Since 1997, ocean color satellite images have revealed large-scale blooms of the coccolithophorid Emiliania huxleyi in the eastern Bering Sea. The blooms are often sustained over several months and have caused ecosystem changes in the Arctic Ocean, as well as in the Bering Sea. We examined continental shelf sediment profiles of alkenone, a biomarker for E. huxleyi, covering the past ∼70 years. The alkenone records suggest that large E. huxleyi blooms are a novel feature in the Bering Sea as they have occurred only since the late 1970s. Recent changes in alkenone content were closely related to the 1976–77 climatic regime shift in the North Pacific, implying that warming and freshening of Bering Sea waters promoted E. huxleyi blooms. The production rate of diatoms (total valves in sediment samples), the dominant primary producers in the Bering Sea, also increased during the past several decades. However, the ratio of alkenone content to total diatom valves in the sediments increased as E. huxleyi production increased, suggesting that the increase in the E. huxleyi production rate frequently exceeded the increase in the diatom production rate. Overall, our results indicate a possible subarctic region ecosystem shift driven by climate change.
 The recent retreat of sea ice in the western Arctic Ocean is partly due to the inflow of warm Pacific water through the Bering Strait [Woodgate et al., 2006], an important gateway connecting the North Pacific with the Arctic Ocean, and partly to atmospheric forcing [Overland et al., 1999]. The Bering Sea, which has been called the “sea of silica” because of its high silica content and high diatom production [Tsunogai et al., 1979], is both a substantial sink for atmospheric CO2 and the largest biological pump in the Northern Hemisphere [Takahashi et al., 2002]. High primary production is maintained by an efficient nutrient supply from the deep Bering Sea and North Pacific into the euphotic zone through upwelling, tidal mixing, transverse circulation, and eddies [Mizobata et al., 2002; Mizobata and Saitoh, 2004; Springer et al., 1996]. The seawater temperature record of the 14 years from 1994 to 2008 indicates that, on average, stratification begins to set up on the middle shelf (57–58°N) of the eastern Bering Sea each year about the middle of May [Hunt et al., 2010]. Thus, a substantial supply of nutrients is found in spring on the outer shelf beneath a mixed layer of ∼75 m, and the wintertime intrusion of nutrients onto the middle shelf continues until stratification becomes established [Hunt et al., 2010]. Little interannual variation in the seasonality of the nutrient supply to the outer and middle shelf occurs, but the nutrient dynamics (the extent of nutrient re-supply in winter) on the inner shelf are not well known [Hunt et al., 2010].
 Since the late 1980s, the Bering Sea ecosystem, especially the standing stock of benthic organisms, has changed substantially [Grebmeier et al., 2006], and the impact of these changes is already being felt on the Pacific side of the Arctic Ocean [Grebmeier et al., 2010]. However, their impacts on the Bering Sea's status as a “sea of silica” and on its functions as a CO2 sink and biological pump are not known. Since 1997, when data from the NASA SeaWiFS ocean color sensor first became available [Sukhanova and Flint, 1998], large-scale blooms of the coccolithophorid Emiliania huxleyi (the only known haptophyte species forming blooms in the region) lasting for several months have been observed over the eastern continental shelf of the Bering Sea [Stockwell et al., 2001]. A combination of summer weather anomalies over the Bering Sea – unusually calm winds, clear skies, and warm air temperatures – resulted in nutrient depletion of the subpycnocline nutrient reservoir in 1997 and has been linked to the blooms in that year [Harada et al., 2003; Napp and Hunt, 2001]. Sustained E. huxleyi blooms were also observed in 2000 and 2006, possibly because light penetration into the stratified surface and subsurface waters was sufficient for bloom development. In 2000, this stratification was maintained by a large cold water mass on the continental slope [Shin et al., 2002]. The response of the Bering Sea ecosystem to the changes in atmospheric conditions (e.g., calm winds, clear skies, and warm air temperatures) suggest that the E. huxleyi blooms may have occurred in response to local or global atmospheric forcing. Ecological factors, such as reduced grazing pressure by microzooplankton, may also contribute to the blooms [Olson and Strom, 2002]. If large-scale E. huxleyi blooms become more frequent in the Bering Sea, or if the dominant primary producer switches permanently from diatoms to coccolithophorids, the sea's efficiency as a biological carbon pump will be reduced because assimilation of atmospheric CO2 by photosynthesis would be partly canceled out by coccolith production [Murata, 2006].
 The goals of the present study were (1) to elucidate recent environmental changes causing the E. huxleyi blooms in the Bering Sea, (2) to examine the possible effects of climate change on the dominant primary producer, and (3) by molecular characterization of the E. huxleyi blooms to explore how the ecosystem changes in the Bering Sea might affect Arctic and Atlantic ecosystems. We used records of C37 unsaturated alkyl ketone (alkenone), a biomarker for E. huxleyi in bottom sediments to investigate the spatial and temporal extent of E. huxleyi blooms on the eastern continental shelf of the Bering Sea, and we also compared how E. huxleyi and diatom production changed in response to the changes in environmental conditions.
2. Materials and Methods
2.1. Sediment Cores
 Sediment cores (one in August 2000 and 11 in August–September 2006) were collected from the eastern Bering Sea with a multiple core sampler capable of extracting eight sub-cores simultaneously. The eight sub-cores collected at each site were used for various chemical and biological (diatom assemblage) analyses, with the longest sub-core being used for organic matter analysis (Figure 1 and Table 1). A coccolithophorid bloom-specific algorithm [Iida et al., 2002] was used to identify large coccolithophorid blooms over the middle and inner continental shelf on SeaWiFS images (Figure 1). This algorithm allows accurate estimation of the spatial and temporal distributions of coccolithophorid blooms, although areas of “bright” color observed by SeaWiFS are not always indicative of coccolithophorid blooms because the water color produced by substantial empty and broken diatom frustules can mimic the appearance of E. huxleyi blooms [Broerse et al., 2003]. This algorithm was used to over come this problem.
Table 1. Sediment Sampling Site Locations and Water Depth in the Bering Sea
 Sub-cores sliced at 1-cm intervals were freeze-dried and ground into a homogeneous powder with a mortar and pestle. Dried samples weighing 2–6 g were packed into plastic vessels for excess 210Pb (210Pbex) and 137Cs radionuclide activity analysis. Radionuclides were measured with an IGW 14023–16 high-performance germanium well detector (Princeton Gamma Tech, Rocky Hill, NJ, USA) and a Quantum 8000 multichannel gamma-ray analyzer (Princeton Gamma Tech) as previously described [Oguri et al., 2012]. The plastic vessels containing the powdered sediments were stored for more than two months to allow radioactive equilibrium to be established between 226Ra and 222Rn before analysis. Total 210Pb and 137Cs concentrations were determined by measuring their characteristic gamma-ray energy peaks at 46.5 and 661.6 keV, respectively. To calculate 210Pbex, 214Pb (351.9 keV) was measured and subtracted from the 210Pb concentration under the assumption that radioactive equilibrium of the uranium decay series had been established in the sediments. Count times were from 180 000 to 360 000 s. The Quantum MCA software package (Princeton Gamma Tech) was used to analyze the spectra.
2.3. Analysis of Organic Carbon, Total Nitrogen, and Alkenone
 To determine organic carbon (Org. C) and total nitrogen (TN) contents, bulk sediment samples sliced at ∼1 cm intervals were freeze-dried and ground into a homogeneous powder. Approximately 10–20 mg of each sample was placed in a tin capsule for total carbon (TC) and TN measurements, or into a silver capsule for Org. C measurements. The Org. C samples were decalcified with concentrated HCl vapor for 8 h, then neutralized in an atmosphere of granular NaOH in a desiccator for a few days prior to analysis. All samples were analyzed with an elemental analyzer (CHN Analyzer NCS2500, Thermo Quest). The standard errors of the mean of replicate samples for Org. C and TN were ≤2% and ≤6%, respectively. We calculated the marine organic carbon (MOC) content of each sediment sample from its total Org. C content and C/N mole-ratio (C/N ratio). The C/N ratio can be used to discriminate between marine [Libes, 1992] and terrigenous organic materials [Guo and Macdonald, 2006] in a sediment sample because it differs according to source. We used the average of typical C/N ratios of marine primary producers (5–7) [Libes, 1992] and the mass balanced mean C/N ratio of the terrestrial organic matter (37.1) transported by the Yukon River (C/N ratios: colloidal organic carbon, 46.8; low-molecular-weight dissolved organic carbon, 26.7; and particulate organic carbon, 26.7) [Guo and Macdonald, 2006] to estimate the MOC content of each sample.
 In the present study, the C37 alkenone content normalized by the MOC content served as a biomarker for coccolithophorids. Bulk organic compounds were extracted from 1 to 3 g of freeze-dried sediments with an accelerated solvent extractor (ASE-200, DIONEX Japan Ltd.). The extracts were saponified, and the neutral fraction was separated by silica gel column chromatography using an automatic solid-preparation system (Rapid Trace SPE Workstation®, Zymark Corp.). The solvents and sub-fractionation steps were as previously described [Harada et al., 2003]. Each alkenone fraction was analyzed by capillary gas chromatography (Agilent Technologies, Inc.), and C37 alkenone content was calculated as the sum of C37:2, C37:3, and C37:4. The standard error of the mean of the replicate sediment samples for alkenone concentration was 5%. The supply of terrestrial organic materials derived from the Yukon River on this study area of the shelf would be limited, because substantial terrestrial organic materials from the Yukon River deposits on only close to the river mouth and in Norton Sound [McManus et al., 1974; Nelson and Creager, 1977]. From the C/N ratio, we estimated that 68–100% of Org. C in the shelf sediments was of marine origin (see auxiliary material). Although some of the terrestrial organic materials are likely to be present in the shelf sediments based on the C/N ratio, no peaks reflecting terrestrial influences interfered with the C37 alkenone peaks on the chromatogram. Alkenone is a biomarker not only of marine haptophyte algae [Marlowe et al., 1984; Volkman et al., 1980] but also of lacustrine haptophyte algae from saline and alkaline lakes on the Asian and North American continents [Liu et al., 2011; Toney et al., 2010]. The alkenone concentration of the suspended particles in water samples collected from a location (63°30′N, 165°30′W close to the mouth of the Yukon River was under the detection limit [Harada et al., 2003], which suggests that it is unlikely that alkenone originating from terrestrial lakes contaminated our samples.
 The alkenone unsaturation index, U37K′ = [C37:2]/[C37:2 + C37:3], was used to estimate sea surface temperature (SST). Alkenone temperatures have been widely used for regional and global reconstruction of SST [e.g., Conte et al., 2006; Müller et al., 1998; Prahl and Wakeham, 1987]. We used the calibration equation of Sikes et al. , T (°C) = 15.1 − 4.54 × ln(1/U37K′ − 1), which provides the closest estimates of in situ SSTs in the Bering Sea [Harada et al., 2003]. Alkenone temperatures estimated from duplicate sediment analyses differed by ±0.5°C (1σ, n = 23).
2.4. Analysis of the Mitochondrial Genes of E. huxleyi
 Mitochondrial gene analysis was performed on E. huxleyi filtered from 10 L of seawater collected in 2006 at 10 m depth in a bloom patch at site 11. DNA isolated by phenol–chloroform extraction was used as a template for polymerase chain reaction (PCR) amplification. PrimeSTAR GXL DNA polymerase (TAKARA BIO, Ohtsu, Japan) and the primer pair EGcox1-F2/EGatp4–16959R were used to amplify the region containing the cytochrome oxidase 1b (cox1b) and adenosine triphosphate 4 (atp4) genes of the mitochondrial genome. The first PCR product was diluted (1:50), then subjected to second-round PCR with the primer pairs EGcox1-F2/EGcox1-R5 and EGcox1-F3/EGatp4–16959R to amplify two overlapping fragments covering cox1b and atp4. The amplified fragments were cloned into a pGEM T-easy vector (Promega) and sequenced with DNA sequencer model 3130 and BigDye Terminator version 3.1 (Applied Biosystems, Foster City, CA, USA). The primer sequences are reported elsewhere [Hagino et al., 2011]. The raw cox1b sequence data have been registered with GenBank (Accession Numbers AB623144 to AB623148 and AB623206 to AB623209).
2.5. Analysis of Diatom Assemblages
 Sub-cores from sites 4 and 11 were used to examine diatom assemblages. Freeze-dried sediment samples (∼50–300 mg) were placed in beakers containing 10 mL of 10% hydrogen peroxide solution and 3 mL of 1 N hydrochloric acid, and heated for 1.5 h. A surfactant (hexametaphosphate; Calgon®) was then added to disaggregate the particles. The samples were dried at 40°C for 24 h, and then mounted on microslides with Pleurax and observed by light microscopy at 1000 × magnification. At least 250 valves were counted on each slide, including resting spores of an unidentified species of the genus Chaetoceros.
3. Results and Discussion
3.1. Estimates of Sediment Accumulation Rate Based on 210Pb and 137Cs Profiles
 Since surface sediments are often disturbed by benthic organisms and strong bottom currents [Field et al., 1981], we investigated the degree of sediment disturbance and sediment accumulation rate in the cores using 210Pbex. 210Pb in the atmosphere, which is produced by the decay of 222Rn, falls to the ocean surface and settles to the sea bottom, where it is incorporated into the sediments; 210Pbex decreases at a constant rate with sediment age. We selected six of the 12 sediment cores, those from sites 1, 4, 6, 7, 9, and 11, for this study because their 210Pbex results indicated minimal disturbance. The 210Pbex profiles in the top sediment layer of the four cores 1, 7, 9 and 11 showed gradually decreasing values with increasing core depth and few vertically homogeneous intervals that would indicate mixing (Figure 2). The core from site 1 showed two activity peaks of 137Cs (Figure 2), a byproduct of nuclear weapons testing and accidents; these peaks correspond to nuclear weapons testing during the 1960s and to the 1986 Chernobyl accident. Thus, in this core, the 137Cs profile provided an independent chronological marker [Baskaran and Naidu, 1995], and the sediment accumulation rate at site 1 estimated using the 210Pbex and 137Cs profiles was 2.5 mm yr−1 (Table 1). The estimated sediment accumulation rates at the other sampling sites ranged from 2.4 to 4.1 mm yr−1 [Oguri et al., 2012]. In areas where sedimentation rates exceed 1 cm per 100 years, statistically significant information is preserved at the 1-cm scale, even with bioturbation effects to 5–10 cm depth [Chapman and Shackleton, 1998]. The estimated sedimentation rates indicate that the sediments of the four cores at sites 1, 7, 9, and 11 represented approximately the past 30–70 years. Following Oguri et al. , we assumed that the age of the surface layer in each of these four cores corresponded to the collection date. At sites 4 and 6, the surface mixed layer of the cores was thicker, and Oguri et al.  estimated it to represent 78 and 50 years, respectively. However, the high alkenone concentrations in these surface mixed layers must have derived from the recent E. huxleyi blooms, suggesting that each represents a shorter time period than that estimated by Oguri et al. . Therefore, we assumed that the age of the surface layer in these cores was the collection date.
 In summary, we discuss here data from the cores collected at sites 1, 4, 6, 7, 9, and 11 because among the 12 cores, they show minimal disturbance. The sediment accumulation rates estimated using the 210Pbex and 137Cs profiles at the six sites ranged from 2.4 to 4.1 mm yr−1, indicating that the core sediments were deposited during approximately the past 30–70 years.
3.2. E. huxleyi Blooms and Alkenone as a Biomarker of Production
 The cell density of E. huxleyi coccospheres was as high as 5 × 106 L−1 in the 2000 bloom [Harada et al., 2003], and ranged from 0.5 × 106 to 3.5 × 106 L−1 in the 2006 bloom. These cell densities are comparable to the densities observed in the 1997 bloom (2.1–2.8 × 106 L−1) [Sukhanova and Flint, 1998]. In the eastern Bering Sea blooms, E. huxleyi was present as morphotype A, whereas near site 7 in the central Bering Sea, where no bloom was observed, it was presented as morphotype O [Hagino et al., 2011]. The results of a culture experiment showed that the coccolith number differs between the exponential and stationary growth phases, with fewer coccoliths per cell in the exponential phase, which is the growth phase of a bloom [Paasche, 1998]. Morphotype A is generally characterized by fewer coccoliths per cell (Figure 3) compared with most other morphotypes, which may be advantageous for bloom formation.
 The life cycle of E. huxleyi includes a diploid, coccolith-bearing stage that forms blooms, and a haploid, non-calcified stage [Green et al., 1996] that is resistant to viral infection [Frada et al., 2008]. At both stages E. huxleyi contains alkenone, but culture experiments have shown a wide range in alkenone content (1–17 pg cell−1), depending on the nutrient conditions, rate of cell division, and the growth phase (exponential versus stationary) [Epstein et al., 1998]. Thus, although alkenone can be used to estimate production for both the calcified and non-calcified life stages, the cellular content cannot be assumed to be the same in the two stages.
 Culture experiments have shown that some bacteria preferentially degrade tri- and tetra-unsaturated alkenones over di-unsaturated alkenone, resulting in an apparent increase in temperature estimates based on the alkenone unsaturation index [Rontani et al., 2005, 2008; Zabeti et al., 2010]. If significant preferential degradation of alkenones occurred in the sediments studied here, the relative abundance of C37:4 alkenone would be expected to decrease with increasing sediment age and UK′37 values would be expected to increase. Although UK′37 did increase with sediment age at sites 4, 6, 9, and 11, the relative abundance of C37:4 did not (see auxiliary material), indicating that preferential degradation did not significantly affect our results.
 In summary, the cell density of E. huxleyi coccospheres in the blooms in the eastern Bering Sea was on the order of 106 L−1. Because the life cycle of E. huxleyi includes a haploid, non-calcified stage, the alkenone content is more suitable for the estimation of E. huxleyi production than the count of E. huxleyi coccospheres, because both the calcified and non-calcified stages contain alkenone, which is also relatively robust against degradation.
3.3. E. huxleyi Blooms and Environmental Conditions in the Bering Sea
 Previous alkenone-based temperature estimates for the eastern continental shelf of the Bering Sea correspond more closely (±1°C) to SSTs measured during the main algal growth period in September than to the annual average SST [Harada et al., 2003]. Therefore, we assumed that the alkenone temperatures determined in the present study (Figure 4) represent late summer temperatures. The range of alkenone temperatures at site 1 (8.1–9.6°C) is typical of SSTs on the eastern continental shelf of the Bering Sea in September (Figure 4). The alkenone temperatures prior to the 1980s at sites 4, 6, 9, and 11 (11–13°C) were about 3°C higher than the typical late summer SST in the eastern Bering Sea, and corresponded more closely to late summer temperatures of waters in the Gulf of Alaska and near the Alaskan coast (http://www.pac.dfo-mpo.gc.ca/science/oceans/data-donnees/sst-tsm/index-eng.htm). Thus, alkenone in sediments deposited on the continental shelf prior to the 1980s may show that an amount of allochthonous alkenone transported from other areas likely from the Gulf of Alaska was relatively larger than the autochthonous alkenone production, because autochthonous alkenone production was quite low on the continental shelf itself. At site 7, on the continental slope, the alkenone temperature remained relatively constant throughout the sampled period.
 The relative abundance of tetra-unsaturated alkenone (%C37:4) is an indicator of salinity [e.g., Rosell-Melé, 1998; Sicre et al., 2002], although it can also be correlated with SST [Bendle et al., 2005; Rosell-Melé, 1998]. During an E. huxleyi bloom in the Bering Sea, %C37:4 of suspended particles showed a negative linear correlation with sea surface salinity (SSS) and NO3− + NO2− concentration [Harada et al., 2003]. Results of the present study show that %C37:4 was negatively correlated with both SST and SSS (Figure 5). Thus, periods of low %C37:4 are characterized by high SST, SSS, and nutrients, while periods of high %C37:4 are characterized by low SST, SSS, and nutrients. However, the use of %C37:4 to reconstruct water mass characteristics in the Bering Sea prior to the 1980s is limited by the presence of allochthonous alkenones, likely from the Gulf of Alaska.
 The alkenone content of the sediment core samples showed similar minimum values but a large range of peak values (14.0–326 μg gMOC−1) among the six sampling sites (Figure 6 and auxiliary material). Alkenone was under the detection limit at depths deeper than 12.5 cm at site 1, and Harada et al.  suggested that bacterial decomposition might be responsible for the failure to detect any alkenones in the deeper sediments. However, at the other sites, alkenone was detected in all samples. Org. C content, furthermore, showed almost no decrease with depth at any of the sites (Figure 6 and auxiliary material). Therefore, the lack of alkenone in the subsurface sediments at site 1 likely reflects very low production of coccolithophorids during the period of sediment deposition rather than subsequent degradation.
 Alkenone content in sediments at sites 9 and 11 on the southern continental shelf increased strongly beginning in the late 1970s (Figure 6). At site 11, alkenone content increased by two orders of magnitude compared with the previous decade. At site 1, in the northern Bering Sea, no alkenone was detected until the late 1970s, and alkenone content increased dramatically around 2000. In contrast, at sites 4 and 6, on the northern continental shelf, alkenone content did not begin to increase until the late 1990s, when the first coccolithophorid bloom was detected by SeaWiFS [Napp and Hunt, 2001]. An increase in E. huxleyi might at least partly explain the large reduction in benthic macrofaunal biomass observed in the northern shelf region from the late 1990s to the beginning of the 2000s [Grebmeier et al., 2006], because it suggests possible competition between E. huxleyi and diatoms, which are a better food source for benthic macrofauna.
 In summary, although at site 7, on the continental slope, alkenone content gradually increased beginning around 1960 and showed no dramatic peak indicating a large bloom, the profiles of alkenone-derived temperature and alkenone content at the other five sites reflect the increasing prominence of E. huxleyi blooms in the Bering Sea since the late 1970s. This period of prominence followed a major climate regime shift in the North Pacific that occurred in 1976–77 [Hare and Mantua, 2000].
3.4. Environmental Conditions Associated With E. huxleyi Blooms in the Bering Sea
 To understand the possible mechanisms leading to E. huxleyi blooms in the eastern Bering Sea, we compared the alkenone records obtained in the present study with two climate indices (Figure 6), the North Pacific Index (NPI) [Trenberth and Hurrell, 1994], which is the area-weighted sea level pressure anomaly over the region 30°N–65°N, 160°E–140°W during November–March 1900–2007 relative to the 1967–2000 mean published by NCEP (National Centers for Environmental Prediction)/NCAR (National Center for Atmospheric Research) (http://www.beringclimate.noaa.gov/data/index.php); and the Pacific decadal oscillation (PDO) [Mantua et al., 1997], estimated from monthly mean SST anomalies during 1900–2007 from 20°N to the boreal polar region (http://www.beringclimate.noaa.gov/index.html). We used the NPI to estimate the intensity of the winter Aleutian Low. When the NPI is negative, the Aleutian Low is intense and the inflow of relatively warm Pacific water to the Bering Sea is strengthened. This inflow of warm water causes a positive PDO and a relatively warm winter in the eastern Bering Sea. Storm tracks from Siberia to the Bering Sea were enhanced during 1950–2003, and these storms may have increased winter warming over the Bering Sea [Rodionov et al., 2007]. According to Rodionov et al. , the change in the position of the Aleutian Low also affects the cooling or warming of the Bering Sea in winter.
 Sea level pressure (SLP) anomalies can be used to detect regional atmospheric circulation corresponding to periods of anomalous weather. According to Hunt et al. , mean SLP anomalies during the spring and summer (May–August) were weak in 2003–2005, suggesting that the relatively warm summertime ocean temperatures (and warm air temperatures) during that period can be attributed largely to the effects of the relatively warm winter (November–March) persisting through the following spring and summer. It seems likely that the climate in winter is tightly linked to that during the following spring and summer through the regional atmospheric circulation. The PDO in the eastern Bering Sea was predominantly positive during the late 1970s to 1980s and from the late 1990s to the early 2000s, periods when alkenone content of the sediments was high, suggesting that calm weather and the inflow of warm water produced a stratified water column and suitable light conditions for E. huxleyi production. Fresh water from sea ice melting, especially in the north, would also have contributed to the water mass stratification.
 Insolation and wind mixing also influence the regional atmospheric circulation. We used NCEP2 data to determine the monthly mean cloud covered area (%) over the eastern Bering Sea (52–66°N, 155–180°W) during 1979–2010 (http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis2.gaussian.html). We found that the area of cloud cover was almost constant, without any significant decreasing or increasing trend, throughout these years, and it ranged from 60% to 70% from July to October (which includes the E. huxleyi bloom season). This result implies that the interannual variation of insolation was quite small and cannot account for the anomalous warm weather in the eastern Bering Sea.
 Comparison of seasonal salinity data for 2003–2007 (Argo floats) with WOD05 climatology data for 1960–1989 reveals that salinity in the northern North Pacific, including the eastern Bering Sea, has recently decreased, although the change is small, only ∼0.1 practical salinity units [Hosoda et al., 2009]. Hosoda et al.  attributed recent changes in global ocean salinity, including the reduction in SSS in the Bering Sea, to an intensified hydrological cycle in recent decades, perhaps due to global climate change. The lower salinity might have favored E. huxleyi production because under lower salinity, less energy and materials are needed to adjust osmotic pressure and drive the Na/H anti-porter [Chen and Jiang, 2009], which might directly increase cell growth and coccolith size [Bollmann et al., 2009].
 In summary, late summer blooms of E. huxleyi from the late 1970s may be related to the major winter climate regime shift that occurred in the North Pacific in 1976–77. The occurrence of blooms may have been promoted by the enhanced surface water stratification, caused by increasing SST and decreasing SSS associated with a change in Aleutian Low activity and an intensified hydrological cycle in recent decades.
3.5. Changes in Diatom Taxa in the Eastern Bering Sea
 The edge of the continental shelf in the eastern Bering Sea is often referred to as the “Green Belt” [Springer et al., 1996] because of high primary production (annual average, 175–275 gC m−2 yr−1), dominated by marine diatoms. We investigated the diatom taxa in the sediments at site 4 (on the northern inner shelf) and site 11 (on the southern outer shelf) (Figure 1). Site 4 is covered by sea ice every winter, whereas site 11 is near the green belt and in an area that is only occasionally covered by sea ice [Hunt et al., 2010]. In spring (April and May) the outer shelf region is characterized by high nitrate concentrations, whereas the inner shelf is characterized by very low nitrate concentrations throughout the surface and subsurface waters [Hunt et al., 2010]. In the late summer (September), nutrient concentrations are similar and low at both sites [Harada et al., 2006].
 The total number of diatom valves at the northern site 4 gradually increased from about 1990 to 2006 (Figure 7a). Nearly all of the major taxa showed this increasing trend, including Chaetoceros resting spores and Thalassionema nitzschioides, which are indicators of high productivity [Sancetta, 1982]; Fossula arctica, Fragilariopsis cylindrus, and Thalassiosira hyalina, which are Arctic neritic species associated with sea ice [von Quillfeldt, 2000]; Rhaphoneis surirella; and Paralia sulcata, a benthic species living in shallow coastal areas and associated with the mixing of water masses by strong winds, possibly indicating off-shelf transport of diatoms [Blasco et al., 1980]. The only exception was Fragilariopsis oceanica, which maintained a relatively constant density. In the northern Bering Sea, the rise in spring SST and the consequent earlier onset of sea ice melting might explain these recent ecosystem changes [Grebmeier et al., 2006]. Earlier melting of the sea ice would also improve light penetration and result in higher diatom productivity.
 In contrast to northern site 4, the total number of diatom valves at southern site 11 increased in the 1980s, gradually decreased until the first half of the 1990s, and then increased again in the 2000s (Figure 7b). Increases in T. nitzschioides, Chaetoceros resting spores, and Neodenticula seminae accounted for most of the increase during the 1980s, and decreases in the cold water taxa Thalassiosira nordenskioeldii and T. gravida [Sancetta, 1982] and in the coastal species P. sulcata accounted for the subsequent decline. Chaetoceros resting spores, T. nitzschioides, and N. seminae continued to increase, however, accounting for the increasing trend in the 2000s. The depth-averaged SST in spring and summer showed a warming trend from 1995, and the extent of the sea ice cover abruptly decreased in the southern Bering Sea after 1977 [Iida and Saitoh, 2007], consistent with the positive PDO at the end of the 1970s. Earlier retreat of the sea ice in spring might also have enhanced total diatom production at the southern site. Sancetta , who investigated surface sediments from the outer shelf of the southern Bering Sea, reported that a high abundance of T. nordenskioeldii was associated with a short sea ice duration. In this study, abrupt increases in T. nordenskioeldii in the 1980s and 2000s also corresponded to a shortened sea ice duration (http://www.beringclimate.noaa.gov/data/BCresult.php). These results suggest that a high or low relative abundance of T. nordenskioeldii in the sediments does not always relate to an expansion or reduction of sea ice [Caissie et al., 2010; Katsuki et al., 2009], although this species lives in cold water near sea ice and decreases in the abundance of this species in the late 1990s and 2000s were concurrent with decreases of sea ice (http://www.beringclimate.noaa.gov/data/BCresult.php). Improvement of light penetration might be another key factor influencing production of T. nordenskioeldii.
 In summary, at northern site 4, the total number of diatoms of all major taxa gradually increased over the last several decades. In contrast, at southern site 11, the total number of diatom valves increased in the 1980s, gradually decreased until the first half of the 1990s, and then increased again in the 2000s. The trend of high-productivity-related species was similar to the trend in total numbers of diatoms. The recent enhanced diatom productivity can be attributed to improved light penetration due to the reduction of the sea ice extent and its earlier retreat in the spring.
3.6. Significance of E. huxleyi Blooms in the Bering Sea
 Despite recent increases in the production rate of diatoms in both the northern and southern Bering Sea, changes in the ratio of alkenone content to total diatom valves in the sediments closely followed changes in E. huxleyi production, particularly at northern site 4 (Figure 8). Although E. huxleyi blooms in the eastern Bering Sea have declined temporally and spatially since 2003, as indicated by a slight reduction in satellite-observed chlorophyll a levels [Grebmeier et al., 2010], coccolithophorids are probably continuously transported from the Bering Sea to the Arctic Ocean, where acidity is increasing [Yamamoto-Kawai et al., 2009]. Higher pCO2 of seawater promotes greater coccolith production in E. huxleyi [Iglesias-Rodríguez et al., 2008; Irie et al., 2010]. If other conditions are favorable for E. huxleyi growth (i.e., no sea ice cover and a stratified water column), a large increase in coccolith production in the Arctic Ocean might further accelerate acidification.
 Coccolithophorids on the Pacific side of the Arctic Ocean may affect the North Atlantic ecosystem. Mitochondrial sequences of E. huxleyi from the Bering Sea blooms have been analyzed and compared with sequences of other E. huxleyi strains from around the world [Hagino et al., 2011]. Multiple sequences of the mitochondrial genome of E. huxleyi from the 2006 Bering Sea bloom (data not shown) were identical to those of E. huxleyi strains BG10–2 (49.30°N, 10.30°W; collected in August 2007), BGI-1 (51.20°N, 10.29°W; August 2007), and PLY92A (50.20°N, 4.22°W; July 1957) from the Celtic Sea, and to strain MM-1 from near Bergen, Norway (60.50°N, 4.50°E; collection date unknown) [Hagino et al., 2011] (Figure 9). Migration of plankton between the North Pacific and North Atlantic oceans is known to have occurred in the past: several decades ago, the North Pacific diatom N. seminae appeared in the North Atlantic for the first time in 780 kyr [Reid et al., 2007], having apparently migrated during the recent ice-free summer months from the Pacific to the Atlantic through the so-called Northwest Passage [Corbyn, 2007]. If E. huxleyi migrated along a similar route, its migration must have occurred prior to the recent sea ice reduction in the Arctic Ocean, because its mitochondrial gene sequences match those of strain PLY92A, a Celtic Sea strain collected in 1957. When and how marine organisms, including E. huxleyi, migrated from the North Pacific to the North Atlantic is still a mystery. Emiliania huxleyi might have arrived in the North Atlantic just after the Little Ice Age (i.e., from about 1900 onwards) or a longer time ago. Further monitoring and observations are necessary to understand the future of the Bering Sea–Arctic marine ecosystem, including the dynamic migration of organisms between the Pacific and Atlantic oceans.
 In summary, the ratio of alkenone content to total number of diatom valves in the sediment increased as E. huxleyi production increased, suggesting that the increase in the production rate of coccolithophorids frequently exceeded that of diatoms. The multiple mitochondrial gene sequences of E. huxleyi from the 2006 Bering Sea bloom were identical to those of E. huxleyi strains collected in the North Atlantic. The oldest E. huxleyi strain in the North Atlantic with mitochondrial sequences identical to those of the Bering Sea strain was collected in 1957, suggesting that the migration of E. huxleyi from Pacific side to the Atlantic side occurred earlier than the recent significant sea ice reduction in the Arctic Ocean.
 Sedimentary alkenone records indicate that E. huxleyi blooms have become more prominent on the eastern continental shelf of the Bering Sea since the late 1970s. Comparisons with Aleutian Low activity and the Pacific Decadal Oscillation index over the past several decades suggest that the onset of these enhanced blooms is related to a major climate regime shift that occurred in 1976–77 in the North Pacific. The blooms were likely promoted by enhanced surface water stratification caused by increasing SST and decreasing SSS. Total numbers of diatoms, the dominant primary producers in the Bering Sea, have gradually increased during the last few decades. Nevertheless, increases in the ratio of alkenone content to total number of diatom valves in the sediment corresponded to increases in E. huxleyi production, suggesting that increases in E. huxleyi production rate exceeded the increases in the diatom production rate. Overall, our results indicate that the E. huxleyi blooms resulted from recent large environmental changes in the Bering Sea–Arctic Ocean system: enhancement of the intensity of the hydrologic cycle, a decline in seasonal sea ice, greater water column stratification and stability, and improving light conditions. Multiple mitochondrial sequences of E. huxleyi from the 2006 Bering Sea bloom were identical to those of an E. huxleyi strain collected in 1957 in the Celtic Sea in the North Atlantic, suggesting that E. huxleyi migrated from the Pacific to the Atlantic side before the recent significant sea ice reduction in the Arctic Ocean. If no sea ice cover and a stratified water column, which is the favorable condition for E. huxleyi growth, occurs in the Arcitic Ocean, E. huxleyi could extend its biogeographic distribution in near future.
 We are grateful to Captain Akamine and the crew of R/V Mirai for their help with sediment collection and water sampling during the MR00-06 and MR06-04 cruises. This paper benefited from important comments by Eric Sundquist, Editor in Chief and two anonymous reviewers. This work was supported by the Japan Agency for Marine-Earth Science and Technology and the Japan Society for the Promotion of Science, KAKENHI (B) 19310018 and (S) 22221003.