Site U1356 (63∘18.6138′ S, 135∘59.9376′ E) is located at 3992 m water depth in front of the glaciated margin of the eastern Wilkes Land Coast of East Antarctica and penetrated 1006 m into the flank of a levee deposit in the transition between the lower continental rise and the abyssal plain (Escutia et al., 2011; Fig. 1). Overall recovery was 35 % with sediments dated to between the early Eocene and Pliocene, but several intervals provide good stratigraphic control (Escutia et al., 2011; Tauxe et al., 2012). The Oligocene section was recovered between 895 and 430.8 m b.s.f. (cores U1356-95R-3 83 cm to U1356-46R). Our study focuses on the relatively high-recovery (78.2 %) interval within the late Oligocene, which spans 689.4 to 641.4 m b.s.f. (cores U1356-72R to -68R). The sediments from this interval are part of shipboard lithostratigraphic Unit V, which is characterized by light greenish-grey, strongly bioturbated claystones and micritic limestones interbedded with dark brown, sparsely bioturbated, parallel- and ripple-laminated claystones with minor cross-laminated interbeds (Escutia et al., 2011). The bioturbated and calcareous claystones and limestones were broadly interpreted as representing pelagic sedimentation superimposed on the background hemipelagic sedimentary input (Escutia et al., 2011). The laminated claystones and ripple cross-laminated sandstones were interpreted as likely resulting from variations in bottom current strength and fine-grained terrigenous supply (Escutia et al., 2011). In addition, a notable absence of iceberg-rafted debris (IRD) (> 250 µm) in this interval relative to underlying and overlying strata was also recorded.

The late Oligocene depositional setting of Site U1356 was, however, different to that of today. The stratigraphic evolution of the region testifies to the progradation of the continental shelf taking place after continental ice sheet buildup during the Eocene–Oligocene Transition (EOT; 33.6 Ma; Eittreim et al., 1995; Escutia et al., 2005, 2014), which resulted in (1) seismic and sedimentary facies on the continental rise becoming more proximal up-section (Hayes and Frakes, 1975; Escutia et al., 2000, 2005, 2014) and (2) high sedimentation rates during the Oligocene (Escutia et al., 2011; Tauxe et al., 2012). In this context, the studied late Oligocene sediments from Site U1356 record distal continental rise deposition in an incipient/low-relief levee of a submarine channel. As progradation continued, a complex network of well-developed channels and high-relief levee systems developed on the continental rise (Escutia et al., 2000) from the latest Oligocene onwards.

Today, Site U1356 lies close to the southern boundary of the Antarctic Circumpolar Current, near the Antarctic Divergence at ∼ 63∘ S (Orsi et al., 1995; Bindoff et al., 2000) (Fig. 1). However, the paleo-latitude of Site U1356 was around 58.5 ± 2.5∘ S (van Hinsbergen et al., 2015) during the late Oligocene, i.e., more northerly than today. Scher and Martin (2008) and Scher et al. (2015) reconstructed the position of the early Oligocene Antarctic Divergence to be located around 60∘ S (Fig. 1), based on the distribution of terrigenous and biogenic (calcareous and siliceous microfossils) sedimentation, Nd isotopes, and Al / Ti ratios through a core transect across the Australian–Antarctic basin in the Southern Ocean. According to these interpretations, Site U1356 lay far to the north of the Antarctic Divergence zone and was closer to the polar front during the Oligocene.

The age model for Site U1356 was established on the basis of the magnetostratigraphic datums constrained by marine diatom, radiolaria, calcareous nannoplankton, and dinocyst biostratigraphic control (Escutia et al., 2011; Tauxe et al., 2012; Bijl et al., 2018a). The late Oligocene interval contains three magnetostratigraphic datums (Table 1): (1) Chron C8n.1n (o) between 643.70 and 643.65 m b.s.f. (U1356-68R-2), (2) C8n.2n (y) between 652.60 and 652.55 m b.s.f. (U1356-69R-2), and (3) C8n.2n (o) between 679.90 and 678.06 m b.s.f. (U1356-71R). For this study, the age model by Tauxe et al. (2012), which was calibrated to the GTS2004 Time Scale (Gradstein et at., 2004), has been updated using the GTS 2012 Astronomic Age Model (Vandenberghe et al., 2012). Based on this calibration, the age of sediments between 678.98 and 643.37 m b.s.f. is 25.99 and 25.26 Ma, respectively (Fig. 2; Table 1).

Detailed facies analyses provide a stratigraphic framework on which we base our sedimentary processes and paleoenvironmental interpretations. Lithofacies are determined on the basis of detailed visual logging of the core during a visit to the IODP Gulf Coast Repository (GCR), expanding on the lower-resolution preliminary descriptions in Escutia et al. (2011). We logged the lithology, sedimentary texture (i.e., shape, size, and distribution of particles), and structures with a focus on the contacts between the beds and on bioturbation at a millimeter- to centimeter-scale resolution in cores expanding from 896 to 95.4 m b.s.f. (cores U1356-95R to -11R) (see Figs. S1 and S2 in the Supplement). Physical property data were measured during IODP Exp. 318 using the Whole Round Multisensor Logger. Magnetic susceptibility measurements were taken at 2.5 cm intervals, and natural gamma radiation (NGR) was measured every 10 cm (Escutia et al., 2011). In this paper, we focus on the interval between 689.4 and 641.4 m b.s.f., which comprises cores 72R to 68R (Fig. 2).

X-ray computed tomography scans (CT scans) measure changes in density and allow for the analysis of fine-scale stratigraphic changes and internal structures of sedimentary deposits in a nondestructive manner (e.g., Duliu, 1999; St-Onge and Long, 2009; Van Daele et al., 2014; Fouinat et al., 2017). To further characterize the different facies in our cores, selected intervals of Core U1356-71R-6 (678.11 to 676.91 m b.s.f.) and Core U1356-71R-2 (672.8 to 671.35 m b.s.f.) were CT-scanned at the Kochi Core Center (KCC) (Japan), with the GE Medical systems LightSpeed Ultra 16. 2-D scout (shooting conditions at 120 Kv with 100 mA; 3-D helical image with 120 Kv and 100 mA and FOV = 22.0). Image spatial resolution consists of 0.42 mm pixel-1 with 0.625 mm of slice thickness (voxel spatial resolution of 0.42 × 0.42 × 0.625 mm).

The type and composition of biogenic and terrigenous particles, particle size, and morphology of each lithofacies was characterized with a high-resolution scanning electron microscope (HR-SEM) at the Centro de Instrumentación Científica (University of Granada, Spain).

Detailed bulk-chemical composition records acquired by XRF core scanning allow the accurate determination of sedimentological changes and the assessment of the contribution of the various components in the biogenic and lithogenic fraction in marine sediments (Croudace et al., 2006). This nondestructive method yields element intensities on the surface of split sediment cores and provides statistically significant data for major and minor elements (Richter et al., 2006; O'Regan et al., 2010; Wilhelms-Dick et al., 2012). The data are given as element intensities in total counts.

XRF core scanning measurements were collected every 2 cm down-core over a 1 cm2 area with a split size of 10 mm, a current of 0.2 mA (Al–Fe) and 1.5 mA (all other elements), and a sampling time of 20 s, directly at the split core surface of the archive half with XRF Core Scanner III at the MARUM – Center for Marine Environmental Sciences, University of Bremen, Germany. Prior to the scanning, cores were thermally equilibrated to room temperature; the surface was cleaned, flattened, and covered with 4 µm thin SPEXCerti Prep Ultralene1 foil to protect the sensor and prevent contamination during the scanning procedure. Scans were collected during three separate runs using generator settings of 10 kV for the elements Al, Si, S, K, Ca, Ti, Mn, Fe; 30 kV for elements such as Br, Rb, Zr, Mo, Pb; and 50 kV for Ba. The data reported here were acquired by a Canberra X-PIPS silicon drift detector (SDD; Model SXD 15C-150-500) with 150 eV X-ray resolution, the Canberra Digital Spectrum Analyzer DAS 1000, and an Oxford Instruments 100 W Neptune X-ray tube with rhodium (Rh) target material. Raw data spectra were processed by the “Analysis of X-ray spectra by Iterative Least square software” (WIN AXIL) package from Canberra Eurisys. Data points from disturbed intervals in the core face (i.e., slight fractures and cracks) were removed.

The light elements Al, Si, and K show large element variations (intra-element variations of 1 order of magnitude or more; Fig. 2). Similar variations have been previously described in sediment cores to indicate substantial analytical deviations due to physical sedimentary properties (i.e., Tjallingii and Röhl et al., 2007; Hennekam and de Lange, 2012). Accordingly, for this study we have discarded the continuous records of Al, Si, and K and concentrated our interpretations on Al, Si, and K values from the XRF analyses in discrete samples (see below). As Titanium (Ti) is restricted to the terrigenous phase in sediments and is inert to diagenetic processes (Calvert and Pedersen, 2007), we utilized Ti to normalize other chemical elements for the terrigenous fraction. Linear correlation (r Pearson) above standardized values has been done in order to find statistical relationships among the variables.

In addition, we conducted measurements of a total of 50 major and minor trace elements in 25 discrete sediment samples collected at 0.4 and 1 m spacing to determine their chemical composition. For this, we used a Pioneer-Bruker XRF spectrometer S4 at the Instituto Andaluz de Ciencias de la Tierra (CSIC-UGR) in Spain, equipped with a Rh tube (60 kV, 150 mA) using internal standards. The samples were prepared in a Vulcan 4Mfusion machine and the analyses performed using a standard-less spectrum sweep with the Spectraplus software.

We selected key environmental indicators from XRF core scanner data and elemental ratios (i.e., Zr / Ba, Ba, Zr / Ti, Ca / Ti, magnetic susceptibility (MS)) to conduct spectral analyses on the data from the interval between 689.4 to 641.4 m b.s.f. (cores U1356-72R to 68R). We performed an evolutionary spectral and harmonic analysis on each dataset using the Astrochron toolkit in the R software (Meyers, 2014). A detailed methodology is provided in the Supplement, following the Astrochron code of Wanlu et al. (2016). This method allows the detection of nonstationary spectra variability within the time series. The time series were analyzed on the depth scale and then, applying the Frequency domain minimal tuning (Meyers et al., 2001), we converted spatial frequencies to sedimentation rates using an average period of 41 kyr to transform them to an age scale, on the basis of the already resolved age model. The evolutionary average spectral misfit method was then used to resolve unevenly sampled series and changing sedimentation rates (Meyers et al., 2012). This method is used to test a range of plausible timescales and simultaneously evaluate the reliability of the presence of astronomical cycles (Sect. S2).

The revised Oligocene facies log (Figs. S1, S2) includes the high-recovery interval between 689.4 and 641.4 m b.s.f. (Fig. 2). The integration of our lithofacies analyses, with physical properties (MS), CT scans, and HR-SEM analyses, characterize an alternation between two main facies (Facies 1 and 2) (Figs. 2, 3, 4). Although these two facies were already visually identified on shipboard, our analyses allow us a more detailed characterization and interpretation of the depositional environments and the processes involved in their development.

Facies 1 (F1) consists of slightly bioturbated greenish claystones with sparse (Fig. 3a) to common laminations (Figs. 2, 3a–f; Table 2). Laminae, as described on shipboard, vary from 0.1 to 1 cm thick and, based on nonquantitative smear slide observations, are composed of well-sorted silt to fine sand-size quartz grains (Escutia et al., 2011). Laminations can be planar and wavy, with ripple cross lamination structures (Escutia et al., 2011), and show faint internal truncation surfaces, mud offshoots, and internal erosional surfaces (Fig. 3a–f). HR-SEM analyses of the claystones show that the matrix is composed of clay-size particles and clay minerals (Fig. 3g, i). In addition, they show rare calcareous nannofossils that are partially dissolved (Fig. 3g, i). Authigenic carbonate crystals are also identified (Fig. 3i). Bioturbation in F1 is scarce, ichnofossils in the sediments are dominated mainly by Chondrites (Fig. 3d). CT scans also show the presence of Skolithos, with their vertical thin tubes filled with high-density material suggesting they are pyritized (Fig. 3b). Pyrite was also observed in shipboard smear slides in small abundances from the laminated facies in the studied interval (Escutia et al., 2011). Magnetic susceptibility values within the laminated facies are low, between 40 and 70 MS instrumental units (iu), with higher values when silt laminations are more abundant (Figs. 2, 4). NGR is anticorrelated with MS, with high values in F1 varying between 50 and 65 counts per second (cps) (Fig. 2).

Facies 2 (F2) is composed by light greenish grey, strongly bioturbated claystones and silty claystones (Figs. 2, 3; Table 2) with variable carbonate content varying between 5 and 16 % based on our XRF analyses. No primary structures are preserved due to the pervasive bioturbation (Fig. 3a–c). Burrows are backfilled with homogeneous coarse material (silt/fine sand). Different types of ichnofossils are present with Planolites and Zoophycos being the most abundant (Fig. 3a, b). HR-SEM images show (1) silt-size grains containing quartz grains with conchoidal fractures in the corners and impact marks on the crystal faces, indicative of high-energy environments and (2) biogenic carbonate consisting of moderately to poorly preserved coccoliths, which exhibit dissolution of their borders, and to a minor degree detrital carbonate grains (Fig. 3h–j). A total of 13 carbonate-rich layers have been observed within the studied interval F2, and they range in thickness from 10 to 110 cm. Facies 2 CT-scan images show an increase in density (i.e., gradation towards lighter colors in the scan) towards the top of each bioturbated interval (Fig. 3b). MS values are higher in F2 compared to F1. Values vary from 50 to 150 instrumental units (iu) and exhibit an inverse grading or a bigradational-like morphology (Figs. 2, 4), while NGR is inversely correlated with minimum values occurring in F2 (between 35 and 55 cps) (Fig. 2).

Contacts between the two facies are sharp and apparently nonerosive, with minimal omission surfaces or lags (Figs. 3, 4). However, when bioturbation is present, gradual contacts in the transition from F1 to F2 also occur (Fig. 3b). Both sharp and transitional contacts are well imaged on the MS plots (Fig. 2).

In addition, where available, the CT-scan images confirm the shipboard and our own visual observations regarding the absence of outsized clasts and coarse sand grains in F1 and F2. Hauptvogel (2015), however, reports grains that are > 150 µm in size (fine sand fraction) as IRD. He argues that grains of that size could only reach Site U1356 through ice rafting given the distance of the site to shore, unless they were delivered by gravity flows. Thick and coarse-grained mass transport deposits (MTDs) during the latest Oligocene at Site U1356 (Escutia et al., 2011), argue for coarse material being delivered to the site by gravity flows. In addition, fine sand grains to gravel-size clasts have been reported from channels on the lower continental rise off the Wilkes Land margin transported by gravity flows, including turbidity flows (Payne et al., 1972; Escutia et al., 2000; Busetti et al., 2003). Given that during the late Oligocene, Site U1356 is located on a low-relief levee of a submarine channel, one can expect delivery of fine-grained sand and even coarser sediment to the site. In any case, even if some background IRD is present in our record, we argue it is minimal compared to elsewhere in the core.

Down-core changes in the log ratios of various elements have been plotted against the facies log (Figs. 2, 4). In addition, in order to determine geochemical element associations, we performed a Pearson correlation coefficient analysis of major elements on the whole XRF-scanner dataset (Table 3). This analysis highlights two main groups that are used as proxies for terrigenous (i.e., Zr, Ti, Rb, Ba) vs. biogenic (i.e., Ca = carbonate) sedimentation.

Titanium (Ti), Zirconium (Zr), and Rubidium (Rb) are primarily derived from terrigenous sources, where Ti represents the background terrigenous input. During sediment transport Zr, Rb, and Ti tend to become concentrated in particular grain-size fractions due to the varying resistance of the minerals in which these elements principally occur. Zr tends to become more concentrated in fine sand and coarse-silt fractions, Ti in somewhat finer fractions, and Rb principally in the clay-sized fraction (Veldkamp and Kroonenberg 1993; Dypvik and Harris 2001). The lack of correlation between Zr and Ti (Fig. 2; Table 3) implies that they are settled in different minerals and processes. The Zr / Rb ratio has been applied as a sediment grain-size proxy in marine records (Schneider, et al., 1997; Dypvik and Harris 2001; Croudace et al., 2006; Campagne et al., 2015). Zr / Al has been interpreted as an indicator for the accumulation of heavy minerals due to bottom currents (Bahr et al., 2014). In our cores, Zr / Rb and Zr / Ti ratios have a near-identical variability down-core (Fig. 2). We utilize the high-amplitude Zr / Ti signal in our records as an indicator of larger grain size and current velocity (Fig. 2). The Zr / Ti ratio varies between 0.1 and 1 and exhibits maximum values within F2 showing an increasing upwards or bigradational patterns (Fig. 2). Although minimum Zr values (cps) are found in F1, laminations with coarser-grained sediment within this claystone facies are also characterized by elevated Zr values similar to those in F2 (Figs. 3, 4; Table 3). The Zr / Ti pattern is positively correlated with magnetic susceptibility throughout the studied interval (Fig. 2).

The Zr / Ti, Zr / Rb, and Zr / Ba ratios covary characterizing the laminations within F1 and the alternation between F1 and F2 by defining the contacts between them (Figs. 2, 4). They also mark the coarsening upwards or bigradational tendency in F2 (Fig. 4). Of the three ratios, the Zr / Ba ratio is the one that highlights these patterns best (Figs. 2, 4).

Barium (Ba) is present in marine sediments mainly in detrital plagioclase crystals and in the form of barite (BaSO4; Tribovillard et al., 2006). In the studied sediments, Ba and Ti have a correlation factor of r2 = 0.66 (Table 3), which is taken to indicate that Barium is predominantly present as a constituent of the continental terrigenous fraction and/or that biogenic barite was sorted by bottom currents. Ba has maximum values (10 000 total counts) at the base of F1 and decreases upwards in a sawtooth pattern, reaching minimum concentrations within F2 (5000 total counts) (Fig. 2; Table 3). The detrital fraction of Ba in the open ocean has been used in other studies as a tracer of shelf waters (Moore and Dymond, 1991; Abrahamsen et al., 2009; Roeske, 2011), and the Ba record is also affected by current intensity in other depositional contourite systems (Bahr et al., 2014) preventing its use as paleo-productivity proxy in environments dominated by contour currents.

Variations in Ca, Mn, and Sr are strongly intercorrelated (Fig. 2) with r2 > 0.87 (Table 3). Biogenic calcite precipitated by coccoliths and foraminifera have greater Sr concentration than inorganically precipitated calcite or dolomite (Hodell et al., 2008). The positive Ca and Sr correlation could therefore potentially be used to differentiate between terrigenous Ca sources (e.g., feldspars and clays) and biogenic carbonates (e.g., Richter et al., 2006; Foubert and Henriet, 2009; Rothwell and Croudace, 2015). Based on these observations, we interpret the Ca in our sediments as being mainly of biogenic origin (CaCO3). This interpretation is supported by HR-SEM images taken from carbonate-rich intervals of F2, which show abundant coccoliths (Fig. 3d). Peaks in Ca in our record (Fig. 2) coincide with the carbonate-rich layers listed in the previous section. Additional peaks in the record may indicate carbonate-rich layers that we have been unable to identify visually.

In order to estimate the CaCO3 content continuously throughout the studied interval we use a calibration (rU13562= 0.81) between the natural logarithm (ln) of the Ca / Ti ratio (ln(Ca / Ti)) from the XRF core scanner data and the XRF discrete CaCO3 measurements (weight %) from Site U1356 as applied in other studies (Zachos et al., 2004; Liebrand et al., 2016) (Fig. 5). “CaCO3 est.” is used throughout the text to refer to carbonate content estimated by the ln(Ca / Ti) ratio. CaCO3 est. concentrations are generally low (between 0 and 16 %). Carbonates are mostly present in F2, varying between 5 and 16 %, although small contents (from 0 to 5 %) can be seen in the intervals of F1 with scarce laminations (Fig. 4). CaCO3 est. peaks in some intervals have a particular morphology producing a double peak in the beginning and/or the end of bioturbated F2 (Figs. 2, 4).

Mn(II) is soluble under anoxic conditions and precipitates as Mn(IV) oxyhydroxides under oxidizing conditions (Tribovillard et al., 2006). Manganese is frequently remobilized to the sedimentary pore fluids under reducing conditions. Dissolved Mn can thus migrate in the sedimentary column and (re)precipitate when oxic conditions are encountered (Calvert and Pedersen, 1996). As such, large Mn enrichments primarily reflect changing oxygen levels at the sediment–water interface (Jaccard et al., 2016). The strongly correlated peaks of Mn and Ca (Fig. 2; Table 3) suggest that at least some of the Mn is present in the studied interval as Mn carbonates and/or Mn oxyhydroxides under the oxic sediment–water interphase (Calvert and Pedersen, 1996, 2007; Tribovillard et al., 2006).

Br / Ti has been previously used as an indicator of organic matter in sediments (e.g., Agnihotri et al., 2008; Ziegler et al., 2008; Bahr et al., 2014). Br / Ti in our record shows generally low values (Fig. 2), most likely because the organic matter content in both facies types is relatively low (< 0.5 %, Escutia et al., 2011). However, it exhibits some variability (0.01 to 0.05 Br / Ti ratio) within the two facies, with higher ratio values in F1. Darker colored sediments in F1 are in agreement with these higher Br / Ti values inside F1.

In addition to the elemental analyses of the XRF-scanned data, we use the detrital Al / Ti ratio in discrete XRF bulk sediment samples to reflect changes in terrigenous provenance (Kuhn and Diekmann, 2002; Scher et al., 2015). The Al / Ti ratio varies between 17 and 21, with the highest values found within F1 and the lowest in F2 (Fig. 2).

To detect periodical signals, a spectral analysis of time series was performed on the Zr / Ba and other elemental proxies (i.e., Ba, Zr / Ti, CaCO3, magnetic susceptibility) using the Astrochron R software (Meyers, 2014; Figs. 6, S3–S10).

Multiple-taper spectral analysis method (MTM) in Zr / Ba show a clear and statistically significant (> 90 %) cyclicity every 2 m (0.5 cycles m-1) and at 4.67 m (0.21 cycles m-1) and a less significant one (> 80 %) at 1 m (0.94 cycles m-1) (Fig. S3). On the basis of a linearly calculated sedimentation rate between the two extreme tie points (Table 1), we obtained a sedimentation rate of approximately 5 cm kyr-1. Within this sedimentation rate, the 0.5 cycles m-1 peak corresponds to the 41 kyr obliquity frequency, and the 0.21 and 0.94 cycles m-1 correspond to the 95 and 21 kyr shorter eccentricity periods and precession frequencies, respectively.

After an initial analysis, we ran an evolutive harmonic analysis (EHA) (Astrochron; Meyers, 2014) with three data tapers for the untuned Zr / Ba in depth domain with 2 cm resolution (Fig. S3). The statistical significance of spectral peaks was tested relative to the null hypothesis of a robust red noise background (AR(1) modeling of median smoothing) at a confidence level of 95 % (Mann and Lees, 1996). Despite a short core gap in the middle of the time series, obliquity (41 kyr) dominates throughout the time series (Fig. 6). The sedimentation rates obtained by this method vary between 4.6 and 5.4 cm kyr-1 for the studied section, similar to those obtained with linearly calculated sedimentation rates. Additionally, the Nyquist frequency for Zr / Ba data is 1 m-1 (0.5 kyr), which implies the site is sampled sufficiently to resolve precessional-scale variations; however, core gaps prevent identification of long eccentricity cycles (Fig. S6). Time series were anchored to the more robust paleomagnetic tie point in the U1356 age model, which is 25.99 Ma at 678.78 m b.s.f. (Fig. S7).

Apart from obliquity, spectral analyses of the tuned age model reveal an alignment of the eccentricity and precession bands (Figs. 6, S8). For example, a marked cyclicity at the obliquity periods of 41 kyr is seen at Ba and Zr / Ti (99 % confidence); eccentricity at 100 kyr and precession at 20 kyr (95 % confidence) are also seen (Fig. S9). We also observe coherent power above the 90 % significance level at ∼ 54 and ∼ 29 kyr periods, which are secondary components of obliquity. The anchored age model provides an unprecedented 500 yr resolution (2.5 cm sampling) of the data during the late Oligocene. Orbital frequencies were tested in each core section individually in the Zr / Ba dataset on the depth scale in order to assure that cyclicity is not an artifact related to the gaps in the series (Fig. S10).

Laminated claystones (F1) from Site U1356 were originally interpreted by the shipboard science team to have formed during glacial times relating to variations in bottom current strength and fine-grained terrigenous supply. Conversely, the bioturbated claystones and micritic limestones (F2) were interpreted as resulting from mostly hemipelagic sedimentation during interglacial times (Escutia et al., 2011). Alternations between laminated deposits and bioturbated hemipelagic deposits, similar to those in F1 and F2, have been previously reported to characterize Pleistocene and Pliocene glacial–interglacial continental rise sedimentation, respectively, on this sector of the Wilkes Land margin (Escutia et al., 2003; Patterson et al., 2014). Gravity flows (mainly turbidity flows) are the dominant process during glacial times, resulting in laminated deposits. Interglacial sedimentation is dominated by hemipelagic deposition with higher opal and biogenic content (Escutia et al., 2003; Busetti et al., 2003). Erosion and redeposition of fine-grained sediment by bottom contour currents has also been reported as another important process during Pleistocene and Plio-Quaternary interglacials (Escutia et al., 2002, 2003; Busetti et al., 2003).

Despite being sparse, the occurrence of bioturbation in our laminated sediments in F1, which slightly affects both claystones and silt laminations, indicates slow and continuous sedimentation. This is not consistent with instantaneous turbidite deposition, which would be expected at the Site U1356 located on the left low-relief levee of a contiguous channel during the late Oligocene. It is, however, consistent with fine-grained turbidite overbank deposits being consequently entrained by bottom currents. Silt layer sedimentary structures similar to those described by Rebesco and Camerlenghi (2008) and Rebesco et al. (2014) indicate that there is current reworking of the sediments. For example, silt layers can be continuous or discontinuous with wavy and irregular morphologies, and within layers, sedimentary structures such as cross laminations are common (Fig. 3c–f). Within the cross laminae, mud offshoots, and internal erosional surfaces are distinctive features of fluctuating currents where successive traction and suspension events are superimposed, indicating bottom current sedimentation as the principal process for the F1 laminated claystones (Shanmugam et al., 1993; Stow, 2002). Based on these observations, we interpret F1 as glacial laminated muddy contourites following the classification of Stow and Faugères (2008). The F1 sedimentary structures suggest bottom currents with fluctuating intensities, which result in laminations and internal structures forming during peak current velocities (Lucchi and Rebesco, 2007; Martín-Chivelet et al., 2008; Rebesco et al., 2014). Laminated, fossil-barren, glaciogenic deposits, consistent with those of Facies F1, have been observed on younger sedimentary sections in glaciated margins and interpreted as contour current-modified turbidite deposits and as muddy contourites (Anderson et al., 1979; Mackensen et al., 1989; Grobe and Mackensen, 1992; Pudsey, 1992; Gilbert et al., 1998; Pudsey and Howe, 1998; Pudsey and Camerlenghi, 1998; Anderson, 1999; Williams and Handwerger, 2005; Lucchi and Rebesco, 2007, Escutia et al., 2009). This particular type of contourite facies is associated with glaciomarine deposition during times of glacial advance and has been interpreted as resulting from unusual, climate-related, environmental conditions of suppressed primary productivity and oxygen-poor deep waters (Lucchi and Rebesco, 2007).

Bioturbated sediments in F2 were previously interpreted as interglacial hemipelagic deposits (Escutia et al., 2011). In this study, we interpret F2 as hemipelagic and overbank deposits reworked by bottom currents. The coarser grain size in F2 compared to F1 (silty clay matrix as seen in HR-SEM, Fig. 3g–j), the distribution of heavy minerals as indicated by the Zr / Ba, and the elevated values of the magnetic susceptibility record with a bigradational pattern within the facies (Figs. 2, 4) support the notion that interglacial sediments of F2 have been heavily modified by bottom currents. Hemipelagic sediments are expected to be homogeneous in terms of grain size and grading is not expected. Current winnowing of hemipelagic deposits and the removal of the fine-grained fraction can produce the higher accumulation of heavy (indicated by the Zr) and ferromagnetic (indicated by MS) minerals observed in F2 compared to F1 (Fig. 2; Table 2). High MS values result from stronger bottom current deposition and/or increased terrigenous input (e.g., Pudsey, 2000; Hepp, 2007). Also, bigradational trends have been previously described in contourite sediments and interpreted as recording an increase followed by a decrease in the current velocities (e.g., Martín-Chivelet et al., 2008). The bigradational patterns in the Zr / Ba and MS plots (Figs. 2, 4) are therefore interpreted as depicting a constant and smooth increase followed by a decrease in current velocity with little gradual changes in flow strength. In addition, the presence of grains of quartz with conchoidal fractures and reworked coccolithophores with signs of dissolution (Fig. 3h, j) support the reworking of background hemipelagic and turbidite overbank sediments by bottom currents in a high-energy environment (Damiani et al., 2006). Following the classification by Stow and Faugères (2008), we interpret that F2 has more silty massive contourites resulting from higher and more constant bottom current velocity compared to F1.

Transitions between the F1 and F2 facies are characterized by glacial-to-interglacial contacts that may be sharp or diffuse due to bioturbation and are characterized by a gradual change in physical and geochemical sediment parameters (Figs. 3, 4; Table 3). Interglacial-to-glacial contacts (F2 to F1), on the other hand, are characterized by an apparently non-erosional sharp lithological boundary. The sharp lithological boundaries between interglacial to glacial transitions can be explained by maximum current intensities achieved at the end of the interglacials (Shanmugam, 2008; Rebesco et al., 2014).

Early Oligocene and post-mid Miocene climate transition sediments from Site U1356 contain granule and larger clasts (> 2mm) interpreted as IRD (Escutia et al., 2011; Sangiorgi et al., 2018; Fig. S1). In addition, dinocyst assemblages indicate the presence of sea ice (Houben et al., 2013). Based on this, one could expect the site to be within the reach of icebergs calving from an expanded ice sheet grounded at the coast or beyond in the late Oligocene. This is supported by Pliocene–Pleistocene sedimentary sections in adjacent continental rise sites containing IRD (Escutia et al., 2011; Patterson et al., 2014). In addition to the paucity of IRD in our studied interval, the absence of sea-ice-loving species Selenopemphix antarctica, and abundant gonyaulacoid phototrophic dinocysts suggest warm to temperate surface waters (Bijl et al., 2018b). A sea-ice-free scenario during the late Oligocene is also supported by elevated sea surface temperatures (i.e., average summer temperatures are ∼ 19 ∘C) based on biomarker sea surface temperatures (TEX86 data in Hartman et al., 2018). Furthermore, the presence of in situ terrestrial palynomorphs suggests that during the late Oligocene, margins nearby were in part free of ice sheets and covered by a cool to temperate vegetation with trees and shrubs (Salzmann et al., 2016; Strother et al., 2017). All these observations suggest a reduced ice sheet and partly ice-free margins in the Wilkes margin during the late Oligocene.

These observations are consistent with the iceberg survivability modeling in the Southern Ocean for the warm Pliocene intervals, which shows the distance that icebergs could travel before melting was significantly reduced (Cook et al., 2014). Warm Pliocene summer sea surface temperatures up to 6 ∘C warmer than today during interglacials and prolonged Pliocene warm intervals have been reported in the Ross Sea (e.g., Naish et al., 2009; McKay et al., 2012) and other locations around Antarctica (Whitehead and Bohaty, 2003; Whitehead et al., 2005; Escutia et al., 2009; Bart and Iwai, 2012). Contrary to what we observe in our late Oligocene record and in the Miocene Climatic Optimum (Sangiorgi et al., 2018), abundant IRD was delivered to continental rise sites adjacent to Site U1356 during the warm Pliocene (Escutia et al., 2011; Patterson et al., 2014). This was interpreted by Cook et al. (2017) to suggest that a considerable number of icebergs (iceberg armadas) had to be produced in order to reach the site under these warm Pliocene conditions. We argue that the lack of IRD delivery to Site U1356 during the studied warm late Oligocene interval can result from the different WSB late Oligocene paleo-topographic setting. Paleo-topographic reconstructions from 34 Myr ago (Wilson et al., 2012) and the early Miocene (Gasson et al., 2016) show the WSB to be an area of lowlands and shallow seas in contrast to the overdeepened marine basin that it is today (Fretwell et al., 2013). This paleo-topographic configuration would have precluded widespread marine ice sheet instability during the Oligocene. This difference is important, as an ice sheet grounded on an overdeepened continental shelf can experience marine ice sheet instability, a runaway process relating to ice sheet retreat across a reverse slope continental shelf (Weertman, 1974), which is proposed to be a driver for retreat of the EAIS in the WSB during the warm Pliocene (Cook et al., 2013). Conversely, a shallower continental shelf allows for the potential expansion of grounded ice sheets into the marine margin during warmer-than-present climates (Wilson et al., 2012), and thus direct records are required to assess the climate threshold for such an advance.

In comparison to the distal U1356 Wilkes Land margin record, the Ross Sea Embayment ice proximal sediments obtained by the CRP contain Oligocene to Early Miocene palynomorphs, foraminifera, and clay assemblages that point to a progressive decrease in fresh meltwater, cooling, and intensifying glacial conditions (Leckie and Webb, 1983; Hannah et al., 2000, 2001; Raine and Askin, 2001; Thorn, 2001; Ehrmann et al., 2005; Barrett, 2007). Therefore, the coastal CRP sediment record does not support a significant loss of ice or warming during the late Oligocene (Barrett, 2007). The high sedimentation rates during the late-Oligocene–early-Miocene recorded at Deep Sea Drilling Project (DSDP) Site 270 were interpreted as reflecting turbid plumes of glaciomarine sediments derived from polythermal-style glaciers or ice sheets that were calving into an open Ross Sea, without an ice shelf (Kemp and Barrett, 1975). In addition, seismic data indicate that during the late to mid-Oligocene widespread expansion of a marine-based ice sheet onto the outer Ross Sea shelf did not take place, but instead glaciers and ice caps drained from local highs and advanced only into shallow marine areas rather than there being a whole-scale marine ice sheet advance (Brancolini et al., 1995; De Santis et al., 2013; Bart and De Santis, 2012).

Combined, this evidence suggests that during the late Oligocene, marine-terminating glaciers, ice caps, and glaciers persisted along the Transantarctic Mountain front reaching the Ross Sea coastal areas, but it may have been more confined within a warmer WSB margin. This is also supported by vegetation reconstructions derived from fossil pollen from both margins, which, for the middle Miocene and late Oligocene, indicate higher terrestrial temperatures and more tree taxa at Wilkes Land (Salzmann et al., 2016; Sangiorgi et al., 2018) than the Ross Sea (Askin and Raine, 2000; Prebble et al., 2006). This is consistent with the ice sheet modeled configuration for Miocene topographies with CO2 scenarios of 500–840 ppm (Gasson et al., 2016; Levy et al., 2016; Fig. 7).

Sediment physical properties and geochemical signatures of F1 and F2 are here related to changes in bottom-water–sediment interphase oxygenation/ventilation during successive glacial and interglacial periods (Table 2). We interpret that these changes are linked to shifts in water masses driven by a north–south displacement of the position of the westerlies and associated changes in the intensity of frontal mixing or location of the polar front and Antarctic Divergence (Fig. 7). Based on our observations, we propose a model to explain the interpreted changes in bottom water conditions at Site U1356 during successive glacial and interglacial times (Fig. 7).

The first spectral analysis on late Oligocene sediments from the eastern Wilkes Land margin at Site U1356 shows that glacial–interglacial cycles, resulting in changes in the oceanic configuration off Wilkes Land, are paced with variations in Earth's orbit and seasonal insolation. Although the data are somewhat discontinuous due to gaps in our record, they clearly show that the glacial–interglacial cyclicity (every 2 m or 41 kyr) discussed above has a persistent obliquity pacing throughout the late Oligocene interval studied (26–25 Ma) in the Wilkes Land. Consequently, this obliquity-paced cyclicity modulates the amount of deep-water production in the Southern Ocean and exerts a major control on oceanic configuration and current strength. Bottom current velocity fluctuations and ventilation of bottom sediments respond to the forcings applied by the strength of the Southern Hemisphere westerlies, by the position of the polar front (PF) in respect of the site, and consequently by the water mass occupying the bottom of the basin at each time. In addition to obliquity, precession is also present, which implies a dynamic response of the EAIS and offshore oceanic water masses to orbital forcing.

East Antarctic ice volume fluctuations at orbital periodicities in the obliquity band in the Wilkes Land margin have been previously reported from early warm Pliocene (3–5 Ma) sediments obtained from Site U1361 (Patterson et al., 2014). In the Ross Sea, cyclicity in sediments collected by the CRP from the late Oligocene, the late Miocene, and the early warm Pliocene period was also paced by obliquity (Naish et al., 2001, 2009; McKay et al., 2009). Similar orbital variability in the deep-water circulation patterns has also been inferred to have occurred with the growth of the EAIS during the middle Miocene between 15.5 to 12.5 Ma (Hall et al., 2003). In addition, other studies have linked changes in Atlantic meridional overturning (Lisiecki et al., 2008; Scher et al., 2015) and Antarctic circumpolar ocean circulation (Toggweiler and Russell, 2008) to obliquity forcing. An interglacial mechanism has been proposed whereby the southward expansion of westerly winds and associated Ekman transport is compensated for by enhanced upwelling of warmer, CO2-rich CDW (Toggweiler et al., 2008), which also promotes atmospheric warming. In the equatorial Pacific, Pälike et al. (2006) also report strong obliquity in the benthic δ13C isotopic record between 26 and 25 Myr, implying that changes in the carbon cycle (pacing glacial /interglacial periods) are triggered in the high southern latitudes and transferred to the global deep-ocean through the bottom water masses.