Remobilization of Old Permafrost Carbon to Chukchi Sea Sediments During the End of the Last Deglaciation.

Climate warming is expected to destabilize permafrost carbon (PF-C) by thaw-erosion and deepening of the seasonally thawed active layer and thereby promote PF-C mineralization to CO2 and CH4. A similar PF-C remobilization might have contributed to the increase in atmospheric CO2 during deglacial warming after the last glacial maximum. Using carbon isotopes and terrestrial biomarkers (Δ14C, δ13C, and lignin phenols), this study quantifies deposition of terrestrial carbon originating from permafrost in sediments from the Chukchi Sea (core SWERUS-L2-4-PC1). The sediment core reconstructs remobilization of permafrost carbon during the late Allerød warm period starting at 13,000 cal years before present (BP), the Younger Dryas, and the early Holocene warming until 11,000 cal years BP and compares this period with the late Holocene, from 3,650 years BP until present. Dual-carbon-isotope-based source apportionment demonstrates that Ice Complex Deposit-ice- and carbon-rich permafrost from the late Pleistocene (also referred to as Yedoma)-was the dominant source of organic carbon (66 ± 8%; mean ± standard deviation) to sediments during the end of the deglaciation, with fluxes more than twice as high (8.0 ± 4.6 g·m-2·year-1) as in the late Holocene (3.1 ± 1.0 g·m-2·year-1). These results are consistent with late deglacial PF-C remobilization observed in a Laptev Sea record, yet in contrast with PF-C sources, which at that location were dominated by active layer material from the Lena River watershed. Release of dormant PF-C from erosion of coastal permafrost during the end of the last deglaciation indicates vulnerability of Ice Complex Deposit in response to future warming and sea level changes.

Introduction

Arctic carbon cycling is particularly susceptible to climate warming due to the amplification of global warming in the Arctic and the large amounts of organic carbon (OC) perennially frozen in near‐surface deposits. Permafrost soils and other deposits today store about 1,300 ± 200 Pg carbon (PF‐C; Hugelius et al., 2014), an amount twice the size of the preindustrial atmospheric carbon pool (Ciais et al., 2012). Previous studies have highlighted the potential for increasing greenhouse gas emissions (e.g., CO2 and CH4) from the Arctic as rising temperatures are expected to thaw and destabilize dormant PF‐C pools in the next decades (e.g., Koven et al., 2011; Schuur et al., 2015). Deepening of the seasonally thawed active layer can gradually mobilize PF‐C, whereas thermokarst formation and coastal erosion abruptly release deeper stores of older PF‐C (Olefeldt et al., 2016; Vonk et al., 2012). Especially abrupt PF‐C release processes are poorly represented in current earth system and climate models, stressing the limited understanding of PF‐C release rates and thresholds in a warming climate (Schneider Von Deimling et al., 2015; Vonk & Gustafsson, 2013).

During the last glacial maximum (LGM; 26,500–19,000 cal years before present [BP]; Clark et al., 2009), PF‐C stocks might have been substantially larger than today (Ciais et al., 2012; Zimov et al., 2009). In this period, the circum‐Arctic landmass between eastern Siberia and Alaska was not covered by ice sheets but by a cold tundra‐steppe, which formed thick ice‐ and carbon‐rich deposits (Ice Complex Deposit [ICD]; e.g., Schirrmeister et al., 2013). The current shallow sea north of Siberia was then above sea level, exposed to the atmosphere and hosted permafrost‐dominated tundra landscapes (Hubberten et al., 2004; Romanovskii et al., 2000; Shakhova et al., 2017). Eustatic sea level rise of ~120 m during the last deglaciation (Lambeck et al., 2014) subsequently submerged this tundra between 15,000 and 7,000 cal years BP, forming the East Siberian Arctic Shelf (ESAS, Figure 1) including the Laptev, East Siberian, and Russian part of the Chukchi Sea (Bauch et al., 2001; Stein & Macdonald, 2004). Earlier studies of sediment cores from the ESAS (e.g., Bauch et al., 2001; Keskitalo et al., 2017) have suggested that landmass flooding caused erosion of coastlines during periods of fast sea level rise (e.g., during meltwater pulse 1A; Lambeck et al., 2014). It is also known that the postglacial warming stimulated large‐scale degradation and thaw of ground ice in shore‐based ICD and created thermokarst (Walter‐Anthony et al., 2014). In the near‐coastal areas, the sea level transgression and erosion of coastal permafrost may have either flooded and buried or remobilized ancient PF‐C into the marine carbon cycle; the latter is similar to what is observed at present along the ESAS coasts where mobilization rates are in the order of 22 Tg ICD‐OC per year (Vonk et al., 2012).

Figure 1

Paleotopographic scheme of the East Siberian Arctic Shelf ~11,000 years BP with the location of the 4‐PC1 core (SWERUS‐C3 expedition 2014) in the Chukchi Sea (CS). The light gray shades show the International Bathymetric Chart of the Arctic Ocean v3 bathymetry between 0‐ and 50‐m depth that approximately corresponds with the land exposed 11,000 cal years BP with an eustatic sea level of ~50 m below present (Jakobsson et al., 2012; Lambeck et al., 2014). The dark gray area with a black outline shows the current position of the coastline. Orange shades show the possible extent of ice and carbon‐rich deposits in Arctic and sub‐Arctic lowlands of Siberia and Alaska, as well as mapped and potential subsea ice and carbon‐rich deposits (Strauss et al., 2017). Target symbols in the LS and ESS show coring sites where historical permafrost carbon remobilization has been studied previously (LS/PC‐23 = Tesi, Muschitiello, et al., 2016; ESS/GC‐58 = Keskitalo et al., 2017). BS = Bering Strait; BFS = Beaufort Sea; ESS = East Siberian Sea; LS = Laptev Sea; bsl = below sea level.

A few previous studies, based either on modeling or observations at lower latitudes, have hypothesized that part of the PF‐C that accumulated during the last glaciation might have been released back to the atmospheric pool of CO2 during the last deglaciation (Ciais et al., 2012; Crichton et al., 2016; Köhler et al., 2014). Starting from an initial postglacial temperature increase around 17,500 cal years BP, the Bølling‐Allerød interstadial (14,700 to 12,900 cal years BP) was the first warming period after the LGM, followed by a short but cold regression during the Younger Dryas (from ~12,900 to 11,700 cal years BP; Carlson, 2013), and then abrupt warming at the end of the Younger Dryas and the transition to the early Holocene (from 11,700 cal years BP; the current interglacial). Atmospheric CO2 levels rose by 80 ppmv between 17,500 and 11,000 cal years BP, representing the full glacial‐interglacial anomaly (Marcott et al., 2014). A portion of old carbon—possibly from permafrost—equivalent to 10 ppmv CO2 entered the active carbon cycle rapidly within a century after the Bølling‐Allerød onset around 14,600 cal years BP and caused a drop in atmospheric Δ14C (Köhler et al., 2014).

Despite the obvious implications for global radiative forcing, observational paleorecords for the release of carbon from Arctic permafrost are scarce, with inferences relying largely on carbon cycle models (Ciais et al., 2012; Crichton et al., 2016; Köhler et al., 2014; Simmons et al., 2016). Direct reconstructions of large‐scale PF‐C remobilization, including the remobilization of PF‐C from the ESAS during the last deglaciation, are rare in part due to the availability of only very few suitable geoarchives. However, a study of late deglacial sediments (11,690–11,140 cal years BP) from a Laptev Sea core (PC‐23, Figure 1; Tesi, Muschitiello, et al., 2016) showed that temperature increases during the Younger Dryas‐Holocene transition caused OC fluxes to Laptev Sea sediments much higher (101 ± 18 g·C·m−2·year−1) than contemporary fluxes (up to 25 g·C·m−2·year−1). These fluxes were mostly associated with active layer deepening in the large Lena River catchment and are estimated to have released 31 ± 9 Pg PF‐C to the Laptev Sea during the Younger Dryas/early Holocene warming (Tesi, Muschitiello, et al., 2016). In contrast, sediments of the Holocene climate optimum (9,500–8,200 cal years BP) from the East Siberian Sea received almost no river‐transported PF‐C (Keskitalo et al., 2017). This suggests that fluvial PF‐C from active layer sources may be distributed rather locally. It is unknown if deglacial permafrost thawing and OC mobilization also occurred in other parts of the Siberian Arctic or during other earlier warming periods than the early Holocene. An increased geographical coverage of studies reconstructing past PF‐C mobilization is needed to understand the vulnerability of permafrost to environmental change and climate warming after the LGM.

In this study we assess the delivery of PF‐C to Chukchi Sea sediments from the late Allerød at 13,000 cal years BP to the early Holocene at 11,000 cal years BP and compare this late deglacial period with the late Holocene from 3,650 cal years BP until present. The main objectives are (i) to reconstruct and quantify postglacial PF‐C fluxes to the Chukchi seabed, (ii) to distinguish PF‐C from different sources (i.e., ICD vs. soil active layer), (iii) and to describe the degradation status of deposited PF‐C, at relatively high temporal resolution (>10 samples/1,000 years). For this purpose, we characterized the isotopic (Δ14C and δ13C) and molecular (lignin phenols) composition of OC in a well‐dated sediment core covering the regional environmental history of the past 13,000 years. An inverse isotope mixing model of end members from the published literature was used to distinguish and quantify OC fluxes derived from ICD, active layer, and marine sources to the Chukchi Sea sediment core at the end of the last deglaciation. This study contributes toward a better understanding of the dominating permafrost thaw processes and the fate of PF‐C in periods of a warming climate and sea level change.

Material and Methods

Study Area and Environmental Context

The Russian sector of the Chukchi Sea is the easternmost margin of the world's largest shelf sea system—the ESAS—that stretches from the Taymyr Peninsula in the West to the Bering Strait in the East (Figure 1). With an area of 620,000 km2 and an average water depth of 80 m (Jakobsson, 2002), the Chukchi Sea is slightly deeper than the Laptev (48 m) and East Siberian Seas (52 m). The Herald Canyon connects the deeper Central Arctic Ocean with the shallow Bering Strait (~53‐m sill depth) to the Pacific Ocean and is an area for recent paleoceanographic studies (Cronin et al., 2017; Gemery et al., 2017; Jakobsson et al., 2017; Pearce et al., 2017).

Since the opening of the Bering Strait 11,000 ± 200 cal years BP (Jakobsson et al., 2017), nutrient‐rich Pacific water masses have turned the Chukchi Sea into a region of high primary production (Stein et al., 2004). Even though none of the largest Arctic rivers enter it directly, the Chukchi Sea receives some sediments from the Yukon and Anadyr Rivers, whose outflow plumes partially enter through the Bering Strait. Furthermore, terrestrial material may originate from Siberian Rivers that drain directly to the East Siberian Sea (i.e., Kolyma and Indigirka). The Siberian Coastal Current encounters the Pacific Water inflow around 165°E resulting in a frontal zone with river‐dominated waters in the western East Siberian Sea and a more productive marine environment in the eastern East Siberian and Chukchi Sea (Semiletov et al., 2005). The combination of high surface water productivity and the transport of Pacific and shelf waters results in high sedimentation rates of up to 300 cm/kyr in the Barrow (Darby et al., 2009) and the Herald (Pearce et al., 2017) Canyons.

Continuous permafrost dominates the hydrologic catchment of the Chukchi Sea. Accordingly, surrounding continental land masses, that is, East Siberia and Alaska, are covered by prominent Arctic periglacial landscape features, including cryo‐lithological facies such as ICD (also referred to as Yedoma), younger permafrost deposits in wetlands, and hillslope and basin thermokarst landscapes (Olefeldt et al., 2016). The coastlines are susceptible to thermo‐erosion, with variable retreat rates ranging from stable conditions to erosion rates of >2 m/year (Lantuit et al., 2012).

Sampling

The 6.1‐m‐long piston core SWERUS‐L2‐4‐PC1 (further abbreviated “4‐PC1”) was collected in the Herald Canyon (Chukchi Sea), during Leg 2 of the SWERUS‐C3 Expedition on the Swedish icebreaker Oden in August 2014 (26 August 2014; 120‐m water depth; 72°50.32′N, 175°43.63′W). Main criteria for site selection were continuous sediment stratification and high temporal resolution indicated by subbottom profiling using a chirp sonar (Jakobsson et al., 2017). After the core was retrieved on deck, basic geophysical measurements and initial sampling were performed shipboard. Subsamples for total OC (TOC), δ13C, and Δ14C of OC as well as biomarker analyses were taken at 10‐cm resolution along the core; these samples were stored separately and frozen at −18 °C until freeze‐drying in the shore‐based laboratories.

Shipboard measurements of bulk density were carried out using a Multisensor Core Logger (Geotek UK). To determine dry bulk density (ρdry), we measured the fractional porosity and grain density of 10 selected freeze‐dried samples and calculated dry bulk density (ρdry) using the equation ρdry = (1‐porosity) × grain density.

14C and 13C Analyses

Carbonaceous Core Chronology

Mollusk fossils were picked onboard and sent to the National Ocean Sciences Accelerator Mass Spectrometry Facility (NOSAMS; Woods Hole Oceanographic Institution, Massachusetts, USA) for 14C analyses. We extended the earlier‐published 4‐PC1 age model by Cronin et al. (2017) using 14C‐dated benthic foraminifera (mainly Elphidium sp.) from two depth intervals in the base meter of the core (supporting information Figure S1 and Table S1). Bulk samples for collecting benthic foraminifera were stored refrigerated and were wet sieved onboard. Foraminifera were picked at the U.S. Geological Survey (Reston, Virginia) and shipped to NOSAMS for 14C analyses.

As for the 4‐PC1 core chronology presented in Cronin et al. (2017), all ages (including the new 14C‐ages of benthic foraminifera) were calibrated with the Marine 13 calibration curve using the software Oxcal 4.2 (Bronk Ramsey, 2008, 2009). We used two different marine radiocarbon reservoir ages to account for the opening of the Bering Strait and a change in water depth and water mass origin that bathed the Herald Canyon during the late deglaciation and the late Holocene. For the core section older than 11,000 cal years BP, the model applies a ΔR = 50 ± 100 years, which follows from a modern value for the shallow shelf in the Laptev Sea (Bauch et al., 2001) where Pacific water masses are also absent. By contrast, for the younger core section, a ΔR = 300 ± 200 years was applied to account for the inflow of relatively old Pacific water and the presence of Atlantic‐sourced waters in the deeper part (>100 m) of the Herald Canyon.

14C and 13C Analyses of Sediment TOC

All δ13C‐OC data have previously been published in Jakobsson et al. (2017). A subset of 14 samples was sent to NOSAMS for HCl acid vapor treatment and subsequent radiocarbon analyses. Radiocarbon values were initially reported as Δ14C values (Stuiver et al., 1977) and were then normalized to the year of deposition according to the age‐depth model (i.e., corrected for sediment reservoir age). This produces ΔΔ14C values that are the 14C age offset between isotopic equilibration with atmospheric CO2 (photosynthesis by terrestrial plants or marine phytoplankton) and (re)deposition on the sea floor. Sediment OC and isotope data for all samples are listed in Table S3.

Isotope‐Based Bayesian Source Apportionment of OC

A dual‐isotopic mass balance approach is used to resolve the relative contributions of ICD PF‐C, active layer PF‐C, and marine OC to bulk organic matter in the 4‐PC1 sediment core. The statistical source apportionment relies on knowledge of the isotopic source profiles of the three end members, which are ICD PF‐C, active layer PF‐C, and marine OC (Karlsson et al., 2016; Tesi, Muschitiello, et al., 2016; Vonk et al., 2010). The ICD PF‐C and active layer PF‐C end members were derived from an extensive literature database (extended from Tesi, Muschitiello, et al., 2016; Vonk et al., 2012) of published δ13C and Δ14C values of ICD and active layers of permafrost soils in the terrestrial catchment areas of the Laptev, East Siberian, and Chukchi Seas (including Yukon catchment). For the active layer, we thus calculated Δ14C‐OC of −173 ± 163‰ (n = 40 values from the literature; mean ± standard deviation, Table S6) and δ13C‐OC of −27.1 ± 1.0‰ (n = 35) and for ICD, Δ14C‐OC of −962 ± 61‰ (n = 415, Table S7), and δ13C‐OC of −26.3 ± 0.7‰ (n = 374), which are assumed to be constant over the length of the core.

The marine end member builds on literature values for marine phytoplankton with a particle size above 10 μm. We applied two separate isotope value ranges (mean ± standard deviation) to account for the environmental change attributed to the Bering Strait opening 11,000 cal years BP (Jakobsson et al., 2017). An open‐marine setting similar to the Chukchi Sea today is assumed for the last 11,000 cal years, and the marine end member for this period is approximated based on phytoplankton samples from the eastern East Siberian Sea that is characterized by a similar open‐marine setting (Δ14C‐OC = −50 ± 12‰, n = 5; δ13C‐OC = −21.0 ± 2.6‰, n = 31; Table S8). Before 11,000 cal years BP, the southern Chukchi Sea did not receive Pacific water inflow through the Bering Strait (Jakobsson et al., 2017) but was under stronger terrestrial influence. Under such conditions, the photosynthetic refixation of terrestrially derived carbon can lead to lower δ13C values of phytoplankton, as observed today in the western East Siberian Sea and Laptev Sea (Tesi et al., 2017). We therefore considered a wider range of phytoplankton measurements for the older core section that includes samples from the open‐marine setting of the eastern East Siberian Sea, as well as samples from more terrestrially influenced settings in the western East Siberian Sea and Laptev Sea (Δ14C‐OC = −16 ± 53‰, n = 12; δ13C‐OC = −23.2 ± 3.5‰, n = 43; Table S9). A full description of isotopic end members including all references is provided in Tables S6–S9.

A Bayesian approach, similar to the one applied in Keskitalo et al. (2017), was used to account for parameter uncertainty. We took advantage of the low variability of our 62 δ13C data points and introduced a time dependence of the end member proportions based on our 14 ΔΔ14C‐OC points. The exact approach employed in the current study is described in further detail in Text S1 (Plummer, 2016; R Core Team, 2017; Reimer et al., 2013).

Lignin Phenol Biomarker Analysis

Lignin phenols were used as molecular biomarkers for terrestrial organic matter (Table S2). Lignin phenols are breakdown products of lignin biopolymers that are produced only by vascular plants. We applied an alkaline CuO oxidation protocol using a microwave‐based extraction as described in Goñi and Montgomery (2000). Briefly, freeze‐dried subsamples (~500 mg) were loaded into Teflon tubes, mixed with 300‐mg cupric oxide (CuO) and 50‐mg ammonium iron (II) sulfate hexahydrate ((NH4)2Fe(SO4)2·6H2O), and suspended in N2‐purged 2M NaOH solution. The extraction was carried out with an UltraWAVE Milestone 215 microwave digestion system at 130 °C for 90 min. An internal recovery standard (ethyl‐vanillin and cinnamic acid) was added to the extracts at known concentration, and concentrated HCl was added until pH 1 was reached. Samples were extracted twice using ethyl acetate (EtOAc), residual water was removed using anhydrous sodium sulfate (NaSO4), and samples were dried in a CentriVap (Christ RVC 2‐25) at 60 °C for 1 hr and then redissolved in pyridine. Sample aliquots were derivatized with bis‐trimethylsilyl trifluoroacetoamide + 1% trimethylchlorosilane to silylate exchangeable hydrogen atoms. Finally, all samples were analyzed on a gas chromatograph with mass spectrometer detector (7820A, Agilent Technologies, United States) using a DB1‐MS column (30 m × 250 μm; 0.25‐μm film thickness). The temperature program started with an initial temperature of 50 °C, followed by a ramp of 10 °C/min until reaching 300 °C. Concentrations of individual lignin phenols and other CuO oxidation products were quantified against six‐point concentration curves of commercially available external standards (a complete list of analytical standards is shown in Table S2). Total lignin concentrations for each sample are the sum of the individual lignin phenol compounds (Sl, Sn, Sd, Vl, Vn, Vd, pCd, and Fd; abbreviations defined in Table S2). We report TOC‐normalized concentrations (mg·g−1·OC) and the corresponding flux rates (mg·m−2·year−1) for each sample depth in Table S4.

Results and Discussion

Chukchi Shelf Sedimentation History

The 4‐PC1 Chronology and Timing of the Late Deglaciation

Core 4‐PC1 provides a well‐dated record of regional sedimentation history with a basal age of 13,120 years at 613 cm core depth (Figure 3). Ten carbonaceous 14C ages of mollusk shells constitute the basis of the chronology for this core, which has been presented and discussed in previous studies (Cronin et al., 2017; Jakobsson et al., 2017). This study adds two new 14C ages of benthic foraminifera, pooled from integrated depth sequences to the previously unconstrained section below 499 cm (12,100 ± 220 cal years BP; Cronin et al., 2017), between (i) 543–560 cm (13,020 ± 120 cal years BP) and (ii) 598–610 cm (12,830 ± 120 cal years BP; Table S1). The age model described by Cronin et al. (2017) was here revised to include these new 14C dates. The chirp sonar profiles of the Herald Canyon show well‐stratified sediments at the 4‐PC1 coring site above that depth. Furthermore, the age model reconstructs continuous sedimentation for the late deglacial period between 417 and 613 cm, including part of the late Allerød warming period from 13,120 to 12,900 cal years BP, the entire cold Younger Dryas (from ~12,900 to 11,700 cal years BP; Carlson, 2013), and the early Holocene. The upper boundary of the late deglacial unit marks the opening of the Bering Strait at 11,000 ± 200 cal years BP (Jakobsson et al., 2017). A condensed interval, or hiatus, separates this late deglacial archive from the abovelying late Holocene unit.

Figure 3

Record of carbon burial in the 4‐PC1 geoarchive compared with Greenland ice core data and sea level history; (a) North Greenland Ice Core Project members (2004) ice core data; (b) rate of sea level change (Lambeck et al., 2014); (c) 14C activity of OC at the time of deposition (ΔΔ14C); (d) δ13C of OC from the 4‐PC1 sediment core; (e) lignin phenol fluxes; (f) source fractions of ancient carbon‐rich permafrost (ICD), active layer and marine plankton (posterior mean ± standard deviation ranges from isotope‐based source proportion calculations are shown) and (g) resulting OC fluxes for the different organic matter source compartments. OC = organic carbon; PF = permafrost; ICD = Ice Complex Deposit.

Shelf transgression and the opening of the Bering Strait changed the environment from a near‐coastal environment to a more open‐marine depositional setting and allowed inflow of nutrient‐rich Pacific water into the Chukchi Sea (Cronin et al., 2017; Jakobsson et al., 2017). One mollusk at 399 cm core depth revealed a 14C age of 8,850 ± 150 cal years BP, another at 337 cm dates to 3,000 ± 150 cal years BP, indicating a sudden transition above 400 cm depth that is accompanied by changing physical properties of the sediment (i.e., bulk density and P‐wave velocity; Jakobsson et al., 2017). This points toward nonstable sedimentation and probably a hiatus between 400 and 337 cm depth and is dated between 11,000 and 3,650 cal years BP, representing a time frame for which the 4‐PC1 allows no interpretations for time‐related OC dynamics to Chukchi Sea sediments. The remaining, top 337 cm of the 4‐PC1 core, date from the late Holocene and represent more stable deposition (126 cm * 1,000 years−1) over the past 3,600 cal years BP, which is consistent with results from a second Herald Canyon core (Pearce et al., 2017).

Composition and Source of Organic Matter in Historical Chukchi Sea Shelf Sediments

Isotopic Indications of Carbon Provenance

Terrestrial and marine OC can be distinguished using the isotopic composition of bulk OC (δ13C from Jakobsson et al., 2017; ΔΔ14C from this study; Table S3). The stable isotope fingerprint (δ13C) reveals two episodes of contrasting OC sources in 4‐PC1 (Figure 2). Sediments deposited during the end of the deglacial period have depleted δ13C values of −25.1 ± 0.2‰ (average ± standard deviation [s.d.]). These values fall within the typical δ13C range of terrestrial C3 plants (between −23‰ and −30‰; Fry & Sherr, 1989) and suggest that terrestrial carbon was an important source for Chukchi Sea sediments during the late deglacial warming (Jakobsson et al., 2017).

Figure 2

Dual‐isotope composition of OC in 4‐PC1 samples compared with end member ranges used for source apportionment. Active layer refers to the seasonally thawed permafrost surface and ICD to Pleistocene Ice Complex deposits (mean ± standard deviation). The marine OC end member changes at 11,000 cal years BP with the onset of Pacific water inflow as described in section 2.3.3. Note that the radiocarbon fingerprint of the deposited OC was corrected for the time since deposition at the site (i.e., Δ of the Δ14C). OC = organic carbon; BS = Bering Strait; BP = before present; ICD = Ice Complex Deposit.

In contrast, the late Holocene core unit of 4‐PC1 shows more enriched δ13C values (−22.4 ± 0.4‰) that resemble those typical for marine phytoplankton (−18‰ to −24‰ according to Fry & Sherr, 1989). These values agree with published δ13C values of contemporary Chukchi Sea surface sediments of approximately −21.5‰ (Naidu et al., 2000; Stein et al., 2004) and demonstrate that primary production was the dominant source of OC during the middle and late Holocene (Naidu et al., 2000; Stein et al., 2004). The absence of C4 plants in the catchment relevant for 4‐PC1 (Still et al., 2003) excludes that the heavier δ13C‐OC composition in the late Holocene core unit is due to C4‐based photosynthesis.

The radioisotope 14C allows a more detailed distinction of OC sources because 14C contents differ between contemporary pools and preaged reservoirs. Marine phytoplankton and the resulting sediment OC are known to be in equilibrium with the atmospheric concentration of radiocarbon (Tesi et al., 2017) and would be characterized by ΔΔ14C values close to zero in the 4‐PC1, whereas both active layer and ICD PF‐C are preaged and characterized by more depleted ΔΔ14C values (Figure 2). OC deposited during the end of the deglaciation had an average 14C age of 9,590 years at the time of deposition (i.e., residence‐time corrected ΔΔ14C = −693‰), in contrast to OC deposited during the late Holocene that contained more modern carbon (ΔΔ14C = −326‰, 3,180 years). Taken together, the isotopic composition of OC in 4‐PC1 sediments suggests a stronger deposition of preaged terrestrial OC during the late deglacial period than in the late Holocene.

Terrestrial Biomarkers

Lignin is exclusively produced by higher plants and can provide further information about the source of land‐derived material. This terrestrial biomarker has been widely applied in many environments, including Arctic Ocean sediments (e.g., Goñi et al., 2000; Opsahl & Benner, 1995; Tesi et al., 2014). Lignin fluxes to the Chukchi Sea (4‐PC1) were highest between 13,000 and 12,400 cal years BP (36.7 ± 17.8 mg·lignin·m−2·year−1) and decreased towards 11,000 cal years BP (7.6 ± 4.9 mg·lignin·m−2·year−1). Lignin fluxes were 9.9 ± 5.1 mg·lignin·m−2·year−1 during the late Holocene (Figure 3e and Table S4).

Lignin phenols and other products of CuO oxidation also provide qualitative information about the degradation status of organic matter. The ratios of syringic acid over syringaldehyde (Sd/Sl) and of vanillic acid over vanillin (Vd/Vl) increase with oxic degradation (Opsahl & Benner, 1995). The ratio of 3,5‐dihydrobenzoic acid over vanillyl lignin phenols (3,5‐Bd/V) is known to increase with overall degradation of plant organic matter (Dickens et al., 2007; Farella et al., 2001; Otto & Simpson, 2005; Prahl et al., 1994) such as due to increasing transport distance from the coastal zone (Bröder et al., 2016; Tesi et al., 2014). Decomposition proxies analyzed throughout the 4‐PC1 indicate that increasingly degraded lignin was deposited during the late deglacial and onward in the late Holocene (Figure S2). Overall, sediments from the late deglacial period exhibit proxy ratios (Sd/Sl = 0.35 ± 0.04; Vd/Vl = 0.53 ± 0.06; 3,5‐Bd/V = 0.25 ± 0.08) significantly lower than sediments from the late Holocene (Sd/Sl = 0.40 ± 0.05; Vd/Vl = 0.68 ± 0.09; 3,5‐Bd/V = 0.30 ± 0.06; Student's t test Sd/Sl and Vd/Vl p value < 0.001; 3,5‐Bd/V p value < 0.05). This points toward an alternative explanation for the strong differences in OC sources between the late deglacial and the late Holocene core sections. The decrease of lignin and PF‐C fluxes (Figures 3e and 3g) between the period at the end of the deglaciation and the late Holocene may thus be related to the increased transport distance and degradation on the way, given that the coast line moved landwards due to the sea level rise.

The decrease in lignin fluxes, as well as the increase in the degradation state of lignin between 13,000 and 11,000 cal years BP (Figure S2) might also be related to the increasing distance from the shore line caused by continuous sea level rise. As carbon isotopes do not suggest a significant change of OC sources during the end of the deglaciation, lignin might have been rapidly degraded during transport (Bröder et al., 2018; Tesi et al., 2014) or retained to coastal areas by physical sorting (Bröder et al., 2018; Tesi, Semiletov, et al., 2016).

Source Apportionment of Carbon Remobilized to the Chukchi Sea

We quantified OC fluxes from different sources to the Chukchi Sea sediments during the end of the last deglaciation (13,000–11,000 cal years BP). Bayesian source apportionments show that 66 ± 8% (mean ± s.d.) of the total OC deposited during the late deglacial period originated from ICD‐PF, 16 ± 10% from the permafrost active layer and 18 ± 7% from marine sources (Figure 3f and Table S5). This implies that the lateral thaw of deep, ice‐rich permafrost (ICD), likely by coastal erosion due to deglacial sea level rise, mobilized vast amounts of ancient PF‐C. Part of this was clearly redeposited as sediments, whereas another part was likely mineralized to CO2. In contrast, late Holocene sediments of the 4‐PC1 contained on average 58 ± 8% marine OC, 26 ± 4% ICD, and 15 ± 10% active layer PF‐C, suggesting that environmental change—involving sea level change to the present day level, the Bering Strait opening, and enhanced marine productivity—completely altered the relative contribution to OC deposition from terrestrial and aquatic sources by 3,650 cal years BP. Core 4‐PC1 source fractions during the end of the deglacial period are similar to those of East Siberian Sea sediments deposited between 9,500 and 8,200 cal years BP, where ICD accounted for 78% of deposited OC (Keskitalo et al., 2017). In contrast, ICD accounted for only 16% of OC that was deposited off the Lena River paleo channel in the Laptev Sea during the late deglaciation (11,690–11,140 cal years BP; Tesi, Muschitiello, et al., 2016). At that location, a fraction of 67% was instead derived from PF‐C in the active layer, indicating extensive permafrost active layer thaw in the Lena catchment at that time (Tesi, Muschitiello, et al., 2016). Yet in spite of some differences in the relative source fractions of remobilized PF‐C, erosion of permafrost coast lines also took place in the Laptev Sea (Bauch et al., 2001) and caused lateral ICD mobilization at absolute flux rates comparable to the other shelf seas (Keskitalo et al., 2017; this study).

Flux Estimates for Old Carbon Remobilization

Combining source apportionment with sediment accumulation rates from the sediment chronology allows quantifying absolute PF‐C fluxes from ICD and active layer sources to the 4‐PC1. For sediments from the late deglacial, about 82 ± 7% of the OC were attributed to PF‐C (ICD + active layer) which results in an average flux of 9.9 g·C·m−2·year−1 to the 4‐PC1 location. The PF‐C fraction in the late Holocene interval was only 42 ± 8%, which translates into a flux of 4.9 g·C·m−2·year−1 and corresponds to only half of the late deglacial rate. For the period between 13,000 and 11,000 cal years BP, ancient ICD‐sourced PF‐C contributed 8.0 ± 4.6 g·C·m−2·year−1 (Figure 3g), which is similar to ICD fluxes reconstructed for the early Holocene (11,690–11,140 cal years BP) for Laptev Sea sediments (15.3 g·C·m−2·year−1; Tesi, Muschitiello, et al., 2016), and ICD fluxes for central East Siberian Sea sediments (5.4 g·C·m−2·year−1; Keskitalo et al., 2017) during the period 9,500–8,200 cal years BP. Furthermore, previous lake sediment‐based studies have demonstrated inland permafrost thaw and redeposition of permafrost sediments in Beringia following warming episodes during the last deglaciation (Gaglioti et al., 2014, 2017; Walter‐Anthony et al., 2014). We conclude that coastal erosion, in addition to inland permafrost thaw, was a large‐scale phenomenon, enduring over several millennia at the end of the last deglaciation, affecting coastal permafrost along the entire ESAS with remobilization of large amounts of old ICD‐OC.

The PF‐C fluxes resulting from active layer remobilization show little variability throughout the 4‐PC1 (average 1.9 g·C·m−2·year−1), from 13,000 cal years BP to the present. Similarly constant active layer PF‐C fluxes have affected the East Siberian Sea over the past 9,500 cal year (0.6 g·C·m−2·year−1; Keskitalo et al., 2017). In contrast, much higher active layer PF‐C fluxes were reported for the Laptev Sea core in the early Holocene (67 g·C·m−2·year−1) between 11,690 and 11,140 cal years BP, which decreased to 0.04 g·C·m−2·year−1 8,300 years BP (Tesi, Muschitiello, et al., 2016). However, it should be noted that this Laptev Sea core may be biased by its location near the mouth of the Lena River at that time. Similarities in fluxes of thermo‐eroded ICD PF‐C to the Laptev, East Siberian, and Chukchi Sea cores are complemented by spatial differences in the strength of riverine delivery of active layer PF‐C.

Biomarker‐Based Characterization of Terrestrial OC Sources

Lignin data provide complementary information about PF‐C sources. The ratios between syringyl to vanillyl phenols (S/V) and cinnamyl to vanillyl phenols (C/V) are commonly used to characterize the source of organic matter from vascular plants. Syringyl phenols (S) are exclusively produced by angiosperms (Hedges et al., 1988) and high S/V ratios throughout the core 4‐PC1 indicate that organic matter from angiosperm tissues (i.e., grasses and herbs) constituted a key fraction of remobilized PF‐C in Chukchi Sea sediments (S/V = 0.62 ± 0.05; average ± s.d.). Furthermore, C/V ratios (0.21 ± 0.04) demonstrate that the angiosperm tissues originated from a mixture of woody and nonwoody sources (Goñi et al., 1998). Both S/V and C/V ratios are congruent with typical biomarker ratios of ICD of the Siberian Arctic (Tesi et al., 2014), a deposit that was formed in a landscape dominated by grasses and herbs (Andreev et al., 2011; Schirrmeister et al., 2013). The lack of an obvious age trend suggests that terrestrial sources changed very little over the past 13,000 cal years, indicating negligible changes in the relative contribution of different sources also during any putative lignin degradation (Opsahl & Benner, 1995). The lignin composition throughout the 4‐PC1 is similar to those of contemporary surface sediment samples in the Chukchi Sea, for example, in the Barrow Canyon and the Bering Strait (Goñi et al., 2013).

Some differences in sources emerge when comparing lignin source ratios from the Chukchi Sea sediments with those from the East Siberian and Laptev Seas (Figure 4). Lower S/V and higher C/V ratios in East Siberian Sea samples suggest that the contribution of nonwoody gymnosperm tissues for that region was higher over the past 9,500 cal years (Keskitalo et al., 2017). By contrast, the Laptev Sea has received sediments with much lower S/V and C/V ratios since the early Holocene (11,690 cal years BP; Tesi, Muschitiello, et al., 2016). This can be explained by continuously higher contributions of riverine‐transported PF‐C to the Laptev Sea originating from locations further south where gymnosperm trees dominate the boreal forest biome (Winterfeld et al., 2015). In summary, lignin biomarkers support the picture drawn by the carbon isotopes and emphasize that PF‐C remobilized from ICD was the dominating terrestrial OC source to Chukchi Sea sediments at the end of the deglaciation and continues to dominate until the present day.

Figure 4

Lignin fingerprints of 4‐PC1 samples in comparison with Chukchi Sea surface sediments (Goñi et al., 2013), GC58 samples of the East Siberian Sea (Keskitalo et al., 2017), and PC23 samples of the Laptev Sea (Tesi, Muschitiello, et al., 2016), as well as of coastal ICD in Eastern Siberia (mean ± standard deviation from Tesi et al., 2014; Winterfeld et al., 2015) and active layer permafrost (Tesi et al., 2014). The ratio of syringyl to vanillyl lignin phenols (S/V) distinguishes between angiosperm and gymnosperm tissues and the ratio between cinnamyl and vanillyl phenols (C/V) between woody (e.g., stems) and nonwoody tissues (e.g., leaves). Source constraints were adopted from Goñi and Hedges (1995) and Hedges and Mann (1979). ICD = Ice Complex Deposit; BP = before present.

Permafrost Carbon Remobilization During the Cold Younger Dryas

The late deglacial core section of the 4‐PC1 covers the time frame between 13,000 and 11,000 cal years BP and thus encompasses also the short but cold regression of the Younger Dryas (from ~12,900 to 11,700 cal years BP; Carlson, 2013), which interrupted the deglacial warming in many Arctic regions. Our record provides no clear evidence for any mitigation of PF‐C input to the Chukchi seabed due to the Younger Dryas cold event, starting at ~12,900 cal years BP (Carlson, 2013). This may be due to marine shelf sediments providing an integrated signal from a large spatial catchment that would smoothen remobilized PF‐C over a certain period. It is known from the Laptev Sea that the transport time of terrestrial OC across the continental shelf may be several thousand years (Bröder et al., 2018). It is also important to note that, in spite of the cold Younger Dryas regression and advancing glaciers in some regions, evidence of the Younger Dryas causing colder climate in the Chukchi Sea region is contrasting (Kokorowski et al., 2008; Meyer et al., 2010) and may have only affected the winter season, whereas most of the coastal remobilization of PF‐C would be expected in the sea ice free period during late summer. Furthermore, the sea level continued to rise throughout the Younger Dryas, driving the remobilization of PF‐C.

Conclusions and Implications

This study investigates the historical transport of PF‐C to ESAS sediments using a well‐dated sediment core (4‐PC1) from the Herald Canyon, Chukchi Sea. Our results show that PF‐C fluxes to the Chukchi seabed were twice as large during the period of climate warming and sea level rise at the end of the last deglaciation (13,000–11,000 cal years BP; 9.9 g·C·m−2·year−1) as in the late Holocene (4.9 g·C·m−2·year−1; 3,560 cal years BP to today). Using dual isotope‐based source apportionment (δ13C, Δ14C), we identified PF‐C from ICD as the dominant source (66 ± 8%) and active layer PF‐C as a smaller source (16 ± 10%) of OC to Chukchi Sea sediments during the last deglacial period. Analysis of lignin phenol biomarkers independently confirmed high terrestrial OC input between 13,000 and 11,000 cal years BP and showed that organic matter deposited during that period was only poorly degraded. This indicates proximate locations of OC sources and thus relatively short transport times, supporting that thermal collapse and coastal erosion mobilized large amounts of old PF‐C as an effect of sea level rise due to climate warming at the end of the deglaciation. For the late Holocene with a present day sea level, the transport time and distance for PF‐C was larger and may cause some underestimation of the remobilization of PF‐C for the period 3,650 cal years BP to today. More research on PF‐C loss during cross‐shelf transport will be necessary to estimate the additional degradation of PF‐C due to sea level changes. This Chukchi Sea sediment record stands in contrast to the strong deglacial transport of PF‐C from active layer deepening by the Lena River to the Laptev Sea observed in a previous study (Tesi, Muschitiello, et al., 2016). Nevertheless, all three cores from the ESAS (Tesi, Muschitiello, et al., 2016; Keskitalo et al., 2017; this study), so far investigated at this level of detail, support PF‐C mobilization by thermo‐erosion and coastal retreat due to postglacial sea level rise, likely resulting in the loss of vast PF‐C stocks.

The current and previous studies (Keskitalo et al., 2017; Tesi, Muschitiello, et al., 2016) combine to provide an observational record of intensive permafrost destabilization along with release and redeposition of old carbon during the end of the last deglaciation, a period that was—similar to the earlier periods of the deglaciation—characterized by rapid increases in the atmospheric CO2 level (+30 ppmv between 13,000 and 11,000 cal years BP; Marcott et al., 2014). The post‐LGM rise in atmospheric CO2 may in part be explained by PF‐C remobilization and degradation during the marine transgression on today's Arctic Ocean shelves (Ciais et al., 2012; Köhler et al., 2014), supported by the growing number of empirical records of this process in the Arctic. We conclude that thermo‐erosion of ESAS coastlines together with inland active layer deepening exposed large amounts of PF‐C to remobilization, likely resulting in both CO2 emissions and lateral relocation. These findings about historical OC cycling suggest a high vulnerability of coastal ICD permafrost in the flat landscapes of Siberia and Alaska to climate warming and sea level rise. To date, Arctic permafrost still holds twice as much carbon as the global atmospheric carbon pool (Ciais et al., 2012). The current anthropogenic climate warming is expected to increase both direct thermal forcing and a sea level rise of almost 1 m by the year 2100 (RCP 8.5; Intergovernmental Panel on Climate Change, 2014), both mechanisms may again lead to accelerated remobilization of permafrost carbon and enhanced release of greenhouse gases to the atmosphere.

Supporting Information S1

Click here for additional data file.

Table S1

Click here for additional data file.

Table S2

Click here for additional data file.

Table S3

Click here for additional data file.

Table S4

Click here for additional data file.

Table S5

Click here for additional data file.

Table S6

Click here for additional data file.

Table S7

Click here for additional data file.