The East Sea is a marginal sea surrounded by the east Asian continent and Japanese islands (Fig. 1; Kang et al., 2010; Liu et al., 2010). The UB located in the southwestern part of the East Sea is a bowl-shaped deep basin (2000–3000 m depth) (Fig. 1) delimited by continental slopes of Korean Peninsula and the southwestern Japanese archipelago on the west and south, respectively, and by the Korea Plateau and the Oki Bank on the north and east, respectively (Chough et al., 2000).

Shipboard experiments were conducted in June 2009 at two sites on the continental slope (station M1, hereafter M1) and in the center (station D3, hereafter D3) of the UB (Fig. 1, Table 1). Surface sediments consist of fine-grained clay with a mean grain size less than 0.004 mm in diameter (Cha et al., 2007). Two stations were characterized by two contrasting sediment colors. The Mn oxide-enriched surface sediment at the basin site (D3) was reddish-brown, whereas at the slope site (M1) it exhibited the typical gray-brown color of muddy continental margin sediments (Fig. 1). Further environmental properties are listed in Table 1.

Sediment samples were collected with a box corer. Onboard, duplicate or mostly triplicate sub-samples for geochemical analyses were collected using acrylic cores (6–9 cm in diameter and 30–40 cm in length). The sub-cores for geochemical analyses were immediately sealed with butyl rubber stoppers and transferred to a N2-filled glove bag for sectioning and loading into polypropylene centrifuge tubes that were then tightly capped and centrifuged for 15 min at 5000 × g. After reintroduction into the N2-filled glove bag, pore waters were sampled and filtered through 0.2 µm cellulose ester syringe filters (ADVANTEC, Toyo Rashi Kaisha, Ltd). One to 2 mL of pore water to determine NH4+ was fixed with saturated HgCl2 and subsequently frozen. For determination of Fe2+, Mn, SO42- and Ca2+, 2 mL of the pore water were acidified with 12 M HCl and stored at 4 ∘C. Pore water for sulfide analysis was preserved with Zn acetate (20 %). Sediments for solid-phase analysis were frozen at -25 ∘C for future analyses.

Anaerobic carbon mineralization rates and dissimilatory Mn and Fe reduction rates were determined in batch incubations based on the procedures of Canfield et al. (1993b) and Thamdrup and Canfield (1996). Sediment cores were transferred to a N2-filled glove bag and sliced in 2 cm intervals to a depth of 10 cm. Sediment from parallel sections was pooled, mixed, and loaded into gas-tight plastic bags (Hansen et al., 2000). The bags were sealed without gas space and incubated in the dark at near in situ temperature (ca. 1–2 ∘C) in larger N2-filled bags to ensure anoxic conditions. Over a period of 18 days of incubation, sub-samples to determine the accumulation of total dissolved inorganic carbon (DIC) and Mn in pore water were withdrawn on days 0, 1, 3, 5, 9, and 18. Two 50 mL centrifuge tubes per bag were filled completely with sediment in a N2-filled glove bag, and pore water was extracted as described above. For DIC analysis, we collected 1.8 mL aliquots into glass vials without head space, fixed with 18 µL of HgCl2 (125 mM), and stored at 4 ∘C until analysis within 4 weeks. Samples for Mn analysis were acidified with 12 M HCl and stored at 4 ∘C. Sediment remaining after the collection of pore water was frozen at -25 ∘C for later analysis of oxalate-extractable solid Fe(II).

Total dissolved inorganic carbon (DIC) and NH4+ were measured by flow injection analysis with conductivity detection (Hall and Aller, 1992). Nitrate was measured spectrophotometrically (Parsons et al., 1984). Dissolved Fe2+ was determined by colorimetric method with ferrozine (Stookey, 1970). Dissolved Mn2+ and Ca2+ were analyzed in acidified pore water by inductively coupled plasma–atomic emission spectrometry (ICP-AES, Optima 3300DV, PerkinElmer Co.) and flame atomic absorption spectrometer (SpectrAA 220/FS, Varian), respectively (Thamdrup and Canfield, 1996). Dissolved sulfide was determined by the methylene blue method (Cline, 1969). Sulfate concentrations were measured using ion chromatography (Metrohm 761). The detection limit of H2S, Ca2+, Mn2+, and Fe2+ was 3, 1.8, 3, and 1 µM, respectively. Reproducibility of DIC and NH4+ was better than 10 %. Precision of NO3- was 1–2 %.

Total oxalate-extractable Fe [Fe(II) + Fe(III)] was extracted from air-dried sediment in a 0.2 M oxic oxalate solution (pH 3) for 4 h (Thamdrup and Canfield, 1996), and Fe(II) was extracted from frozen sediment in anoxic oxalate (Phillips and Lovley, 1987). The total oxalate-extractable Fe and Fe(II), hereafter total Fe(oxal) and Fe(II)(oxal), were determined as described for the pore-water analysis of Fe2+. Oxalate-extractable Fe(III), hereafter Fe(III)(oxal), was defined as the difference between total Fe(oxal) and Fe(II)(oxal). This fraction represents poorly crystalline Fe(III) oxides. Particulate Mn, hereafter Mn(DCA) was extracted with dithionite–citrate–acetic acid (DCA; pH 4.8) for 4 h from air-dried sediment and was determined by inductively coupled plasma–atomic emission spectrometry (ICP-AES, Optima 3300DV, PerkinElmer Co). The DCA extraction aims at dissolving free Mn oxides and authigenic Mn(II) phases. The reproducibility of the measurements was better than 10 % and the detection limits were 3 µM for Mn and 1 µM for Fe. For the determination of total reduced sulfur (TRS) that includes acid-volatile sulfide (AVS = FeS + H2S and small amounts of other metal sulfides; see Rickard and Morse, 2005; Luther III, 2005) and chromium-reducible sulfur (CRS = S0 + FeS2), sediment samples were fixed with Zn acetate, and sulfide was determined according to the method of Cline (1969) after a two-step distillation with cold 12 M HCl and boiling 0.5 M Cr2+ solution (Fossing and Jørgensen, 1989). The contents of total organic carbon (TOC) and nitrogen (TN) in the surface sediment were analyzed using a CHN analyzer (CE Instruments, EA 1110) after removing CaCO3 using 12 M HCl.

Oxygen profiles were measured at 50 µm resolution using Clark-type microelectrodes (Unisense, OX-50) while stirring the overlying water. Microelectrodes were calibrated between 100 % air-saturated in situ bottom water and N2-purged anoxic bottom water. Three profiles were measured at each site. The diffusive boundary layer (DBL) and sediment–water interface (SWI) were determined according to Jørgensen and Revsbech (1985). To estimate the volume-specific oxygen consumption rate, we used the PROFILE software (Berg et al., 1998).

The diffusive oxygen uptake (DOU) was calculated from the calibrated oxygen micro-profiles. DOU=-Do×(ΔC/Δz), where Do (1.07×10-9 m2 s-1 at M1 and 1.03×10-9 m2 s-1 at D3) is the temperature-corrected molecular diffusion coefficient estimated from Schulz (2006), and C is the oxygen concentration at depth z within the diffusive boundary layer (DBL) (Jørgensen and Revsbech, 1985).

The volume-specific O2 consumption rates exhibited a bimodal depth distribution (see Sect. 3.2) with activity peaks near the SWI and the oxic–anoxic interface, respectively. Thus, O2 consumption rates by aerobic organotrophic respiration were defined as the O2 consumption rate near the SWI, whereas the oxygen consumption at the oxic–anoxic interface was assigned to re-oxidation of reduced inorganic compounds (Rasmussen and Jørgensen, 1992; Canfield et al., 2005).

Total anaerobic Corg mineralization rates were determined by linear regression of the accumulation of total DIC with time during the anoxic bag incubations (Fig. 3) after correcting for CaCO3 precipitation (Thamdrup et al., 2000). Briefly, CaCO3 precipitation was calculated from decreasing dissolved Ca2+ concentration during the anoxic bag incubation: ΔCaCO3=Δ[Ca2+]sol×(1+KCa), where KCa is the adsorption constant for Ca2+ (KCa=1.6) (Li and Gregory, 1974). Then the DIC production rate corrected for CaCO3 precipitation was calculated as DIC production=DIC accumulation+CaCO3precipitation. Fe(III) reduction rates were determined by linear regression of the increase in solid-phase Fe(II)(oxal) content with time during anoxic bag incubations. The dissimilatory microbial Fe(III) reduction rate was derived by subtracting abiotic Fe reduction coupled to the oxidation of sulfide produced by sulfate reduction (Gribsholt et al., 2003): Dissimilatory microbial Fe(III) red.=total Fe(III) red.-abiotic Fe(III) red., assuming that abiotic Fe reduction coupled to H2S oxidation occurred at a stoichiometry of 2 Fe(III) per 3 H2S (Pyzik and Sommer, 1981; Melton et al., 2014): 2FeOOH+3H2S(producedbySR)=2FeS+S0+4H2O. Finally, to estimate the Corg oxidation by microbial Fe reduction, the 4 : 1 stoichiometry of iron reduction coupled to Corg oxidation was used from the stoichiometric equation (Canfield et al., 1993a): CH2O+4FeOOH+8H+=CO2+4Fe2++7H2O.

Mn reduction rates were determined from linear regression of the production of dissolved Mn2+ with time during the anoxic bag incubations. Similar to previous studies (e.g., Canfield et al., 1993a, b; Thamdrup and Dalsgaard, 2000), we assumed that accumulating dissolved Mn was Mn2+. This ignores a potential contribution from Mn3+, which in some cases can constitute a substantial fraction of the dissolved Mn pool at the upper boundary of the zone with soluble Mn accumulation in marine sediments (Madison et al., 2013). Further studies of the dynamics of soluble Mn3+ are required to evaluate its potential importance in anoxic incubations. Such studies pending, we find justification for our assumption in the good agreement observed in the previous studies between Mn reduction rates calculated based on the assumption that soluble Mn is Mn2+ (Eq. 7) and independent estimates of rates of carbon mineralization through dissimilatory Mn reduction based on DIC or NH4+ accumulation. Due to strong adsorption of Mn2+ to Mn oxide surfaces (Canfield et al., 1993b), the Mn reduction rates were estimated after compensating for the adsorption effect of Mn2+ to Mn oxides according to Thamdrup and Dalsgaard (2000): Mn reduction rate=Mn2+accumulation rate×(1+K∗Mn2+×(1-Φ)×Φ-1×δs), where Φ is porosity, δs is density of sediment, K∗Mn2+=4.8+0.14×[Mn(IV)] (mL g-1), and [Mn(IV)] is the content of Mn(IV) (µmol g-1).

We here assume that extracted Mn(DCA) represents Mn(IV) as observed in surface sediments of another Mn-rich site (Canfield et al., 1993b; Thamdrup and Dalsgaard, 2000). Small levels of Mn(DCA) remaining at depth further suggest that little Mn(II) accumulates in the solid phase (see Results). Corg oxidation by dissimilatory Mn(IV) reduction was calculated from the stoichiometric equation (Canfield et al., 1993a): CH2O+2MnO2+4H+=CO2+2Mn2++3H2O. Sulfate reduction rates were determined using the radiotracer method of Jørgensen (1978). Sediment cores (35 cm long with 2.9 cm i.d.) were collected in triplicate, injected horizontally at 1 cm vertical interval with 5 µL radiolabeled sulfate (35S-SO42-, 15 kBq µL-1, Amersham), diluted in sterilized NaCl solution (3.0 %), and incubated for 12 h at in situ temperature. At the end of the incubation, the sediment was sliced into sections, fixed in Zn acetate (20 %), and frozen at -25 ∘C until processed in the laboratory. The reduced 35S was recovered using distillation with a boiling acidic Cr2+ solution according to Fossing and Jørgensen (1989). Background radioactivity of 35S was 32.4 ± 3.7 cpm cm-3 (n=10) at site D3 and 87.5 ± 38.7 cpm cm-3 (n=10) at site M1. Detection limits of the sulfate reduction rate (SRR), estimated from the double standard deviation of the blank value (i.e., 7.4 and 77.4 cpm) according to Fossing et al. (2000), ranged from 0.79 to 2.62 nmol cm-3 d-1. To elucidate the contribution of sulfate reduction in anaerobic carbon oxidation, the SRRs (Fig. 5b, g) were converted to carbon oxidation using a stoichiometric equation (Thamdrup and Canfield, 1996): 2CH2O+SO42-+2H+=2CO2+H2S+2H2O.

The depth distributions of NH4+, NO3-, Mn2+, and Fe2+ in the pore water as well as solid-phase Mn, Fe, and S for the two stations are shown in Fig. 2. NH4+ concentrations at M1 increased steadily with depth (Fig. 2a), whereas at D3 it decreased down to 3 cm depth before it increased below (Fig. 2f). Highest concentrations of nitrate were measured at 0 to 1 cm sediment depth at the two stations and concentrations decreased below a background level (< 2 µM) below 1 cm at both M1 and D3 (Fig. 2a, f). Dissolved Mn2+ concentrations differed widely between the sites showing a maximum of 56 µM between 0 and 3 cm depth and not exceeding 10 µM below at M1 (Fig. 2b), whereas at D3 concentrations increased to a maximum of 286 µM at 10–12 cm depth (Fig. 2g). Conversely, dissolved Fe2+ concentrations at M1 increased from 11 µM at 0–0.5 cm to 32 µM at 6–7 cm depth, and stayed constant below (Fig. 2c), whereas at D3, concentrations were uniformly low, showing a slight increase to 12 µM at 15 cm (Fig. 2h).

Extractable Mn (Mn(DCA)) contents were low (< 3 µmol cm-3) in the upper 20 cm at the slope site (M1) (Fig. 2b), but up to 200 µmol cm-3 in the upper 4 cm depth of the sediment at the center of the basin (D3), with a sharp decrease to near depletion (∼ 1 µmol cm-3) below 10 cm (Fig. 2g). At the slope site (M1), contents of Fe(III)(oxal) decreased slightly with increasing depth from 28 µmol cm-3 near the surface to 13 µmol cm-3 at 20 cm depth, mirroring an increase in Fe(II)(oxal) (Fig. 2d). At the center of the basin (D3), Fe(III)(oxal) increased slightly from 67 µmol cm-3 at 0–0.5 cm to 90 µmol cm-3 at 4–6 cm depth, and it decreased steeply below to 4.8 µmol cm-3 at 12–14 cm depth (Fig. 2i). Of total Fe(oxal), Fe(III)(oxal) comprised > 98 % at 0–2 cm and > 97 % at 0–8 cm depth at M1 and D3, respectively. The fraction of Fe(III)(oxal) in Fe(oxal) then decreased to 40 % at 10–12 cm depth at both sites. Acid-volatile sulfur (AVS) exhibited a slight increase with depth at M1 from 0.8 µmol cm-3 at the surface to 7.2 µmol cm-3 at 20 cm depth (Fig. 2e) but was not detected at D3 (Fig. 2j). Chromium-reducible sulfur (CRS) contents at M1 increased rapidly with depth from 1.9 µmol cm-3 at 0–0.5 cm to 21.8 µmol cm-3 at 20 cm depth (Fig. 2e), whereas the CRS contents remained < 0.1 µmol cm-3 at D3 (Fig. 2j).

Oxygen penetrated less than 4 mm into the sediments (Fig. 3), and rates of diffusive oxygen uptake (DOU) were 7.1 and 6.0 mmol O2 m-2 d-1 at M1 and D3, respectively (Table 2). Oxygen consumption by aerobic respiration estimated from the O2 micro-profiles (area I and II in Fig. 3) was higher at the M1 in the slope site (4.0 mmol O2 m-2 d-1) than at the D3 in the center of the basin (2.5 mmol O2 m-2 d-1). O2 consumption by re-oxidation of reduced inorganic compounds indicated by increased activity at the oxic–anoxic interface (area III in Fig. 3) accounted for 43 and 57 % of the DOU at M1 and D3, respectively. From the profiles of geochemical constituents (Fig. 2), O2 consumption was mainly attributed to the re-oxidation of sulfide and Fe2+ at M1 and of Mn2+ at D3.

Changes in concentrations of DIC, Ca2+, and dissolved Mn2+ and solid Fe(II)(oxal) contents over time during anoxic bag incubations from sediment of 0–2, 2–4, 4–6, and 6–8 cm depth intervals are presented in Fig. 4. The DIC concentrations increased linearly over time during incubations of sediment in all bags from M1 and D3, except the bag from 6–8 cm at D3. The DIC accumulation rates were generally higher at the slope site (M1) than at the basin site (D3) (Table 4). The concentrations of Ca2+ decreased with time at all depth intervals of M1, whereas a decrease in Ca2+ was observed only for the 2–4 cm depth interval at D3. The decrease in Ca2+ indicates CaCO3 precipitation, which consequently underestimates DIC accumulation, especially at M1.

Coinciding with high solid Mn(DCA) contents (Fig. 2g), prominent Mn2+ accumulation appeared at 0–6 cm depth of D3, whereas no increase in Mn2+ was observed at M1 except a slight accumulation at 0–2 cm interval (Fig. 4). Solid Fe(II)(oxal) contents increased linearly with time at 0–4 cm depth of M1, whereas highest Fe(II)(oxal) accumulation was observed at 4–6 cm depth at D3. An increase in Fe(II)(oxal) was not discernible in the Mn oxide-rich surface sediment (0–2 cm) of D3.

At the slope site (M1), SRR increased from 18 nmol cm-3 d-1 at the surface to 97–103 nmol cm-3 d-1 at 1.5–2 cm depth, and decreased below to 12.5 nmol cm-3 d-1 at 20 cm depth (Fig. 5b). In contrast, SRR at the manganese oxide-rich basin site (D3) ranged from 1.7 to 8.7 nmol cm-3 d-1, and did not vary with depth (Fig. 5g). Depth-integrated SRR down to 10 cm depth was 10 times higher at M1 (4.3 mmol m-2 d-1) than at D3 (0.4 mmol m-2 d-1) (Table 3).

Vertical profiles of the DIC production rate that were derived from the linear regression of the DIC production measured in anoxic bag incubation (Fig. 4) after correcting for CaCO3 precipitation are presented in Fig. 5c and h for M1 and D3, respectively. At M1, the DIC production rates decreased with depth from 280 nmol cm-3 d-1 (0–2 cm depth) to 69 nmol cm-3 d-1 (8–10 cm depth) (Fig. 5c), whereas the DIC production rates at D3 were relatively similar across the upper 6 cm ranging from 86 to 136 nmol cm-3 d-1, and decreased to 8–15 nmol cm-3 d-1 at 6–10 cm (Fig. 5h). The integrated DIC production rate within 10 cm depth of the sediment was twice as high at M1 (14.0 mmol m-2 d-1) as at the D3 (7.2 mmol m-2 d-1) (Table 4).

The accumulation of Mn2+ presented evidence that manganese reduction was occurring in the surface sediment (0–6 cm) of D3 (Fig. 4). The manganese reduction rate (MnRR) derived from Mn2+ accumulation with correction for adsorption ranged from 7.5 nmol cm-3 d-1 (0–2 cm depth) to 198 nmol cm-3 d-1 (2–4 cm depth) at D3 (Fig. 5i). In contrast, MnRR at M1 was indiscernible except for low activity (2.2 nmol cm-3 d-1) at 0–2 cm depth (Fig. 5d). Depth-integrated MnRR at D3 (8.21 mmol m-2 d-1) was 200 times higher than the MnRR at M1 (0.04 mmol m-2 d-1) (Table 3). The iron reduction rate (FeRR), derived from Fe(II)(oxal) accumulation, at M1 was highest in the 0–2 cm interval (237 nmol cm-3 d-1), and then decreased with depth to 38 nmol cm-3 d-1 at 8–10 cm depth (Fig. 5e). In contrast, Fe reduction was not detected in the surface sediment at D3, but increased to its maximum rate of 240 nmol cm-3 d-1 at 4–6 cm depth. The FeRR then decreased with depth to 12 nmol cm-3 d-1 at 8–10 cm (Fig. 5j), where a few data points were adopted to derive the line of best-fit regression. Depth-integrated total FeRR was slightly higher at M1 (11.4 mmol m-2 d-1) than at D3 (7.53 mmol m-2 d-1) (Table 3). The ratio of microbial Fe reduction, Fe red.(microbial), to abiotic Fe reduction coupled to sulfide oxidation, Fe red.(abiotic), ranged from 1.14 (8–10 cm at M1) to 52.3 (2–4 cm at D3), which indicated that the Fe reduction at Mn- and Fe oxide-rich basin site was mostly a microbiologically mediated process (Table 3).

One of the most prominent features revealed from the vertical distributions of geochemical constituents at the basin site (D3) was that electron acceptors such as O2, nitrate, and Mn and Fe oxides were systematically distributed with discrete zonation according to the order of decreasing energy yield for Corg oxidation (Fig. 5f). Such biogeochemical zones are not sharply separated in most aquatic sediments due to, for example, sediment heterogeneity and mixing resulting from bioirrigation, bioturbation, and bottom turbidity currents. The profiles of dissolved and solid-phase geochemical constituents in the sediment provide indications as to specific diagenetic reactions prevailing (Froelich et al., 1979). However, reoxidation of reduced inorganic compounds often mask the primary reactions involved in carbon oxidation (Sørensen and Jørgensen, 1987; Hines et al., 1991). Together with the discrete geochemical zonation of the electron acceptors, the independently executed metabolic rate measurements (Fig. 5) allowed us to evaluate the relative contribution of each terminal electron-accepting pathway with sediment depth.

Previous experimental studies that have quantified pathways of anaerobic carbon oxidation in subtidal marine sediments have generally determined the contributions of Mn and Fe reduction indirectly from the difference between rates of DIC production and sulfate reduction converted to carbon equivalents (e.g., Canfield et al., 1993b; Thamdrup and Canfield, 1996; Vandieken et al., 2006). The inferred rates of Mn and Fe reduction were further supported by the depth distribution of metal oxides and patterns of Mn2+ and Fe2+ accumulation in the pore water, but this could not be verified because the accumulation of particulate Mn(II) and Fe(II) – which represents the overwhelming fraction of the reduced pools – was not quantified. Here, we combined the indirect approach with independent determination of Mn and Fe reduction rates. Thus, we obtained two separate estimates of anaerobic carbon oxidation rates: based on DIC production and on the sum of sulfate, Fe, and Mn reduction converted to carbon equivalents, respectively (Table 4). At M1, within the 0–10 cm depth interval, the average ratio between total anaerobic Corg oxidation rate (10.7 mmol C m-2 d-1) and the Corg oxidation from DIC production (14.0 mmol C m-2 d-1) was 0.77 (Table 4). Similarly, at D3, the average ratio between total anaerobic Corg oxidation (6.79 mmol m-2 d-1) and anaerobic DIC production (7.22 mmol m-2 d-1) was 0.94. There was a good agreement between the two estimates with a ratio of total anaerobic Corg oxidation by Mn + Fe + sulfate: DIC production for individual depth intervals of 0.8–1.2 (Table 4) with the exception at the 0–2 cm depth of the slope site (M1), where the ratio was slightly lower, 0.66, possibly due to a contribution from the Corg oxidation by nitrate reduction. The similarity of the two estimates across all incubations spanning a range of redox conditions provides confidence in our approach for calculating dissimilatory Mn and Fe reduction rates. Specifically, the good agreement indicates that the underlying assumptions concerning Mn adsorption and reactions of Fe(III) and sulfide are valid as first-order approximations. The general agreement further supports the validity of previous determinations of dissimilatory Mn and Fe reduction rates based on the difference between DIC production and SO42- reduction (Canfield et al., 1993a, b; Thamdrup et al., 2000; Vandieken et al., 2006, 2014).

To elucidate the contribution of sulfate reduction in anaerobic carbon oxidation, the SRRs (Fig. 5b, g) were converted to carbon oxidation (Thamdrup and Canfield, 1996) and then compared to the DIC production rates from anoxic bag incubation (Fig. 5c, h). At the slope site (M1), the fraction of anaerobic Corg oxidation coupled to sulfate reduction increased with depth from 48 % at 0–2 cm to 80 % at 8–10 cm (Table 4). Thus, the excess Corg oxidation in the upper layers should be coupled to other electron-accepting processes. Indeed, the Corg oxidation by Fe reduction (0.96 mmol m-2 d-1) accounted for most of the remaining anaerobic Corg oxidation (12–18 % of DIC production) at 0–8 cm depth, consistent with the distribution of Fe(III) decreasing from > 25 µmol cm-3 near the surface (Fig. 6, Table 4). Mn reduction was of minor importance at M1 because of the low content of Mn oxide (< 3 µmol cm-3). Carbon oxidation coupled to aerobic respiration was estimated to 3.1 mmol m-2 d-1, corresponding to 18 % of the total aerobic + anaerobic oxidation, while the contributions of Fe and sulfate reduction to this total were 12 and 50 %, respectively (Table 4). As mentioned above, nitrate reduction/denitrification may contribute part of the unexplained 19 % of carbon oxidation, but most of this imbalance likely reflects the combined uncertainties in the estimates of the individual pathways. Additionally, our partitioning of carbon oxidation pathways could be biased towards the anaerobic electron acceptors due to the use of the diffusive oxygen uptake (DOU) rather than total oxygen uptake (TOU), which will exceed DOU if bioirrigation is active (Glud, 2008). Bioirrigation was not determined at our sites, but the pore water profiles show no indication of strong irrigation (Fig. 2). An average DOU / TOU ratio of ∼ 0.6 has been reported for sediments at 1.5–2.5 km depth (Glud, 2008). Using this ratio, and assuming that TOU is partitioned similarly as DOU between aerobic carbon oxidation and reoxidation, aerobic carbon oxidation would account for 25 %, while Fe and sulfate reduction would account for 11 and 46 % of carbon oxidation, respectively. Thus, the potential bias from using DOU is not expected to affect the ranking of electron acceptors by quantitative importance (SO42- > O2 > Fe(III)), and, as discussed further below, the partitioning of Corg oxidation at M1 falls within the range previously reported for continental margin sediments.

In contrast to M1, Corg oxidation by sulfate reduction at the basin site (D3) accounted for only a small fraction (< 11 %) of anaerobic Corg oxidation at 0–6 cm interval and it only dominated carbon oxidation at 8–10 cm (Fig. 5h, Table 4). Oxygen and NO3- were depleted within 3.6 mm and 1 cm depth of the sediment surface, respectively (Fig. 5f), while Mn and Fe(III) oxides were abundant at 0–4 and 0–6 cm, respectively. Consistent with the abundance of electron acceptors, high rates of Mn and Fe reduction (Fig. 5i and j) implied Mn and Fe reduction as the most significant Corg oxidation pathways to 6 cm depth. At 0–2 cm depth, Corg oxidation by aerobic respiration and Mn reduction accounted for 53 and 43 % of total Corg oxidation, respectively (Fig. 6). At 2–4 cm, Mn reduction accounted for 73 % of total Corg oxidation and 92 % of anaerobic Corg oxidation (Table 4, Fig. 6). Its importance decreased to 22 % at 4–6 cm due to lower Mn contents, while microbial Fe(III) reduction here contributed 51 %, and the partitioning of sulfate reduction increased to 11 % (Fig. 6). Consequently, the relative distribution of each Corg oxidation pathway with depth at D3 (Fig. 6) matched well with the depth distribution of the respective electron acceptors (Fig. 5f). Overall, within the 10 cm depth sediment interval, Mn reduction and Fe reduction were the dominant Corg oxidation pathways comprising 45 and 20 % of total carbon oxidation, respectively, at the Mn and Fe oxide-rich site in the center of the UB (Table 4). Correction for a potential underestimation of TOU, as discussed for M1, would reduce the contributions of Mn and Fe reduction slightly to 41 and 18 %, respectively.

Despite the high Fe oxide content at 0–4 cm at D3 (Fig. 5f), no solid Fe(II)(oxal) accumulation was observed at this depth range (Fig. 4). This indicates that Fe(III) reduction may not occur under these Mn oxide-rich conditions. Indeed, using acombination of 16S rRNA-stable isotope probing and geochemical analysis in three manganese oxide-rich sediments including the UB, Vandieken et al. (2012) identified bacteria related to Colwellia, Oceanospillaceae and Arcobacter as acetate-oxidizing bacteria that potentially reduce manganese, whereas no known iron reducers were detected in the Mn-rich sediment. Similarly, Thamdrup et al. (2000) found, in Mn oxide-rich Black Sea sediment, that the abundance of viable Fe-reducing bacteria in most probable number counts was low in comparison to Mn reducers and the addition of ferrihydrite did not stimulate Fe reduction, which implied that Fe reduction should be outcompeted by the Mn reduction process.

As manganese reduction is thermodynamically more favorable than iron and sulfate reduction, the Mn2+ liberation (Fig. 4) likely resulted from dissimilatory Mn reduction. Nonetheless, Mn reduction estimated from the increase in Mn2+ at 0–4 cm interval at D3 (Fig. 4) could be due to oxidation of Fe2+ or sulfide. Fe2+ may readily react with Mn oxides (Meyers and Nealson, 1988; Lovley and Phillips, 1988) by the reaction 2Fe2+ + MnO2 + 4H2O = Mn2+ + 2Fe(OH)3 + 2H+. However, in the Mn oxide-rich sediment of the Skagerrak, Canfield et al. (1993b) revealed that the addition of Ferrozine, a strong complexation agent for Fe2+, had no inhibitory effect on the Mn2+ liberation, indicating that the chemical reaction of MnO2 with Fe2+ generated by Fe reduction was not responsible for the accumulation of Mn2+.

Despite the anoxic conditions and nitrate depletion during the bag incubation, Mn reduction rates at 0–2 cm depth (Fig. 5i) based on Mn2+ accumulation were substantially lower than the rates inferred from DIC accumulation (Fig. 5h). A similar discrepancy was previously observed for the uppermost part of the Mn reduction zone (Thamdrup et al., 2000) and is likely explained by particularly strong sorption of Mn2+ to fresh Mn oxide surfaces, which is not included in the adsorption coefficient used here. Low Mn2+ together with the rapid decrease in nitrate at 0–2 cm depth at D3 (Fig. 2f, g) also suggested that dissolved reduced manganese might act as a reducing agent for nitrate, as was suggested by Aller et al. (1998) in the Panama Basin and Mogollón et al. (2016) in the deep-sea sediment of the Clarion–Clipperton fracture zone in the northeast equatorial Pacific.

Previous estimation of denitrification in 0–2 cm depth of the UB ranged from 0.01 to 0.17 mmol N m-2 d-1 (Lee, 2009), which is equivalent to a Corg oxidation of 0.013–0.213 mmol C m-2 d-1 using the stoichiometric equation of 4H++ 5CH2O + 4NO3- = 5CO2 + 2N2 + 7H2O. Based on the average, the contribution of carbon oxidation by denitrification (0.11 mmol C m-2 d-1) should be minor at the basin site (≤ 3 % of total Corg oxidation at 0–2 cm; ∼ 1 % of integrated Corg oxidation). This is consistent with the general consensus that Corg oxidation by denitrification is of little importance in most marine sediments (Sørensen et al., 1979; Canfield et al., 1993a; Trimmer and Engström, 2011). Denitrification may be even further suppressed in Mn-rich sediments due to competitive inhibition from Mn reduction (Trimmer et al., 2013).

Microbial Fe reduction has been quantified directly in sediments of various coastal oceans (Gribsholt et al., 2003; Kostka et al., 2002a, b; Hyun et al., 2007, 2009b) and indirectly in deeper continental margins (Thamdrup and Canfield, 1996; Jensen et al., 2003; Kostka et al., 1999). Earlier estimation from 16 different continental margin sediments indicated that Fe(III) reduction contributed 22 % on average to anaerobic carbon oxidation (Thamdrup, 2000). Thus, the contributions from Fe(III) reduction of 12 and 20 % of anaerobic Corg oxidation on the slope (M1) and in the basin (D3) of the UB (Table 4) fall in the range of the previous indirect estimates.

Unlike Fe reduction, direct estimation of manganese reduction rates is not easy, mainly because of the restriction of the process to a thin surface layer (Sundby and Silverberg, 1985), the rapid reduction of manganese oxides with H2S and Fe2+ (Postma, 1985; Burdige and Nealson, 1986; Kostka et al., 1995; Lovley and Phillips, 1988), and the adsorption of Mn2+ to Mn oxide surface (Canfield et al., 1993b). For that reason, only two studies, from the Skagerrak and Black Sea, are available for direct comparison on the partitioning of Mn reduction. The process has also been indicated to be of importance in the Panama Basin based on diagenetic modeling (Aller, 1990) and at some Arctic shelf sites where it was, however, not quantified separately from Fe reduction (Vandieken et al., 2006; Nickel et al., 2008). Mn reduction was responsible for over 90 % of total Corg oxidation at 600 m depth in the Skagerrak (Canfield et al., 1993b), and accounted for 13–45 % of anaerobic Corg oxidation in the Black Sea shelf sites at 60–130 m of water depth (Thamdrup et al., 2000). To our knowledge, this report of Corg oxidation dominated by Mn reduction comprising 45 % of total Corg oxidation and 57 % of anaerobic Corg respiration in the center of the UB (Table 4) represents the first from deep-offshore basin of the eastern Asian marginal seas.

The difference in partitioning of Mn reduction in Corg oxidation between the UB, Black Sea, and Skagerrak reflects the close relationship between Mn oxide content in the sediment and Mn reduction (Thamdrup et al., 2000). From the vertical distribution of electron acceptors (Fig. 5f) and contribution of each Corg oxidation pathway with depth (Fig. 6), it is apparent that the availability of Mn(IV) largely controls the relative contribution to C oxidation. In the Skagerrak, the Mn oxides are abundant in high content down to 10 cm depth (Canfield et al., 1993b), whereas Mn oxides in the Black Sea and the Ulleung Basin were enriched only down to 2 and 4 cm, respectively (Thamdrup et al., 2000; Fig. 2g). Using the available data set for the three marine sediments, we further plotted the relative contribution of manganese reduction to anaerobic carbon oxidation as a function of Mn oxide content to expand data from Thamdrup (2000) (Fig. 7). The plot indicates saturation kinetics with a close correlation between Mn oxide content and the importance of Mn reduction at low contents. Curve fitting yields a content of MnO2 at 50 % of contribution of manganese reduction to total Corg oxidation (Ks) of 8.6 µmol cm-3 similar to the approximately 10 µmol cm-3 suggested before (Thamdrup et al., 2000). This indicates that Mn reduction can be a dominant Corg oxidation process even at low contents of Mn oxides compared to those found at UB. Manganese enrichments of this magnitude have been reported for several locations on the continental margins and in deep basins (Murray et al., 1984; Gingele and Kasten, 1994; Gobeil et al., 1997; Haese et al., 2000; Mouret et al., 2009; Magen et al., 2011; Macdonald and Gobeil, 2012; Mewes et al., 2014) in addition to the relatively few places where dissimilatory Mn reduction was already indicated to be of importance, as discussed above. Thus, the process may be of more widespread significance, particularly in deep basin settings such as UB that allow geochemical focusing of manganese.

The strong enrichment of Mn in the UB surface sediment is primarily of diagenetic origin as indicated by just slightly higher Mn contents at depth in the sediment at D3 (mean 1.1 µmol cm-3 at 10–20 cm depth) compared to M1 (0.45 µmol cm-3) (Fig. 2) combined with higher sediment accumulation rates at the slope (0.15–0.3 cm year-1) than in the basin (0.07 cm year-1; Cha et al., 2005). Thus, the burial flux of Mn, and thereby the net input assuming steady-state deposition, is similar or higher at M1 compared to D3. Furthermore, Mn is likely subject to geochemical focusing in the basin as Mn depositing at shallower depths is reductively mobilized and incompletely oxidized in the thin oxic surface layer, resulting in release to the water column and net down-slope transport, as inferred in other ventilated basins (Sundby and Silverberg, 1985; Canfield at al., 1993b; Schaller and Wehrli, 1997). A diagenetic source of Mn enrichment was also concluded in previous studies (Yin et al., 1989; Cha et al., 2007; Choi et al., 2009). The Mn remaining and being buried at M1 likely represents unreactive detrital forms to a larger extent than at D3 (Cha et al., 2007). Adopting the sediment accumulation rate of 0.07 cm year-1 in the UB determined at a station 50 km from D3 (Cha et al., 2005), the average Mn(DCA) content of 1.1 µmol cm-3 at 10–20 cm depth (Fig. 2g) corresponds to a flux for permanent burial of 0.002 mmol m-2 d-1 or just 0.03 % of the Mn reduction rate (Table 3) – i.e., an Mn atom is recycled 3800 times before it finally gets buried, first by stripping from the particles that settle to the seafloor and subsequently, over and over, by reductive dissolution of the Mn oxides that from by reoxidation in the oxic surface layer (or, potentially, in the nitrate zone; Aller et al., 1998; Mogollón et al., 2016). This is a much more extensive recycling than found in the Mn sediment of Skagerrak (130–260 times; Canfield et al., 1993b). The difference results mainly from a much higher burial flux of Mn (as authigenic Mn[II]) in the Skagerrak (∼ 40 µmol cm-3; Canfield et al., 1993b). The reason that little, if any, authigenic Mn(II) is buried in the UB is not clear.

As noted in previous studies (Aller, 1990; Canfield et al., 1993b), high contributions of Mn and Fe reduction to carbon oxidation in offshore sediments require physical mixing, which typically occurs through bioturbation. This is also the case for the UB, where the burial flux from the oxic surface layer into the Mn reduction zone corresponded to 0.4 mmol m-2 d-1 or 5 % of the Mn reduction rate (213 µmol cm-3 × 0.07 cm year-1). Bioturbation has previously been inferred, but not quantified, from 210Pb profiles in the UB (Cha, 2002), and thin polychaete worms were observed during our sampling. Assuming bioturbation to be a diffusive process, we estimate, in a similar manner as in the previous studies and based on the average gradient in Mn(DCA) from 0.5–1 to 7–8 cm, that the Mn reduction rate would be supported at a biodiffusion coefficient of 9.5 cm2 year-1. This value is 3.6 times lower than the coefficient estimated for the Skagerrak (Canfield et al., 1993b) and consistent with estimates for other sediments with similar deposition rates (Boudreau, 1994). The estimated biodiffusion coefficient (Db) of 9.5 cm2 year-1 at Site D3 corresponds to ∼ 2 % of the molecular diffusion coefficient of oxygen (388 cm2 year-1). Judging from the absence of major fauna in the UB sediments, the mixing is brought about by small organisms with each individual affecting only a small area relative to the size of our cores, and the Db averaging many of these small and local but frequent events. Under such conditions, bioturbation can drive Mn cycling in the UB without substantial smearing of the redox zonation. Similarly, Hyacinthe et al. (2001) found that well-defined profiles can be observed in both sediments with low and high bioactivity in the Bay of Biscay.

The SRRs measured in this study (0.43–4.29 mmol m-2 d-1) are higher than those measured in productive systems such as the Benguela upwelling system in the southeastern Atlantic (0.14–1.39 mmol m-2 d-1; Ferdelman et al., 1999), and even comparable to those reported at the Chilean (2.7–4.8 mmol m-2 d-1; Thamdrup and Canfield, 1996) and Peruvian upwelling system (5.2 mmol m-2 d-1; Fossing, 1990) at a similar depth range of 1000–2500 m. The total anaerobic DIC production rates at the slope (14.0 mmol m-2 d-1) and basin site (7.2 mmol m-2 d-1) were also comparable to those measured at the same depth range of a Chilean upwelling site (9.2–11.6 mmol m-2 d-1) (Thamdrup and Canfield, 1996). Since rates of benthic carbon oxidation are largely controlled by the supply of Corg (Canfield et al., 2005), a high Corg flux reflected in the high Corg content (> 2.5 %, dry wt.) in the sediment of the UB (Table 1) is likely to explain the high metabolic activities. A similar high Corg content as in the UB is rarely found in deep-sea sediment underlying oxic bottom water at depths exceeding 2000 m, except for a Chilean upwelling site (Lee et al., 2008). This high Corg content in the UB is mainly associated with the combination of enhanced biological production resulting from the formation of coastal upwelling (Hyun et al., 2009a), enhanced new production in summer (Kwak et al., 2013), occurrence of an intrathermocline eddy resulting in the extraordinary subsurface chlorophyll a maximum (Kim et al., 2012), high Corg accumulation rates exceeding 2 g C m-2 year-1 (Lee et al., 2008), and high export production (Kim et al., 2009). Consequently, high benthic mineralization resulting from the high Corg in the sediment implied that the UB is a biogeochemical hotspot where significant turnover of organic matter and nutrient regeneration occur.

Recent oceanographic observations revealed that the gradual deoxygenation and warming of the bottom water of the East Sea over the last 30 years have resulted in an ∼ 10 % decrease in dissolved oxygen and ∼ 0.04 ∘C increase in potential temperature (Kim et al., 2001; Gamo, 2011; Gamo et al., 2014). Benthic metabolism and respiratory Corg oxidation coupled to various terminal electron-accepting processes in the sediments are largely controlled by the combination of O2 content, temperature, and biological production overlying the water column (Canfield et al., 2005). It is thus important to monitor any changes in the rates and partitioning of Corg oxidation to better understand and predict the variations in biogeochemical cycles of carbon, nutrients, and metals potentially associated with long-term climatic changes in the UB, the biogeochemical hotspot of the East Sea.