Introduction

The West Antarctic Ice Sheet (WAIS) is largely marine-based (grounded below sea level) and, therefore, highly sensitive to climatic and oceanographic changes1. Over the past decades, the WAIS has been undergoing dramatic mass loss due to subglacial melting and contributed to sea-level rise at a faster rate than any other continental ice sheet on Earth1,2. Model-based simulations predict a complete collapse of the WAIS in the near future that would result in a global sea-level rise of 3.3-4.3 metres3,4. However, these model predictions differ for both past and future stability of ice sheets and have large uncertainties associated with the timing of the ice sheet collapse and resulting sea-level rise5. Thus, these numerical model-based projections require rigorous validation. A complementary approach based on geological evidence from proximal West Antarctic Ice Sheet margins can be utilised to constrain ice sheet sensitivity to climate forcing(s), particularly during geologic times when climatic conditions were similar to modern-day and/or the near future. Arguably, the best analogue in Earth’s history for the near-future global climate is the Pliocene (5.33–2.58 Ma)6. This epoch experienced 2–3 °C higher global mean temperatures7 and atmospheric carbon dioxide concentrations (pCO2) between 350 and 450 ppm, which are comparable to the early 21st century and 25–60% higher than pre-industrial values8,9,10,11. The transition period from the warm Pliocene to a cold Pleistocene climate is a critical interval to investigate the dynamics of ice sheet behaviour, given its analogous ice sheet boundary conditions of the Pleistocene to the modern period, as well as those of the Pliocene to likely near-future conditions in terms of atmospheric pCO2, sea surface temperature and sea-level12,13. A recent study14 has demonstrated that Antarctic Ice Sheet dimensions are tied to a multitude of temperature thresholds beyond which ice loss is essentially irreversible, previously referred to as a “tipping point” due to the hysteresis behaviour of the ice sheet: that is, the currently observed ice-sheet volume and extent is not regained even if temperatures were reversed to present-day levels. Reconstructions of a WAIS collapse or a substantially smaller WAIS in the geological past will enable us to test ice sheet models that aim to predict future hysteresis behaviour14. Few attempts have been made earlier based on sediment cores retrieved from the western Ross Sea under the Antarctic Drilling (ANDRILL) project. However, these ANDRILL results15 are not representative of the WAIS in the Amundsen Sea sector, where maximum ice mass loss is taking place today16. The Amundsen Sea in West Antarctica is a key region for monitoring ice-sheet stability since the continental ice sheet in this sector is currently being lost at an accelerated pace, leading to rapid ice sheet retreat of Pine Island, Thwaites, and neighbouring glaciers16,17. Numerical model-based studies suggested that the Amundsen Sea sector is a key region for major West Antarctic Ice Sheet (WAIS) retreat or potential collapse during the most intense warm periods of the Plio-Pleistocene over the past 5 Ma3,4. However, evidence for such a collapse or smaller dimensions of the WAIS is, so far, indirect and mainly based on far-field data, such as sea-level records18. Direct geological evidence from ice-proximal records to WAIS is limited.

Here, we provide a record that constrains the behaviour of the WAIS in the Amundsen Sea sector during the late Pliocene and Pleistocene, based on sedimentological and geochemical data from marine sediment core Site U1532 (68°36.683′S, 107°31.500′W, 3961 m water depth) retrieved during IODP Expedition 379 in 201919 (Fig. 1, Supplementary Information). This site is located in close proximity to the WAIS, having recorded processes directly related to past WAIS dynamics. Based on the age-depth model, the cores of the present study represent the depositional history of the Pliocene-Pleistocene interval (Supplementary Fig. 1 and Supplementary Information).

Fig. 1: IODP Expedition 379 Site U1532 in the Amundsen Sea sector.
Fig. 1: IODP Expedition 379 Site U1532 in the Amundsen Sea sector.
Full size image

a Digital Elevation Map (DEM) of Antarctica63. The colour bar is in meters. The red star indicates the core site of the U1532. b The West Antarctic Ice Sheet primarily comprises the Thwaites (TG) and Pine Island (PIG) glaciers basins, delineated by the grey lines; Filled blue circle indicate core location ANT34/A2-1020. Colour bar represents the ice flow rate. The Red dashed line marks the ice-sheet extent simulated for the warm Pliocene4.

Results and discussion

To constrain the dynamics of the West Antarctic Ice Sheet, we produced geochemical and radiogenic neodymium (Nd) and lead (Pb) isotope records, providing key information on climate-state dependent sediment sourcing from the late Pliocene into the Pleistocene based on detrital marine sediment signatures from IODP Site U1532. We used radiogenic isotope compositions of neodymium (143Nd/144Nd, expressed as εNd, which is the deviation of measured 143Nd/144Nd from the Chondritic Uniform Reservoir in parts per 10,000) and lead isotope ratios (here 208Pb/204Pb and 208Pb/207Pb) to trace sediment provenance. In conjunction, Nd and Pb isotopic signatures of the bulk sediments contain unique information about the continental source areas and are relatively conservative during transport and deposition, and thus can be used to delineate the sources. Our working hypothesis proposes that a major ice sheet destabilization would shift the subglacial erosion regime further inland, leading to erosion and sediment supply from new geological source regions to our sites, which would ultimately be recorded in proximal offshore marine sediments. The detrital εNd record based on a total of 140 sediment samples at IODP Site U1532 shows a major shift from less radiogenic values (-7 ± 1) during the late Pliocene to substantially more radiogenic values (-3 ± 1) during the Pleistocene. It is noteworthy to highlight that except for four anomalous data points (three being more radiogenic with an average εNd of ~ -0.66 ± 0.28 and one being less radiogenic ~ -6, Fig. 2a), all the detrital εNd values (n = 38) show a narrow range -4.3 to -2.0 with an average εNd of -3 ± 1; almost invariant at across all sampled Pleistocene glacial-interglacial (G-I) cycles (Fig. 2a). The detrital Pb isotope records also followed similar trends with higher ratios during the late Pliocene and lower ratios in the Pleistocene (Fig. 2b, c). When shown in concert with a previously published West Antarctic sediment record (site ANT34/A2-1020, covering ~770 ka) ~421 nautical miles away from our core site in the northern edge of the Amundsen Sea (Fig. 1b), our detrital εNd record from Site U1532 show moderate and consistent variations throughout sampled late Pleistocene G-I periods (Fig. 2a). The overall trend derived from the εNd records of these two core sites clearly demonstrates a gradual and distinct shift from the supply of substantially less radiogenic Pliocene sediments (~ -8) to more radiogenic Pleistocene sediments (-3 ± 1, n = 49) and, thereafter, largely invariant sediment provenance signature during the entire Pleistocene (Fig. 2a). The physical properties, lithological and geochemical parameters of bulk sediments exhibit consistent changes with the radiogenic isotope profiles, especially during the Pliocene-Pleistocene transition (Fig. 2). Since no other Nd and Pb isotope records extended into the Pliocene, these pronounced changes in sediment sourcing have not been previously detected.

Fig. 2: Late Pliocene and Pleistocene geochemical records from Site U1532.
Fig. 2: Late Pliocene and Pleistocene geochemical records from Site U1532.
Full size image

a Bulk detrital sediment Nd isotope compositions (εNd) (error bars are 2 SD external reproducibility) record from IODP Site U1532 (this study) together with earlier published records of bulk detrital records (<63 μm fraction) ANT34/A2-10 (Lat: 125.58°W, Long: 67.03°S)22, (b, c) Bulk detrital sediment Pb isotope records from the site U1532. The error bars are smaller than the symbols in the Pb isotope data. d Ratio of total organic carbon and nitrogen percentage (TOC/TN). e Sediment reflectance a*. f Average sedimentation rates. The data of the panels (d−f) are published in the IODP proceedings27 and are also available in the IODP data portal (https://web.iodp.tamu.edu/OVERVIEW/). g Global benthic δ18O (LR04) curve44. Colour bank indicates climate transitions; iNHG-intensification of Northern Hemispheric Glaciations and mPWP-mid Pliocene Warm Period. h Lithological log46 of sediment core U1532A, B.

West Antarctic Ice sheet growth during the Plio-Pleistocene transition

This study investigates past erosion history and its connections to ice sheet dynamics and configuration during the Plio-Pleistocene climate transition and late Pleistocene glacial-interglacial cycles. Pine Island Glacier (PIG) and Thwaites Glacier (TG) are the two major ice streams draining around 32% of the West Antarctic Ice Sheet (WAIS) into Pine Island Bay within the Amundsen Sea Embayment (ASE) (Fig. 1, Supplementary Fig. 3). These glaciers, collectively termed the “weak underbelly” of the WAIS20, represent the most vulnerable part due to rapid current mass loss and are susceptible to rapid future collapse. These two glaciers supply glaciogenic detritus to the core sites (Fig. 1b). Glacially eroded fine-grained detritus delivered offshore by PIG is characterised by less radiogenic εNd (~-9) and is isotopically distinct from sediments supplied by TG (~-4)20,21 (Supplementary Fig. 3c, d). There is some overlap in εNd values between the Antarctic Peninsula (AP) and Western Amundsen Sea (AW) and western PIG and TG (Supplementary Fig. 3b, d, and Table S2); however, these can be distinguished with Pb isotope signatures (Fig. 3c, d, Supplementary Information). The Pb and Nd isotope data of Site U1532 in 208Pb/204Pb vs. 208Pb/207Pb and εNd vs. 208Pb/207Pb, show a binary mixing trend between the two distinct sources (Fig. 3e, f). Therefore, the substantial shift in εNd, 208Pb/204Pb, and 207Pb/208Pb records during the Pliocene-Pleistocene climate transition implies changes in regional sediment sourcing over this interval (Fig. 2a–c, Fig. 3, Supplementary Table 2).

Fig. 3: Deciphering sediment provenance using Nd and Pb isotope compositions.
Fig. 3: Deciphering sediment provenance using Nd and Pb isotope compositions.
Full size image

a Geological map highlighting exposed major rocks and ice flow. The exposed geology map was generated using the data from Cox et al. (2023)64 and Quantarctica software (https://www.qgis.org/en/site/about/case_studies/antarctica.html) (b) Detrital εNd signature of surface sediments along the west Antarctic margin and its hinterland lithology20,21,36. Star represents IODP Site U1532. Closer insight of surface εNd distribution in the Amundsen Sea Embayment36. c, d Pb isotope signatures of surface sediments along the coastal regions of western Antarctica and the Antarctic Peninsula65,66 and hinterland lithology67,68 are shown in these panels. Nd and Pb isotope data of Site U1532 are plotted in the cross plot between detrital (e) 208Pb/204Pb and 208Pb/207Pb ratios and (f) εNd and 208Pb/207Pb ratios. The Pliocene and Pleistocene core records of Pb and Nd isotopes are plotted in red and blue, respectively. The arrow indicates that the range of the PIG ²⁰⁸Pb/²⁰⁷Pb ratio exceeds the plotting limits in both directions, reflecting the full variability of the Pb isotope endmember values. Filled coloured inverted triangles indicate major sources of sediments such as Ross Sea (RS), Bellingshausen Sea (BS), Amundsen Sea West (AW) and East (AE), Pine Island Glacier (PIG) and Antarctic Peninsula (AP). The trend in mixing plots shows a distinct shift in the sediment provenance during the Pliocene-Pleistocene transition.

The prominent shift in the radiogenic isotope profiles can either be explained by (i) variable mineralogical and grain size sorting, (ii) sediment dispersal and transport associated with transient ocean currents and (iii) changes in sediment provenance and erosion regime associated with ice-sheet dynamics and/or a combination of these factors. Detrital εNd measurements in the bulk samples of the present core U1532 and fine fraction (<63 μm) of the samples from the proximal core site ANT34/A2-1020 show similar values within their measurement uncertainty (Fig. 2a) and, thus, rule out potential effects of mineralogical or grain size sorting on the isotope compositions. The three anomalous radiogenic values (average of -0.66 ± 0.28) and one less radiogenic value (-6) during the Pleistocene do not fit in the overall trend in the εNd profile. These radiogenic values could be due to the supply of volcanic material deposited directly over the region during the eruptions and/ or transported by Ice Rafted Debris (IRD >250 μm detrital) sediments22 (Supplementary Fig. 4, 5). The geochemical fingerprints of glacially eroded bedrock from West Antarctica based on available radiogenic isotope and trace element data as evidenced via Sr-Nd isotope data in previous studies20,21,22,23,24 of glacial-marine sediments reveal a distinct contrast in IRD provenance between the AP, Bellingshausen Sea (BS), and the Sulzberger Bay (SB) sectors. The detrital thermochronology (hornblende 40Ar/39Ar ages) and bulk (<63 μ) Sm/Nd isotope systematics data25 from glacial-marine sediments in the AP sector showed relatively young ages compared to BS and AS (Supplementary Fig. 5a) and more radiogenic values that can be traced back to the west coast of the AP, while IRD from the BS and SB contributed less radiogenic sediments during interglacials (Supplementary Fig. 5). Detailed investigation22 suggested that the reorganization of ocean currents during MIS 16 significantly influenced IRD trajectories and sediment sources (see Supplementary Information, Section 3.1, and Supplementary Fig. 7). This shift contributed to a supply of less radiogenic sediments at site ANT34/A2-10 before MIS 16, which shifted to more radiogenic sediments thereafter.

The bathymetric map26 of the Pacific sector of Antarctica includes a network of seismic lines (black) in the Amundsen Sea sector27 (Supplementary Fig. 1a). It suggests that glaciogenic detritus originating from the continental shelf supplied most of the terrigenous drift sediments via gravity-driven transport through the channels, with bottom-current capture and subsequent deposition of their suspended load on the drift28,29. Hence, this core site sensitively recorded processes directly related to past WAIS dynamics in the form of sediment deposition after downslope transport, but also indirectly by current-related transport and deposition. The study site U1532 is located close to the Southern Antarctic Circumpolar Current Front, with its bottom waters influenced by Antarctic Bottom Water (AABW) derived from the western Ross Sea (Supplementary Fig. 5). Therefore, the down-core isotopic changes in Pliocene sediments at Site U1532 likely reflect a combined influence of past variations in ocean currents and detrital input from various source regions, linked to the extent of the WAIS. The ACC flows eastward at intermediate to abyssal water depths around Antarctica, while the Antarctic Slope Current (ASC) flows in the opposite direction along the Antarctic continental shelf and rise. Both these currents could influence sediment transport and deposition at our core site as a function of their respective strength and position. A recent study30 reconstructed ACC strength variations over the past ~5 Ma (Supplementary Fig. 6), revealing significant variations in its vigour during the Pleistocene glacial-interglacial periods. The reconstructed ACC record does not show any significant relation with these two detrital εNd records presented here that are distributed at geographically widespread locations within the Amundsen Sea at almost similar depths (Supplementary Figs. 5b, 6b). This indicates that the ACC did not play the dominant role in sediment transport and dispersal. However, the westward-flowing ASC could also have played an important role as it dominates iceberg transport near the coast31,32. The anomalous εNd values during the Pleistocene (~-6 and -0.66 ± 0.28) are attributed to changes in IRD sources and iceberg transport by ocean currents. Nevertheless, IRD constitutes only a small fraction of the total sediment deposited at the core site, with hinterland erosion and downslope transport identified as the primary contributors. As a consequence, ocean currents only play a secondary role in controlling the observed isotopic signatures (see Supplementary Information, “Methods”). Thus, the provenance record from core site U1532 essentially reflects ice-sheet dynamics associated with the growth and retreat and/or configuration of the ice sheet further inland. The large variations in TOC/TN ratio (Fig. 2d) suggest a mixed input of terrestrial- and marine-derived matter, with a higher ratio indicating a greater contribution of terrestrial organic matter during the Pliocene33. Higher sedimentation rates (Fig. 2f) during the Pliocene warm period, together with higher TOC/TN, suggest enhanced erosion from the continental hinterland due to extended warming that resulted in an ice sheet margin retreat further inland and the supply of more terrestrial sediments from the interior, possibly under a stronger hydrological cycle. Based on the seismic correlation of sediment layers from the drill sites to the continental shelf, grounding zone wedges have been identified in the shelf sediments that predominantly formed during ice sheet retreat within an extended Pliocene warm period from ∼4.2– 3.2 Ma28. These findings from the seismic study indicate that the WAIS was highly dynamic and a regionally less extensive ice sheet along the Amundsen Sea margin during the Pliocene. Warmer regional climatic boundary conditions during the Pliocene would have resulted in a smaller and less extensive WAIS, which is consistent with the ice sheet model simulations for the Pliocene4,34,35. Consequently, substantial amounts of glacially eroded sediments were deposited in front of the Pliocene ice shelf front during glacial-interglacial fluctuations of the ice sheet and ultimately delivered over the continental shelf edge (Supplementary Fig. 8). This situation gradually changed with the expansion of the WAIS preceding the transition to the early Pleistocene cold climate during which large part of the marine-based ice-sheet extended to the ocean. Therefore, glacial-interglacial fluctuations at the terminal end of the extended marine ice sheet were mainly within the ocean and thus resulted in less sediment production associated with these fluctuations (Supplementary Fig. 8d, e). Sediment supply was reduced by almost fivefold compared to the Pliocene (Fig. 2f) due to extensive ice cover over the hinterland. Together, these traceable sedimentary features bear witness to the pronounced expansion of the WAIS and changes in its configuration during the Pliocene-Pleistocene transition. Last but not least, the reduced and relatively invariant or muted responses of Pleistocene sedimentation compared to the Pliocene also indicate that the terminal position of the ice sheet during the Pleistocene never retreated for prolonged interval to ice sheet margin positions near the substantially less radiogenic (older) source areas inland of West Antarctica during the Pliocene.

West Antarctic ice sheet dynamics during Pleistocene glacial-interglacial cycles

Sediments from two core sites in the Amundsen Sea (U1532 and ANT34/A2-10) reveal relatively low variability (-3 ± 1, n = 49) in detrital εNd records during the Pleistocene glacial-interglacial over the past ~750 ka (Fig. 4a). This contrasts sharply with the strong transient changes from ~ -8 to ~-3 observed over ~550 ka across the Pliocene-Pleistocene transition. Distinct εNd signatures from Pine Island (εNd ~-9) and Thwaites (εNd ~-4) basins, along with subglacial lithologies in major ice drainage basins21 (Supplementary Fig. 3), suggest that almost uniform εNd values of the late Pleistocene reflect relatively stable and distinct sediment provenance compared to the Pliocene-Pleistocene transition. This isotopic consistency likely results from almost uniform sediment sources with a homogeneous εNd signature in the hinterland. Significant isotopic changes would require ice sheet retreat into geologically distinct terranes, such as the Ellsworth-Whitmore Mountains (EWM, εNd ~ -8)36, which is located far interior at the southern flank of the West Antarctic Rift system (Fig. 3a, b). The absence of such variability implies that the ice sheet margins did not retreat far enough inland to erode and transport the sediments from the EWM during some of the Pleistocene interglacial sections captured by the detrital εNd record. Due to limited exposure of the bedrock and poorly constrained isotope signature in subglacial geology, it is difficult to quantify the exact extent of ice sheet retreat. However, the radiogenic isotopes of Nd and Pb detected in these sediments are sufficiently sensitive to record large-scale changes, such as seen during the Pliocene-Pleistocene transition. Prior to the Pliocene-Pleistocene transition, less radiogenic εNd values (~-8) suggest a WAIS of substantially smaller dimensions with an ice sheet margin being positioned deep in the interior of West Antarctica, reaching the EWM and delivering sediments with lower εNd values. In contrast, the Pleistocene data show no evidence of such extensive retreat during any of those interglacials sampled for the detrital records in the present study.

Fig. 4: Pleistocene erosion records from the Amundsen Sea.
Fig. 4: Pleistocene erosion records from the Amundsen Sea.
Full size image

a Detrital εNd records from four core sites U1532 (this study) and ANT34/A2-1020. b Magnetic susceptibility measured in bulk sediments of the U1532 core. c Elemental ratio K/Al and (d) Principal Component 1 (PC1) derived from principal component analysis (PCA) of elemental ratios, explaining 70% of the variability, primarily associated with lithogenic variations (Supplementary Fig. 2). e Global benthic δ18O record (LR04)44. Red coloured vertical shades indicate Pleistocene interglacial periods.

To investigate the relatively stable Pleistocene εNd signals, we re-examined the variability with much lower amplitude compared to that occurring during the Pliocene-Pleistocene transition. Although the resolution of the detrital εNd record in the Pleistocene interval is not sufficient to resolve all the glacial-interglacial cycles, these do not show changes at a near comparable extent that occurred during the Plio-Pleistocene transition. The high resolution records of magnetic susceptibility and geochemical proxies, including the first principal component (PC1) from PCA analysis of elemental ratios and K/Al ratio, primarily reflect changes in lithogenic inputs during MIS5 (Fig. 4c–e). The comparison of these proxies suggests a different pattern of ice sheet retreat during the extreme Eemian interglacial (MIS5e) compared to other Pleistocene interglacials. Unfortunately, the compiled two records could not capture the MIS 5e interval at the desired temporal resolution. In this context, it is noteworthy to mention that another two records (PC493 and PS58/254)37 from the proximity to our core site, present detrital records which captured the interval of the MIS5e. These records reveal significant variation compared to other interglacials, yet are much smaller in extent compared to that observed during the Pliocene-Pleistocene transition. Further, the relatively muted response during other Pleistocene interglacials characterized by the high insolation38 implies minimal or insignificant ice sheet retreat during most of the interglacials of the late Pleistocene compared with those of the Pliocene. Previous studies based on geomorphological evidences and multiple cosmogenic in-situ 26Al, 10Be, and 21Ne derived exposure ages of quartz-bearing erratics from the southern Ellsworth Mountains suggest that the WAIS has fluctuated only modestly for at least the last 1.4 million years, while a regional ice sheet centered on the Ellsworth-Whitmore uplands may have survived Pleistocene warm periods39, which support our findings. In summary, we conclude that the WAIS exhibited less extensive retreat during Pleistocene interglacials compared to the WAIS preceding the Pliocene-Pleistocene transition implied by the Nd and Pb isotopic constraints presented here. These findings do not align with ice-sheet models predicting widespread WAIS collapse during Pleistocene interglacials4,40. Instead, our findings of the Pliocene ice sheet behaviour align more closely with ice-sheet models, highlighting the relative resilience of the WAIS during the Pleistocene. This muted response, which contrasts with modern rapid WAIS mass loss, underscores the need to re-evaluate the underlying controls of past and future WAIS stability.

Forcing factor(s) for the Plio-Pleistocene ice sheet extent and dynamics

To investigate the forcing factors and determine the threshold of the boundary conditions for the retreat and growth of the WAIS, we compare the WAIS history inferred from previous provenance records (Fig. 5) with reconstructions of greenhouse/radiative forcing, specifically atmospheric pCO2, and orbital forcing (solar insolation) (Fig. 5b, c). The WAIS was substantially smaller than its modern configuration at a pCO2 of ~350–500 ppm during the Pliocene. No comparable retreat is apparent during extreme interglacial climate stages of the Pleistocene despite large variations in atmospheric temperature ( + 4 to -10 °C) and atmospheric CO2 concentrations (~180–280 ppm) during glacial-interglacial cycles. We made use of a change point detection analysis for the temperature record revealing a clear cooling trend between 3.25 Ma to 2.75 Ma (Supplementary Fig. 9a, d) with an onset of ~40 ka glacial-interglacial cycles paced by the orbital forcing (sharp increase in temperature variance at 32−64 ka band, Fig. 5d, Supplementary Fig. 9c) that coincided with changes in the configuration of the WAIS from a smaller ice sheet to a more expanded ice sheet extent afterward. This observation is consistent with previous findings reported from the Ross Ice Shelf based on AND-1B sediment core records retrieved under the ANDRILL Project that suggested 40 ka cyclic variations in ice-sheet extent during the Pliocene were linked to obliquity cycles15. Further, the AND-1B core record equally provides evidence for a smaller WAIS41, prolonged marine conditions in the Ross Embayment, and reduced sea ice extent before 3.3 Ma. Subsequent Southern Ocean cooling of ∼2.5 °C (Supplementary Fig. 10) in concert with a stepwise expansion of sea ice in the Ross Sea between 3.3 and 2.75 Ma coincided with the drawdown of atmospheric pCO2 from ∼400–280 ppm (Fig. 5b). We note that the major expansion of an ice sheet in the Ross Sea sector also initiated at ∼3.3 Ma41. This orbital forcing paced cooling increased the seasonal persistence of Antarctic sea ice and led to an expansion of the westerly wind belt, resulting in the northward migration of oceanic fronts in the Southern Ocean42,43. The Pliocene records in the proximity of the WAIS margin in the Amundsen Sea and the Ross Sea provide direct evidence for continent-wide changes in the WAIS configuration. The evidence is consistent with an ice-sheet/ice-shelf model4 that simulates fluctuations in Antarctic ice volume associated with the loss of the WAIS, in response to ocean-induced melting paced by obliquity.

Fig. 5: Comparison of erosion records with forcing factors, regional climate, and global sea level records.
Fig. 5: Comparison of erosion records with forcing factors, regional climate, and global sea level records.
Full size image

a Bulk detrital εNd records from the Amundsen Sea sector; U1532 (this study) and ANT34/A2-1020 (error bars are 2 SD external reproducibility). b Atmospheric CO2 records; the blue curve represents CO2 records based on Ice core record69 and boron isotope of foraminifera derived record70. c Record of the southern hemisphere (at 70° S) insolation38. d Antarctic ice core temperature anomaly (ΔT, difference from mean values of the last millennium) derived from deuterium isotopes at EPICA Dome C (EDC)71 (black curve). The red curve is the reconstructed temperature based on the deep temperature and atmospheric temperature72. e Sea level record derived from benthic oxygen isotopes proxy record34.

Surprisingly, no substantial changes were observed even during the glacial-interglacial cycles of the late Pleistocene, when global mean temperatures fluctuated substantially, ranging from +4 to -10 °C relative to pre-industrial levels for the duration of 2−6 ka (Supplementary Fig. 10). We observed a muted response even during some of the intense super-interglacials44, which were subjected to the higher insolation values38 the than modern. This suggests that these extreme super interglacial conditions of the Pleistocene might have exerted insufficient thermal forcing for the collapse of the WAIS, which is contrary to previous reports45 from the Ross Sea Embayment. Further, our observation of only moderate Pleistocene ice sheet variability is also in contrast to the findings from the East Antarctic ice sheet13, which shows ice margin retreat or thinning in the vicinity of the Wilkes Subglacial Basin of East Antarctica during warm late Pleistocene interglacial intervals. Therefore, we suggest that the present mass loss of the West Antarctic Ice sheet is not comparable to the long-term climate trend of the last 2.7 Ma. Instead, it appears to be a recent phenomenon resulting from the rapidly progressing concomitant changes in atmospheric circulation and changes in the incursion strength and pattern of the circumpolar deep water circulation.

In conclusion, our data suggest a more stable WAIS during the Late Pleistocene relative to the Pliocene. Prolonged near total loss of its marine-based portions, with the ice sheet margin retreating near the Ellsworth-Whitmore Mountains, is unlikely to have occurred, even during the most intense interglacials of the late Pleistocene. Ice-sheet model predictions3,4,35 for the Pliocene are consistent with our proposition. However, it does not align with some of the model predictions for the Pleistocene interglacial periods3, and does not support earlier suggestions of WAIS contribution to the Pleistocene sea level change. Our findings suggest that the tipping point for the growth and stability of the WAIS may lie within the oceanic and atmospheric boundary conditions of the Pliocene-Pleistocene transition. These findings have significant implications for predicting future ice sheet behaviour and its contribution to sea level rise in a warming world.

Methods

IODP Expedition 379 drill sites and sediment cores

The IODP expedition 37927 in 2019 aimed to drill at six sites in the Amundsen Sea Embayment; however, the drilling operation was hindered by the excessive discharge of icebergs, rendering drilling operations impossible on the shelf in Amundsen Embayment. Finally, two sites such as U1532 (68°36.683′S, 107°31.500′W, 3961.5 m water depth) and U1533 (68°44′S, 109°0′W, water depth ∼4180 m water depth) were successfully drilled (Supplementary Fig. 1a). These sites are located in close proximity to the WAIS, having recorded processes directly related to past WAIS dynamics. Out of these two sites, the U1532 site was chosen for this study because of its higher sedimentation rates (Fig. 2f) and, thus, better temporal resolution19. Sediment recovery from Site U1532 was ~90 %46. This site is located on the western upper flank of a large sediment drift (Resolution Drift) on the continental rise, 270 km north of the Amundsen Sea Embayment shelf edge. This core site is the deepest drill site in close proximity to maximum modern ice mass loss, revealing sedimentary processes only affected by WAIS dynamics. Site U1532 contains a sequence of Pliocene and Pleistocene sediments with an almost complete paleomagnetic record for constraining high-resolution, suborbital-scale climate variability28. It provides continuous records of glacially eroded sediments, representing valuable archives of climatic and oceanographic changes throughout the Plio-Pleistocene glacial cycles.

The various cores of holes A to G of Site U1532 comprise mainly silty clay and clay with different amounts of biogenic content (Supplementary Fig. 1c). The cores comprise a single lithostratigraphic unit in which two subunits IA (0 – 92.6 mbsf, Pliocene to recent), IB (92.6 – ~130 mbsf, Pliocene) are targeted47 in this study (Supplementary Fig. 1b). Physical property data measured on board were used to carry out a core-log seismic integration with drill records. To tie core-log information to seismic profiles, gamma ray attenuation bulk density and P-wave velocity were used to calculate synthetic seismograms. Furthermore, colour reflectance (a*), green/grey colour ratio, age constraints, sedimentation rate, diatom abundance, and lithological descriptions were obtained by the IODP Exp. 379 shipboard scientific party27 (Fig. 2, Supplementary Fig. 1c, d and e).

X-Ray Fluorescence (XRF) scanning for geochemical data

Post-cruise X-ray fluorescence (XRF) scans of the archive halves of sediment core sections were conducted at the IODP Gulf Coast Repository (College Station, Texas, USA)48 using a Avaatech XRF core scanner system. The scans were performed at a 2-cm resolution and analysed for a range of elements, providing semi-quantitative elemental data for Site U1532 sediment cores. Elemental ratios exhibit pronounced variations related to glacial-interglacial cycles. To extract common signals among the variations in elemental ratios, Principal Component Analysis (PCA) was performed using online Excel STAT software (https://www.xlstat.com/en/) (Supplementary Fig. 2). The first principal component (PC1) accounts for ~70% of the variance, related to lithogenic and biogenic variations during the glacial-interglacial cycles. Notably, the excursion observed in PC and PC2 reached their highest amplitude during the MIS 5e interglacial, much higher than those of other late Pleistocene interglacials (Supplementary Fig. 2).

Age model of IODP Site U1532

Age constraints used in this study were taken from the shipboard age-depth models, which were established based on biostratigraphy age datums in correlation with magnetostratigraphic polarity zones47. Based on shipboard analyses of Holes U1532A–U1532G, an age-depth model for Site U1532 was constructed using bio-stratigraphic age datums (Supplementary Table 1) to constrain correlation of magnetostratigraphic polarity zones at Site U1532 correlated to the Gradstein et al. (2012) geological timescale (GTS2012)49 (see Table T5 in the Expedition 379 methods chapter46). Bio-stratigraphic age control for Site U1532 is based on diatom and radiolarian datums46. Sufficient microfossils for bio-stratigraphic age assignment were only present in the upper ~0-10 m of the studied section (i.e. from 0 to 92 m; Lithostratigraphic Subunit IA) which provides an age of middle Pleistocene to recent (0–0.60 Ma). Based on diatom and radiolarian biostratigraphy, the interval between ~92 and 156 m are assigned a mid-to-late Pliocene age of 3.2–3.8 Ma. Using these bio-stratigraphic age control points, the interpreted correlation of magnetostratigraphic polarity reversals identified at Site U1532 (Table T16R)27 to the GTS2012 is relatively unambiguous. The interval between 0 and ~45 m is assigned a Pleistocene age, and the interval between ~45 and 150 m represents Pliocene age. Because of an absence of siliceous microfossils, there is no bio-stratigraphic age control between ~10 and 92 m in Hole U1532A-B27. For Hole U1532A, a reliable shipboard magneto-stratigraphy consisting of four normal and four reversed polarity intervals was obtained. These intervals are the Brunhes–Matuyama polarity transition (0.781 Ma), the termination and beginning of the Olduvai Subchron (1.778 and 1.945 Ma), the Matuyama–Gauss polarity transition (2.581 Ma), the termination and beginning of the Kaena Subchron (C2An.1r; 3.032 and 3.116 Ma) and the termination and beginning of the Mammoth Subchron (C2An.2r; 3.207 and 3.330 Ma, respectively) (see Supplementary Table 1). Paleomagnetic measurements in Hole U1532B identified the beginning of the Mammoth Subchron (C2An.2r; 3.330 Ma) and the Gauss–Gilbert polarity transition (3.596 Ma). Total eight magnetostratigraphic tie points were found between 0 and ~100 CSF-A(m) and two tie points between 101 and 146 CSF-A(m) for the basis of the chronology (Supplementary Table 1, Supplementary Fig. 1b). To improve the shipboard chronology further, we have identified more tie points in the present study based on the correlation between geochemical parameters and global benthic δ18O curve (LR04)44. Elemental ratios of (a) Ba/Al, (b) Ba/Rb, and (c) PC1 (first principal component) derived from Principal Component Analysis (PCA) of multiple elemental ratios were compared with the LR04 curve (Supplementary Fig. 1f–i). In addition to previously determined tie points, nine additional new tie points were identified based on the excursions in the geochemical parameters and their correspondence with the glacial-interglacial signals in the LR04 δ18O curve (shown in Supplementary Fig. 1f–i). Finally, the Matlab™-based programme “Undatable”50 was used to generate an age-depth model with uncertainty. The advantage of using Undatable over other available age-depth modelling programmes is that it allows for the input of uncertainties in both age and depth and for a series of different types of age control points to be incorporated. Undatable was run using 100,000 simulations50. A probability density cloud illustrating the uncertainty envelope around the modelled age-depth points was computed using Monte Carlo iterations. Blue and black dotted lines show 1σ and 2σ uncertainty, respectively. This revised model is based on sixteen tie-points within the study interval (0–3.5 Ma, Supplementary Fig. 1j, and Supplementary Table 1). The age-depth model for the Pleistocene interval is better constrained than the Pliocene interval due to the availability of more tie points/ages and therefore Pleistocene glacial-interglacial variations can be resolved with confidence.

Detrital sediment Nd and Pb isotope measurements

Nd isotopes were measured in the detrital phases of the bulk sediments following the method adopted from earlier studies13,51. We first conducted mild leaching for a shorter duration (ten seconds) without attacking the detrital fraction followed by a second leach for a duration of 24 h to ensure complete removal of the authigenic fractions52,53,54. Approximately 0.3 g of powdered and homogenized sediment samples were agitated with a mixture solution of 0.005 M hydroxylamine hydrochloride, 1.5% acetic acid, and 0.003 M Na-EDTA, buffered to pH 4 with NaOH. Then leachates were removed by centrifugation and washed with Milli-Q water thoroughly. Residual sediment samples were decarbonated by treating with 0.6 N HCl at 80 °C for ~30 min and then the supernatant was discarded, and the acid-free residue was washed with Milli-Q water. The above steps were repeated to ensure complete decarbonation of the samples and washed with Milli-Q water and checked with pH paper to ensure that carbonates are removed from the sediments. Subsequently, samples were washed thoroughly with Milli-Q water and dried in the oven at 80 °C. Dried samples were then ashed at 600 °C to remove organic matter. Approximately 50 mg of ashed samples were digested in precleaned Teflon vials using HF-HNO3-HCL mixture at 120 °C. Standard Reference Materials BCR-2 and BHVO-2 were also digested along with the samples. The digested samples were divided into two aliquots for the column chromatography of Nd and Pb. The dissolved samples were subjected to purification by column chromatography. Samples were passed through columns filled with cation-exchange resin AG50W-X8 (200–400 micron) to separate Rare Earth Elements (REE). Then REE fractions were passed through the column filled with Eichrom LN specTM (50–100 micron) resin to separate Nd from the REEs55.

Nd isotope analysis

Nd isotope ratios were measured using a Multi-Collector Inductively Coupled Plasma Mass Spectrometer (Thermo Fisher Scientific, Neptune Plus) at the National Centre for Polar and Ocean Research (NCPOR), Goa. Instrumental mass fractionation was corrected using the ratio of 146Nd/144Nd = 0.7219 and an exponential law. A correction for direct 144Sm interference was also applied, with all samples which was below (<0.1% of the 144Nd signal). Every five samples were bracketed with concentration matched Nd JNdi-1, and all the mass bias corrected 143Nd/144Nd ratios were normalised to the accepted JNdi-1 standard 143Nd/144 value of 0.51211556. The obtained long-term average ratios of measured JNdi-1 over the period of the analysis of these samples at 10 analytical seasons were 0.512115 ± 0.000049 (2σ, n = 51). The external reproducibility calculated for each session was 0.12–0.27 εNd (2σ) units based on the repeated measurements of JNdi-1. Several procedural blanks were also processed along with the samples and ascertained an average blank of ~70 pg for the samples (n = 9), which is several orders of magnitude lower than the total Nd typically analysed in samples. Hence, no blank correction was applied. Replicates have an average εNd variation of ±0.17 (n = 16). To ensure the quality of the measurements, International reference rock standards BCR-2 and BHVO-2 were processed along with the samples and measured which gave 143Nd/144Nd = 0.512641 ± 0.000030 (2σ, n = 7) and 143Nd/144Nd = 0.512997 ± 0.000027 (2σ, n = 6) respectively, in agreement with reported values57 (Supplementary Data 2).

Pb isotope analysis

Pb isotope was analysed in the same aliquot in which Nd isotope was measured. The Pb cuts in the detrital dissolved were purified by column chromatography filled with ∼80 μL AG1-X8 resin. Pb isotopes were measured using a Thermo Scientific Neptune Plus MC-ICP-MS at GEOMAR Helmholtz Centre for Ocean Research Kiel. Mass bias correction during Pb isotope measurements was performed externally using the Tl-doping technique52,58 with added NIST997 Tl standard solution and a Pb/Tl ratio of ~4. Given that Pb and Tl isotopes undergo slightly different mass-dependent isotopic fractionation during measurements, the adequate SRM 997 Tl isotope composition was chosen on a session-by-session basis so that the offset of all six Pb isotope ratios (206Pb/204Pb, 207Pb/204Pb, 208Pb/204Pb, 208Pb/206Pb, 207Pb/206Pb, 208Pb/207Pb) for NBS SRM 981 Pb isotope standard were zero. Detailed information about the measurement procedure can be found in Süfke et al. 58 and Huang et al. 52. The resultant external intermediate reproducibility of all SRM NBS981 standard solutions were as follows: 206Pb/204Pb = 0.0011, 207Pb/204Pb = 0.0010, 208Pb/204Pb = 0.0029, 208Pb/206Pb = 0.00007, 207Pb/206Pb = 0.00002, 208Pb/207Pb = 0.00006 (n = 39). The total procedure blanks were ~500 pg which represents <0.1% of the sample signal. Blank corrected and uncorrected ratios do not show significant differences and hence blank uncorrected data is reported in this study. The reproducibility of the rock standard BCR-2 and BHVO-2 is listed in the Supplementary Data 2. As shown in the table, all measured standard Pb isotopic ratios are within the error of published compositions.

Robustness of the detrital sediment Nd and Pb isotope records

Unlike other radiogenic isotope systems (for example, Sr and Hf), Nd isotopes are relatively insensitive to grain size variations59, making them a robust indicator of sediment provenance. Their insensitivity to grain size variations makes them more reliable tracers of sediment provenance. However, it is crucial to consider potential complications in provenance interpretations if changes in sediment transport processes result in changing the isotope compositions. In such cases, a link between grain size and Nd isotopes could emerge, impacting the reliability of using these isotopes for provenance analysis. To assess the robustness of Nd isotopes, Simões Pereira (2018)37 analysed Nd isotopes in fine-grained fraction (<63 μm) of sediment cores from two locations (PS58/254 and PC493) near our core site U1532 (Fig. 1b). The results of this investigation ruled out any significant influence of grain-size variations on the Nd isotope compositions, reinforcing the suitability of Nd isotopes as a dependable proxy for sediment provenance in this particular study area. Among various combinations of Pb isotope ratios (six in total), we have used 208Pb/204Pb and 208Pb/207Pb ratios as these displayed the most distinctive, clearest, and smoothest signal compared to other ratios, and both these ratios co-vary in concert with the εNd record. Observation of changes in the paired Nd-Pb isotope records is more robust to decipher the changes in the sediment provenances than either of the two trace metal isotope compositions alone. This finding further validates the robustness of using these isotopes as tracers of sediment sources and transport processes in the specific study region. Pertinent to this, it is important to mention that Pliocene erosional history and ice loss from the East Antarctic Ice Sheet (EAIS) were successfully reconstructed previously using detrital εNd records from the Adélie Land sector of East Antarctica12,13. Distinct sediment sourcing associated with the growth and retreat of the ice sheet margin and multiple collapse events establish a close link between ice sheet dynamics, erosion, and variable regional sediment supply. Therefore, our extensive analysis in the present study and successful demonstration in the previous study from this region provide confidence in the applicability of detrital radiogenic isotope compositions of Nd and Pb provide confidence for deciphering sediment sources and transport processes in the studied area.

Wavelet and scaled-average variance of wavelet power analysis

Wavelet transforms decompose time series into time-frequency space and can therefore find localized intermittent periodicities. We used Continuous Wavelet Transform (CWT) i.e. Morlet wavelet to decompose the time series into time-frequency space which enable us to identify the modes of variability and how those modes vary with time60. Statistical significance is estimated against a red noise model. Monte Carlo methods were used to assess the statistical significance against red noise backgrounds61. Wavelet analysis of the temperature record was performed using online Matlab codes http://grinsted.github.io/wavelet-coherence/ to identify nonstationary power at certain frequency bands60,61. Temperature variability at ~40 ka and ~100 ka bands is shown in the wavelet analysis. More detailed information on wavelet analysis is provided elsewhere60,61. Further, to examine fluctuations in power over a range of scales (a band), we used scale-averaged wavelet power which is a time series of the average variance in a certain band. This can be used to examine the modulation of one time series by another, or the modulation of one frequency by another within the same time series. We have performed this analysis using online Matlab codes http://paos.colorado.edu/research/wavelets/. The scaled average variance of the temperature at 32−64 ka and 64 -100 ka band captured the wavelet analysis show how the ~40 ka and ~100 ka periodicities evolved with time (Supplementary Fig. 9c).

Change-point detection using a Bayesian Ensemble Algorithm (BEAST)

Abrupt changes in the temperature time series were detected using a MATLAB based Bayesian model averaging time-series decomposition algorithm (BEAST)62 (Supplementary Fig. 9d). This algorithm offers a robust method to detect abrupt change points and nonlinear trends in the Antarctic temperature record. The code and software are available at https://github.com/zhaokg/Rbeast. The Bayesian algorithm—BEAS decomposition time series into three contrasting components: abrupt change, periodic/seasonal change, and trend. Mathematically, BEAST is rigorously formulated, with its key equations being analytically tractable. Practically, BEAST provides an estimate of probabilities of change point occurrence, detects both large and low-magnitude disturbances, and uncovers complex nonlinear trend dynamics. The details of the Bayesian Ensemble Algorithm (BEAST) are discussed elsewhere62.