Summary of expedition results

Overview of ages and sedimentation rates

Expedition 323 in the Bering Sea focused on analyzing long-term ocean and climate trends and the evolution of higher frequency glacial–interglacial to millennial-scale oscillations through the Pliocene and Pleistocene. As such, our primary drilling objective was to obtain sediments whose components could be used to elucidate the detailed evolutionary history of climate, surface ocean, and intermediate water conditions since the earliest Pliocene in the Bering Sea, where amplified high-resolution changes of climatic signals are recorded. In addition, we sought to explore subseafloor microbial activity in the Bering Sea because of its extremely high surface productivity.

Our objectives, in terms of acquiring the sediment core samples as well as conducting shipboard research, have been adequately met because our targeted drill sites had extremely high sedimentation rates (Fig. F5) containing abundant microfossils and other paleoceanographic proxies.

Among the three drill sites explored in the Bowers Ridge region, both of the deepest holes drilled, Holes U1340A (605 m uncompressed core depth below seafloor [CSF]) and U1341B (600 m CSF), represent the time spans between the Holocene to ~5 Ma in the Pliocene (Fig. F5). The expedition's initial goal of penetrating to ~5 Ma was adequately accomplished at both sites despite the failure of the extended core barrel (XCB) cutting shoe spacing sub in Hole U1340A, whose target depth of penetration was 700 m. In the gateway region sites (at the Bering slope), two deep holes were drilled: Holes U1343E (744 m CSF) and U1344A (745 m CSF). Hole U1343E reached ~2.1 Ma, whereas Hole U1344A reached ~1.9 Ma (Fig. F5). At other drill sites, the bottom ages or the sedimentary sequences based on biomagnetostratigraphy are as follows: Site U1339 (Umnak Plateau: 0.74 Ma), Site U1342 (Bowers Ridge: 1.2 Ma [with the exception of the middle Miocene sediments just above the basement]), and Site U1345 (gateway: 0.5 Ma).

The sediments recovered from Bowers Ridge display high sedimentation (~12 cm/k.y. at Sites U1340 and U1341) without apparent hiatuses and are generally appropriate for high-resolution Pliocene–Pleistocene paleoceanography with adequate calcareous benthic foraminiferal preservation in the Pleistocene but lower preservation in the Pliocene. On the other hand, sediments at these sites are generally barren of planktonic foraminifers and calcareous nannofossils except in the section between ~2.5 and ~3 Ma. The abundance of all siliceous microfossils is generally high, enabling good biostratigraphy and paleoceanographic reconstruction. Furthermore, the upper part of Site U1340 (~20–150 m uncompressed core composite depth [CCSF-A]) had obvious soft-sediment deformation due to mass movement possibly caused by local seismic activity. Although such deformation hinders the continuous reconstruction of late Pleistocene high-resolution paleoceanography at this site, information from other drill sites can readily fill the gap.

In the region of the Arctic gateway sites proximal to the Bering slope, the observed sedimentation rates were overwhelmingly high: Hole U1343E had sedimentation rates of 21–58 cm/k.y. and Hole U1344A had rates of 29–50 cm/k.y. Sedimentation rates were so high, in fact, that drilling reached ages of only 2.1 and 1.9 Ma, respectively, despite penetration to 745 m CSF at each site. Such high sedimentation rates stem from the deposition of silt and clays transported by the Yukon and other rivers as well as the terrigenous sediments once deposited on the shelf. In spite of the high percentage of terrigenous materials, pertinent biotic proxies including benthic foraminifers and siliceous microfossils are adequately preserved, enabling future studies. Therefore, the overall coverage of excellent cores to ~5 Ma in the Bowers Ridge region and ~2 Ma in the gateway region makes detailed, continuous high-resolution paleoceanographic studies relevant to global climate change possible.

Lithostratigraphic summary

The seven sites drilled during Expedition 323 provide a continuous high-resolution record of the evolution of marine sedimentation in the marginal Bering Sea (Fig. F6). Onboard, lithologic, and sedimentologic analysis of the core sediments was carried out with a combination of visual core description, smear slide analysis, and, only for selected samples, X-ray diffractometry. Overall, the sediments recovered in the Bering Sea are a mixture of three components: biogenic (mainly diatom frustules with varying proportions of calcareous nannofossils, foraminifers, silicoflagellates, and radiolarians), siliciclastic (mainly silt and, secondarily, sand and isolated IRD clasts of pebble to cobble size), and volcaniclastic (mainly fine ash). Other accessory lithologies identified include authigenic carbonates (dolomite and aragonite) and sulfides. The most prominent sedimentary features observed were decimeter- to meter-scale bedded alternations of sediment color and texture reflecting alternations in lithology. The sediments were generally highly bioturbated. However, fine-scale lamination preserving alternations between millimeter-scale laminae of biogenic and terrigenous material was also present in several of the drilled sites (Fig. F6).

The distributions of the sedimentary components and sedimentary structures and their variability both within and between the Expedition 323 sites account for changes of the biogenic, glaciomarine, terrigenous, and volcanogenic sediment sources and the environmental conditions present during sediment deposition. The scales of these lithologic variations indicate that sedimentation in the Bering Sea has recorded long-term trends that include the critical period of reorganization of Earth's climate from the warm early Pliocene and the transition into the ice ages. Milankovitch-scale (bedding) and sub-Milankovitch-scale (lamination) variability is also recorded. The physiographic setting of the different sites, their depths, and their location relative to the sediment source areas (continents, ice sheets, volcanoes, and upwelling centers) account for the marked regional differences in sediment composition, especially between the Pleistocene sections of the Bowers Ridge and Bering slope sites.

The preliminary results of Expedition 323 suggest that the history of sedimentation in the Bering Sea is broadly characterized by three main sedimentary phases that occurred between ~5 and ~2.7 Ma, ~2.7 and ~1.74 Ma, and ~1.74 Ma to recent (Fig. F6).

The oldest portion of the sedimentary record (between ~5 and ~2.7 Ma) was retrieved only at Bowers Ridge Sites U1340 and U1341. Sedimentation during the early middle Pliocene was relatively high (Fig. F5) and characterized by biogenic diatom ooze with minor amounts of diatom silt, sponge spicules, and vitric ash. Although the Pliocene sediment is commonly bioturbated, distinct intervals characterized by extensive lamination also occur. The oldest laminated intervals (<3.8 Ma) were observed at Site U1341. Stratigraphic intervals where laminations are concentrated are indicative of the depth of the OMZ. Isolated IRD pebbles were observed in sediments older than 3.8 Ma only at Site U1340. Limited dropstone occurrence prior to 2.7 Ma was also reported at two sites drilled in the northern Pacific during Leg 145 (ODP Sites 881 and 883) and in the Yakataga Formation in Alaska (Lagoe et al., 1993), which suggests the development of ice sheets prior to the onset of NHG (Krissek, 1995).

The middle section of Sites U1340 and U1341 (between ~2.7 and 1.74 Ma) is characterized by beds of diatom ooze with minor amounts of calcareous nannofossils and foraminifer ooze alternating with diatom silt beds. The latter are composed of subequal proportions of siliciclastic (silt-sized quartz, feldspar, and rock fragments and/or clay) and biogenic (mainly diatom and, secondarily, calcareous nannofossils, foraminifers, silicoflagellates, and sponge spicules) components and minor volcaniclastic components. Dropstone occurrence is common—indicating a peak in siliciclastic deposition that has also been observed at Leg 145 sites—and coincides with the beginning of NHG. However, the dramatic drop in paleoproductivity recorded at Site 882 (Haug et al., 1999) is not present at the Bowers Ridge sites where, conversely, the biogenic component is high throughout the late Pliocene and Pleistocene.

All sites drilled during Expedition 323 preserve a record of sedimentation ranging from the late Pliocene through the Holocene (between 1.74 Ma and recent). Lithologies and sedimentation rates vary between the different sites, as indicated by a basin-wide comparison of the evolution of sedimentation in the Bering Sea during this period (Fig. F6). The slowest sedimentation rates (only 4.5 cm/k.y.) were observed at Site U1342, where laminated foraminifer-rich diatom ooze beds alternate with silty clay beds at scales ranging in the Milankovitch band. The same temporal interval corresponds to a much thicker section at Sites U1340 and U1341, where the bedding alternations are less distinct and the abundance of IRD is higher. Although lamination is common at Sites U1342 and U1340, laminae are virtually absent at the deeper Site U1341, suggesting a shallowing of the depth of the OMZ during most of the Pleistocene. At the Bering Sea slope site, sedimentation rates are about three times higher than at the Bowers Ridge sites. At Sites U1339, U1343, and U1344, siliciclastic-rich beds (mostly diatom-rich silt clay) and mixed siliciclastic-biogenic beds (clay, silt, and diatom ooze with varying abundance of foraminifers, nannofossils, and sponge spicules) alternate rhythmically. The sections are pervasively bioturbated and laminated intervals are rare. Overall, sedimentation on the Bering slope is characterized by higher influence of both siliciclastic material delivered by ice sheets as well as terrigenous sedimentation derived from the continental shelf and slope, which are indented by some of the largest submarine canyons in the world. However, because of their proximity to the continents, it is not clear whether the sediments characterized by high siliciclastic content are recording periods of ice sheet expansion (stadials) or increased runoff (interstadials). IRD is a common feature at all sites during this time period and increases significantly around 1 Ma, as is also observed in coeval sediments from the North Pacific based on the results of Leg 145 (Krissek, 1995).

Cyclical siliciclastic deposition and its relationship to climate change

Expedition 323 in the Bering Sea is the first expedition to recover deep continuous sections of sediment, providing an opportunity to study sediment distributions in time and space in the Bering Sea. The plethora of climatic and sedimentary structure information in the cores will allow us to develop an understanding of the influence of climate changes on sedimentation. Clearly, siliciclastics are a dominant component of all Bering slope sites and are less dominant at the Bowers Ridge sites. At all sites, because the sediments are primarily comprised of mixtures of siliciclastics and diatom frustules, the physical magnetic properties and natural gamma ray (NGR) measurements made on the tracks and with the logging tools generally provide information on the relative proportions of clays/silts versus diatoms (although ash is a significant component of the sediment at Site U1339). The NGR records (Fig. F6) indicate clear cyclicity; the amplitude and wavelength change markedly with depth at the Bering slope site but not at the Bowers Ridge sites. Specifically, at the Bering slope sites (Sites U1339, U1343, U1344, and U1345) in sediment younger than ~1 Ma the NGR record appears to vary with lower amplitude and across a wide range of frequencies higher than 1/40 k.y., than it does in sediment older than 1 Ma (Fig. F6; see NGR records of Sites U1343 and U1344). Because Sites U1339 and U1345 are short records younger than 1 Ma, longer term trends in the character of the variability can only be evaluated in the longer NGR records of Sites U1343 and U1344; however, the relatively high frequency variability of the NGR data from the younger sediment is evident at all four sites.

This change in the character of the variability is also apparent in the logging data from Sites U1343 and U1344 (Fig. F7). The downhole logging data provide our only opportunity to derive continuous records of lithologic variability in the deeper sections where only one hole was drilled and where continuous composites of data generated on the cores cannot be spliced together as they are in the sections with multiple drill holes. The potassium concentrations (K%) measured by the logging tool are generally related to the terrigenous clay concentrations, although variations in mineralogy of the terrigenous fraction could have a secondary influence. In this sedimentary environment, it appears that NGR represents the amount of terrigenous material relative to biogenic material in the sediment. Although much more research is needed, the shipboard data indicate that high siliciclastic relative to biogenic deposition probably occurred during colder periods or possibly during deglaciations.

Applying the shipboard age model to the K% record at Site U1343 and performing spectral analyses on the record (not shown) indicate that variability in K% dominantly occurs with 40 k.y. periodicity in periods older than ~1 Ma. In the portion of the record younger than 1 Ma, the variance is dispersed among many frequencies; 40 k.y. periodicity does not dominate the spectrum. Benthic δ18O records (Lisiecki and Raymo, 2005) indicate that variability of ice volume and therefore of eustatic sea level occurs primarily with 40 k.y. periodicity prior to ~1 Ma (referred to as the "40 K world"). As such, it appears that sea level may control the timing of pulses of sedimentation at the Bering slope sites. Sea level highstands in warm periods may be times when deposition of sediments on shelves occurred at the expense of sediment deposition on the slope. As ice volume increased, sea level fell, the coastline moved seaward, and the flux of terrigenous material to the slope may have increased. This simple explanation of how sea level controlled long-lived pulses of sediment at Sites U1343 and U1344 may explain how global climate change indirectly controlled the timing of sediment deposition at the slope sites during the 40 K world. After 1 Ma the variability of the sediment composition changed, as reflected by logging K%. NGR core measurements do not appear to be simply related to sea level change, which occurred dominantly with ~100 k.y. variability; rather, the pulses of sedimentation appear to occur with much higher frequency. It is unclear why this happened, but the increase in sea ice diatom and sea ice dinoflagellate forms (discussed in the next section) at ~1 Ma may indicate that high-frequency variability in siliciclastic deposition related to sea ice processes may be an important process—perhaps more important than global ice volume and concomitant sea level changes, which occur with lower frequency.

A close look at magnetic (Fig. F8) and density (not shown) data provides additional details about processes that may dictate terrigenous deposition. At first glance the record of relative magnetic paleointensity of the sediment at Sites U1343 and U1344 agrees with the K% records and indicates a change in depositional style and cyclicity at ~1 Ma (Fig. F8), with lower frequency and higher amplitude changes prior to 1 Ma at the Bering slope sites. Natural remanent magnetization (NRM)/Chi is an index that attempts to normalize intensity by the concentration of magnetic grains. However, changes in grain size distributions may also impact this ratio; low NRM/Chi values indicate that magnetic grains are coarser than those in sediment with higher NRM/Chi values. Grain-size differences can be related to a variety of climatically controlled factors as well as to diagenetic dissolution of the smallest clay-sized grains. The fact that NRM and Chi each show distinct variability with the same length scale representing 40 k.y. cycles suggests that the primary factor controlling the magnetic susceptibility and paleointensity records is related to the relative amount of terrigenous material. This also indicates that diagenetic overprints may alter but not obliterate the primary signal. The fact that the amplitude of changes in the paleointensity record (Fig. F9) downcore are even more extreme than those seen in the NGR and K% data suggests that, if anything, diagenetic overprints may amplify the contrast between sediments with varying amounts of siliciclastics grains.

After 1 Ma, during the time period of large 100 k.y. ice volume cycles, gamma ray attenuation (GRA) bulk density records vary across a spectrum of frequencies. However, they more clearly show long-period variance characteristic of the 100 k.y. ice age cycles than they do other physical parameter properties. For example, after 1 Ma paleointensity (Fig. F8) and NGR (Fig. F6) variability have power distributed across low (orbital) and suborbital frequencies. This indicates that terrigenous deposition at the Bering slope sites in the late Pleistocene was likely a combination of many processes such as bottom water current deposition of fine-grained material (drift), IRD from sea ice and icebergs, and mass sediment transfer from the shelf to the lower slope and abyss. Superimposed on this are possibly independent or nonlinearly related changes in biogenic fluxes and postdepositional processes such as diagenesis and precipitation of authigenic minerals. The periodic and stochastic forces that drive each of these processes could be different, resulting in rich records of lithologic variability in the latest Pleistocene.

On Bowers Ridge at Sites U1340, U1341, and U1342, the trends and variability in siliciclastics appear to be notably different than at the Bering slope sites, as indicated by the NGR (Fig. F6), K% (Fig. F7), and NRM/Chi (Fig. F8) records. Unlike the Bering slope sites, there are some marked long-term trends in the Bowers Ridge records. A notable change in Site U1341 logging K% data occurs at ~425 m depth (Fig. F7) along with an increase in the observed silt content and dropstones in the sediment in smear slides (Fig. F6). This level in the section is equivalent in time to ~3 Ma and may signify that the Bering Sea regional expression of NHG includes an increase in clays and silts delivered via sea ice or icebergs to Bowers Ridge. The fact that K% decreases above this level in the section even while the ice ages persisted is difficult to explain. One possibility is that the warmer climate of the Pliocene facilitated the formation of K-rich clays on land that were shed into the Bering Sea once the ice ages began. K-rich clay formation could have then decreased when the climate cooled. The NGR, magnetic susceptibility, and sedimentological smear slide and descriptive data indicate increasing amounts of terrigenous material upsection with a pronounced increase at ~1 Ma, probably due to a greater supply of ice-rafted siliciclastics to Bowers Ridge once large 100 k.y. glaciations occurred.

As at the Bering slope sites, the variability of physical properties in sediments younger than 1 Ma at Bowers Ridge appears to include higher frequency variability than the 100 k.y. ice volume cycles. Again, higher frequency climate variability, including changes in the distribution of sea ice and icebergs delivering sediment to Bowers Ridge, may dominate the mode of sediment deposition in the Bering Sea after 1 Ma. However, unlike at the Bering slope sites the variability of all lithologic parameters studied—NGR, magnetic susceptibility, and GRA bulk density—in sediment older than 1 Ma at Bowers Ridge is not dominantly paced by 40 k.y. ice volume cycles. One explanation is that even in the 40 K world sediment delivery to Bowers Ridge was paced at least in part by processes other than sea level change. Spectral analyses of K% data from Site U1341 at Bowers Ridge indicate that variability occurs across a spectrum of frequencies, with notable concentrations of variance at 40 k.y. and 23 k.y. periodicities. During the 40 K world, cycles and episodes of deposition of IRD on Bowers Ridge were possibly impacted by regional climate changes responding to both obliquity (40 k.y.) and precession (23 k.y.) solar radiation changes. In contrast, at the Bering slope, sea level variations that occur dominantly with 40 k.y. periodicity were the main factor controlling the transport of sediment from the continent/shelf to the slope. The observation that high-latitude climate was responding to precession forcing during the 40 K world has important implications, not only for theories that explain ice volume changes, but ultimately for our understanding of ice sheet dynamics and our ability to predict ice sheet response to radiative forcing.

History of sea ice development

The investigation of the evolutionary history of climate and surface ocean conditions is primarily approached by studying microfossils, sedimentology, and physical properties of sediments. This is particularly true for the influence of sea ice formation and the flow of the relatively warm oligotrophic Pacific surface waters into the Bering Sea.

One of the most striking findings of Expedition 323 is the general sea ice distribution history of the Bering Sea for the past 5 m.y. As described in "Overview of ages and sedimentation rates," the first sign of sea ice is the presence of pebbles, which are thought to be transported as IRD starting at >3.8 Ma at Site U1340 (Fig. F6), indicating the formation of sea ice or iceberg transportation to the Bowers Ridge region. The bulk of the evolution of sea ice distribution has been interpreted from shipboard analyses of sea ice diatoms and sea ice–related dinoflagellate taxa and to a lesser extent by other diatom taxa and intermediate water–dwelling radiolarians (see below for details).

The details of sea ice evolution are derived from changes in the relative abundance of sea ice diatom taxa, which are represented mainly by Thalassiosira antarctica spores (Abelmann, 1992a) and sea ice dinoflagellates (Fig. F9). The first signs of sea ice diatoms and dinoflagellates are subtle increases in their abundance at Bowers Ridge at Site U1340, starting at ~3.4 Ma for dinoflagellates and ~2.7 Ma for diatoms, coincident with NHG (Maslin et al., 1996). After ~2 Ma the sea ice assemblage signals become progressively stronger into the present, up to values of ~10%–20%. In contrast to the Bowers Ridge sites, sea ice cover at the Bering slope sites is markedly severe, indicated by sea ice assemblage percentages significantly higher than the values observed at Bowers Ridge. Sea ice diatom values range from ~10% to 50% during the latest Pliocene and increase from ~30% to 70% during the Pleistocene. The abundance of sea ice dinoflagellates also increases but not to the same extent as sea ice diatoms. Notably, sea ice diatom and sea ice dinoflagellate assemblages clearly show a significant increase in both abundance and amplitudes of variability at around the mid-Pleistocene Transition (MPT) (Fig. F9). Analogous to sea ice–associated diatom and dinoflagellate taxa, a clear increasing trend in abundance of intermediate water–dwelling radiolarian taxa at the MPT is also observed at Sites U1343 and U1344. This is consistent with the interpretation that the surface water became gradually more affected by the formation of sea ice as climate progressively cooled; in the presence of sea ice, surface-dwelling radiolarians disappeared, and, as a consequence, the relative percentages of intermediate water dwellers such as Cycladophora davisiana were higher (Abelmann, 1992b; Okazaki et al., 2003).

At Sites U1343 and U1344, which are located closer than the Bowers Ridge sites to the southern boundary of today's seasonal sea ice maximum extent, a dramatic change in the dominance of dinoflagellate cyst assemblages from autotrophic to heterotrophic taxa is evident at ~1.2–1.5 Ma. This suggests that sea ice formation occurred well before the time when the abundance of sea ice taxa significantly increased at ~1 Ma. Heterotrophic dinoflagellates feed mainly on diatoms (Jacobson and Anderson, 1986) and their abundance is most likely the result of food availability; hence, they can persist in ice-laden environments. On the other hand, the low production of their autotrophic counterparts, which dwell mainly in open waters without sea ice, could be related to these taxa being out-competed in the presence of blooming diatom populations in regions of seasonal sea ice cover. Furthermore, heterotrophic dinoflagellate cysts dominate the assemblages in polar areas where sea ice occurs up to 12 months per year (e.g., Rochon et al., 1999). The sea ice–associated species Islandinium minutum, known to be very abundant or dominant in the Arctic Ocean and subarctic regions with pronounced sea ice cover (Head et al., 2001; Hamel et al., 2002), appeared to increase at ~1 Ma at both Sites U1343 and U1344 (Fig. F9). Along with significant increases of both sea ice dinoflagellates and sea ice diatoms (e.g., Thalassiosira antarctica spores) at ~1 Ma at both sites, a rather significant drop in the abundance of Neodenticula seminae, a typical pelagic diatom, is also seen. All of these biotic events are within the time interval of the MPT, which spans from ~1.2 to 0.8 Ma and marks the transition from 41 k.y. obliquity ice volume cycles to larger ice age cycles that vary at ~100 k.y.

As noted above, the Bowers Ridge and Bering slope regions show distinct differences in the extent of sea ice cover throughout the last ~2.1 m.y., the time period covered by the sedimentary records of both regions. The extent of sea ice cover of the latter is substantially greater than that of the former because of the distal locations of the three Bering slope sites, which are most prone to perennial sea ice cover in the Bering Sea. The spatial differences in sea ice cover today are mainly attributed to the surface water circulation pattern (Fig. F2); this spatial difference appears to have persisted since at least 2.1 Ma, implying that the surface water circulation patterns have also persisted (see below for details with N. seminae and other taxa). This is further supported by the fact that at Site U1339 on Umnak Plateau, the site closest to an entry point of Alaskan Stream subarctic Pacific water, the sea ice taxa suggest lower levels of sea ice cover compared to the Bering slope sites farther north, which are not as heavily impacted by Alaskan Stream water. Furthermore, Katsuki and Takahashi (2005) illustrated that during the low sea level stand of the Last Glacial Maximum, the relatively warm subarctic Alaskan Stream water entered into the Bering Sea through the western Aleutian passes rather than the eastern passes because of their shallow water depth. After entry it turns eastward, flowing counter-clockwise around Bowers Ridge, and consequently reaches the southeast corner of the Bering Sea, close to Site U1339 on Umnak Plateau, before turning northwest. It is surprising that the spatial pattern of sea ice and surface water circulation observed in the last glacial period, including the stronger influence of warm Alaskan Stream water at Umnak Plateau compared to at the northern Bering slope, persists over long timescales.

Changes in biological productivity and the influence of subarctic Pacific surface water mass flowing into the Bering Sea

Based on the spatial distributions of long-term temporal changes of three diatom taxa (Coscinodiscus marginatus, Neodenticula, and Actinocyclus curvatulus), it is clear that the influence of subarctic Pacific waters, which are relatively warm and less eutrophic than Bering Sea waters, has typically been strongest at the Bowers Ridge sites, followed by the Umnak site; the weakest influence of this warm water mass has occurred at the Bering slope sites (Fig. F10). This is the same pattern found by Katsuki and Takahashi's (2005) study of past water mass circulation patterns, which they inferred from sea ice distributions from the last glacial period. The longer records from Expedition 323 indicate that as climate cooled through the Pleistocene, pelagic water influence at all the sites progressively weakened. Furthermore, the sites closest to straits through which pelagic water flows into the Bering Sea have consistently higher abundances of subarctic diatom species than those downstream in the counter-clockwise circulation pattern of the surface water masses.

From the bottom of the holes upward in the Bowers Ridge sites, a marked drop in Coscinodiscus marginatus was seen at ~3 Ma at Site U1341 and at ~2.6 Ma at Site U1340 (Fig. F10). This can be interpreted as resulting from a sharp reduction in supply of nutrients due to the development of upper layer stratification. It is apparent that the diatom taxon C. marginatus requires a relatively high nutrient supply and tolerates low light intensity. This is substantiated by the fact that (1) today this diatom taxon dwells in the lower euphotic zone off Spain (Nogueira et al., 2000; Nogueira and Figueiras, 2005), and (2) it occurs during early winter (~November–January) in the subarctic Pacific and the Bering Sea based on time-series sediment trapping (Takahashi, 1986; Takahashi et al., 1989; Onodera and Takahashi, 2009). Thus, high demand of nutrient supply and low demand of light intensity is apparent for this taxon. This timing of 3–2.6 Ma coincides approximately with the so-called end of "opal dump" observed in the subarctic Pacific at ~2.7 Ma, which is coincidental with the onset of NHG (Maslin et al., 1996). Although the reduction in C. marginatus around the time of NHG persisted, an overwhelmingly continuous presence of diatom ooze and interbedded diatom ooze and silt sediments accumulated throughout the Pliocene–Pleistocene in the Bering Sea. This clearly suggests that a high amount of opal sedimentation continued after the onset of NHG well into the Pleistocene.

The 5 m.y. long-term trend of Neodenticula (N. kamtschatica, N. koizumii, N. seminae, and Neodenticula sp.) in the Bowers Ridge region shows the following two patterns. Generally higher percent values are observed from the hole bottoms up until ~2.9–2.7 Ma, depending on the hole, than are observed for the sections above. From there onward there seems to be a decline in Neodenticula with sizable fluctuations that indicate surface water stratification progressively developed as the climate cooled from the Pliocene into the Pleistocene. The general antiphase relationship of this taxon versus Actinocyclus spp. (Fig. F10) suggests subtle and perhaps periodic changes in stratification/water mass circulation as the colder climate developed. This is generally true for Sites U1340 and U1341 at Bowers Ridge and Site U1343 off the Bering slope. The data from Site U1344, however, provide a more complicated picture than this and do not show a clear antiphase relationship trend, probably because of the overwhelmingly high abundance of sea ice–associated diatoms such as Thalassiosira antarctica spores (Abelmann, 1992a) (~25% during 4–0.9 Ma; ~40% during 0.9–0 Ma) (Fig. F10). Furthermore, an even more clearly visible antiphase relationship of Actinocyclus spp. versus N. seminae is seen during ~1.9 Ma and 1.1 Ma at Site U1343. Their percentages increased at ~1.9 Ma. Marked declines are seen for both taxa at slightly different times (Actinocyclus spp. at ~1.1 Ma and N. seminae at 0.9 Ma), both of which are around the MPT (Fig. F10). The upheaval of both taxa during this time interval is identical. During intervals of high eutrophic levels suggested by high Actinocyclus spp., cooler, low trophic subarctic Pacific waters, indicated by N. seminae, were introduced into the Bering Sea. This is based on the geographic and seasonal distributions of Actinocyclus curvatulus, which appears to be ecologically similar to Thalassiosira trifulta and, to a lesser extent, N. seminae (Sancetta, 1982; Takahashi, 1986; Takahashi et al., 1989). As surface waters became increasingly stratified, especially after ~0.9 Ma with Milankovitch-scale 100 k.y. climatic cyclic regimes, N. seminae declined with the emerging sea ice diatoms.

Changes in bottom and intermediate water conditions

Much of the Pacific water entering the Bering Sea is matched by outflow through the Aleutian Islands. The most significant outflow is through the Kamchatka Strait, which has a maximum water depth of 4420 m. If some component of North Pacific Intermediate or deep water formed in the Bering Sea in the past, particularly when sea level was lower, it would have flowed out through the Kamchatka Strait or a secondary outlet near the Commander–Near Strait at 2000 m. One of the scientific objectives of Expedition 323 was to elucidate the history, temporal variability, and intensity of NPIW and deep water formation in the Bering Sea and its links to surface water processes. This objective was partially fulfilled during the expedition but would have been more adequately addressed had we obtained permission to drill two important sites located in Russian territorial waters, Shirshov Ridge Site SHR-3B at 2232 m water depth and Site KST-1B inside the Kamchatka Strait at 3435 m water depth. The main objective at these sites was to monitor deep water masses flowing out of the Bering Sea to the Pacific Ocean. Nevertheless, seven sites were successfully cored and complete records of environmental variability of intermediate and deep water sites spanning the Pleistocene to the Pliocene were recovered during Expedition 323. The sites ranged from 818 to 3174 m and allow for the characterization of past vertical water mass distribution and for the reconstruction of the history and distribution of the OMZ in the region (Fig. F4).

Previous observations made in glacial records from the Bering Sea and just outside the Bering Sea on the Detroit Seamount in the North Pacific suggest a source of ventilated intermediate water coming from the Bering Sea and/or the Detroit Seamount region (Gorbarenko, 1996). Furthermore, other paleoceanographic reconstructions of the Bering Sea, made possible by the R/V Hakuhou-Maru piston coring survey cruise in 1999, show past NPIW formation during four different time slices, reconstructed based on the high oxygen indicator and intermediate water–dwelling radiolarian species Cycladophora davisiana. According to these studies, the role of the Bering Sea in NPIW formation is apparently visible during the cold intervals.

Shipboard analyses of sediment samples recovered during Expedition 323 show continuous recovery of Pliocene to Holocene deep sea benthic foraminifers and midwater radiolarians at all sites (Figs. F11, F12), although calcareous benthic foraminifers appear to be rare in the middle and early Pliocene. The benthic foraminifer faunal composition displays large assemblage changes and shifts in species dominance. These changes are likely related to variability in local oxygen concentration in the bottom waters associated with surface water productivity and/or deepwater ventilation on Milankovitch timescales. For example, Bulimina aff. exilis, a common species in the Bering Sea samples, is generally regarded as a low oxygen/deep infaunal species and has been found in samples associated with high productivity and low sea ice (e.g., Bubenshchikova et al., 2008; Kaiho, 1994).

Bering Sea benthic foraminifers show affinities to assemblages found within or near the OMZ in the Sea of Okhotsk and also to more common deep Pacific Ocean assemblages. The faunal similarity between the two marginal seas and the deep Pacific Ocean will allow for the reconstruction of the physico-chemical characteristics of deep and intermediate water masses as a result of changes in surface productivity, deepwater ventilation, sea ice coverage, and continental glaciation.

Previous piston core studies showed a large increase in the intensity of the OMZ during the last deglacial at Umnak Plateau (Okazaki et al., 2005), suggesting a relationship between productivity and terrestrial nutrient supply from melting. However, there is no information regarding the longer timescale relationship through the Pleistocene. Analysis of faunas from the newly drilled Bering Sea sites will be particularly important in extending this record through the entire Pliocene (at Bowers ridge) and Pleistocene (at Bowers ridge and the Bering slope) and will allow us to decipher the onset and evolution of the OMZ and provide further insight into NPIW production at the marginal sea. Furthermore, Site U1344 at ~3200 m is presently located below the OMZ and also has the potential to monitor past deepwater changes.

A striking finding of the expedition was the relatively low oxygen content of intermediate water mass conditions at most sites during the last 5 m.y., as indicated by the presence of episodic laminated sediment intervals throughout the entire sections. However, the benthic foraminiferal faunas recovered from deep cored Sites U1340 and U1341 suggest that in general the Bering Sea intermediate and deep waters were better oxygenated in the early Pliocene, as suggested by the presence of the agglutinated foraminifer Martinottiella communis (Fig. F12), and less oxygenated during the last ~2 m.y., as suggested by the increase in deep infaunal benthic foraminifers (Fig. F11). This assemblage is composed of abundant calcareous benthic genera (e.g., Bulimina, Globobulimina, Islandiella, Nonionella, and Valvulineria) that are typically indicative of reduced oxygen conditions (Bubenshchikova et al., 2008).

High sediment accumulation rate Sites U1339 and U1345, located within the OMZ today (Fig. F4), reveal high-amplitude variability in the relative abundance of the deep infaunal assemblage since 0.8 Ma. This appears to be associated with interglacial–deglacial cyclicity, showing higher abundance of deep infaunal species (reflecting the lowest bottom water oxygen conditions) during interglacials, in particular strong interglacials marine isotope Stages (MIS) 1, 5, and 11. Higher bottom water oxygen conditions appear to correlate with glacials. Furthermore, Sites U1340, U1343, and U1344 contain well-preserved foraminifers over the last 2 m.y. with increasing absolute abundances of benthic and planktonic taxa across the MPT (~0.8–1.1 Ma) in association with an increase in abundance of the polar planktonic foraminifer Neogloboquadrina pachyderma sinistral. This cooling trend was also observed as an increase in the abundance of sea ice dinoflagellates and diatoms and intermediate water–dwelling radiolarians (e.g., C. davisiana). Cooling of the surface waters would have enhanced ventilation of the intermediate waters during glacials and increased density stratification during interglacials, contributing to the drop in oxygen content in the intermediate and bottom waters at these times, which is suggested by the increase in deep infaunal benthic foraminifer taxa.

Microbiology in high surface productivity environments

The microbiological objectives of Expedition 323 were to constrain global models of subseafloor biomass and microbial respiration by quantifying subseafloor cell abundance and pore water chemistry in an extremely high productivity region of the ocean. We also sought to determine how subseafloor community composition is influenced by high productivity in the overlying water column.

To meet these objectives, high-resolution sampling for microbiological analyses and pore water chemistry took place at five sites throughout the Bering Sea. Each site was selected based upon its distance from land and its levels of marine productivity determined by annual chlorophyll-a concentrations in the water column.

Four cores were drilled using the APC system to ~40 m CSF at each microbiology-dedicated hole. Contamination tests using perfluorocarbon tracer (PFT) and fluorescent beads (Smith et al., 2000) were performed in each hole, showing that contamination from the drill fluid was insignificant.

Sediment samples were taken for determination of cell abundance measurements (a proxy for microbial biomass) and diversity and structure of the microbial community. In general, these samples were taken every 25 cm for Sections 1H-1 to 2H-3 (or 2H-6), and every 75 cm for Core 3H. Below Core 3H, the sampling resolution was one per core. Interstitial water whole-rounds were taken with the same resolution adjacent to microbiology whole-rounds. In addition, 10 cm whole-rounds were taken for interstitial water and microbiological analyses at an interval of one per core until APC refusal. Interstitial water sampling continued at this resolution in XCB cores, and microbiology samples for cell abundance, diversity, and community structure were collected at ~400 and ~700 m CSF. PFT analyses showed no drill-fluid contamination at these depths.

Interstitial water samples were squeezed immediately on board and aliquots were fixed for both shore-based and shipboard analysis. The shipboard aliquots were used to determine concentrations of dissolved inorganic carbon (DIC), alkalinity, sulfate, sulfide, ammonium, phosphate, major ions (e.g., Ca, Na, and K), and minor ions (e.g., Fe and Mn). These data along with formation factor data allow for a quantitative determination of microbial respiration rates (D'Hondt et al., 2002).

The geochemical data obtained during the expedition show that the microbial activity along the slope sites (i.e., Sites U1339, U1343, U1344, and U1345) is substantially higher and more diverse in terms of respiration pathways than at Site U1342 at Bowers Ridge. At the slope sites, the concentrations of microbial respiration products such as DIC, ammonium, and phosphate are approximately an order of magnitude higher than at Bowers Ridge (Fig. F13). A shallow sulfate–methane transition zone (SMTZ) (~6–11 m CSF) is also present (Fig. F13), indicating that both methanogenesis and sulfate reduction based on methane oxidation occur in these sediments. Interstitial water data further suggest the presence of microbial-mediated Fe and Mn reduction. Interestingly, the geochemical profiles even suggest significant microbial activity as deep as 700 m CSF. In contrast, at Bowers Ridge sulfate penetrates to the basement and is almost unaltered with depth, suggesting only very low rates of microbial-mediated sulfate reduction. Methane is below the detection limit. The difference in microbial activity at these sites might be caused by different rates of water column productivity and sedimentation.

We expect that the differences in the geochemical parameters between the slope and ridge sites will be reflected in microbial abundance and diversity. A larger and more diverse microbial community at the slope sites is likely. Specifically, we expect elevated cell abundance and a consortium of bacteria and archaea at the SMTZ. At the slope sites, geochemical profiles suggest that methanogens, iron reducers, manganese reducers, and sulfate reducers exist throughout the sediment column. At Bowers Ridge, geochemical profiles indicate that only sulfate reducers and nitrate reducers are present.