Edinburgh Research Explorer Deglacial history of the Pensacola Mountains, Antarctica from glacial geomorphology and cosmogenic nuclide surface exposure dating.

The retreat history of the Antarctic Ice Sheet is important for understanding rapid deglaciation, as well as to constrain numerical ice sheet models and ice loading models required for glacial isostatic adjustment modelling. There is particular debate about the extent of grounded ice in the Weddell Sea embayment at the Last Glacial Maximum, and its subsequent deglacial history. Here we provide a new dataset of geomorphological observations and cosmogenic nuclide surface exposure ages of erratic samples that constrain the deglacial history of the Pensacola Mountains, adjacent to the present day Foundation Ice Stream and Academy Glacier in the southern Weddell Sea embayment. We show there is evidence of at least two glaciations, the ﬁ rst of which was relatively old and warm-based, and a more recent cold-based glaciation. During the most recent glaciation ice thickened by at least 450 m in the Williams Hills and at least 380 m on Mt Bragg. Progressive thinning from these sites was well underway by 10 ka BP and ice reached present levels by 2.5 ka BP, and is broadly similar to the relatively modest thinning histories in the southern Ellsworth Mountains. The thinning history is consistent with, but does not mandate, a Late Holocene retreat of the grounding line to a smaller-than-present con ﬁ guration, as has been recently hypothesized based on ice sheet and glacial isostatic modelling. The data also show that clasts with complex exposure histories are pervasive and that clast recycling is highly site-dependent. These new data provide constraints on a reconstruction of the retreat history of the formerly-expanded Foundation Ice Stream, derived using a numerical ﬂ owband model.


a b s t r a c t
The retreat history of the Antarctic Ice Sheet is important for understanding rapid deglaciation, as well as to constrain numerical ice sheet models and ice loading models required for glacial isostatic adjustment modelling. There is particular debate about the extent of grounded ice in the Weddell Sea embayment at the Last Glacial Maximum, and its subsequent deglacial history. Here we provide a new dataset of geomorphological observations and cosmogenic nuclide surface exposure ages of erratic samples that constrain the deglacial history of the Pensacola Mountains, adjacent to the present day Foundation Ice Stream and Academy Glacier in the southern Weddell Sea embayment. We show there is evidence of at least two glaciations, the first of which was relatively old and warm-based, and a more recent cold-based glaciation. During the most recent glaciation ice thickened by at least 450 m in the Williams Hills and at least 380 m on Mt Bragg. Progressive thinning from these sites was well underway by 10 ka BP and ice reached present levels by 2.5 ka BP, and is broadly similar to the relatively modest thinning histories in the southern Ellsworth Mountains. The thinning history is consistent with, but does not mandate, a Late Holocene retreat of the grounding line to a smaller-than-present configuration, as has been recently hypothesized based on ice sheet and glacial isostatic modelling. The data also show that clasts with complex exposure histories are pervasive and that clast recycling is highly site-dependent. These new data provide constraints on a reconstruction of the retreat history of the formerly-expanded Foundation Ice Stream, derived using a numerical flowband model.

Background and rationale
The Antarctic Ice Sheet is the largest potential contributor to future sea-level rise. It is currently losing mass Shepherd et al., 2012) and some studies suggest that the rate of mass loss is accelerating (Harig and Simons, 2015;Velicogna et al., 2014;Williams et al., 2014). Understanding the past history of the ice sheet is important because: it can inform how the ice sheet has responded to past environmental changes, and record its trajectory preceding the observational record; it allows us to test ice sheet models by hindcasting; and it provides us with inputs for models of glacial isostatic adjustment (GIA), which are needed to interpret satellite gravimetric measurements of ice mass loss (Bentley, 2010).
The Weddell Sea sector of the ice sheet has not seen the same level of attention as other areas such as the Amundsen Sea and Ross Sea. This is despite some studies suggesting that the area is particularly susceptible to ice shelf thinning (Hellmer et al., 2012) and grounding line retreat (Ross et al., 2012;Wright et al., 2014). Part of this sensitivity derives from the fact that the southern part of the area is overdeepened, particularly in the east, where several subglacial basins are comparable in depth to other deep but grounded parts of the West Antarctic Ice Sheet, for example, along the Amundsen Sea coast. The largest trough in the Weddell Sea, the Foundation-Thiel Trough, is 1300e1500 m deep and extends northsouth for >1000 km, right across the Weddell Sea continental shelf to the shelf break (Fig. 1). Its position, size and extent mean that it was once occupied by a major ice stream (the "Thiel Trough Ice Stream") draining ice through an area east of Berkner Island. This ice stream is likely to have exerted a key control on regional ice elevation.
At present there are two alternative models for the Last Glacial Maximum (LGM) extent of ice in the Weddell Sea embayment (Bentley et al., 2014;. The first, based largely on marine geological evidence is an extensive model with ice grounded over the outer continental shelf Larter et al., 2012). The implication of this model is that the Thiel Trough Ice Stream would be grounded as far as the shelf break. The second, based largely on terrestrial evidence of minor elevational change of the ice sheet at the LGM Hein et al., 2011;Mulvaney et al., 2007) is a restricted model with the grounding line of the Thiel Trough Ice Stream confined to the midto inner-shelf Le Brocq et al., 2011;Whitehouse et al., 2012).
In addition to the debate on ice sheet extent, the timing of postglacial thinning in the Weddell Sea is not yet well understood. This hampers the development of ice loading models for GIA modelling, and the understanding of recent (Late Holocene) change in the region. For example, although most studies have assumed a simple retreat from LGM to present, a recent study has demonstrated that a Late Holocene re-advance of the ice sheet may explain some formerly puzzling observations from GPS and glaciological surveys (Bradley et al., 2015). Such a readvance has important implications for our understanding of ice sheet stability, in that it implies that some grounding lines on inward dipping bedrock beds may be advancing (Bradley et al., 2015). Such 'unstable advance' has been suggested on theoretical grounds (Schoof, 2007). We require further observational data on ice sheet retreat timing to test the validity of such ideas.
Our aim in this paper is to determine former ice sheet extent and elevation change adjacent to the southern extension of the Thiel Trough Ice Stream (Fig. 1). The glacial geology of this area contains a record of ice thickness change that can yield information on the extent of ice along the Foundation-Thiel Trough, and the timing of its thinning from the local last glacial maximum (LLGM) (Clark et al., 2009) to its present configuration. This paper describes the geomorphological evidence of former thicker ice levels in the  (Fretwell et al., 2013). Inset shows location within the southern Weddell Sea embayment. BI¼Berkner Island; EM ¼ Ellsworth Mountains; FTT¼Foundation-Thiel Trough; IIS¼Institute Ice Stream; SR¼Shackleton Range.
northern Pensacola Mountains and then provides a chronology of deglacial ice sheet thinning using cosmogenic nuclide surface exposure dating. We also compare our results to a study of similar sites in the Pensacola Mountains (Balco et al., 2016). A companion paper to this one (Whitehouse et al., 2016) uses a glaciological model to explore the implications of the data presented both here and in Balco et al. (2016) with regard to the former extent of ice in the Foundation-Thiel Trough.

Study area and methods
The study area is the northern Pensacola Mountains, located at the southern extension of the Foundation-Thiel Trough (Fig. 1). The Pensacola Mountains are split into the Neptune Range in the north, and the Patuxent Range in the south. The Foundation Ice Stream, fed by the tributary Academy Glacier, flows south to north through the area and becomes afloat close to the Schmidt Hills. The Foundation Ice Stream is by far the largest West Antarctic Ice Sheet outlet in the south-eastern Weddell Sea: indeed today it contributes more ice to the Weddell embayment than any other outlet except the Evans Ice Stream (Joughin and Bamber, 2005). The margins of the Academy Glacier and Foundation Ice Stream have several blue-ice areas with abundant supraglacial morainic debris, and this coupled to the presence of numerous nunataks along their flanks means that there is a rich record of past ice sheet elevational changes preserved.
We studied the geomorphology and sampled altitudinal transects of erratic sandstone clasts at several sites : Schmidt Hills,Williams Hills,Mt. Bragg and Mt. Harper,and Thomas Hills (Figs. 1 and 2.). Striation directions are corrected to true north using declination values that vary across the study area from 23 to 31 east of north.

Sampling of erratics and cosmogenic nuclide analysis
The geology of the Pensacola Mountains is dominated by quartzbearing lithologies: dominantly a mix of folded sandstones, mudstones, conglomerates and limestones, with a more minor series of lava flows and pillow lavas in the western part (Schmidt et al., 1978) and so a large proportion of erratics are suitable for exposure dating using 10 Be and 26 Al contained in quartz. The presence of abundant erratics left behind by a thinning ice sheet was noted by Schmidt et al. (1978) who described erratics on 'most rock slopes' up to 1000 m above present day ice. Erratic samples were taken whole or sampled from larger boulders with a hammer and chisel. Our particular focus was in understanding the deglacial (thinning) history of the different areas studied and we sampled erratics along altitudinal transects for this reason. The sampling strategy for cosmogenic nuclides was designed to reduce the chance of nuclide inheritance, and exclude the possibility of nuclide loss through erosion. Specifically, we sampled erratics that were perched on bedrock, felsenmeer or drift surfaces, and avoided any samples that were embedded in drift, or sitting at the base of slopes in order to minimise problems of post-depositional movement and selfshielding. Because our focus is on the most recent deglacial thinning we selected the freshest erratics, and so avoided erratics with weathering such as ventification, tafoni, spallation, weathering rinds or patina. We targeted sub-glacially derived clasts with striated surfaces and sub-angular to sub-rounded shapes. Topographic shielding was measured using an abney level and compass.
The sample and cosmogenic nuclide data are presented in Tables 1 and 2, and Fig. 2. All exposure ages discussed are based on 10 Be ages because the production rate is better constrained; 26 Al was used to provide a check on a range of selected samples from different sites and in particular to determine, within the errors of the approach, that the younger ages were concordant within errorand therefore less likely to be reworked. In the case of the older ages (>100 ka) we used 26 Al analysis of a sub-set of samples to determine whether they had experienced periods of burial, and were therefore more likely to record complex exposure histories or reworking over more than one interval of ice sheet thickening and thinning.

Laboratory and analytical techniques
Whole rock samples were crushed and sieved to obtain the 250e710 mm fraction. Be and Al were selectively extracted from the quartz component of the whole-rock sample at the University of Edinburgh's Cosmogenic Nuclide Laboratory following established methods (Bierman et al., 2002;Kohl and Nishiizumi, 1992). 10 Be/ 9 Be and 26 Al/ 27 Al ratios were measured in 20e30 g of quartz at the Scottish Universities Environmental Research Centre (SUERC) Accelerator Mass Spectrometry (AMS) Laboratory in East Kilbride, UK. Measurements are normalised to the NIST SRM-4325 Be standard material with a revised (Nishiizumi et al., 2007) nominal 10 Be/ 9 Be of 2.79 Â 10 À11 , and the Purdue Z92-0222 Al standard material with a nominal 26 Al/ 27 Al of 4.11 Â 10 À11 , which agrees with the Al standard material of Nishiizumi (2004) with half-life of 0.705 Ma (Xu et al., 2010). SUERC 10 Be-AMS is insensitive to 10 B interference (Xu et al., 2013) and the interferences to 26 Al detection are well characterised . Process blanks (n ¼ 19) and samples were both spiked with 250 mg 9 Be carrier (Scharlau Be carrier, 1000 mg/l, density 1.02 g/ml). Blanks were spiked with 1.5 mg 27 Al carrier (Fischer Al carrier, 1000 ppm) and samples were spiked with up to 1.5 mg 27 Al carrier (the latter value varied depending on the native Al-content of the sample). Blanks range from 5 Â 10 À15 e 3 Â 10 À14 [ 10 Be/ 9 Be] (less than 5% of total 10 Be atoms in all but the youngest samples); and 2 Â 10 À15 e 2 Â 10 À14 [ 26 Al/ 27 Al] (less than 2% of total 26 Al atoms in all but the youngest samples). Concentrations in Table 1 are corrected for process blanks; uncertainties include propagated AMS sample/lab-blank uncertainty, a 2% carrier mass uncertainty, and a 3% stable 27 Al measurement (ICP-OES) uncertainty.

Exposure age calculations
For exposure age calculations we used default settings in Version 2.0 of the CRONUScalc programme . The CRONUS-Earth production rates  with the nuclide-dependent scaling of Lifton-Sato-Dunai (Lifton et al., 2014) were used to calculate the ages presented in the paper. Sea level and high latitude production rates are 3.92 ± 0.31 atoms g À1 a À1 for 10 Be and 28.5 ± 3.1 atoms g À1 a À1 for 26 Al. The alternative use of Lal/Stone (Lal, 1991;Stone, 2000) scaling does not change the conclusions of the paper despite the approximately 3% and 6% older exposure ages for 10 Be and 26 Al, respectively. Rock density is assumed 2.7 g cm À3 and the attenuation length used is 153 ± 10 g cm À2 . No corrections are made for rock surface erosion or snow cover and thus exposure ages are minima. Finally, we make no attempt to account for the relatively minor production-rate variations that would be caused by elevation changes associated with glacial isostatic adjustment of the massif through time (Stone, 2000;Suganuma et al., 2014). All ages are apparent exposure ages, which make the assumptions that all exposure has been achieved in a single episode at the current location of the sample, that the sample has not been covered by snow or sediment, and has experienced zero erosion.

Results
From the sediments, moraines and erosional landforms preserved in the Neptune and Patuxent Ranges we have identified evidence for ice sheet expansion (thickening) and retreat (thinning). The area shows abundant evidence of thicker ice, with sandstone erratics strewn over most surfaces visited (Fig. 3a). A small number of features provide evidence of former ice limits (e.g., Fig. 3b), and in some cases show complex and repeated interaction of different ice masses (Fig. 3c). In all cases we sampled erratics over altitudinal profiles from the summits to the present ice margin. The exposure ages of these erratics yields a thinning history for the Foundation Ice Stream and Academy Glacier, which is presented after a description of the glacial geomorphology of the sample sites.

Geomorphology
The erratic lithologies in the Neptune Range are dominated by Dover Sandstone (white to yellow, medium to coarse-grained, finely bedded to massive, quartz-rich) (Schmidt et al., 1978). Erratics are found on at least the lower parts of almost every nunatak visited, and in many cases are abundant. There are fewer Dover Sandstone erratics on nunataks in the Patuxent Range: here erratic lithologies tended to be darker-coloured, harder and finer-grained sandstones. Although not generally sampled, other erratic lithologies seen on nunataks include sandstone with conglomeratic layers (including granodiorite pebbles), basalt (fine-grained, weathered, rounded, ventifacted), dolerite-gabbro (coarse, angular blocks), granodiorite (pebble and small cobbles only), granite, gneiss and limestone. Some sandstone erratics show one or more of granular disaggregation, spalling and tafoni. Whilst overall across the study area the erratics are of diverse lithologies, we focused sampling on the Dover Sandstone clasts as they were the commonest clasts at most localities and had a fresher appearance than some of the basalt (ventifacted) or other sandstone (tafoni weathering) clasts. A few Dover Sandstone clasts had a darker yellow weathered outer surface; these were avoided in sampling.
Many of the slopes in the area are mantled by a veneer of debris, much of it made up of a layer (>30 cm thick) of pale yellow-brown (fresh) to khaki brown (weathered) silty diamict, with locallyderived clasts, and open pore spaces up to a few mm across. The diamict has a hard crust and is also partially-cemented. Surface salt encrustation of this diamict is common. Some slopes have small terraces, 2e4 m across with risers of 30e50 cm. Exposures of the diamict are common in these risers or in frost cracks that cut through the terracing. Such features are particularly common downslope of snow patches, probably due to availability of liquid water. This diamict underlies talus slopes, desert pavement cover and many of the erratics sampled by us, but it is not ubiquitous and some of the nunataks, especially in the Schmidt Hills and parts of the Thomas Hills, do not have exposures of a diamict. In these locations there is a felsenmeer of locally-derived bedrock on lowangle slopes.
Many nunataks have areas of striated bedrock. For example, in Williams Hills, the bedrock on Pillow Knob (NE end) is well-striated with two generations of striations. The first is finely-spaced, mm-   scale, regular and extends across bedrock surfaces for 10s of cm (Fig. 3d). These have a distinct weathering sheen or patina over the surface. Their trend is 050e230 . These striations are found commonly on surfaces all over the upper ridge and on the summit. The second, cross-cutting set of striations are irregular, cm-scale 'gouges' (up to 1 cm wide, but usually 2e5 mm wide) that cut into the weathering patina, and are oriented approximately 100e280 . They incise (1e2 mm) the patina and are themselves not weathered (Fig. 3d). These gouges are not continuous (unlike the earlier striation set) and can be identified as individual gouges up to 12e15 cm long, but usually only up to 5 cm. We follow Atkins et al. (2002) and Bentley et al. (2006) in interpreting the two sets of striations as evidence for an early warm-based ice sheet (closelyspaced, polished, weathered striations) followed by a later coldbased ice sheet (unweathered, less regular gouges). However, at one location on the NE side of Pillow Knob both sets appear more closely-spaced, regular and equally weathered, and so we cannot rule out that the second ice sheet was warm-based, at least in parts. Bedrock throughout the area is usually weathered with a dark red surface patina. Locally-derived talus is common on steeper slopes and in places this overlies either the bedrock or the silty diamict. Both the weathered bedrock and the silty diamict extend right down to the present ice surface.

Schmidt Hills
The Schmidt Hills are a line of nunataks along the east side of the Foundation Ice Stream, close to its grounding line (Fig. 2a). Spurs from these nunataks extend west down to the ice margin, at 200e300 m asl. Erratics are rare on all nunataks in the Schmidt Hills. For example on Dimmo Peak, only two true erratics (one sandstone, one granitic) were found in the 400 m elevational range between the ice edge on the SE side of the nunatak and the summit, and both were found at a relatively low elevation. At~860 m on the W ridge of Mt. Nervo there is a local concentration of sandstone and conglomerate erratics that are of sufficient lithological and colour/ weathering diversity for us to be confident that they are not originally derived from a single clast.
No striations were found on summits and virtually all the bedrock surfaces have weathered surfaces (Fig. 3e). Lower down (below~600 m asl) there are many striated surfaces on a spur W of Dimmo Peak, with trend 036e216 to 056e236 . Lower parts of the spurs in Schmidt Hills also appear more rounded than the summits. Diamict occurs over parts of the nunataks but is more commonly exposed on the lower spurs, where the debris cover is modified by surficial mass movement. In several locations on the S side of Mt. Nervo (e.g., 83 14.494 S, 057 56.440 W, 975 m) there are bedrock fractures that have been filled by diamict.
The degree of weathering increases upwards on Mt. Nervo and Mt. Coulter such that near the summits there are features such as tafoni weathering, spalling, and thick weathering rinds. The ridge crests themselves are weathered such that cleaved sedimentary bedrock forms soft friable surfaces, whilst basalts have weathered into nodules. Bowl-shaped depressions a few 100 m across on the flanks of these nunataks contain extensive periglaciated (polygons, stripes, terraces, deep frost cracks) deposits containing pale yellow silty diamict, >30 cm thick. At the mouth of the bowl on the SW flank of Mt. Nervo there are prominent morainic deposits forming a mantle of debris over linear bedrock features sub-parallel to the Foundation Ice Stream margin.

Williams Hills
The Williams Hills have abundant erratics at all elevations, lying on the yellow silty diamict, on felsenmeer, on summit plateaux, or on bedrock (Fig. 2b). The vast majority of these are erratics of Dover sandstone. There is also evidence of ice transport of local lithologies (translocation of clasts from host outcrops on flat surfaces or uphill). Weathered, pervasively striated bedrock shows 'old' overriding (warm-based) glaciations trending 035e215 (Mt. Hobbs) or 050e230 (Pillow Knob). Fresher striations trend 100e280 and are scattered and more gouge-like (cold-based). The yellow diamict is ubiquitous over all surfaces, from the modern ice edge to the highest summit (Mt. Hobbs).

Mt. Harper and Mt. Bragg
Along the northern margin of the Academy Glacier there are a series of nunataks that have a geological record of glacier fluctuations on their southern flanks. The westernmost of these is Mt. Harper, with Mt. Bragg to the east (Fig. 2c). There are extensive blue ice areas with supraglacial moraines along this margin of the Academy Glacier.
Mt. Harper has a series of low knolls directly south of the main nunatak. These are mantled in a veneer of drift that is composed mainly of the local basaltic lithology but with abundant freshly weathered sandstone clasts. There is an underlying sandy diamict present but this is generally confined to pockets in bedrock.
The most prominent glacial feature on Mt. Bragg is a distinctive line of boulders located about 200 m above the present ice surface and which can be traced wrapping continuously around the slopes of Mt. Bragg for up to 2 km (Fig. 3b). The boulder moraine is composed of off-white to yellow erratics. To the east (up-ice) it cannot be traced beyond a talus slope directly below the summit of Mt. Bragg. To the west (down-ice) the limit descends parallel to the modern glacier margin below. There is no apparent difference in weathering of bedrock on either side of the boulder limit, and a small number of fresh sandstone erratics are found on the slopes immediately above it. Bedrock on Mt. Bragg has a reddish-brown to purple weathering sheen, and in places has polish and striations. The striations are mm-scale, finely-spaced, trend 077e257 and are widely distributed.
Higher up, the west summit (1200 m) of Mt. Bragg is highly weathered with an orange-brown patina and ubiquitous frostshattering of bedrock. The number of erratics mantling the surface is much less than lower on the nunatak, and the erratics themselves are more weathered than those lower down. In particular volcanic erratic clasts have polish, weathering patina and ventifaction.
On the south side of the Academy Glacier we sampled former lake shorelines in the Mt. Lowry region: these may record ice sheet thinning (see Hodgson and Bentley (2013) for full description). The lake shorelines are covered by sub-fossil microbial mats that have desiccated into grey papery deposits that are found preserved under cobbles and boulders.

Thomas Hills
The Thomas Hills comprise a series of NW-SE trending ridges that flank the east side of the Foundation Ice Stream, upstream of its confluence with the Academy Glacier (Fig. 2d). Here we report the geomorphology from the spurs 1 of Mt. Yarborough, Sugden Ridge, Clapperton Ridge, and Bentley Peak.

Mt. Yarborough
Mt. Yarborough sits towards the SW end of the Thomas Hills. Most of the nunatak is mantled in morainic debris and numerous erratics. Striations near the summit of Mt. Yarborough and close to 1 Only a few of the spurs in the Thomas Hills are named on the 1960s USGS map and so in 2011 the UK Antarctic Place Names Committee named the previously unnamed spurs. the ice margin trend 119e299 , and lower slopes are mantled in a silty white diamict. Erratics are numerous and were sampled from the summit down to the edge of the Foundation Ice Stream.

Sugden Ridge
The glacial geomorphology preserved on this ridge records the interaction between the Foundation Ice Stream and more locallysourced ice spilling over from the MacNamara Glacier to the south-east. The Foundation Ice Stream is higher than the area between the spurs of the Thomas Hills and so lobes of the Foundation Ice Stream flow downhill into the embayments, each of them terminating in an area of blue-ice, and hence net ablation. At the northern end of the ridge there is a small lobe of Foundation Ice Stream ice that flows down to the SE into the valley between Sugden and Clapperton Ridges, and former positions of this lobe are marked by a series of lateral moraines on the NE flank of Sugden Ridge. Each of these moraines consists of distinct lines of pale sandstone boulders, except for the lowest moraine close to the present Foundation Ice Stream margin, which constitutes a substantial ridge of diamict a few metres high. The moraines range in elevation from just a few metres above the present ice margin (370 m asl) to 150 m above the present ice, and have gradients consistent with formerly expanded Foundation Ice Stream ice. Similarly, former positions of a lobe derived from the MacNamara Glacier that spilled over the col between the two ridges and flowed NW towards the Foundation Ice Stream are also recorded on the same slope. Some of these moraine sequences intersect and overlap (Fig. 3c). Finally there is an area of patterned ground e probably representing former snowpatches e that descends NE-wards from the middle of the ridge forming Sugden Ridge (Fig. 3c). The northern ridge summit above the moraines is striated whilst the southern parts of the ridge are mantled in silty diamict with a widespread surface salt encrustation.

Clapperton Ridge
The sandstone bedrock of Clapperton Ridge is deeply weathered, striated in places and an overlying silty diamict is widespread, including overlying striated bedrock. Striations in a col on the ridge trend 129e309 , and are of the extensive, finely-spaced (warmbased) type. No scattered striations or gouges (cold-based) were seen cutting them.

Bentley Peak
This minor peak lies at the NE end of Thomas Hills. Striae are visible at all altitudes on this nunatak, and trend 130e310 . We sampled a transect of erratics and quartz bedrock veins from the summit (715 m) down to the Foundation Ice Stream margin at 490 m. The lowest part of the nunatak is mantled in a thick diamict that has an unweathered appearance and contains fresh, unweathered erratics.

Geomorphological interpretation
There is a consistent pattern of glacial geomorphology in the Pensacola Mountains. Specifically there is evidence of at least two major glacial phases or configurations: the first of these was a warm-based glaciation that striated bedrock across the field area and flowed oblique to present-day ice flow. For example, in the Williams Hills, modern ice flow is northwards, but striations show former flow aligned NE-SW. This may have been the same glacial phase that deposited a near-ubiquitous silty yellow diamict across the field area and which occurs directly over many of the striated surfaces. The weathering of the striated surfaces and of the diamict itself suggests that this glaciation was old, and we cannot rule out that the diamict is an equivalent of the Sirius Group diamictons or other several-million year old diamicts found elsewhere in the Transantarctic Mountains. A second glacial phase constituted a cold-based ice sheet expansion that over-rode and striated several summits, for example, in the Williams Hills. There is little other evidence of erosion, and ice flowed obliquely to the earlier phase. This second phase is associated with the deposition of erratics across almost all surfaces, as well as a fresh diamict in a few places. Thinning from this glaciation left boulder moraines recording former ice margins along the side of Mt. Bragg and Thomas Hills, and erratics located from summits to present ice elevations. During this period, there were interactions between an expanded Foundation Ice Stream/Academy Glacier and more locally-sourced ice, such as is recorded in the Thomas Hills.
The Schmidt Hills are anomalous in this regional pattern in that they contain few erratics. Where they do occur they tend to cluster in highly localized concentrations, such as on the W ridge of Mt. Nervo. In the Schmidt Hills and in some of the Thomas Hills there is a consistent pattern of summits being significantly more weathered than lower reaches.

Cosmogenic nuclide exposure ages
We sampled and analysed 105 erratics to derive their exposure ages using cosmogenic nuclides. The exposure ages ranged from 2.5 ka to 3 Ma but with significant clustering of ages <10 ka. When plotted against their height above present day ice the younger ages in two of the four areas studied show clear thinning trajectories in the Holocene (Table 2, Fig. 4): Williams Hills and the closely adjacent sites of Mt. Bragg and Mt. Harper. Other sites show more complex histories of ice sheet fluctuations (Fig. 4). In all four areas the analysed samples include a component of older material, a common feature of cosmogenic isotope-derived chronologies of ice sheet thinning in Antarctica (Balco, 2011;Balco et al., 2016;Bentley et al., 2006Bentley et al., , 2010Hein et al., 2011Hein et al., , 2014Mackintosh et al., 2011) caused by individual clasts being recycled or overridden during ice sheet thickening-thinning cycles. In the case of the subset of older samples analysed for 26 Al these all yielded discordant ages implying a complex exposure history with at least one period of burial.
3.5.2.1. Schmidt Hills. In the Schmidt Hills, of the eight erratic samples analysed, there are no samples younger than 236 ka, and most are substantially older, up to 3 Ma (Fig. 4a). Whilst there may be a partial record of fluctuations over successive glacial cycles the site does not provide any constraints on deglaciation from the LLGM.

Williams Hills.
Of the 20 cosmogenic exposure dates 17 yield Holocene ages and show that thinning of the ice sheet occurred from at least 450 m above present ice elevations. This thinning was progressive from the Early to mid-Holocene and reached 50 m above present-day ice by 5 ka. Notably there are no samples younger than 5 ka (Fig. 4b and c).
3.5.2.3. Mt. Bragg and Mt. Harper. We analysed 35 samples from Mt Bragg and Harper, and these range from 2.5 ka to 395 ka but with a significant number, especially at Mt Harper, yielding Holocene ages. The data show that the ice sheet thinned from at least 380 m above present ice with the majority of the samples deposited below 240 m above present ice (Fig. 4d and e) from the early to mid-Holocene. Thinning may have been underway at 17.6 ka but was certainly occurring progressively by 7.9 ka. The youngest age of 2.5 ka, close to present-day ice shows that present ice levels were reached by 2.5 ka. Samples from the boulder moraine on the sides   , LSD scaling (Lifton et al., 2014), no erosion correction; attenuation length 153 ± 10 g cm 2 ; 26 Al and 10 Be production rates from CRONUS-Earth Project . b (int) ¼ Internal uncertainties; includes only concentration uncertainties based on lab/AMS measurements. c (ext) ¼ External uncertainties; includes internal uncertainties plus scaling and production rate uncertainties. d (full) ¼ Full uncertainties; includes external uncertainties plus uncertainties on thickness (0.5 cm), pressure (10 hPa), attenuation length (10 g cm-2), and density (0.05 g cm À3 ). e Sample elevations are normalised to the elevation of the ice surface at the foot of the spur or peak on which the sample lies, and so the plots in Fig. 4 are in terms of elevation above present ice.  of Mt. Bragg (~200 m above present ice) yielded a mixed population of ages (12, 37, 41, 74, 161, 206 ka Al exposure ages on a subset of samples show discordant ages, implying a complex exposure history, and thus confirm this view of at least one period of burial. 3.5.2.5. Sugden Ridge. Sugden Ridge (Figs. 3c and 4g) yields a complex set of exposure ages on the cross-cutting moraines, and does not provide a clear thinning history for the Foundation Ice Stream. Each moraine has yielded a significant range of ages. The highest parts of Sugden Ridge yield pre-LGM exposure ages of 96 ka (peak) and 563 ka (col). The highest Foundation Ice Stream moraine yields ages of 51.1 and 32.5 ka. It is notable that all of the moraines below this yield mixed ages but that ages in the range 28e33 ka occur in every moraine. The most extensive and distinct moraine yields one age of 12.8 ka, which is the youngest age anywhere on this spur. The lowest elevation moraine provides mixed ages between 380 and 28.7 ka.

Discussion
The data reported here show that the Foundation Ice Stream and its tributary Academy Glacier thickened by 100s of metres during the LLGM. The data do not constrain the maximum thickening but it was >380e450 m close to the confluence of the Academy Glacier and the Foundation Ice Stream. The data do show progressive thinning during the Holocene, and this is consistent with recent data from Balco et al. (2016) that show similar thinning trajectories from the northern Pensacola Mountains (Fig. 3).

Foundation Ice Stream and Academy Glacier thinning histories
The most detailed thinning histories come from the north side of the Academy Glacier where the datasets from Williams Hills and Mt. Bragg and Mt. Harper give similar, but not identical, deglacial trajectories (Fig. 4f) of progressive thinning from the early-to mid-Holocene. Specifically the Williams Hills data have no samples younger than 5 ka whereas at Mt. Bragg and Mt. Harper there are younger samples. Moreover the rate of thinning may have been faster at Williams Hills than at Mt. Bragg/Mt. Harper (Fig. 4b, d, f). Mt. Hobbs, the highest summit of the Williams Hills, yielded an exposure age of 8.85 ka at 450 m above present ice. On Mt. Bragg, the highest sample taken, 383 m above present ice, yielded an age of 17.6 ka. The data do not directly constrain either the extent or timing of the maximum glaciation at either site.
The Williams Hills exposure ages are similar to the exposure ages reported from the same area by Balco et al. (2016), which also show progressive thinning through the first half of the Holocene and no ages <4.18 ka. Fig. 4b and f shows that the thinning trajectories of the two datasets are closely coincident. Farther south, in the Thomas Hills our data do not record a clear post-glacial thinning trajectory apart from a single age of 4.2 ka at Bentley Peak and a single age of 12.8 ka amongst a mixed age population from a single moraine on Sugden Ridge. However, the dataset from Balco et al. (2016) show that ice was >250 m thicker at the southern end of the Thomas Hills in the early Holocene and that the ice thinned rapidly prior to 5 ka (Fig. 4i), consistent with the sites along the north side of the Academy Glacier.

Lack of recently exposed erratics in the Schmidt Hills
The lack of erratics younger than 236 ka in the Schmidt Hills, despite widespread evidence for LGM glaciation elsewhere in the Pensacola Mountains, is similar to the pattern found by Balco et al. (2016) from the same area, where all of the 10 Be apparent exposure ages in the Schmidt Hills were from prior to the last glacial cycle. Both our data and the data presented in Balco et al. (2016) show clear evidence for ice cover up to 450 m thicker than present in the Williams Hills, which lie only a few tens of kilometres south of the Schmidt Hills (Fig. 1), and so this difference is puzzling. Balco et al. (2016) undertook multiple isotopic analyses ( 10 Be, 21 Ne, 3 He, and 14 C) on their samples and showed that there were inconsistencies between the different isotopes. They argued that although their in situ 14 C apparent exposure ages showed a clear long-term exposure with no significant intervals of ice cover above 500 m elevation (corresponding to ice~350 m thicker than present), the dataset could not be reconciled with simple Holocene thinning histories. They suggested that to explain the ages would require either significant unrecognized analytical errors (of up to 25%), or a complex and unlikely configuration of LGM ice, which transported the samples from prior exposure on higher summits in the Schmidt Hills. Balco et al. (2016) concluded that they could not clearly resolve the exposure history of the Schmidt Hills but that the in situ 14 C concentrations were clear in showing that any LGM ice cover of the area reached <500 m asl elevation. We concur that the lack of dated evidence for LGM glaciation of the Schmidt Hills is difficult to explain. Perhaps it reflects some effect of lithology on analytical procedures or more probably an effect of patterns of local ice flow. It is notable that an expanded Childs Glacier, coming off the Iroquois plateau may have diverted the debris-bearing ice that was flowing NW over the Williams Hills to the west before it reached the Schmidt Hills (Fig. 1). The Balco et al. (2016) 14 C data require that the ice did not thicken more than~350 m above present at the Schmidt Hills. Given that the Foundation Ice Stream thickened at least by 450 m only a few tens of km to the south then this would require a significant steepening of the Foundation Ice Stream gradient, or that the centreline of the Foundation Ice Stream thickened significantly more than the margins at Schmidt Hills (Balco et al., 2016). The implications of the Schmidt Hills data for former configuration of an expanded Foundation Ice Stream are explored further using a numerical flowband model in Whitehouse  2016), where it is demonstrated that differential thickening of the centre and margins of the Foundation Ice Stream could potentially explain the Schmidt Hills data.

Mixed age populations on boulder moraines and reworking of clasts
Conventionally, boulder moraines provide a useful fix on the position of a formerly expanded ice margin and rigorous dating of such a limit should yield a well-dated reconstruction of that margin. However, in the Pensacola Mountains we found that each of the boulder moraines we mapped and sampled (Mt. Bragg, Sugden Ridge) yielded mixed age populations. For this reason we have been unable to assign ages to any of the boulder moraines. Each of these boulder limits is associated with a present-day ice margin that is dominantly composed of blue-ice regions. The mixed ages imply that either the reworking of erratics along the blue-ice margins of Foundation Ice Stream may be pervasive, or that the debris has been emerging at the ice margin for millennia.
The cross-cutting moraines at Sugden Ridge reflect complex behaviour of the Foundation Ice Stream and locally-sourced Mac-Namara Glacier. There are a number of features of the mixed age populations of these moraines that are worthy of discussion. At face value, the mixed ages e ranging from 12.8 ka to 380 ka e are difficult to reconcile with the Williams/Bragg/Harper trajectory of an expanded Foundation Ice Stream at the LGM which thinned through the early to mid-Holocene. There are at least three possible explanations. Firstly, all fluctuations at Sugden Ridge predate the LLGM and have then been over-ridden by thicker LLGM ice. Overriding of landforms without modification by cold-based ice has been noted elsewhere in Antarctica (Sugden et al., 2005). In this case the youngest age of 12.8 ka may be the closest representative thinning age. Second, all fluctuations are post-LLGM but the majority of samples are reworked. This could explain the substantial scatter on every moraine sampled. Third, the data may reflect the real timing of fluctuations (dated by the youngest sample on each moraine) but would imply very small thickness changes at each site, and also ice sheet thickening that is inconsistent with the timing of ice sheet change elsewhere in the study area. The latter two explanations are very difficult to reconcile with the Bragg/ Harper/Williams dataset and with evidence of the Foundation Ice Stream thinning by 125 m since 4.18 ka at Bentley Peak, and so we favour the first explanation, namely that the cross-cutting moraine sequences on Sugden Ridge have been over-ridden.

Comparison to other Weddell Sea deglacial chronologies
Our data can be compared with previous work on the glacial history of the Antarctic Ice Sheet in the Weddell Sea embayment. Bentley et al. (2010) showed modest LLGM thickening of 230e480 m in the southern Ellsworth Mountains followed by progressive thinning through the Holocene to reach present ice levels by c. 2 ka. This is similar to the glacial history reported here and both sites sit in similar configurations close to the present-day grounding line where thickening is likely to have been significant during past glacial periods. In the Shackleton Range, farther north along the Thiel Trough Ice Stream, Hein et al. (2011) showed that there had been minimal thickening at the LGM. The implication of this is that an expanded Thiel Trough Ice Stream could not have thickened significantly at the vicinity of the Shackleton Range (Hein et al., 2011;. Ice core data from Berkner Island showed that this ice rise, with a current altitude of 886 m asl was not overridden by thick interior ice during the LGM and that it remained an independent ice centre throughout the last glacial cycle Mulvaney et al., 2007).

4.5.
A late Holocene ice sheet readvance in the Weddell Sea? Bradley et al. (2015) suggested that a number of puzzling glaciological and geophysical observations could be explained if the ice sheet in the Weddell Sea had retreated to smaller-than-presentlimits in the mid-to Late Holocene and had subsequently advanced to its present configuration. Specifically, this would help explain observations of grounding lines existing in an apparently stable configuration on reverse slopes, and also low GPS-measured crustal uplift rates, including an observation of subsidence from at least one GPS receiver located in the interior of the ice sheet inboard of the grounding line. Moreover, a readvance from smaller-thanpresent would also explain the glaciological observations of significant Late Holocene flow re-organisation of the Institute and M€ oller Ice Streams . Data from the Ellsworth Mountains are consistent with the concept of a Late Holocene ice sheet readvance since the youngest exposure ages of erratics close to the present margin are 2e3 thousand years old Hein et al., 2016a). Our new data from the Foundation Ice Stream are consistent with the data from the Ellsworth Mountains and glaciological observations from the Institute and M€ oller Ice Streams: specifically, in the Williams Hills there are no exposure ages younger than 5ka (our data) or 4 ka (Balco et al., 2016). Thus the ice margin in the vicinity of the Williams Hills may have thinned below present ice levels any time after 4ka, and only subsequently re-thickened. It is notable that the Williams Hills data show a very abrupt lower limit to exposure ages and that the data would be consistent with rapid ice sheet thinning adjacent to the Williams Hills at 4e5 ka. In the area of Mt. Bragg and Mt. Harper the youngest ages are slightly younger, with the youngest age above the present margin being 3.1 ka (an age at the present margin yields 2.5 ka but because it sits right at the present margin we cannot rule out it being reworked by the margin of a thickening ice sheet). In summary, our data are consistent with, but do not mandate, a scenario of Holocene thinning and subsequent thickening in the interval after 4 ka (Williams Hills) or 3.1e2.5 ka (Mt. Bragg/Mt. Harper).

Sampling of ice stream margins in Antarctica for thinning histories
In Antarctica, blue ice regions are regions of net ablation, often due to sublimation under katabatic wind regimes, in what would otherwise be the accumulation area of the ice sheet. They are a prime source of glacially-deposited material because ablation brings englacial material to the ice sheet surface. At face value, this makes them attractive areas for sampling for studies of former thinning histories. However, the problems of sampling for cosmogenic nuclide surface exposure dating from areas adjacent to blue ice moraines have been apparent in several studies and this is no exception. In areas adjacent to blue ice on the present-day Foundation Ice Stream or Academy Glacier, the apparent exposure ages we retrieved have a high component of scatter (e.g., Mt. Bragg, Sugden Ridge, Mt. Yarborough). As in several previous studies, the surface weathering characteristics or appearance of erratics cannot be used as a reliable guide to relative age: fresh-looking erratics regularly yield 'old' ages (see discussion in Balco (2011) andHein et al. (2014)). Moreover, comparison of our work with that by Balco et al. (2016) shows that the proportion of younger ages can change markedly in closely adjacent sites: for example in the Thomas Hills we sampled Mt. Yarborough and found no samples younger than 255 ka (n ¼ 12), yet sampling by Balco et al. of the adjacent nunataks on either side ('Nance Ridge' and 'South Mainland') yielded an age population where 50% of the ages were post-LGM in age (n ¼ 16). This implies significant variations in spatial distribution.