Sedimentary record from the Canada Basin, Arctic Ocean: implications for late to middle Pleistocene glacial history

Sediment core ARC4-BN05 collected from the Canada Basin, Arctic Ocean, covers the late to middle Quaternary (Marine Isotope Stage – MIS – 1–15, ca. 0.5–0.6 Ma) as estimated by correlation to earlier proposed Arctic Ocean stratigraphies and AMS14C dating of the youngest sediments. Detailed examination of clay and bulk mineralogy along with grain size, content of Ca and Mn, and planktic foraminiferal numbers in core ARC4–BN05 provides important new information about sedimentary environments and provenance. We use increased contents of coarse debris as an indicator of glacier collapse events at the margins of the western Arctic Ocean, and identify the provenance of these events from mineralogical composition. Notably, peaks of dolomite debris, including large dropstones, track the Laurentide Ice Sheet (LIS) discharge events to the Arctic Ocean. Major LIS inputs occurred during the stratigraphic intervals estimated as MIS 3, intra-MIS 5 and 7 events, MIS 8, and MIS 10. Inputs from the East Siberian Ice Sheet (ESIS) are inferred from peaks of smectite, kaolinite, and chlorite associated with coarse sediment. Major ESIS sedimentary events occurred in the intervals estimated as MIS 4, MIS 6 and MIS 12. Differences in LIS vs. ESIS inputs can be explained by ice-sheet configurations at different sea levels, sediment delivery mechanisms (iceberg rafting, suspension plumes, and debris flows), and surface circulation. A longterm change in the pattern of sediment inputs, with an apparent step change near the estimated MIS 7–8 boundary (ca. 0.25 Ma), presumably indicates an overall glacial expansion at the western Arctic margins, especially in North America.


Introduction
The advances and decays of continental ice sheets play a significant role in the alteration of global climatic system, such as changing atmospheric circulations, creating largearea albedo anomalies and regulating the global sea level fluctuations (Clark et al., 1990).Reconstruction of the history of ice sheets is therefore important not only for a better understanding of feedbacks of the future climate change and its impact on regional climates but also for getting insights into the mechanisms of abrupt climate change.
Studies of Pleistocene glaciations around the Arctic Ocean dealt mostly with the late Quaternary history of the Eurasian Ice Sheet during Marine Isotope Stages (MIS) 1-6 (e.g., Svendsen et al.,2004;Larsen et al., 2006) or the Laurentide Ice Sheet (LIS) with a special attention to the Last Glacial Maximum (LGM) (e.g., Dyke et al., 2002;England et al., 2009).In addition to terrestrial data, studies of sediment cores from the Arctic Ocean are critical for comprehending the history of glacial advances and retreats (e.g., Polyak et al., 2004Polyak et al., , 2009;;Spielhagen et al., 2004;Stein et al., 2012;Kaparulina et al., 2016).However, the long-term history of circum-Arctic glaciations is still poorly understood, especially with respect to the western Arctic including the North America and East Siberia.A major impact of the North Beaufort Gyre (BG), Transpolar Drift (TPD) and Bering Strait (BS).Schematic geological map shows the distribution and prevailing lithology of the main terrains adjacent to the Arctic Ocean (Fagel et al., 2014).
American ice sheets on circulation and depositional environments in the Arctic Ocean is indicated by various marine and terrestrial data (e.g., Phillips and Grantz, 2001;Stokes et al., 2005), whereas the East Siberian Ice Sheet (ESIS) remained largely hypothetical until recently.While terrestrial data are limited and remain to be better investigated (Grosswald, 1989;Basilyan et al., 2010;Ivanova, 2012), seafloor mapping data now provide ample evidence for the existence of considerable ice masses on the East Siberian margin (Niessen et al., 2013;Dove et al., 2014;Jakobsson et al., 2014Jakobsson et al., , 2016)), but the timing and extent of these glaciations is virtually unknown.Marine sedimentary records from the Arctic Ocean adjacent to the East Siberian margin could add valuable information to this intriguing paleoglaciological problem.
In this paper, we present a multiproxy study of glacialinterglacial changes during the late to middle Pleistocene based on sediment core ARC4-BN05 from the Canada Basin north of the Chukchi Plateau and east of the Mendeleev Ridge (Fig. 1).This location can be affected by the two main Arctic Ocean circulation systems, the Beaufort Gyre and the Transpolar Drift, which carry sea ice, icebergs, and sediment discharge from North America and Siberia, respectively.As this circulation along with sedimentary environments and sources varied greatly during the Pleistocene climate cycles, resulting variations in sediment delivery and deposition make for a valuable paleoclimatic record for the western Arctic.Biogenic proxies (such as foraminifers) have uneven and overall limited distribution in Arctic Ocean sediments, while the terrigenous component provides a more consistent mate-rial for paleoceanographic studies (e.g.Stein, 2008;Polyak et al., 2009).As sediments in the Arctic Ocean are primarily transported by sea ice and/or icebergs during glacial events, sediment composition yields important information not only on the provenance and transport pathways but also on the attendant glacial and paleoclimatic history (e.g.Spielhagen et al., 1997;Vogt et al., 2001;Knies et al., 2001).By using clay and bulk mineralogy, along with grain size and the content of major elements Ca and Mn, we reconstruct depositional environments and sediment provenance to provide clues to the history of western Arctic ice sheets and their interaction with the Arctic Ocean.

Regional background
The Arctic Ocean is surrounded by land masses composed of an assortment of lithologies and situated in a variety of climatic, tectonic, and physiographic settings.Figure 1 depicts a schematic geological map showing the main terrains and associated lithologies (Fagel et al., 2014).The West Siberian Basin and East Siberian platform of the Eurasian continent are mainly composed of terrigenous sediment (Fagel et al., 2014).The Siberian (Putorana) traps constitute one of the largest flood basalts in the world (Sharma et al., 1992).The western Okhotsk-Chukotsk volcanic belt contains acidic to intermediate rocks, whereas intermediate to basic rocks are more characteristic of the eastern side (Viscosi-Shirley et al., 2003).The Kara Plate and the Taymyr foldbelt, as well as the Ural and Novaya Zemlya foldbelt, are mainly composed of intrusive and metamorphic rocks (Fagel et al., 2014).
The geology of outcropping terraines of Alaska mainly includes Canadian-Alaskan Cordillera, Brooks Range, and part of the North American platform containing mostly intrusive, metamorphic, and some clastic rocks (Fagel et al., 2014).The outcrops of the Canadian Arctic Archipelago are mainly composed of carbonate and clastic rocks (Phillips and Grantz, 2001;Fagel et al., 2014), whereas intrusive and clastic rocks are mostly characteristic for Greenland (Fagel et al., 2014).
Dissolved and suspended matter is transported to the Arctic Ocean by voluminous rivers, with the Lena and Mackenzie rivers being the largest on the Siberian and North American side, respectively, both directly affecting the western Arctic Ocean.The transported material is further distributed across the Arctic Ocean in water and/or ice by currents.The two main surface, wind-driven circulation systems are the clockwise Beaufort Gyre (BG) in the western Arctic and the Transpolar Drift (TPD), which carries water and ice from the Siberian margin to the Norwegian-Greenland Sea (e.g., Rudels, 2009).The strength and trajectories of these current systems may vary depending on changes in atmospheric pressure fields known as the Arctic Oscillation (Rigor et al., 2002).
Sedimentation in the Arctic Ocean is strongly controlled by sea ice that acts as sediment carrier but can also suppress sediment deposition under thick and persistent ice cover (Darby et al., 2006;Polyak et al., 2009).During glacial/deglacial events, multiple icebergs discharged into the Arctic Ocean from the termini of marine-based ice sheets and strongly affected sediment dispersal and deposition (e.g., Spielhagen et al., 2004;Polyak et al., 2009).Fine-grained sediments can also be transported by subsurface and deepwater currents, such as the Atlantic water (Winkler et al., 2002), but their role in the overall Arctic Ocean sedimentation is not well understood.

Materials and methods
Gravity core ARC4-BN05 (referred to hereafter as BN05) was collected from the Canada Basin in the vicinity of the Mendeleev Ridge (80 • 29.04 N, 161 • 27.90 W; 3156 m water depth) (Figs. 1 and 2) on the fourth Chinese National Arctic Research Expedition (CHINARE-IV) in 2009.The BN05 site was chosen in close proximity to earlier investigated cores FL224 and PS72/392-5 (Clark et al., 1980;Stein et al., 2010a) to enable robust correlation with the established stratigraphies.A total of 119 samples were taken at 2 cm intervals over the 238 cm BN05 length, and kept frozen until analyzed.
For age constraint within the radiocarbon range, accelerator mass spectrometry 14 C dating was performed on 1000-1200 tests of planktic foraminifers (Neogloboquadrina  England et al., 2009;Svendsen et al., 2004;Niessen et al., 2013).Proposed flow lines for grounded ice sheets and ice shelves (red and white arrows, respectively) are after Niessen et al. (2013).
For grain-size analysis, ∼ 2 g sediment samples were successively treated with 15 mL of 15 % H 2 O 2 , 5 mL of 3 mol L −1 HCl, and 20 mL of 1 mol L −1 Na 2 CO 3 for removing organic matter, biogenic carbonates, and biogenic silica, respectively.Grain-size measurements in the range of 0.02 to 2000 µm were performed on a Malvern Mastersize laser particle sizer (Mastersizer 2000) at the First Institute of Oceanography, SOA, China.
Coarse sediment > 63 µm was sieved from ∼ 10-15 g samples and counted under the microscope for foraminiferal and mineral grain numbers; planktonic foraminiferal amounts were expressed as percent of the total grain numbers (at least 300 grains per sample counted).
Concentrations of major elements, such as Ca and Mn, were determined on point samples by ICP-OES (iCAP6300) at the First Institute of Oceanography, SOA, China, following the standard procedures.For a more detailed downcore distribution, relative elemental abundances were obtained at 1 cm resolution using the Itrax XRF core scanner at the Polar Research Institute of China, set at 20 s count times, 10 kV X-ray voltage and 20 mA X-ray current.A good match of the ICP-  S1 for data used.
OES and Itrax XRF data (Fig. 3) verifies the consistency of results.To account for the dilution effects on the background sedimentation, such as by coarse debris and biogenic processes, element contents were normalized to Al (e.g., März et al., 2011).
Color reflectance was measured using a hand-held Minolta CM-2002 spectrophotometer at 1 cm intervals.Only the grayscale lightness index (L * ) is used in this paper.
A total of sixty 2 cm thick samples were collected at 4 cm intervals for paleomagnetic measurements performed at the Paleomagnetism and Geochronology Laboratory of the Institute of Geology and Geophysics, Chinese Academy of Science.Magnetic susceptibility was measured using the KLY-4s Kappabridge instrument.Subsequently, stepwise alternating-field (AF) demagnetization of natural remanent magnetization (NRM) was conducted using the 2-G Enterprises model 760-R cryogenic magnetometer (2G760) installed in a magnetically shielded (< 300 nT) space.AF demagnetization steps of 5-10 mT were used up to a maximum AF of 100 mT.
For bulk sediment mineralogy ∼ 5 g samples were dried, pulverized, passed through a 63 µm sieve, and loaded into aluminum holders.Samples were X-rayed from 5 to 65 • 2θ with Cu K-alpha radiation (40 kV, 100 mA) using a step size of 0.02 • 2θ and a counting time of 2 s per step on a D/max-2500 diffractometer equipped with a graphite monochromator with 1 • slits in the laboratory of the First Institute of Oceanography, SOA, China.Prior to the analysis, the instrument was blank-corrected and all samples were measured under the same conditions.Peak areas were estimated from X-ray diffraction (XRD) traces using Jade 6.0 software, and semi-quantitative estimates of bulk mineral percentages were calculated following Cook et al. (1975) (Table 1).
Samples for clay minerals determination (∼ 5 g) were first treated with H 2 O 2 (10 %) and HCl (1 mol L −1 ) to oxidize the organic matter and remove biogenic carbonates, respectively.Clay fractions (< 2 µm) were obtained by the Atterberg settling tubes method according to Stoke's law.Each sample was transferred to two slides by wet smearing.Samples were then air-dried prior to XRD analysis.One sample slide was The intensity factors are determined in 1 : 1 mixtures with quartz by obtaining the ratio of the diagnostic peak intensity of each mineral with that of quartz, which is assigned a value of 1.00.The detection limit in weight percent of the minerals in a siliceous or calcareous matrix can be obtained by multiplying the intensity factor by 0.12 (Cook et al., 1975).air dried at 60 • C for 2 h and analyzed.The second sample was solvated with ethylene glycol in an underpressured desiccator for at least 24 h at 60 • C. Every ethylene-glycol solvated sample was measured twice: the first scanning was done from 3 to 30 • 2θ with a step size of 0.02 • , and the second scanning from 24 to 26 • 2θ with a 0.01 • step.The latter was run as a slow scan to distinguish the 3.54/3.58Å kaolinite/chlorite double peak.Clay minerals were also identified by XRD using a D/max-2500 diffractometer with CuKα radiation (40 kV and 100 mA) in the laboratory of the First Institute of Oceanography, SOA, China.Peak areas representing the clay mineral groups were estimated from glycolated XRD traces using the 17 Å smectite, 10 Å illite, and 7 Å chlorite plus kaolinite peaks.Chlorite (004) was identified at 3.54 Å and kaolinite (002) at 3.58 Å (Biscaye, 1964), respectively.Semi-quantitative estimates of clay mineral percentages were calculated by means of Biscaye's (1965) factors.
To enhance the identification of potential contributions from various sediment sources, and thus the interpretation of downcore proxy distributions, principal component analysis (PCA) was performed in MATLAB (MathWorks, 2014).To account for proxies potentially indicative of sediment provenance and depositional processes and environments, PCA included all analyzed mineralogical proxies along with main grain-size groups (clay, silt, fine to medium sand (63-250 µm), and coarser grains), Ca and Mn concentrations, and foraminiferal numbers (Table S1 in the Supplement).A combined use of various sedimentological and geochemical data gives informative results in PCA application to paleoclimatic research, including studies of Arctic marine sediments (Pelto, 2014;Simon et al., 2014).The choice of variables for PCA performance was tested by Pearson correlation coefficients (Table S3).

General stratigraphy
As common for sediment cores from the Arctic Ocean (e.g., Jakobsson et al., 2000;Polyak et al., 2004Polyak et al., , 2009;;Spielhagen et al., 2004;Stein et al., 2010a, b), core ARC4-BN05 displays distinct cycles in sediment color and composition expressed in interlamination of dark-brownish and lightercolored grayish muds (silty clays, clay silts and sandy silt), with coarser dropstones occurring in several layers.The color cyclicity is approximated by changes in sediment lightness that largely mirrors the content of Mn (Fig. 3), consistent with other studies from the Arctic Ocean (e.g., Jakobsson et al., 2000;Polyak et al., 2004;Löwemark et al., 2008;Adler et al., 2009).We identify 18 distinctly brown units, from B1 to B18, characterized by elevated content of Mn (Fig. 3).Another prominent lithostratigraphic feature in the western Arctic Ocean, widely used for core correlation, is pink-white to whitish layers (PW) rich in detrital carbonates (e.g., Clark et al., 1980;Polyak et al., 2009;Stein et al., 2010a, b).We identify three major PW layers expressed both visually and in high Ca content (Fig. 3).Lower Ca peaks occur throughout the record without being clearly expressed in the core macroscopic appearance.Foraminiferal abundances are generally high (mostly > 50 % of > 63 µm grains) in brown units, except for B11-B13 and below B17-B18, and are very low to absent in gray units.This pattern is consistent with foraminiferal stratigraphy reported in earlier studies from the western Arctic Ocean (e.g., cores NP-26, HLY0503-JPC6 and 8, P1-92AR-P23 and 39: Polyak et al., 2004Polyak et al., , 2013;;Adler et al., 2009;Cronin et al., 2013;Lazar and Polyak, 2016).While only planktic foraminifers have been counted, data from correlative records indicate similar downcore variability in relative abundance of benthic foraminifers.

AMS 14 C dating
The measured AMS 14 C ages of core ARC4-BN05 were calibrated to calendar ages based on calibration using CALIB 7.10 (http://calib.org/calib/calib.html)(Table 2).The reservoir corrections of 790 and 1400 years were applied to Holocene and glacial-age samples, respectively, according to  Coulthard et al. (2010) and Hanslik et al. (2010).The same corrections have also been applied to 14 C ages in core 03M03 from the Chukchi Abyssal Plain (Wang et al., 2013; see Fig. 2 for location).

Paleomagnetic stratigraphy
While detailed paleomagnetic investigation is not an objective of this paper, we utilize the inclination data for an independent stratigraphic constraint in line with earlier studies (e.g., Jakobsson et al., 2000;Spielhagen et al., 2004;Polyak et al., 2009).Paleomagnetic inclination in core ARC4-BN05 shows mostly positive values oscillating around +70 • in the upper part of the core, with a major polarity change occurring at ∼ 120 cm (Fig. 3).A similar inclination drop has been identified in multiple sediment cores across the Arctic Ocean in the same stratigraphic position within estimated MIS 7, although the nature of this change in paleomagnetic characteristics is not well understood (e.g., Jakobsson et al., 2000;Polyak et al., 2009;Xuan and Channell, 2010).

Grain size and dropstones
Based on the results of grain-size analysis, sediment in core BN05 can be generally classified as sandy, poorly sorted mud (e.g., Blott and Pye, 2012).Overall, silt and clay predominate grain-size composition (33-60 and 23-61 %, respectively), but coarser particles also make a considerable contribution, with up to > 30 % peak contents of sand (> 63 µm) (Fig. 4a).We note that 4 µm was used as a cutoff size between clay and silt to account for overestimation of fine sediment diameters by laser diffraction, especially in the presence of platy particles (Beuselinck et al., 1998;Ramaswamy and Rao, 2006).Coarse size fractions (63-250 and > 250 µm) mostly co-vary downcore.
with the fine-and/or coarse-sand mode, but also forming very fine-grained intervals (e.g., at 37 cm, Fig. 4b).Mode 2 (7-7.5 µm) is common in the lower part of the core (below ∼ 175 cm), where it mostly co-occurs with mode 1 and coarse-grain tail, and also in distinct gray units around 30-40 and 90-100 cm in combination with the fine-sand mode 3 (e.g., 39 and 93 cm, Fig. 4b).
Several core intervals contain large rock fragments > 5 mm (dropstones).These rock fragments are mostly poorly rounded, subangular to angular in shape.Composition of sampled dropstones is illustrated in Fig. 4c.Most dropstones are represented by dolomite and low metamorphic quartz sandstone fragments of up to 5 cm in diameter.Also found were individual dropstones composed of volcanic rock and shale, as well as a few greisen dropstones near the base of the core.

Sediment mineralogy
The clay assemblage in samples from core ARC4-BN05 mainly consists of illite, chlorite, kaolinite and smectite (Fig. 5).The illite group is overall the major constituent of the clay mineral fraction, ranging between 43 and 73 %.Its downcore distribution pattern is opposite to that of the three other major clay-mineral groups -kaolinite, chlorite, and smectite (mostly present in very low contents).These three groups largely co-vary except for some lithostratigraphic intervals, such as PW layers.Elevated content of these clay minerals is characteristic for grayish sedimentary units.
The bulk mineral assemblage in core ARC4-BN05 mainly consists of quartz, K-feldspar, plagioclase, calcite, dolomite and pyroxene (Fig. 5).Quartz is generally the most abundant mineral, ranging between 20 and 51 % and typically peaking in grayish sediment units.K-feldspar, plagioclase and pyroxene (mainly augitic) mostly co-vary, with peaks in gray units in the upper part of the core, but more in brown units in the lower part starting from unit B10.Calcite has a high content in brown units of the upper part and much lower values below unit B9.Dolomite distribution shows distinct peaks reaching up to 53 %, with the highest peaks occurring in or adjacent to the PW layers.Similar to other minerals, the pattern of dolomite distribution changes around unit B10, with maxima in thick gray units below and in thin interlayers within brown units above this stratigraphic level.

Principal component analysis
The first five principal components identified by PCA with a Varimax rotation account for 77 % of the total variance, with relatively evenly distributed communalities (Table 3).This pattern presumably reflects a complexity of multi-proxy variables characterizing sedimentary environments and provenance, as well as their strong variability occurring over multiple climatic cycles.To further test the PCA performance, we have also run a factor analysis with the maximum likelihood extraction, which produced similar factor loadings and variance explained, thus indicating the robustness of the results.

Stratigraphic framework
As no single existing chronostratigraphic method can comprehensively constrain the age of the Arctic Ocean Pleistocene sediments, the age model for core ARC4-BN05 was developed by correlating multiple proxies (such as paleomagnetic, foraminiferal, and lithological; see Figs. 3 and 5), combined with 14 C ages in the youngest part of the record, to earlier established Arctic Ocean stratigraphies (e.g., Adler et al., 2009;Polyak et al., 2009Polyak et al., , 2013;;Stein et al., 2010a).Core PS72/392-5, raised very close to ARC4-BN05 and investigated in much detail (Stein et al., 2010a), was used to exemplify the correlation (Fig. 6).
The two 14 C dates from the uppermost, 10 cm thick brown sedimentary unit (B1) in core ARC4-BN05 clearly identify its Holocene age (Table 2; Fig. 3).Compilations of 14 C ages from the surficial and downcore sediments in the western Arctic Ocean (Polyak et al., 2009;Xiao et al., 2014) indicate that the age of this unit extends from ∼ 2-3 ka on top to ∼ 10-11 ka at the bottom contact, although an accurate estimate is impeded by the uncertainties with the reservoir ages.
Two 14 C dates of ca.42-44 ka from the brown unit B2 (Table 2; Fig. 3) apparently fall into MIS 3, consistent with earlier stratigraphic results (e.g., Polyak et al., 2004Polyak et al., , 2009;;Adler et al., 2009;Stein et al., 2010a).These ages should be, however, considered as crude estimates as they are close to the 14 C dating limit, and the age distribution in B2 has common inversions (e.g., Polyak et al., 2009).In cores with relatively elevated sedimentation rates this unit occurs as two distinct brown layers, indicated in some papers as B2a and B2b (e.g., Stein et al., 2010a, b;Wang et al., 2013).In core ARC4-BN05 this partitioning is less apparent due to low sedimentation rates, but the brownish sediment on top of the coarse detrital carbonate peak PW/W3, typically located between B2a and B2b, probably corresponds to B2a.
An abrupt increase in sediment age between closely spaced B1 and B2 in core ARC4-BN05 suggests a very con- densed section or a hiatus between MIS 1 and MIS 3.This age distribution is common for the western Arctic Ocean and has been attributed to very low to no sedimentation due to a very solid sea-ice cover or an ice shelf during the LGM in MIS 2 (e.g.Polyak et al., 2009;Wang et al., 2013).Below the range of 14 C ages the age model is based entirely on proxy correlations with earlier developed Arctic Ocean stratigraphies (e.g., Fig. 6).This correlation is enabled by the cyclic nature of sediment lithology and attendant proxies, where brown and grayish units generally correspond to interglacial (or major interstadial) and glacial climatic intervals, respectively (e.g., Jakobsson et al., 2000;Polyak et al., 2004Polyak et al., , 2009;;Adler et al., 2009;Stein et al., 2010a, b).In addition, correlation tie points are provided by rare or unique events, such as prominent detrital carbonate peaks (PW/W), major paleomagnetic inclination swings, and changes in foraminiferal assemblages and abundance pattern.
According to this approach, we identify foraminifera-and Mn-rich brown units B3-B7 and B8-B10 as warm substages of MIS 5 and 7, respectively (Figs. 3 and 6).This age assignment is corroborated by the prominent detrital carbonate peaks PW 2 and 1 near the bottom of MIS 5 and 7, respectively.Furthermore, the principal drop in paleomagnetic inclination in core ARC4-BN05 occurs in the lower part of MIS 7, consistent with many cores from the Arctic Ocean (e.g., Jakobsson et al., 2000;Spielhagen et al., 2004;Adler et al., 2009;Polyak et al., 2009).A solidly grayish, foraminifera-and Mn-poor unit separating brown units B2 and B3 is accordingly considered as related to glacial MIS 4, and a similar unit between B7 and B8 related to MIS 6.It is possible, however, that most of the fine-grained, grayish sediment was deposited during deglaciations following the actual glacial intervals, which may have been very compressed, similar to the LGM.
Stratigraphy below MIS 7 has been less investigated in prior studies, and is more difficult to address due to often less distinct units and scarce to absent foraminifers, probably resulting from stronger dissolution (e.g., Lazar and Polyak, 2016).Therefore the age model for the lower part of the core is more tentative.Nevertheless, a prominent oldest foraminiferal peak in units B14-B15 (Fig. 3) allows us to identify these units as MIS 11 by comparison with other microfaunal records reported from the western Arctic Ocean (e.g., Cronin et al., 2013;Polyak et al., 2013).While individual species have not been counted in ARC4-BN05, predominant planktic foraminifers in this peak are identifiable as Turborotalitaegelida, constituting a unique event in the Arctic stratigraphy (see Cronin et al., 2013Cronin et al., , 2014, for more detail).MIS 13 and 15 have been tentatively assigned to units B17 and B18 underlying a prominent gray interval attributed to MIS 12. Overall, the record in core ARC4-BN05 is estimated to represent the last ca.0.5-0.6Ma, that is, most of the middle to late Quaternary with an average sedimentation rates of 4-5 mm kyr −1 .

Depositional environments and sediment provenance
Distribution of various terrigenous components in Arctic sediment records carries information on sediment sources and depositional environments, and thus paleocirculation and changes in paleoclimatic conditions, such as connection to other oceans and build-up/disintegration of ice sheets (e.g., Bischof and Darby, 1997;Krylov et al., 2008;Polyak et al., 2009;Stein et al., 2010a, b;Yurco et al., 2010;Fagel et al., 2014).We utilize the data on clay and bulk minerals along with the grain size and total Ca and Mn distribution in core ARC4-BN05 to reconstruct changes in glacial conditions and circulation in the western Arctic Ocean during several glacial cycles extending to estimated ca. 0.5-0.6Ma.In this work we capitalize on earlier studies on the distribution of bulk and/or clay minerals in surface and downcore Arctic Ocean sediments (e.g., Vogt, 1997;Stein, 2008;Krylov et al., 2014;Zou, 2016), corroborated by more targeted provenance proxies, such as radiogenic isotopes (Fagel et al., 2014;Bazhenova et al., 2017), heavy minerals (Stein, 2008;Kaparulina et al., 2016), composition of coarse debris (Bischof et al., 1996;Wang et al., 2013), and iron oxide grains (e.g., Bischof and Darby, 1997;Darby et al., 2002).To optimize the PCA results for clarifying relationships between various sedimentary proxies, we plotted the leading PC loading scores as biplots in the PC 1-2 and PC 3-4 space (Fig. 7a).These plots help to identify several sedimentary variable groups with high loadings (> 0.7 average) in at least one of the lead-  3 for loading data and Table S3 for correlation between variables).Sedimentary variable groups revealed by the loading distribution are enclosed by ellipses and numbered, with the closest groupings highlighted in gray.(b) Downcore distribution of sedimentary variable groups plotted using combined PC 1-2 and PC 3-4 scores (see Table S4 for score data).
ing PCs.Group 1 consists of various proxies characteristic of brown layers (primarily Mn, foraminifera, calcite, and clay, with an apparent affinity to chlorite).The opposing group 2 includes most clay minerals except chlorite, with a proximity to sand and, to a lesser extent, silt size fractions.Bulk mineralogy proxies are largely represented by the opposing groups 3 and 4. Group 3 comprises feldspar, pyroxene, and more distant quartz and plagioclase.Group 4 builds around dolomite that has high loading scores in both PC 3 and 4, along with bulk Ca, and shows affinity to Kfsp/Plag and Qz/Fsp indices.In addition to these groups revealed by the leading PC biplots, PC 5 (10 % variance) shows high scores for sand and coarser sediment, with silt as the main opposing variable.
To gain insight into stratigraphic changes in sedimentary environments and provenance, we plotted the distribution of the identified variable groups 1-4 using the combined downcore scores of PC 1-2 and PC 3-4 (Fig. 7b).A combination of the PC group composition and downcore variability provides useful guidance for interpreting major sedimentary controls and their stratigraphic evolution.

Grain size and depositional processes
A variable, mostly multimodal distribution of grain size in core BN05 indicates multiple controls on sediment delivery and/or deposition.The prevailing mode 1 at ∼ 4 µm (Fig. 4b), often in variable combinations with the fine-sand mode, is common for brown units, except for the oldest layers (B16-B18; estimated MIS 13/15 and partly 11; Fig. 6).This granulometric pattern is similar to grain-size distribution with an average mode at ∼ 3.4 µm reported for Holocene sediments across the Arctic Ocean (Darby et al., 2009).Furthermore, sediment in core BN05 with the same mode also makes up the most fine-grained intervals in glacial/deglacial units, such as MIS 4 and 6 (at ∼ 30-40 and 100-110 cm; Fig. 4b).We infer that mode 1 represents some combination of deposition from sea ice and from suspension that could result from winnowing of fines from the basin margins and ridges during interglacials, as well as overflow plumes discharged by retreating glaciers during glacial/deglacial intervals.An occurrence of apparently similar grain-size pattern in interglacial and fine-grained glacial/deglacial intervals might indicate a convergence of glacial erosion processes with those related to sea-ice formation and transportation.A similar grain-size interpretation has been earlier proposed for sediment from the Canada Basin with the principal mode at ∼ 2 µm (Clark et al., 1980).This apparent discrepancy may be related to the methodological offset between grain size determined by the pipette method vs. laser diffraction, where the latter produces larger diameters for fine sediment, especially in the presence of platy particles (Beuselinck et al., 1998;Ramaswamy and Rao, 2006).
Additionally, modes 1 and 2 make up a bimodal distribution in the lowermost part of the core -mostly in estimated MIS 13/15 and near the bottom of MIS 11.The predominant stratigraphic position in brown units makes the glacigenic origin of this sediment unlikely.We hypothesize that this grain-size pattern reflects a combination of "normal" interglacial environments with winnowed silts deposited by downwelling of shelf waters enriched in dense brines.Although no observational evidence exists for such waters penetrating deeper than the halocline (∼ 200 m) under modern Arctic conditions, periods of stronger cascading in the past have been inferred from sediment distribution on the slopes (Darby et al., 2009) and some sedimentary proxies, such as radiogenic isotopes (Haley and Polyak, 2013;Jang et al., 2013).The bimodal distribution of fine sediment in the lower part of the record is accompanied in most samples by a small but consistent coarse-sand mode (400-450 µm), likely indicating the presence of iceberg rafting.
Coarse sediment, up to dropstones of several centimeters large, is a consistent feature in core BN05.In the apparent absence of strong current control on sedimentation, except for some shelf areas, and a pervasive presence of floating ice, coarse sediment in the Arctic Ocean is typically attributed to ice rafting, including sea ice and icebergs (e.g., Stein, 2008;Polyak et al., 2010, and references therein).Sedimentological studies in areas of sea-ice formation or melting and in ice itself indicate that sediment carried by sea ice in the Arctic Ocean is predominated by silt and clay, while coarser fractions are of minor importance (Clark and Hanson, 1983;Nürnberg et al., 1994;Hebbeln, 2000;Darby, 2003;Dethleff, 2005;Darby et al., 2009).Some studies suggest a higher content of sand in ice formed at the sea floor (anchor ice) (Darby et al., 2011), but the contribution of this source still needs to be evaluated.Furthermore, the role of sea ice on sedimentation in the Arctic Ocean is not clear for glacial intervals, when most of the sediment entrainment areas were exposed or covered by ice sheets.In contrast, in iceberg-rafted sediment, deposited mostly in glacial/deglacial environments, the content of large size fractions, from sand to boulders, is typically high, in excess of 10-20 % (Clark and Hanson, 1983;Dowdeswell et al., 1993;Andrews, 2000).Thus, elevated content of coarse sediment can be regarded as a good indicator of intense iceberg rafting.Such events are not probable during full interglacials, exemplified by modern conditions, but most likely occurred at times of instability and disintegration of ice sheets that extended to the Arctic Ocean in the past (e.g., Spielhagen et al., 2004;Stokes et al., 2005;Polyak et al., 2009).
In core BN05, coarse fractions (from coarse silt to sand) measured at different sizes show very similar distribution patterns (Fig. 4a), indicating the same predominant delivery mechanism, that is, iceberg rafting.This pattern is reflected in a good correlation of fine to medium sand (63-250 µm) with coarser, > 250 µm fractions, that defines one of the principal components (PC 5: Table 3).Increased coarsegrain content mostly characterizes grayish units, especially near gray-to-brown sediment transitions, and the PW layers, but also occurs in brown units in the upper part of the record.The latter peaks enriched in detrital carbonates (high dolomite and total Ca) represent interstadial or incomplete www.clim-past.net/13/511/2017/Clim.Past, 13, 511-531, 2017 interglacial conditions, such as MIS 3, MIS 5a, and parts of MIS 7 (Fig. 6).
A common occurrence (separate or combined) of two coarse grain modes, around 85-90 and 400-450 µm, may indicate different sources for iceberg-rafted material or different thresholds for glacial disintegration of various rock types.While a more thorough interpretation requires further research, we note that grain-size mode 1 may co-occur with both fine-and coarse-sand modes, mode 2 only with the finesand mode, and bimodal 1/2 sediment type only with the coarse-sand one.

North American provenance
One of the most robust sedimentary variable groups is distinctly characterized by high loadings of dolomite along with total Ca content (group 4: Fig. 7a, Table 3).Dolomite has been proposed as the main contributor of Ca in sediment cores from the western Arctic Ocean, with an especially high content in multiple coarse-grain peaks of detrital carbonates (Bischof et al., 1996;Phillips and Grantz, 2001;Polyak et al., 2009;Stein et al., 2010a, b).A high correlation (r = 0.81) and consistent PC grouping of dolomite and total Ca (Fig. 7a) corroborates their affinity in ARC4-BN05, although calcite can also contribute to Ca content (r = 0.58), especially in interglacial intervals with high foraminiferal numbers.We note that total Ca may be a redundant proxy in the presence of dolomite and calcite data; however, it is convenient for a comparison with a growing number of cores analyzed for elemental composition using XRF scanners (e.g., Löwemark et al., 2008;Polyak et al., 2009).
The main western Arctic source for dolomite is the extensive, carbonate-rich Paleozoic terrane in northern Canada (North American Platform;Fig. 1;Okulitch, 1991;Harrison et al., 2008).During the Pleistocene this terrane was repeatedly impacted by the LIS with a subsequent transport of eroded material into the western Arctic Ocean (e.g., Stokes et al., 2005;England et al., 2009).The distribution of dolomite in Arctic sediment cores is thus a robust indicator of the North American provenance and can be used for reconstructing the history of the LIS sedimentary inputs.
Consistent with other cores from the western Arctic Ocean, overall high dolomite content in core ARC4-BN05 has major peaks corresponding to visually identifiable PW/W layers enriched in coarse debris (Fig. 5).As has been suggested in earlier studies (e.g., Stokes et al., 2005;Polyak et al., 2009), we infer that the dolomite peaks are related to pulses of massive iceberg discharge from the LIS during the periods of its destabilization and disintegration.Furthermore, radiogenic isotope studies demonstrate that fine sediment in the dolomitic peaks also has North American provenance (Fagel et al., 2014;Bazhenova et al., 2017).These results indicate that dolomite may have been transported not only by icebergs but also in meltwater plumes coming during deglaciations from the Canadian Archipelago or the Mackenzie River.
As noted above, a change in the stratigraphic pattern of dolomite distribution occurs around unit B10 estimated to correspond to the lower part of MIS 7 (Fig. 6).In older sediments dolomite maxima co-occur with glacial (predominantly gray) intervals, whereas in the younger stratigraphy dolomite peaks in brown sediment or grayish interlayers within brown units (MIS 3, 5, and 7), presumably corresponding to transitional paleoclimatic environments, such as interstadials or stadials within complex interglacial stages.
Other potential mineral indicators related to the North American provenance are quartz/feldspar and Kfeldspar/plagioclase ratios as exemplified by the BN-05 PCA results (group 4: Fig. 7a), consistent with earlier studies (e.g., Vogt, 1997;Zou, 2016;Kobayashi et al., 2016).High Qz / Fsp ratio has been related to a considerable presence of sedimentary rocks enriched in quartz, but not feldspar, in the Canadian Arctic in comparison with the Siberian margin (Vogt, 1997;Zou, 2016;Kobayashi et al., 2016).In core ARC4-BN05 the distribution of this index is generally similar to dolomite (Fig. 5), except for some coarse-grain peaks, notably low Qz/Fsp values in PW 1 and 3, resulting in an overall lower correlation (r = 0.46).This pattern may be related to grain-size variance or might reflect provenance differences.Low plagioclase content has also been identified for intervals with high detrital inputs from the Canadian Arctic (Vogt, 1997;Zou, 2016).Especially high Kfsp/Plag values accompany dolomitic peaks in the older glacial intervals corresponding to MIS 12 and 10 (Figs. 5 and 6).

Siberian provenance
Mineral proxies potentially linked to Siberian provenance make two distinct groups, as reflected in the PCA results (groups 2 and 3: Fig. 7a, Table 3).Group 3 comprises primarily pyroxene, feldspar, and plagioclase, and strongly anticorrelates with the North American proxies, primarily dolomite.The downcore distribution pattern of this group changes from the affinity to interglacials in the lower part of the record to peaks in glacial/deglacial intervals related to MIS 4 and 6 (Fig. 7b).The major source for pyroxene in the Arctic Ocean is the Siberian trap basaltic province that drains to the Kara Sea and western Laptev Sea (Fig. 1; Washner et al., 1999;Schoster et al., 2000;Krylov et al., 2008).However, basaltic rocks related to the Okhotsk-Chukotka province (Fig. 1) may have also provided a significant source of pyroxenes, as exemplified in surface sediments by a relative pyroxene enrichment in the Chukchi Basin on the background of overall low values in the western Arctic Ocean (Dong et al., 2014).Distributions of feldspar and plagioclase at the Siberian margin show elevated contents occurring both in the western Laptev Sea and the East Siberian Sea (Zou, 2016).
Based on a considerable affinity of the pyroxene-feldspar group to brown units and a lack of correlation with coarse sediment fractions, we infer that it is primarily related to sea-ice transport during interglacial/deglacial intervals, with sources potentially including the East Siberian margin and more westerly areas.The difference in both the sources and delivery processes from the LIS proxies may explain an especially strong opposition of these groups.Multiple studies suggest that sea ice from the Kara and Laptev seas may transport sediments to the Canada Basin under favorable atmospheric conditions, such as the positive phase of the Arctic Oscillation (Behrends, 1999;Darby, 2003;Darby et al., 2004Darby et al., , 2012;;Yurco et al., 2010), although it remains to be investigated to what extent this circulation pattern could have provided a significant sediment source for the western Arctic Ocean in the Pleistocene.

East Siberian Ice Sheet
Another leading sedimentary variable group comprises primarily clay minerals smectite, kaolinite, and chlorite, and shows affinity to coarse sediment, especially consistently to fine sand (63-250 µm) (group 2: Fig. 7a, Table 3).This composition is especially characteristic for intervals estimated as MIS 4, 6, and 12.The association of clay minerals with coarse sediment (correlation reaching as high as r = 0.65 for kaolinite) is unusual and suggests that they may have been derived by glacial erosion of source hard rocks.This linkage has been elaborated for kaolinite distribution in the Barents Sea and central Arctic Ocean (Junttila, 2007;Vogt and Knies, 2009;Krylov et al., 2014).While kaolinite sources, such as Meso-Cenozoic paleosols and shales, are mostly known in the western Arctic from northern Alaska and Canada (Naidu et al., 1971;Darby, 1975;Dalrymple and Maass, 1987), kaolinite weathering crusts have also been described from the East Siberian margin (Slobodin et al., 1990;Kim and Slobodin, 1991).Smectite, which is typically related to chemical weathering of basic rocks has been mostly associated in Arctic sediments with delivery from Siberian trap basalts (Fig. 1) as reflected in the surface sediments, suspended particulate material, and sea-ice samples from the Kara Sea and western Laptev Sea (Stein et al., 1994;Wahsner et al., 1999;Schoster et al., 2000;Dethleff et al., 2000).Peaks of smectite related to that source are especially charcteristic for deglacial intervals in sediment cores from the eastern Arctic Ocean (Vogt and Knies, 2008).However, considerable sources of smectite also exist further east along the Siberian margin due to basaltic outcrops related to the Okhotsk-Chukotka volcanic province (Fig. 1), resulting in a high content of smectite in surface sediments of the East Siberian and Chukchi seas (Naidu et al., 1982;Viscosi-Shirley et al., 2003;Nwaodua et al., 2014).Chlorite is also common in surface sediments and suspended particulate material at the East Siberian margin (Dethleff et al., 2000;Viscosi-Shirley et al., 2003).Modern and Holocene sediments on the Chukchi shelf are especially enriched in chlorite due to advection from the North Pacific at high sealevel stands (Kalinenko, 2001;Ortiz et al., 2009;Nwaodua et al., 2014;Kobayashi et al., 2016); however, this mechanism is only applicable to interglacial periods.
We infer that sediment with a concerted enrichment in smectite, kaolinite, and chlorite clay minerals associated with coarse fractions was transported to the Canada Basin primarily in relation to the existence of large ice sheets in northern East Siberia during glacial periods.Radiogenic isotope signature in upper Quaternary records from the Mendeleev Ridge also indicates that the Okhotsk-Chukotka volcanic rocks provided one of the principal end members, especially during MIS 4 and 6 (Fagel et al., 2014;Bazhenova et al., 2017).This sediment had to be transported into the Arctic Ocean directly from the East Siberian/Chukchi margin as the alternative pathway via the Bering Sea only operated at high interglacial sea levels, when the Bering Strait was open for throughflow (e.g., Keigwin et al., 2006;Ortiz et al., 2009).Considering an affinity of the kaolinitesmectite-chlorite group with sediments coarser than clays, corresponding to grain-size modes 2 and 3, their distribution across the basin was likely related to iceberg rafting and glacial underflows, as discussed above in Sect.5.2.1.A relatively fast and direct delivery mechanism by debris flows and ensuing turbidites may explain a good preservation of fragile clay minerals, normally not resistant to physical erosion.Some early paleoglaciological studies proposed the existence of a thick Pleistocene ice sheet centered over the East Siberian shelf (Hughes et al., 1977;Grosswald and Hughes, 2002).The inference of former ice sheets/shelves in this region is now corroborated by multibeam bathymetry and sub-bottom data revealing multiple glacigenic features on the top and slopes of the Chukchi and East Siberian margin (Polyak et al., 2001(Polyak et al., , 2007;;Jakobsson et al., 2008Jakobsson et al., , 2014Jakobsson et al., , 2016;;Niessen et al., 2013;Dove et al., 2014).ESIS has also been reproduced by numerical paleoclimatic modeling for a large Pleistocene glaciation exemplified by MIS 6 (Colleoni et al., 2016).Sedimentary proxies indicative of the Okhotsk-Chukotka provenance in cores from the Canada Basin may provide an additional tool for reconstructing the ESIS history.

Interglacial signature
Data points from brown units make up a distinct sedimentary variable group with Mn, foraminiferal numbers, calcite, and fine sediment as lead variables (group 1: Fig. 7a; Table 3).This composition is consistent with the modern-type Arctic Ocean environments characterized by predominant controls of sediment deposition by sea ice, considerable biological activity in summer, and high sea levels.The last of these is important for providing supply of Mn from the surrounding shelves (März et al., 2011;Löwemark et al., 2014).The same condition may also control biological production, and thus foraminiferal numbers, via export of nutrients from the marginal seas (e.g., Xiao et al., 2014), although interaction of this factor with sea-ice conditions still needs to be clarified.
www.clim-past.net/13/511/2017/Clim.Past, 13, 511-531, 2017 We note that the absence (dissolution) of foraminiferal tests in brown units corresponding to MIS 9 and below MIS 11 likely weakens their relationship to other interglacial proxies.Nevertheless, the foraminiferal variable shows a consistent proximity to Mn, clay, and calcite in the PCA results (Fig. 7a).
The mineral having the closest distribution to the main constituents of PC group 1 is illite, consistent with a predominant occurrence in brown, interglacial/major interstadial units (Figs. 5 and 7a).Illite is atypical high-latitude clay mineral, mainly supplied by physical weathering of metasedimentary and plutonic rocks (Chamley, 1989;Junttila, 2007).High illite concentrations in surficial Arctic Ocean sediments have been found in many areas including the Alaska margin and adjacent Canada Basin (Dong et al., 2014;Kobayashi et al., 2016), East Siberian Sea and the adjacent part of the Laptev Sea (Wahsner et al., 1999;Kalinenko, 2001;Viscosi-Shirley et al., 2003;Dethleff, 2005;Zou, 2016), and northern Greenland and Svalbard regions (Stein et al., 1994).In core ARC4-BN05 illite has consistently high values in generally fine-grained brown units (Fig. 5), although peak values may not exactly coincide with those of Mn or foraminiferal numbers.In addition, illite shows a prominent peak in a very finegrained interval at ∼ 35 cm within glacial/deglacial sediment of estimated MIS 4. This distribution is consistent with the pattern in both surface sediments and sediment cores, where illite is characteristic for fine-grained sediment indicative of transportation by sea ice or in the water column (Krylov et al., 2014).As shown by sediment-core studies, these mechanisms can provide high illite levels under both interglacial (this study) and glacial/deglacial environments (Knies and Vogt, 2003;Yurco et al., 2010).The latter is probably associated with deposition of fine sediment from glacial overflows, as exemplified by the fine-grained part of MIS 4 deglaciation.
High contents of calcite in core ARC4-BN05 mostly cooccur with high numbers of foraminifers (Fig. 7a; Table 3), indicating that calcite in these sediments is to a large extent biogenic, consistent with earlier results from the study area (Stein et al., 2010a).Nevertheless, in the lower part of the record, where calcareous fossils are mostly not preserved, calcite shows a considerable affinity to dolomite, which corroborates a mixed, biogenic and detrital nature of calcite in Arctic Ocean sediments (e.g., Vogt, 1997).

Evolution of sedimentary environments
The stratigraphically changing pattern of sediment delivery and deposition, including cyclic glacial-interglacial fluctuations and longer-term changes, indicates complex interactions of climatic and oceanographic factors controlling depositional environments in both glacial and interglacial intervals.A long-term trend in interglacial environments is indicated by a shift from predominantly Siberian to more North American provenance, especially strong in MIS 5 and 1, and increasingly high scores of interglacial proxies (group 1), with a threshold around the bottom of MIS 7 (Fig. 7b).Glacial environments show an apparently more complex provenance change, with Siberian sources predominating MIS 4 and 6, and Laurentide provenance controlling MIS 8 and 10 (Fig. 7b).Earlier glaciations, exemplified by a prominent MIS 12 unit, have a mixed signature of high smectite and dolomite contents, likely reflecting a combination of East Siberian and LIS inputs.In addition, an interglacial-type signature (group 1) characterizes some intervals in MIS 4 and 6 as well as intermittent (stadial) intra-MIS 3, 5, and 7 events.We note that MIS 2 is not represented in these data due to its very compressed nature.

Glacial environments
The identified changes in sedimentary environments and provenance can be explained by several types of controls, including configuration of ice sheets against sea level and climatic conditions, sediment delivery mechanisms, and circulation.Ice sheet sites and geometry at specific time intervals dictate the timing and location of major sediment discharge events into the Arctic Ocean.Transportation mechanisms, such as by icebergs, debris flows, or suspension plumes, further control sediment delivery to specific sites.Finally, oceanic circulation affects the distribution of sediment across the oceanic basins.This may include surface circulation driving sea ice, icebergs, and surface plumes; deep circulation affecting turbidite/contourite pathways; and downwelling of sediment-laden dense waters.
We infer that sedimentary variations observed in core BN05 and correlative records from the western Arctic Ocean can be explained by the evolution of surrounding ice sheets and associated changes in oceanic conditions, such as circulation, sea ice, and biota.It has been known from early studies (e.g., Clark et al., 1980;Winter et al., 1997) that glacial, notably LIS, impact on the western Arctic Ocean has been steadily increasing over the time span covered by sediment cores from this region.A recent investigation utilizing a more up-to-date stratigraphic paradigm estimated the timing of a step increase in LIS inputs as ca.0.8 Ma (Polyak et al., 2013), consistent with the onset of major glaciations in the Northern Hemisphere (Head and Gibbard, 2015).Core BN05 provides a record of sediment deposition in the Canada Basin, and thus glacial inputs into the western Arctic Ocean during most of the time interval to follow.
Considering the overall gradual growth of Pleistocene Arctic ice sheets, we infer that the shift from Siberian to North American sources between MIS 12 and 10 was primarily related to the expansion of the LIS, especially the northwestern Keewatin sector that discharges into the western Arctic Ocean.However, its further growth may have had an opposite effect due to a more massive ice sheet that required warmer climatic conditions and/or higher sea levels to destabilize it.Based on data for the last glacial cycle, the Keewatin sector of the LIS rested mostly on relatively elevated terrane of the Canadian Archipelago and adjacent mainland, fringed by a narrow continental shelf and dissected by numerous channels providing conduits for ice streams and evacuation of icebergs at rising sea levels (Stokes et al., 2005(Stokes et al., , 2009;;England et al., 2009;Margold et al., 2015).The latter events are illustrated in BN05 data by intra-MIS 5 stadials with a consistent LIS signature (group 4: Fig. 7b).Especially high LIS scores characterize PW layers 2 and 3 attributed to MIS 5d and late MIS 3, respectively.A similar, LIS-dominated pattern likely represents the last deglaciation as indicated by a number of provenance studies (e.g., Stokes et al., 2005;Jang et al., 2013;Bazhenova et al., 2017).
In comparison to the LIS, a presumably much smaller ESIS, formed on a broad and overall flat East Siberian/Chukchi margin (Niessen et al., 2013;Dove et al., 2014;Colleoni et al., 2016), had to be responsive to sea-level changes even at low levels.It may be possible that the ESIS also increased in size by MIS 6, known as a time of a dramatic increase in glacial inputs from the Barents-Kara Ice Sheet into the eastern Arctic Ocean (e.g., O'Regan et al., 2008).A synchronous MIS 6 expansion of both North American and Siberian ice sheets and related ice shelves might explain the deep-keel glacial erosion of the Lomonosov Ridge at modern water depths exceeding 1000 m (Jakobsson et al., 2016, and references therein).
A concurrent interpretation can be proposed with a focus on sediment transportation processes as deposits of some glacial intervals, notably MIS 12 and parts of MIS 4 and 6, are associated with grain-size mode 2, potentially indicating glacial debris flow/turbidite emplacement.Large debris flows entering the Chukchi Basin and continuing as turbidites into Canada Basin, as exemplified by sub-bottom sonar profiles (Niessen et al., 2013;Dove et al., 2014), may have overprinted deposition from icebergs.We note that deposits of MIS 4 and 6 also contain intervals where Siberian provenance is combined with interglacial positive scores (group 1: Fig. 7b) due to their fine-grained composition along with high illite content.These sediments likely represent deposition from suspension plumes, potentially marking especially strong deglacial meltwater discharge.A prominent fine-grained, finely laminated interval within MIS 4 deglaciation (possibly extending into MIS 3) has been reported from multiple cores across the Chukchi Basin-Mendeleev Ridge area (Adler et al., 2009;Matthiessen et al., 2010;Wang et al., 2013;Bazhenova et al., 2017).
Under modern conditions the BN05 site is mostly controlled by the Beaufort Gyre current circulation system, although it can also be affected by the Transpolar Drift during strong shifts in the Arctic Oscillation (Rigor et al., 2002).This setting probably applies to the Holocene and comparable interglacial conditions (Darby and Bischof, 2004).Some authors have suggested that during glacial periods the surface circulation that controls pathways of iceberg and sea-ice drift may have been considerably different from the modern pattern, with both North American and Siberian sources short-cutting the Arctic Ocean towards the Fram Strait (Bischof and Darby, 1997;Stärz et al., 2012).These changes would have potentially affected the study area, possibly making it more exposed to the Siberian provenance than under present conditions.However, the existing reconstructions based on very limited records with only crude stratigraphic controls, need to be elaborated by spatially and stratigraphically more representative data constraining past circulation changes.In particular, glacial maxima may be elusive, especially in the western Arctic Ocean, due to extremely low sedimentation rates or a hiatus, as exemplified by the LGM (Polyak et al., 2009;Poirier et al., 2012).
An overall integration of potential controls on sediment deposition in the study area during major identified types of glacial environments are illustrated in Fig. 8.More studies are needed to discriminate between different controls, including proxy records providing higher resolution for target intervals as well as modeling experiments to test spatial and stratigraphic variability in such factors as iceberg and meltwater discharge and their ensuing distribution pathways.

Interglacial environments
The long-term trend in interglacial environments reflected in a shift from negative to increasingly positive scores of interglacial proxies (group 1: Fig. 7b), with a threshold around the bottom of MIS 7, can be partially explained by the absence of calcareous foraminifers in the lower part of the record.However, even MIS 11, which has abundant foraminifers, has low interglacial scores, suggesting more controls.One possibility is that this trend was related to the evolution of circum-Arctic ice sheets that would have inevitably incurred changes in oceanic conditions, such as circulation and sea ice.An expansion of perennial sea ice in the western Arctic Ocean near the MIS 7 bottom has been proposed based on foraminiferal assemblages (Polyak et al., 2013;Lazar and Polyak, 2016).This step change has been tentatively attributed to the LIS growth that may have affected sea-ice conditions via increased albedo and/or higher meltwater inputs.This inference is consistent with a coeval change from mostly Siberian (group 3) to North American (group 4) provenance during interglacials in BN05 (Fig. 7b).In addition to a more lingering LIS during interstadials/interglacials, this shift in provenance could be related to a strengthening of the Beaufort Gyre as more sea ice filled the western Arctic Ocean.
More limited sea-ice cover in the older part of the middle Pleistocene could have also enhanced the production of dense brines at the Siberian margin, resulting in a deeper convection and cascading of shelf sediments to the deep basin.This scenario would explain an unusual grain-size composition of sediments in the older interglacials combining mode 2, indicative of winnowed silt, with a typical interglacial fine-grained mode 1.

Summary and conclusions
Sediment core ARC4-BN05 was collected from the Canada Basin in the vicinity of the Chukchi Plateau and the Mendeleev Ridge, Arctic Ocean, on the fourth Chinese National Arctic Research Expedition (CHINARE-IV).Based on correlation to earlier proposed Arctic Ocean stratigraphies (e.g., Adler et al., 2009;Stein et al., 2010a;Polyak et al., 2013) and AMS 14 C dating of the youngest sediments, the BN05 record covers the late to middle Quaternary (MIS 1-15, ca.0.5-0.6Ma).The core was investigated for multiple sedimentary proxies including clay and bulk mineralogy, grain size, paleomagnetism, elemental content, and planktonic foraminiferal numbers with an average estimated age resolution of 4-5 ka per sample.This study, facilitated by principal component analysis of major paleoceanographic variables, provides important new information about sedimentary environments and provenance in the western Arctic Ocean on glacial timescales.The results enhance our knowledge on the history of Arctic glaciations and interglacial conditions.
Glacially derived sediment can be discriminated between the North American and Siberian provenance by their mineralogical and textural signature.In particular, peaks of dolomite debris, including large dropstones, track the Laurentide Ice Sheet (LIS) discharge events, while the East Siberian Ice Sheet (ESIS) inputs are inferred from combined peaks of smectite, kaolinite, and chlorite associated with coarse sediment.Siberian provenance is also identified from high content of pyroxene, feldspar, and plagioclase, unrelated to coarse sediment.This sedimentary signature is interpreted to indicate sea-ice transport from the Siberian margin during interglacial/deglacial intervals.Full interglacial environments are characterized by overall fine grain size, high content of Mn (and resulting dark-brown sediment color), and elevated contents of calcite and chlorite.Foraminiferal tests are abundant in interglacial units in the upper part of the record (MIS 1-7) and estimated MIS 11 but have very low numbers in other interglacials older than MIS 7, apparently due to dissolution.
In addition to glacial-interglacial cyclicity, the investigated record indicates variable impacts of LIS vs. ESIS on sediment inputs at different glacial events, along with a longterm change in middle to late Quaternary sedimentary environments.Based on the age model employed, major LIS inputs to the study area occurred during MIS 3, intra-MIS 5 and 7 events, MIS 8, and MIS 10, while ESIS signature is characteristic for MIS 4, MIS 6 and MIS 12.These differences may be related to ice-sheet configurations at different sea levels, sediment delivery mechanisms (iceberg rafting, suspension plumes, and debris flows), and surface circulation.A long-term shift in the pattern of sediment inputs shows an apparent step change near the estimated MIS 7/8 boundary (ca.0.25 Ma), consistent with more sea-ice growth in the Arctic Ocean inferred from benthic foraminiferal assemblages (Lazar and Polyak, 2016).This development of Arctic Ocean paleoenvironments possibly indicates an overall glacial expansion at the western Arctic margins, especially in North America.Such expansion may have affected not only glacial but also interglacial conditions via increased albedo and/or higher meltwater inputs, as well as a strengthening of the Beaufort Gyre circulation as more sea ice filled the western Arctic Ocean.

Figure 1 .
Figure 1. Background map showing the location of core ARC4-BN05,the main Arctic rivers and the two major surface current systems: Beaufort Gyre (BG), Transpolar Drift (TPD) and Bering Strait (BS).Schematic geological map shows the distribution and prevailing lithology of the main terrains adjacent to the Arctic Ocean(Fagel et al., 2014).

Figure 2 .
Figure 2. Index map showing the location of core ARC4-BN05 (yellow circle) and other cores from previous studies mentioned in this paper (red circles).LR, MR, AR, and NR are Lomonosov, Mendeleev, Alpha, and Northwind ridges, respectively; NGS is Norwegian-Greenland Sea.White lines show maximal Pleistocene limits reconstructed for Greenland, Laurentide, Eurasian, and East Siberian ice sheets (GIS, LIS, EAIS and ESIS;England et al., 2009;Svendsen et al., 2004;Niessen et al., 2013).Proposed flow lines for grounded ice sheets and ice shelves (red and white arrows, respectively) are afterNiessen et al. (2013).

Figure 3 .
Figure 3. Lithostratigraphy and major proxies in core BN05: core photograph with brown layer indices, lightness, Ca and Mn content (bulk XRF -gray line, ICP-OES -black line), paleomagnetic inclination, magnetic susceptibility, planktic foraminiferal abundance, and AMS 14 C datings.Predominantly dark-brown intervals B1-B18 are highlighted in gray; high-Ca, pink-white layers are marked by purple lines.The main inclination drop is marked by the orange line.See TableS1for data used.

Figure 6 .
Figure 6.Stratigraphic correlation of core BN05 with PS72/392-5 (Stein et al., 2010a) based on sediment lightness, magnetic susceptibility, calcite and dolomite content.See Fig. 3 for other stratigraphic proxies and lithostratigraphy explanation.The vertical magenta bar indicates the position of the foraminiferal peak in B14-15.

Figure 7 .
Figure 7. (a) Biplots of principal component loading scores in PC 1-2 and PC 3-4 space (see Table3for loading data and TableS3for correlation between variables).Sedimentary variable groups revealed by the loading distribution are enclosed by ellipses and numbered, with the closest groupings highlighted in gray.(b) Downcore distribution of sedimentary variable groups plotted using combined PC 1-2 and PC 3-4 scores (see TableS4for score data).

Figure 8 .
Figure 8. Schematic reconstruction of glacial environments in the western Arctic Ocean and factors controlling sedimentation at the BN05 site (yellow circle): surface circulation (red and green arrows), glacioturbidites (orange filled arrow), and relative ice-sheet size (red and green crosses).See Fig. 1 for modern circulation.(a) High ESIS inputs: MIS 4, 6, 12, and 14; (b) high LIS inputs: MIS 8 and 10; (c) especially high LIS inputs: intra-MIS 5 and 3.

Table 1 .
Minerals actively sought in diffraction data analysis.

Table 3 .
Loading scores for variables used in the PCA.