A 17.3 m long sediment core CSH1 (31∘13.7′ N, 128∘43.4′ E; water depth: 703 m) was collected from the northern OT, close to the main stream of Tsushima Warm Current (TWC) (Fig. 1b) and within the depth of the NPIW (Fig. 1c) using a piston corer during Xiangyanghong09 Cruise in 1998, carried out by the First Institute of Oceanography, Ministry of Natural Resources of China. This location is enabling us to reconstruct millennial-scale changes in the properties of the TWC and NPIW. Core CSH1 mainly consists of clayey silt and silt with occurrence of plant debris at some depth intervals (Ge et al., 2007) (Fig. 3a). In addition, three layers of volcanic ash were observed at depths of 74–106, 782–794 and 1570–1602 cm. These three intervals can be correlated with well-known ash layers, Kikai-Akahoya (K-Ah; 7.3 ka), Aira-Tanzawa (AT; 29.24 ka) and Aso-4 (roughly around MIS 5a) (Machida, 1999), respectively. The core was split and subsampled at 4 cm interval and then stored in the China Ocean Sample Repository at 4 ∘C until analysis.

Previously, paleoceanographic studies have been conducted and a set of data has been investigated for core CSH1, including the contents of planktic foraminifers as well as their carbon (δ13C) and oxygen isotope (δ18O) compositions (Shi et al., 2014), pollen (Chen et al., 2006), paleomagnetism (Ge et al., 2007) and CaCO3 (Wu et al., 2004). An age model for this core has been constructed by using 10 Accelerator Mass Spectrometry (AMS) 14C dates and six oxygen isotope (δ18O) age control points. The whole 17.3 m core contains ca. 88 ka long record of continuous sedimentation (Shi et al., 2014).

Notably, the original age model, which used constant radiocarbon reservoir ages throughout core CSH1 are suitable to reveal orbital-scale Kuroshio variations (Shi et al., 2014) but insufficient to investigate millennial-scale climatic events. The occurrence of a higher abundance of Neogloboquadrina pachyderma (dextral), e.g., during warmer intervals, such as the Bölling-Alleröd (B/A), has been challenging to explain. On the other hand, paired measurements of 14C/12C and 230Th ages from Hulu Cave stalagmites suggest magnetic field changes have greatly contributed to high atmospheric 14C/12C values at HS4 and the Younger Dryas (YD) (Cheng et al., 2018). Thus a constant reservoir age (ΔR=0) assumed when calibrating foraminiferal radiocarbon dates using CALIB 6 software and the Marine 13 calibration dataset (Reimer et al., 2013) for core CSH1 may cause large chronological uncertainties.

Here, we recalibrated the radiocarbon dates using updated CALIB 7.04 software with Marine 13 calibration dataset (Reimer et al., 2013). Moreover, on the basis of significant correlation between planktic foraminifera species Globigerinoides ruber δ18O and Chinese stalagmite δ18O (Cheng et al., 2016), a proxy of summer EAM related to SSS of the ECS, we improve the age model for core CSH1 (Fig. 3b–d). Overall, the new chronological framework is similar to the one previously reported by Shi et al. (2014) but with more dates. In order to compare with published results associated with ventilation changes in the North Pacific, here we mainly report the history of sedimentary oxygenation in the northern OT since the last glacial period. Linear sedimentation rate varied between ∼10 and 40 cm ka-1 with higher sedimentation rates (around 30–40 cm ka-1) between ∼24 and 32.5 ka. Variation in sedimentation rate has been attributed to changes in eustatic sea level, summer EAM intensity, path and/or intensity of Kuroshio Current. Generally, sea level is thought to be the first-order factor for controlling linear sedimentation rate changes (Beny et al., 2018; Li et al., 2015; Zhao et al., 2017). The new age control points are shown in Table 2.

Sediment subsamples for geochemical analyses were freeze-dried and ground to a fine powder with an agate mortar and pestle. Based on the age model, 85 subsamples from core CSH1, representing a temporal resolution of about 600 years (every 4 cm interval) were selected for detailed geochemical analyses of major and minor elements, and total carbon (TC), organic carbon (TOC) and nitrogen (TN) contents. The pretreatment of sediment and other analytical methods have been reported elsewhere (Zou et al., 2012).

TC and TN were determined with an elemental analyzer (EA; Vario EL III, Elementar Analysensysteme GmbH) in the Key Laboratory of Marine Sediment and Environment Geology, First Institute of Oceanography, Ministry of Natural Resources of China, Qingdao. Carbonate was removed from sediments by adding 1 M HCl to the homogenized sediments for a total organic carbon (TOC) analysis using the same equipment. The content of calcium carbonate (CaCO3) was calculated using the equation: CaCO3=TC-TOC×8.33, where 8.33 is the ratio between the molecular weight of carbonate and the atomic weight of carbon. A national reference material (GSD-9), a blank sample and replicated samples were used to control the analytical process. The relative standard deviation of the GSD-9 for TC, TN and TOC is ≤3.4 %.

About 0.5 g of sediment powder was digested in double-distilled HF:HNO3 (3:1), followed by concentrated HClO4, and then redissolved in 5 % HNO3. Selected major and minor elements such as aluminum (Al) and manganese (Mn) were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES; Thermo Scientific iCAP 6000, Thermo Fisher Scientific), as detailed elsewhere (Zou et al., 2012). In addition, Mo and U were analyzed with inductively coupled plasma mass spectrometry (ICP-MS; Thermo Scientific XSERIES 2, Thermo Fisher Scientific), as described in Zou et al. (2012). Precision for most elements in the reference material GSD-9 is ≤5 % relative standard deviation. The excess fractions of U and Mo were estimated by normalization to Al as follows:

Excess fraction =totalelement-(element/Alaverageshale×Al), with U/Alaverageshale=0.307×10-6 and Mo/Alaverageshale=0.295×10-6 (Li and Schoonmaker, 2014).

In addition, given the different geochemical behaviors of Mn and Mo and co-precipitation and adsorption processes associated with the redox cycling of Mn, we calculated the ratio of Mo to Mn, assuming that a higher Mo/Mn ratio indicates a lower oxygen content in the depositional environment and vice versa. In combination with the concentration of excess uranium, we infer the history of sedimentary oxygenation in the subtropical North Pacific since the last glaciation.

The content of CaCO3 varies from 8.8 % to 35 % (Fig. 4a), and it mostly shows higher values with increasing trends during the last deglaciation. In contrast, the content of CaCO3 is low and exhibits decreasing trends during the late MIS 3 and the LGM (Fig. 4a). TN content shows a larger variation compared to TOC (Fig. 4b), but it still strongly correlates with TOC (r=0.74, p<0.01) throughout the entire core. Concentration of TOC ranges from 0.5 % to 2.1 %, and it shows higher values with stable trends during the last glacial phase (MIS 3) (Fig. 4c). Molar ratios of TOC/TN vary around 10, with higher ratios at the transition into the LGM (Fig. 4d), corresponding to a higher linear sedimentation rate (Fig. 4e).

Both TOC and CaCO3 have been used as proxies for the reconstruction of past export productivity (Cartapanis et al., 2011; Lembke-Jene et al., 2017; Rühlemann et al., 1999). Molar C/N ratios of >10 (Fig. 4c) suggest that terrigenous organic sources significantly contribute to the TOC concentration in core CSH1. The TOC content therefore may not be a reliable proxy for the reconstruction of surface water export productivity during times of the LGM and late deglaciation, when maxima in C/N ratios co-occur with decoupled trends between CaCO3 and TOC concentrations.

Several lines of evidence support CaCO3 as a reliable productivity proxy, particularly during the last deglaciation. The strong negative correlation coefficient (r=-0.85; p<0.01) between Al and CaCO3 in sediments throughout core CSH1 confirms the biogenic origin of CaCO3 against terrigenous Al (Fig. 4f). Generally, terrigenous dilution decreases concentrations of CaCO3. An inconsistent relationship between CaCO3 contents and sedimentation rates indicates a minor effect of dilution on CaCO3. Furthermore, the increasing trend in CaCO3 associated with a high sedimentation rate during the last deglacial interval indicates a substantial increase in export productivity (Fig. 4a and d). The high coherence between the CaCO3 content and alkenone-derived sea surface water (SST) (Shi et al., 2014) indicates a direct control on CaCO3 by SST. Moreover, a detailed comparison between CaCO3 concentrations and the previously published foraminiferal fragmentation ratio (Wu et al., 2004) shows, apart from a small portion within the LGM, no clear covariation between them. These pieces of evidence suggest that CaCO3 changes are driven primarily by variations in carbonate primary production and are not overprinted by secondary processes, such as carbonate dissolution through changes in the lysocline depth and dilution by terrigenous materials. Likewise, a similar deglacial trend in CaCO3 is also observed in core MD01-2404 (Chang et al., 2009), indicating a ubiquitous, not local picture in the OT. All these lines of evidence thus support CaCO3 of core CSH1 as a reliable productivity proxy to a first-order approximation.

Figure 4 shows time series of selected redox-sensitive elements (RSEs) and proxies derived from them. Mn shows higher concentrations during the LGM and HS1 (16–22 ka) and middle–late Holocene but lower concentrations during the last deglacial and Preboreal periods (15.8–9.5 ka, Fig. 4g). Generally, concentrations of excess Mo and excess U (Fig. 4j and l) show coherent patterns with those of Mo and U (Fig. 4i and k), but both are out of phase with Mn over the last glacial period (Fig. 4h). Pronounced variations in U concentration after 8.5 ka are related to the occurrence of discrete volcanic materials. A significant positive Eu anomaly (Zhu et al., 2015) confirms the occurrence of discrete volcanic materials and its dilution effects on terrigenous components since 7 ka. Occurrence of discrete volcanic material is likely related to an intensified Kuroshio Current during the middle–late Holocene, as supported by higher hydrothermal Hg concentrations in sediments from the middle OT (Lim et al., 2017). A negative correlation between Mn and Moexcess during the last glaciation and the Holocene and the strong positive correlation between them during the LGM and HS1 (Fig. 5a and b) further corroborate the complex geochemical behaviors of Mn and Mo. A strong positive correlation between Moexcess and Mn (Fig. 5b) may be attributed to co-precipitation of Mo by manganese oxyhydroxide under oxygenated conditions. Here, we thus use the Mo/Mn ratio, instead of an excess Mo concentration to reconstruct variations in sedimentary redox conditions in our study area. Overall, the Mo/Mn ratio shows a similar downcore pattern to that of Moexcess with higher ratios during the last deglaciation but lower ratios during the LGM and HS1. A strong correlation (r=0.69) between Mo/Mn ratio and excess U concentration (excluding Holocene data, due to the contamination with volcanic material; Fig. 5c) further corroborates the integrity of Mo/Mn as an indicator of sedimentary oxygenation changes.

Rapidly decreasing Mo/Mn ratios indicate a well-oxygenated sedimentary environment after ∼8 ka (Fig. 4h). Both higher Mo/Mn ratios and excess U concentrations, together with lower Mn concentrations, suggest suboxic depositional conditions during the late deglacial period (15.8–9.5 ka), whereas lower ratios during the LGM, HS1 and HS2 indicate relatively better-oxygenated sedimentary conditions. A decreasing trend in the Mo/Mn ratio and the excess U concentration from 50 to 25 ka also suggest higher sedimentary oxygen levels.

In general, three different terms, hypoxia, suboxia and anoxia, are widely used to describe the degree of oxygen depletion in the marine environment (Hofmann et al., 2011). Here, we adopt the definition of oxygen thresholds by Bianchi et al. (2012) for oxic (>120 µmolkg-1 O2), hypoxic (<60–120 µmolkg-1 O2) and suboxic (<2–10 µmolkg-1 O2) conditions, whereas anoxia is the absence of measurable oxygen.

Proxies associated with RSEs, such as sedimentary Mo concentration (Lyons et al., 2009; Scott et al., 2008) have been used to constrain the degree of oxygenation in seawater. Algeo and Tribovillard (2009) proposed that open-ocean systems with suboxic waters tend to yield Uexcess enrichment relative to Moexcess, resulting in a sediment (Mo/U)excess ratio less than that of seawater (7.5–7.9). Under increasingly reducing and occasionally sulfidic conditions, the accumulation of Moexcess increases relative to that of Uexcess leading the (Mo/U)excess ratio either being equal to or exceeding that of seawater. Furthermore, Scott and Lyons (2012) suggested a non-euxinic condition with the presence of sulfide in pore waters when Mo concentrations range from >2 µgg-1 to <25 µgg-1. Given that the northern OT is located in an open-ocean setting, we use these two proxies to evaluate the degree of oxygenation in sediments.

Both the bulk Mo concentration (1.2–9.5 µgg-1) and excess (Mo/U) ratio (0.2–5.7) in core CSH1 suggest that oxygen-depleted conditions have prevailed in the deep water of the northern OT over the last 50 ka (Fig. 4m). However, increased excess Mo concentrations with higher Mo/U ratios during the last termination (18–9 ka) indicate more reducing conditions compared to the Holocene and the last glacial period, though Mo concentrations were less than 25 µgg-1, a threshold for euxinic deposition proposed by Scott and Lyons (2012).

The relative abundance of benthic foraminifera species that thrive in different oxygen concentrations has also been widely used to reconstruct variations in bottom water ventilation, such as the enhanced abundance of Bulimina aculeata, Uvigerina peregrina and Chilostomella oolina found under oxygen-depleted conditions in the central and southern OT from 18 to 9.2 ka (Jian et al., 1996; Li et al., 2005). An oxygenated bottom water condition is also indicated by abundant benthic foraminifera species Cibicidoides hyalina and Globocassidulina subglobosa after 9.2 ka (Jian et al., 1996; Li et al., 2005) in cores E017 (1826 m water depth) and 255 (1575 m water depth) and high benthic δ13C values (Wahyudi and Minagawa, 1997) in core PN-3 (1058 m water depth) from the middle and southern OT during the postglacial period. The poorly ventilated deep water in the middle and southern OT inferred by benthic foraminiferal assemblages during the last deglaciation correlates with the one in the northern OT referring to our RSEs (Fig. 4). A link thus can be hypothesized between deepwater ventilation and sedimentary oxygenation in the OT. Overall, a combination of our proxy records of RSEs in core CSH1 with other records shows oxygen-rich conditions during the last glaciation and middle and late Holocene (since 8.5 ka) intervals but oxygen-poor conditions during the last deglaciation.

Our observed pattern of RSEs in core CSH1 suggests that drastic changes in sedimentary oxygenation occurred on orbital and millennial timescales over the last glaciation in the OT. In general, four factors can regulate the redox condition in the deep water column and are as follows: (i) O2 solubility, (ii) export productivity and subsequent degradation of organic matter, (iii) vertical mixing and (iv) lateral supply of oxygen through intermediate and deeper water masses (Ivanochko and Pedersen, 2004; Jaccard and Galbraith, 2012). These processes have been invoked in previous studies to explain the deglacial Pacific-wide variations in oxygenation by either one or a combination of these factors (Galbraith and Jaccard, 2015; Moffitt et al., 2015; Praetorius et al., 2015). Our data also suggest drastic variations in sedimentary oxygenation over the last 50 ka. However, the mechanisms responsible for sedimentary oxygenation variations in the basin-wide OT and its connection with ventilation of the open North Pacific remain unclear. In order to place our core results in a wider regional context, we compare our proxy records of sedimentary oxygenation (Uexcess concentration and Mo/Mn ratio) and export productivity (CaCO3) (Fig. 6a–c) with the abundance of Pulleniatina obliquiloculata (an indicator of Kuroshio strength) and sea surface temperature in core CSH1 (Shi et al., 2014), the bulk sedimentary nitrogen isotope (an indicator of denitrification) in core MD01-2404 (Kao et al., 2008), benthic foraminifera δ13C (a proxy for ventilation) in cores PN-3 and PC23A (Rella et al., 2012; Wahyudi and Minagawa, 1997) and the abundance of benthic foraminifera (an indicator of hypoxia) in core E017 (Li et al., 2005) and ODP Site 1017 (Cannariato and Kennett, 1999) (Fig. 6d–k).

One of the characteristic climate features in the Northern Hemisphere, in particular the North Atlantic is millennial-scale oscillation during glacial and deglacial periods. These abrupt climatic events have been widely thought to be closely related to the varying strength of the Atlantic Meridional Overturning Circulation (AMOC) (Lynch-Stieglitz, 2017). One of dynamic proxies of ocean circulation, 231Pa/230Th reveals that severe weakening of the AMOC only existed during Heinrich stadials due to increased freshwater discharges into the North Atlantic (Böhm et al., 2015; McManus et al., 2004). On the other hand, several mechanisms, such as a sudden termination of freshwater input (Liu et al., 2009), an increase in atmospheric CO2 concentration (Zhang et al., 2017), an enhanced advection of salt (Barker et al., 2010) and changes in background climate (Knorr and Lohmann, 2007), were proposed to explain the reinvigoration of the AMOC during the B/A.

Our RSEs data in the Northern OT and endobenthic δ13C in the Bering Sea (Fig. 7a–c) both show a substantial millennial variability in intermediate-water ventilation in the subtropical North Pacific. Notably, enhanced ventilation during HS1 and HS2 and oxygen-poor conditions during the B/A, respectively, correspond to the collapse and resumption of the AMOC (Fig. 7d). Such an out-of-phase millennial-scale pattern is consistent with the results of various modeling simulations (Chikamoto et al., 2012; Menviel et al., 2014; Okazaki et al., 2010; Saenko et al., 2004), although these models had different boundary conditions and causes for the observed effects in GNPIW formation and ventilation ages derived from B–P 14C (Freeman et al., 2015; Max et al., 2014; Okazaki et al., 2012). These lines of evidence confirm a persistent link between the ventilation of the North Pacific and North Atlantic climate (Lohmann et al., 2019). Such links have also been corroborated by proxy data and modeling experiment between the AMOC and East Asian monsoon during the 8.2 ka event (Liu et al., 2013), the Holocene (Wang et al., 2005) and 34–60 ka (Sun et al., 2012). The mechanism linking East Asia with the North Atlantic has been attributed to an atmospheric teleconnection, such as the position and strength of the westerly jet and Mongolia-Siberian High (Porter and Zhisheng, 1995). However, the mechanism behind such an out-of-phase pattern between the ventilation in the subtropical North Pacific and the North Atlantic deep water formation remains unclear.

Increased NPIW formation during HS1 may have been caused by enhanced salinity-driven vertical mixing through higher meridional water mass transport from the subtropical Pacific. Previous studies have proposed that intermediate-water formation in the North Pacific hinged on a basin-wide increase in sea surface salinity driven by changes in strength of the summer EAM and the moisture transport from the Atlantic to the Pacific (Emile-Geay et al., 2003). Several modeling studies found that freshwater forcing in the North Atlantic could cause a widespread surface salinification in the subtropical Pacific Ocean (Menviel et al., 2014; Okazaki et al., 2010; Saenko et al., 2004). This idea has been tested by proxy data (Rodríguez-Sanz et al., 2013; Sagawa and Ikehara, 2008), which indicated a weakened summer EAM and reduced transport of moisture from Atlantic to Pacific through the Isthmus of Panama owing to the southward displacement of the Intertropical Convergence Zone caused by a weakening of the AMOC. Along with this process, as predicted through a general circulation modeling, a strengthened Pacific Meridional Overturning Circulation would have transported more warm and salty subtropical water into the high-latitude North Pacific (Okazaki et al., 2010). In accordance with comprehensive Mg/Ca ratio-based salinity reconstructions, however, Riethdorf et al. (2013) found no clear evidence for such higher-salinity patterns in the subarctic northwest Pacific during HS1.

On the other hand, a weakened AMOC would deepen the wintertime Aleutian Low based on modern observation (Okumura et al., 2009), which is closely related to the sea ice formation in the marginal seas of the subarctic Pacific (Cavalieri and Parkinson, 1987). Once stronger Aleutian Low, intense brine rejection due to sea ice expansion, would have enhanced the NPIW formation. Recently modeling-derived evidence confirmed that enhanced sea ice coverage occurred in the southern Okhotsk Sea and off the east Kamchatka Peninsula during HS1 (Gong et al., 2019). In addition, stronger advection of low-salinity water via the Alaskan Stream to the subarctic NW Pacific was probably enhanced during HS1, related to a shift in the Aleutian Low pressure system over the North Pacific, which could also increase sea ice formation, brine rejection and thereafter intermediate-water ventilation (Riethdorf et al., 2013).

During the late deglaciation, ameliorating global climate conditions, such as a warming Northern Hemisphere and a strengthened East Asian summer monsoon, are a result of changes in insolation forcing, greenhouse gases concentrations and variable strengths of the AMOC (Clark et al., 2012; Liu et al., 2009). During the B/A, a decrease in sea ice extent and duration was indicated by combined reconstructions of SST and mixed layer temperatures from the subarctic Pacific (Riethdorf et al., 2013). At that time, the rising eustatic sea level (Spratt and Lisiecki, 2016) would have supported the intrusion of the Alaska Stream into the Bering Sea by deepening and opening glacially closed straits of the Aleutian Islands chain, while reducing the advection of the Alaska Stream to the subarctic Pacific Gyre (Riethdorf et al., 2013). In this scenario, saltier and more stratified surface water conditions would have inhibited brine rejection and subsequent formation and ventilation of the NPIW (Lam et al., 2013), leading to a reorganization of the Pacific water mass, closely coupled to the collapse and resumption modes of the AMOC during these two intervals.

One of the striking features of RSEs data is the presence of higher Mo/Mn ratios and excess U concentrations across the B/A, supporting an expansion of the oxygen minimum zone in the North Pacific (Galbraith and Jaccard, 2015; Jaccard and Galbraith, 2012; Moffitt et al., 2015) and coinciding with the termination of atmospheric CO2 concentration rise (Marcott et al., 2014) (Fig. 7e). As described above, it can be related to the upwelling of nutrient- and CO2-rich Pacific Deep Water due to resumption of the AMOC and enhanced export production. Notably, boron isotope data measured on surface-dwelling foraminifera in core MD01-2416 situated in the western subarctic North Pacific did reveal a decrease in near-surface pH and an increase in pCO2 at the onset of B/A (Gray et al., 2018), indicating that the subarctic North Pacific was a source of the relatively high atmospheric CO2 concentration at that time. Here we cannot conclude that the same processes could have occurred in the subtropical North Pacific due to the lack of well-known drivers to draw out of the old carbon from the deep sea into the atmosphere. In combination with published records from the North Pacific (Addison et al., 2012; Cartapanis et al., 2011; Crusius et al., 2004; Galbraith et al., 2007; Lembke-Jene et al., 2017; Shibahara et al., 2007), an expansion of the oxygen-depleted zone during the B/A suggests an increase in respired carbon storage at mid-depth waters of the North Pacific, which likely stalls the rise of atmospheric CO2. Our results support the findings by Galbraith et al. (2007). Given the sizable volume of the North Pacific, potentially, in the past the respired carbon could have been emitted to the atmosphere in stages, which would have brought the planet out of the last ice age (Jaccard and Galbraith, 2018).