Cosmogenic nuclide exposure age scatter records glacial history and processes in McMurdo Sound, Antarctica

The preservation of cosmogenic nuclides that accumulated during periods of prior exposure but were not subsequently removed by erosion or radioactive decay complicates interpretation of exposure, erosion, and burial ages used for a variety of geomorphological applications. In glacial settings, cold-based, non-erosive glacier ice may fail to remove inventories of inherited nuclides in glacially transported material. As a result, individual exposure ages can vary widely across a single landform (e.g., moraine) and exceed the expected or true depositional age. The surface processes that contribute to inheritance remain poorly understood, thus limiting interpretations of cosmogenic nuclide datasets in glacial environments. Here, we present a compilation of new and previously published exposure ages of multiple lithologies in local Last Glacial Maximum (LGM) and older Pleistocene glacial sediments in the McMurdo Sound region of Antarctica. Unlike most Antarctic exposure chronologies, we are able to compare exposure ages of local LGM sediments directly against an independent radiocarbon chronology of fossil algae from the same sedimentary unit that brackets the age of the local LGM between 12.3 and 19.6 ka. Cosmogenic exposure ages vary by lithology, suggesting that bedrock source and surface processes prior to, during, and after glacial entrainment explain scatter. 10Be exposure ages of quartz in granite, sourced from the base of the stratigraphic section in the Transantarctic Mountains, are scattered but young, suggesting that clasts entrained by sub-glacial plucking can generate reasonable apparent exposure ages. 3He exposure ages of pyroxene in Ferrar Dolerite, which crops out above outlet glaciers in the Transantarctic Mountains, are older, which suggests that clasts initially exposed on cliff faces and glacially entrained by rock fall carry inherited nuclides. 3He exposure ages of olivine in basalt from local volcanic bedrock in the McMurdo Sound region contain many excessively old ages but also have a bimodal distribution with peak probabilities that slightly pre-date and post-date the local LGM; this suggests that glacial clasts from local bedrock record local landscape exposure. With the magnitude and geological processes contributing to age scatter in mind, we examine exposure ages of older glacial sediments deposited by the most extensive ice sheet to inundate McMurdo Sound during the Pleistocene. These results underscore how surface processes operating in the Transantarctic Mountains are expressed in the cosmogenic nuclide inventories held in Antarctic glacial sediments.

Abstract. The preservation of cosmogenic nuclides that accumulated during periods of prior exposure but were not subsequently removed by erosion or radioactive decay complicates interpretation of exposure, erosion, and burial ages used for a variety of geomorphological applications. In glacial settings, cold-based, non-erosive glacier ice may fail to remove inventories of inherited nuclides in glacially transported material. As a result, individual exposure ages can vary widely across a single landform (e.g., moraine) and exceed the expected or true depositional age. The surface processes that contribute to inheritance remain poorly understood, thus limiting interpretations of cosmogenic nuclide datasets in glacial environments. Here, we present a compilation of new and previously published exposure ages of multiple lithologies in local Last Glacial Maximum (LGM) and older Pleistocene glacial sediments in the McMurdo Sound region of Antarctica. Unlike most Antarctic exposure chronologies, we are able to compare exposure ages of local LGM sediments directly against an independent radiocarbon chronology of fossil algae from the same sedimentary unit that brackets the age of the local LGM between 12.3 and 19.6 ka. Cosmogenic exposure ages vary by lithology, suggesting that bedrock source and surface processes prior to, during, and after glacial entrainment explain scatter. 10 Be exposure ages of quartz in granite, sourced from the base of the stratigraphic section in the Transantarctic Mountains, are scattered but young, suggesting that clasts entrained by sub-glacial plucking can generate reasonable apparent exposure ages. 3 He exposure ages of pyroxene in Ferrar Dolerite, which crops out above outlet glaciers in the Transantarctic Mountains, are older, which suggests that clasts initially exposed on cliff faces and glacially entrained by rock fall carry inherited nuclides. 3 He exposure ages of olivine in basalt from local volcanic bedrock in the McMurdo Sound region contain many excessively old ages but also have a bimodal distribution with peak probabilities that slightly pre-date and post-date the local LGM; this suggests that glacial clasts from local bedrock record local landscape exposure. With the magnitude and geological processes contributing to age scatter in mind, we examine exposure ages of older glacial sediments deposited by the most extensive ice sheet to inundate McMurdo Sound during the Pleistocene. These results underscore how surface processes operating in the Transantarctic Mountains are expressed in the cosmogenic nuclide inventories held in Antarctic glacial sediments.

Introduction: the problem of inherited cosmogenic nuclides
In situ cosmogenic nuclides, which accumulate in nearsurface materials during exposure to cosmic radiation, can be measured across a wide range of environments and timescales to quantitatively describe earth surface processes, including quantifying erosion rates (Portenga et al., 2019), dating landforms and exposed bedrock surfaces (Christ et al., 2021;Wells et al., 1995), and reconstructing changes in climate over millions of years Schaefer et al., 2016;Shakun et al., 2018), among many other applications. However, inherited nuclides -those which accumulated in surface material during periods of prior exposure but were not subsequently removed by erosion or through the decay of radioactive nuclides -can produce results that are difficult to interpret in surface exposure dating samples. Nuclide inheritance affects nearly all geomorphologic settings: it complicates the interpretation of alluvial fan ages in deserts (Owen et al., 2011), river terrace exposure ages used to determine canyon incision rates (Cook et al., 2009), shore platform erosion rates in coastal settings (Trenhaile, 2018), and glacial chronologies using exposure ages of boulders on moraines or glacially sculpted bedrock (Briner et al., 2014;Corbett et al., 2016;Davis et al., 1999;Young et al., 2016). Inherited nuclides are prevalent in glacial sediments, from alpine glacier moraines (Briner et al., 2005;Heyman et al., 2011) to glacial erratics transported by polar ice sheets (Brook et al., 1995;Corbett et al., 2019;Hein et al., 2014). Surface exposure ages of glacial deposits are assumed to record the duration of exposure following emplacement and/or glacial retreat, allowing a reconstruction of former ice extent at specific time slices (Dunai, 2010). In practice, exposure chronologies can yield highly scattered exposure ages that vary widely across a single, contemporaneously deposited landform (i.e., a moraine) and often greatly exceed the expected or true depositional age Heyman et al., 2011). Cosmogenic nuclide scatter in glacial environments is likely due to clasts or surfaces that contain nuclides that accumulated during periods of prior exposure but were not subsequently removed by non-erosive, cold-based glacial ice (Briner et al., 2006;Corbett et al., 2019;Koester et al., 2020). Cold-based glaciation occurs where ice thickness and/or basal ice temperatures are insufficient to produce meltwater and basal sliding, both of which facilitate subglacial erosion and therefore removal of inherited nuclides in formerly exposed surfaces (Bennett and Glasser, 1996). The basal thermal regime below ice can be highly heterogeneous, resulting in selective erosion of the landscape (Briner et al., 2006;Jamieson et al., 2010;Sugden et al., 2005). Consequently, glacial sediments can be entrained, transported, and deposited by both warm-and cold-based glacial ice and ultimately produce scattered cosmogenic surface exposure ages Heyman et al., 2011).
The problem of inherited nuclides and exposure age scatter is exacerbated in Antarctica, where large polythermal ice sheets, hyper-arid polar desert climate conditions, and extremely low sub-aerial erosion rates have persisted since the Miocene (Lamp et al., 2017;Lewis and Ashworth, 2016;Lewis et al., 2008;Shakun et al., 2018;Sugden and Denton, 2004;Sugden et al., 2006). Glacial sediments in ice-free areas of Antarctica preserve direct geological constraints on former ice sheet configurations from Holocene to Miocene times (Anderson et al., 2020;Balco et al., 2013;Balter-Kennedy et al., 2020;Bromley et al., 2010;Christ and Bierman, 2020;Hall et al., 2015;Jones et al., 2015Jones et al., , 2021Nichols et al., 2019;Spector et al., 2017). Cosmogenic nuclides are often the only method available for constraining a numerical chronology of Antarctic glacial deposits. Radiocarbon dating is usually not applicable due to the absence of organic remains and/or because the sedimentary archives are > 50 kyr. Low erosion rates have resulted in long-lived exposure of high-elevation surfaces that are later covered and/or entrained by cold-based glacier ice that fails to erode and remove inherited nuclides Hein et al., 2014;Stone et al., 2003;Sugden et al., 2014). Additionally, debris entrainment via rockfall may introduce nuclide scatter due to differential erosion on cliff faces (Lamp et al., 2017;Mackay et al., 2014). However, large outlet glaciers and ice streams have also deeply eroded and modified parts of the landscape (Lewis and Ashworth, 2016;Sugden et al., 2005) and transported sediments to terrestrial margins (Stuiver et al., 1981) and onto the continental shelf (Anderson, 1999;Bentley et al., 2014). As a result, many cosmogenic nuclide exposure chronologies from Antarctic terrestrial glacial deposits contain scatter with exposure ages that can span tens to hundreds of thousands of years for individual landforms (Brook et al., 1993;Hein et al., 2014;Joy et al., 2014;Staiger et al., 2006;Swanger et al., 2011).

McMurdo Sound: a unique setting to examine cosmogenic nuclide inheritance
Here, we focus on cosmogenic nuclide analyses from icefree areas surrounding McMurdo Sound, Antarctica (78 • S, 165 • E), as it is uniquely suited to examine scatter in exposure age datasets from glacial sediments (Fig. 1a). McMurdo Sound hosts one of the largest ice-free areas in Antarctica and contains Pleistocene glacial sediments on volcanic islands and coastal valleys of the Transantarctic Mountains (Christ and Bierman, 2020;Denton and Marchant, 2000;Hall et al., 2015;Hall and Denton, 2000;Jackson et al., 2018;Stuiver et al., 1981) (Christ and Bierman, 2020;Denton and Marchant, 2000;Greenwood et al., 2018;Hall et al., 2015;Stuiver et al., 1981) (Fig. 1). These incursions of grounded ice into McMurdo Sound transported and deposited lithologies from the Transantarctic Mountains, including Granite Harbor Intrusives, Ferrar Dolerite, Koettlitz Group metamorphic rocks, Beacon Sandstone, and other rocks from the East Antarctic interior (Christ and Bierman, 2020;Denton and Marchant, 2000;Talarico et al., 2012Talarico et al., , 2013 (Fig. 1b). Specifically, lithologies present in the AND-1B drill core and erratics in ice-free areas of McMurdo Sound correspond primarily to rocks found along the Mullock and Skelton glaciers (Talarico et al., , 2013. The diversity of lithologies in glacial sediments in the McMurdo Sound region provides targets for multiple cosmogenic nuclides to calculate exposure ages; yet previous efforts in the McMurdo Sound region yielded a complex exposure history Brook et al., 1995;Joy et al., 2017). The ice-free islands and peninsulas of McMurdo Sound contain at least two different Pleistocene glacial sedimentary units (Denton and Marchant, 2000). At lower elevations (< 510 m), the local Last Glacial Maximum (LGM) glacial deposit, known locally as Ross Sea drift, is characterized by a high concentration of erratic lithologies and un-weathered clasts that lack iron staining or ventifaction (Christ and Bierman, 2020;Denton and Marchant, 2000;Hall et al., 2015;Stuiver et al., 1981). Unlike many Antarctic glacial sediments, Ross Sea drift has a robust radiocarbon chronology of fossil algae embedded in moraines and ice-marginal sediments, which indicates that grounded ice maintained its local LGM extent between 12.3 and 19.6 calibrated thousands of years before present (ka cal BP) (Christ and Bierman, 2020;Hall et al., 2015;Hall and Denton, 2000;Jackson et al., 2018;Stuiver et al., 1981). A pioneering application of cosmogenic 10 Be, 26 Al, and 3 He in several lithologies from Ross Sea drift produced a highly scattered exposure age dataset ranging from 8 to 106 kyr (Brook et al., 1995). A recent elevation transect of 10 Be in quartz ( 10 Be quartz ) exposure ages of granite boulders along eastern Mount Discovery indicates that ice thinned from its maximum extent no earlier than 14.1 ka to present conditions by 7.6 ka . At higher elevations on the ice-free regions of volcanic islands in McMurdo Sound, other glacial sediments remain largely unmapped and undated. These glacial deposits are characterized by a lower concentration of erratic lithologies relative to Ross Sea drift and are presumably older, as indicated by enhanced surface weathering characteristics including chemically altered surfaces and ventifacted and cavernously weathered boulders (Denton and Marchant, 2000). Earlier reconnaissance mapping on Mount Discovery indicated that these deposits extend up to 770 m elevation (Denton and Marchant, 2000;Stuiver et al., 1981). Other similar higher-elevation glacial deposits exist above Ross Sea drift in the Royal Society Range and, while they have not been mapped in detail, have some exposure age constraints that range widely between 104-567 kyr (Brook et al., 1995).
There are several advantages to examining exposure ages of glacial deposits in the ice-free areas surrounding Mc-Murdo Sound. Exposure ages of clasts from Ross Sea drift can be compared against the independently constrained radiocarbon chronology of fossil algae embedded in moraines and ice-marginal sediments from the same deposit (Figs. 2 and 3) (Christ and Bierman, 2020;Hall et al., 2015;Hall and Denton, 2000;Jackson et al., 2018;Stuiver et al., 1981). The diversity of lithologies in glacial sediments in the ice-free areas surrounding McMurdo Sound provides targets for multiple cosmogenic nuclides by which to calculate exposure ages and these different lithologies in glacial sediments can be tied to their bedrock sources in the Transantarctic Mountains (Talarico et al., , 2013. We compiled new and previously published Brook et al., 1995) exposure ages of different lithologies in Ross Sea drift to quantify the magnitude of exposure age scatter and investigate surface processes that contribute to, prevent, or reduce nuclide inheritance in Antarctic terrestrial glacial sediments. With the magnitude and pattern of exposure age scatter known, we then examine previously unmapped and undated glacial deposits that delineate the maximum extent of grounded ice on Mount Discovery.

Glacial geologic mapping
We built upon the detailed mapping and descriptions of Ross Sea drift on the volcanic islands and peninsulas (Christ and Bierman, 2020;Denton and Marchant, 2000) with new surveys of higher-elevation glacial deposits on Mount Discovery (hereafter referred to as the Upper Discovery deposit). The Upper Discovery deposit was mapped at several sites on Mount Discovery at elevations between 450 and 775 m. GPS data points were collected using a handheld Garmin GPS unit (precision < 10 m) on moraines and glacial limits. We complemented on-the-ground field mapping with analysis of aerial photographs, high-resolution (< 1.85 m per pixel) Dig-italGlobe, Inc. imagery, and a stereo-photogrammetric digital elevation model (2 m per pixel) created by the Polar Geospatial Center (Howat et al., 2019).

Sample collection
Erratic clasts of Granite Harbor Intrusives (granite) and Ferrar Dolerite were collected for cosmogenic 10 Be in quartz ( 10 Be qtz ) and 3 He in pyroxene ( 3 He pyx ) exposure dating, respectively, from moraine crests and the limits of Ross Sea drift and the Upper Discovery deposit on the volcanic islands of McMurdo Sound. Clasts that were clearly perched, either on top of other clasts or on the surface of glacial sediment, were collected over those embedded into sediment to avoid "young" exposure ages due to possible exhumation. Samples were collected from local topographic highs (i.e., moraine ridge crests) to avoid local shielding, clast overturning, rotation, exhumation, or downslope movement. The GPS coordinates, topographic shielding, clast orientation, clast dimensions, weathering characteristics, and geomorphic setting were recorded for each sample. Samples from larger clasts were collected from the upper 10 cm of rock using a hammer drill and wedges and shims. Whole-rock samples of smaller cobble-sized clasts were taken.

Cosmogenic nuclide analyses
The upper ∼ 5 cm of each sample was cut using a rock saw at Boston University. Samples were shipped to the Cosmogenic Nuclide Dating Group at the Lamont-Doherty Earth Obser-  (Arndt et al., 2013) and the Reference Elevation Mosaic of Antarctica generated by the Polar Geospatial Center (Howat et al., 2019); generalized bedrock geology adapted from (Grindley and Laird, 1969;Warren, 1969). (b) Generalized stratigraphy of southern Victoria Land, Antarctica adapted from (Fitzgerald, 2002). vatory for further sample processing and in situ nuclide extraction.

Cosmogenic 3 He in pyroxene
Samples were crushed to 63-125 µm to separate pyroxene mineral grains, rinsed thoroughly with deionized (DI) water to remove fines, then leached in a 10 % phosphoric acid solution overnight to remove oxidation and other contaminants from mineral surfaces. The samples were then density-separated in a centrifuge using sodium polytungstate heavy liquid at a density of 3.0 g cm −3 ; the sinking fraction was rinsed with deionized water and dried overnight. Magnetic minerals were removed using a Frantz magnetic separator set between 0.4 and 0.5 amps. The non-magnetic fraction was then leached in a 2 % HNO 3 /2 % HF solution on a shaker table overnight, then rinsed, dried, and packed in aluminum foil for analysis. Helium isotopes were measured with the LDEO MAP 215-50 noble gas mass spectrometer using an air standard. Previous cosmogenic nuclide studies show that pyroxene from Ferrar Dolerite contains non-trivial amounts of non-cosmogenic, nucleogenic 3 He, which is relevant for LGM-aged sediments (Ackert, 2000;Balter-Kennedy et al., 2020;Kaplan et al., 2017;Margerison et al., 2005). We subtracted a non-nucleogenic 3 He correction of 3.3 × 10 6 atoms g −1 (Balter-Kennedy et al., 2020) to all 3 He measurements, as this is the most up-to-date correction measurement and higher correction values (5-  (Christ and Bierman, 2020;Hall et al., 2015;Hall and Denton, 2000;Jackson et al., 2018). All exposure ages from this study and previously published data Brook et al., 1995;Joy et al., 2017) were recalculated using the same LSDn scaling scheme  and shown here as the exposure age ± 1σ internal uncertainty from previous studies. Extent of Fig. 4 shown in white for detailed mapping on Mount Discovery. AND-1B drill core (red circle) shown for reference . Base map: Landsat Image Mosaic of Antarctica and 500 m contours derived from digital elevation models created by the Polar Geospatial Center from DigitalGlobe, Inc. imagery (Howat et al., 2019). 7 × 10 6 atoms g −1 ) generate some exposure ages that are unreasonably young. 3 He in pyroxene ( 3 He pyx ) exposure ages were calculated using Version 3 of the online exposure age calculator hosted by the University of Washington (https: //hess.ess.washington.edu/, last access: 6 September 2021) (Balco et al., 2008) assuming a density of 2.9 g cm −3 , corrected for shielding and thickness, the Antarctica-specific atmospheric model (ANT) (Stone, 2000), the global 3 He production rate (120 ± 9.4 atom g −1 yr −1 ) (Goehring et al., 2010), and the LSDn scaling scheme . We employed the LSDn scaling scheme, which is time dependent, because the compiled dataset spans a wide timescale over the past 500 kyr. We note that exposure ages calculated using the St or Lm scaling schemes generate slightly older exposure ages, but do not change observed patterns in the wider dataset.

Cosmogenic 10 Be in quartz
Samples were crushed to 500-125 µm, then froth flotation was used to separate the quartz grains. Quartz separates were purified by leaching in a HNO 3 /HF acid solution, and then 10 Be qtz was extracted from purified quartz, both according to standard protocols in the Cosmogenic Nuclide Dating Group and established methods (Brown et al., 1991;Kohl and Nishiizumi, 1992). 10 Be/ 9 Be ratios were measured at the Center for Accelerator Mass Spectrometry at Lawrence Livermore National Laboratory and normalized to 07KNSTD3110, which has an assumed 10 Be/ 9 Be ratio of 2.85 × 10 −12 (Nishiizumi et al., 2007). 10 Be qtz exposure ages were calculated using Version 3 of the online exposure age calculator hosted by the University of Washington (https://hess.ess.washington.edu/) (Balco et al., 2008) assuming a rock density of 2.7 g cm −3 corrected for shielding and thickness, the Antarctica-specific atmospheric model (ANT) (Stone, 2000), the global 10 Be production rate (4.15 atoms g −1 yr −1 ) (Martin et al., 2017), and the LSDn scaling scheme  for the same reasons explained for 3 He pyx above. Three procedural blanks were processed with samples to correct background 10 Be/ 9 Be (Table 3). Blank corrections were applied to samples according to the batch in which they were processed. Two blanks (BLK1-2016Jun03, BLK2-2016Jun03) were averaged (4.6 ± 1.7 × 10 −16 ) and applied to 10 Be/ 9 Be measurements from samples ACX-13-18, ACX-14-26, and ACX-13-56. A blank correction of 4.9 ± 1.4 × 10 −16 (BLK-2016Apr12) was applied to 10 Be/ 9 Be measurements from samples ACX-14-08 and ACX-14-16.

Compiled exposure chronology
New and previously published cosmogenic nuclide concentrations from previous studies in the McMurdo Sound region Brook et al., 1995;Joy et al., 2017) were compiled from the ICE-D database (http://antarctica. ice-d.org, last access: 6 September 2021) . Exposure ages for all samples were calculated using Version 3 of the online exposure age calculator hosted by the University of Washington (https://hess.ess.washington.edu/) using the same atmospheric model, LSDn scaling scheme, and global production rates as used for the new samples in this study scheme . Only samples collected from the maximum extent of Ross Sea drift were selected from previous studies; we do not include exposure ages collected from elevation transects and used for deglacial thinning chronologies. All exposure ages of older glacial deposits, however, were included as geomorphologic mapping of these deposits is less complete. Only 10 Be qtz in granite, 3 He in olivine ( 3 He olv ) in basalt, and 3 He pyx in dolerite are considered here; too few exposure ages in other erratic lithologies (such as gneiss and sandstone) and 26 Al qtz exist to be compared against the larger dataset. Additionally, we exclude previously published 3 He qtz exposure ages due to diffusion uncertainties inherent in that specific nuclide system (Brook et al., 1995).

Glacial geologic mapping
The Upper Discovery deposit reaches a maximum elevation of 775 m and decreases in elevation from east to west along Mount Discovery (Fig. 4). The limit of the Upper Discovery deposit is nearly continuous and well-defined across the northern face of Mount Discovery and is draped onto cinder cones, volcanic vents, and intervening valleys. Windswept snow fields and alpine glaciers presently cover and cross over the glacial limit in a few locations. The elevation of the glacial limit descends to ∼ 450 m elevation on western Mount Discovery above Koettlitz Glacier, where it becomes poorly defined due to disturbance from solifluction.  Table 1). Four of the five 3 He pyx exposure ages of Ferrar Dolerite in Ross Sea drift on eastern Mount Discovery and Brown Peninsula range between 14.3 ± 1.0 ka to 36.0 ± 1.8 ka, while the remaining exposure age from a sharp-crested moraine on Brown Peninsula was much older: 361.6 ± 10.7 ka. 3 He pyx exposure ages (n = 4) from the Upper Discovery deposit were old; three ages are between 229.2 ± 8.0 ka and 245.9 ± 13.6 ka; the other exposure age was even older (347.9 ± 11.4 ka).
4.2.2 New 10 Be qtz in granite exposure ages 10 Be qtz exposure ages of granite sort according to stratigraphic age and display less scatter than 3 He pyx exposure ages (Table 2). 10 Be qtz exposure ages of clasts from Ross Sea drift on Mount Discovery and Brown Peninsula had two younger and similar ages (12.6 ± 0.3 ka, 12.8 ± 0.3 ka) and two older ages (20.2 ± 0.6 ka, 34.2 ± 0.7 ka). A single exposure age of a cavernously weathered boulder in the Upper Discovery deposit yielded an age of 144.5 ± 1.9 ka.

New insights from inherited nuclides and exposure age scatter
In former and active ice sheet environments, cosmogenic nuclides in glacial sediments often do not yield simple, easy-tointerpret exposure ages; instead, cosmogenic nuclides record the integrated effect of many surface processes operating on a landscape through time. Rather than focusing exclusively on the age of ice extent or retreat, cosmogenic nuclides can also be used to disentangle how glacial sediment is entrained from bedrock sources, transported by glacier ice, and deposited on the landscape. By taking advantage of the relatively simple geology of the Transantarctic Mountains, an independent (albeit rare) radiocarbon chronology, and analysis of multiple nuclides and lithologies in Ross Sea drift, we can better quantify the magnitude of and processes responsible for exposure age scatter. While scatter in exposure ages from Ross Sea drift prevents precise age constraints of maximum ice extent, it is possible to assign ages at glacial-interglacial timescales. With this information, we then consider the age of older Pleistocene glacial deposits in McMurdo Sound.

Scattered exposure ages of Ross Sea drift
The narrow range of calibrated radiocarbon ages (12.3-19.6 ka cal BP) and scattered range of exposure ages that date to within the last glacial period confirm that Ross Sea drift is the youngest glacial deposit recording ice sheet expansion into McMurdo Sound. However, the compiled cosmogenic nuclide exposure chronology derived from multiple lithologies is highly scattered (Table 4), and most individual exposure ages exceed the timeframe of local LGM ice extent  a Assumed density for all samples = 2.9 g/cm 3 . b Corrected for 3.3 × 10 6 atoms g −1 contribution of nucleogenic 3 He in Ferrar Dolerite (Balter-Kennedy et al., 2020). c Exposure ages were calculated using Version 3 of the online exposure age calculator hosted by the University of Washington (https://hess.ess.washington.edu/) (Balco et al., 2008) using the Antarctica-specific atmospheric model (ANT) (Stone, 2000), the global 3 He production rate (Goehring et al., 2010), and the LSDn scaling scheme . dd: decimal degrees.
established by the radiocarbon chronology. While many ages are "too old" and suggest that these clasts carry an inventory of inherited nuclides, most of the exposure ages (with three exceptions) are between the early Holocene (Marine Isotope Stage (MIS) 1) and last glacial period (MIS 2-4). Upper-elevation, weathered deposits yield much older Pleistocene ages (MIS 5-12) that are incompatible with a local LGM depositional age. This exposure age dataset underscores the difficulty of using surface exposure dating in Antarctica to constrain precise ages of individual landforms or glacial deposits from the local LGM using 10 Be qtz , 3 He pyx , or 3 He olv . The most extreme example of scatter comes from two exposure ages from the same moraine ridge on northern Brown Peninsula. This sharp-crested moraine ridge is independently dated to 17.9 ka cal BP (Christ and Bierman, 2020), but nearly adjacent clasts yield exposure ages that are both too old and widely different. A perched granite cobble produced an exposure age of 20.2 ± 0.5 ka (ACX-14-08) and a nearby dolerite cobble has an exposure age of 382.8 ± 10.9 ka (ACX-14-05), the oldest of any sample from Ross Sea drift. On eastern Mount Discovery (512 m), the glacial limit is marked by a stark lithological contrast between erratic-rich Ross Sea drift lying at lower elevations than volcanic-rich glacial deposits Christ and Bierman, 2020;Denton and Marchant, 2000). Here, granite exposure ages coincide with the end of the local LGM in McMurdo Sound (12.6 to 14.4 ka, n = 3) (this study; Anderson et al., 2017), but two of the three dolerite exposure ages of dolerite from the same location are older (ACX-13-09: 29.6 ± 1.4 ka; ACX-13-12: 31.9 ± 1.4 ka). Clearly, exposure ages from individual landforms from the same local LGM deposit in the McMurdo Sound region contain clasts with different exposure histories.

Surface processes contributing to exposure age scatter
We suggest that scatter is related to bedrock source and that specific surface processes prevent, enhance, or reduce the likelihood of nuclide inheritance prior to and during glacial entrainment and deposition. Exposure age scatter sorts according to lithology and nuclide: granite 10 Be qtz exposure ages are younger, but scattered, dolerite 3 He pyx ages are consistently too old, and basalt 3 He olv ages slightly pre-date and post-date the local LGM. These patterns can be explained by examining how clasts were glacially entrained and incorporated into Ross Sea drift.
graphic section for much of the Transantarctic Mountains in southern Victoria Land (Fig. 1). The large EAIS outlet glaciers in the Transantarctic Mountains south of Mc-Murdo Sound -Byrd Glacier, Mullock Glacier, and Skelton Glacier -have all incised deep troughs that cut into Granite Harbor Intrusives (Fig. 1). Ice sheet models simulate that these outlet glaciers thicken and expand into the Ross Sea during glacial periods; they are also wet-based and erode large volumes of sediment (Golledge et al., 2013). Rock surfaces at the bottom of outlet glacier troughs are up to 1000 m below sea level and could not experience exposure to cosmic rays. In addition, the Pleistocene stability and persistence of EAIS outlet glaciers in the Transantarctic Mountains likely prevented prior exposure of the bedrock troughs (Kaplan et al., 2017). Taken together, this means that Granite Harbor Intrusives plucked from the base of outlet glacier troughs are less likely to contain cosmogenic nuclides inherited from periods of exposure prior to local LGM deglaciation in Ross Sea drift. The resulting exposure ages are more likely to record clast emplacement at maximum ice extent. Ferrar Dolerite bedrock is present below many outlet glaciers draining the EAIS and tributary alpine glaciers in the Transantarctic Mountains. Presumably, dolerite clasts are also sub-glacially plucked and entrained into overriding ice, which would generate exposure ages that reliably record clast emplacement on moraines. However, it appears that dolerite clasts in Ross Sea drift are more likely to have older than expected exposure ages that indicate a complex exposure history and nuclide inheritance.

Cliff exposure, rockfall, and supraglacial transport increases nuclide inheritance
Old exposure ages of several granite and nearly all dolerite clasts in Ross Sea drift may reflect prior cliff exposure and entrainment by rockfall along outlet glacier margins. Granite Harbor Intrusive rocks are exposed in cliff faces above the glacier surface closer to the coast and along outlet glacier margins in the Transantarctic Mountains (Fig. 1). During previous interglacial periods or glacial low-stands when outlet glaciers were thinner (Jones et al., 2017), exposure on granite cliff faces would have accumulated cosmogenic nuclides prior to later glacial entrainment. The effect of prior cliff exposure is likely more important for exposure ages of Ferrar Dolerite because these rocks form many of the peaks and cliff faces above glacier surfaces in the Transantarctic Mountains at elevations > 1000 m (Fig. 1). Prolonged cliff exposure combined with low erosion rates (Mackay et al., 2014) at high elevations (where the cosmogenic nuclide production rate is also higher) will increase cosmogenic nuclide inventories in dolerite more than granite clasts. Additionally, 3 He pyx is a stable cosmogenic nuclide that does not decay during burial (Dunai, 2010); this further increases the probability of nuclide inheritance in dolerite clasts. Even with the nucleogenic 3 He correction applied, dolerite 3 He pyx exposure ages still tend to be too old, suggesting nuclide inheritance. Despite increased outlet glacier surface elevations during glacial periods, cliff faces may not be subjected to wet-based glacial erosion due to insufficient ice thickness. Selective linear erosion below polythermal ice sheets can lead to wetbased, erosive glacial conditions in outlet glacier troughs but cold-based, non-erosive glacial conditions along valley walls Stone et al., 2003;Sugden et al., 2005). Clasts sourced from rockfall events onto the glacier surface or reworked from older ice marginal sediments in these regions of cold-based ice will not be sufficiently eroded to remove inherited nuclides. Ferrar Dolerite may be particularly susceptible to cold-based, non-erosive glacial transport because many high-elevation cliff faces are exposed over thin alpine and valley glaciers that are tributaries to larger outlet glaciers. Dolerite clasts sourced from these areas likely carry inherited nuclides that are not removed by glacial erosion.
When rockfalls from cliff faces deliver granite or dolerite onto glacier surfaces below, inherited nuclides remain in some clasts. While rockfall onto glaciers may shatter large volumes of rock and expose fresh surfaces, some clasts may retain the original surface exposed on the cliff face. Clasts that land directly on the ice surface will accumulate additional nuclides if they remain exposed during supraglacial transport. Rockfall that does not land directly on the glacier surface but accumulates along outlet glacier margins could be subjected to additional exposure prior to a glacial highstand that entrains these clasts. Additionally, clasts may be Cosmogenic nuclide data from Anderson et al. (2017), Brook et al. (1995), and Joy et al. (2017) accessed using the ICE-D database . Exposure ages calculated with Version 3 of the online exposure age calculator hosted by the University of Washington (https://hess.ess.washington.edu/) using the Antarctic atmospheric model (Stone, 2000), global production rates (Goehring et al., 2010;Martin et al., 2017), and LSDn scaling scheme . Sample coordinates from Brook et al. (1995) are approximate.
repeatedly recycled by multiple glacial advance and retreat events.

Ice sheet reworking of local volcanic rocks causes and prevents prior exposure
3 He olv exposure ages of basalt clasts, sourced from the local McMurdo Volcanic Suite, record local landscape exposure prior to and following ice sheet advance (Brook et al., 1995). The bimodal distribution of 3 He olv in basalt exposure ages pre-dates maximum ice sheet extent in McMurdo Sound (25.0 ka) and post-dates ice thinning (9.6 ka). Clasts with inherited nuclides and excessively old exposure ages may have been initially sub-aerially exposed during a glacial low-stand prior to the local LGM or the last interglacial (MIS 5). As ice advanced into McMurdo Sound during the local LGM, volcanic clasts with inherited nuclides may have been reworked into Ross Sea drift. However, clasts that lack previous exposure could be entrained from submarine sources that lack any previous exposure and could be transported along the base of the ice sheet in McMurdo Sound. As ice thinned and downwasted, basal clasts would be exposed last, possibly producing exposure ages younger than the onset of ice thinning (12.3 ka), a process which could also explain some of the younger 10 Be qtz exposure ages of granite in Ross Sea drift. The 3 He olv exposure ages are compatible with ice thinning on eastern Mount Discovery from its maximum extent after 14.0 ka to near present elevations at 7.3 ka . Likewise, 3 He olv exposure ages from the McMurdo Sound region agree with rapid lowering of outlet glaciers during the Early Holocene Jones et al., 2015Jones et al., , 2021Spector et al., 2017) as grounded ice in the Ross Sea retreated (Halberstadt et al., 2016;McKay et al., 2016).

Upper Discovery deposit: maximum glaciation of McMurdo Sound
The Upper Discovery deposit marks the maximum limit of glaciation of McMurdo Sound. The decreasing elevation of the Upper Discovery limit around Mount Discovery from 776 m in the east to ∼ 450 m in the west suggests that a larger ice sheet in the Ross Sea overflowed into McMurdo Sound during a glacial period prior to the local LGM (Fig. 4).
As there are no erratic lithologies present above this limit, the Upper Discovery deposit delineates the largest and thickest ice sheet in the western Ross Sea to inundate McMurdo Sound since Mount Discovery formed at 5.5 to 4.5 Ma (Kyle, 1990). Taking into consideration exposure age scatter, nuclide inheritance, and the few exposure ages available in our dataset, it is possible that the Upper Discovery deposit dates to glacial Termination III at the transition from the late MIS 8 glacial period to the early MIS 7 interglacial period (Fig. 5). At the local LGM limit on Mount Discovery, Ross Sea drift 3 He pyx ages exceed the 10 Be qtz ages by as much as ∼ 15 kyr. If a similar magnitude of nuclide inheritance is assumed for the Upper Discovery deposit, three 3 He pyx exposure ages in Ferrar Dolerite from the Upper Discovery deposit date within early MIS 7 to late MIS 8. One older 3 He pyx dolerite exposure age from this deposit (ACX-13-61: 347.9 ± 11.6 ka) is compatible with this MIS 7-8 age; as observed with Ross Sea drift, outlier 3 He pyx exposure ages can be preserved in the same deposit, possibly due to clast recycling. The 10 Be qtz exposure age (ACX-13-52: 153.3 ± 2.0 ka, MIS 6) is too young. However, this sample was collected above a cavernous weathering pit on a granite boulder, which clearly indicates that the boulder was eroding and thus losing some of the inventory of 10 Be that accumulated since deposition; this would yield a young-biased exposure age. Even if lithologyspecific boulder erosion rates for granite (13 mm kyr −1 ) and dolerite (19 mm kyr −1 ) (Marrero et al., 2018) in Antarctica are considered, the difference between the erosion-corrected exposure ages of the granite boulder (147 ka) and the dolerite boulders (∼ 226-356 ka) is too much to be explained by the difference in erosion rates between lithologies. This suggests that this particular granite boulder was weathering much faster than regional average rates, or it may have been affected by post-depositional movement to generate an anomalously young age. Additional exposure age constraints from the Upper Discovery deposit as well as other higherelevation pre-LGM glacial sediments in the McMurdo Sound region will further test if a glacial high-stand occurred during MIS 8.
While our current dataset is limited by the low number of samples and exposure age scatter, maximum ice sheet expansion during the end of MIS 8 is compatible with other Pleistocene glacial records from the McMurdo Sound region. Many of the exposure ages from pre-LGM glacial sediments at lower elevations in McMurdo Sound are either younger or older than those from the Upper Discovery deposit; this suggests that older Pleistocene ice sheets in McMurdo Sound were thicker than during the local LGM but were less extensive than during MIS 8. The AND-1B core in McMurdo Sound records around five glaciations during the last 500 kyr . While poor core recovery in the upper part of AND-1B limits the interpretation of ice sheet behavior during the last glaciation and interglacial, deformed till recovered from the third-to-last glacial period could reflect greater ice thickness in McMurdo Sound during MIS 8 .
There is little evidence for an MIS 8 glacial high-stand preserved in sediments along the margins of EAIS outlet glaciers that flow into McMurdo Sound during glacial periods. Clasts from the head of the Skelton Glacier drainage basin likely record multiple episodes of exposure during interglacials and burial below ice during glacial periods from the Pliocene to Pleistocene, which is compatible with increased ice thickness during MIS 8 (Jones et al., 2017). At lower elevations along Skelton Glacier, most samples record ice-thinning fol-  (Jouzel et al., 2007). (c) Exposure ages (± 1σ error) of Ross Sea drift (yellow circles) and older pre-LGM deposits (orange circles), and the Upper Discovery deposit (purple). Shapes correspond to target nuclide and lithology of each sample. lowing the LGM (Anderson et al., 2020). Due to ice cover, no cosmogenic nuclide studies from Mullock Glacier exist. Although Hatherton and Darwin glaciers transport less material into McMurdo Sound during glacial periods (Talarico et al., 2013), there is also no evidence of MIS 8 glacial sediments preserved along the margins of those glaciers (Joy et al., 2014;King et al., 2020;Storey et al., 2010). In general, exposure ages of pre-LGM glacial sediments above outlet glacier margins record glacial high-stands during MIS 6 and earlier, but there is no apparent evidence for elevated outlet glaciers during MIS 8 (Anderson et al., 2020;Bromley et al., 2010;Hillenbrand et al., 2014;Joy et al., 2014;Staiger et al., 2006;Storey et al., 2010).
At a wider scale, maximum ice sheet extent in McMurdo Sound during MIS 8 may reflect the unique character of this glacial period relative to others of the past 800 kyr. During MIS 8, global ice volume was lower and average eustatic sea level was higher than the LGM (Spratt and Lisiecki, 2016); yet the marine-based ice sheet in McMurdo Sound was larger than its local LGM configuration. In the EPICA Dome C ice core, the glacial termination following MIS 8 (T3) is characterized by an earlier warming between 256-249 ka that coincided with an insolation minimum in the Northern Hemisphere (unlike other terminations of the past 500 kyr), followed by brief cooling between 248-249 ka and then rapid warming until 243 ka (Bréant et al., 2019). The subsequent MIS 7e interglacial, while warm in Antarctica, was globally cooler than other interglacials of the past 450 kyr (Past Interglacials Working Group of PAGES, 2016). It is possible that warmer Antarctic temperatures increased accumulation and that reduced ocean forcing supported a thicker marine ice sheet in the western Ross Sea at the end of MIS 8, similar to the mechanism proposed for the persistence of grounded ice in McMurdo Sound during the last glacial termination (Hall et al., 2015).

Conclusions
Exposure ages of glacial deposits in the McMurdo Sound region record multiple surface processes operating in the Transantarctic Mountains prior to, during, and following glacial entrainment throughout the Pleistocene. Ross Sea drift contains clasts that yield exposure ages during the early Holocene and last glacial period but generally exceed the timing of local LGM ice extent (12.3-19.6 ka) established by an independent radiocarbon chronology. We attribute the scatter and prior exposure in this dataset to surface processes related to bedrock sources for glacially eroded and transported material. Granite clasts, sourced via sub-glacial plucking and/or rock fall, produce scattered exposure ages that are less likely to record prior exposure. Ferrar Dolerite clasts, sourced from exposed cliff faces and high-elevation peaks in the Transantarctic Mountains above outlet glaciers, are more likely to contain inherited nuclides, due to entrainment along cold-based glacier margins. Local volcanic rocks from the McMurdo Sound region produce exposure ages that record landscape exposure prior to ice sheet advance and following ice sheet thinning, which reflects reworking of previously exposed clasts along with clasts derived from submarine sources that lack any previous exposure. 3 He pyx exposure ages of higher-elevation, weathered glacial deposits on Mount Discovery suggest that the largest Pleistocene ice sheet advanced into McMurdo Sound during the end of MIS 8 and early MIS 7. Although the small number of samples, prior exposure, and boulder erosion limits a precise age of such older deposits, it may be rare geologic evidence of Antarctic ice sheet volume during MIS 8, a glacial period marked by generally warmer Antarctic temperature and higher global sea level. Data availability. All data presented in this paper are hosted in a repository at the Open Science Framework (https://doi.org/10.17605/OSF.IO/XH5EW; Christ, 2021).
Author contributions. AJC designed the study, conducted field work, analyzed data, and wrote the paper with contributions from all co-authors. PRB edited the paper and provided mentorship. JLL, JMS, and GW conducted laboratory analyses and edited the paper.
Competing interests. The contact author has declared that neither they nor their co-authors have any competing interests.
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.