Ocean‐Forced Instability of the West Antarctic Ice Sheet Since the Mid‐Pleistocene

Evidence on West Antarctic Ice Sheet (WAIS) instability through Pleistocene glacial/interglacial cycles can provide fundamental constraints on interactions between the climate system and cryosphere. To explore such ice sheet‐ocean‐climate processes on orbital timescales over the last ∼770 ka, we provide continuous records of iceberg‐rafted debris (IRD) content and clay mineralogy, supported by detrital Sr‐Nd isotopes from the pronounced IRD peaks, in gravity core ANT34/A2‐10 from the Amundsen abyssal plain. The IRD record reveals interglacial WAIS instability since ∼770 ka, while comparison to the clay mineralogy record and published records of regional oceanic and atmospheric forcing suggests a temporal link with a strengthened Antarctic Circumpolar Current, enhanced deepwater ventilation, and poleward‐shifted southern westerly winds. In addition, the Sr‐Nd isotope signature of the detrital sediments indicates a shift in provenance around Marine Isotope Stage (MIS) 16, potentially linked to regional oceanic circulation changes. We suggest that an expanded Ross Gyre was important for controlling iceberg trajectories and sediment transport to the site before MIS 16, whereas modern‐like iceberg trajectories were established after MIS 16, probably related to a poleward shift of the Amundsen Sea Low after the end of the Mid‐Pleistocene Transition. This reorganization of the ocean and atmospheric circulation was followed by an interval of enhanced WAIS variability during MIS 15 to 13, which was linked to strong orbital and ocean forcing. These insights into the role of ocean‐atmosphere forcing on the past behavior of the WAIS may improve our framework for understanding future changes in this region.

Here, to explore the connections between high-latitude ocean-atmosphere processes and WAIS dynamics on orbital timescales, we present semi-quantitative iceberg-rafted debris (IRD) abundance (wt% and grains/g) and clay mineralogy records from a sediment core in the Pacific sector of the Southern Ocean spanning the last ∼770 ka. By comparing these records to published benthic carbon isotope (δ 13 C) records (Hodell et al., 2003;Ullermann et al., 2016), deep ocean ventilation proxies (Tang et al., 2022), Antarctic Circumpolar Current (ACC) flow-speed reconstructions (Toyos et al., 2020;S. Wu, Lembke-Jene, et al., 2021), and Antarctic ice core data constraining Southern Westerly Wind (SWW) migration (Fudge et al., 2013;R. F. Anderson et al., 2009), we assess the role of ocean forcing in driving destabilization of the WAIS. In addition, we use the combination of clay minerals and Sr-Nd isotope data on bulk detrital sediments and IRD samples to provide constraints on iceberg sourcing and pathways and to investigate changes in the high-latitude coupled atmospheric-oceanic circulation patterns during those peak IRD intervals. 4 of 27 S. Jacobs et al., 2012), while the Antarctic Coastal Current in the surface ocean is driven by the easterly winds and flows westward parallel to the Antarctic coastline ( Figure 1b).
Most layers of the ACC are occupied by CDW (Orsi et al., 1995;Rickli et al., 2014), a relatively warm, high-salinity water mass that generally circulates along the pathways of the ACC and the Ross Gyre (Orsi et al., 1995;Figure 1b). Some CDW upwells south of the SBACC onto the Ross Sea shelf and other narrower continental shelf areas, where it is ultimately modified by sea-ice formation and feeds into the formation of AABW (Orsi et al., 1999). Moreover, the modification of the incoming CDW near the outer edge of the Ross Sea produces MCDW (Figure 1b), which is warmer (temperatures up to 0.95°C) than surrounding waters and has a salinity that is intermediate between the surface and deep water masses (Orsi & Wiederwohl, 2009). This MCDW is the primary source of heat, salt, and nutrients to the Ross Sea continental shelf region (Budillon & Spezie, 2004;Hiscock, 2004;W. O. Smith et al., 2006). Such local incursions of CDW and MCDW far onto the inner continental shelves toward the WAIS margins in the Amundsen and Ross Seas (Das et al., 2020;S. Jacobs et al., 2012;Figure 1b) make the South Pacific marine-based WAIS highly sensitive to ocean thermal forcing.

Material
Gravity core ANT34/A2-10 is 4.54 m long and was split lengthwise and logged in detail by visual examination. Its lithology is characterized by mainly pelagic silty clay and occasional volcaniclastic material, with a foraminifer-rich layer at 11-29 cm, and frequent occurrence of radiolarians from 0 to 110 cm ( Figure 2). The entire core is dominated by silty clay with no significant sedimentary structures, except for thin planar laminations (less than 3 mm thick) at 401-454 cm depth, which may reflect sorted silts or possibly diatom mat deposition ( Figure 2). There is no clear evidence of turbidite sedimentation or mass redeposition, with the possible exception of the laminated section from 401 to 454 cm as noted above. Large clasts were found at 234-238, 338-340, and 346-350 cm ( Figure 2). The entire core was segmented at 2 cm intervals to analyze IRD content and clay mineralogy, and Sr-Nd isotopic composition was measured on selected bulk detrital samples.

Analytical Methods
Approximately 5-10 g of original wet samples from core ANT34/A2-10 were reacted with 10% acetic acid and 3.5% hydrogen peroxide to remove carbonate and organic matter, respectively, before being dried, weighed, and separated by wet sieving (150 μm) to obtain the operationally defined IRD component (Caniupán et al., 2011). The >150 μm fraction was dried under 55°C, and the 150 μm-2 mm fraction was obtained by dry sieving through a 2 mm mesh. The semi-quantitative IRD weight percentage (IRD wt%) was calculated by dividing the weight of the coarse fraction (150 μm-2 mm) by the dried bulk carbonate-and organic matter-free sample weight (Caniupán et al., 2011;L. Wu, Wilson, et al., 2021). In addition, the number of particles in the 150 μm-2 mm fraction was counted under a binocular microscope (LEICA S8AP0), with subangular to subrounded quartz and feldspar grains and rock fragments considered to represent the main IRD components (Kanfoush et al., 2000;Starr et al., 2021;Teitler et al., 2010;Watkins et al., 1974); these data are reported as the semi-quantitative IRD content (IRD grains/g). Although radiolarian shells with sizes of 150 μm-2 mm frequently appeared in some layers of ANT34/A2-10, their abundance and weights were too low to contribute significantly to the IRD wt% record. Furthermore, the opal content in the bulk sediment was typically ∼12% and always less than 20%, so any changes in opal content do not significantly contribute to the downcore patterns in the IRD proxies. We also note Figure 1. Geographic and oceanographic setting of the study area: (a) location map of the relevant core sites and regional sediment sources; (b) three-dimensional topographic map showing the core sites and major ocean currents; (c) modern 10 m sea surface winds in the study area, from selected days in austral summer (left) and winter (right) (vectors indicate wind direction; color shading indicates wind strength); and (d) hydrographic section showing temperature structure of the Amundsen Sea. The ocean fronts in panels (a and d) are from Orsi et al. (1995). The dashed line in panel (a) represents the boundary for panel (b). The wind stress data in panel (c) are from the website (https://coastwatch.pfeg.noaa.gov), and the black arrows represent the approximate wind stress direction. Temperature data in panel (d)  that volcanic particles and micronodules were not a significant component in most layers of core ANT34/A2-10 and are unlikely to contribute significantly to downcore variations in the IRD proxy records. This point is also supported by the good agreement between the IRD counts, where such components could be manually excluded, and the IRD wt% records.
For clay mineralogy, 227 samples were processed to obtain the <2 μm fraction, with separation based on Stokes' settling velocity principle, after removing carbonate and organic matter by reacting with acetic acid and excess H 2 O 2 (S. Wan et al., 2010). X-ray diffraction analysis of the samples was performed with a Brucker D8 ADVANCE diffractometer, using oriented mounts and CuKα radiation (40 kV, 40 mA), in the Key Laboratory of Marine Geology and Environment at the Institute of Oceanology, Chinese Academy of Sciences. Each sample was measured two times after ethylene-glycol solvation (i.e., (a) 3°-30° 2θ, 0.02° steps; (b) 24°-26° 2θ, 0.01° steps) to identify the double peak of 3.58/3.54 Å (chlorite/kaolinite) (S. Wan et al., 2006Wan et al., , 2007. Semi-quantitative estimates of the major clay mineral groups (smectite, kaolinite, illite, and chlorite) were based on the integrated peak areas of the  (Lisiecki & Raymo, 2005), and sedimentation rates (on age scale). Vertical gray dashed lines are age-depth tie points from Tang et al. (2022). Gray crosses: tie points based on Ba/Ti. Red cross: organic carbon AMS 14 C date on bulk acid-insoluble organic carbon. Blue cross: foraminiferal AMS 14 C date. See Tables S1 and S2 in Supporting Information S1 for further details on the age model. characteristic basal reflections (smectite-17 Å, illite-10 Å, and chlorite/kaolinite-7 Å) in the glycolate state, using Topas 2P software with the experimental factors of (Biscaye, 1965). The relative proportions of chlorite and kaolinite were determined based on the ratio of the 3.58/3.54 Å peak areas (chlorite/kaolinite) in the glycolate state. Relative clay mineral abundances are given in percent, with the analytical precision (relative standard deviation) for the abundance of each clay mineral estimated to be approximately ±5% (S. Wan et al., 2010). The illite chemical index was calculated from the ratio of the 5 and 10 Å illite peak areas (Esquevin Index) in the glycolate state (Esquevin, 1969). Ratios higher than 0.4 represent Al-rich illite formed under strong hydrolysis, whereas ratios lower than 0.4 correspond to Fe-Mg illite, which is a product of physical rock weathering (Ehrmann, 1998;Gingele et al., 2001). The illite and smectite crystallinity were estimated from the full width at half maximum (FWHM) of the illite 10 Å peak and smectite 17 Å peak, respectively. Generally, high FWHM values indicate poor crystallinities, whereas low FWHM values indicate good crystallinities.
Detrital strontium (Sr) and neodymium (Nd) isotopes were analyzed on 11 bulk sediment samples (∼100 mg), predominantly taken from IRD peak intervals, and on 3 IRD (>250 μm) samples (∼50 to 100 mg) from IRD peak intervals. The samples were processed in the Key Laboratory of Marine Geology and Metallogeny, First Institute of Oceanography, Ministry of Natural and Resources (FIO). First, the samples were ground and leached to remove any authigenic ferromanganese oxide phases using 1 M hydroxylamine hydrochloride/25% buffered acetic acid at 90°C for 3 hr. They were then transferred to Teflon beakers and digested in a mixture of 2 mL HF, 1.5 mL HNO 3 , and 0.2 mL HClO 4 on an electric heating plate at 120°C for about a week until completely dissolved. After evaporation at 180°C, the samples were dissolved in 2.5 M HCl and the supernatant after centrifugation was processed through ion-exchange columns, with Sr separated using AG50W-X12 resin and Nd using LN-B50-A resin.
The Sr and Nd isotopic compositions were analyzed on a NuInstruments Plasma 1 HR multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS), in the FIO, or on a ThermoScientific Neptune MC-ICP-MS in the Key Laboratory of Submarine Science and Prospecting Technology, Ministry of Education, Ocean University of China (OUC;  (Jacobsen & Wasserburg, 1980). Typical bulk detrital samples yielded 2-3 μg Nd, whereas typical Nd blanks in this laboratory are ∼1 pg Nd, so blank contributions are negligible.

Age Model
Previous studies of Pleistocene sediments from the Antarctic margin have developed orbital-scale age models using correlations between surface-water productivity proxies (such as biogenic opal, silica, Ba/Ti ratios, or XRF-scanning Ba/Ti count ratios) and the LR04 benthic δ 18 O stack (Ceccaroni et al., 1998;Hillenbrand et al., 2009;Presti et al., 2011;Wu et al., 2017), often in combination with biostratigraphy or magnetostratigraphy. In a recent study (Tang et al., 2022), the age model of core ANT34/A2-10 was established by correlating the XRF-scanning Ba/Ti ratios to the LR04 benthic δ 18 O stack (Lisiecki & Raymo, 2005). The age model uses 26 age control points based on comparison of the Ba/Ti ratios to the LR04 δ 18 O stack and two AMS 14 C age control points (planktonic foraminifera and bulk organic carbon) (Figure 2), and was not further modified here. For detailed information on the age model, see Tables S1 and S2 in Supporting Information S1.
According to the age model, core ANT34/A2-10 spans the last ∼770 ka, with an average sedimentation rate of 0.59 cm/kyr ( Figure 2). In comparison, the average sedimentation rate over this interval in core PS58/254 was 1.8 cm/kyr, and sedimentation rates in the Pleistocene intervals of cores U1532 and U1533 were 2.0 and 1.4 cm/ kyr, respectively (Gohl et al., 2021). All of these three cores were collected nearby, at similar depths but closer to the shelf break than our site (Figures 1a and 1b). We note that we are unable to provide biostratigraphic constraints on core ANT34/A2-10, which could lead to some uncertainty in the age model, particularly in its deeper sections or where sub-orbital timescales are considered. Nevertheless, Tang et al. (2022) verified the reliability of this age model based on the consistency between the downcore magnetic susceptibility record of core ANT334/A2-10 and the dust concentration record from the EPICA Dome C (EDC) ice core (Tang et al., 2022). In addition, the similar glacial/interglacial cyclicity of the sedimentary records between core ANT34/A2-10 and core PS58/254 (Hillenbrand et al., 2009;Konfirst et al., 2012) provides further reasonable validation for the reliability of this age model over glacial/interglacial timescales (see Text S1 and Figure S1 in Supporting Information S1).

Variability in IRD Inputs Through Glacial/Interglacial Cycles
The record of IRD counts (grains/g) is similar to the record of IRD wt% (150 μm-2 mm) ( Figure 3a), with R 2 = 0.72. Most of the IRD consists of sand-sized quartz and feldspar grains, with only occasional minor amounts of gravel ( Figure 2). Both records show higher abundance during interglacial periods than glacial periods, with peak counts of up to ∼640 grains/g and ∼3.6 wt% ( Figure 3a). The highest amplitude variations and largest IRD peaks occurred at ∼474 to 530 ka (Marine Isotope Stage [MIS] 13) and ∼550 to 580 ka (MIS 15).
The coretop age of ANT34/A2-10 is ∼8.8 ka (Table S1 in Supporting Information S1), indicating a hiatus that likely reflects sample loss or disturbance during piston coring. As such, the IRD abundance in the Holocene layer should be taken with some caution and may not provide a good baseline for understanding the past behavior. We therefore identified IRD peak intervals' based on the assumptions that (a) the high IRD content predominantly relates to warm interglacial periods and can be compared to a glacial baseline and (b) the IRD peak intervals should be characterized by a higher IRD content than the average IRD content of glacial periods (i.e., 146 grains/g and 0.74 wt%) (Figures 4a and 4b).
Having identified the major orbital-scale IRD peak intervals, labeled as I1 to I12 (Figure 4a), we determined their boundaries by comparison to the glacial average values for the IRD proxies. Nearly all IRD peak intervals occur during interglacial periods, with only two IRD peak intervals appearing to occur during glacial periods; these latter two IRD peaks (I11 and I12) are found near the base of the core during MIS 18, according to the age model.
Smectite and illite crystallinity values show only minor fluctuations before MIS 6, but exhibit higher frequency and higher amplitude fluctuations since MIS 6 ( Figures 3f and 3g). The smectite and illite crystallinity values range from 1.17 to 1.82 Δ°2θ (average 1.38) and 0.20 to 0.44 Δ°2θ (average 0.28), respectively, indicating high to moderate crystallinity for smectite and very high to high crystallinity for illite (Figures 3f and 3g;Ehrmann et al., 2005). The illite chemical index is consistently less than 0.3, with an average of 0.21, indicating strong physical rock weathering in the source area ( Figure 3h; Ehrmann, 1998;Gingele et al., 1998;S. Wan et al., 2006). Furthermore, glacial/interglacial trends in clay mineralogy and chemical parameters (Figures 3b-3h) show systematic variations that coincide with the IRD peak intervals (Figure 3a), with high IRD content generally coinciding with high smectite and low illite content.

Detrital Strontium and Neodymium Isotopes
The Sr and Nd isotopic compositions of bulk detrital samples and the IRD samples (>250 μm) from the IRD peak intervals (sampling shown on Figure 4a) are reported in Table 1. The bulk detrital Sr-Nd isotopic compositions . Records of iceberg-rafted debris (IRD) content and Sr-Nd isotopic data from selected IRD peak intervals in core ANT34/A2-10. Records of (a) IRD abundance (grains/g), (b) IRD abundance (wt%), and (c) LR04 benthic δ 18 O stack are shown with 95% bootstrap confidence intervals, calculated using a "loess" smoothing method and 1,000 bootstrap simulations in Acycle (M. S. Li et al., 2019). The IRD peaks are shaded red and labeled I1-I12, where "I" represents IRD peak. The gray lines in panels (a and b) indicate the average IRD content in the glacial intervals of the record (i.e., 146 grains/g in panel (a) and 0.74 wt% in panel (b)). The yellow circles and green crosses in panel (a) indicate the samples analyzed for Sr-Nd isotopes in bulk detrital sediments (plotted in panel (d)) and IRD grains (plotted in panel (e)). For Nd isotopes, uncertainties (2 SD) are plotted as error bars or are smaller than the symbol size where not shown. For Sr isotopes, uncertainties are smaller than the symbol size for all samples. Glacial periods are numbered with Marine Isotope Stage numbers along the top.
range from 87 Sr/ 86 Sr = 0.710267-0.713400 and ε Nd = −2.5 to −5.0. The IRD sample compositions range from 87 Sr/ 86 Sr = 0.707030-0.717786 and ε Nd = +1.2 to −8.0. As such, the IRD samples show a greater range of isotopic variability between samples, and typically differ from the contemporaneous bulk detrital sediment, but both records indicate similar trends through time (Figures 4d and 4e). Before MIS 16, 87 Sr/ 86 Sr values were higher (more radiogenic), and ε Nd values were lower (less radiogenic) in both fractions compared to the interval since MIS 16 (Figures 4d and 4e). We also note that the bulk detrital Sr-Nd isotope signature of the glacial MIS 2 sample ( 87 Sr/ 86 Sr = 0.710295, ε Nd = −3.5) is within the range of bulk detrital samples from the interglacial IRD peak intervals after MIS 16 ( 87 Sr/ 86 Sr = 0.710267-0.710815, mean of 0.710571; ε Nd = −2.5 to −3.9, mean of −3.3).

Interpretation of the IRD Record
The abundance of IRD in Southern Ocean sediment cores is usually considered to reflect Antarctic Ice Sheet dynamics (Kanfoush et al., 2000;Nielsen et al., 2007;L. Wu, Wilson, et al., 2021;Wilson et al., 2018). However, the IRD content is a mixed signal that is ultimately determined by the flow of icebergs containing debris but that gets further modulated by transport processes and climate; as such, it is not a direct record of ice margin instability. Close to Antarctica, and depending on the local setting, IRD deposition is generally highest during deglacial and interglacial ice-sheet retreat (Hillenbrand et al., 2021;Weber et al., 2014;Wilson et al., 2018) or during glacial ice-sheet advance and collapse (Hillenbrand et al., 2002(Hillenbrand et al., , 2009L. Wu, Wilson, et al., 2021). Nevertheless, apart from iceberg calving rates, IRD records can also be influenced by other factors, such as iceberg trajectories, iceberg drift distance, sea surface temperatures, and variable debris content linked to different iceberg source areas (e.g., L. Wu, Wilson, et al., 2021 and references therein), which need to be considered on a site-by-site basis.
Modern icebergs travel anti-clockwise around Antarctica within the Antarctic Coastal Current (Gladstone et al., 2001). The Coriolis force helps to keep them entrained in that current close to the edge of the Antarctic continent (K. L. Smith, 2011), although topographically steered ocean currents allow some of them to stray north . A similar trajectory probably existed during previous interglacial periods because the Antarctic continental margin geographically constrains the SBACC in the Amundsen Sea sector (Yamazaki et al., 2021;Figure 1b), while the easterly winds driving the Antarctic Coastal Current are largely insensitive to climate .
Ocean thermal forcing could have induced enhanced ice melting at the grounding line at times of warm CDW or MCDW incursion, but this process would not have significantly influenced the mass loss of drifting icebergs. Icebergs may lose some mass during long-distance drifting (e.g., through wave erosion, friction, or sea-ice collisions), but the mass-loss rates remain low until they reach the warmer waters of the ACC . Site ANT34/A2-10 is located close to the SBACC, so debris-laden icebergs from the Ross and Amundsen Seas (Hillenbrand et al., 2002) could readily survive to reach site ANT34/A2-10, even under slightly warmer global climate states than modern, before their melting rate increases and they release most of their debris in response to the warmer sea surface temperatures of the ACC.
Whereas elevated IRD contents during the past interglacials probably reflect enhanced calving, the lower IRD content of glacial intervals could be explained by a combination of both restricted iceberg calving under colder Antarctic temperatures and reduced iceberg melt rates during transport in colder waters. At such times, icebergs may have survived transport to more distal locations at lower latitudes (Bigg, 2020). In addition to regional cooling, an equatorward shift of the SWW during colder intervals could have led to the northward migration of the SBACC and SACCF (Civel-Mazens et al., 2021). This scenario could cause local sea surface temperatures at site ANT34/A2-10 to be as cold as those in the Antarctic Coastal Current, thereby increasing the iceberg survivability and decreasing the amount of debris released as they pass site ANT34/A2-10.
It is also important to consider that icebergs derived from different settings (e.g., Alpine-style glaciers vs. large ice shelves) may contain very different IRD contents (J. A. Smith et al., 2019;J. B. Anderson, 1999). Of relevance to the present study is that Alpine-style glaciers from the hinterlands along Dry Valleys/North Victoria Land, Ruppert and Hobbs Coasts, Marie Byrd Land, and the Antarctic Peninsula ( Figure 1) contain significant supraglacial/englacial debris, so past retreat of the marine-based WAIS would be expected to lead to a major discharge of debris-laden icebergs. As such, we consider that the IRD content should be a useful marker for large-scale instabilities, while being relatively insensitive to changes in the ice shelves, since these tend to release relatively debris-free icebergs. Isotopic provenance analysis can also be used to place constraints on how changes in iceberg sourcing could have affected the IRD records.
Although intense ACC flow speeds, particularly during past interglacials (Toyos et al., 2020; Figure 5c), could potentially remove silt and clay particles and thus concentrate particles in the >150 μm size fraction, site ANT34/A2-10 is bathed by AABW and hence should be insensitive to such a direct influence from the ACC (Figure 1b). Furthermore, there is also no evidence for a strong winnowing influence on the record, because sedimentation rates during interglacial IRD peak intervals do not appear unusually low (Figure 2). Therefore, although we have not calculated IRD mass accumulation rates, we consider the abundance of 150 μm-2 mm particles (counts/g or wt%) to be a robust proxy indicator of IRD inputs to this core. In addition, this site is sufficiently far from the Antarctic continental margin (Figure 1) to prevent a single outlet glacier or a single iceberg dumping event from exerting a major control on the IRD records, while a minimal influence from turbidity currents is also envisaged based on its relatively distal setting and the regional seafloor bathymetry (Figure 1b). Previous studies from the adjacent site PS58/254 indicate that its IRD record contains significant orbital variations, thereby confirming that iceberg calving in the study area is not simply a stochastic process (Hillenbrand et al., 2009). Therefore, we suggest that the IRD abundance peaks in core ANT34/A2-10, which occurred predominantly during interglacials (Figure 5a), were driven by increased iceberg fluxes related to the past instability of the WAIS.

Clay Mineral Provenance Changes Driven by Ocean Current Strength
Two important assumptions in interpreting past clay mineral variations in the study area are that the clays are overwhelmingly sourced from the Antarctic region (Hillenbrand & Ehrmann, 2005;Perez et al., 2021) and that modern potential source areas (PSAs) are a good indication of past sources. Stability of source areas is a reasonable approximation for the period of interest (i.e., since ∼770 ka) given a lack of notable tectonic activity in Antarctica since the early Pleistocene (Hillenbrand & Ehrmann, 2005;Perez et al., 2021). Our interpretation of a dominant Antarctic source of clay minerals is also supported by the narrow range of smectite crystallinity (1.17-1.82 Δ°2θ), illite crystallinity (0.20-0.44 Δ°2θ), and illite chemical index (0.12-0.34) in core ANT34/A2-10 (Figures 3f-3h), which are consistent with derivation from a cold and arid environment. Notably, these values are comparable to the smectite and illite crystallinity and illite chemical index in clays from the Transantarctic Mountains, which range from 0.68-2.87 Δ°2θ, 0.20-0.94 Δ°2θ, and 0.05-1.75, respectively . They are also comparable to the illite crystallinity and chemical index in coretop samples of West Antarctic continental margin clays, which range from ∼0.3-0.6 Δ°2θ and ∼0.1-0.55, respectively (Ehrmann et al., 2011;S. Wu et al., 2019).
In the study area, clay minerals are transported into the marine environment both by entrainment in ocean currents and by drifting sediment-laden icebergs (Hillenbrand et al., 2002), with little dust input from the West Antarctic (Petschick et al., 1996). Site ANT34/A2-10 is sufficiently far from the hinterland of the Bellingshausen, Amundsen, and Ross Seas (Figure 1a) that clay minerals are not supplied directly by glaciers, but via transport in the vigorous ACC and in sediment-laden icebergs. Therefore, past changes in the ACC, or in iceberg provenance and trajectories, could influence the clay mineral assemblages in the core.
We use a ternary diagram with endmembers of (smectite + kaolinite)-illite-chlorite to identify the provenance of clay minerals at site ANT34/A2-10 by comparison to PSAs from the Pacific sector of Antarctica, including the relatively proximal Amundsen and Bellingshausen Seas, the West Antarctic Peninsula to the east, and the Ross Sea and East Antarctica to the west (Figure 6a). The clay mineral assemblages in core ANT34/A2-10 lie close to the clay endmembers from the Amundsen, Bellingshausen, and Ross Seas, suggesting their derivation from a mixture of these sources, while the chlorite content is too low to accommodate a significant West Antarctic Peninsula source (Figure 6a). During glacial intervals with low IRD content, the clay mineralogy is well-explained by mostly local sources from the Amundsen and Bellingshausen Seas (probably dominated by the Amundsen Sea). In contrast, the interglacial IRD peak intervals were accompanied by a clear trend toward higher smectite + kaolinite content, consistent with clays supplied from a region to the west of site ANT34/A2-10, such as the WRS (e.g., CRP-1; Ehrmann et al., 2005) and/or offshore Adelie Land in East Antarctica (e.g., Site U1359; Pant et al., 2013; Figure 6a), which are located upstream in the pathway of the ACC relative to site ANT34/A2-10. The clays in the WRS Embayment initially originated from the Transantarctic Mountains and are enriched in smectite with very low kaolinite contents , while the Amundsen Sea and Bellingshausen Sea sources have lower smectite content and more kaolinite, as well as a higher illite content (Ehrmann et al., 2011;Hillenbrand et al., 2003). Although the clay mineral assemblages in Pleistocene sediment from Site U1359 (Pant et al., 2013) show higher smectite + kaolinite values than sediments from the Amundsen and Bellingshausen Seas, they are not sufficiently high on their own to explain the values recorded in core ANT34/A2-10 (Figure 6a), while their smectite/kaolinite ratios are too low for them to be a major contributor to ANT34/A2-10. Overall, we consider that both smectite/kaolinite and (smectite + kaolinite)/(illite + chlorite) ratios could provide a tracer for west-sourced clays relative to local clay inputs, and suggest that the former could reflect a mixture of predominantly Ross Sea with potentially some East Antarctic sources.
A recent study indicates a high smectite and illite content in surface sediments from Terra Nova Bay, adjacent to Victoria Land in the WRS (Figure 1a), which were linked to sources from its surrounding volcanic rocks and glacial meltwater, respectively (Figure 6a; Jung et al., 2021). However, the clay mineral compositions recorded during both glacial and interglacial intervals in core ANT34/A2-10 are quite different from the clay mineral endmember in Terra Nova Bay (Figure 6a). A lack of sediment input to ANT34/A2-10 from this source is probably due to the deep channel in Terra Nova Bay (Jung et al., 2021), which may trap fine-grained detritus and minimize its transport out of the Ross Sea. Moreover, the outlet glaciers draining into McMurdo Sound and the Dry Valleys (adjacent to the CRP-1 core) could provide a greater input to the Ross Sea than the more localized outlet glaciers supplying Terra Nova Bay, so that it may be variability in those inputs that drives the variability in our clay mineral record at times when the Ross Ice Shelf has significantly retreated.
In another recent study, Andrews and LeMasurier (2021)  Therefore, these studies appear to further support our interpretation of these clay mineral ratios in terms of clay mineral provenance changes between the more proximal sources in the Amundsen and Bellingshausen Seas and the more distal sources to the west.
We compare the temporal changes in the smectite/kaolinite and (smectite + kaolinite)/(illite + chlorite) ratios at site ANT34/A2-10 ( Figure 5b) with the sortable silt record from core PS97/93-2 (Figure 5c), which is located in the pathway of the ACC in the Pacific sector of the Southern Ocean but further downstream (Figure 1a; Toyos et al., 2020). This comparison reveals a clear correlation between the sortable silt and clay mineralogy records, indicating that stronger eastward ACC transport during interglacials was associated with increased inputs from westerly sources such as the Ross Sea, that is, higher smectite/kaolinite and (smectite + kaolinite)/(illite + chlorite) ratios (Figures 5b and 5c). As a corollary, weaker ACC flow during glacial periods would have supplied less west-sourced clay to our site, leading to a greater dominance of more local terrigenous inputs with higher kaolinite content derived from the Amundsen Sea shelf. The supply of such material to the Amundsen Sea continental slope during Late Pleistocene glacial periods when terrigenous ice crossed the Amundsen Sea shelf has previously been proposed, based on the observation of low smectite/kaolinite ratios in fine-grained (<2 μm) detrital sediment records from that region (Hillenbrand et al., 2002(Hillenbrand et al., , 2003. In addition, contour currents of the glacial Ross Sea Bottom Water may have provided a mechanism to help deliver this low smectite/kaolinite signal from the continental margin of the Amundsen Sea to site ANT34/A2-10.

Detrital Sr-Nd Isotopic Signature Since MIS 16
The Sr-Nd isotopic composition of bulk detrital sediments and IRD samples in ANT34/A2-10 are compared to data from regolith, glacial drift, sediment detrital fractions (<63 μm), and bulk rocks representing PSAs from the Antarctic margin in the South Pacific sector in Figures 6b and 6c. This comparison allows us to determine the possible terrigenous sources to this site, and ultimately their relation through time to oceanic and atmospheric circulation patterns.
Overall, the detrital continental margin sediments reflecting the PSAs of West Antarctica (circles in Figure 6b) have ε Nd values ranging from −11.9 to +1.3 and 87 Sr/ 86 Sr ratios ranging from 0.705 to 0.750 (Simões Pereira et al., 2018, 2020. Note that there is also heterogeneity between samples collected from different locations within a given coastal region. The bulk detrital Sr-Nd isotopic compositions analyzed from IRD peak intervals since ∼620 ka (I1, I2, I3, I5, I6, I7, and I8) and from MIS 2 (∼20 ka) in core ANT34/A2-10 (Table 1) lie close to the signatures of the EAS (Thwaites Glacier), WAS, WRS, and Bellingshausen Sea (Figure 6b). Hence, these data appear to implicate a mixture of sediments from some combination of those marginal seas. In principle, transport could occur by grounded ice flow and turbidity currents from those shelf regions to the continental rise (J. B. Anderson et al., 1984;Lowe & Anderson, 2002), followed by bottom current reworking and redeposition on the deep continental rise. Since the bottom contour current and the ACC experienced significant variations during glacial/interglacial cycles, these mechanisms might be expected to have generated differences in the proportions of sediment derived from different PSAs through time. For example, the clay mineralogy results suggest that a weak ACC and vigorous AABW Figure 6. Sediment provenance analysis of clay minerals and Sr-Nd isotopes in core ANT34/A2-10. (a) Ternary diagram of (smectite + kaolinite)-illite-chlorite, showing clay mineral compositions in core ANT34/A2-10 during glacial intervals with low iceberg-rafted debris (IRD) content and during IRD peaks within interglacial periods, and a comparison to published clay data. (b) Sr-Nd isotopic compositions of bulk detrital sediment and IRD samples (>250 μm) in core ANT34/ A2-10 from IRD peak intervals and one bulk detrital sediment sample from Marine Isotope Stage 2, compared to published detrital Sr-Nd data. (c) Sr-Nd isotopic compositions of IRD samples in core ANT34/A2-10 during IRD peak intervals compared to published bulk rock Sr-Nd isotopic data from selected West Antarctic sources. Note the break of scale on the x axis for panels (b and c). Endmember data for clay minerals are from previous studies (Diekmann et al., 2004;Ehrmann et al., 2005Ehrmann et al., , 2011Hillenbrand, 2000Hillenbrand, , 2001Hillenbrand et al., 2003;Jung et al., 2021;Pant et al., 2013;Setti et al., 2004). Sr-Nd isotope data are from previous studies (Adams et al., 2005;Blakowski et al., 2016;Farmer et al., 2006;Futa & Le Masurier, 1983;Hart et al., 1997;Korhonen et al., 2010;McCarron & Smellie, 1998;Pankhurst et al., 1993;Riley et al., 2001;Ryan, 2007;Simões Pereira et al., 2018, 2020 flow during the last glacial period at ∼20 ka (Basak et al., 2018;S. Wu, Lembke-Jene, et al., 2021) could have led to less Ross Sea-sourced and more Amundsen and Bellingshausen Seas-sourced fine-grained detritus reaching this site compared to during the interglacial IRD peak intervals (Figures 5b and 5c). However, such a difference is not observed in Sr-Nd isotopes, which are similar between these periods (Figures 4d and 6b), which may suggest that deep-ocean current transport had a minor influence on the bulk detrital Sr-Nd isotopic compositions. This scenario could simply reflect the shorter sediment transport distance of the bulk detrital sediment compared to the fine-grained clay fraction, but it could also implicate a more significant (and perhaps dominant) role for icebergs in supplying the coarser detritus (>2 μm) to site ANT34/A2-10.
In addition to the above PSAs, the bulk detrital Sr-Nd isotopic compositions during IRD peak intervals and MIS 2 also overlap with the compositions of glacial drift from the Dry Valleys and North Victoria Land, which show similar Sr isotopic compositions of 0.709-0.712 and a wide range of ε Nd values between −8.6 and +2.5 (Blakowski et al., 2016) (diamonds in Figure 6b). These results could indicate supply to site ANT34/A2-10 from Alpine-type glaciers in the Dry Valleys/North Victoria Land regions, consistent with the WRS source proposed above. However, it is important to note that it is hard to distinguish these WRS sources from the WAS/ EAS sources. The glaciers in the Dry Valleys/North Victoria Land regions would have been advanced into the Ross Sea Embayment in the form of outlet glaciers during glacial periods (Heath et al., 2022), while their retreat during warm periods could supply such glacial drift via calving icebergs. Overall, the consistent Sr-Nd isotope signatures since MIS 16 implicate a mixture of terrigenous detritus supplied predominantly by icebergs from the Amundsen and Bellingshausen Seas and from the Dry Valleys/North Victoria Land (WRS) that probably drifted through modern-like iceberg trajectories to our site (Figures 1c and 7a).
This modern-like iceberg trajectory is also supported by comparing the Sr-Nd isotope signature of the IRD in this core (from I7 and I8) to bulk rock PSAs in the South Pacific sector of West Antarctica (Figure 1a). The IRD sample from I8 ( 87 Sr/ 86 Sr = 0.711, ε Nd = −3.1) has a similar composition to the contemporaneous bulk detrital sediments, whereas the IRD sample from I7 has considerably lower 87 Sr/ 86 Sr ratios (0.707) and higher ε Nd values (+1.2) than the bulk detrital sediment from I7 (but note that these samples were from different core depths; Figure 4a). Overall, these IRD data lie close to the composition of sediments from the West Antarctic margins and the Ross Sea (Figure 6b), as well as specifically matching certain rock endmembers, such as those from Graham Land and Alexander Island (Figure 6c). The more radiogenic Nd and less radiogenic Sr isotope signatures in the I7 sample suggest a dominant volcanic contribution, which could potentially reflect local volcanic sources from Marie Byrd Land (Figure 6c), which influence sediments along the Ruppert and Hobbs Coasts (Figure 6b), or more distal pyroclastic sources from Alexander Island (Figure 6c). Such a mixture of iceberg sources from these PSAs is consistent with modern-like iceberg trajectories after MIS 16 (Figure 7a), with detritus carried northwestward by iceberg rafting from the Amundsen and Bellingshausen Sea Embayments by wind-and tide-driven currents (Hillenbrand et al., 2003;Simões Pereira et al., 2018), with a further potential influence from the westward-flowing Antarctic Coastal Current (Hillenbrand et al., 2003;Simões Pereira et al., 2018;Tournadre et al., 2015).

Detrital Sr-Nd Isotopic Signature Before MIS 16
Rather than a mixed signal from the Ross, Amundsen, and Bellingshausen Sea shelf regions (as seen since MIS 16), the bulk detrital Sr-Nd isotopic compositions from the I11 and I12 peaks (before MIS 16) lie close to the signatures of detrital sediment from the Ross Sea shelf region, specifically overlapping with the WRS sources and trending toward ERS and central Ross Sea sources (Figure 6b). While it is hard to constrain a unique provenance solution given the wide range of variability in the PSAs, the bulk detrital Sr-Nd isotopic compositions from the I11 and I12 peaks may indicate an increased contribution, or possible dominance, of sediment from Ross Sea sources during this time period (Figure 6b). Such a change could reflect the importance of iceberg-delivered detritus, with the shift in provenance arising from differing iceberg trajectories before and after MIS 16. Such changes may relate to a proposed shift in the position of the Amundsen Sea Low (ASL), an important and highly dynamic regional low-pressure system, at ∼620 ka (Konfirst et al., 2012; Figure 7).
The ASL is a crucial driver of modern West Antarctic climate variability and controls the meridional winds directed toward West Antarctica Turner et al., 2013;Wang et al., 2020). Furthermore, variability of the ASL may play a role in accelerating glacial ice loss (Hosking et al., 2016), due to its influence on the heat budget (Dotto et al., 2020), the cavity circulation (Hattermann et al., 2021), the upwelling of warm CDW in polynyas , and the import of CDW beneath the ice shelves (Holland et al., 2019). All those processes could act to induce ocean-forced ice-sheet instability in this region. Moreover, intensification of the ASL can be linked to a positive Southern Annular Mode (SAM) and poleward migration of the SWW (Hosking et al., 2016;O'Connor et al., 2021;Turner et al., 2013). This situation could generate a negative sea surface height anomaly center in the northeast Ross Sea and be expected to expand and strengthen the Ross Gyre (Dotto et al., 2018), increasing its impact on iceberg trajectories in this region. Konfirst et al. (2012) suggested that the ASL had a more northeastward and stable annual position before ∼620 ka than since that time, which could lead to the negative sea surface height anomaly being centered further northeastward ( Figure 7b). We speculate that this condition would have led to a long-term state involving an expanded Ross Gyre, thus favoring transport of coastal icebergs from the Ross Sea Embayment to site ANT34/A2-10 ( Figure 7b). Conversely, once a southwestward shift of the ASL established the modern-like southwestward sea surface height anomaly center, a contracted Ross Gyre could explain the mixed pattern of detrital sources from the Ross, Amundsen, and Bellingshausen Seas (Figure 7a). In addition, this latter scenario could lead to enhanced ice-sheet instability in those regions of the WAIS in response to these atmosphere-ocean changes (Steig et al., 2012;Thoma et al., 2008), which could have contributed to the increased supply of IRD from those local WAIS margins since MIS 16.
The Sr-Nd isotope signature of IRD samples was only measured in one sample (I10) from before MIS 16, so any interpretations of that data point should be taken as preliminary. Nevertheless, it shows the same sense of changes as the bulk detrital data from before MIS 16 (i.e., more radiogenic Sr isotopes and less radiogenic Nd isotopes) (Figures 4d, 4e, and 6c), and overlaps with central Ross Sea sediments (Figure 6b), so could also reflect a dominantly Ross Sea source. In addition, its composition is similar to the signature of Paleozoic/Mesozoic granitoids from Adie Inlet/Fomalhaut Nunatak (Figure 6c), making transport from Palmer Land to site ANT34/A2-10 another possibility. However, it differs from the composition of pyroclastic sources in Alexander Island (Figure 6c), which is the node of iceberg transport from Palmer Land to site ANT34/A2-10, which may argue against this alternative iceberg pathway. More Sr-Nd-Pb isotope data from both detrital and IRD samples from this interval, as well as single-mineral 40 Ar/ 39 Ar and U/Pb evidence, should be explored to better constrain the changes in iceberg trajectories at this time and test these hypotheses.

Evidence for Ocean-Forced WAIS Instability Since ∼770 ka
Today, the incursion of relatively warm CDW and/or MCDW onto the shelf regions in the South Pacific sector of the WAIS (Figure 1b) contributes to the loss of buttressing ice shelves, enhanced drainage of outlet glaciers, and grounding-line retreat, accompanied by iceberg calving (Adusumilli et al., 2020;Das et al., 2020;Hofmann et al., 2012;Lowe & Anderson, 2002;Nakayama et al., 2018;S. S. Jacobs et al., 1996S. S. Jacobs et al., , 2011. These changes in CDW circulation are probably related to latitudinal shifts in the SWW (Hillenbrand et al., 2017;Steig et al., 2012;Thoma et al., 2008), possibly in response to changes in the SAM and teleconnections to the tropical Pacific climate (Steig et al., 2012). This section assesses how past changes in WAIS dynamics evidenced in our IRD and provenance records from core ANT34/A2-10 were related to atmosphere-ocean forcing operating in the Southern Ocean on orbital timescales.
Several proxies have been used to provide evidence of past changes in atmospheric and ocean circulation in the Southern Ocean over orbital timescales. Based on the idea that deepwater stratification in the Southern Ocean would increase the regenerated nutrient content in the deep ocean and reduce the preformed nutrient content in southern-sourced intermediate waters (Toggweiler et al., 2006), benthic foraminiferal δ 13 C gradients between the deep and intermediate ocean (Δδ 13 C deep-int ) can be used to reconstruct the vertical chemical stratification (Charles et al., 2010;Hall et al., 2001;Hodell et al., 2003;Ullermann et al., 2016). Meanwhile, reconstructions of sortable silt mean grain size in core PS97/93-2 from the Drake Passage provide a record of ACC strength in the Pacific sector (Toyos et al., 2020), which may be related to changes in the SWW and surface buoyancy forcing (S. Wu, Lembke-Jene, et al., 2021). In addition, the bottom water oxygenation proxies (XRF Mn/Ti and excess uranium) in core ANT34/A2-10 reflect deep Southern Ocean ventilation (Tang et al., 2022), while ice accumulation rates in the EDC ice core provide evidence on SWW-driven upwelling (Fudge et al., 2013).
A comparison of these records reveals a clear correlation in which the IRD peak intervals coincide with reduced benthic δ 13 C gradients (Δδ 13 C deep-int ), peaks in EDC ice accumulation rate, elevated mean sortable silt in core PS97/93-2, and shifts in clay mineralogy and bottom water oxygenation proxies in core ANT34/A2-10 (Figures 5a-5f). These results indicate a temporal link between iceberg calving from the WAIS and regional conditions of weak deep stratification, well-ventilated CDW, strengthening ACC, and intense SWW-driven upwelling. Such agreement is consistent with the persistence of an ocean-forced WAIS instability in the South Pacific sector of the Southern Ocean on orbital timescales since at least ∼770 ka.
In detail, the co-variation between these proxies supports the hypothesis that during the interglacial periods, a poleward shift in the SWW accompanied by an intensification of the ASL (in accord with positive SAM)  led to incursions of warm CDW onto the Antarctic shelves, causing grounding line retreat and ice mass loss from the WAIS (Hofmann et al., 2012;Schmidtko et al., 2014;S. S. Jacobs et al., 1996;Thoma et al., 2008; Figure 8a). Furthermore, the influx of meltwater to the surface of the Southern Ocean (England et al., 2020) could have stabilized the upper water column, enabling warming of the stratified subsurface ocean and causing further disintegration of the WAIS (Fogwill et al., 2015;Liu et al., 2015;Menviel et al., 2010;Schmidtko et al., 2014;S. S. Jacobs et al., 2011;Walker et al., 2007;Figure 8a). In addition to local iceberg input from the Amundsen and Bellingshausen Seas, icebergs calved from glaciers in the WRS could have been transported northward by a clockwise Antarctic Coastal Current and then been driven eastward by the SWW to reach site ANT34/A2-10, following modern-like iceberg trajectories (Figure 7a; Baines & Fraedrich, 1988).
The opposite scenario is envisaged during cold glacial periods, when the absence of IRD peaks at site ANT34/A2-10 may have been linked to the deeply stratified Southern Ocean (Wu et al., 2018) (Figure 8b). An additional factor in the weak ocean forcing during glacial periods may have been the suppression of MCDW incursions into the Ross Ice Shelf by the presence of dense shelf waters, related to the strong formation of Ross Sea-sourced AABW (Basak et al., 2018). . Schematic diagram of (a) ocean-forced West Antarctic Ice Sheet instability during iceberg-rafted debris peaks within interglacial periods, compared to (b) more stable glacial conditions. See main text for discussion. The location of the core site is shown with the red dot.

Enhanced Instability of the WAIS During MIS 13 and 15
The largest IRD peaks in the record from core ANT34/A2-10 occurred during MIS 13 and MIS 15 (Figure 5a), which may point to a maximum in WAIS instability during this interval. A depositional anomaly during MIS 13-15 was also observed in several sedimentary properties in core PS58/254 from the Amundsen Sea margin (Figures 1a and 1b) and was interpreted in terms of possible retreat or collapse of the WAIS during MIS 13 and MIS 15 (Hillenbrand et al., 2009). In the Antarctic ice core record, those interglacials were cooler than the more recent interglacials since the Mid-Brunhes Event at ∼420 ka, and were separated by the unusually short and weak glacial period of MIS 14 (Figure 5g; Hao et al., 2015;Jouzel et al., 2007). Therefore, the relatively warmer conditions during MIS 14 may have contributed to different initial boundary conditions upon the onset of MIS 13 and led to extra ice mass loss during MIS 13. However, it is also possible that the cooler Antarctic temperatures during MIS 13 and MIS 15 than in the more recent interglacial periods could have enhanced iceberg survivability. Whether such an influence could help explain the high IRD peaks (by allowing more icebergs to reach the site), or instead could mean that they represent an under-estimate of iceberg calving changes (by shifting the locus of iceberg melting to lower latitudes) is not known. However, there was no distinct difference in sediment provenance between these interglacials and the more recent interglacials (Figures 4d and 4e), which could point to similar iceberg sourcing and no significant influence of survivability changes, such that the elevated IRD content probably represents greater calving during MIS 13 and MIS 15. The previous glacial history could also matter for sediment supply to icebergs from inland glacial troughs, so the decline in IRD in subsequent interglacial intervals could also partly represent the removal of such debris from glacial troughs during this period of strong instability.
Moreover, the southward movement of the SWW and ASL after ∼620 ka (Figure 7a), as previously proposed (Konfirst et al., 2012) and supported by bulk detrital sediment provenance shifts in core ANT34/A2-10 ( Figure 4d; Section 5.2.3), may have played a role in the enhanced WAIS instability during MIS 13 and MIS 15, by enabling CDW upwelling near the shelf break (Steig et al., 2012;Thoma et al., 2008). Interestingly, despite the lukewarm nature of MIS 13 and MIS 15 in ice core records (Figure 5g), these intervals were still characterized by small Δδ 13 C deep-int gradients in the Southern Ocean (Figure 5d) that were comparable to more recent interglacials, which is consistent with strong deep ocean ventilation and strong ocean forcing during these intervals. Despite cooler interglacial temperatures near the Subantarctic Front during this interval than since the Mid-Brunhes Event (Ho et al., 2012), the surface ocean near Antarctica may have been warmer at this time due to local ocean dynamics (Konfirst et al., 2012). Future work should further investigate ocean temperatures near Antarctica during this interval and their implications for the ice sheet behavior.
Spectral analysis can be used to characterize the pacing of past variability in the IRD inputs from the WAIS and to assess potential links to orbital forcing, as well as whether these aspects changed in relation to the proposed circulation shift at ∼620 ka or could explain the enhanced IRD during MIS 13-15. Analysis of the IRD abundance (grains/g) reveals a significant peak in the ∼100 kyr eccentricity band, with power also in the ∼41 kyr obliquity band ( Figure S2a in Supporting Information S1), while the low sedimentation rate and age model uncertainty restrict the potential to resolve power in the precession band. Wavelet analysis of the IRD abundance shows significant power only before ∼400 ka in the 41 kyr band, whereas there is significant power in the 100 kyr band throughout the record ( Figure S2b in Supporting Information S1).
We also used the "cross wavelet and wavelet coherence toolbox' for MATLAB (Grinsted et al., 2004) to perform cross-wavelet coherence (XWT) analysis between the IRD record and both eccentricity and obliquity (Laskar et al., 2004). For eccentricity, the XWT analysis reveals a significant coherent in-phase relationship with the IRD record in the 100 kyr band ( Figure S2d in Supporting Information S1), with no lead or lag resolvable ( Figure S2e in Supporting Information S1). Additionally, the correlation coefficient (Pearson r) between eccentricity and IRD abundance is above 0 in 100 kyr sliding time windows through the full record (Figure 5h). Except for IRD peaks I2 and I4, the IRD peaks in all other interglacials coincide with a decrease in Pearson r and an increase in eccentricity (Figure 5h), which indicates the linkage between warm interglacials and WAIS instability. Variability in the Antarctic Ice Sheet with a period of 100 kyr has previously been inferred for earlier geological intervals (Levy et al., 2019;Patterson et al., 2014) and has typically been linked to a dominance of insolation forcing rather than ocean forcing. However, in the mid to Late Pleistocene, ∼100 kyr periodicity could also arise from ocean forcing changes linked to glacial-interglacial cycles. Despite the strong eccentricity during MIS 13 and 15 (Figure 5h), such eccentricity forcing seems unable on its own to explain the anomalously high IRD abundance during these times, because subsequent intervals with similarly high eccentricity (e.g., MIS 5 and MIS 7; Figure 5h) were not accompanied by such elevated IRD peaks (Figure 5a).
Low obliquity usually relates to cooler surface conditions at high latitudes, enhancing iceberg survivability and leading to contemporaneous IRD peaks in lower latitudes and low IRD signals in high latitudes (Starr et al., 2021). This phenomenon could help explain the linkage between low obliquity, low IRD content, and positive Pearson r during glacials in core ANT34/A2-10 ( Figure 5i). In contrast, high obliquity is linked to increased annual mean insolation at high latitudes (Loutre et al., 2004), which could drive additional ice mass loss. We observe a positive and significant Pearson r between obliquity and IRD abundance during MIS 13 to 15 (Figure 5i), which was accompanied by the most significant coherence between obliquity and IRD abundance seen in the record ( Figure  S2c in Supporting Information S1). These results may indicate that obliquity paced the IRD peaks during MIS 13-15 after the obscure pacing period of the Mid-Pleistocene Transition (Starr et al., 2021), which may also relate to a non-linear response to orbital forcing (Starr et al., 2021;Willeit et al., 2019). We therefore suggest that either obliquity or mean annual insolation forcing (Patterson et al., 2014;Pollard & DeConto, 2009) may have played a role in driving additional mass loss of the WAIS during this interval.

Conclusions
This study provides sedimentological, mineralogical, and geochemical evidence for the evolving interaction between ocean-atmosphere forcing and WAIS instability in the South Pacific sector of the Southern Ocean from ∼770 ka to present. We identify a series of intervals with elevated IRD content in core ANT34/A2-10, which appear to indicate persistent interglacial instability of the WAIS since ∼770 ka. Over orbital timescales, increased IRD abundance during interglacials coincided with a shift in clay mineralogy in the fine fraction, which we interpret in terms of increased Ross Sea-sourced clay transported by the ACC. Together with evidence from previous studies for increased deepwater ventilation, poleward-shifted southern westerly winds, and a strengthened ACC during these intervals, our study supports a predominant control of ocean forcing on WAIS variability. Specifically, changes in the upwelling of CDW onto the shelves would have controlled basal melting, iceberg calving, and grounding line retreat.
In addition, Sr-Nd isotopes measured on bulk detrital sediments and the IRD samples (>250 μm) suggest a shift in provenance at around MIS 16, which we link to coupled atmosphere-ocean circulation changes in the region. We propose that changes in the intensity and position of the ASL and Ross Gyre led to a switch in iceberg trajectories and hence detrital provenance from predominantly Ross Sea sources (before MIS 16) to a combination of Ross, Amundsen, and Bellingshausen Sea sources (after MIS 16). However, categorically distinguishing between all potential provenance sources is difficult and other interpretations would be possible, so further work using multiple isotope systems and single-mineral 40 Ar/ 39 Ar and U/Pb evidence would be required to confirm this hypothesis. Interestingly, MIS 13 and MIS 15 were characterized by the highest IRD abundance and greatest IRD variability in the record, which may reflect a strong response of the WAIS to these changes in the ASL and Ross Gyre dynamics, while strong orbital forcing at this time and an influence from the weak intervening glacial period of MIS 14 could have driven additional ice loss.