Main

The EAIS response to anthropogenic warming and contribution to global sea level are the largest uncertainties in climate models because EAIS formation, evolution, and behaviour during past warm climates are poorly understood10,11. Deep-sea benthic foraminifer oxygen isotopes (δ18O) indicate that during the early Eocene (53–51 million years ago; Myr ago), Earth experienced the warmest conditions1,4,17,18 of the past 65 Myr. This warmth was followed by about 15 Myr of cooling, declining atmospheric CO2, tectonic reorganizations, and the development of continental-scale Antarctic ice sheets by the earliest Oligocene (33.6 Myr ago)1,2,3,4,17,18,19. As atmospheric CO2 declined through the Oligocene and Miocene, deep-sea δ18O and far-field (far from Antarctica) sea-level records suggest that ice sheets advanced to and retreated from Antarctica’s continental shelves in response to astronomically paced changes in solar insolation3,4,18,20,21. These records also suggest larger Antarctic ice sheets with less pronounced growth and decay cycles after the middle Miocene (about 13.8 Myr ago)1,4, when global climate was cool and atmospheric CO2 concentrations low, relative to the Eocene and Oligocene4,17. While far-field records provide a general framework for understanding Cenozoic Antarctic cryosphere development, these records provide little direct evidence for the ice location and extent or the thermal conditions required to assess the climate forcings and feedbacks involved in Antarctic cryosphere and global climate evolution; these records are also complicated by Northern Hemisphere ice volume in the Plio-Pleistocene epoch1,3,4.

The East Antarctic continental margin and Southern Ocean sediments provide direct evidence of EAIS evolution, indicating that there existed regional marine-terminating ice in the late Eocene22,23,24 and astronomically paced glacial–interglacial cycles through the Pliocene5,6,14. However, existing ice-proximal records are geographically limited and temporally discontinuous, making regional comparisons difficult. Recent ice-sheet models provide additional insight into EAIS evolution10,11,25. Outputs indicate that EAIS catchments with deep landward-dipping subglacial topography and surface meltwater, including the Aurora subglacial basin (ASB), may be sensitive to climate perturbations (for example, atmospheric and/or oceanic temperatures, atmospheric CO2, sea level)9,10,11,25. However, outputs depend on poorly constrained initial boundary conditions17,24,25, feedbacks18, and retreat mechanisms11. Thus, uncertainties remain regarding EAIS evolution that can only be resolved with well dated ice-proximal marine geological and geophysical data1,19.

To improve predictions of future EAIS response to warming and contribution to global sea level rise10,11, knowledge of EAIS evolution in catchments with large potential sea-level contributions is critical. The low-lying glacially sculpted ASB catchment (containing ice equivalent to a 3–5 m rise in sea level9,15,26; Fig. 1a) drains ice from the Gamburtsev Mountains to the Sabrina Coast via the Totten Glacier, which is experiencing the largest mass loss in East Antarctica27 and is influenced by warm subsurface (deeper than 400 m) ocean waters at its grounding line28. The ASB catchment consists of several over-deepened basins15,26 and hosts an active subglacial hydrological system that drains basal meltwater to the ocean29, suggesting that regional outlet glaciers may be susceptible to both progressive retreat13 and changing subglacial hydrology29. Thus, regional glacial dynamics and, ultimately, sea-level contribution during a given warm interval depends on both catchment and glacier boundary conditions (for example, subglacial topography, substrate, and meltwater presence and volume) coupled to atmospheric and oceanic forcings.

Figure 1: ASB elevations and Sabrina Coast bathymetry.
figure 1

a, ASB elevations9 and Sabrina Coast shelf study location (black rectangle) with NBP14-02 (yellow) and published (white)30 seismic lines indicated. The inset shows the ASB location (black rectangle) in Antarctica. ASB highlands (brown), reduced ice glacial pathways (white arrows)26, approximate retreated Oligo–Miocene (solid line) and late Miocene–Pleistocene (dashed line) grounding-line locations15. b, NBP14-02 seismic lines (black), bathymetry and BEDMAP2 bathymetry9 with interpreted seismic lines (red) and jumbo piston cores (JPC; white) indicated. The inset shows regional bathymetry9 and seismics. The Sabrina Coast coastline (blue outline), Moscow University Ice Shelf (MUIS), and shelf break are indicated.

PowerPoint slide

Full size image

We present the first ice-proximal marine geophysical and geological records of ASB glacial evolution (Methods; Figs 1b, 2, 3a). To document regional glacial development, ice dynamics, and the timing of major environmental transitions, we integrate seismic reflection and sedimentary data from the Sabrina Coast continental shelf, at the outlet of the ASB (Fig. 1b). This margin formed during Late Cretaceous period rifting of Antarctica and Australia, with tectonic subsidence continuing through the Palaeogene period8. The present-day continental shelf is about 200 km wide, about 600 m deep, and slopes landward (Fig. 1b). We imaged around 1,300 m of dipping sedimentary strata that overlie acoustic basement on the inner continental shelf (Methods; Fig. 1b). We identified three distinct packages of sedimentary rocks bounded by basement, regionally extensive unconformities, and the sea floor, termed Megasequences I–III (MS-I, MS-II and MS-III; Figs 2a, c, 3a, Extended Data Fig. 1). Glacial erosion of the sea floor allowed us to recover and date strata near the top of MS-I and at the base of MS-III (Methods; Figs 2a, c).

Figure 2: Sabrina Coast seismic and piston core biostratigraphy.
figure 2

VE, vertical exaggeration. a, Seismic line 17 with core locations in MS-I (pre-glacial to pro-glacial; grey shading). MS-II (meltwater-rich glacial) overlies the first expression of grounded ice (dark-blue horizon). MS-III (polar glacial) overlies regional unconformity (light-blue horizon). b, Biostratigraphy for JPC-55, based on pollen and benthic foraminifers (late Palaeocene; brown shading), and JPC-54, based on pollen (early to middle Eocene; beige shading). c, Seismic line 13 with core locations relative to regional unconformity (light-blue horizon). d, JPC-30 and JPC-31 diatom biostratigraphy, with conservative (beige shading, red line) and preferred (brown shading, blue line) ages.

PowerPoint slide

Full size image
Figure 3: Composite Sabrina Coast section with glacial surfaces and climate indicators.
figure 3

a, Composite seismic line with pre- to pro-glacial MS-I (black), glacial MS-II with erosion surfaces (first dark-blue, then grey) overlaid by a non-glacial interval and regional unconformity (light-blue), and polar glacial MS-III. b, Cenozoic atmospheric CO2 reconstructions17,18 with two-standard-deviation error bars (black). c, Composite high-latitude benthic foraminifer δ18O record with blue uncertainty band generated/calculated as in ref. 4, reflecting global ice volume and deep ocean temperatures4; VPDB, Vienna Pee-Dee belemnite standard. d, New Jersey margin sea-level reconstructions (black) with minimum uncertainty (grey envelope) and best estimates (blue line)3.

PowerPoint slide

Full size image

The deepest unit, MS-I, overlies basement and consists of an approximately 620-m-thick, seaward-dipping sequence of low-amplitude discontinuous reflectors that increase in amplitude and lateral continuity upsection (Methods; Fig. 2a, Extended Data Fig. 1). No evidence of glacial erosion exists within these strata (Figs 2a and 3a, Extended Data Fig. 1). On the middle shelf, we imaged two intervals of inclined stratal surfaces (clinoforms), indicating times of high sediment flux to an unglaciated continental margin (Fig. 2a). The jumbo piston core NPB14-02 JPC-55 (1.69 m) recovered mica-rich silty sands 15–20 m below the upper clinoform (Methods; Fig. 2a, Extended Data Figs 1b, 2a, Extended Data Table 1). Terrestrial palynomorphs and benthic foraminifers indicate that these marine sediments are late Palaeocene in age (Methods; Fig. 2b, Extended Data Figs 1, 2, 5, Extended Data Tables 2, 3), confirming the pre-glacial seismic interpretation of MS-I.

Above the upper clinoform, within MS-I, are a series of moderate- to high-amplitude, laterally variable reflectors (grey shading; Methods; Fig. 2a, Extended Data Fig. 1). Piston core NBP14-02 JPC-54 (1.2 m; Methods; Fig. 2a, Extended Data Figs 1b, 3), recovered from this interval, contains centimetre-scale lonestones interpreted to be ice-rafted debris. Terrestrial palynomorphs indicate that these sediments are of early-to-middle Eocene age (Methods; Fig. 2b, Extended Data Table 2). Laterally variable reflectivity without chaotic seismic facies, with ice-rafted debris, and no evidence for cross-shelf glacial erosion indicate that marine-terminating glaciers were present at the Sabrina Coast by the middle Eocene, but grounded ice had not yet advanced across the shelf (Figs 2a, 3a).

MS-I strata reveal episodes of enhanced sediment flux from the ASB, followed by the early to middle Eocene arrival of marine-terminating glaciers to the Sabrina Coast. Models and observations indicate that Antarctica’s ice sheets nucleated in the higher elevations of the Gamburtsev Mountains and first reached the ocean near the Sabrina Coast and Prydz Bay19, increasing sediment flux to the Australo-Antarctic Gulf30. Within the ASB are a series of topographically constrained basins that probably hosted progressively larger ice volumes15,26 as ice expanded in the catchment (Fig. 1a). We speculate that after the early Eocene climate optimum (53–51 Myr ago), as regional and global temperatures cooled and atmospheric CO2 declined17,18 (Fig. 3b, c), glacial ice breached the northern ASB highlands26, allowing marine-terminating glaciers to deliver ice-rafted debris to the Sabrina Coast shelf by the early to middle Eocene (Figs 1a, 2a, Extended Data Fig. 2). This finding is important and indicates both the presence of substantial East Antarctic ice volume by the early to middle Eocene and the relatively early arrival of marine-terminating glaciers at the Sabrina Coast, compared with late Eocene arrivals in Prydz Bay and the Weddell Sea22,23,24. Owing to the relative paucity of Eocene data from Antarctica’s margins, it is not clear if this early arrival is unique to the Sabrina Coast, or if equivalent data have not yet been recovered.

Upsection (around 13 m) from core JPC-54, the deepest regionally mappable roughly eroded surface (dark-blue horizon; Figs 2a, 3a, Extended Data Fig. 1) separates MS-I strata from MS-II strata and provides the first preserved evidence of grounded ice on the Sabrina Coast shelf (Methods). MS-II is up to 675 m thick with ten additional erosive surfaces (grey numbered horizons; Fig. 3a, Extended Data Fig. 1) that truncate reflectors and exhibit rough morphology and channels indicative of glacial erosion in a meltwater-rich environment7,12,29,30 (Methods). Between erosive surfaces, we observe strata with parallel high-amplitude reflectivity and prograding strata of varying thickness (Methods), which indicate open marine conditions and intervals of high sediment flux7,8, respectively, between the 11 glacial advances and retreats from the ASB.

Unlike previously imaged East Antarctic shelf sequences7,8, Sabrina Coast MS-II reveals multiple erosive surfaces (2–6, 8 and 9) with U-shaped channels carved into sedimentary strata (Fig. 3a). Their geometry and size (Methods; ≤170 m deep; about 1,150 m wide) indicate that these channels are consistent with subglacial tunnel valleys observed in surface meltwater-rich sub-polar glacial systems12. The most prominent channels are associated with surfaces 3–5, 8 and 9 (Fig. 3a, Extended Data Fig. 1e). Overlying erosive surface 11 (Fig. 3a, Extended Data Fig. 1e) is an approximately 330-m-thick sequence of seaward-dipping strata devoid of rough erosional surfaces, indicating prolonged continental shelf progradation and high sediment flux in an open marine setting7,8. A regional landward-dipping angular unconformity truncates seaward-dipping MS-II (and in some places, MS-I) strata (light-blue horizon; Figs 2a, c, 3a, Extended Data Fig. 1). Late Miocene to earliest Pliocene diatomites were recovered from and immediately above the unconformity (Fig. 2c, d, Extended Data Figs 1, 4, 6). Because we may not have recovered sediments below the unconformity, we consider the late Miocene (around 7–5.5 Myr ago) the youngest possible age for the MS-III base (Fig. 2d).

Ice advanced across the Sabrina Coast continental shelf at least 11 times from the early to middle Eocene to the late Miocene (Fig. 3a), when average atmospheric CO2 concentrations (Fig. 3b), global temperatures (Fig. 3c), and global sea levels (Fig. 3d) were similar to or higher than at present4,17,18. Without additional age constraints, the pacing of these glaciations is unknown, but far-field and ice-proximal records indicate cryosphere sensitivity to astronomically paced insolation changes during the Oligo–Miocene5,20,21. The scale of the Sabrina Coast shelf tunnel valleys and the presence of similar channels within the ASB catchment, about 400 km from the present grounding line, suggest that regional subglacial hydrologic systems were fed by large volumes of surface meltwater during Oligo–Miocene glacial–interglacial cycles (Methods)11,15. Thus, during climates similar to or warmer than present, surface-derived meltwater may play an important part in EAIS behaviour29, as indicated by models11. The prograding sequence at the top of MS-II is similar to middle to late Miocene sequences in Wilkes Land and Prydz Bay, which reflect the transition from sub-polar to polar glacial regimes7,8 (Fig. 3a, Extended Data Fig. 1e).

Above the regional unconformity, MS-III consists of a 0–110-m-thick veneer of sub-horizontal to landward-dipping strata that thicken landward, indicating substantial glacial erosion of MS-II or lower regional sediment flux and onset of ice loading by the late Miocene8 (Methods; Figs 2a, c, 3a, Extended Data Fig. 1). MS-III strata contain no visible channels, suggesting reduced regional surface meltwater influence and more diffuse basal meltwater flux12,15,26,29. High-amplitude reflectors (Fig. 3a) within acoustically chaotic MS-III strata indicate erosional surfaces in late Miocene to Pleistocene tills (Methods) and advance or retreat of an expanded EAIS15. Open marine sediments are present, but the lack of accumulation and preservation suggests limited regional ice retreat or shorter interglacials since the late Miocene (Methods; Figs 2c and 3a, Extended Data Fig. 1).

An expanded polar EAIS occupied the ASB catchment and the Sabrina Coast continental shelf since the late Miocene15, coincident with global climate, carbon and hydrologic cycle reorganizations4,13, continent-wide ice-sheet expansion and stabilization7,8,13, Antarctic Circumpolar Current intensification, Southern Ocean cooling, and modern meridional thermal gradient development (Fig. 3b)1,13. Atmospheric cooling probably limited the amount of regional surface ablation, resulting in ice expansion and reduced surface-derived meltwater in the ASB catchment. Although open marine conditions intermittently existed on the shelf, the relative MS-III thickness and patterns of erosion within the ASB catchment suggest a maximum grounding-line retreat15 of about 150 km from its present location since the late Miocene. Thus, in contrast to the adjacent Wilkes subglacial basin14, the ASB did not contribute substantially to sea level rise during Pliocene warmth15,16.

Sabrina Coast shelf records reveal the importance of atmospheric temperatures and surface-derived meltwater to Antarctica’s ice mass balance. Although deeper, more continuous sampling of these sediments is required to assess the timing, magnitude and rates of EAIS evolution in the ASB, the ice-proximal Sabrina Coast shelf record confirms model predictions of the region’s long-term sensitivity to climate10,11,15,25,26. Critical for future global sea-level rise scenarios is the potential for ASB catchment glaciers to revert from the extensive polar system of the past 7 Myr to the surface meltwater-rich sub-polar system of the Oligo-Miocene (Fig. 3a), when average global temperatures and atmospheric CO2 concentrations were similar to those anticipated under current warming projections (Fig. 3b, c)10,11,17. At present, the Totten Glacier is thinning faster than any other East Antarctic outlet glacier11,27,28 owing to ocean thermal forcing28. Our findings suggest that ice in the ASB catchment may respond dramatically to anthropogenic climate forcing if regional atmospheric warming results in surface meltwater production.

Methods

Seismic data acquisition, processing and interpretation

The 750 km of 3-m resolution multichannel seismic data were acquired using dual 45-cubic-inch generator–injector guns and a 75-m-long, 24-channel streamer in 2014 in heavy ice conditions aboard the RV/IB N. B. Palmer. Data processing followed standard steps of filtering, spherical divergence correction, normal moveout correction, and muting, but no deconvolution was required, because of the quality of the generator–injector source. All sediment thicknesses are presented in meters and based on a velocity of 2,250 m s−1; seafloor depths are based on a velocity of 1,500 m s−1. The acoustic basement is the limit of our reflectivity and interpreted as crystalline rock.

Seismic megasequences were identified on the basis of the presence or absence of erosional surfaces and seismic facies of the mappable units within the sediment packages. Seismic facies observed include: (1) stratified (laminated) or semi-stratified intervals interpreted as open marine, (2) relatively continuous layers with variable reflectivity interpreted as open marine conditions influenced by ice-rafting, and (3) chaotic, discontinuous or transparent intervals interpreted as glacial to periglacial conditions. MS-I exhibits stratified, semi-stratified and variably reflective (grey shading) intervals. MS-II exhibits chaotic and discontinuous intervals related to erosive surfaces, stratified or semi-stratified intervals, and prograding intervals. Rough, undulatory surfaces are indicators of glacial advance on continental shelves31. MS-III consists of chaotic or acoustically transparent intervals with thin intervals of stratified facies. The thickness of stratified intervals between erosional surfaces may be a proxy for duration of open water conditions and extent of ice retreat (and thus time and distance required for readvance)32. Therefore, MS-II includes extensive and long-lasting glacial retreats, whereas MS-III records localized or relatively short-lived retreats.

All identified horizons are regionally mappable within the seismic survey area. Glacial erosion surfaces are interpreted based on roughness, extent of down-cutting, and association with overlying chaotic or discontinuous facies. Tunnel valley determinations are based on comparisons with imaged tunnel valleys from the North Sea, Alaska, and Svalbard32,33,34,35,36,37,38. Hydrologic modelling suggests that tunnel valleys form only when meltwater exceeds the capacity of flow through porous glacial substrate and any sheet flow at the base of a glacier39. Because tunnel valley size is expected to relate to discharge of subglacial meltwater, the Sabrina Coast glacial erosion surfaces 3–5, 8 and 9 are interpreted as meltwater-rich glaciations that probably required surface-derived meltwater. In the Ross Sea, a widespread regional unconformity that separates prograding shelf strata from glacial tills was interpreted to indicate widespread ice-sheet expansion and the onset of ice loading, as suggested for the Sabrina Coast angular unconformity underlying MS-III40. Uninterpreted versions of the seismic profiles in Figs 2a, c, and 3a, including individual lines from the regionally representative cross-shelf composite line (Fig. 3a), are included in Extended Data Fig. 1.

Marine sediment collection and physical properties

Marine sediments were collected in 440–550 m of water 100–150 km offshore, on the Sabrina Coast continental shelf (Extended Data Table 1). Geophysical data guided the recovery of a suite of four <2-m-long piston cores that targeted outcropping reflectors on the continental shelf (Figs 1b, 2a, c, Extended Data Figs 2, 3, 4). Seismic data, lithology, benthic foraminifers, diatoms, and bulk sediment geochemistry confirm that these sequences were deposited in open marine to subglacial settings. Sediment cores were transported (unsplit and at 4 °C) to the Antarctic Marine Geological Research Facility at Florida State University, where they were split, photographed, visually described, X-rayed, and the GEOTEK Multi-sensor Core Logger instrument’s data were collected following standard protocols. The radiographs were interpreted in Adobe Photoshop with the contrast adjusted for each image. Organic carbon, δ13C and δ15N analyses of bulk sediments were conducted using a Carlo Erba 2500 Elemental Analyzer coupled to a continuous flow ThermoFinnigan Delta Plus XL IRMS at the College of Marine Science, University of South Florida, following standard methods. Lithologic, physical properties, and geochemical data are shown in Extended Data Figs 2, 3, 4 and provided in Supplementary Information.

Core JPC-55 (1.69 m) contains two distinct lithologic units (Extended Data Fig. 2b). The upper unit (0–0.4 m; Unit 1) consists of Quaternary-recent diatom-rich sandy silt with relatively high magnetic susceptibility (Supplementary Information) overlying a more consolidated lower unit (0.4–1.69 m; Unit 2) of homogeneous black micaceous silty fine sands with organic detritus, rare pyrite nodules, macro- and microfossils, and an approximately 10-cm-diameter spherical siderite concretion nucleated around a monocot stem (Extended Data Fig. 2b, d and e; Supplementary Information).

Core JPC-54 (1.21 m), collected above the youngest clinoform, contains two distinct lithologic units (Fig. 2a, Extended Data Fig. 3b). The lithology of the upper unit (0–0.2 m; Unit 1) in JPC-54 is similar to that of JPC-55 and overlies a lower unit (0.2–1.21 m; Unit 2) composed of structureless gravel-rich sandy silts to silty coarse sands with centimetre-scale angular lonestones throughout Units 1 and 2 (Extended Data Fig. 3b). A conservative approach to interpretation of ice-rafted debris was undertaken in these sediments so that only angular lonestones exceeding 1 cm were interpreted to be ice-rafted debris. Lighter-coloured sediment with modern diatoms, visible on the right-hand side of JPC-54, indicates flow-in below approximately 0.8 m, probably due to a partial piston stroke; flow of dark sediment along the right side of the upper core is consistent with this interpretation and on-deck or transport disturbance (Extended Data Fig. 3b). However, angular lonestones are observed throughout and a majority are surrounded by the dark-coloured sediments.

Core JPC-30 (0.52 m plus cutter nose) contains diatom-bearing sandy muds, with intervals of well sorted sands (0–0.25 m). Sub-angular diatomite clasts are present in a mud matrix between 0.25 m and 0.52 m. In the core cutter nose, we recovered stratified diatomite and gravelly diatom-bearing sandstone and sandy diatomite above a sharp contact with sandy diamictite below (Extended Data Fig. 4b).

Core JPC-31 (0.47 m) contains an upper unit of muddy diamicton (0–0.29 m). Between 0.29 m and 0.47 m, angular diatomite clasts are present, which may have been fractured during coring (Extended Data Fig. 4c).

Biostratigraphic methods

Palynology. Nine samples from JPC-54 and eight samples from JPC-55 (Extended Data Figs 2b, 3b, Extended Data Table 2) were processed at Global Geolab Limited, Alberta, Canada, using palynological techniques suited for Antarctic sediments. Approximately 5 g of dried sediment were weighed and spiked with a known quantity of Lycopodium spores to allow computation of palynomorph concentrations. Acid-soluble minerals (carbonates and silicates) were removed via digestion in the acids HCl and HF. Residues were concentrated by filtration through a 10-μm sieve and mounted on microscope slides for analyses. Analysis was conducted under 100× oil immersion objective with a Zeiss Axio microscope. For samples with sufficient palynomorph abundance, a minimum of 300 palynomorphs were tabulated per sample. For samples with low abundance, the entire residue was tabulated. A database of all palynomorphs recovered was prepared and key species were photographically documented. The taxonomic evaluation was completed based on the type specimen repository and library at the Louisiana State University Center for Excellence in Palynology (CENEX). Palynological results are presented in Extended Data Table 2.

Benthic foraminifera

Benthic foraminifer counts and biostratigraphic data were generated for 16 depths in JPC-55 and 5 depths in JPC-54 using standard protocols (Extended Data Figs 2b, 5, Extended Data Table 3). Sediment samples between 20 cm3 and 30 cm3 were washed over a 63-μm sieve with deionized water. Sample residues were dried at 50 °C for 24 h, transferred to labelled vials, dry-sieved into 250-μm and 150-μm fractions, and examined using a Zeiss Stemi 2000-C stereomicroscope with a 1.6× lens and 10× eyepiece (magnification 10.4–80×). Genus and species identifications were refined using a scanning electron microscope (SEM) at the University of South Florida College of Marine Science (Extended Data Fig. 5). All benthic foraminifer individuals present in each JPC-55 sample are tabulated in Extended Data Table 3. In JPC-55, preservation of aragonite and calcium carbonate tests, determined both visually and via SEM, ranges from poor to excellent (Extended Data Fig. 5). Five species of well-preserved aragonitic and calcareous benthic foraminifers were observed throughout JPC-55 Unit 2 (Extended Data Table 3). No foraminifers were observed in JPC-54.

Diatoms

Diatom biostratigraphy was conducted on two sets of NBP14-02 samples: (1) JPC-30 cutter nose diatomites and (2) diatomite clasts from the bottom of JPC-31 (Extended Data Figs 4b, d, 6). Quantitative slides were prepared at Colgate University for diatom assemblage studies and biostratigraphic evaluation using a settling technique that results in a random and even distribution of frustules41; sub-samples were sieved at 10-μm and 63-μm to concentrate unbroken frustules for examination. Photographic documentation at 1,000× magnification using oil immersion on Olympus BX50 and BX60 microscopes was completed at Colgate University (Extended Data Fig. 6). The beige shading in Fig. 2d represents the conservative zonal assignment and age range, whereas the brown shading represents the refined age interpretation. Age constraints for key diatom bioevents are derived from the statistical compilation and analysis of average age ranges for Southern Ocean taxa42.

Chronology

Palynological biostratigraphic zonation scheme

Palynological biostratigraphic zonation of cores NBP14-02 JPC-54 and JPC-55 is based on the presence of a few key species and limited data available from Antarctica and surrounding regions (for example, Australia and New Zealand). Gambierina edwardsii and Gambierina rudata are known as Cretaceous to Palaeocene species. A recent study published a robust last appearance date (LAD) for these species at the Palaeocene/Eocene boundary on the East Tasman Rise (ODP Site 1172)43. However, in southeastern Australia, the two Gambierina species observed in the Sabrina Coast sequence range into the earliest early Eocene44. Extended ranges for the Gambierina sp. (dashed lines; Fig. 2b) are based on the palynological analysis of ODP Site 1166 in Prydz Bay45,46, where abundant well preserved Gambierina specimens were observed and not considered reworked. Consequently, the Gambierina sp. range was extended into the early to mid-Eocene in East Antarctica45,46.

Microalatidites paleogenicus has a Palaeogene to Neogene range46. Although we are adopting their range herein, there is some controversy with this range. The first occurrence of Microalatidites paleogenicus is listed as Senonian in Australia and New Zealand47, but there is no robust evidence supporting an extended range in Antarctica. In the Fossilworks database (PaleoDB taxon number 321781 at http://fossilworks.org/), Microalatidites paleogenicus is listed in the range 55.8–11.608 Myr ago.

Nothofagidites lachlaniae ranges from the Palaeogene to modern times while Nothofagidites flemingii–rocaensis ranges from the Palaeogene to the Neogene46. The range for N. lachlaniae in New Zealand is listed as Late Cretaceous to present and is similar to other forms48. In the Palaeocene and Eocene, there is climatically induced variability observed in the Nothofagidites ranges. For example, broad regional vegetation changes (for example, the abundance of Nothofagidites lachlaniae in western Southland (the Ohai, Waiau and Balleny basins) and its scarcity in other Eocene sections (Waikato, the Taranaki basin, and the west coast of New Zealand’s South Island) may be related to palaeoenvironmental factors49. The type material is Pliocene50, but the distinction of this species from other Fuscospora pollen (including N. brachyspinulosa and N. waipawaensis) is problematic. If N. waipawaensis and N. senectus are excluded, then the New Zealand first appearance date (FAD) of other Fuscospora pollen would be late Palaeocene.

In Southern Australia, the FAD of N. flemingii is in the upper part of the Lygistepollenites balmei Zone (late Palaeocene)51,52. However, in a detailed study of Palaeocene–Eocene transition strata in western Victoria, N. flemingii is not reported53. In New Zealand, the N. flemingii FAD is reported as middle Eocene54,55. However, in well dated early Eocene New Zealand localities, occasional small N. flemingii-like specimens are observed; their identification is under debate. Owing to the relative geographic proximity of East Antarctica and Southern Australia in the Palaeogene, we follow refs 52 and 53 and place the FADs of both species in the late Palaeocene.

Proteacidites tenuiexinus has a range from 66.043 Myr ago to 15.97 Myr ago (PaleoDB taxon number 277519 at http://fossilworks.org/). We adopt the published late Palaeocene Proteacidites tenuiexinus FAD in southeastern Australia51, but acknowledge that the FAD could be as early as the early Palaeocene.

Two pollen species present in core JPC-54 were not observed in core JPC-55, Nothofagidites cranwelliae and Nothofagidites emarcidus. Most verified references for Nothofagidites cranwelliae and Nothofagidites emarcidus (for example, those with specimens properly identified; the Nothofagidites group is diverse, complex and easily misidentified) place the FAD of both of these species in the early Eocene, at the earliest56,57. The latter species was also found in the Eocene of Western Australia58.

Diatom preservation and biostratigraphy

The diatom assemblages present in cores JPC-30 and JPC-31 are indistinguishable, although preservation is better and abundance higher in the JPC-31 diatomites compared to the sandy diatom muds recovered in JPC-30. Overall, preservation is moderate to good in JPC-31 and poor to moderate in JPC-30 (Extended Data Fig. 6). Diatoms in both cores suffer from a high degree of fragmentation. Large centric taxa, such as Actinocyclus spp. and Thalassiosira spp., are generally broken, whereas the smaller centric and pennate specimens are well preserved (Extended Data Fig. 6). Denticulopsis specimens are generally well preserved, although the longer specimens of D. delicata are typically broken. Rouxia spp. occur mostly in fragments, making identification more problematic. Similarly, specimens of Fragilariopsis spp. are mostly present as broken specimens, with the exception of a few F. praecurta specimens (Extended Data Fig. 6).

Many of the diatom species present in both JPC-30 and JPC-31 have long age ranges and do not provide good biostratigraphic age constraints (for example, Coscinodiscus marginatus, Trinacria excavata). However, the presence of several common taxa provides robust support for a late Miocene–earliest Pliocene age. These taxa include: Actinocylus ingens var. ovalis, Denticulopsis delicata, Fragilariopsis praecurta, Thalassiosira oliverana var. sparsa, Thalassiosira torokina (large form), and silicoflagellates in the Distephanus speculum speculum ‘pseudofibula plexus’ group. There is very little evidence of reworking of older material, with only one specimen of Pyxilla sp. and a fragment of Hemiaulus sp. observed; all species within both of these genera are typical of the Eocene and Oligocene.

Age determination for JPC-55 sediments

On the basis of our conservative pollen zonation scheme, we favour a late Palaeocene to earliest early Eocene age for the exceptionally diverse JPC-55 in situ fossil pollen assemblage; this assemblage is easily distinguishable from reworked Cretaceous microfossils present in the sediments (Fig. 2b). The presence of middle bathyal benthic foraminifer species Gyroidinoides globosus and Palmula sp., both of which went extinct at the Palaeocene–Eocene boundary59, enables us to refine the pollen-based age designation further to the late Palaeocene (Extended Data Fig. 5). Although its first occurrence may be diachronous, the presence of aragonitic Cenozoic benthic foraminifer species Hoeglundina elegans59,60,61,62,63 indicates that these sediments are Cenozoic in age, confirming the interpretation that co-occurring Cretaceous microfossils are reworked. H. elegans and other aragonitic benthic foraminifers are most common in upper to middle bathyal assemblages along the southern Australian margin and the Australo-Antarctic Gulf during the Palaeocene and Eocene60,63. Thus, we conservatively designate an age of late Palaeocene to sediments in the lower unit (Unit 2) of JPC-55 (Fig. 2b, Extended Data Fig. 2b).

Age determination for JPC-54 sediments

Pollen bistratigraphy constrains the depositional age of JPC-54 Unit 2 sediments to the early to mid-Eocene (Fig. 2b, Extended Data Fig. 3b). No foraminifers are observed in the lower lithologic unit of JPC-54. Thus, based on the pollen assemblage alone, we favour an early to middle Eocene age for sediments in the lower unit (Unit 2) of JPC-54.

Age determination for JPC-30 and JPC-31 sediments

Based on Southern Ocean diatom ages42, the FAD of T. oliverana var. sparsa (8.61 Myr) and the LAD of T. ingens var. ovalis (4.78 Myr) provide a conservative age estimate for the diatom assemblages present in JPC-30 and JPC-31 (8.61–4.78 Myr; Fig. 2d). A more restricted age interpretation is possible if the presence of Shionodiscus tetraoestrupii (FAD 6.91 Myr) and the absence of the typical early Pliocene taxon Thalassiosira inura (FAD 5.59 Myr) are considered, indicating an age of 6.91–5.59 Myr (Fig. 2d). This more restricted age should be considered tentative, since precise calibrations for many Southern Ocean diatom bioevents in the Chron C3–C3A (4.2–7.1 Myr ago) interval are compromised by multiple short hiatuses at many drill sites and poor magnetostratigraphy. Further age refinement for the JPC-30 and JPC-31 samples will become possible as diatom biostratigraphic data are published for expanded late Neogene sections recovered on the Wilkes Land margin64. However, the more conservative age estimate (8.61–4.78 Myr ago; Fig. 2d) is well supported by the presence of several taxa with well calibrated ages in the Southern Ocean, including Rouxia naviculoides (FAD 9.84 Myr ago), Thalassiosira oliverana (FAD 9.73 Myr ago), and Thalassiosira torokina (FAD 9.36 Myr ago). The absence of Denticulopsis dimorpha (LAD 9.75 Myr ago, Denticulopsis ovata (LAD 8.13), Thalassiosira complicata (FAD 5.12 Myr ago), Fragilariopsis barronii (FAD 4.38 Myr ago), and Fragilariopsis interfrigidaria (FAD 4.13 Myr ago) support this age assessment.

Data availability

The seismic data from the study are available in the Academic Seismic Datacentre at the University of Texas Institute for Geophysics (http://www-udc.ig.utexas.edu/sdc/cruise.php?cruiseIn=nbp1402). Sediment cores are archived in the NSF-funded Antarctic Core Repository at Oregon State University. All other data that support the findings of this study are available within the paper and its Supplementary Information; these data may also be downloaded from the US Antarctic Program Data Center (http://www.usap-dc.org).