Quartz grain microtextures illuminate Pliocene periglacial sand fluxes on the Antarctic continental margin

On high‐latitude continental margins sediment is supplied from land to the deep sea through a variety of processes, including iceberg and sea‐ice rafting, and bottom current transport. The accurate reconstruction of sediment fluxes from these sources through time is important in palaeoclimate reconstructions. The goal of this study was to assess a shift in the intensity of glacial processes, iceberg and sea‐ice rafting during the Pliocene through an investigation of coarse sediment deposited at the AND‐2A site in the Ross Sea and at International Ocean Discovery Program Site U1359 on the Antarctic continental rise. Terrigenous particle‐size distributions and suites of quartz grain microtextures in the sand fraction of the deep‐sea sediments were compared to those from Antarctic glaciomarine diamictites as a baseline for proximal glacial sediment in its source area. Using images acquired through Scanning Electron Microscopy, and following a quantitative approach, fewer immature and potentially glacially transported grains were found in Pliocene deep‐sea sand fractions than in ice‐contact sediments. Specifically, in the lower Pliocene interval silt and fine sand percentages are elevated, and microtextures in at least half of the sand fraction are inconsistent with a primary glacial origin. Larger numbers of chemically altered and abraded grains in the deep‐sea sand fraction, along with microtextures that are diagnostic of periglacial environments, suggest a role for eolian sediment transport. These results highlight the anomalous nature of high‐latitude sediment fluxes during prolonged periods of ice retreat. Furthermore, the identification of a significant offshore sediment flux during Antarctic deglaciation has implications for estimated nutrient supply to the Southern Ocean and the potential for high‐latitude climate feedbacks under warmer climate states.


| INTRODUCTION
Polar ice sheets are key variables in understanding the Earth's climate history. Traditionally, high-resolution sediment records of ice-rafted debris (IRD) have been widely used as a proxy for glacial activity. These records involve interpretations of sand-dominated coarse fraction fluxes to the deep-marine environment. Iceberg-rafted debris (IBRD) upon its initial release is expected to have a similar composition to a till deposited directly from glacial ice. One approach is based on the assumption that the accumulation of debris with a grain diameter larger than 150 or 250 µm is proportional to the supply of glacially eroded coarse fraction from melting icebergs, a proxy called IBRD. Another is to use end-member modelling of grain-size distributions to assess IBRD abundance (Prins et al., 2002). Either approach implies that the supply of coarse debris by icebergs is constant in volume and size range and that all other processes that supply sand or coarser material to the deep sea are well-constrained. However, these conditions pose a challenge in analysing pre-Pleistocene stratigraphic records representing altered sediment fluxes under different climate states (Gilbert & Domack, 2003;Westerhold et al., 2020).
Here the transport history of the sand components recovered from Pliocene Antarctic drillcores is evaluated through detailed analyses of terrigenous particle-size distributions along with scanning electron microscopy (SEM) of quartz grain microtextures. Previous work has shown that high-resolution records of the abundance of coarse sand recovered from deep-sea drillcores on the Wilkes Land margin indicate a response to orbital forcing in the obliquity and precession bands (Hansen et al., 2015;Patterson et al., 2014). However, the contribution of the Neogene Antarctic Ice Sheet to global ice volume changes in the precession band is debated (Caballero-Gill et al., 2019;DeVleesschouwer et al., 2017), warranting investigation of the precise nature of the ice-rafted sediment fluxes (Gilbert, 1990).
An existing Pliocene high-resolution laser particlesize record (Hansen et al., 2015) from International Ocean Discovery Program (IODP) Site U1359 is extended across the Mid-Pliocene Warm Period (MPWP) and the relative proportions of the sand fluxes attributed to ice rafting and other sediment transport mechanisms for this deep-water site are investigated. Furthermore, a baseline for glacially derived material is provided by analyses of discrete intervals of Neogene ice-proximal diamictites recovered from the ANDRILL Site 2A (AND-2A) core in the Ross Sea (Fielding et al., 2008). The Ross Sea is considered to be the primary provenance area for Pliocene IBRD recovered from the Wilkes Land continental rise (Cook et al., 2017).

| METHODS AND MATERIALS
The AND-2A drillhole was completed in 2007 in ca 380 m water depth on a sea-ice platform within 10 km of the East-Antarctic coast in the Ross Sea ( Figure 1). In the upper ca 250 m, drilling recovered Neogene massive diamictites interbedded with minor stratified diamictites, sandstones and conglomerates . Another interval of massive and stratified diamictite was recovered at ca 650 m below sea floor (mbsf). Based on macroscopic evidence of ice-proximal deposition, 46 samples were selected for particle-size analyses. Four diamictite intervals ranging from ca 6 to 14 m thick were selected for sampling at 0.5-1.0 m spacing to capture the vertical variability (Hansen, 2011). The selected intervals include diamictites with shear structures and clastic dikes, and range in age from Miocene to Pleistocene (Figure 2) (Fielding et al., 2008;Passchier et al., 2011).
The IODP Site U1359 was drilled in 4,003 m water depth within 100 km of the shelf break on the Wilkes Land margin and recovered Pliocene sediments between ca 50 and 150 mbsf (Figure 1). At Site U1359, as part of this study, 110 samples were analysed across the MPWP, and merged with the previously published record of Hansen et al. (2015). Pliocene sediments recovered at IODP Site U1359 consist primarily of silty clays and diatomaceous silty clays with dispersed clasts (>2 mm) with up to ca 43% biogenic silica (Hansen et al., 2015). Site U1359 was sampled at ca 3 kyr temporal resolution using an age model from Tauxe et al. (2012), which translated into a sample spacing of approximately 15 cm. Most samples were acquired from Hole U1359A. In one interval between 104.63 and 114.63 mbsf-A, where Hole U1359A had recovery gaps and poor core quality, equivalent strata were sampled in Hole U1359B (Hansen et al., 2015). Each sample consisted of an intact wedge of sediment from a ca 1.5 cm thick core interval, which was divided into subsamples prior to disaggregation.

| Particle-size data
Particle-size distributions were measured on a Malvern Mastersizer 2000 laser particle-size analyser. The Udden-Wentworth grain-size classification was used (Wentworth, 1922). For indurated sediments, samples were first soaked in Millipore water to create a slurry. Samples were wetdisaggregated using a rubber cork, or pestle applying gentle vertical motion to the sample slurry to avoid breaking grains. Smear slides of samples were checked under a microscope to assess proper disaggregation before analysing them on the laser particle sizer. Ultrasonic treatment was used as a last resort to aid in the disaggregation. This process is labourintensive, but has proven to be very effective given the excellent correlation between mud percent and geophysical properties data for the AND-2A core  Geosphere, their Figure 2B). To obtain the terrigenous fraction, organic matter and carbonate were digested through the addition of 30% H 2 O 2 and 10% HCl to a 50-100 ml suspension on a hot plate. For IODP Site U1359, in order to assess the particle-size distribution of the siliciclastic sediment supplied from land, the data were collected on the terrigenous fraction of the marine sediment after dissolution of the biogenic silica, which typically consists of diatoms in the silt size range. Using data from the terrigenous fraction, allows for a direct comparison between particle-size distributions of ice-contact diamictite from the Antarctic continental shelf and the ice-distal sand fractions, including IRD. Following centrifuge cycles to remove supernatant with excess chemicals, at Site U1359 biogenic silica was removed by mixing sediment with a 0.2 N NaOH solution in an 85-degree hot bath for 1 h (Passchier, 2011). This standard method of wet-alkaline digestion is sufficient to remove most biogenic silica in hemipelagic sediments with low amounts of biogenic silica, while minimising loss of the lithogenic clay fraction (Cardinal et al., 2007;DeMaster, 1981;Ragueneau et al., 2005).  The Malvern Mastersizer 2000 is equipped with a Hydro 2000 MU wet sample dispersion unit. Prior to instrument analysis, the terrigenous fractions were dispersed using a heated sodium pyrophosphate solution. The Mastersizer 2000 uses Mie theory to provide calculations of the fine sediment fractions, which requires an estimate of the refractive index of the material. For the diamict samples, a standard operating protocol with refractive index 1.544 (quartz), absorption coefficient of 0.9, and rotor speed 2,200 rpm was followed. For the deep-marine samples, a refractive index of 1.6 (illite), absorption coefficient of 0.9 and rotor speed of 2,000 rpm was used. The rotor speed is a tradeoff between the proper dispersion of coarse materials and the increased incidence of air bubbles at higher rotor speeds. Other Malvern instrument variables were the same for all samples. Standard operating protocols are based on experiments with fine-grained sediments by Sperazza et al. (2004) and the authors own experiments with diamict aliquots. Quality control was performed through repeat analysis of a fine-grained industrial standard QAS3002 and an in-house natural fine sand standard (Sandy Hook Dune 4). Eight replicates of a fully dispersed till sample (D 50 = 25 µm; Uniformity = 3.4) show that results per 0.25 phi size class are typically reproducible within an uncertainty of <10%, as long as obscuration values are kept between 15% and 50%. Eleven replicates of the coarser dune F I G U R E 2 Core images of AND-2A drillcore sections sampled for both SEM and particle-size analysis. Depth range is in metres below sea floor (mbsf). Samples originate from the center of each of the intervals. Sample depth below sea floor and lithology is listed in Table 1 568 | PASSCHIER Et Al. standard (D 50 = 354 µm; Uniformity = 0.3) show a typical relative uncertainty of <15% per 0.25 phi size class. However, larger relative uncertainty is observed at the edges of the grainsize distributions, where there is a large difference in vol.% between two adjacent size classes.

| SEM analysis of quartz grain microtextures
For the microtexture analysis, 786 SEM images of quartz grains from sieved >63 µm sand fractions of 20 samples were examined, 10 from each site. In the ANDRILL core, SEM samples were chosen based on their proximity to shear structures or other evidence of ice-contact deposition (Hansen, 2011; Table 1; Figure 2). At Site U1359, samples with large quantities of terrigenous sand and some samples with lower quantities were selected for SEM analysis to investigate the sediment transport history for the sand fraction (Hansen, 2016;Rosenberg, 2014; Table 1; Figure 3). Samples were wet sieved over 63 µm and ca 40 grains were picked quickly and without examination from the >63 µm fraction. The vast majority of the grains that were picked were in the 200-500 µm size range. All grains were coated with a thin layer of gold-palladium. The images were created in secondary electron (SE) mode at 12kV with a ca 10 mm working distance on a Hitachi S3400N Scanning Electron Microscope at the Microscopy and Microanalysis Research Laboratory (MMRL) at Montclair State University. Grain composition was checked using a Bruker X-Flash energy dispersive X-ray spectrometer (EDS) system and only the 786 grains that had a SiO 2 composition were analysed for microtextures. A quantitative approach to microtexture analysis was used, focusing on combinations of microtextures on individual grains for palaeoenvironmental interpretations (Cowan et al., 2014;Damiani et al., 2006;Dunhill, 1998;Hart, 2006;Hodel et al., 1988;Mahaney et al., 2001). Initially, the observation and characterisation of individual microtextures involved comparison of textures on grains of unknown origin to grains from known sedimentary environments (Helland & Holmes, 1997;Mahaney, 2002 and other references therein). Microtextures were observed visually following a checklist by the second author (Hansen, 2011(Hansen, , 2016. The work on each site was part of a different project. The checklists for each of the palaeoenvironments were slightly different because of variability in the microtextures that were encountered in the samples derived from the diamictites versus the glacially influenced deep-marine environment. In a sensitivity study, Culver et al. (1983) demonstrated that operator variance existed in the classification of individual microtextures, but also that the environmental interpretations based on blind surveys of sets of multiple microtextures on each grain, along with grain roundness and relief, were consistent and accurate for five different operators. Therefore, the use of combinations of microtextures is preferred in environmental discrimination (Mahaney et al., F I G U R E 3 Core images of IODP Site U1359 drillcore sections sampled for SEM and particle-size analysis. Samples originate from the center of each of the intervals. Sample depth below sea floor and lithology is listed in Table 1 570 | PASSCHIER Et Al. 2001). Furthermore, the presence of grains recycled from pre-existing sediments and sedimentary rocks can create a strong provenance-related overtone in the texture tallies in studies of glacial environments (cf. Mazullo & Ritter, 1991). To avoid this problem, observations on the various layers of textures that overprint each other are also made on the entire grain surface with the aim to separate textures produced during the multiple cycles of surface exposure, erosion, sediment transport and deposition (cf. Mahaney et al., 2001;Molén, 2014;Woronko & Pisarska-Jamroży, 2016). In the workflow of SEM studies of quartz grain populations, it is uncommon that microtextures are observed separately from the context of the entire grain, including its other textures and the grain shape. Therefore, the microtextures that are tallied for each grain are typically not independent observations and, in such cases, the use of statistical methods on microtexture checklists to underpin environmental interpretations is not a valid approach.
For these reasons, SEM images were reanalysed by the first author to classify grains visually into grain types, with criteria established based on expected source-to-sink sediment transport histories and with emphasis on the multi-cyclic nature of sediment transport before deposition (Mahaney et al., 2001;Molén, 2014). Individual grains were attributed to one of 12 grain types based on a survey of combinations of microtextures, grain roundness, and relief. Each of the 12 grain types was interpreted as a product of a specific sediment transport path and depositional environment based on the genetic interpretation of combinations of microtextures, and how textures overprinted each other on the surfaces of the grains.
Image file ID names and grain sources were randomised using a script prior to analysis and the origin of each image, that is, which AND-2A or U1359 core sample, was unknown to the operator to avoid subconscious bias. The entire batch of 786 grains was classified into grain types before image file names were reinstated. This process created a raw dataset with 786 completely independent observations. Prior to analysis, values were standardised by dividing the frequency for each grain type by the average frequency for that grain type. The statistical analysis was carried out with the Principal Coordinates Analysis module using Bray-Curtis Similarity with a transformation coefficient of 2 in PAST v. 3.26b (Hammer et al., 2001). The algorithm is from Davis (1986). The Bray-Curtis Similarity was chosen over the Euclidian Distance or other methods because of its better handling of 'zero' counts in observations. Principle Coordinates Analysis  is more commonly known as Metric Multidimensional Scaling (MDS) and is used to determine whether a collection of observations represents a single population or a mixture of several populations.

| RESULTS AND INTERPRETATIONS
Particle-size distributions show variability in sand content in the four densely sampled intervals of Miocene, Pliocene and Pleistocene diamictites in the AND-2A core in the Ross Sea ( Figure 1). Stratified muddy diamict (at 648-663 and 238-250 mbsf) is homogeneous over a depth range and predominantly silt-rich, consisting of gravel-sized (>2-mm diameter) clasts within a matrix of glacial rock flour (Figure 1). In contrast, Upper Miocene to Pleistocene massive diamictites (134-142 and 64-70 mbsf) show both fine and coarse sand modes, even in repeat analysis of the same samples. Such till matrix heterogeneity can be attributed to differences in the efficiency of glacial comminution processes, where grain crushing dominates over abrasion under high effective pressures (Cowan et al., 2014;Hiemstra & Van der Meer, 1997). At deep-water Site U1359, coarse sand dominates the >125 µm dispersed sand fraction in the upper Pliocene section ( Figure 4A), with a complete absence of fine sand in the entire interval sampled between 48 and 70 mbsf (Figures 1  and 4B). In contrast, lower Pliocene hemipelagic and diatomrich silty clays have variable terrigenous grain-size distributions with mostly fine sand modes and only occasionally coarse sand ( Figure 4C,D). Observations of number of grains carrying each microtexture in each of the checklists for sample intervals and individual samples are tallied here as bar graphs ( Figure 5). Noteworthy is that fractured plates are present on >50% of grains and v-shaped impact pits on >25% of grains in the diamictites, along with microtextures, such as straight and arcuate steps, conchoidal fractures and grooves. Grains exhibiting upturned plates and mechanical impact microtextures, such as impact craters, chattermarks, abrasion features and rounded edges were observed in the samples from the glacially influenced deep-marine environment (Site U1359).
In the reanalysis, 12 grain types were distinguished based on the examination of co-existing microtextures on each grain ( Figure 6 and Table 2). Samples within close proximity in a core interval, are plotted together in the histograms, and have similar grain-type distributions (Figure 7). Modified glacially derived grains with high relief, steps, parallel fractures, conchoidal fractures, microblocks and edge rounding (Type D) were the most common grain type found overall at both sites (Figures 7 and 8). In contrast, low-medium relief subrounded to rounded grain types G and H (Figure 6), which resemble those found in lowland rivers such as the Nile and Loire (Manickam & Barbaroux, 1987;Vos et al., 2014) were rare in these Antarctic samples. Medium relief rounded grains with v-shaped and dishshaped percussion marks, conchoidal fractures and grain breakage (Type F), indicative of high-intensity subaqueous transport are also rare, except for one interval of sandy diamictite in AND-2A at 250.11 m, and one interval above a normally graded lamina in U1359 at ca 62 mbsf (sample U1359A-8H4). Not unexpectedly, the diamictites from the Ross Sea continental shelf exhibited the largest proportion of high-relief angular Type C grains with steps, parallel fractures, conchoidal fractures or microblocks (Figure 8), surface textures that are found on grains from modern and recent glacial environments (Hodel et al., 1988;Mahaney et al., 1996;Molén, 2014;Whalley & Krinsley, 1974;Whalley & Langway, 1980). In contrast, Type I grains that are almost entirely covered with polygenetic upturned plates, saw-tooth fractures or broken plates (Figure 9) are surprisingly abundant in deep-marine sediments at Site U1359, but rare in the diamictites ( Figure 7A). The hemipelagic sediments also contain common high-relief Type E grains with smoothed/abraded fracture surfaces, dish-shaped depressions, edge rounding and precipitation (Figure 9). These grain types are found in periglacial or high elevation desert environments where perennial lake, river and/or eolian sediment transport processes are impacting otherwise immature quartz grains (Hao et al., 2019;Kalińska-Nartiša et al., 2017;Li et al., 2020;Mahaney, 2015;Margolis & Krinsley, 1971;Shrivastava et al., 2014;Wellendorf & Krinsley, 1980;Woronko & Hoch, 2011).
The MDS statistical analysis reveals a correlation between macroscopically defined sedimentary facies and microtextures ( Figure 7B). The first two coordinates explain 51% of the variance. Samples plotting near each other on the ordination plot are more similar to each other than samples plotting far away from each other. Diamicts plot in a narrow range on the first coordinate axis (MDS 1) with limited overlap between massive and stratified diamicts. Most deep-water samples plot in a narrow range on the | 573 PASSCHIER Et Al.  second coordinate axis (MDS 2; Figure 7B). However, upper Pliocene samples from IODP Site U1359, including sample U1359A-7H4-90-92, have greater similarity in grain microtextures to stratified diamicts than lower Pliocene samples, as is also evident from the grain-type frequency distributions ( Figure 7A). The least similar samples to diamictites are from Cores U1359B-13H and U1359A-15H. Samples from these cores also yielded 5 grains of altered and abraded vesicular glass as part of the sand fraction investigated via SEM analysis (Figure 10).

| DISCUSSION
To investigate the relative importance of Pliocene iceberg rafting from the Ross Sea as a sediment dispersal mechanism of coarse sediment, suites of quartz grain microtextures are compared between glacially influenced IODP Site U1359 on the Antarctic Wilkes Land continental rise and its IRD source represented by the AND-2A diamictites. Subsequently, sand fluxes are evaluated using terrigenous particle-size distributions.

| Microtexture frequencies
One challenge with the environmental interpretation of raw frequencies of quartz grain microtextures ( Figure 5) is the strong provenance signal in the texture distributions in grains from glacial environments (Mazullo & Ritter, 1991;Moss & Green, 1975). In Cenozoic tills deposited in valleys in the uplands of the Transantarctic Mountains, conchoidal and sublinear fractures are abundant, and v-shaped impact pits are also present (Mahaney et al., 1996). In samples of AND-2A diamictites the high frequencies of fractured plates, breakage blocks, in addition to v-shaped impact pits ( Figure 5) points to glacial erosion and transport of quartz derived from crystalline and/or immature sedimentary rocks (Mazullo & Ritter, 1991). Rounding of grains and an increase in v-shaped impact pits on glacially sourced grains have been observed elsewhere after only ca 80 km of subaqueous transport (Sweet & Brannan, 2016). However, in this study these textures are more consistent with inheritance of grain morphologies through glacial plates and mechanical impact microtextures, such as impact craters, chattermarks, abrasion features, saw-tooth fractures and rounded edges ( Figure 5). These textures are more consistent with transport in different kinds of flows by wind and water (Costa et al., 2013;Lindé & Mycielska-Dowgiałło, 1980). The microtexture checklist surveys show differences in the observed frequency distributions of microtextures per sample between ice-contact diamictites and glacially influenced deep-marine sediments with evidence of ice rafting. In sediment transport from source to sink, the likelihood that the original texture inherited from the source rock is preserved decreases, unless grains stay embedded in the glacial ice or an iceberg until deposition.

| Grain-type classification and palaeoenvironmental interpretation
The classification of entire grain surfaces allows for the characterisation of multiple consecutive sediment transport modes and palaeoenvironment (Table 2). This is achieved by separating the inherited from contemporaneous textures within the context of how textures coexist and overprint each other on individual grains (Molén, 2014). Sand dispersal to the Antarctic continental rise takes place as IBRD, sea-icerafted debris (SIRD) and hyperpycnal flows (Cowan et al., 2008;Damiani et al., 2006;Hansen et al., 2015;Patterson et al., 2014). Sediment can be supplied to sea ice through eolian transport, as freeze-on, anchor ice floatation, frazil ice formation and wash overs (Powell & Domack, 2002). Grain types recovered from subglacial tills, that is, the AND-2A massive diamictites with shear structures, represent the population in the ice source for ice rafting, and modification of grain surfaces by eolian and subaquatic sediment transport is expected in the proglacial or glaciomarine environment (Dunhill, 1998;Hodel et al., 1988;Kalińska-Nartiša et al., 2017;Sweet & Brannan, 2016).
In deep-marine sediment samples from Site U1359, between 23% and 58% of grains have either fresh or modified glacial microtextures, such as steps, parallel and/or conchoidal fractures, or microblocks (Types B, C and D; Figure 8), compared to between 48% and 75% of grains in the Ross Sea diamictites (Figure 7). The lower number of grains with primary glacial microtextures (Types B, C and D) than the baseline ice-proximal diamictite is inconsistent with a sole IBRD source for the sediment at deep-marine Site U1359. Between 6% and 39% of grains from the deep-water site are of Type I and have textures, such as upturned plates and saw-tooth fractures with dissolution ( Figure 9). In contrast, Type I grains were rare (<10%) in the survey of diamicts from the AND-2A core. Upturned plates were also rare in tills from the Transantarctic Mountains (Mahaney et al., 1996).
Only a low abundance of grains with mechanical impact textures are present in the upper Pliocene samples in Cores 7H, 8H and 9H. The upper Pliocene continental rise samples contain common grains of the glacial grain types B, C and D, and, as a result plot in the MDS near the glaciomarine stratified diamictites from the continental shelf ( Figure 7B). It is possible that glacial transport followed by minor current or wave impact on some grains produced similar grain types in both the continental shelf and deep-water glaciomarine settings. These comparisons suggest that the sand fractions retrieved from the upper Pliocene interval represent IBRD from ice sheets entraining material similar to shelf diamictites prior to calving.
On the other hand, samples from lower Pliocene diatomrich silty clays with dispersed gravel in Cores U1359B-13H and U1359A-15H are distinct in their microtexture distributions as is also evident from the position of these samples on the MDS plot (Figure 7). Chemically rounded grains (Types E and G), and grains with saw-tooth fractures and upturned plates (Type I) represent between 30% and 70% of grains in these samples, whereas the percentages for these grain types combined are <20% in the diamictites. Silt and fine sand laminae (example in Figure 3E) are present in the lower Pliocene interval of Site U1359 ( Figure 1B). Smear slides show that some laminae also contain diatom debris and are not completely clastic (Expedition 318 Scientists, 2011). Given the lack of evidence for subaqueous bedload transport in the microtextures, along with the reduced terrigenous, and the elevated diatom-component in the sediment, only winnowing or suspension transport as part of density flows could have been responsible for the development of the laminae. Furthermore, Hansen et al. (2015) pointed out that, locally, sand and gravelsized coarse fraction is dispersed within the graded laminated mud facies in this interval, which was explained by a combination of iceberg rafting and sediment lofting from hyperpycnal flows. Since microtextures that would support bedload transport by bottom currents, such as v-shaped impact pits, are lacking, it is assumed that sediment was deposited after transport in suspension or via gravity settling only. Chemical smoothing and rounding and upturned plates are generally attributed to in-situ weathering, cryoturbation, and eolian transport under a periglacial setting, which must have taken place prior to transport to the deep sea (Moss & Green, 1975;Wellendorf & Krinsley, 1980;Woronko & Hoch, 2011).
Microtextures consistent with preweathering and edge rounding induced by subaerial surface processes were also found by Cowan et al. (2008) on quartz grains from Pliocene continental rise sediments off the Antarctic Peninsula. There, the grains that lacked features of glacial transport were interpreted as supraglacial debris supplied by iceberg calving from a thinner Pliocene ice sheet, with greater exposure of glacial valley walls. Supraglacial debris is not well-documented on top of the outlet glaciers and ice streams in the Ross Sea.

| 577
PASSCHIER Et Al. However, isolated drapes of sorted eolian sand have been described from Wright Glacier in the Dry Valleys and other areas (Hambrey & Glasser, 2012). In the Ross Sea, eolian sediment with a prominent fine sand mode is also observed to collect onto sea ice near shore, and it is derived from the McMurdo Ice Shelf (Atkins & Dunbar, 2009). For a warmer climate setting in the Antarctic Peninsula region, Gilbert and Domack (2003) explain the origin of eolian sediment on the sea floor through the deposition of wind-blown sediment in melt ponds on top of ice shelves and release of eolian sediment at the onset of ice shelf break-up as the melt ponds drain. Therefore, for sediment sourced from the Ross Sea embayment during the warm Pliocene eolian sediment transport followed by either iceberg or sea-ice rafting can be regarded as a viable transport mechanism for portions of the sand fraction found offshore.

| Eolian sediment supply
The quartz component of modern pelagic sediment deposited away from fluvial sources is typically of eolian origin (Hodel et al., 1988;Leinen et al., 1986). The largest proportions of wind-blown sediment occur offshore sourced from unvegetated regions, typically deserts, and areas of deposition reflect the dominant wind patterns. Today, Antarctica's periglacial surfaces have limited exposure due to extensive glaciation, which limits eolian sand supply. In the modern Ross Sea, eolian sediment concentrations on sea ice are observed to be low more than 15 km offshore although some wind-blown sand can be transported large distances (>100 km) over sea ice via saltation during high-wind events (Chewings et al., 2014).
The annual eolian sediment flux can be expected to be a function of sediment availability and wind regime and these boundary conditions were markedly different during the early Pliocene in this area of Antarctica. East Antarctic ice retreat exposed larger areas of East Antarctica within the Wilkes and Aurora Subglacial basins that were elevated above sea level due to glacio-isostatic rebound and differences in dynamic topography (Austermann et al., 2015;Cook et al., 2013;Dumitru et al., 2019) (Figure 1C). The bulk geochemistry of Pliocene sediments from IODP Site U1358 on the Wilkes Land continental shelf shows limited chemical weathering of the mudrocks (Orejola et al., 2014), implying that the Antarctic source areas were cold and arid despite the deglaciation. Compared to modern sea floor sediments, Pliocene dispersed sand on the Wilkes Land margin contains a large proportion of quartz (Cook et al., 2017) and quartz is particularly resistant to the extreme attrition of grains in eolian transport. Furthermore, atmospheric modelling shows that upon deglaciation in the Pliocene, Site U1359 was in the pathway for eolian transport from the exposed source terrains with seasonally averaged summer wind speeds in excess of 10 m/s extending across the ocean surface (Scherer et al., 2016). During high wind events, wind speeds were probably sufficient to entrain sand under dry conditions and transport it onto ice shelves or over a sea-ice surface via saltation (Chewings et al., 2014;Gilbert, 1990;Gilbert & Domack, 2003).
Modern conditions in the Ross Sea are unlikely to be representative of past interglacials with the exposure of emerging coasts as discussed above. Therefore, the modern interglacial setting of the glaciated Canadian shield is used as an additional partial analogue for the early Pliocene interglacials. Field experiments on a modern emerging coastline in eastern Canada show that, under dry conditions with unlimited supply, fine to medium sand is typically mobile at wind speeds of 10 m/s (Davidson-Arnott et al., 2008). At higher latitude, eolian sediment transport is evident in the Arctic Coastal Plain of Alaska, where large sand dunes have formed through thermokarst, wind erosion and transport of exposed Pleistocene marine sediments (Carter, 1981). Furthermore, North of Alaska, eolian sediments contribute a large proportion of the sand fraction on the continental shelf and half the grains in a sample of frazil ice in this region were found to have rounding and textures typical of eolian transport histories (Hodel et al., 1988). Microtextures on grains retrieved from modern Arctic sea ice floes also show common edge rounding, silica dissolution and precipitation, upturned plates and microlayering (described as mechanical and chemical layer separation), in addition to microtextures typical of glacial sediments, such as conchoidal and step-like fractures, striations, gouges and breakage blocks (Dunhill, 1998;St. John et al., 2015). These microtexture populations are also indicative of a combination of glacial, periglacial and eolian processes (Moss & Green, 1975;Wellendorf & Krinsley, 1980;Woronko & Hoch, 2011).
At IODP Site U1359, Hansen et al. (2015) interpreted a concentration of gravel (>2 mm) at ca 4.5 Ma as a pulse of IBRD due to calving, followed by glacial retreat. However, comparison of grain-size distributions and microtextures with ice-proximal diamictites in the source area, indicate that approximately half of the sand grains in this interval were supplied through eolian sediment transport. It is envisioned that sediment was entrained or transported onto ice shelves or sea ice near the coast, with wind-driven sea-ice drift (Chewings et al., 2014;Holland & Kwok, 2012), followed by iceberg and sea-ice rafting and deposition at Site U1359. Sediment may have been eroded and transported by glaciers initially, followed by exposure, entrainment and modification in subaerially exposed periglacial environments or while embedded in the ice shelves, icebergs, or sea ice. Even though this scenario could imply a different provenance for this microtextural group as compared to the IBRD from the Ross Sea, a different provenance alone cannot explain the pervasive alteration of grain surfaces observed via SEM.
Export to the sea floor on the continental rise may have occurred via settling through the water column upon tipping of icebergs releasing their supraglacial load, from seasonal sea ice via transport from the continental shelf through highdensity saline flows originating from a shelf polynya, or winddriven circulation. The exact mechanisms remain uncertain. Ice shelf and sea-ice palaeorecords are sparse for the early Pliocene and indicate some spatial and temporal variability in extent (Taylor-Silva & Riesselman, 2018;Whitehead et al., 2005). Nevertheless, regardless of the exact scenario, the early Pliocene increase in the offshore sediment supply in the silt and sand fraction may have been primarily governed by the eolian sediment fluxes and not rates of iceberg calving, even in the presence of concentrations of gravel-sized IBRD.

| Broader implications
The implications of this finding are twofold. First, these results confirm challenges in using different grain-sizes in the sand fraction as a proxy for glacial activity (Gilbert, 1990). The glacial sediment load in icebergs as characterised by the multimodal particle-size distributions of diamictites, can be expected to have a variable sand mode ( Figure 1A). Furthermore, with glacial retreat the contribution of periglacial sediment cannot be ignored in the deposition of hemipelagic sediment with IRD. Dunhill (1998) noted that the unique property of IBRD in the modern Arctic is the presence of gravel, which is absent from SIRD. Therefore, gravel abundances are probably the most reliable first-order proxy for iceberg flux in continental margin settings where multiple sediment fluxes operate concurrently, whereas particle-size distributions and microtexture studies illuminate the nature of the other sediment fluxes.
Second, the discovery of enhanced wind-driven nutrient fluxes to the surface ocean could alter discussions of the biological carbon pump and other carbon cycle perturbations originating in the Southern Ocean during periods of deglaciation (Caballero-Gill et al., 2019;Chewings et al., 2014). Southern Ocean productivity is severely nutrient limited with nutrient supply from sources outside Antarctica focused during glacials, not periods of ice retreat. This study highlights a different scenario for the past, during the Pliocene Climatic Optimum, when the Earth system operated under a different climate state (Westerhold et al., 2020).

| CONCLUSIONS
It has been demonstrated here using a combination of sedimentological techniques that Pliocene dispersed sand within deep-marine sediments off the Wilkes Land margin, Antarctica, in comparison to diamictites with the same source, can contain a substantial non-glacial component. In contrast to the upper Pliocene IBRD, the lower Pliocene sand fraction has a distinct character with a larger silt to fine sand mode and a proportion of grains that are chemically weathered or exhibit polygenetic upturned plates. These types of grains are found predominantly in periglacial environments where frostweathering and eolian sediment transport processes prevail (Margolis & Krinsley, 1971;Woronko & Hoch, 2011). The lower Pliocene IRD maxima partially represent enhanced eolian fine sand fluxes via sea ice or supraglacial debris in icebergs with a possible role for export through density flows, but not bedload transport. Separating IBRD, the supraglacial component, and SIRD by grain-size alone is difficult: the variable sand modes of ice-proximal diamictites show that IBRD pulses can be misinterpreted when gravel counts and the fine sand fraction are omitted from the proxy. Microtexture analyses of quartz grains, in addition to particle-size analysis, and gravel counts provides greater insight into the relative contributions of sediment fluxes in a glaciomarine environment, and enhances the accuracy of palaeoclimate reconstructions.