Corresponding author: Christian März, AG Mikrobiogeochemie, Institut für Chemie und Biologie des Meeres (ICBM), Postfach 2503, 26111, Oldenburg, Germany. (email@example.com)
 Intergrated Ocean Drilling Program Expedition 323 recovered a sediment record covering the last ~4.3 Ma from the Bering Sea (Integrated Ocean Drilling Program Site U1341, Bowers Ridge, 2177 m water depth). To resolve Pliocene-Pleistocene paleoenvironmental changes in this marginal basin, ~190 sediment samples were analyzed for their bulk element composition. Aluminium contents in Bowers Ridge sediments are variable but overall higher toward younger sediments, probably related to the intensification of the Northern Hemispheric Glaciation and increasing sea ice transport in the Bering Sea. The gradual increase of terrigenous input is mirrored by decreasing SiO2 and excess Si (Sixs) contents, but the overall Si enrichment of the deposits reflects continuous opal deposition since the Pliocene at Bowers Ridge. Unlike in the North Pacific, the Sixs record at Site U1341 does not support a dramatic decrease in opal export following the onset of the Northern Hemispheric Glaciation around 2.7 Ma, but SiO2xs have higher accumulation rates (up to ~8 g/cm2/ka) between ~2.6 and ~1.8 Ma BP. During this period, the major oceanic opal deposition centers shifted globally from open marine high latitude regions to upwelling areas. We here discuss how the onset of North Pacific stratification at ~2.7 Ma BP may have caused leakage of nutrient-rich deep/intermediate North Pacific water into the Bering Sea via the deep Kamtchatka Strait, leading to increased opal deposition—and likely reactive organic matter export—at Bowers Ridge. As a result, magnetic and geochemical records were overprinted by intensified diagenesis, significantly affecting their applications as paleoceanographic proxies.
 Over the last decades, scientific ocean drilling has been crucial to improve our understanding of the Earth's climate system and its variabilities [Bickle et al., 2001]. However, there still is a lack of high-quality sediment records from the (sub-) Arctic Ocean basins, defying the critical importance of these regions as both amplifiers and archives of global climate change. In summer 2009, the Integrated Ocean Drilling Program (IODP) Expedition 323 retrieved around 5.7 km of sediment cores from the subarctic Bering Sea [Expedition 323 scientists, 2010; Takahashi et al., 2011] covering the last ~4.3 Ma, including the Pliocene-Pleistocene transition (2.588 Ma BP). The Bering Sea is a key location for reconstructing subarctic climate variability on various timescales [Takahashi, 2005]. The periodic closure of the Bering Strait (the only northern Pacific-Atlantic connection) and nutrient cycling in one of the world's most productive ocean regions [Springer et al., 1996] might have exerted major effects on global climate variations [De Boer and Nof, 2004]. In 1973, the Deep Sea Drilling Project Leg 19 drilled the first long sediment cores covering the Quaternary of the Bering Sea, but recovery and quality of the core material were rather poor [Scholl and Creager, 1973]. Still, this expedition allowed for a first examination of the ages and characteristics of Bering Sea sediments.
 In 1992, the Ocean Drilling Program (ODP) Leg 145 drilled several sites in the North Pacific just south of the Aleutian Islands [Rea et al., 1993]. Subsequent studies revealed that the North Pacific played an important role in Pliocene-Pleistocene climate variations by storing and releasing CO2 and nutrients dissolved in its deep waters on glacial-interglacial timescales [Keigwin, 1998; Haug et al., 1999; Jaccard et al., 2005; Galbraith et al., 2007; Jaccard et al., 2009]. In particular, the major onset of Northern Hemispheric Glaciation (NHG; ~2.75 Ma BP) was accompanied by an enhanced stratification, a crash in diatom opal accumulation, and an increase of ice-rafted debris (IRD) deposition in the North Pacific [Haug et al., 1995; Maslin et al., 1996; Haug et al., 1999].
 To investigate if similar paleoceanographic signals were also recorded in the the Bering Sea, we studied sediments recovered at IODP Site U1341 on Bowers Ridge (a drowned volcanic arc in the southern Bering Sea) using inorganic-geochemical methods (Figure 1). We here show how detrital input and biosilica productivity developed at this setting over the last ~4.3 Ma. In addition, we will briefly discuss early diagenetic reactions that overprinted the sediment composition and bias paleoceanographic proxies (e.g., magnetic susceptibility [MS] and Ba records). More details on diagenetic processes at Site U1341 will be presented in a forthcoming publication [Wehrmann et al., accepted].
2 Material and Methods
 The studied sediment cores were retrieved from the United States Implementation Organization drill ship JOIDES Resolution by the Advanced Piston Coring (to 458 m drilling depth below sea floor, DSF) and the Extended Coring Barrel (to 600 m DSF) systems. A composite record was established as a splice of Holes 1341A, B, and C (2177 m water depth). Sediment depths are given in meters corrected core composite depth below sea floor (m CCSF-B), which is a close approximation of the actual drilled interval in the sediment column [Expedition 323 Scientists, 2010]. The age model used here is an update of the shipboard preliminary age model, established by a combination of magneto- and biostratigraphy [Expedition 323 Scientists, 2010; Takahashi et al., 2011]. Data were revised through detailed onshore biostratigraphic studies, and interpolation between these age points (assuming linear sedimentation rates) is the basis for the age model used in this study. Extrapolating sedimentation rates beyond the oldest diatom-based datum (~3.87 Ma) result in an estimated maximum sediment age of ~4.3 Ma at Site U1341, whereas it is ~4.0 Ma based on the oldest radiolarian datum [Ikenoue et al., 2011; Onodera et al., accepted]. Other shipboard investigations relevant for this study are detailed in the IODP Expedition 323 Preliminary Report and Proceedings volumes [Expedition 323 Scientists, 2010; Takahashi et al., 2011].
 Sediments at Site U1341 are a variable mixture of detrital, biogenic, and authigenic components, with a major contribution of biogenic opal and detrital clay/silt, and subordinate occurrence of sand layers, ash beds, semilithified authigenic carbonate layers, and biogenic carbonate. Shipboard geochemical analyses included pore water extraction by pressure squeezing of whole rounds and subsequent determination of sulphate concentrations. Squeeze cakes were analyzed for total sulfur, total inorganic carbon, and total organic carbon (TOC). Respective methods are detailed in the IODP Expedition 323 Preliminary Report and Proceedings [Expedition 323 Scientists, 2010; Takahashi et al., 2011].
 For shore-based geochemical analyses, around 190 sediment samples were obtained on board from whole-round pore water squeeze cake residues (~10 cm thickness) and discrete plastic scoop samples (~2 cm thickness). Samples were taken directly after core retrieval and core opening, respectively, and frozen until further processing. No samples were taken from visible sand, ash or authigenic carbonate layers, or finely laminated intervals. Onshore, sediment samples were freeze-dried and ground in an agate ball mill. The water content (in wt%) was determined by weighing the samples before and after freeze-drying. Each sample (700 mg) was mixed with 4200 mg di-lithiumtetraborate (Li2B4O7, Spectromelt A10), preoxidized at 500 °C with ~1.0 g NH4NO3 (p.a.), fused to homogenous glass beads, and analyzed for Al, Ba, Ca, K, Mg, Mn, Ni, P, Si, and Zr contents by wavelength-dispersive X-ray fluorescence (Panalytical PW 2400). Analytical precision and accuracy were better than 5% as checked by in-house and international standards. Complementary to shipboard total inorganic carbon, TOC, and total sulfur analyses, we determined total carbon (TC) and sulfur contents of our samples by combustion analysis with an Eltra C/S analyser. Comparing the onboard and onshore TC contents reveals that both methods are in excellent agreement. Total sulfur contents were corrected for the contribution of pore water sulfate using water contents of the samples and an average sea water sulfate concentration of 0.271 wt% (29 mM). Mass accumulation rates (MARs; g/cm2/ka) were calculated by multiplying dry bulk densities from onboard “moisture and density” measurements (g/cm3) [Expedition 323 Scientists, 2010; Takahashi et al., 2011] with (excess) contents of the respective element (wt%) and linear sedimentation rates (cm/ka). Despite the higher resolution of GRAPE density data, these were affected by gas expansion in the cores (D. Scholl, personal communication), so we used moisture and density data of discrete samples taken close (within 20 cm distance) to our geochemistry samples.
 The dominanting element in sediments at Site U1341 is silica, with SiO2 contributing 60–80 wt% to the total sediment mass (Figure 2). Despite significant superimposed variability especially in the lower and middle part of the record, there is a clear long-term SiO2 trend of decreasing values toward younger sediments. Al2O3 (2–15 wt%) shows an opposing trend (Figure 2), illustrated by a clear negative SiO2-Al2O3 correlation (R2 = 0.86; Figure 2), which indicates mutual dilution of these compounds.
 This relationship is also evident in a ternary SiO2-CaO-Al2O3 diagram, with the endmember components quartz and/or opal, calcium carbonate, and Al-oxide (representing a major structural component of clay minerals) (Figure 3a) [Brumsack, 1989]. Most of the samples fall on a dilution line between SiO2 and the composition of upper continental crust (UCC) [Wedepohl, 2004], but not average shale [Wedepohl, 1971, 1991]. Several samples deviate from this mixing line toward higher Ca contents (gray box), but not in a systematic way, and are excluded from the following records because they are strongly affected by authigenic carbonate precipitation [Wehrmann et al., 1995]. Figures 2 and 3a show that the element records at Site U1341 are affected by variable dilution of two main geochemical components, i.e., SiO2 (quartz and/or opal) and UCC-like material. To test if the SiO2 dilution was caused by enrichments of quartz or opal, Si contents are plotted against Zr, resulting in a clear negative relationship (Figure 3b).
 From Figure 3a, we can draw the conclusion that the most terrigenous background material has a composition similar to UCC. The additional contribution of andesitic or basaltic material can be estimated using a MgO-K2O-Al2O3 ternary diagram (Figure 3c) [Böning et al., 2004]. Although the similarity of the Site U1341 sediments to UCC is confirmed, some samples show slight trends toward overall lower K and Mg contents, approaching the composition of average Meso-Cenozoic andesite [Condie, 1993].
 To extract the temporal variability of, e.g., biogenic silica contents from bulk geochemical records, we need to correct for the variable dilution by fine-grained siliciclastic material. This is done by calculating element contents in excess of the UCC composition [elementxs = elementsample − Alsample × (element / Al)UCC] [Brumsack, 2006], assuming that Al is exclusively bound to siliciclastic material. The resulting Sixs plot (Figure 4; gray line and symbols = analyzed data, black line = 5-point runing mean) shows overall higher values in the lower part of the record. Between 210 and 380 m CCSF-B and below 500 m CCSF-B, Sixs peaks reach between 20% and 35%. The SiO2xs MAR record shows a very similar downcore pattern. Values in the oldest part of the record (600–410 m CCSF-B) are mostly around 3–5 g/cm2/ka. Between 410 and 377 m CCSF-B, there is a short drop in accumulation rates to <2 g/cm3/ka, followed by an abrupt increase to values around 3–8 g/cm2/ka (377–210 m CCSF-B). Between 210 and 150 m CCSF-B, the SiO2xs MAR fluctuates between 6 and 1 g/cm2/ka and drop to <3 g/cm2/ka above 150 m CCSF-B.
 The Baxs record at Site U1341 (Figure 4) shows enrichments relative to UCC (Ba/Al (104) = 73), with Baxs contents exceeding 200 ppm in most samples (>1000 ppm near the top of the record). However, between ~310 and 430 m CCSF-B, Baxs contents are consistently less than 200 ppm, and sometimes even depleted relative to the UCC. This Ba-poor interval corresponds to lowest abundances of calcareous nannofossils, benthic, and planktonic foraminifera over the whole sedimentary record [Expedition 323 Scientists, 2010; Takahashi et al., 2011].
 Figure 4 further illustrates the record of sedimentary Sxs and the shipboard magnetic susceptibility record. Sulfur is significantly enriched in the sediments relative to UCC (697 ppm) [Wedepohl, 2004], especially below ~220 m CCSF-B, with Sxs maxima >1% occurring between ~310 and 400 m CCSF-B. Also, the MS record (Figure 4), which (simplified) is a function of the relative proportions of ferrimagnetic magnetite and paramagnetic pyrite [Karlin and Levi, 1983], can be subdivided accordingly. The upper part of the record (above ~220 m CCSF-B) shows highly variable MS values (0–400 × 10−6 SI), and the lower part (below ~220 m CCSF-B) shows values mostly below detection limit.
 The record of Mn at Site U1341 shows a slight but consistent depletion relative to UCC (Figure 4), which is most strongly expressed between ~320 and 400 m CCSF-B. Also, P is depleted in most samples relative to UCC (Figure 4). Laminated intervals [Expedition 323 Scientists, 2010; Takahashi et al., 2011] mostly occur in the central part of the record, between ~200 and 420 m CCSF-B, roughly coinciding with maximum Sixs contents and SiO2xs MARs as well as low MS.
 Between 310 and 430 m CCSF-B, TC contents (Figure 5) show a slight but consistent enrichment synchronous to the Sixs peak (Figure 4). Shipboard data reveal that below ~100 m CCSF-B, the sediments are almost carbonate-lean, and total sedimentary C is dominated by organic matter (OM; Figure 5) [Expedition 323 Scientists, 2010; Takahashi et al., 2011]. We therefore suggest that below 100 m CCSF-B, our TC contents basically record organic carbon.
4.1 The Composition of the Terrigenous Background Material at Bowers Ridge
 As shown by the SiO2-Al2O3 relationship (Figures 2 and 3a), the sediments at IODP Site U1341 largely represent a two-component system of fine-grained siliciclastic material and biogenic opal, except for some Ca-rich layers [Aiello and Ravelo, 2013]. The composition of the lithogenic background material (Figure 3a) seems to be more similar to UCC [Wedepohl, 2004] than to average shale [Wedepohl, 1971, 1991], which is expected for a high-latitude setting dominated by physical weathering. Figure 3c indicates that the K depletion of most samples is related to additional volcanic material in the Bowers Ridge deposits. There are two potential volcanic sources to the southern Bering Sea: the Aleutian Island volcanic arc to the south (geochemically variable, but mostly andesites) and the Bering Sea basalt province to the northeast [e.g., DeLong, 1974; Kay, 1978; Kay et al., 1978; Winer et al., 1992]. The clustering of the Site U1341 samples in Figure 3c as well as the low Ni contents (not shown) of the sediments (7–49 ppm, average = 23 ppm, compared with 56 ppm in UCC) are typical features of Aleutian Island andesites (3–36 ppm) [DeLong, 1974].
 It can be assumed that finely dispersed andesitic material was transported from the Aleutian Island arc to the Bowers Ridge by currents, wind, and/or sea ice. An additional contribution of Asian dust (e.g., loess) to the detrital input at Site U1341 is also likely, especially after ~2.7 Ma BP [Bailey et al., 2011], but is difficult to quantify due to the geochemical similarities of Chinese loess and UCC [Taylor et al., 1983].
4.2 Paleoproductivity Reconstructions from Bering Sea Sediments
 The Bering Sea is often cited as one of the most productive ocean regions [e.g., Walsh et al., 1989; Springer et al., 1996; Sambrotto et al., 2008], although recent studies question this traditional view [Brown et al., 2011]. As the importance of high-productivity regions for atmospheric CO2 sequestration largely depends on the export of OM to the sea floor, reconstructing this export productivity from sedimentary proxies is an important issue. Various geochemical proxies are used to reconstruct paleoproductivity from marine sediments [review by Wefer et al., 2007], and we will discuss their applicabilities at Site U1341.
 Biogenic carbonate is very low in Bering Sea sediments because (a) the ecosystem is dominated by biosiliceous organisms and (b) carbonate is dissolved at and below the sea floor [e.g., Wehrmann et al., 1995]. The sediments recovered at Site U1341 (2177 m water depth) are also consistently poor in organic carbon, with TOC contents <1 wt% in almost all samples (mostly around 0.5 wt%; Figure 5) [Takahashi et al., 2011]. The fact that biogenic opal is the dominating sediment constituent at Site U1341 suggests that despite high opal productivity, the long-term organic carbon burial efficiency of the Bering Sea sediments at Bowers Ridge was low. Probably remineralization processes in the water column, at the sediment-water interface, and/or in the sediment decoupled the organic carbon from the biogenic silica record [e.g., Ragueneau et al., 2000; Henson et al., 2012]. In addition, Wehrmann et al.  showed that along the eastern Bering Sea slope, organic matter degradation simultaneously produced methane and dissolved bicarbonate, which were eventually fixed in the sediments as authigenic carbonate phases. Thus, both TOC and carbonate contents seem to be poor recorders of paleoproductivity in Bering Sea sediments. Still, the Site U1341 record depicts some intervals with higher TC and TOC contents (e.g., around 350 m CCSF-B), likely documenting enhanced OM export and/or preservation (Figure 5).
 As the Bering Sea is dominated by opal productivity, the biogenic silica record at Site U1341 may be considered as an adequate paleoproductivity proxy. However, shipboard mineralogical analyses [Takahashi et al., 2011] imply that opal-A to opal-CT transformations already occur at relatively shallow burial depths. This recrystallization potentially overprints the original biogenic opal record, and especially the fraction of opal extractable from the sediment by conventional alkaline leaching techniques [Expedition 323 Scientists, 2010; Takahashi et al., 2011]. The bulk SiO2xs contents should be much less affected by recrystallization than the leachable opal fraction, as the opal-A to opal-C/T transformation occurs in situ and usually does not involve diffusion of dissolved silica through the pore space [Leinen, 1977; Leinen et al., 1986; McDonald et al., 1999].
 Figure 3a demonstrates that all samples at Site U1341 fall on a dilution line between UCC [Wedepohl, 2004] and pure SiO2. However, this SiO2 dilution could be either caused by biogenic silica (e.g., diatoms) or by Si-rich detrital material (e.g., quartz silt and sand). We thus compare the Si and Zr records (Figure 3b), as Zr in marine sediments is dominantly bound to the heavy mineral zircon and associated with higher quartz contents. As both coarse-grained and heavy minerals are enriched under higher depositional energies, Zr may serve as a proxy for elevated heavy mineral and quartz contents [Schnetger et al., 2000; Calvert and Pedersen, 2007]. The negative correlation of Si with Zr at Site U1341 (Figure 3b) proves that the Si enrichment is not related to detrital quartz, which is in agreement with Aiello and Ravelo  who report that the sand fraction at Site U1341 is mainly composed of whole centric diatom valves. We conclude that the Sixs record at Site U1341 is a valuable proxy for biogenic opal contents over the last ~4.3 Ma.
 A paleoproductivity proxy frequently used in the Pacific region is biogenic barite [e.g., Church, 1970; Dehairs et al., 1980; Bishop, 1988; Francois et al., 1995; Paytan and Griffith, 2007]. According to the most widely accepted model, biogenic barite forms due to barite oversaturation within OM-rich aggregates sinking through the water column [Dehairs et al., 2000; Ganeshram et al., 2003]. The Ba content of bulk sediments has been used to estimate the productivity-related (biogenic) barite contribution, applying a factor to eliminate Ba in the detrital background material [e.g., Dymond et al., 1992; Dean et al., 1997; Bonn et al., 1998; Jaccard et al., 2009]. Consequently, Ba enrichments relative to UCC at Site U1341 (Figure 4) would imply high paleoproductivity and OM export, whereas very low Baxs contents between ~310 and 420 m CCSF-B would document extremely low paleoproductivity (Figure 4). However, this low-productivity scenario is not supported by the records of Sixs and TC that both reach maximum values within this interval (Figures 4 and 5). Instead, we must consider diagenetic processes to be responsible for Ba depletion in this interval, specifically the transfromation of biogenic into authigenic barite [Brumsack, 1986; Von Breymann et al., 1992; Torres et al., 1996; McManus et al., 1998]. In high-productivity settings, intense bacterial sulphate reduction can cause sulphate depletion of pore waters and promote the redistribution of barite within the sediment column [e.g., McManus et al., 1994; Arndt et al., 2006, 2009; Hendy, 2010]. We suggest that such processes almost completely leached biogenic barite from the interval between ~310 and 420 m CCSF-B at Site U1341, as supported by the pore water sulphate profile [Expedition 323 Scientists, 2010; Takahashi et al., 2011]. Further evidence for a strong diagenetic overprint at Site U1341 less than ~220 m CCSF-B comes from the low MS values, which document that the deposits were affected by intense sulphate reduction and pyrite formation. Also, the depletion of Mn in the lower part of the record and the occurrence of laminated sediments (Figure 4) support syn-sedimentary oxygen depletion shortly below the sediment-water interface at Site U1341 [Froelich et al., 1979; Brumsack, 2006], suggesting that bacterial sulphate reduction dominated organic matter remineralization. Therefore, the Ba record at Site U1341 has been strongly overprinted by diagenesis and should not be used as a paleoproductivity proxy.
 In sediments from high-productivity areas, P is often enriched in organic matter, fish remains, or authigenic apatite [e.g., Brumsack, 1989; Delaney, 1998; Brumsack, 2006; Calvert and Pedersen, 2007; Filippelli, 2008]. However, as shown in Figure 4, the sediments at Site U1341 are slightly depleted in P relative to UCC and do not correlate with the Sixs record. Although this lack of P enrichment might indicate that all P is bound to terrigenous material with a lower P content than UCC, it could also result from the remobilization of diagenetically mobile P phases. As indicated by the Mn depletion and laminated sediment intervals (Figure 4), the surface sediments at Site U1341 were oxygen-depleted over much of the past 4.3 Ma. Under such conditions, P is efficiently recycled from the sediments into the water column [e.g., Ingall et al., 1993; Ingall and Jahnke, 1997; Anderson et al., 2001; Algeo and Ingall, 2007]. In any case, bulk P contents cannot be used as paleoproductivity indicator at Site U1341.
 In summary, the Sixs contents and SiO2xs MARs at Site U1341 are the most reliable proxies to reconstruct paleoproductivity (i.e., the buried fraction of biogenic opal) over the past ~4.3 Ma, whereas organic carbon, carbonate, barium, and phosphorus records are strongly overprinted by postdepoitional processes.
4.3 A Long-term trend of Opal Deposition on Bowers Ridge: Potential Nutrient Leakage from the North Pacific
 As shown by Haug et al. [1995, 1999] and Maslin et al. , a dramatic decrease in opal contents and opal accumulation rates and an increase in MS marked the onset of the NHG around 2.7 Ma BP in the North Pacific at ODP Site 882. The abrupt onset of North Pacific salinity stratification prevented the delivery of nutrient-rich abyssal water masses to the photic zone and dramatically reduced diatom productivity. Higher MS values were supposedly caused by an increase of winter sea ice formation and the resulting delivery of terrigenous material (IRD) rich in magnetite to the open North Pacific. Also, the increased input of volcanic ash from Aleutian and Kamchatka volcanoes, potentially in combination with Asian dust, was suggested as an explanation for the elevated MS of the sediments after ~2.7 Ma BP [Rea et al., 1995; Prueher and Rea, 1998; Prueher and Rea, 2001; Bailey et al., 2011].
 To examine the Bering Sea record at IODP Site U1341 for potential NHG signals, we here focus on biogenic opal and MS records. When comparing the opal contents and MARs of ODP Site 882 to the respective SiO2xs records at IODP Site U1341 (Figure 6), a decline of biogenic silica deposition around 2.7 Ma BP can be observed at both locations. However, opal deposition at Site U1341 quickly recovered, and the highest SiO2xs enrichments and MARs are recorded after 2.5–2.6 Ma BP. This finding is supported by highest total numbers of diatom valves between 2.4 and 2.6 Ma BP as reported by Onodera et al. [accepted]. Only after ~1.8 Ma BP, SiO2xs contents, and MARs decreased to overall lower values. The shape of the Site U1341 MS profile resembles the one of ODP Site 882. However, the abrupt increase of the MS record at U1341 did not occur at ~2.7 Ma BP, but only around 1.8 Ma BP (~220 m CCSF-B) [Expedition 323 Scientists, 2010; Takahashi et al., 2011]. This finding is counterintuitive, as one would expect an earlier onset of the NHG and related IRD input in the Bering Sea, which is located further north and closer to potential sea ice sources than the North Pacific.
 The almost synchronous decline of SiO2xs and increase of MS around 1.8 Ma BP supports a coupling of opal deposition and magnetic properties of the sediments via early diagenetic processes. Specifically, the MS record at IODP Site U1341 was probably affected by diagenetic dissolution of magnetic phases (e.g., magnetite, Fe3O4) below ~220 m CCSF-B [Expedition 323 Scientists, 2010; Takahashi et al., 2011]. Ferromagnetic magnetite is known to react with H2S produced during bacterial sulphate reduction, leading to the formation of paramagnetic pyrite (as indicated by the high Sxs values in the lower part of the Site U1341 record). Such changes to the diagenetic regime, and resulting effects on the magnetic properties, are common in marine sediments [e.g., Karlin and Levi, 1983; Canfield and Berner, 1987; Leslie et al., 1990; Poulton et al., 2004; Riedinger et al., 2005; März et al., 2008]. In many cases, including Site U1341, the diagenetic overprint on MS records appears to be related to increased primary productivity and reactive organic matter export to the sea floor [Tarduno, 1994; Dickens and Owen, 1996; Florindo et al., 2003; Hepp et al., 2009]. In the specific case of the Bering Sea, we conclude that the MS record does not appear to be a reliable proxy for IRD input, and at this point, no alternative IRD record exists at Site U1341 for a direct comparison with the open North Pacific records.
 The marked discrepancies between ODP Site 882 and IODP Site U1341 biosilica records might be due to their different oceanographic settings. Although Site 882 was drilled in the deep North Pacific (3244 m water depth), Site U1341 is located on the slope of the upwelling-influenced Bowers Ridge (2177 m water depth). Cortese et al.  and Bolton et al.  noted that the global oceanographic reorganization around the Pliocene-Pleistocene transition, and gradual cooling from ~2.7 Ma BP onwards, also affected marine biogeochemical cycles worldwide. In detail, the sites of major opal deposition shifted from the high latitudes (Southern Ocean, North Pacific) to upwelling systems (e.g., off Namibia, California), possibly due to “leakage” and upwelling of excess silicic acid from high latitudes via exported intermediate water masses. We therefore suggest that the sedimentary Sixs record at Site U1341 might document a shift from a “high latitude–type” system (as the North Pacific and Southern Ocean) to an “upwelling-type” system (as the Namibian and Californian upwelling areas) [Maslin et al., 1996; Lange et al., 1999; Cortese et al., 2004; Etourneau et al., 2009; Bolton et al., 2011]. To further evaluate this global pattern, Figure 6 compares the SiO2xs MAR record at Site U1341 with the Pliocene-Pleistocene opal MAR records of several high-latitude and upwelling settings. Despite potential age uncertainties, some clear patterns emerge: in sediments older than ~2.7 Ma, the SiO2xs MARs at Site U1341 match the high opal MARs at ODP Sites 882 and 885 (open North Pacific) [Maslin et al., 1996; Haug et al., 1999] and ODP Site 1096 (Southern Ocean) [Hillenbrand and Cortese, 2006]. In contrast, after ~2.7 Ma BP, the SiO2xs record at Site U1341 matches increased opal MARs at ODP Sites 1018 and 1082 (upwelling systems off Namibia and California) [Lyle et al., 1997; Janecek, 2000; Perez et al., 2001; Etourneau et al., 2009], with consistently high opal deposition between ~2.6 and 1.8 Ma BP, but lower values after ~1.8 Ma BP. These results indicate that the global reorganization of oceanic opal deposition centers as stated by Cortese et al.  also affected the Bering Sea and the open North Pacific.
 A potential explanation for the shifting opal deposition from the open North Pacific to the Bering Sea after ~2.7 Ma BP could be the increased North Pacific salinity stratification and the formation of sea ice that caused the “opal crash” at ODP Site 882 [Haug et al., 1999]. Under similar conditions of intensified stratification, the accumulation of excess nutrients in deeper water masses has been suggested as an export mechanism of excess silicic acid from the Southern Ocean to mid-latitudes during Pleistocene glacials [Chase et al., 2003; Kohfeld et al., 2005; Beucher et al., 2007; Dubois et al., 2010]. In the case of the Bering Sea, the stratification could have induced the export of deep/intermediate North Pacific water masses enriched in nutrients (Si, N, P) to the Bering Sea through the deep Kamchatka Strait (~4420 m water depth). Because of upwelling at the bathymetric elevation of the Bowers Ridge, this “nutrient leakage” might have caused increased diatom productivity and opal deposition at its western slope shortly after the open North Pacific became more stratified. Indeed, increasing Sixs contents and MARs at Site U1341 after ~2.7 Ma BP (Figure 4) are paralleled by changing diatom assemblages [Expedition 323 scientists, 2010; Takahashi et al., 2011; Onodera et al., accepted], supporting variations in the nutrient regime at Bowers Ridge during this time. Also, Lawrence et al.  and Bolton et al.  suggested nutrient leakage from the increasingly stratified North Pacific to explain a productivity increase in the Eastern Equatorial Pacific (EEP) starting around 3–2.7 Ma BP. Although the Bering Sea and the EEP were certainly affected by different paleoceanographic conditions over the Plio- and Pleistocene, both oceanographic setting might have been affected by the increased availability of nutrients as the North Pacific became increasingly stratified. Notably, Aiello and Ravelo  interpret a systematic decrease in diatom valve preservation at Site U1341 between ~2.6 and 2.0 Ma BP as a result of lower silica concentrations at Bowers Ridge, which does not agree with the nutrient leakage hypothesis. This co-occurence of low diatom valve preservation but at the same time higher total diatom valve counts [Onodera et al., accepted] and higher SiO2xs MARs (this study) after ~2.6 Ma BP will certainly require further investigation.
 In contrast, in the Sea of Japan (ODP Leg 127), White and Alexandrovich  described an opal dissolution zone in sediments deposited at ~2.3–2.9 Ma BP, coinciding with the Pliocene-Pleistocene transition and the intensification of the NHG. The authors attributed this low biosilica preservation to a decreasing supply of silica-rich North Pacific surface waters through the shallow straits (~130 m water depth) into the Sea of Japan, inducing silica undersaturation and opal dissolution in this marginal basin. The salinity stratification of the North Pacific that developed around 2.7 Ma BP can explain this paleoceanographic scenario: water exchange between the North Pacific and the Sea of Japan was limited to the relatively fresh, nutrient-poor surface layer due to the lack of deepwater connections. In contrast, nutrient-rich intermediate to deep North Pacific water masses could have been exported into the Bering Sea through the deep Kamchatka Strait. From this difference in biogenic Si records between the Bering Sea and the Sea of Japan, we conclude that any potential nutrient leakage from the open North Pacific into its marginal basins would have strongly depended on the existence of deepwater connections [Vashchenko, 2011].
 Based on these results, it is a distinct possibility that increasing stratification in the North Pacific related to the NHG shut down opal production across the subarctic Pacific Gyre but triggered enhanced productivity both in the Bering Sea and in the EEP, by the export of intermediate to deep, nutrient-rich waters. In the case of the Bering Sea, the topography of Bowers Ridge might have facilitated the upwelling of these deeper water masses. From this study, it is difficult to constrain if enhanced opal deposition between ~2.7 and ~1.8 Ma BP was restricted to Bowers Ridge or also took place along the Bering Sea shelf. In any case, the stable stratification of the Bering Sea as a whole after ~2.7 Ma BP [e.g., Swann, 2010] is not supported by our data. Overall, we agree with Bolton et al.  that marine biogeochemical cycles were deeply affected by the onset of the NHG on a global scale. We also emphasize the role that the geometries and bathymetries of different ocean basins may have played for the distribution of nutrients and the regional development of primary productivity.
 Opal deposition in the southern Bering Sea (IODP Site U1341) overall declined since ~4.3 Ma BP, partly due to an increasing dilution by detrital material that reached Bowers Ridge either through wind transport or drifting sea ice. However, an “opal crash” related to the intensification of the NHG around 2.7 Ma BP, as reconstructed for the open North Pacific (ODP Leg 145), did not occur in the Bering Sea. Excess SiO2 MARs, applied as reliable and diagenetically stable proxies for opal deposition, declined slightly around 2.7 Ma BP but recovered thereafter and showed persistently high values up to ~1.8 Ma BP. These high Early Pleistocene SiO2xs contents and MARs at Bowers Ridge are consistent with a global shift in opal depositional centers from open-marine high-latitude settings to upwelling areas. Nutrient leakage due to sea ice formation and ocean stratification in the high latitudes may have contributed to the high biosilica productivity in upwelling systems. At the same time, additional processes resulting from the global cooling and aridification (e.g., an increased bioavailable Fe input by wind or sea ice) [Martin, 1990; Bailey et al., 2011] could have supported the hypothesized macronutrient leakage. The future inclusion of additional biogeochemical factors in the paleoenvironmental reconstruction at Bowers Ridge after the onset of the NHG will certainly require further investigation. This might contribute to explaining the discrepancy between high diatom counts [Onodera et al., accepted] and SiO2 MAR but low diatom valve preservation [Aiello and Ravelo, 2013] at Site U1341 between ~2.6 and 2.0 Ma BP.
 The high export productivity scenario at Bowers Ridge at least between 4.3 and 1.8 Ma BP is clearly supported by the effects that intensified diagenetic processes had on geochemical and geophysical sediment properties. There are clear relationships between the records of SiO2xs, Sxs, and MS and the occurrence of laminated sediment intervals, implying that increased opal export to the sea floor (documented by high SiO2xs MARs) triggered early diagenesis in the surface sediments (dissolution of Mn/Fe oxides, P recycling to the water column, sulphate reduction, and pyrite formation). This promoted the enhanced preservation of laminated sediments, sedimentary sulfur accumulation, slight Mn and P depletion, and dissolution of biogenic barite. The latter issue is of major importance for the reconstruction of export productivity from Bering Sea deposits, as the diagenetic remobilization of biogenic barite clearly limits its potential as a paleoproductivity proxy.
 The authors are most grateful to master, crew, staff, and co-chief scientists and all scientific participants on board JOIDES Resolution for making the IODP Expedition 323 a great success. In particular, discussions with S. Lund, J. Onodera, S. Kender, K. Husum, and L.M. Wehrmann are highly acknowledged. The authors thank C. Lehners, E. Gruendken, M. Schmidt, T. Mandytsch, and S. Asendorf for their analytical assistance. Special thanks to G. Cortese and C.D. Hillenbrand, who kindly provided opal data of several ODP sites. Two anonymous reviewers and Editor A.C. Ravelo gave insightful and constructive comments that improved the manuscript significantly. This project was financed by a DFG Research Fellowship granted to C. März within Priority Program 527 “IODP/ODP” (MA-4791/2-1 and 3-1).