Gravity Core B2-9 was recovered from the northern Bering Sea Slope (59°17′32″N, 178°41′50″W, water depth 2 200 m, Fig. 1) during the 1st Chinese National Arctic Expedition in 1999. The sediment core is 231 cm in length and is primarily composed of dark gray diatom mud. The core was sampled at 2-cm intervals and 115 samples in total were obtained.

All samples were extracted and measured for qualitative and quantitative analysis of organic matter. For this study we mainly analyzed the n-alkanes. The freeze-dried samples were ground manually. After adding internal standards of n-C₂₄D₅₀ and n-C₁₉H₃₉OH, each sample (2–4 g) was extracted ultrasonically 4–5 times with di-chloromethane/methanol (3:1, v/v). The supernatants were concentrated and then saponified with 6% KOH/ methanol at room temperature overnight (approximately 12 h). Neutral components were extracted with n-hexane and then were further separated into two sub-fractions by silica gel column chromatography. Hydrocarbons were eluted with n-hexane and analyzed directly using a gas chromatograph (GC) under the following GC conditions: an HP-1 capillary column (60 m×0.32 mm×0.17 μm), the injector and flame ionization detector (FID) both at 280°C, splitless injection and helium as the carrier gas with a flow rate of 1.0 mL/min (He et al., 2008). Quantification of biomarkers was performed by integration of the relevant peak areas of the biomarkers of interest and compared with those of the internal standards. Test results using internal standards showed that the recovery ratio of this method was up to 85%, and the standard deviation of four parallel experiments was less than 1%. All analyses were performed at the State Key Laboratory of Marine Geology in Tongji University, Shanghai.

The age model for Core B2-9 was established by ten Accelerator Mass Spectrometry (AMS) ¹⁴C dating data, obtained from the organic carbon in sediments and calibrated by Calib 6.0 using a mixed Marine & Northern Hemisphere Atmosphere mixing curve with 75% marine organic carbon (Fairbanks et al., 2005). Linear interpolation was applied between each AMS ¹⁴C age (Fig. 2). The core recovered a sediment sequence from 0.98 ka to 9.59 ka with an average sample resolution of 75 years. All ages are reported as calibrated calendar ages. Based on this age model, we calculated the sedimentation rate of Core B2-9, which had gone through three stages overall (Fig. 2).

In Core B2-9, the n-alkanes range from nC₂₁ to nC₃₅, with obvious odd-carbon dominance (Fig. 3), similar to previous studies in other oceans. In the following sections we present two features that differ from other studies.

First, nC₂₇, rather than nC₂₉ or nC₃₁, shows the highest content among the n-alkanes in Core B2-9, consistent with the studies reported by Lu et al. (2004) and Zhang et al. (2007) in the Bering Sea. The nC₂₇ content ranges from 90.5 ng/g (Fig. 3a) to 1 187.3 ng/g (Fig. 3b), with an average value of 494.2 ng/g (Fig. 3c). A clear stage characteristic (Fig. 4a) is apparent in the variations. The total odd-carbon number n-alkane contents between nC₂₅ and nC₃₃ (referred to as ∑_odd(nC₂₅–nC₃₃)) (Fig. 4b) show good correlation (R²=0.99) to those between nC₂₃ and nC₃₁ (referred to as ∑_odd(nC₂₃–nC₃₁)) (Fig. 4c). nC₂₇ was the uniquely greatest contributor to the total n–alkane content during the Holocene. As shown in Fig. 4d, we take nC₂₇/∑_odd(nC₂₅–nC₃₃) to indicate the contribution ratio of nC₂₇ to total n-alkane content (referred to as nC₂₇ contribution), which ranges from 26% to 44%, and reaches to 32% on average. The high nC₂₇ content also leads to almost the same pattern between the total n-alkane content and nC₂₇, and the correlation coefficient is 0.98 (R²). In regions where grasses dominated, nC₃₁ is the major n-alkane, whereas nC₂₇ and nC₂₉ are more abundant in sediments where trees dominated (Cranwell, 1973). Ternary plots of the relative proportions of nC₂₇, nC₂₉ and nC₃₁ in sediments can be used to reconstruct contributions of grasses, shrubs, coniferous trees and deciduous trees surrounding the area studied (Meyers, 2003; and references therein). Our data suggest that abundant woody plants were distributed in the source region during the Holocene, and that the ratio of woody/herbaceous plants was relatively stable.

Second, several instances of rapid decrease in nC₂₇ may be attributed to climate deterioration in the source region. Alaskan pollen records show a climatic optimum in the early Holocene of 8–10 ka BP, when communities of Populus spp. were more widely distributed than at present. After a transgression in the early Holocene, the forest in eastern Alaska spread to the modern location, and modern vegetation associations had probably developed by that time with the arrival of Pinus pumila at 9 ka BP (Lozhkin et al., 1993; Liu et al., 2004). Hence, the rapid decline in nC₂₇ between ca. 8–7.8 ka BP may be related to the climatic optimum during the early Holocene. Around ca. 8 ka BP, Picea glauca arrived in the northern Alaska Range, and formed a forest tundra association until ca. 6 ka BP. Shrub thickets of Alnus spp. established in the region at ca. 6.5 ka BP and Betula papyrifera arrived at ca. 6 ka BP. Closed P. glauca forests developed at ca. 6 ka BP. P. mariana subsequently became important and replaced P. glauca as the dominant tree species in the region at ca. 4 ka BP. The establishment of P. mariana forests probably reflects the complex response of forest ecosystems to the onset of cooler and moister conditions during the late Holocene (Hu et al., 1996). As the most abundant n-alkane from woody plants, nC₂₇ productivity would inevitably be affected by changes in vegetation structure. Sedimentary nC₂₇ content (Fig. 4a) recorded the progress of rapid cooling at 9.3 ka BP and 8.2 ka BP (Fig. 4f), but the nC₂₇ contribution to the total n-alkane content (Fig. 4d), which showed the different changes during these two rapid cooling intervals (Fig. 4f), was nearly unchanged, especially around ca. 9.3 ka BP. Further explanations of the reasons for these changes in paleoenvironmental and paleoecological reconstructions need more support from multiple proxies, such as pollen records (e.g., Meyers, 2003). The likely cause will be discussed in Section 4.3 and should be considered in conjunction with changes in the molecular indexes (such as ACL and AI).

nC₂₃ in Core B2-9 is maintained at a higher level than nC₃₃ in almost all samples. Its contents range from 57.3 ng/g to 627.3 ng/g, with an average of 234.8 ng/g, much more than nC₃₃ (95.6 ng/g). The nC₂₃ content clearly shows a stepwise decrease from the early to the late Holocene (Fig. 4e), which is similar to nC₂₇, ∑_odd(nC₂₅–nC₃₃) and ∑_odd(nC₂₃–nC₃₁) during the Holocene (Figs 4a–c), with correlation coefficients (R²) of 0.85, 0.83 and 0.88, respectively. Such similar fluctuation patterns suggest that nC₂₃ may be controlled by some factors that also control nC₂₇. However, previous studies have shown that the high abundance of nC₂₃ is mainly produced by some submerged and floating macrophytes (Meyers, 2003; and references therein). For example, the sea-grass known as eelgrass (Zostera marina L.) is widespread in coastal and estuarine waters, and its life cycle is controlled by sea water salinity and temperature (Ye and Zhao, 2002; Nejrup and Pedersen, 2008). To exclude the influence of some uncertain factors and maintain uniformity in calculation between the total n-alkane content and n-alkane indexes (such as carbon preference index (CPI) in the following sections), we remain consistent in this paper in choosing ∑_odd(nC₂₅–nC₃₃) as the total n-alkane content, used as the proxy for terrestrial inputs (Fig. 4b).

To some extent the changes in nC₂₃ indicate changes in the regional climate. According to pollen records from lake sediments in southwestern Alaska, the vegetation structure was similar between southwestern Alaska and north-central Alaska (Hu et al., 1996), and summer temperature during the early Holocene was warmer than at present (Ritchie et al., 1983; Barnosky et al., 1987; Anderson and Brubaker, 1994). As a result, low lake levels and high productivity might have been caused by warm and dry summers (Hu et al., 1996), leading to high nC₂₃ productivity over the same period (Fig. 4e). Similar to the nC₂₇ (Fig. 4a), nC₂₃ content also dropped at ca. 9.3 ka BP and 8.1 ka BP (Fig. 4e), the response of submerged/floating plants to two sudden cold events recorded by δ¹⁸O in the NGRIP ice core from Greenland (Fig. 4f). nC₂₃ content rebounded to a high level at around 7.2 ka BP, different from nC₂₇ (Fig. 4a) and the total n-alkane content (Fig. 4b). Productivity records from the Arolik Lake, southwestern Alaska, clearly show drops during the two cold periods at around 9.3 ka BP and 8.1 ka BP (Hu et al., 2003a). The cooling would also have suppressed the output of nC₂₃ from marine submerged/floating plants. In contrast, the rise in nC₂₃ content at ca. 7.2 ka BP might have been caused by increasing productivity in lakes, and may also reflect the distinction between nC₂₃ and terrestrial n-alkanes as the proxies of climate change (e.g., Ficken et al., 2000).

The total n-alkane content ranged from 275.3 ng/g to 3 694.8 ng/g and 1 681.2 ng/g on average during the past 9.6 ka BP, with an overall decreasing trend from the early to the late Holocene. This kind of decline change has four clearly periodic stages (Fig. 4b). During Stage I, ca. 9.6–7.8 ka BP, total n-alkane content underwent a rapid fluctuation with high amplitude, and its average value was up to 2 657.3 ng/g. In Stage II, ca. 7.8–6.7 ka BP, both the frequency and amplitude of changes in the total n-alkane content were reduced, showing a slightly upward trend, but lower than Stage I, with an average of 1 948.7 ng/g. In Stage III, ca. 6.7–5.4 ka BP, the total n-alkane content decreased further, with an average of 1 304.6 ng/g. Its trend was similar to Stage II, but more stable. The total n-alkane content was relatively steady in Stage IV, from ca. 5.4 BP, with the minimum average value (814.2 ng/g) of the four stages, but during the Holocene this was still significantly higher than in other seas, such as the Southern Ocean (Li and Wang, 2004) and the northern South China Sea (He et al., 2008).

In general, the deposit of n-alkanes in marine sediments is controlled by two factors: the n-alkane productivity of land plants and the environmental conditions of transportation. We argue that, in the study area, the most important factor of terrestrial input changes during the Holocene was sea-level change. Global sea levels rose from about –50 m (compared with the modern level) at ca. 11 ka and reached the modern level at ca. 6 ka (Huang et al., 2009; and references therein). As a result, the shoreline of Alaska and Siberia would have shifted away from the core location and the wide, exposed northeastern continental shelf of the Bering Sea would have been submerged again, causing a decreased vegetation coverage area. During sea-level rise, the plant residues were transported by currents to the slope area, and further into the basin. The rising sea level also moved estuaries such as that of the Yukon River back inland, which extended the delivery distance of terrestrial materials from the estuary to sedimentary sites in the sea. The rising sea level not only reduced the distribution range of terrestrial vegetation, but also extended the distance between river estuary and sedimentary area on the northern slope. Therefore, the decline in total n-alkane input during ca. 9.6–5.4 ka BP (Fig. 4c) was the result of a reduction in terrestrial materials. From ca. 6 ka BP the influence of sea-level change weakened gradually. However, Hopkins (1967) has pointed out that there was a transgression in the Bering Sea between ca. 6 ka BP and 5 ka BP, which was also recorded in sediments of Core B2-9. The total n-alkane content dropped by nearly half in 200 years, and this was the last stepwise decrease during the Holocene (Fig. 4c).

Continuous sea-level rise had a gradual impact on the total n-alkane content; its decrease around ca. 6.7 ka BP and ca. 7.8 ka BP (Fig. 4c) may not only have been controlled by sea level, but also have been related to the climate and vegetation abundance in the source region. Pollen data from lake sediments in southwestern Alaska indicate that winter precipitation and spring temperature both increased from ca. 9.8 ka BP (Hu et al., 1995). During ca. 11–7 ka BP, the Holocene Thermal Maximum occurred both in central Beringia and western Alaska (Kaufman et al., 2004, and references therein). Katsuki et al. (2009) pointed out that the vast area of southwestern Alaska was, overall, warm and wet at this time because of the impact of the Aleutian Low. Such a climate favored terrestrial vegetation growth and density, which produced more n-alkanes. After ca. 7 ka BP the climate of southwestern Alaska changed from warm and wet to cold and wet under the influence of the enhanced Aleutian Low. These changes might not only have caused the shrinkage in the vegetation area, but also altered the strength and direction of the prevailing winds. By then, the n-alkanes would have been delivered mainly by river discharge. Finally, the total n-alkane content dropped at ca. 6.7 ka BP (Fig. 4c). The controlling factor for the rapid decline at ca. 7.8 ka might be slightly different from that of other declines. A variety of paleoenvironmental records from Alaska and adjacent areas in Canada for the period ca. 11–8 ka BP suggest that this region experienced a period of relatively warm summers, presumably caused by a pronounced maximum in summer insolation in northern high latitudes (Hu et al., 1996, and references therein). Hence, in Stage I, the high n-alkanes may be mainly caused by high terrestrial vegetation productivity. Total n-alkane content shows a 200-year episode of extremely low values around 8.0 ka BP (Fig. 4c), even lower than the average value of Stage II. This change is probably a response to the “8.2 ka event” as recorded in the Greenland ice core δ¹⁸O (Alley and Ágústsdóttir, 2005), which was a rapid cold event lasting only about 400 years (Fig. 4f). Another clearly minimum value event in ca. 9.3–9.1 ka BP was the marked response to the 9.3 ka cold event (Fig. 4f). The production of terrestrial vegetation was severely reduced during the cold spells, as confirmed by Hu et al. (1996): the vegetation community appeared to have declined in abundance near the end of the early Holocene between ca. 11 ka BP and 8 ka BP. The impact on terrestrial n-alkane production was irreversible, especially during the 8.2 ka BP event, and directly induced the shift from Stage I to Stage II.

Seasonal sea ice may have been another factor influencing n-alkane input. Seasonal sea ice retreated gradually into the inner continental shelf in the Holocene (Katsuki et al., 2009). During the early Holocene, sea ice retreated significantly in the Bering Sea (Méheust et al., 2016). From the mid-Holocene, sea ice rarely extended to the core site of B2-9 (Katsuki et al., 2009). Thus, although its influence cannot be ignored, sea ice may not be a major controlling factor. Similarly, the influence of sedimentation rate on n-alkane contents may also be limited. According to Fig. 2, the sedimentation rate of Core B2-9 shows a change in phase with three stages, which is different from the characteristically stages of the n-alkanes (Figs 4a, b). Before ca. 5.4 ka BP, the sedimentation rate was relatively stable at ca. 31.4 cm/ka on average, but the nC₂₇ (Fig. 4a) and total n-alkane content (Fig. 4b) dropped at ca. 9.6 ka BP, 7.8 ka BP and 5.4 ka BP. In contrast, the n-alkanes were relatively stable since ca. 5.4 ka BP, while the sedimentation rate underwent several rapid changes between ca. 5.4 ka BP and 4.2 ka BP, and remained at a lower level since ca. 4.2 ka BP. This distinction may suggest that there is no direct relationship between them.

The molecular indexes of n-alkanes based on their compositions, such as CPI, ACL, alkane index (AI) and nC₃₁/nC₂₇ ratio, have been used to estimate the sources of these n-alkanes and to reconstruct the vegetation type and climate of the source region. The odd-carbon preference of the n-alkanes of terrestrial higher plants decreases with degradation and diagenesis, and the CPI of fresh land vegetation n-alkane is generally greater than 3; thus CPI has been used to indicate n-alkane source and its degree of maturation (He et al., 2008, and Table 1 therein). ACL, AI and nC₃₁/nC₂₇ ratios are established on the basis of the modern observation that the most abundant n-alkanes have different carbon numbers in different types of vegetation. As discussed above, both grass and trees produce nC₂₇, nC₂₉ and nC₃₁, but usually nC₃₁ is most abundant in grass while nC₂₇ or nC₂₉ is most abundant in trees. Thus, the increasing values of ACL, AI and nC₃₁/nC₂₇ ratios indicate a higher contribution from grass (Maffei, 1996; Zhang et al., 2006).

CPI ranges from 1.47 to 5.19 in Core B2-9, with an average of 3.63. Most values are greater than 3, indicating that the n-alkanes are primarily derived from terrestrial vegetation, rather than from fossil n-alkanes with high maturation. CPI shows a descending trend overall, suggesting that the contribution of terrestrial n-alkanes to the total n-alkanes gradually reduced. Sea levels reached their modern levels by ca. 6 ka BP, which may be the most important reason for the CPI changes before ca. 5.4 ka BP. Sea-level rise would have reduced the input from terrestrial n-alkanes but rarely affected fossil n-alkanes. From ca. 5.4 ka BP, the nC₂₇ and total n-alkane content remained at a steady level (Figs 4a, b); however, CPI continued its downward tendency (Fig. 5a), although its average value was 3.29 during this stage. Hence, the changes in CPI after ca. 5.4 ka BP may have been caused by increased contributions from fossil n-alkanes, and not by decreasing n-alkanes. At the same time, the variations in ACL and nC₃₁/nC₂₇ ratios show nearly the same stable trend. ACL, nC₃₁/nC₂₇ ratio and AI did not show any special pattern of variation (Figs 5b–d). Considering the contribution of nC₂₇ to total n-alkanes, we argue that vegetation structure in source regions maintained a relatively stable state during the Holocene, and woody plants predominated over herbaceous plants in most of intervals.

Variation patterns in these indexes were not uniform during the Holocene. First, several peaks in the ACL and nC₃₁/nC₂₇ ratio corresponded to low CPI (Fig. 5a) and low nC₂₇ contributions (Fig. 4d) in some short periods, such as ca. 7.9 ka BP and ca. 1.1 ka BP. These peaks reveal that the herbaceous plant contribution increased as the woody plant contribution decreased. Low CPI indicates an increased fossil n-alkane contribution, and low nC₂₇ contribution indicates the declined woody plant n-alkane contribution. However, the nC₂₇ production and total n-alkane content have similar change patterns. Therefore, the growth rate of woody plant n-alkane contribution was lower than that of herbaceous plant n-alkane and fossil n-alkane, even though woody plants still had high productivity. One possible explanation is that, during these short periods, the climate might have become dry and cold in the source regions, the stronger terrestrial weathering provided considerably more fossil n-alkanes and herbaceous plants flourished. There is some evidence for this. According to previous studies the tundra vegetation communities in the Arctic often comprised mosses, lichens, erect shrubs, graminoids and prostrate shrubs, and the vegetation community on the land north of the Bering Sea was mainly tundra (Walker et al., 2005; Bratsch et al., 2016). The Yukon watershed in Alaska may be taken as an example. The Yukon River is the main terrestrial input source of the Bering Sea: approximately 20% of the catchment is covered by spruce forest such as Picea glauca and black P. mariana (Brabets et al., 2000), about 40% by grassland, 20% by shrubland and 8% by open water and wetlands associated with the lowland areas (Amon et al., 2012; and references therein). However, further research will be needed to provide more evidence for this explanation.

Second, similar to the nC₂₇ contribution (Fig. 4d), these n-alkane indexes did not show any significant changes during the “8.2 ka event” and “9.3 ka event” intervals (Figs 5a–d). The most likely reason for such discrepancies was the effect of drastic cooling over a short time. According to Alley and Ágústsdóttir (2005), the cooling event at ca. 8.2 ka BP might have caused the low water level and a declined spruce in Alaska (Andreev and Peteet, 1997; Mason et al., 2001), indicating a decrease in aquatic productivity and woody plant production. Mason et al. (2001) further suggested that extensive weathering and erosion in these regions, which supported an increase in fossil n-alkane production at that time. However, glacial advance in the St. Elias Mountains culminated at about the same time (Denton and Karlén, 1973), which might indicate that the transportation of fossil n-alkanes was restricted to some extent. Moreover, Brubaker et al. (2001) indicated that the herb pollen percentage decreased more than the percentage of xylophyta in sediments from six lakes in southwestern Alaska at ca. 9.3 ka BP. As a result of these changes, although the production of alkanes was limited during these two rapid cooling periods, the contribution of nC₂₇ (Fig. 4d), CPI, ACL, nC₃₁/nC₂₇ ratio and AI showed no significant decrease or increase. Climate signal transmission between sea and land, and early diagenesis of sediments, may also have caused these differences.