The delivery of ice-rafted debris (IRD) from glaciated margins is a function of ice sheet dynamics. Shifts in supply and sourcing of IRD can therefore identify episodes of ice sheet instability; however, records can be difficult to correctly interpret because the subglacial geology of the catchment areas, which controls IRD composition, may be obscured. Importantly, variations can also result from shifts in erosion sites due to changes in the basal ice sheet conditions. This study evaluates where subglacial erosion has occurred in catchments that flow into the southern Weddell Sea, Antarctica, by determining the Pb isotopic compositions of individual ice-rafted feldspars from late Holocene marine sediments. Feldspar compositions match those of rock units inferred (through extrapolation of outcrop, magnetic, and gravity data) to compose areas where ice velocity, bed roughness, and shear stress are high. Significantly, signals from areas where ice velocities are high but bed roughness and shear stresses are low were not recorded, suggesting that there is reduced bedrock erosion in these regions. Major variations in IRD composition in the Weddell Sea can result from changing the loci of subglacial erosion, and do not necessarily correspond with major ice sheet instability.


Determining the source of ice-rafted debris (IRD) in marine sediments can provide independent tests for reconstructing ice sheet responses to environmental changes (e.g., Hemming, 2004). For instance, the geochemical and geochronological signature of mineral grains contained within the so-called “Heinrich” IRD layers deposited in the North Atlantic during the Pleistocene indicate that this detritus was carried via iceberg armadas as the Laurentide ice sheet episodically collapsed through the Hudson Strait (e.g., Broecker et al., 1992). In Antarctica, abrupt shifts in IRD provenance during the Neogene also have been attributed to episodes of ice sheet instability (Williams et al., 2010; Pierce et al., 2011). Antarctic IRD provenance is problematic because, due to limited rock outcrop, the potential source regions are poorly characterized. Consequently, IRD grains from marine sediment core tops collected from the continental margin around Antarctica are used to infer subglacial geology within various drainage sectors (e.g., Roy et al., 2007). However, it is unclear to what extent sedimentary recycling, lithological bias, and mineral fertility control the IRD signal and thus how representative this signal is of the geology within a catchment area. The influence of these factors can be better understood if the sites of subglacial erosion are constrained. Doing so is desirable because the extent and loci of erosion within a drainage basin vary as the ice sheet responds to external and internal forcing (Livingstone et al., 2011; Ehrmann et al., 2011), indicating caution against assuming that a modern-day provenance signal for a particular catchment has been constant through time.

This study evaluates from where recently deposited IRD originated and was eroded beneath ice-stream catchment areas for the southern Weddell Sea, Antarctica. The crust here is varied in its age and composition (Veevers and Saeed, 2012), and several large, potentially highly erosive ice streams flow over a glacially modified topography in the area (Jamieson et al., 2010). By comparing the Pb isotopic compositions of feldspar in IRD adjacent to the Luitpold and Caird Coasts and the front of the Filchner Ice Shelf (Fig. 1) with those from adjacent rock exposures (Flowerdew et al., 2012) and their inferred subglacial extension (interpreted from magnetic and gravity data; Studinger and Miller, 1999; Golynsky and Aleshkova, 2000; Riedel et al., 2012), the areas within the catchments where erosion is focused can be constrained.


Samples of IRD (fraction >150 μm) were separated from seafloor surface sediments retrieved in box cores collected during the James Clark Ross cruise JR244 from the front of the Filchner Ice Shelf (northeast from Berkner Island; BC571, BC580, BC591) and in bays along the Luitpold Coast (BC566, BC607) and Caird Coast (BC628; Fig. 1). The surface sediments from the front of the Filchner Ice Shelf are soft foraminifera-bearing muds whereas sediments collected from the coastal sites are diatom-rich oozes. The interpretation that the sediments reflect (near-) modern deposition with coarser material representing recently deposited IRD is supported by AMS 14C dates from benthic carbonate or planktonic foraminifera from cores BC591, BC607, and BC628, which yielded uncorrected ages between ca. 800 and 2000 yr B.P. These dates are consistent with the modern marine radiocarbon reservoir effect in Antarctic waters that spans ∼800–1300 yr (Berkman and Forman, 1996; see the GSA Data Repository1). An older age of ca. 5800 yr B.P., determined from acid-insoluble organic matter in sample BC566, probably reflects contamination with reworked fossil organic carbon and thus records an erroneously old age (Rosenheim et al., 2008). It is assumed that the IRD along the Luitpold and Caird Coasts was derived from restricted glacial catchments within Coats Land, while the IRD in front of the Filchner Ice Shelf originates from the catchments of Bailey Ice Stream and the Support Force, Recovery and Slessor Glaciers (Swithinbank et al., 1988). Previous mineralogical and petrographical studies on ice-rafted sand- and gravel-sized grains, clasts, and pebbles deposited in these areas have demonstrated that the IRD is derived from local ice and not from far-traveled icebergs (Anderson et al., 1991; Diekmann and Kuhn, 1999). Moreover, cores on the Luitpold and Caird Coasts were collected from small, geographically restricted embayments, making it unlikely that icebergs, calved elsewhere, drifted into these bays and deposited IRD.

The Pb isotopic composition of feldspar from the IRD fraction from the marine sediments, as well as from gravel collected from the englacial Brunt Ice Shelf moraine (Clarkson 1988; Z.1301.29 in Fig. 1), were determined by laser ablation–multicollector–inductively coupled plasma–mass spectrometry (LA-MC-ICP-MS), using a New Wave 193 nm Excimer laser attached to a Thermo-Fisher Neptune instrument housed at the National Centre for Isotope Geochemistry (NCIG), University College Dublin, Ireland. The method followed Flowerdew et al. (2012), and data, together with further analytical details, are given in the Data Repository.


The analyzed feldspars form two broad but distinct populations. One population (206Pb/204Pb between 17.4 and 18.5, and 207Pb/204Pb between 15.4 and 15.7) comprises the majority of feldspars from the front of the Filchner Ice Shelf, and corresponds predominantly to the Shackleton and Pensacola provinces (Fig. 2). The second population is narrower in its isotopic range (206Pb/204Pb between 16.5 and 17.3, and 207Pb/204Pb between 15.2 and 15.7); most of the feldspars from the Luitpold and Caird Coasts and the Brunt Ice Shelf moraine compose this population (Fig. 2). Additionally, there is a shift toward lower Pb isotopic values of the feldspars from north to south along this coastal transect. Feldspars from the englacial moraine Z.1301.29 and core BC628 have high 207Pb/204Pb values (15.4–15.7) and overlap with the Dronning Maud Land province. Feldspars from BC566 and BC607 generally have significantly lower 207Pb/204Pb values (15.35–15.4) which overlap with the composition from the Ellsworth-Whitmore Mountains province (Haag Nunataks), although the feldspars with the lowest 207Pb/204Pb values (15.25–15.3) have no known exposed correlative.


Major tectonic junctions exist in the study area, and the inferred geological terranes are shown in Figure 1. In the southwest, the Cambrian Ross orogeny affected both sedimentary rocks and crystalline basement akin to the Paleoproterozic gneisses exposed in the Read Mountains. In the northeast, the late Neoproterozoic East African–Antarctic orogeny affected the late Mesoproterozoic Maud belt (Jacobs, 2009). In the northwest lies the poorly exposed, magnetically distinct Coats Land block (Studinger and Miller, 1999), a possible fragment of Laurentia that has remained attached to Antarctica (Loewy et al., 2011). In the southeast, the geology is unknown but magnetic anomalies and detrital mineral signatures suggest it partly comprises rocks that are Paleoproterozoic or older (the Recovery and South Pole terranes; Fig. 1).

The glaciers delivering feldspar to the Luitpold and Caird Coasts drain the Coats Land block exclusively. However, none of the feldspars collected from surface sediments along these coasts or the Brunt Ice Shelf have a Pb isotopic composition matching that from the only exposures of the Coats Land block (undeformed late Mesoproterozoic volcanic rocks and granophyres of the Littlewood Nunataks; blue crosses in Fig. 2). Variations in feldspar Pb isotopic compositions from south to north along the Luitpold and Caird Coasts identify a geochemical boundary within the Coats Land block that may have tectonic significance, an observation consistent with petrological studies on clasts in glaciomarine and subglacial sediments (Anderson et al., 1991). The composition of feldspar derived from the hinterland of the Caird Coast suggest the geology in this region comprises late Mesoproterozoic crust unaffected by later orogeny and thus resembles the Vardeklettane terrane west of the Heimfront shear zone in western Dronning Maud Land (Jacobs, 2009). Ice-rafted feldspars derived from the hinterland of the Luitpold Coast have a bimodal distribution. The feldspars with high 207Pb/204Pb values lie within the Haag Nunataks field of the Ellsworth-Whitmore Mountains province, and suggest correlative rocks exist in Coats Land. The second group of feldspars is not represented in any exposed rock from adjacent Antarctica and may represent basement upon which the volcanic rocks exposed at Littlewood Nunataks rest.

Feldspars derived from the Bailey Ice Stream and the Recovery, Slessor, and Support Force glacier catchments have Pb isotopic compositions that indicate they originate from regions where the crust was affected by the Ross or East African–Antarctic orogenies (see below).


Ice streams usually erode a significant quantity of detritus from the underlying substrate and transport this basal debris downstream (Bennett, 2003); however, subglacial erosion is dependent on basal ice-sheet conditions which vary both geographically and temporarily across any catchment area (Smith and Murray, 2009). Streaming areas often have relatively warm ice at the base (Jamieson et al., 2010), or they are underlain by a deforming bed of soft unconsolidated sediments (Alley et al., 1986). The mechanisms for subglacial erosion include plucking and abrasion of bedrock (Glasser and Bennett, 2004), basal freeze-on (Christoffersen et al., 2010), meltwater channel erosion (King et al., 2004), and soft till deformation and shearing (Smith and Murray, 2009). Aside from rare, spatially restricted geophysical in-situ surveys (Smith and Murray, 2009), insights on the sites and mechanisms of subglacial erosion on a regional scale are based on modeling basal shear stress and subglacial roughness from integrating remote sensing with in-situ data. High basal shear stresses (Joughin et al., 2006) and subglacial roughness (Bingham and Siegert, 2009) indicate areas where erosion of a hard lithified sedimentary or crystalline bedrock substrate may occur.

Considering the present ice drainage pattern (Swithinbank et al., 1988), it would be anticipated that (1) IRD at site BC571 is derived mainly from Berkner Island and Support Force Glacier, and (2) IRD at sites BC580 and BC591 are predominantly supplied from Recovery Glacier. The compositions of detrital feldspars recovered along the front of the Filchner Ice Shelf are entirely consistent with erosion from sites where ice velocities, basal shear stresses, and bed roughness are all high within these catchments (Fig. 1). For Recovery Glacier, such conditions are found near its grounding zone and just east of the Shackleton Range. For the Support Force catchment (BC571), such conditions exist immediately south of the Pensacola Mountains. The bedrock beneath the candidate erosion sites is inferred to contain feldspar with compositions that overlap either the Shackleton or Pensacola Pb isotope provinces, and the majority of the feldspars have such compositions (Fig. 2). It follows that areas with high basal shear stress (Joughin et al., 2006) and bed roughness (Bamber et al., 2006), coupled with high ice velocities in the other catchments, might also be the source of the ice-rafted feldspar (Fig. 1). The breadth of compositions recorded are, in this case, not consistent with marine basal freeze-on (Nicholls et al., 2012) being the dominant mechanism of feldspar capture and delivery.

Additional supporting evidence comes from the absence of ice-rafted feldspars with (1) low μ values (derived from rocks with low 238U/204Pb ratios) at site BC571 extensive to the interior portions of the Support Force Glacier catchment (Flowerdew et al., 2012); and (2) Read Mountains Pb isotopic composition (Fig. 2) at sites BC580 and BC591. South of the Shackleton Range, the Recovery Glacier flows over highly magnetic crust with geophysical properties resembling bedrock exposed in the Read Mountains (Golynsky and Aleshkova, 2000) suggesting that Read Mountains–type crust extends beneath the Recovery Glacier. This part of the Recovery ice stream is characterized by a high flow velocity and low shear stress, possibly due to the presence of a weak unconsolidated sedimentary substrate (Joughin et al., 2006), which would prevent direct erosion of the underlying bedrock. If these sediments are of marine origin as opposed to reworking of underlying bedrock (Joughin et al., 2006), their erosion and downstream transport would not result in the delivery of Read Mountains–type feldspar to the Weddell Sea, consistent with our findings.

A response of basal ice-stream conditions to changes in external forcing (Livingstone et al., 2011) would modify the main locations of bedrock erosion within a catchment. The age and geochemical signature of IRD could significantly shift if the new sites of erosion corresponded to geologically different bedrock. Rippin et al. (2006) inferred from the pattern of ice-flow variability within the tributaries of Slessor Glacier that such tributaries may have the ability to switch on and off, depending on the distribution of subglacial water and sediment. If correct, then it is possible that crust with a Read Mountains Pb isotopic composition was eroded by the Recovery Glacier in the past. This implies that the IRD signature of a single catchment area may change with time, even if the catchment itself does not. Thus, reconstructing ice sheet evolution histories by interpreting the origin of past IRD based on the modern geochemical fingerprints for catchment areas may not be appropriate in every case.


The Pb isotopic composition of ice-rafted feldspars can be used to constrain sites of subglacial erosion by integrating ice sheet and bed conditions (bed roughness, shear stress, ice velocity) with the geophysical and geochemical properties of the underlying rocks. This approach shows that late Holocene subglacial bedrock erosion from southern Weddell Sea catchments was not uniform but occurred in regions of high bed roughness, shear stress, and ice velocity. Areas where ice velocity is high, but bed roughness and shear stress are low, were identified as sites where bedrock erosion was reduced. These regions are underlain by rocks with a distinct feldspar Pb isotopic composition, one which was not recorded in the recently deposited IRD. The results thus have wider implications for using the age and composition of IRD when investigating the response of ice sheets to environmental change: IRD signals delivered from catchments can change with shifts in subglacial erosion sites which do not necessarily correspond with episodes of ice sheet instability.

This study is part of the British Antarctic Survey Polar Science for Planet Earth program funded by the UK Natural Environment Research Council. Tyrrell is funded by a Griffiths Geoscience Award. NCIG is funded mainly by Science Foundation Ireland. Help from the crew and staff on cruise JR244 are greatly appreciated. Tom Culligan and Mike Tabecki are thanked for their skilled sample preparation. Comments by C.-D. Hillenbrand on an early version, and by J. Jacobs, two anonymous reviewers, and editor Ellen Thomas are greatly appreciated.

1GSA Data Repository item 2013036, Tables DR1 and DR2, is available online at www.geosociety.org/pubs/ft2013.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.