The marine transgression starting at the end of the last glacial period flooded a complex permafrost landscape with intact and thermokarst-affected ICD, lakes, and rivers3. Terrestrial sites near the Buor-Khaya Bay show a sequence of deposits, with Holocene deposits at the top that are frequently underlain by ICD (10–50 ka)2, but also fluvial/alluvial sands from the MIS 4/5a-d (50–110 ka), thermokarst deposits formed during the MIS 5e interglacial (115–130 ka), and older ICD from the MIS 6/7 stadial (130–200 ka)2,6.
Optically stimulated luminescence dating together with grain sizes are consistent with the hypothesis that subsea permafrost of the Buor-Khaya Bay largely does not represent ICD but instead older material of different origin. Depositional ages of core 4D-12 fall between 162 ka at 51 m depth and 51 ka at 17 m depth below the sea floor (Fig. 2 and Supplementary Table 1). Although this period covers a considerable range in climatic conditions, a close, linear correlation of age and depth (Pearson’s correlation, R2 = 0.98, n = 5) suggests a rather constant deposition at least at the low temporal resolution available. The measured age of 8.5 ± 0.6 ka at 15 m depth falls outside this pattern and reflects potentially a period of low deposition or an erosional event. Even excluding this single observation, the comparatively young OSL ages observed here at great depths of core 4D-12 contrast to previously determined radiocarbon ages of >40 ka in ICD above sea level around the Buor-Khaya Bay, including on Muostakh Island34,35. This could indicate inconsistencies in OSL vs radiocarbon-based ages, or a large shift in depositional environment between coastal and subsea permafrost sediment sequences.
The subsea permafrost cores were characterized by high variability in grain size distribution. Cores 4D-12 and 2D-13 showed alternations between finer (silt, clay) and coarser-grained (sand) deposits with mostly unimodal grain size distributions (Fig. 2). These grain size distributions contrast to the typically multimodal distribution of ICD with peaks in both silt and fine sand fractions35,36, and rather suggest fluctuations between predominant wind- and water-related deposition. Core 4D-13 showed largely bimodal distributions with silt and sand at the same depth. Predominantly sandy material has been previously described for core BK-2 from the eastern Buor-Khaya Bay28 where subsea permafrost also developed by coastal erosion and inundation as for our locations near Muostakh Island. In Ivashkina Lagoon where subsea permafrost was formed by salinization of a coastal thermokarst lake, mostly sandy material was observed for 1D-14, and a transition from silt to sand with depth for 5D-13. Overall, pronounced differences in grain size distribution point at high spatial variability of deposition regimes even within the small study area of the Buor-Khaya Bay.
Organic matter sources
Organic matter properties at the current thaw front of subsea permafrost were compared with previously published data from terrestrial deposits in northeast Siberia that could resemble the original state of subsea permafrost before inundation. These include (1) Pleistocene ICD, (2) Pleistocene and Holocene thermokarst deposits, (3) Holocene peat permafrost, and (4) Holocene active layers. We further included translocated organic matter in our comparison, such as (5) Pleistocene fluvial/alluvial deposits, (6) contemporary river-suspended material where data from fluvial/alluvial deposits were not available, and (6) contemporary, marine surface sediments from the Buor-Khaya Bay that receives strong input from the Lena River.
Predominant fluvial/alluvial deposition of organic matter at the thaw front of subsea permafrost near Muostakh Island was indicated by the content and 13C isotopic composition of organic carbon and a suite of biomarker source proxies. Organic carbon contents averaged 0.8 ± 0.3% (mean ± standard deviation) for 4D-13, 2D-13, and 4D-12, as well as 0.5 ± 0.7% as previously described for core BK-228. Organic carbon contents showed no systematic trends across these cores and fell in the same range as those of fluvial/alluvial sediments deposited during different periods of the Pleistocene2 (Fig. 3). Organic carbon contents of subsea permafrost were in general lower than in Pleistocene ICD, thermokarst deposits2, and near-by Buor-Khaya Bay surface sediments14,37. Cores 5D-13 and 1D-14 from Ivashkina Lagoon showed higher organic carbon contents at the surface (Supplementary Table 2) that likely reflect Holocene sediments, with an overall average of 1.4 ± 1.4% (previous38 plus own data; one observation of 26% excluded as an outlier). The 13C isotopic composition of organic carbon provides additional information on organic carbon sources. Eastern Siberia is dominated by C3 vegetation with δ13C values commonly between −25‰ and −30‰39. Terrestrial permafrost in the region, such as ICD and thermokarst deposits, fall at the lower end of this range2, and fluvially or marine translocated deposits at the upper end2,37 (Fig. 3). This gradient reflects changes in the isotopic composition of organic matter by processing during transport, together with mixing with comparatively 13C-enriched organic matter from aquatic primary production. Average δ13C values of 4D-13, 2D-13, and 4D-12 were in line with organic matter modification during fluvial translocation (−25.0 ± 1.0‰). Cores 5D-13 and 1D-14 from Ivashkina Lagoon showed more depleted values at the surface leading to higher down-core variability (−25.4 ± 1.7‰), and core BK-2 had overall more depleted δ13C values of −26.3 ± 0.8‰ as previously described40. Similarly, mass ratios of total organic carbon over total nitrogen (OC/TN) were higher at the surface and more variable in 5D-13 and 1D-14 (15.1 ± 4.8) compared to 4D-13, 2D-12, and 4D-12 (9.4 ± 2.5; Fig. 3).
Biomarker proxies based on n-alkanes and lignin phenols permit a more detailed assessment of organic matter sources to subsea permafrost near Muostakh Island. These support contributions from boreal forests, tundra and—to a small extent—peatlands, as well as the modification of terrigenous organic matter during aquatic transport. Concentrations of terrestrial high-molecular-weight (HMW) n-alkanes, n-alkanoic acids and n-alkanols, as well as lignin phenols amounted to 2.7 ± 2.6, 1.5 ± 1.8, 1.0 ± 0.5, and 5.3 ± 5.2 mg g−1 OC, respectively (Supplementary Fig. 1).
The modification of terrigenous organic matter during aquatic transport is supported by terrigenous-to-aquatic ratios (TAR), complementing organic carbon δ 13C values. The TAR builds on the much higher abundance of HMW relative to low-molecular-weight (LMW) n-alkanes in higher plants compared to lower plants such as algae and mosses41. In northeast Siberia, TAR sharply delineates terrestrial vs aquatic deposits following the abundance of higher vs lower plants in these systems (Fig. 4). Ice Complex and thermokarst deposits showed much higher TAR than Buor-Khaya Bay sediments that however still by far exceeded 1 (19 ± 3)42. This indicates a higher contribution of n-alkanes from lower plants than in ICD and thermokarst deposits, but still a dominance of higher-plant n-alkanes due to high input of land-derived organic matter by the Lena River. The TAR values of subsea permafrost near Muostakh Island fall in the same range and are in line with the TAR signal of terrigenous organic matter after aquatic translocation, rather than with that of the terrigenous source itself. An even weaker terrestrial signal has been previously reported for subsea permafrost in Ivashkina Lagoon43.
Lignin proxies and C25/(C25 + C29) n-alkane ratios allow to assess the contribution of organic matter from different terrestrial sources. Low C25/(C25 + C29) n-alkane ratios of subsea permafrost indicate a low contribution of peat-forming Sphagnum moss. The range observed in subsea permafrost was similar to that of mineral deposits in Siberia, including non-peat active layer44, ICD36,44, and Holocene thermokarst36 (Fig. 4). By comparison, active layer and permafrost material from Siberian peatlands was characterized by higher C25/(C25 + C29) ratios45. Intermediate ratios were observed in contemporary material influenced by aquatical translocation such as Buor-Khaya sediments and suspended material in the Kolyma river, indicating varying contribution of peat material from the respective drainage basins42,46,47.
Lignin proxies allow to apportion organic matter from different terrestrial higher plants. Syringyl/vanillyl (S/V) and cinnamyl/vanillyl (C/V) lignin phenol ratios reflect the relative contribution of lignin from angiosperm vs gymnosperm and non-woody vs woody plant tissues. The wide S/V and C/V range of 0.59 ± 0.18 and 0.34 ± 0.25 at the thaw front of 4D-13, 2D-13, and 4D-12 indicates a contribution of both woody gymnosperm and non-woody angiosperm tissues (Fig. 5), with changes in their relative proportion over time (Supplementary Fig. 1). Non-woody angiosperm vegetation is characteristic for tundra landscapes with abundant shrubs or grasses that dominated late Pleistocene Beringian landscapes as well as today’s coastline. This is reflected also in high S/V and C/V ratios in Pleistocene ICD, Holocene peat permafrost, and tundra active layer soils14,48. By contrast, woody gymnosperm material was likely translocated by rivers from boreal forests in the south. Vegetation reconstructions suggest that trees have been sparse in the study region since at least the MIS 5e interglacial49,50 ca. 115 ka before present, compared to an OSL-dated age of 65–72 ka for the section of core 4D-12 that was analyzed for biomarkers. Similar long-range translocation of forest organic matter can be observed under present conditions in eastern Siberia. Organic material suspended in the Lena River and deposited in Buor-Khaya Bay sediments is characterized by a high contribution of woody gymnosperm tissues that has been inferred to reflect transport from boreal forests far south of the Lena delta14,48 (Fig. 5).
Organic carbon storage and thaw
Subsea permafrost represents a large organic matter pool that, if thawed and microbially degraded, might be an increasing source of CO2 and CH4 to the atmosphere. The average organic carbon content of subsea permafrost cores 4D-13, 2D-13, 4D-12, 1D-14, 5D-13, and BK-2 amounted to 0.7 ± 0.3% or 9.3 ± 3.6 mg cm−3, based on the dry sediment mass per total volume of cores 2D-13 and 4D-12 (1.3 ± 0.2 g dry weight cm−3), and excluding the upper-most meter that might represent Holocene sediments. The rates of subsea permafrost thaw of 14 ± 3 cm year−1 in the study area16 thus correspond to the thaw-out of 1.3 ± 0.6 kg OC m−2 year−1. By comparison, the gradual thaw of terrestrial permafrost in northeast Siberia by active layer deepening is here estimated to yield 0.14 kg OC m−2 year−1 (standard deviation 0.32 kg m−2 year−1). This estimation is based on the average active layer deepening rate of 0.4 cm year−1 in central and eastern Siberia between 1990 and 2020 (stations with minimum 4 years of observation)20,21, the relative distribution of permafrost soil suborders9, and their average organic carbon density51 between the minimum and maximum active layer depth20,21.
Degradation state of organic matter at the thaw-front
Organic matter now in subsea permafrost may have been degraded at the site of origin, during transport, after re-deposition before freeze-down, and again after re-thaw. Lignin- and lipid-based degradation proxies suggest that terrestrial organic matter in translocated deposits such as river suspended material and marine sediments14,42,46,47 is more degraded than in terrestrial (source) deposits14,36,44,45,48 (Fig. 6). Although bulk organic matter and biomarker source proxies indicate the aquatic translocation and modification of organic matter at the thaw front of subsea permafrost in the Buor-Khaya Bay, differences between biomarker signatures give an inconsistent overall picture of its degradation state. Lignin-based degradation proxies (acid/aldehyde ratios of syringyl and vanillyl phenols, 3,5-dihydroxybenzoic acid/vanillyl ratios) suggest a lower degree of decomposition of organic matter found in subsea permafrost than in terrestrial deposits; lipid-based degradation proxies (carbon preference indices of HMW n-alkanes and HMW n-alkanoic acids, HMW n-alkanoic acid/HMW n-alkane ratios) indicate the opposite (Fig. 6). Correlations among subsea permafrost samples show that strong degradation indicated by lipid-based proxies coincided with low degradation based on lignin-based proxies and vice versa (Supplementary Table 3). It is possible that the observed discrepancies between degradation proxies reflect lower degradation of lignin than lipids, for instance, due to the limited abundance of specialized lignin-degrading microorganisms. Alternatively, the differences might stem from the stability of lignin degradation proxies during decomposition under anoxic conditions52 or abiotic alteration of proxies over long time frames. Other possibilities include an effect of hydrodynamic sorting. A previous study has observed differences in lignin- and lipid-based degradation proxies between ESAS sediment fractions, with more degraded, lipid-rich organic matter associated with fine-grained sediments, and less degraded, lignin-rich organic matter in coarser sediments53; changes in depositional regime could thus differently affect lignin- and lipid-based degradation proxies.
High-resolution samples from above and below the IBPT represent a continuum from organic matter that has been constantly frozen since the Pleistocene, to organic matter that has thawed in recent decades. Linear interpolation between IBPT positions measured in 1982/3 and 2013 suggests the onset of thaw at the upper limit of the high-resolution sections only about 10 and 20 years before drilling in cores 4D-13 and 2D-13. We examined potential changes in degradation proxies over this time by comparing organic matter properties above and below the IBPT and testing for correlations with distance from the IBPT in the thawed core part. The few significant effects were non-systematic in direction (Supplementary Table 4), and occurred abruptly (Supplementary Fig. 1). This indicates that the observed variability in degradation proxies was likely driven by differences in the degradation state of organic matter at the time of deposition rather than by decomposition upon thaw. However, the data do not exclude organic matter decomposition after thaw per se; changes in degradation proxies could be masked by source variability or might become detectable only after longer periods of thaw and degradation.
Greenhouse gas production in recently thawed subsea strata
Microbial degradation of organic matter following subsea permafrost thaw might lead to greenhouse gas release. During anoxic incubation at 4 °C, CO2 production from thawed subsea permafrost was highest at the beginning (58–250 µmol g−1 OC d−1) and decreased over time. The average rate was 2.4 µmol g−1 OC d−1 (0.7–6.4 µmol g−1 OC d−1) over the entire period of 600 days (Supplementary Fig. 2 and Supplementary Table 5). Very weak N2O production was observed after the CO2 peak, in the first 1–5 day interval for samples 4D-13, 2D-13, and 4D-12 and in the 9−15 day interval for 1D-14 (2–55 nmol g−1 OC d−1). Nitrous oxide production was followed by net N2O consumption to concentrations below the detection limit for the rest of the incubation (Supplementary Table 5). Methane production was first observed for the 1–5 day interval in the 4D-13 sample, and for the 9–16 day interval in the 2D-13, 4D-13 and 1D-14 samples (4–38 nmol g−1 OC d−1; Supplementary Fig. 2 and Supplementary Table 5). In each case, the first occurrence of CH4 also represented the peak in CH4 production rates that then decreased to <3 nmol g−1 OC d−1 after 600 days. On average, CH4 production amounted to 1.7 nmol g−1 OC d−1 (0.4–4.1 nmol g−1 OC d−1). The sequence of peaks in CO2, N2O, and CH4, each followed by a decline in net production, points at a transition from electron acceptors of higher to lower redox potential during the incubation.
Both CO2 and CH4 production were well described by two-stage decomposition models. Linear correlation of modeled vs observed cumulative CO2 and CH4 production showed R2 > 0.99 for all samples. This indicates that organic matter degradation dynamics over the incubation period could be well approximated by two discrete components. For CH4, the two model components likely reflected two organic carbon pools of different degradability, and the more easily degradable pool accounted for less than 0.001% of initial organic carbon (see Supplementary Table 6 for fitted parameters). For CO2, the transition to N2O and CH4 production after the CO2 peak suggests not the exhaustion of a more easily degradable carbon pool, but rather a depletion of suitable electron acceptors, behind the two modeled stages of decomposition. Less than 0.2% of the initial organic carbon was mineralized to CO2 in the first decomposition stage (Supplementary Table 6).
Subsea permafrost in the Buor-Khaya Bay showed similar organic carbon losses as terrestrial permafrost, and comparatively low CH4 production. During the first year of incubation, subsea permafrost lost on average 1.3% of organic carbon (range 0.4–3.5%). Previous anoxic, cold-temperature (4 °C) incubations have reported similar losses for terrestrial permafrost of low organic carbon content (<5%), including Pleistocene permafrost, Holocene permafrost, and thermokarst material (Fig. 7)10,22,25. During the first year of incubation, CH4 production averaged 0.1% (0.04–0.23%) of organic carbon lost from subsea permafrost, at a rate of 2.1 nmol CH4 g−1 OC d−1 (0.5–5.2 nmol g−1 OC d−1). These ranges are comparable to those of Pleistocene permafrost, but lower than those of Holocene permafrost and thermokarst deposits over the same time frame (Fig. 7)10,22,25. Previous studies on terrestrial permafrost have however also highlighted high variability in CH4 production rates, frequent CH4 production below the detection limit, and multi-month lag times before its onset10,22,25,54. Similarly, high variability in CH4 production has been observed for subglacial deposits of low organic carbon content, with average rates between 0.3 and 1100 nmol CH4 g−1 OC d−1 during 1 year at 4 °C55. We here found the highest CH4 production per organic carbon in samples with a stronger influence of aquatic translocation, indicated by lower organic carbon and nitrogen content, OC/TN and TAR, S/V and C/V lignin phenol ratios suggesting forest sources, as well as higher δ13C values and C25/(C25 + C29) n-alkane Sphagnum proxies (Spearman’s correlation p < 0.1). Potential mechanisms behind variability among subsea permafrost samples and compared to terrestrial permafrost include differences in organic matter degradability, methanogenic microbial community composition and activation after thaw, and pore fluid composition, including electron acceptors of higher redox potential that could inhibit CH4 production or facilitate CH4 consumption.
These constraints on CH4 production by decomposition in thawed subsea permafrost may be compared with estimates of ocean-atmosphere CH4 fluxes. Combining the organic carbon thaw-out rate described above with methanogenesis observed during the experiment results in an estimated production of 3.6 µmol CH4 m−2 d−1 (standard deviation 4.0 µmol CH4 m−2 d−1) during the first 600 days after thaw. If methanogenesis persists longer, or even increases over time as observed for terrestrial permafrost10,22,54, CH4 production by decomposition in thawed subsea permafrost could be substantially larger. The thaw depth of the six cores mentioned in this study, and of two additional cores from the Buor-Khaya Bay16 averages 14 ± 7 m in 2012–2014. Extrapolating the two-stage model beyond 600 days results in the production of additional 129 µmol CH4 m−2 d−1 (standard deviation 167 µmol CH4 m−2 d−1). Since the model converges to constant CH4 production rates once the easily degradable pool is depleted, this extrapolation does not account for the eventual decrease in CH4 production that is expected with the depletion of all carbon pools over longer time frames. While this quantitative assessment establishes a scale for potential CH4 production by decomposition processes in thawed subsea permafrost, the in situ release of CH4 can be influenced by many factors that are not accounted for in our experiment. Methanogenesis rates might be affected by changes of the microbial community over time56, or by intrusion of seawater, potentially transporting microbial substrates that could facilitate or inhibit CH4 production such as sulfate28. A previous meta-analysis reported also a decreasing contribution of CH4 to organic carbon losses with lower temperatures11. Methane production might consequently be lower than in our incubation under the in-situ temperatures around zero. On the other hand, part of the produced CH4 is likely aerobically or anaerobically oxidized within thawed subsea permafrost28,57, in surface sediments and the overlying water, before it reaches the atmosphere. Ocean-atmosphere fluxes of CH4 in the vicinity of the subsea permafrost drilling locations are in the order of 300–1300 µmol m−2 d−1 based on several years of field observations26. The lower rates of CH4 production by subsea permafrost decomposition estimated here, and the likely oxidation of part of this CH4, do not point to a dominant contribution of organic matter decomposition in thawed subsea permafrost to the high emissions observed in the area. We emphasize, however, the high variability of observed CH4 production rates, and the limitations of upscaling from incubations to natural environments. Taken together, the high CH4 emissions ubiquitously observed in the field likely stem from other sources such as preformed CH4 in gas pockets in the subsea permafrost, collapsing CH4 hydrates, or venting of a deep thermogenic CH4 pool.
The stable isotopic values of CH4 generated during our incubation fall in line with the range previously reported for microbial fermentation58. The δD values averaged −300 ± 14‰, and δ13C −65 ± 5‰ (Supplementary Table 7). The here determined fingerprint of CH4 from microbial degradation of subsea permafrost organic matter can be combined with isotopic fingerprints of other potential CH4 sources to calculate the relative contributions of these sources to the CH4 release observed in the field.
In addition to the possible release of the strong greenhouse gas CH4, our incubation experiment suggests that thawing subsea permafrost might be a so-far less considered driver of ocean CO2 emissions and ocean acidification. Using the same approach as for CH4, we estimate the average production of 5.2 mmol CO2 m−2 d−1 (standard deviation 6.2 mmol CO2 m−2 d−1) in the first 600 days after thaw, and an additional 201 mmol CO2 m−2 d−1 (standard deviation 232 mmol CO2 m−2 d−1) over longer time frames. Previous studies have reported high concentrations of CO2 in the near-coastal Laptev Sea, the net release of CO2 to the atmosphere, and strong ocean acidification by CO2 dissolution threatening ocean fauna, and have linked these to the decomposition of terrigenous organic matter to CO2 in the ocean32,59. The respiration of terrigenous organic matter has been estimated to 5.9 mmol C m−2 d−1 for the outer Laptev Sea60. Rates are likely higher in our study area closer to the coast considering the higher concentrations of terrigenous organic matter32 and water column CO259. Hence, the CO2 flux estimate based on subsea permafrost incubation is of a magnitude relevant for CO2 dynamics in ESAS waters. Under natural conditions, however, inorganic carbon produced by organic matter decomposition in subsea permafrost will be present not only as CO2, but also as carbonate and bicarbonate, in pH-dependent proportions. How much CO2 is eventually released to the water column will also depend on dissolved inorganic carbon consumption by microbial processes and the balance of precipitation and dissolution reactions. Nevertheless, our findings suggest that organic matter decomposition in recently thawed subsea permafrost could also play a role in CO2 emissions and ocean acidification in areas of the rapid thaw of subsea permafrost.
Subsea permafrost on the extensive Eurasian Arctic Ocean shelf seas is rapidly thawing due to natural and anthropogenic warming. We here characterize organic matter composition and dynamics in a set of subsea permafrost drill cores from the southeast Laptev Sea that reflect sediment deposition in a heterogeneous and dynamic landscape over the past 160,000 ka. Lignin- and lipid-based biomarker proxies indicate the contribution of organic matter originating from tundra and forest to the current thaw front of subsea permafrost near Muostakh Island, and the modification of this organic matter during aquatic transport. Although organic carbon content was here relatively low (average 0.7 ± 0.3%), the high rates of permafrost thaw of 14 ± 3 cm year−1 at the study site yield a thaw-out of 1.3 ± 0.6 kg OC m−2 year−1. These rates exceed those for terrestrial sites by a factor of 35 for the deepening of the permafrost table, and by a factor of nine for organic matter thaw. Constraining the susceptibility of the vast and rapidly thawing subsea permafrost organic carbon pool to degradation is urgently needed for improving estimates of greenhouse gas emissions from all permafrost compartments. This study provides constraints for subsea permafrost on potential CO2, CH4, and N2O production by organic matter decomposition upon thaw and the isotopic composition of CH4. Our findings point to other sources than microbial degradation of thawing subsea permafrost as the main drivers of the high CH4 emissions in the study area. Subsea permafrost might however be a contributor to strong ocean acidification in the East Siberian Arctic Shelf region that has not been considered so far.