Environmental and Oceanographic Conditions at the Continental Margin of the Central Basin, Northwestern Ross Sea (Antarctica) Since the Last Glacial Maximum

The continental margin is a key area for studying the sedimentary processes related to the advance and retreat of the Ross Ice Shelf (Antarctica); nevertheless, much remains to be investigated. The aim of this study is to increase the knowledge of the last glacial/deglacial dynamics in the Central Basin slope–basin system using a multidisciplinary approach, including integrated sedimentological, micropaleontological and tephrochronological information. The analyses carried out on three box cores highlighted sedimentary sequences characterised by tree stratigraphic units. Collected sediments represent a time interval from 24 ka Before Present (BP) to the present time. Grain size clustering and data on the sortable silt component, together with diatom, silicoflagellate and foraminifera assemblages indicate the influence of the ice shelf calving zone (Unit 1, 24–17 ka BP), progressive receding due to Circumpolar Deep Water inflow (Unit 2, 17–10.2 ka BP) and (Unit 3, 10.2 ka BP–present) the establishment of seasonal sea ice with a strengthening of bottom currents. The dominant and persistent process is a sedimentation controlled by contour currents, which tend to modulate intensity in time and space. A primary volcanic ash layer dated back at around 22 ka BP is correlated with the explosive activity of Mount Rittmann.


Introduction
The high-latitude continental margins are key areas where the sedimentary processes are influenced by the advance and retreat of the ice shelf and by different bottom water current activity on the continental slope between the glacial and interglacial periods [1][2][3].
The Ross Sea continental margins record the variation of the Antarctic Ice Sheet (AIS) due to the presence of the largest ice shelf, the Ross Ice Shelf (RIS), which promotes the production of dense shelf water which is the precursor for Antarctic Bottom Waters (AABW) [4]. The RIS dynamics that occurred during the Pleistocene and Holocene have and J3 (ANTA99−cJ3) studied by [25]. The black line is the axis for the schematic model presented in Section 4.3. This study aims to increase the understanding of the last glacial and deglacial dynamics in the Central Basin slope-basin system, which is an ideal region for studying the oceanographic variations in relation to the RIS dynamics. We used three box cores located in different morphological settings and at different water depths, to highlight the variations in time and space of the environmental changes recorded by the studied sedimentary sequence, using an integrated analysis of sedimentological, micropaleontological and tephrochronological.
We investigated how the different micropaleontological indicators, together with other proxies, respond to the main environmental/climatic variations. This study allows us to provide the spatial and temporal evolution of the most important environmental and oceanographic phases.

Materials and Methods
The three-box cores, KI13−BC02, KI13−BC03, and KI13−BC04 (hereafter BC02, BC03 and BC04) were collected during the KOPRI ANA03B cruise onboard the IBRV ARAON in February 2013 in the framework of the scientific cooperation between Korea and Italy (K-PORT and PNRA/ROSSLOPE Projects) (Table 1, Figure 1 [24,25]). The box cores were split, described and x-rayed at the Department of Mathematics and Geosciences (Trieste). Sediment colours were based on the Munsell colour chart.

Chronology
The chronology of the studied box-cores is based on the thirteen Accelerator Mass Spectrometry (AMS) radiocarbon dates performed at the Poznań Radiocarbon Laboratory, Adam Mickiewicz University (Poland) and with the MICADAS [26] at the Alfred-Wegener-Institut Helmholtz-Zentrum für Polar-und Meeresforschung (AWI, Germany). Eleven AMS radiocarbon analyses were carried out using acid insoluble organic matter (AIOM) and two were carried out on N. pachyderma tests ( Table 2). Uncorrected and calibrated 14 C data are summarised in Table 2. Table 2. Accelerator Mass Spectrometry (AMS) 14 C ages with calibrated calendar ages ± 2σ (year) and applied local contamination offset and reservoir age correction (see text) of the studied core. The calendar ages were calibrated using CALIB REV 7.1 software [27] with a Marine 13 calibration curve [28]. A constant reservoir correction of 1144 ± 121 years with a ∆R of 791 ± 121 years [29] was applied. Taking into consideration the different carbonate and organic carbon matrices, we decided to directly calibrate the carbonate dates and to treat AIOM dates prior to the calibration in 14 C ages. Radiocarbon chronologies using the AIOM fraction from bulk sediments are often compromised by contamination from the reworked old organic carbon derived from glacial erosion and/or from the reworking of unconsolidated sediments (see [30,31] for the discussion). This is evidenced by AIOM dates of Antarctic surface sediments that extend over several thousand years [32,33]. The AIOM dates of the surface levels (0-1 cm) of BC02 and BC04 have been used to correct the respective ages in depth for each box-core. These surficial dates embed the regional marine reservoir effect (MRE) and the local dead carbon contamination offset (LCO), assuming that both MRE and LCO did not change during the Holocene ( [29,30] for details). We also assumed that the age difference between the AIOM 14 C age of the core-top of two box cores and the Marine Reservoir Effect (MRE) represents the local contamination with older organic matter in this area, and hereafter it is referred to as LCO (see discussion in [34]). Before calibrating the 14 C dates, the LCO obtained for this area was subtracted from the AIOM 14 C down box-cores ages (LCO corrected ages), assuming that both MRE and LCO did not change over the Holocene ( [29,30] for details). We used a MRE of 1.1 ± 0.12 ka 14 C suggested by [29] for the Ross Sea carbonate samples as the box-core BC03 was only dated in two subsurface levels using carbonates (Table 2), the surface level is assumed to be of modern age, similar to the other two box cores.

Box
The LCO corrected AMS 14 C dates performed on AIOM and those performed on N. pachyderma tests were converted into calibrated ages by means of the CALIB REV 7.1 calibration program [27], at 95% confidence ranges. The MARINE 13 calibration curve [28], assuming regional marine offset (∆R) 0.79 ± 0.12 ka from the global MRE [29], was used. The new calibration curve (MARINE 20, [35]) was not used since it is not suitable for the calibration of Antarctic sediments, as suggested by [36].

Grain Size Analyses
Grain sizes analyses were performed on 1-cm thick subsamples at 1-2-cm intervals. The fraction >1 mm was separated by sieving to remove the larger grains. Grains were counted as belonging to the same sediment aliquot, and were thus used as an indicator of the coarse detrital component as a whole. The grain size characterisation of the <1 mm fraction was carried out using a Malvern Mastersizer Hydro2000S diffraction laser unit. Sand, silt, and clay fractions were determined using the grain-size classification proposed by [37]. Statistical parameters (mean size Mz, sorting σi, skewness Sk, kurtosis Kg) were determined according to [38]. We did not remove the biogenic silica fraction due to the abundance of siliciclastic particles highlighted during smear slide analyses as suggested by [39]. Moreover, the presence of non-biogenic amorphous silica resulting from volcanic glass or as an output from glacial abrasion should be considered part of the sediment archive. The parameters sortable silt mean size (mean SS) and percentage (SS%) were calculated from the grain size dataset according to [40]. Half-PHI (Φ) grain size data were used for clustering through the non-hierarchical K-means method. The number of iterations was stopped at n = 6, e.g., when achieving a number of grain-size end-members (=clusters) representative of the regular fining/coarsening of the grain-size spectra or the uni-or bi-pluri-modal feature of the distribution. Elaboration was carried out using the Past3 open-source program (version 3.18, [41]).

Cryptotephra Analyses
A cryptotephra, i.e., a layer of volcanic ash not visible at the naked eye, was identified during the micropaleontological investigation at 38-39 cm of sediment depth in BC04. The bulk sediment sample was treated using H 2 O 2 with 40% concentration to dissolve organic matter, washed with deionized water in an ultrasonic bath and dried at 60 • C. The treated sample was then impregnated in epoxy resin, polished and prepared for textural and geochemical (major and minor elements) analyses. The textures and composition of glass particles in the BC04 38-39 sample were studied using a scanning electron microscope (SEM), Zeiss EVO MA coupled with an Oxford-Aztec Energy EDS Analysis System. Standards of volcanic glass (VG2) were analysed for comparison. Cryptotephra analyses were performed at the Institution of Geophysics and Volcanology (INGV) Pisa Lab.

Diatoms
A total of 64 samples were performed according to the methodology described in [42]. A diatom count was performed under a microscope at a magnification of 1000×. For each sample, at least 300 diatom valves were counted for each slide following the counting rules proposed by [43] and taxonomical identification is based on [44]. When the diatom concentration was too low, five hundred fields of view per slide were analysed. The total absolute diatom abundance (ADA) in terms of the number of frustules per gram of dry weight (v/gdw), was calculated using the formula described by [45]. The relative abundance (expressed in %) in each sample was also calculated as the ratio between the diatom species and the total diatom abundance. Relative percentage of each diatom species was estimated as a percentage of the Chaetoceros resting spore free diatom assemblage (e.g., [46]). Species occurring at <2% were considered as statistically not significant [47].

Silicoflagellates
Silicoflagellates were counted from 69 samples on the same slides prepared for diatoms. Counts were performed at 200× magnification with an Olympus BX50 light microscope. Up to 60 specimens of Stephanocha speculum were counted in each sample, to calculate the Absolute Silicoflagellate Abundance (ASA) in terms of silicoflagellates per gram of dry weight (s/gdw) following the same formula used for diatoms. Depending on silicoflagellate density, a slide surface of 8-250 mm 2 was scanned, resulting in a detectability limit of 0.5-1 × 10 4 s/gdw. Each S. speculum specimen was observed at 1000× with immersion oil for morphotype identification, following the taxonomic concepts of [48] and recent review of [49]. Morphotypes were grouped according to their ecological preferences [50] and percent values were calculated only when silicoflagellate density was statistically significant.

Foraminifera
A total of 28 samples were collected to study the planktic and benthic foraminiferal assemblage. The samples were washed through a 63-µm sieve; foraminifera were studied in the sediment fraction >63 µm, using a binocular microscope. When the foraminifera were abundant, the samples were subdivided using a dry splitter until an aliquot containing approximately 300 specimens was obtained. The species counts were carried out only on well-preserved specimens and recorded as the number of specimens of each taxon; these data were subsequently converted into frequency expressed as a percentage and as densities (the number of specimens/g of total sediment). The number of broken and/or etched foraminifera allowed us to calculate the % of fragmentation for each level. All foraminifera were identified at species levels, except for Lagena and Oolina which were identified at the generic level. The identification mainly follows [51][52][53][54]. The online catalogue [55] was used to consult original taxa descriptions.

Age Model
The chronological reconstruction based on the calibrated ages corresponds to time intervals spanning from 24 ka BP to the late Holocene. The age-depth model was calculated using a linear interpolation function between dated levels ( Figure 3). Sedimentation rates are variable among the box cores ( Figure 3). As a general rule, the box cores show the highest sedimentation rates (from 3.62 to 5.99 cm/ky, without considering a temporary drop in BC 04) for the levels approximately older than 15 ka BP, whereas the lowest sedimentation rates (from 0.74 to 1.70 cm/ky) characterise the entire Holocene.

Sediment Grain Size
The raw grain-size data are reported in Table S1 and the grain size box core detailed description can be found in Supplementary S1.
The statistical approach allows for the grouping of samples in six clusters (Table S1) with a marked similarity of two pairs of clusters (2 and 3, as well as 4 and 5). Mean grain size distributions are represented in Figure 4 and the main statistical parameters of each cluster are summarised in Table 3. The continuum given by the sequence of clusters from one to five represents the progressive increase in sand and sortable silt (sensu Mc Cave et al. [40]), both in terms of proportion to the entire <63 µm fraction (SS%) and the mean SS size. Accordingly, the central tendency parameters (mean and median) also increase from cluster 1 to 5, representative of a clear shifting of the primary mode of the grain size distribution of each cluster (see Figure 4), from 10 µm (cluster 1), 20 µm (cluster 2), 30 µm (cluster 3), 35 µm (cluster 4), to 40 µm (cluster 5). Cluster 6 represents a separate group of samples with the coarsest mode (55 µm) and a poorly sorted distribution, markedly asymmetrical. Table 3. Main grain size data and statistical parameters for each grain size cluster. Mz, σ i , Sk and Kg are the mean size, sorting, skewness and kurtosis according to [38], C is the 1st percentile, M is the median, SS is the sortable silt, calculated according to [40]. In order to extrapolate information on selective processes responsible for the different grain size patterns, we tested the sortable silt signal through the classical SS% vs. mean SS relationship ( [39] and references therein). Data are grouped according to the extracted clusters and reported as a whole dataset and in separate relationships, one for each core ( Figure 5). Linear correlations are significant in all cases; data are very highly correlated both for the whole dataset (r = 0.89) and for each core, corresponding to r = 0.93 for BC02, r = 0.89 for BC03 and r = 0.88 for BC04. The grain size trend along the three box cores is represented by the vertical distribution of the clusters, ( Figure 6). In the same graph the number of dropstones (fraction >1 mm) recovered in the fixed amount of sediment used for the grain size analysis is used as a proxy of the vertical flux linked to the Ice Rafted Debris (IRD). BC02 and BC03 show a similar fining and subsequent coarsening trend downcore, although not represented by the same cluster sequence. BC04 is texturally quite homogenous, after a coarser 9-cm thick topmost layer (cluster 5 and 4). BC02 and BC04 feature a significant amount of IRD, either at the base of both cores and at the already defined topmost layer of the latter core.

Diatom Assemblage
The diatoms are well preserved and do not show dissolution features represented by the loss of valve margin, the breakage of the middle part, distorted or/and conjoined pores on diatom valves, and the breakage of areolae, etc. This is also confirmed, in agreement with [43], by the higher presence of Fragilariopsis kerguelensis then Thalassiosira lentiginosa. The ecological significance of different taxa can be found in Supplementary S2.
Total absolute diatom abundance (ADA) in BC02 varies between 5.5 and 120.6 × 10 6 valves per gram of dry sediment (v/gds), in BC03 from 9.7 to 199.1 × 10 6 (v/gds) and in BC04 from 4.2 to 218.4 × 10 6 (v/gdw). The diatom record shows low abundance before to 31 cm in BC02, 29 cm in BC03, and 21 cm in BC04 (ca. 17 ka BP), followed by a rapid increase and then at about 14 cm in BC02, 10 cm in BC03 and 10 cm in BC04 (12 ka BP for BC02 and BC04; 9 ka BP for BC03) ka BP ADA decreases towards the core-top.
In BC02 (Figure 8), F. kerguelensis (avg. 22%) has a positive trend from the core-bottom of the core to 30 cm and then it becomes the dominant species from 30 cm to 8 cm (17 ka BP to 6.5 ka BP), and then it decreases towards the core-top. Relative abundances of E. antarctica (avg. 17%) present the downcore opposite trend to the one of F. kerguelensis over the entire record. Relative abundances of E. antarctica are the highest between the core-bottom and 30 cm (18-17 ka BP) and from 12 cm (10.2 ka BP) to the core-top of the core. Fragilariopsis obliquecostata (avg. 18%) presents a higher value until 27 cm (16.5 ka BP), then decreases until 11 cm (9 ka BP). From 11 cm to 2 cm (9 ka BP to 1 ka BP) F. obliquecostata slightly increases and subsequently decreases. Fragilariopsis curta (avg. 13%) increases from the core-bottom to 19 cm (18-15 ka BP), and then decreases until 12 cm (10.2 ka BP), and it increases again from 12 cm to the core-top (10.2 ka BP-present). The "warm water group", represented by the sum of Thalassiosira lentiginosa, T. oliverana, and T. gracilis (avg. 4%), presents a very low relative abundance, increasing towards the core-top. Actinocyclus actinochilius (avg. 4%) presents the highest value from the core-bottom to 30 cm (18-17 ka BP). Paralia sulcata (avg. 2%) increases from 12 cm to the core-top (10 ka BP to present) with the highest value at the top.
Within S. speculum, the most abundant varieties are var. speculum, var. monospicata, and var. coronata, grouped together as the open ocean varieties. Among these varieties, rare forms with pentagonal/heptagonal outline or with a large apical ring are included. In BC02 (Figure 7), the cumulative abundance of the open ocean varieties is above 50% between 27 and 11 cm (16.5 and 9 ka BP), then drops at 9-8 cm (7.5-6.5 ka BP), increases again to~50% at 7-6 cm (5.6-4.6 ka BP), and then decreases to values below 21% in the upper 4 cm (latest 2.8 ka BP); in BC03 ( Figure 8) it is above 60% from 32 to 5 cm (17 to 2.3 ka BP), then decreases to lower values in the upper 4 cm (latest 1.7 ka BP); in BC04 ( Figure 9), it is above 50% at 20 cm (16.4 ka BP), then decreases and increases again from 17 cm to the core-top (14.7 ka BP-present), remaining high up to the core-top.
S. speculum var. coronata shows in contrast higher relative abundance in the bottom and top samples: in BC02, it displays high relative abundance at 31 cm (17.2 ka BP) then decreases, while in BC04, it increases to peak values at 19 cm (15.8 ka BP), then decreases. A slight increase in its percent values is observed in both BC02 and BC04 at 11 cm for BC02 and 8 cm for BC04 (9.5 ka BP) and a further increase by 6.5 ka BP, with higher percentage values in BC02 as compared to BC04. S. speculum var. coronata also includes morphologies with a large apical ring, typically indicated as var. minuta [61,62]. The latter constitutes up to 10% in BC03 and BC04 and up to 20-38% in the top samples of BC02. Finally, other varieties are represented by forms with an incomplete crown of apical spines and by aberrant morphologies, i.e., forms with an unclosed basal ring, doubled apical ring or forked spines: all these are usually a minor component of the S. speculum assemblage.

Foraminifera
Planktic and benthic foraminifera are present in BC03 and BC04; on the contrary, they are represented only by rare benthic agglutinated and fragmented taxa. Ecological significance of different taxa can be found in Supplementary S2.
Neogloboquadrina pachyderma (adults and juvenile tests) is the only planktic species recorded in the investigated sediments. In BC03, N. pachyderma varies from 1 to 818 specimens/g and is abundant from 36 to 28 cm (18-16 ka BP), reaching its maximum value at 31 cm (16.7 ka BP) ( Figure 11).
In this interval (18-16 ka BP), adult and juvenile specimens are recorded with a comparable abundance. In the sediments above 24 cm (younger than 14.1 ka BP), N. pachyderma is very rare or absent and when present is only represented by adult forms. In BC04, N. pachyderma is present from 35 to 25 cm (21.5-19.2 ka BP) with a density ranging from 26 to 206 specimens/g ( Figure 12). It reaches the maximum presence at 31 cm (20.6 ka BP). Otherwise, it is nearly absent. In this box-core, only adult specimens are present. Tests of N. pachyderma are well-preserved and do not show dissolution or fragmentation.

Sedimentary Dynamics
As previously observed by Hoffmann et al. [63], an important concern in interpreting sedimentary records is "whether the sortable silt reflects only current-driven sorting, or whether other vectors of sediment supply can bias the record". As observed in the box cores, IRD is a relevant sedimentary input and may influence grain size due to the delivery of unsorted material. Figure 4 demonstrates a very good fit of the samples from the Central Basin to the typical linear relationship linked to the selection of the SS according to McCave et al. [39]. Cluster 6 represents sediments mostly affected by the influence of the vertical flux due to IRD as found in BC02 and partially in BC03: these samples are those responsible for the bias in the relationship reported in Figure 4. Excluding cluster 6, the correlation coefficient improves from r = 0.89 to r = 0.93.
The diagram also suggests the goodness of the clustering procedure, since the coarsening (or fining) sequence follows the linear trend, thus inferring the increasing (or reducing) effect of selectivity of the local currents in the sedimentary process affecting the Central Basin.
Given the above considerations, we can deduce that the dominant process is a sedimentation controlled by currents, which tend to modulate intensity in time and space, as seen by the grain size trend (stepwise fining or coarsening) according to different SS modes. The grain sizes conserve a marked modal peak (primary SS subpopulation) which can shift in the order of 5-10 µm from 10 µm to 55 µm. In terms of mean SS each shift of clusters (excluding cluster 6) implies a progressive increase of 1.6 to 2.4 µm. If we consider the sensitivity (relationship between flow speed and mean SS) proposed by McCave et al. [64], the sedimentological changes linked to each cluster step are due to increase in the current speed between 2.2 and 3.3 cm s −1 . Thus, the fining or coarsening grain size trend, which involves the current-related grain size clusters, implies an overall flow speed variation of ca 7.5 cm s −1 .
The mostly regular presence of SS, the lack of sedimentary structures within the sampled thickness, accompanied by low and variable sedimentary rates (between 0.74 and 4.33 cm ky −1 ) not related to changes in grain size, do exclude turbidity currents as well as intermittent sediment supplies.
The sedimentary records seem more likely to be associated to a continuous flux of glacigenic sediments, rich in silt, delivered by the Ross Ice shelf [65], maintained in suspension by currents up to the shelf break and further moved towards the Central Basin, where they are captured by the westerly contour-current. The specific physiography of the Central Basin is likely to cause a local gyre, as inferred by the pattern of currents resulting from the longline fishing analysis carried out by Richard et al. [66]. Despite the limitation of the method, longline data suggest a complex movement of the water masses directly around and west of Iselin Bank (Figure 1), due to the control given by the topographic highs. These are likely to induce a circular current pattern inside the Central Basin responsible for the partitioning of the sediments according to the specific places, more or less exposed to the velocity gradient. The above hypothesis could explain the difference in the stratigraphic signal, as well as the different current intensities among the different cores, as recorded by the SS proxy.
The contourite pattern is altered by the vertical flux linked to the IRD, characterising the entire stratigraphic record. Although icebergs and sea ice are currently sources of IRD to the slope sediments, the presence of IRD in the sediment does not affect the typical grain size signal due to the current selection. The similar vertical pattern of grain size sorting in BC02 and BC03 is an interesting aspect that we can only discuss as a mere hypothesis. These two box cores seem to be correlated according to a typical downcurrent fining, considering BC03 as the more exposed site and BC02 as a basin sink. This implies that the current would flow from BC03 to BC02, thus inverting the original westerly direction while decreasing its velocity. We are unable to confirm this pathway, but future investigations may or may not confirm this hypothesis.

Cryptotephra: Correlation with Volcanic Source
The Glass composition of BC04 38-39 cryptotephra plot in the compositional field of phonolite to trachyte and overlap the compositions of the products of Mount Rittmann [57,58,60,67] (Figure 7c,d). These can be clearly distinguished from the pyroclastic products of the other volcanic complexes of North Victoria Land namely Mount Melbourne and the Pleiades volcanoes or from the products of the Mt Erebus volcano, which is located further south than the Central Basin. BC04 38-39 cryptotephra shares a geochemical affinity with several marine tephra and cryptotephra layers recently found interbedded in the Holocene-late Pleistocene sediments of the Ross Sea and in ice records [57,60,61]. The glass compositions of BC04 cryptotephra match those of the 1254 C.E. (or 1252 ± 2) tephra from Mount Rittmann widespread in the marine sediments from Cape Hallett [31,57] and in the ice records of the entire West Antarctica [58,[68][69][70][71]. BC04 38-39 cryptotephra compositions also match very well with the composition of 11.1 ± 1.4 ka Aviator tephra (AVT tephra) found in the marine sediments of the Aviator Basin ( [60], and again derived from the Mount Rittmann. Finally, except for slight differences in SiO 2 and Al 2 O 3 content, BC04 38-39 cryptotephra shares a broad geochemical affinity, with two marine tephra layers found in the western Ross Sea constrained by 40 Ar- 39 Ar ages at 23.7 ± 5.3 ka BP (ANTA02-NW2-0-13) and 21.2 ± 6.4 ka BP (ANTA02-NW31-42-61), respectively [61], and again correlated to the explosive activity of Mount Rittmann [58]. Unfortunately, it was not possible to directly date BC04 38-39 marine cryptotephra using the radiometric method due to the extremely fine grain size of the particles and the very low abundance of K-feldspar crystals directly derived from the tephra. However, as shown in Figure 2 the age of 22.0 ka BP was obtained by AMS dating of AIOM in BC04 at 37 cm, on top of the cryptotephra. Thus, we can attribute to BC04 38-39 cryptotephra a minimum age slightly older than 22 ka BP. This age is in good agreement with those determined for the two marine tephra found interbedded in the marine sediments of the Drygalski Basin [60], dated by the 40 Ar- 39 Ar method between 23.7 ± 5.3 ka (ANTA02-NW2-0-13) and 21.2 ± 6.4 ka BP (ANTA02-NW31-42-61) mentioned above.
In addition to slight differences in the major element composition, we suggest that BC04 38-39 cryptotephra and ANTA02-NW2-0-13 and ANTA02-NW31-42-61 marine tephra layers could be derived from the same eruption and represent the very distal and medialdistal deposits, respectively, of a highly energetic explosive eruption of Mount Rittmann which occurred at ca. 22 ka BP.

Paleoenvironment Interpretation
Radiocarbon datings are consistent with their stratigraphic position and therefore the effect of sedimentary reworking is negligible.
On the basis of the age model result, the studied sediments recorded the last 24 ka, which correspond to a part of the Last Glacial Maximum (LGM) (26.5-19 ka BP, sensu Clark et al. [72]), the subsequent deglaciation, and the onset of the Holocene. This offers us the possibility to discuss the main phases occurring during this period which are still under debate in terms of its timing, style, and forcing factors (e.g., [9]).
BC04 sampled the oldest sedimentary sequence and allowed for an inference of the environmental and oceanographic condition during the LGM, while BC02 and BC03 record the more recent period and allow for a better understanding of the environmental change during the early stage of the deglaciation and the following Holocene.
The downcore variation in the relative abundance of diatom, silicoflagellate and foraminifera contents and the grain size and tephra data allows for the recognition of three different sedimentary units: Unit 1 (24 ka BP to 17 ka BP), Unit 2 (17 ka BP to 10.2 ka BP), and Unit 3 (10.2 ka BP to present). In Figure 13 we present a schematic sketch of processes occurring in the three sedimentary units.  (Figure 1) [5,9,[73][74][75]. Even if the ice sheet expansion is well documented, the ice shelf cover and its extension still remain poorly constrained. The authors of [9] proposed a model of ice shelf extension starting from 20 ka BP and suggested that the ice shelf covered until the outer sector of the JOIDES basin but the Central Basin has not yet been examined.
In the studied area, Unit 1 (24 ka BP to 17 ka BP) is well represented in BC04 while in BC02 and BC03 only the latest portion of this unit was sampled. This unit is characterised by silty sediments for the most belonging to cluster 3 (BC04), thus suggesting a dominant contour current able to control the sedimentary processes. The unit features a high gravel content and abundant sand derived from IRD supply. The vertical flux locally tends to alter the current-controlled signal, as highlighted by the presence of cluster 6 in the topmost part of the unit (18-17 ka BP) in BC02 and BC03.
Low ADA and absence of silicoflagellates in this interval are a consequence of high terrigenous supply due to strong glacial influence. In this period, the high relative percentage of E. antarctica which indicates iron enrichment due to ice melting (see Supplementary S2) indicates an environment dominated by iceberg transit and iceberg melting.
The relatively high percentage of F. obliquecostata and A. actinochilus among diatoms and S. speculum var. coronata (only detected at 17.2 ka BP in BC02) among silicoflagellates indicate the presence of seasonal sea ice (see Supplementary S2). In BC04, the co-presence of cold and sea ice indicators with T. antarctica indicates its vicinity to the ice shelf edge [75] and suggests the proximity to the calving zone [76][77][78], as well as the ice shelf destabilisation.
Starting from 21 ka BP, the occurrence of calcareous foraminifera is recorded at the site (BC04) closer to the JOIDES Basin. They are characterised by the planktic N. pachyderma (only adult tests) and scattered specimens of benthic taxa (Figure 12). Even if the density of N. pachyderma is significantly high at 20.5 ka BP (Figure 12), the contemporary presence of numerous fragmented tests (mean fragmentation 51%), led us to consider these foraminifera as transported/reworked. This consideration is in agreement with the influence of bottom currents seen in this period.
A further recovery of calcareous-sediment-rich intervals is also recorded at the site of Hallett Ridge site (BC03), from 18.1 to 15.9 ka BP. In contrast to BC04, the foraminifera here are generally well preserved. They are represented by the planktic N. pachyderma, including both adult and juvenile forms. Neogloboquadrina pachyderma is able to live in polar oceans, need glacial-free conditions to better proliferate and to reach the reproduction phase [79]. The co-presence of both growth stages of N. pachyderma can be the effect of accumulation during different seasons [78]. Similarly, we interpret that their occurrence is indicative of a period characterised by the beginning of open water with seasonal sea ice conditions. At the same time, the benthic foraminifera are abundant and are mainly represented by E. exigua, A. weddellensis, Nonionella spp., and subordinately by C. lobatulus, G. biora, G. subglobosa and T. earlandi. This benthic foraminifera association is normally used to represent sub-ice shelf environments, from proximal to distal, although the presence of A. weddellensis and E. exigua suggests a more open marine environment [80]. These two small species could also indicate the occasional influence of CDW for this area, as suggested by [80,81].
Similar results of carbonate-rich intervals have been reported either by [82] in the eastern Ross Sea continental margin in a time period spanning from 28.2 to 17.2 ka BP (as uncorrected ages) or by [15] and by [78] in the western Ross Sea continental margin, at around 20 ka BP. Authors of [82] suggested that this carbonate layer was deposited under a destabilisation of the Ross ice shelf-sea ice system likely caused by meltwater pulse (MWP) at 19 ka BP. Authors of [15] suggested that the deposition of carbonate (mainly planktic foraminifera) layer occurred during the glacial period due to the influence of grounded ice sheets on the continental shelf that inhibited the formation of corrosive shelf water and/or due to intense bottom current that allowed preservation by restricting fine material deposition. In our studied area the presence of carbonate is recorded only in those sites (BC03 and BC04) located on two bathymetric highs, whereas carbonates are absent in BC02 due to the greater depth.
To summarise, the co-presence of benthic foraminifera and diatom cold water indicator F. obliquecostata and T. antarctica, in agreement with the abundance of N. pachyderma, F. curta and A. actinochilus, suggests an environment near the ice shelf calving zone characterised by the presence of IRD and the development of a seasonal sea ice condition, and also confirmed by the presence of the cryptotephra layer in BC04 at 22 ka BP.
The same environmental condition is also reported as a result of the analysis of core J3 collected in the outer section of the JOIDES Basin [25], where the granulated facies representing a lift-off zone or the presence of thin water film [73] is coeval to our unit. The sediments deposited in the Central Basin were therefore likely deposited under strong glacial influence and in the vicinity of the ice shelf calving zone (sensu [77]), as also proposed by [11,15] at nearby locations.
The schematic depositional environment during unit 1 is shown in Figure 13a The Early Deglacial in Antarctica is reported starting from~22 to 21 ka BP in the West Antarctic Ice Sheet divide Ice core (WDC) (West Antarctic Ice Sheet (WAIS) Divide Project members, 2013 [82]). In contrast different studies based on model and ice core data (Epica Dome C) have hypothesised that the atmospheric warming that drove the deglaciation, following the LGM maximum ice extent, began at 18 ka BP [83][84][85][86][87][88]. The timing of the ice sheet and consequently ice shelf retreat did not start at the same time in the Antarctic region [5,[88][89][90][91]. Regarding the western sector of the Ross Sea, the maximum extension of glaciers reached the mid-outer shelf of the JOIDES Basin during the LGM [5,73] and the open water condition began to develop at about 14 ka BP [9].
In our sediments the first evidence of decreasing glacial influence is demonstrated by the inflow of warm waters. This occurs at~17 ka BP in the two offshore sites (BC02 and BC03), and then in BC04, in which the inflow of warm water occurred from~16 ka BP.
In this unit we detected the maximum values of ADA and ASA as previously reported in studies performed on cores from the northwest Ross Sea cores [11,15,25,78,92]. The increase in siliceous microfossil concentration testifies to an increase in primary productivity, also indicated by the presence of the phytodetritus benthic foraminifera E. exigua [93]. The change in productivity is caused by the continuous inflow of oceanic warm water and less terrigenous supply due to ice shelf retreat and thus supports the hypothesis of ameliorate climate conditions. This micropaleontological content matches with siltier sediments and the prevalence of cluster 2 (BC03 and BC04) and cluster 1 (BC02) representative of very slow currents, with the exception of the timing interval 16-15 ka BP in BC03 where cluster 3, representative of a more intense current, prevails in agreement with the peak of T. earlandi. The beginning of the deglaciation phase, recorded in the three box cores, is marked by the replacement of the cold and sea ice diatoms by the open water taxa, F. kerguelensis among diatoms and, the open ocean varieties of S. speculum among silicoflagellates (see Supplementary S2). In the offshore sites (BC02 and BC03), the increase in open ocean diatom and silicoflagellates (see Supplementary S2) reached at 16.5 ka BP and corresponds to a drop in ice-related species. In contrast, in the site (BC04) closer to the continental shelf, the increase in open ocean species is interrupted at 16 ka BP by an increase in ice-related species (F. obliquecostata, A. actinochilus and S. speculum var. coronata) and the percent maxima of F. kerguelensis reached at 14.5-14 ka BP. In the offshore site BC03, the foraminifera strongly decreased from 16 ka BP. Though considerably diminished in absolute abundance, they are still represented by the significant presence of M. earlandi, which becomes the only foraminifera present from 11.2 ka BP and above.
This agglutinated taxon has been documented in the diatomaceous sediments of the Ross Sea troughs by [54,80,94] (as M. arenacea) during the Late Pleistocene-Holocene transition. The exclusive presence M. earlandi and the disappearance of the calcareous foraminifera are normally related to the lower carbonate ion concentration in the deep water, presumably under the influence of the HSSW [14,80].
As pointed out by [30], high abundances of F. kerguelensis, are probably due to the intrusion of the warm and nutrient-rich CDW. The correlation with two cores (ANTA99-cJ5 and ANTA99-cJ3 [25] from the JOIDES basin allow us to identify the same F. kerguelensis abundances at 13.5 ka BP. This correlation highlighted that the CDW intrusion began from offshore to the continental shelf as well as the increasing distance from the ice shelf calving zone. Unit 2 thus represents the start of the Deglaciation characterised by the inflow of CDW into the Central Basin and the retreat of the ice shelf onto the JOIDES Basin. The CDW intrusion is first recorded at the outer Central Basin sites and then at the site near the continental shelf margin (Figure 13b).

Unit 3: From 10.2 ka BP to Present
The Holocene period in the Antarctic region is characterised by the AIS retreat in response to the warming ocean and an increase in temperature [87]. During the Holocene, the ice sheet rapidly retreated from the Ross Sea [5] and this led to the establishment of seasonal sea ice. In this unit the sediments show a coarsening trend, suggesting an increase in the contour current velocity.
The decrease in ADA and ASA content recorded in both BC02 and BC04 since 10.2 ka BP documents unfavourable conditions for phytoplankton bloom. The decreased abundance of CRS, especially in the offshore site, reflects the increasing distance of the ice shelf edge, as already observed during the late phase of deglaciation on the Antarctic Peninsula (Barbara et al., 2010), due to the decreased influence of nutrient-rich water from the melting ice shelf.
In the JOIDES Basin, the increase in ice-related species (F. curta and F. obliquecostata) is observed from 8.2 ka BP [30] and correlates with an increase in ssNa (sea salt Sodium) in ice cores (TALDICE and Taylor Dome) [30], indicating the establishment of a seasonal sea ice cycle and the sea ice extent due to the ice shelf retreats away from the outer sector of the JOIDES Basin. In this time period the JOIDES basin and the Ross Sea Embayment were fully opened as suggested by [95] and later by [30].
As already observed in the previous unit from the early Holocene onwards, calcareous foraminifera are nearly absent in Unit 3 as well and the only benthic species present is M. earlandi. The presence of this agglutinated taxon together with the percentage values of F. kerguelensis indicates the maintenance of open water conditions [54,80,[94][95][96] at least until 6.5 ka, even though a slight increase and oscillating abundance of ice-related and cold-water species (F. obliquecostata, E. antarctica, and S. speculum var. coronata) are observed, especially in the offshore site, suggesting a decreased inflow of CDW. A slight increase in P. sulcata from 10.2 ka BP indicates the onset of bottom transport from the shelf and could coincide with the renewed production of HSSW due to wind forcing and sea ice formation.
An increase in sea ice-related species (F. curta, A. actinochilus and S. speculum var. coronata) and cold-water-related species (F. obliquecostata) is observed from 6.5 ka BP and is associated with low ADA and ASA, low CRS, and a gradual decrease in F. kerguelensis. However, the co-occurrence of open ocean and sea ice related species, along with a slight increase in the warm water diatom group point to conditions of seasonal rather than perennial sea ice and a prolonged open water season.
The presence of P. sulcata from 4 ka BP, particularly marked in BC04, indicates strengthening and persistent transport from bottom waters also confirmed by the dominance of the grain size found in cluster 5.
The increase in sea-ice related diatom species together with S. speculum var. coronata is particularly significant in the last 2.5 ka and is higher in the offshore site suggesting colder water conditions. In the same interval, P. sulcata is particularly abundant reflecting a stronger influence of the HSSW, whose formation can be related to the establishments of the Ross Sea polynya. Furthermore, the last 2 ka BP is characterized the increase in sand and IRD content which may be related to the presence of icebergs, as evidenced by increased abundances of E. antarctica.
To summarise, integrated results suggest that Unit 3 represents the Holocene period characterised by the establishment of seasonal sea ice and intense bottom currents (Figure 13c).

Conclusions
Micropaleontological, sedimentological and cryptotephra analyses performed on three sediment box-cores allow for the reconstruction of the environmental and oceanographic changes during the LGM-deglacial-Holocene period in the Central Basin of the western Ross Sea.
Micropaleontological investigation allows for the recognition of three stratigraphic units. Unit 1 (from 24 to 17 ka BP) is characterised by a strong glacial influence due to the vicinity to ice shelf calving zone, Unit 2 (17-10.2 ka BP) represents the progressive receding of the ice shelf calving zone due to CDW inflow. Unit 3 (10.2 ka BP to present) is characterised by the establishment of seasonal sea ice and a strengthening of the bottom current. Moreover, we suggest that the abrupt increase in F. kerguelensis may be viewed as an indicator of the beginning of deglaciation and the presence of P. sulcata in the offshore area may be viewed as an indicator of the presence of bottom current.
Grain size data highlighted a dominant process of sedimentation controlled by currents, which tended to modulate intensity over time and space, as testified by the grain size central tendency marked by different sortable silt modes. The specific physiography of the Central Basin is likely to cause a local gyre, thus inducing a circular pattern responsible for the partitioning of the sediment according to the specific places, more or less exposed to the velocity gradient.
Finally, this investigation allowed for the identification of a cryptotephra correlated with the explosive activity of Mount Rittmann and dated at c. 22 ka BP. This further confirms the occurrence of highly energetic explosive activity at this volcano in the Late Pleistocene. In addition, this cryptotephra provides a new chronostratigraphic marker for the future dating, correlation and synchronisation of the geological archive in this area.  Data Availability Statement: Micropaleontological data are available on request from the corresponding author. These data are not publicly available as they are objects of PhD thesis. The other data are contained within the article.

Acknowledgments:
We thank the technicians, captain, and crew of IBRV Araon for their efforts in obtaining sediment samples and geophysical data during the ANA03B expedition. We would like to thank the Italian PNRA (Progetto Nazionale di Ricerca in Antartide) and KOPRI (Korean Polar Research Institute) for funding the campaign. This research used samples provided by the Sorting Center-Trieste Section of the Museo Nazionale dell'Antartide-and it is developed in the framework of the STREAM Project (Late Quaternary evolution of the ocean-ice sheet interactions: the record from the Ross Sea continental margin (Antarctica). EC and BK wish to express their appreciation to the Italian Ministry of Foreign Affairs and International Cooperation (MAECI PGR 00822 and 01060) and PNRA project (2010/A2.07), and the National Research Foundation of Korea (2019K1A3A1A25000116) and KOPRI project (21090), respectively. This is contribution number 2047 of the Marine Science Institute-CNR, Bologna. Many thanks to Karolyn Close for the linguistic revision. We are grateful to the three anonymous reviewers for their suggestions on how to improve the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.