. (2012).We use linear interpolation between the modelled position of the grounding line at 20 ka, which is based initially on Bentley et al. (2010) and Le Brocq et al. (2011), and the present-day position of the grounding line to infer its location at 15 ka, 10 ka and 5 ka.It is important to note that the reconstructed grounding-line positions are not therefore based on marine geological evidence but instead are inferred, based on glaciological modelling to remain consistent with terrestrial geological data.Full details of this approach are given in Whitehouse et al. (2012).
Scenario B assumes that the dates from the marine sediment cores are a mix of minimum and maximum ages for the last icesheet retreat (i.e. that the old dates were obtained from reworked microfossils that lived before the last ice-sheet advance) and that the most extended of the apparent hiatuses observed in the different parts of the Weddell Sea sector (see Fig. 5b) were caused by grounded ice sheet advance across the core sites.In Scenario B the dates constraining the termination of the hiatus between w31.0 and w21.5 cal ka BP observed north of the FilchnereRonne and Riiser-Larsen ice shelves are ages close to the last grounding-line retreat.The corresponding dates (taken from shelf cores only) are displayed in Fig. 7.According to Scenario B, grounded ice did extend to the shelf break north of the FilchnereRonne Ice Shelf during the LGM.To ensure consistency with the terrestrial data this scenario requires very thin, low profile ice on the continental shelf.This ice may have been just thick enough for grounding and may have remained grounded for only several hundred to a few thousand years (Bentley et al., 2010;Le Brocq et al., 2011;Hillenbrand et al., 2012;Larter et al., 2012).For the various time-slices displaying the ice-sheet extent according to Scenario B from 25 cal ka BP to 5 cal ka BP (Figs. 12e16), we give different certainty levels for the grounding-line positions.These levels indicate whether the grounding-line position is (i) constrained by nearby subglacial bedforms of unknown age (Fig. 2), (ii) constrained by nearby sediment cores that recovered subglacial/over-consolidated deposits (Fig. 3), for which no or only limiting ages are available, or (iii) simply inferred.
One crucial limitation for the palaeo-grounding line reconstructions in both scenarios is the lack of marine geophysical and geological information for the middle and outer shelf offshore from the Ronne Ice Shelf (Figs. 2 and 3).Here, no data on subglacial bedforms exist and no dates have been obtained from the only two cores (IWSOE68-2 and IWSOE68-11), which recovered less than 40 cm of glaciomarine sediments (Supplementary Table 1).The maximum grounding-line position in this part of the WSE predicted by Scenario A is inferred from the relationship between icesheet thickness constrained by the terrestrial data from the hinterland and shelf bathymetry, thereby using a modelled ice-sheet surface profile (for details see Whitehouse et al., 2012).The reconstruction shown in Scenario B is based on the assumption that ice draining the WAIS and APIS at the LGM advanced onto the outer shelf as it did in their other drainage sectors along the Pacific margin (e.g.Anderson et al., 2002;Livingstone et al., 2012).

Scenario A
20 ka: The ice sheet was at or close to its maximum thickness in the Ellsworth Mountains, was at a maximum thickness in the SE Antarctic Peninsula and was at its present level or thinner in the Shackleton Range.Berkner Island was an independent ice dispersal centre and thus not over-ridden by inland ice (Fig. 8).Glaciological modelling of the ice-sheet grounding line to remain consistent with onshore glacial geological data suggests that the 20 ka grounding line had reached close to the continental shelf break at the mouth of Hughes Trough and in the region immediately north of Berkner Island.In the Filchner Trough and Ronne Trough grounded ice was much less extensive and was confined to the inner-or mid-shelf parts of these troughs (Fig. 8).On the shelf east of Filchner Trough the grounding line was located on the mid-shelf and did not reach the continental shelf break.Although the grounding line is shown with a deep embayment within two of the Weddell Sea troughs, in reality we expect there to have existed either extensive ice shelves or lightly-grounded ice across these regions, both of which could have supported the rapid streaming ice flow which typically occurs along major ice-sheet outlets (Whitehouse et al., 2012).Modelling of EDML ice core isotopic data suggest accumulation-driven thickening began at this time in the Dronning Maud Land region and continued through to w5 ka.Mumiyo ages from the easternmost part of the sector and adjacent regions suggest that there ice may have been close to its present-day thickness since w33 corrected 14 C ka BP.
15 ka: The lower trimline in the Ellsworth Mountains was abandoned by the thinning ice sheet at or around 15 ka, which continued through the Holocene.In the Shackleton Range the ice was at its present level or thinner.According to the model of Whitehouse et al. (2012), the grounding line had retreated landward along troughs and away from the continental shelf break north of Berkner Island (Fig. 9).On the shelf east of Filchner Trough the grounding line had retreated back onto the inner shelf such that it was close to or at the present-day grounding line.
10 ka: The grounding line had continued its retreat and was located on the inner shelf everywhere, except immediately north of Berkner Island (Fig. 10).Ice elevations in the Shackleton Range were at present-day levels or thinner.
5 ka: The grounding line was at or close to the present-day grounding line, and so for example in the southernmost Weddell Sea was only a few tens of kilometres from the modern grounding lines of Foundation Ice Stream, Support Force Glacier and Institute Ice Stream (Fig. 11).In the Ellsworth Mountains ice elevations were <160 m above present, and most LGM ice had been lost by ca 4 ka.In the SE Antarctic Peninsula ice was <300 m thicker than present, while ice elevations in the Shackleton Range were at present-day levels or thinner.The precise timing at which the ice elevations in the Ellsworth Mountains and SE Antarctic Peninsula reached present are not tightly-constrained but the data from the Ellsworth Mountains are consistent with this occurring sometime between 2 ka and present.In the Pensacola Mountains ice had largely retreated from Forlidas Valley in the Dufek Massif by 4.3 ka, and ice-margin elevations in the Schmidt Hills lowered by ca 3 m over the last 1200 years.Modelling of isotopic data from EDML suggests that ice-sheet thinning in central Dronning Maud Land began around 5 ka.Many sites there and further east showed continuous accumulation of mumiyo (and thus ice close to present levels) prior to w5 cal ka BP.

Scenario B
25 cal ka BP: Dates from sites 3-7-1 and 3-17-1 (Fig. 4) indicate that grounded ice had retreated from the shelf offshore from the Brunt Ice Shelf and the Quar Ice Shelf (Fig. 12).In the rest of the Weddell Sea sector, the grounding line may have been located at the shelf break or at an outer shelf position.The outer moraine belt observed north of the Riiser-Larsen Ice Shelf (Fig. 2) may mark the grounding-line position in this area at 25 cal ka BP. 20 cal ka BP: The grounding line had retreated from site A85-10, which lies landward of the outer moraine belt north of the Riiser-Larsen Ice Shelf (Fig. 4).The inner moraine belt (Fig. 2) may have been deposited at this time.The chronology of core 2-19-1 indicates that the outermost shelf between Filchner Trough and Hughes Tough had become free of grounded ice at some time before 18.2 cal ka BP (Figs. 4 and 6).Therefore, we assume that the grounding line had started to retreat from the shelf break in most parts of the Weddell Sea sector at around 20 cal ka BP (Fig. 13).
15 cal ka BP: The WAIS and EAIS had retreated from outer shelf locations north of the FilchnereRonne Ice Shelf (Fig. 14).A grounding-zone wedge and linear iceberg furrows on the outermost shelf within Filchner Trough (Fig. 2) suggest that a pause in ice-sheet retreat and a minor re-advance occurred after initial grounding-line retreat (Larter et al., 2012).Offshore from the Riiser-Larsen Ice Shelf, the grounding line may have started to retreat from the inner moraine belt.
10 cal ka BP: The outer shelf on the eastern flank of Filchner Trough (site G2, Fig. 4) and the inner shelf north of the Riiser-Larsen Ice Shelf (site KK9601, Fig. 4) were free of grounded ice (Fig. 15).Ice retreat in the rest of the study area continued.
5 cal ka BP: The grounding line was located landward of most of the core sites, for which chronological information is available (Fig. 16).Only individual small embayments along the Coats Land coast may have remained covered by grounded ice at 5 cal ka BP (e.g.site G17, Fig. 4).In the western part of the Weddell Sea sector, the grounding line may have been located close to the modern calving lines of the FilchnereRonne Ice Shelf.

Recent changes
Satellite radar altimetry measurements indicated that those parts of the EAIS which drain into the Weddell Sea sector to the east of Filchner Trough had thickened by a few centimetres per year from 1992 to 2003 (Davis et al., 2005).Also the catchments of ice streams feeding into the Filchner and Ronne ice shelves thickened during that time period, while their fast moving sections remained unchanged (Joughin and Bamber, 2005).Radar interferometry data collected between 1992 and 2006 suggested a positive mass balance for the Filchner Ice Shelf but the measurements for the Ronne Ice Shelf and the drainage basins east of Filchner Trough were inconclusive (Rignot et al., 2008).More accurate laser altimeter measurements carried out between 2003 and 2008 revealed a thickening of 2e4 cm/yr for most ice shelves in the Weddell Sea sector, a thinning of 1e2 cm/yr for the Quar and Ekstrøm ice shelves and no change for the Fimbul Ice Shelf (Pritchard et al., 2012).Recently, a study using the same data set came to similar conclusions regarding the ice-shelf melting in the eastern part of the Weddell Sea sector but concluded a thinning of 13 AE 10 cm/yr for the Filchner Ice Shelf and 14 AE 10 cm/yr for the Ronne Ice Shelf (Rignot et al., 2013).
No significant advances or retreats of the grounding line have been reported for the Weddell Sea sector over the last few decades.However, major iceberg calving events affected the Filchner and Ronne ice shelves between 1986 and 2000 (e.g.Lambrecht et al., 2007).These recurrent calving events had a complex impact on sea-ice concentration and water mass circulation, and thus on melting and freezing processes in the sub-ice shelf cavity (e.g.Grosfeld et al., 2001;Nicholls et al., 2009).Therefore, minor shifts of the grounding line in response to these calving events cannot be ruled out.

Discrepancies between the reconstructions from marine and terrestrial datasets and possible explanations
The main discrepancies between Scenarios A and B in reconstructing the ice-sheet configuration in the Weddell Sea sector during the last glacial period are (i) the maximum extent of grounded ice on the continental shelf (except for the shelf between the Filchner and Ronne troughs), and (ii) the grounding-line positions in the deep inner shelf parts of the Filchner and Ronne troughs (Figs. 8 and 12).The differences in grounding-line positions during the last deglaciation (Figs.9e11 and 13e16) are direct consequences of these mismatches in maximum ice-sheet size.The discrepancies in the reconstructed maximum ice-sheet configurations are probably larger than for any other Antarctic sector.We discuss possible reasons for this below but it has to be kept in mind that in only a few sectors of Antarctica are both cosmogenic exposure ages and marine deglaciation dates available from the same drainage basin.Examples of such areas are the Mac.Robertson Land coast in East Antarctica and the Marguerite Trough palaeo-ice stream basin on the SW Antarctic Peninsula, where both datasets allowed consistent palaeo-reconstructions (Mackintosh et al., 2011;Bentley et al., 2011b).Nevertheless, more drainage basins should be targeted by both terrestrial and marine dating in order to evaluate whether the apparently inconsistent marine and terrestrial reconstructions in the Weddell Sea sector are exceptional.
If Scenario A were correct, the pristine preservation of subglacial bedforms of pre-LGM age on the WSE shelf would imply that glaciomarine deposition over the last 25 ka was insufficient to bury these features.Elsewhere, it has been recognised during the last few years that even some pristine glacial landforms mapped on the Antarctic continental shelf provide a composite picture resulting from different phases during either the same glacial period or Fig. 12. Grounded ice-sheet extent in the Weddell Sea sector at 25 cal ka BP according to Scenario B. The position of the grounding line (GL) was reconstructed using the ages displayed in Fig. 7.The different certainty levels given for the GL indicate whether its position is (i) constrained by nearby subglacial bedforms of unknown age (Fig. 2), (ii) constrained by nearby sediment cores that recovered subglacial/over-consolidated deposits of unknown age (Fig. 3), or (iii) simply inferred.different glacial periods (e.g.Heroy and Anderson, 2005;Graham et al., 2009;Reinardy et al., 2011).Furthermore, sedimentation rates under Antarctic ice shelves are as low as ca 2e3 cm/ka (e.g.Hemer et al., 2007).Therefore, formation of the subglacial geomorphology on the WSE shelf during the penultimate glacial period (Marine Isotope Stage 6 from ca 191e130 ka) combined with longterm ice shelf coverage throughout the last glacial period could explain its pristine appearance (Larter et al., 2012).Notably, the mismatch between Scenarios A and B in the Weddell Sea sector is not only based on different conclusions from the available terrestrial and marine datasets, but also on different interpretations of the available radiocarbon ages obtained from the marine sediment cores.These interpretations crucially depend on the facies assignment of the sediments the dated samples were taken from (cf.Elverhøi, 1981 with;Anderson and Andrews, 1999;cf.;Anderson et al., 1980 with;Stolldorf et al., 2012).If a date was obtained from microfossils deposited in-situ within a glaciomarine setting, it would give a minimum age for grounded ice-sheet retreat, but if reworked microfossils from a subglacial till were dated, the corresponding age would provide a maximum date for the last advance of grounded ice across the core site.An additional complication in the Weddell Sea sector is that here, in apparent contrast to other Antarctic sectors, glaciomarine sediments may have been overconsolidated after their deposition by overriding grounded ice (e.g.Elverhøi, 1984;Hillenbrand et al., 2012).This problem implies that even if conclusive evidence for the glaciomarine origin of a sample of pre-LGM age is provided, the date does not necessarily rule out grounded ice advance across the core site during the LGM.
Notably the evidence for grounding on the shelf provided by the presence of subglacial bedforms and the occurrence of subglacially over-consolidated as well as subglacially deposited sediments is consistent with a short-lived ice sheet advance that lasted only several hundred to a few thousand years.This raises the possibility that if Scenario B were correct, then the boundary between unweathered and weathered rocks observed in the Shackleton Range (Fogwill et al., 2004) and the Ellsworth Mountains (Bentley et al., 2010) might not indicate the maximum elevation of the LGM icesheet surface, but an intermediate elevation following short-lived LGM ice-sheet thickening (Clark, 2011).Thicker, non-erosive, cold-based ice may have preserved 'weathered' rocks above these limits at the LGM.If the maximum thickening occurred for a short term only, it may not be resolved in the available exposure ages.These explanations were not completely ruled out by Bentley et al. (2011a) and Hein et al. (2011), but considered to be very unlikely, and that there was no terrestrial evidence for such an ice sheet thickening.Both Hillenbrand et al. (2012) and Larter et al. (2012) point out that short-term LGM grounding were consistent with their observations and interpretation of the marine datasets.Evidence is growing that subglacial features formed in a soft substrate on the Antarctic continental shelf may only represent a 'snapshot' of the latest phase of maximum ice-sheet extent (Graham et al., 2009;Reinardy et al., 2011), which is consistent with the rapid formation and erosion of bedforms under contemporary ice streams (e.g.Smith et al., 2007Smith et al., , 2012)).
Whatever the duration of the LGM thickening, at least three scenarios have been suggested that can reconcile the marine and terrestrial datasets.These were summarised by Larter et al. ( 2012): (i) The LGM ice sheet had an extremely low surface gradient and resembled an 'ice plain' (cf.Le Brocq et al., 2011;Hillenbrand et al., 2012).'Ice plains' are observed just upstream of the grounding line of some contemporary ice streams and are characterised by very low basal shear stresses, resulting in surface slope angles with tangents <10 À3 (e.g.Alley et al., 1989;Bindschadler et al., 2005).(ii) The FilchnereRonne ice shelf advanced across the shelf, and a minor thickening combined with the LGM sea-level drop of ca 130 m resulted in a 'touchdown' of the ice shelf/sheet on the seabed.Support for this hypothesis comes from the widespread occurrence of initially glaciomarine sediments that were over-consolidated at some time after their deposition (Fig. 3; Elverhøi, 1984;Haase, 1986;Melles, 1987;Wessels, 1989;Hillenbrand et al., 2012).(iii) At the LGM, Slessor and Recovery glaciers had become cold-based and stagnated, while Support Force Glacier and Foundation Ice Stream had remained warm based and both fed into the palaeo-ice stream draining through Filchner Trough (Fig. 1).Such a flowswitch of Foundation Ice Stream is consistent with both some earlier reconstructions of the LGM drainage pattern (Hughes, 1977) and subglacial topography (Fretwell et al., 2013) indicating the locus of long-term erosion around Berkner Island.As a consequence of these ice-flow changes, LGM ice-sheet thickening in the Shackleton Range may have remained insignificant, which is consistent with the conclusion by Hein et al. (2011), even though there was grounded ice advance in Filchner Trough.However, advance of a grounded ice stream through Filchner Trough should have provided a buttressing back-stress for Recovery and Slessor glaciers.This would have resulted in their thickening because elsewhere in Antarctica downstream ice 'damming' has caused significant glacier thickening (e.g.Anderson et al., 2004).

Consistencies between the reconstructions from marine and terrestrial datasets
Despite all the discrepancies between Scenarios A and B, there are two remarkable consistencies.First, in both scenarios the contribution of ice-sheet build-up in the Weddell Sea sector during the LGM made only a very minor contribution of a few metres to the global sea-level low stand of ca 130 m during this time (cf.Bentley et al., 2010;Le Brocq et al., 2011;Hillenbrand et al., 2012;Larter et al., 2012;Stolldorf et al., 2012;Whitehouse et al., 2012).Consequently, melting of glacial ice in this sector during the last deglaciation cannot have made a dominant contribution to the meltwater pulses of 10e15 m around ca 19.1 cal ka BP (Clark et al., 2004) and of 10e18 m at 14.6 cal ka BP, even though an Antarctic source has been repeatedly proposed for meltwater pulse 1A at 14.6 cal ka BP (Clark et al., 2002;Weaver et al., 2003;Deschamps et al., 2012).
Second, even in Scenario B the seabed offshore from the Brunt and the Quar ice shelves was free of grounded ice by at least 25 cal ka BP (Fig. 12).Thus, both Scenario A and Scenario B indicate diachronous ice-sheet retreat from the continental shelf of the Weddell Sea sector, with at least parts of the EAIS retreating earlier than the WAIS.This conclusion is consistent with earlier reconstructions of post-LGM ice-sheet retreat from the Antarctic continental shelf on both a regional scale (Elverhøi, 1981;Anderson and Andrews, 1999;Stolldorf et al., 2012) and a continental scale (Anderson et al., 2002;Livingstone et al., 2012) but is inconsistent with the conclusions of Clark et al. (2009) and Weber et al. (2011) who argued for synchronous advance and retreat around Antarctica.Furthermore, time-transgressive ice-sheet retreat may help to explain the in-situ survival of Antarctic shelf benthos during glacialeinterglacial cycles (Thatje et al., 2005;Convey et al., 2009).Interestingly, Barnes and Hillenbrand (2010) inferred from the similarity of modern bryozoan assemblages on the Ross Sea shelf and the Weddell Sea shelf (i.e. from samples collected on the seabed offshore from the Brunt, Riiser-Larsen, Quar, Ekstrøm and Jelbart ice shelves) that the seafloor in the two sectors could not have been completely overridden by grounded ice during the last glacial period.This conclusion is in agreement with both geological data from the Ross Sea (Licht et al., 1996(Licht et al., , 1999;;Domack et al., 1999;Shipp et al., 1999;Bart and Cone, 2012) and the different reconstructions for the Weddell Sea sector according to Scenarios A and B presented here.

Recommendations for future research
Given the very limited amount of the currently available terrestrial and marine geomorphological and chronological data for the Weddell Sea sector, new collection of data and samples and their full exploitation together with that of the already existing material are urgently required.Only such a strategy will allow reconstruction of the ice-sheet history in the Weddell Sea sector during the last glacial cycle with some certainty.Apart from the acquisition of new geomorphological and chronological data from some key areas, such as the Pensacola Mountains and the middle and outer shelf parts of the Hughes, Ronne and Filchner troughs, as well as from terrestrial sites, where pilot studies have been carried out, such as the Heimefrontfjella in western Dronning Maud Land, a more detailed analysis of new and existing samples and data seems to be necessary.For example, any new 14 C dates on marine sediment cores from Filchner Trough may help to verify or rule out the existence of the proposed hiatus from w34.0 to w18.5 cal ka BP (Fig. 5b) and thus to test the validity of Scenario B, while any new exposure dates on erratics collected from above the trimlines interpreted to indicate the maximum ice-sheet elevations at the LGM may help to test the validity of Scenario A and/or short-lived thickening events.Swath bathymetry maps covering core locations, where over-compacted glaciomarine sediments were recovered, have the potential to show bedforms that will help to clarify, whether the observed over-consolidation was caused by icebergscouring or ice-sheet overriding.This important distinction is almost impossible on the basis of sedimentological data and acoustic subbottom profiles alone (e.g.Fütterer and Melles, 1990;Melles and 1993).
Novel and refined analytical approaches are required to distinguish subglacial from glaciomarine facies in sediment cores and to evaluate the reliability of the 14 C dates obtained from the Weddell Sea shelf.One such method was proposed by Stolldorf et al. (2012) who used subtle grain-size changes to distinguish unsorted and poorly sorted subglacial deposits from better sorted glaciomarine sediments.If available, acoustic subbottom and seismic profiles from core locations should always be considered for the stratigraphic interpretation of sedimentary units and the 14 C dates obtained from these units.For example, core 3-3-1 from the eastern flank of the inner shelf part of Filchner Trough (Fig. 4) recovered proximal glaciomarine sediments at its core top, which provided a very old age of 47.7 cal ka BP (Fig. 6; Stolldorf et al., 2012) and may, in fact, be much older (cf.Takada et al., 2003).When the core site is projected onto a nearby seismic profile (Fig. 5 in Anderson et al., 1983or Fig. 3 in Anderson et al., 1991), it becomes clear that the core probably recovered sediments from the westward dipping reflectors described by Elverhøi and Maisey (1983).This observation alludes to the possibility that the 14 C date was obtained from calcareous foraminifera tests that had been reworked from the old dipping strata.Further improvement of the reliability of the 14 C ages from the marine sediments may be achieved by dating calcareous benthic foraminifera tests from obviously 'unmixed' assemblages typical for modern glaciomarine environments (Stolldorf et al., 2012) and removing possibly reworked foraminifera tests from the samples before AMS dating (Bart and Cone, 2012).Furthermore, the suitability of d 18 O records from foraminifera-bearing sediments on the outer WSE shelf for oxygen isotope stratigraphy (e.g. at sites PS1609 and PS1420; Hillenbrand et al., 2012) should be evaluated by obtaining down-core AMS 14 C dates from these cores.
Future research should also focus on testing the hypotheses developed by Hillenbrand et al. (2012) and Larter et al. (2012) for reconciling an LGM ice-sheet advance to the shelf break within Filchner Trough with the limited thickening in the WSE hinterland documented by the terrestrial evidence (Fogwill et al., 2004;Bentley et al., 2010;Hein et al., 2011).For example, ice-sheet model runs could explore the plausibility of bed conditions required for an ice plain to extend all the way from the modern Filchner Ice Shelf front to the shelf break.Provenance studies on Holocene glaciomarine sediments and pre-Holocene subglacial tills from inner shelf cores recovered to the east and west of Berkner Island (Fig. 3) should be carried out to detect possible flowswitches of Foundation Ice Stream in the past.Finally, qualitative insights from the past ice-flow changes in the Weddell Sea sector should be utilised to estimate the risk of possible future deglaciation in this and other sectors of the Antarctic Ice Sheet and the magnitude of associated sea-level rise.For example, the palaeorecord from the WSE can be used for validating the sensitivity of ice-sheet retreat to reverse bed gradients (e.g.Schoof, 2007;Katz and Worster, 2010;Jamieson et al., 2012), and the outcome can be implemented in numerical ice-sheet models.

Conclusions
Even though the data base of marine and terrestrial geological records from the Weddell Sea sector and its hinterland has significantly increased over the last few years, the LGM to Holocene glacial history of this sector is still poorly known when compared to other sectors of the Antarctic Ice Sheet.Subglacial bedforms recorded in high-resolution bathymetric maps and seismic profiles from the Weddell Sea continental shelf document that the grounding lines of the WAIS and EAIS had advanced across the shelf in the past, probably during the Late Pleistocene.The glacial geomorphological record in the hinterland of the Weddell Sea sector, surface exposure ages derived from cosmogenic nuclides and changes in ice sheet-thickness archived in the Berkner Island ice core are best explained by no or only minor thickening of the WAIS and EAIS during the last glacial period, suggesting that ice did not ground in the deepest parts of the palaeo-ice stream troughs north of the FilchnereRonne Ice Shelf.Available radiocarbon dates on calcareous microfossils from sediment cores recovered from the continental shelf and uppermost slope can be interpreted to indicate that the last advance of grounded ice occurred either before the last glacial period or at the LGM.This contradicting interpretation originates from (i) the low number of available ages, (ii) a lack of the geomorphological and seismostratigraphic context for most of the dated cores, (iii) the problem of a reliable distinction between subglacial facies and glaciomarine facies, (iv) our inability to clearly identify glaciomarine sediments, which were over-compacted after their deposition by an overriding grounded ice sheet as opposed to an iceberg, (v) a lack of information, whether 14 C dates were obtained from autochthonous or reworked allochthonous microfossils, and (vi) the difficulty of evaluating the reliability of ages obtained from sediments recovered on the continental slope in constraining the timing of grounded ice-sheet advance/retreat on the adjacent shelf.Grounded ice-sheet advance onto the outer shelf of the Weddell Sea during the last glacial period and no/minor ice-sheet thickening in its hinterland can be reconciled by assuming a short-term advance of ice with a thickness close to floatation and a very low slope gradient and ice-flow changes in the drainage basins of the Filchner and Ronne ice shelves.All LGM-Holocene reconstructions for the Weddell Sea sector conclude (i) time-transgressive changes in the various drainage basins of the WAIS and EAIS, (ii) no or only minor ice-sheet build-up at the LGM and (iii) no significant contribution of post-LGM ice-sheet melting to global meltwater pulses during the last deglaciation.

Fig
Fig. 1.Overview map over the Weddell Sea sector with shelf bathymetry and ice-sheet surface elevation (in metres above sea level) according to Bedmap2 (Fretwell et al., 2013) and the main physiographic and glaciological features.Ice shelves are displayed in light blue shading.Inset map shows the Weddell Sea sector outlined by the red line within the context of Antarctica, with ice shelves being displayed in grey shading (APIS: Antarctic Peninsula Ice Sheet, EAIS: East Antarctic Ice Sheet, WAIS: West Antarctic Ice Sheet).

Fig. 2 .
Fig. 2. Locations of subglacial bedforms in the Weddell Sea sector mapped by high-resolution bathymetry.The circles highlight the areas for which data have been published by Kristoffersen et al. (2000b), Larter et al. (2012) and Stolldorf et al. (2012).

Fig. 5 .
Fig.5.Conventional radiocarbon dates versus calibrated (or corrected) radiocarbon ages from the cores displayed in Fig.4(for details see Supplementary Table2).5a: All ages grouped for different regions (note: Brunt Ice Shelf dates are exclusively from core 3-7-1, Quar Ice Shelf dates are exclusively from core 3-17-1 and the Fimbul Ice Shelf date is from core 206).Minimum ages are marked with arrows, and dates from cores recovered on the continental slope are underscored.Light grey shading indicates the time span of a

Fig. 8 .
Fig. 8. Grounded ice-sheet extent in the Weddell Sea sector at 20 ka according to Scenario A (GL: grounding line).

Fig. 9 .
Fig. 9. Grounded ice-sheet extent in the Weddell Sea sector at 15 ka according to Scenario A.

Fig. 10 .
Fig.10.Grounded ice-sheet extent in the Weddell Sea sector at 10 ka according to Scenario A.

Fig. 11 .
Fig. 11.Grounded ice-sheet extent in the Weddell Sea sector at 5 ka according to Scenario A.

Fig. 13 .
Fig. 13.Grounded ice-sheet extent in the Weddell Sea sector at 20 cal ka BP according to Scenario B.

Fig. 14 .
Fig. 14.Grounded ice-sheet extent in the Weddell Sea sector at 15 cal ka BP according to Scenario B.

Fig. 15 .
Fig. 15.Grounded ice-sheet extent in the Weddell Sea sector at 10 cal ka BP according to Scenario B.

Fig. 16 .
Fig. 16.Grounded ice-sheet extent in the Weddell Sea sector at 5 cal ka BP according to Scenario B.