Grain sizes were analyzed using a laser particle sizer (LS 200, Beckmann-Coulter) with a measuring range of 0.375 to 1000 µm (Figs. 2, S1 in the Supplement). Grain-size calculations were done following Folk and Ward (1957) using Gradistat v8 (Blott and Pye, 2001). A detailed description of these analytical techniques is given in the Supplement (Sect. S1.1 in the Supplement).

To determine the total elemental carbon and total nitrogen (TN) content, the samples were measured by a carbon–nitrogen–sulfur analyzer (Vario EL III, Elementar). TOCwt% was measured with a TOC analyzer (Vario Max C, Elementar). The volumetric TOC content (TOCkgm-3) was calculated according to Strauss et al. (2013). A detailed description of this technique is given in the Supplement (Sect. S1.2).

The TOC / TN (C / N) ratio has been used as a general indicator of the degree of organic-matter decomposition (Stevenson, 1994). Based on the assumption that organic-matter components are degraded selectively, degradation modifies elemental compositions and hence C / N in deposits. Because a decrease in the C / N ratio has been observed in aerated deposits with microbial immobilization of TN (nitrogen stays in the system) accompanied by the remineralization of TOC (Sollins et al., 1984) and CO2 emission, this ratio is used in the following way: the higher the C / N ratio, the lower the degree of decomposition.

BD (bulk density) was calculated using Eq. (1). BD[103kgm-3]=sample dry weight[103kg]sample volume[m3] Estimating the BD is required in order to convert the measured-weight-based TOCwt% content per sample to a volume-based value. Thus, the TOCkgm-3 was calculated according to Eq. (2): TOCkgm-3=BD[103kgm-3]×TOCwt%100.

Stable TOC carbon isotopes were determined with a Finnigan MAT Delta-S mass spectrometer combined with a FLASH elemental analyzer and a ConFlo III gas mixing system. A detailed methodology is given in the Supplement (Sect. S1.4). The stable carbon isotopes of OC reflect (1) an initial contribution from different plant species and plant components and (2) subsequent degradation processes (Gundelwein et al., 2007). Assuming constant photosynthetic isotope fractionation in source plants in the region (C3 plants are ubiquitous in the Arctic; Tieszen, 1973), we use δ13C ratios as a degradation proxy. According to Heyer et al. (1976), decomposition discriminates against the lighter isotope (12C), resulting in more negative δ13C ratios. Thus, this proxy is used in the following way: lower (more negative) δ13C values are connected to less degraded material, while higher (less negative) δ13C values reflect greater decomposition.

Ages were determined by radiocarbon dating of selected macroscopic plant remains performed at the Poznań Radiocarbon Laboratory, Poland (Goslar et al., 2004). The presented radiocarbon ages are uncalibrated ages; Table 1 includes calibrated ages as well. Radiocarbon ages are given in year before present (yr BP).

To look more closely at the molecular composition, we used specific lipid biomarkers. Molecular fossils or biomarkers were studied by chromatography methods coupled with mass spectrometers. Characteristic fractions like n-alkanes, n-fatty acids, sterols, and hopanes were isolated. Because the TOCwt% in the profiles is not equally distributed, we calculated and visualized the biomarker concentration as µg gTOCwt%-1 and µg gSediment-1 (µg gSed-1). For the results, we focus on µg gTOCwt%-1. Every radiocarbon-dated sample and additional samples were used for biomarker analysis. In total 25 biomarker samples were analyzed. Independently of TOCwt%, the sample selection for biomarkers was based on their stratigraphic position with the aim of covering the maximum time period.

For lipid biomarker analyses, 2–12 g of ground sediment was weighed in an extraction cell with an accelerated solvent extractor (ASE 200, Dionex). Samples were extracted with dichloromethane / methanol (99:1). Each sample was held in a static phase for 20 min at 75 ∘C (after 5 min heating, no preheating) at a pressure of 5 MPa. Afterwards, the dissolved compounds were concentrated with a TurboVap (Zymark) closed-cell concentrator and further dried by evaporating the solvent in a stream of nitrogen gas. After that, internal standards (5α-androstane for the aliphatic fraction, ethylpyrene for the aromatic fraction, 5α-androstan-17-on for nitrogen-, sulfur-, and oxygen- (NSO-) containing compounds and erucic acid for the NSO fatty-acid fraction) were added. The amount of internal standards depended on the TOCwt% content (< 10 wt%: 8 µg; > 10 to ≤ 25 wt%: 20 µg; > 25 wt%: 50 µg). After the removal of the n-hexane-insoluble fraction (by the addition of a large excess of n-hexane, called “asphaltene” precipitation), the hexane-soluble portion of the extract was separated by medium-pressure liquid chromatography (MPLC; Radke et al., 1980) into fractions of different polarity (aliphatic and aromatic hydrocarbons as well as polar hetero- (NSO) components). Afterwards, the NSO fraction was split into a fatty acids and an alcohol fraction using a KOH-impregnated silica gel column (Schulte et al., 2000).

For this study, the focus was placed on the aliphatic fraction (containing n-alkanes and triterpenoid compounds) and the NSO fraction (containing n-fatty acids). The fractions were measured by gas chromatography–mass spectrometry (GC-MS). All compounds of interest were identified using the Xcalibur software (Thermo Fisher Scientific).

The n-alkanes, n-alcohols, hopenes (hop-17(21)-ene), and other triterpenoids (β-amyrin (olean-12-en-3β-ol), olean-12-ene, and olean-13(18)-ene) were measured with a GC-MS system (GC – Trace GC Ultra; MS – DSQ; both Thermo Fisher Scientific). Prior to the measurements, the n-fatty acids were methylated with diazomethane and the alcohols were silylated with N-methyl-N-trimethylsilyltrifluoroacetamide (MSTFA). The GC was equipped with a programmable temperature vaporization (PTV) injector system (starting temperature of 50 ∘C, heating rate of 10 ∘C s-1 to 300 ∘C, isothermal holding time of 10 min, operated in splitless mode) and a fused silica capillary column (SGE BPX5, 50 m length, 0.22 mm inner diameter, 0.25 µm film thickness). For the measurements the GC oven was programmed with a starting temperature of 50 ∘C, a heating rate of 3 ∘C min-1 to 310 ∘C, and an isothermal holding time of 30 min. Helium with a constant flow rate of 1 mL min-1 was used as a carrier gas. For the n-fatty-acid fraction a different temperature program (starting temperature of 50 ∘C, 1 min isotherm, heating rate of 3 ∘C min-1 to 350 ∘C, isothermal holding time 25 min) was used. For compound identification, the gas chromatograph was linked to a mass spectrometer, which was operated in electron impact ionization mode at 70 eV. The temperature of the ion source was set to 230 ∘C. Full scan mass spectra were recorded from m/z 50 to 600 Da at a scan rate of 2.5 scans s-1. For the n-fatty-acid fraction, the scan rate was m/z 50 to 650 Da.

The quantification of n-alkanes, n-fatty acids, and β-amyrin was done in the GC-MS total ion current chromatogram by relating the peak area of the target compound to the peak area of an internal standard of known concentration. Other triterpenoids like olean-12-ene, olean-13(18)-ene, and hopene were quantified using the m/z 191 mass trace relative to the peak area of the β, β-diploptene (in the m/z 191 mass trace), the concentration of which was calculated in the total ion current chromatogram relative to the internal standard (5α-androstane).

The absolute lipid concentration is used as a rough estimator of organic-matter quality for degradation in the following way: the higher the concentration, the better the conservation of the lipid and the better the quality of the organic matter.

The CPI was introduced by Bray and Evans (1961) as the ratio of odd- to neighboring even-numbered alkanes; this is a measure of the alteration of organic matter. Here we use the improved formula from Marzi et al. (1993). In addition, we also applied the CPI for fatty acids in which even-numbered fatty acids predominate over adjacent odd n-fatty acids (Glombitza et al., 2009). CPI=∑i=nmC2i+1+∑i=n+1m+1C2i+12×∑i=n+1m+1C2i, where n is the starting dominating chain length/2, m is the ending dominating chain length/2, i is the index (carbon number), and C is concentration.

The CPI is used as a degradation and alteration proxy by quantifying the odd over even (n-alkanes, Fig. S2) or even over odd (n-fatty acids, Fig. S3) predominance of the carbon chains (Bray and Evans, 1961; Glombitza et al., 2009). A low CPI means mature or degraded organic matter (e.g., CPI of crude oil ∼ 1).

As introduced by Poynter (1989), the n-alkane ACL value is the concentration-weighted mean of different carbon chain lengths in a geological sample. For n-alkanes we use the C23–C33 interval, for n-fatty acids the C20–C34: ACL=∑i×Ci∑Ci, where i is the index (carbon number) and C is concentration.

The ACL is a rough OC source parameter. A schematic showing different chain lengths in different organisms is given in Fig. S4. The higher C3 land plants are expected to have an ACL of ∼ 28–29.

We use hop-17(21)-ene as another marker for low-maturity organic material. The hop-17(21)-ene is produced by bacteria. The assumption here is that during degradation and diagenesis the hop-17(21)-ene will be transformed into saturated hopane (Luo et al., 2012).

The ratio of the major even wax alcohols over the sum of major odd wax alkanes plus even alcohols was introduced by Poynter (1989) as the higher-plant alcohol (HPA) index. It is applied as an indicator for chemical degradation of the wax components. Based on this index, but using fatty acids instead of alcohols, we developed the HPFA index. The general assumption for this index is that it reflects the preservation degree of organic matter due to the higher lability of n-fatty acids in relation to n-alkanes. HPFA=∑n-fatty acidsC24,C26,C28∑n-fatty acidsC24,C26,C28+∑n-alkanesC27,C29,C31 The HPFA ratio cannot be considered an absolute index of degradation but is an indicator of the relative amounts of the more labile fatty acids that remain in a sample. Since n-alkanes are preserved preferentially compared to n-fatty acids, a decrease in this index indicates increased decomposition (the more degraded, the lower the HPFA index).

β-Amyrin (olean-12-en-3β-ol) is a triterpenoid produced by higher land plants. As a first degradation step, β-amyrin is expected to lose its hydroxy group and will be transformed to olean-12-ene. A second step would be a shift of the double-bond-forming olean-13(18)-ene. Thus, fresh organic material is associated with a lower oleanene ratio, while more degraded organic matter is reflected in a higher ratio. This index is calculated as follows: oleanene ratio(%)=olean-12-ene+olean-13(18)-eneβ-amyrin⋅100.

Pore water was obtained from each sample by centrifugation in specific pore water tubes. Water extracts were analyzed twice using ion chromatography with conductivity detection (ICS 3000, Dionex). An analytical column (AS 11 HC, 2 × 250 mm, Dionex) was used at a constant 35 ∘C. The sample was eluted with a KOH solution of varying concentration over time. The initial concentration was 1.4 mM. Between 0 and 6 min, the KOH solution was increased at a constant rate to 1.6 mM. Between 6 and 12 min the solution was increased to 10.0 mM KOH, and a concentration of 15.0 mM KOH was reached at 22 min. After 32 min, 60.0 mM KOH concentration was achieved, maintained for 1 min, and followed by a rapid decrease to 1.4 mM after 33 min, when samples were fixed for 45 min to equilibrate the system. For the quantification of acetate, standards containing the investigated compound were measured. The standard deviation of the sample and of standard quantification was < 5 %. Because acetate can act as excellent feedstock for microbes (Smith and Mah, 1980; Vieth et al., 2008) and it has been shown that acetate was rapidly consumed in the presence of oxygen and nitrate (Kuesel and Drake, 1995), we use the acetate pore water concentrations in the different deposits as a parameter to assess the quality of the organic matter and to compare the potential of the different deposits for future microbial degradation.

For testing the samples concerning their statistical distribution, the Shapiro–Wilk normality test was applied. Because of non-normal distribution, we used the Mann–Whitney–Wilcoxon test for significance testing of Yedoma and thermokarst samples. In order to compare the five different profiles, we used the Kruskal–Wallis rank sum test.

Multivariate statistical techniques, such as the PCA used here, allow the analysis of multiple variables in order to investigate connections between the different degradation proxies. Prior to the PCA, concentration data were transformed using a log (x+1) transformation. As the square root transformation is commonly applied to count data, especially if the values are mostly rather small, we decided to use this weaker (compared to logarithm) transformation for the TOC (wt% and kgm-3) data. Both transformations were applied to reduce right skewness and to put the parameters on the same scale. We performed three PCA runs. First, a PCA of the sediment parameters was implemented to infer differences between Yedoma and thermokarst deposits. Second, a PCA of biomarker proxies was performed. For this purpose, other characteristics were added as supplementary variables (TOCwt%, TOCkgm-3, C / N, δ13C, grain size, BD, ice content, and depth) without being included in the PCA calculation. These supplementary variables have no influence on the PCA and were plotted afterwards in the PCA biplot. Third, a PCA was conducted on samples of the major odd n-alkanes to infer possible changes in the source organisms with the same supplementary variables as described above. Computations were performed using the “vegan” package of R software (Oksanen, 2013).

The radiocarbon ages (Table 1, Fig. S5) of the Yedoma deposits range from infinite ages (> 55 000 yr BP) at the very bottom to 30 100 yr BP at the uppermost sampled Yedoma unit. This is comparable to other Yedoma sequences in the region (Schirrmeister et al., 2011b). The mean grain sizes show a decreasing trend in the Buo-04 lower Yedoma profile, from 28 µm at the bottom to 11 µm in the upper part of Buo-04-A. The Buo-02 Yedoma profile shows no trend but exhibits a more heterogeneous mean grain size, including three maxima at 22.5 m a.s.l. (32 µm), 23.7 m a.s.l. (34 µm), and 25.5 m a.s.l. (33 µm). Nevertheless, all Yedoma deposit samples are classified as poorly sorted medium-to-coarse silts with a stable low clay fraction (< 15 %).

The TOCwt% contents vary from 0.2 wt% at 5 m a.s.l to 24.0 wt% in a peaty paleocryosol horizon at 24 m a.s.l. (Fig. 3). The mean TOCwt% content is 2.4 wt% (median 0.97 wt%). Calculating the TOCkgm-3 according to Strauss et al. (2013), by utilizing the BD (between 0.1 and 1.5 103 kg m-3; 103 kg m-3 = g cm-3) and ice content (without ice wedges; 21 to 90 vol %), the Yedoma sediments contain from 3 to 46 kg C m-3 with a mean of 14 kg C m-3 (median 9 kg C m-3 ). The maxima correspond to the peaty horizons with large TOCwt% contents and a low BD. Within the paleocryosol horizons, located at 6.8, 24.0 to 24.5, 24.8, and 27.8 to 28.9 m a.s.l., maxima in the C / N ratio are observable. The C / N range in these horizons is 8 to 31. In the cryopedolith profile parts, the C / N maximum is reached in the lowermost Buo-04-C sub-profile (17.7 and 16.7). The C / N of the rest of the Yedoma profile falls between 4.1 (at Buo-02-C, 23.7 m a.s.l.) and 14.3 (below the paleocryosol at 23.5 m a.s.l.)

The δ13C of the Yedoma deposits ranges from -29.0 to -24.7 ‰. The minima fit the maxima of the C / N ratio well in the paleocryosol horizons at 6.8, 24.0 to 24.5, 24.8, and 27.8 to 28.9 m a.s.l. The minimum C / N of the Buo-02-C sub-profile corresponds approximately to the δ13C maximum (-25.0 to -24.7 ‰).

A series of long-chain n-alkanes that exhibit a strong odd-carbon preference ranging from n-C21 to n-C33 are identified in all Yedoma samples (Fig. S2). Moreover, the n-alkanes show a unimodal distribution maximizing at the C27, C29, or C31 n-alkane (Fig. S2). The n-fatty acids show strong even over odd carbon number predominance and a bimodal distribution ranging from C14 to C30 (Fig. S3). The maxima are generally located at n-C16 in the lower carbon number range and at n-C24 in the higher carbon number range. Total n-alkane and n-fatty-acid concentrations related to TOCwt% and sediment weight show a homogeneous pattern similar to that of the TOCwt% and C / N values. The n-alkane concentration ranges from 3 to 75 µg gSed-1 (mean 20 µg gSed-1) and from 387 to 1715 µg TOCwt%-1 (mean 1132 µg TOCwt%-1). The n-fatty acids range from 4 to 306 µg gSed-1 (mean 51 µg gSed-1) and from 475 to 4669 µg TOC-1 (mean 2196 µg TOCwt%-1).

This Yedoma series shows a distinct preference between even and odd carbon. The mean CPI values of the n-alkanes (12.2, ranging from 8.3 to 15.9) are higher than the CPI values of the n-fatty acids (4.9, ranging from 3.8 to 7.6). Because n-fatty acids are functional compounds (including a functional group, e.g., a carboxyl group), their degradation rates are much higher compared to those of n-alkanes (Poynter and Eglinton, 1990). This statement is also based on the assumption of similar sources. The ACL of the n-alkanes and n-fatty acids is very stable at around 28.4 (range: 27.6 to 29.2) and 25.0 (range: 23.8 to 25.6), respectively.

Relatively higher hop-17(21)-ene concentrations are used as an indicator for a lower organic-matter degradation state. In the lower Yedoma profile, the hop-17(21)-ene ranges from 0.0 µg gTOC-1 in the lowermost and uppermost samples (4.3 and 18.5 m a.s.l.) to the overall maximum of 19.4 µg gTOC-1 at the Buor-Khaya site (9.1 m a.s.l.). At Buo-02, the hop-17(21)-ene concentration is lower compared to the other Yedoma profile, with a mean of 1.9 µg gTOCwt%-1 and a maximum of 7.7 µg gTOCwt%-1 in the potentially Holocene-contaminated uppermost sample. The HPFA ratio for the Yedoma samples is very stable around the mean value of 0.50 (median 0.54), with a minimum in 18.5 m a.s.l. (0.15) and a maximum at the uppermost sample (0.69) at 29.7 m a.s.l. For Yedoma, the oleanene ratio is 0.0 (except for a ratio of 10.0 in the uppermost sample). The acetate content of the Yedoma sample is between 0.6 and 57.5 mg L-1, with a mean of 6.7 mg L-1 (median 1.2 mg L-1).

The radiocarbon dating shows Holocene ages between 8140 ± 50 and 3665 ± 35 yr BP (Fig. S6, Table 1). The lowermost Buo-05-C profile shows an age inversion for the two samples (0.3 and 2.2 m a.s.l.). The mean grain size at Buo-05 from the bottom to 6.7 m a.s.l. is 13 µm. Above, the mean grain size increases to 19 µm. The Buo-05 clay fraction is stable at a low level (< 15 %). The Buo-01 profile shows a very scattered grain size ranging from 4 to 44 µm mean grain size. For the whole data set, there is a maximum in the clay fraction (35 %) in the peat horizon at 8.7 m a.s.l. Buo-03 shows a slight decrease from 18 to 11 µm. All thermokarst deposits are classified as (very) poorly sorted silts. Similarly to the Yedoma deposits, the BD of the thermokarst deposits is between 0.1 and 1.5 × 103 kg m-3 and the ice content (without the ice wedges) is 23 to 87 wt% (Fig. 4).

The mean TOCwt% contents of the thermokarst deposits, 4.7 wt% (median 1.7 wt%), are higher compared to Yedoma deposits, varying between 0.2 and 43.0 wt%. Minimum and maximum TOCwt% both occur at Buo-01 and exhibit the same scatter as in the grain sizes. TOCkgm-3 ranges between 2.8 and 93.5 kg C m-3 (mean 24, median 19 kg C m-3 ).

At Buo-05 the C / N ratio is stable at around 9 to 10 (Fig. 4), except for a paleocryosol horizon at 9.2 m a.s.l. that shows a value of 22. At Buo-01, the C / N ratio below the paleocryosol horizon is remarkably low, between 2 and 9, followed by the overall maximum in the peaty horizon with a ratio of 34. The Buo-03 cryopedolith samples show C / N ratios around 10, while the paleocryosol samples exhibit C / N ratios from 16 to 19. The δ13C values range between -29.5 and -25.0 ‰, with minima corresponding to the C / N maxima at the paleocryosol horizons (anticorrelated to the C / N, Fig. 5a).

The absolute lipid concentration of n-alkanes are in the same range but slightly higher compared to the Yedoma profiles. The n-alkane average is 1275.7 µg gTOCwt%-1 (median 1260.1 µg gTOCwt%-1), ranging from 599.7 (8.7 m a.s.l.) to 1907.2 µg gTOCwt%-1 (29.5 m a.s.l.). The n-fatty-acid average is nearly double that found in the Yedoma samples. On average, 4096.1 µg gTOCwt%-1 (median 3805.7 µg gTOCwt%-1) are stored in the thermokarst deposits of Buor-Khaya, ranging from 554.5 (uppermost Buo-01 sample) to 11013.3 µg gTOC-1wt% (uppermost Buo-03 sample).

A series of long-chain n-alkanes were recognized in all thermokarst samples with a strong odd carbon number preference ranging from n-C21 to n-C33. Nearly all samples show a unimodal distribution of n-alkanes maximized at C27, C29, or C31 (Fig. S2). Sample Buo-03-A-03 alone does not fit into this scheme because it maximizes at n-C25. Compared to Yedoma samples, the short-chain fraction < n-C27 is more pronounced (Fig. S2). The n-fatty acids show a strong even-carbon-number preference and a bimodal distribution between n-C14 and n-C30 (Fig. S3), but the n-C16 is less pronounced than in the Yedoma deposits. An exception to this is found in sample Buo-01-A-02, where the C16 monomer reaches the overall maximum of the distribution. Apart from that, the maxima are generally located at the C24 n-fatty acid.

The n-alkane CPI of thermokarst averages 9.6 (median 9.3) and is lower compared to the Yedoma deposits, although the CPI values are in the same range (between 7.0 and 15.3). The CPI of the fatty acids ranges from 4.0 to 9.0 (mean 5.3, median 4.9). The ACL of n-alkanes and fatty acids reveal a homogeneous signal between 27.2 and 29.2 (mean 28.3) for n-alkanes and 23.6 to 25.6 (mean 24.8) for n-fatty acids.

Except for the maximum value of 16.1 µg gTOCwt%-1 at 8.7 m a.s.l., the hop-17(21)-ene concentration at Buo-05 varies between 0.1 and 4.9 µg gTOCwt%-1. Buo-01 paleocryosol values are 0.9 (8.7 m a.s.l.) and 8.4 in the lowermost sample (7.8 m a.s.l.). For Buo-03 the hop-17(21)-ene concentration ranges from 5 up to 8 µg gTOCwt%-1.

The HPFA ratio for the Buo-05 thermokarst samples is high, between 0.6 and 0.8; only the uppermost sample (9.3 m a.s.l.) shows a lower value of 0.2. The Buo-01 profile decreases from 0.7 in the lowest sample to 0.2 at the top. Buo-03 shows high parameter values of 0.8 and 0.9. The oleanene ratio for the thermokarst deposits ranges between 0 (Buo-01) and 13.8 (Buo-03). The overall mean oleanene ratio in thermokarst is 3.7 (median 2.2), which is remarkably higher compared to the Yedoma deposits.

The acetate content of the thermokarst samples is between 0.4 and 109.4 mg L-1, with a mean of 23.5 mg L-1 (median 2.8 mg L-1). Large acetate contents are found especially in the middle part of Buo-05, from 3.4 m a.s.l. (74.1 mg L-1) to 6.1 m a.s.l. (109.4 mg L-1), and in the uppermost Buo-03 sample (35.3 mg L-1).

Except for the Yedoma CPI, the Yedoma HPFA and the thermokarst hop-17(21)-ene Shapiro–Wilk normality tests reveal a non-normal distribution. Based on this, we chose non-parametric significance testing. This reveals significant differences for TOC, C / N, δ13C, and HPFA on the stratigraphical level (Yedoma vs. thermokarst, Mann–Whitney–Wilcoxon test, Table S1 in the Supplement). On the profile level, we found significant differences for C / N, δ13C, and HPFA by applying the Kruskal–Wallis test (Table S1).

The first PCA diagram (Fig. 6a) shows that thermokarst sediments, especially at Buo-05, could not be separated from Yedoma deposits. This diagram, including the first two principal components, explains 79 % (pc1 57 %, pc2 22 %) of the total data set variance. The second PCA diagram (Fig. 6b) illustrates that biomarker quality estimators in Yedoma samples have slightly lower variability because they cluster in an area at pc1 and pc2 > 0, while the thermokarst samples do not cluster. In this diagram 53 % of the data set variance is explained. Moreover, this PCA shows that there is good consistency between the CPIalkane quality estimator and the C / N ratio (Fig. 6b). The PCA of the n-alkane chain length (Fig. 6c) shows that the best separating variables for thermokarst are the shorter-chain n-alkanes (C17, C19, and C21), contrary to C29 for the majority of the Yedoma samples. The pc1 explains 39 and pc2 explains 29 % (total 68 %) of the data set variance.