Mapping the degree of decomposition and thaw remobilization potential of soil organic matter in discontinuous permafrost terrain

Authors


Corresponding author: G. Hugelius, Department of Physical Geography and Quaternary Geology, Stockholm University, SE-10691 Stockholm, Sweden. (gustaf.hugelius@natgeo.su.se)

Abstract

[1] Soil organic matter (SOM) stored in permafrost terrain is a key component in the global carbon cycle, but its composition and lability are largely unknown. We characterize and assess the degree of decomposition of SOM at nine sites representing major land-cover and soil types (including peat deposits) in an area of discontinuous permafrost in the European Russian Arctic. We analyze the elemental and stable isotopic composition of bulk SOM, and the degree of humification and elemental composition of humic acids (HA). The degree of decomposition is low in the O-horizons of mineral soils and peat deposits. In the permafrost free non-peatland soils there is enrichment of13C and 15N, and decrease in bulk C/N ratios indicating more decomposed material with depth. Spectral characterization of HA indicates low humification in O-horizons and peat deposits, but increase in humification in the deeper soil horizons of non-peatland soils, and in mineral horizons underlying peat deposits. GIS based maps indicate that less decomposed OM characteristic of the O-horizon and permafrost peat deposits constitute the bulk of landscape SOM (>70% of landscape soil C). We conclude, however, that permafrost has not been the key environmental factor controlling the current degree of decomposition of SOM in this landscape due to relatively recent permafrost aggradation. In this century, active layer deepening will mainly affect SOM with a relatively high degree of decomposition in deeper mineral soil horizons. Additionally, thawing permafrost in peat plateaus may cause rapid remobilization of less decomposed SOM through thermokarst expansion.

1. Introduction

[2] Soils of high latitude terrestrial ecosystems are key components in the global carbon (C) cycle [McGuire et al., 2009]. Large stocks of soil organic matter (SOM) have accumulated in permafrost soils and peatlands, where low temperature and anoxia due to water logging reduce decomposition rates [Davidson and Janssens, 2006]. Tarnocai et al. [2009] estimated the soil organic C (SOC) stocks in the northern permafrost region to be 1024 Pg (Pg = g × 1015) for the upper three meters, with an additional 241 Pg stored in deep deltaic deposits, and 407 Pg in deep Pleistocene loess deposits (Yedoma). Permafrost soils and peatlands have been identified as key global C pools that are vulnerable to remobilization through permafrost thaw, and changes in surface hydrological conditions [Gruber et al., 2004]. Recent warming and thawing of permafrost reported from around the circum-Arctic region [Romanovsky et al., 2010a, 2010b; Smith et al., 2010] may lead to increased remobilization of SOM in permafrost soils. SOM currently stored in permafrost soils is believed to have a relatively low decomposition rate, and the magnitude of C flux resulting from global warming and permafrost thawing depends on the (1) rate and extent of remobilization processes and (2) size and lability of the SOM pools that become available for decomposition — both factors remain largely unknown [Schuur et al., 2008]. In recent years, detailed local scale studies have investigated the lability of SOM in permafrost soils [e.g., Weintraub and Schimel, 2003; Michaelson and Ping, 2003; Dutta et al., 2006; Kaiser et al., 2007; Xu et al., 2009], leading to an increased understanding of SOM dynamics in periglacial environments. These studies are however limited by their spatial extent and temporal resolution, and they often focus on a narrow selection of soil types or only surface/active layer soil horizons.

[3] The quality of SOM in periglacial environments is expected to vary according to its genesis and alteration following the storage history and age. Some key environmental factors that affect the mean residence time of SOM include protection within soil aggregates, flooding (anoxia) and sub-zero temperatures [Davidson and Janssens, 2006]. These changes can be investigated by tracing the bulk characteristics of SOM. For example, changes in elemental ratios (C/N and H/C) and isotopic composition are useful indicators of decomposition in SOM [Zaccone et al., 2007; Diochon and Kellman, 2008; Xu et al., 2009; Ranjan et al., 2010]. Studies of SOM decay in peat indicate that C/N ratios decrease with decomposition as a result of anaerobic decay [Kuhry and Vitt, 1996; McKane et al., 1997]. Likewise, the bulk H/C ratio decreases as the molecular complexity increases during decomposition in peat [Zaccone et al., 2007; Šire et al., 2008]. In aerated (mineral) tundra soils, release of CO2 during decomposition lowers the C/N ratios [Ping et al., 1998]. Kinetic fractionation causes SOM to become enriched in 13C and 15N with increase in depth, age and microbial decomposition [Diochon and Kellman, 2008; Xu et al., 2009].

[4] Previous studies indicate that as complex natural organic polymers decompose, they lose the more labile components in a process known as humification [Stevenson, 1982]. The resulting highly complex recalcitrant products are categorized into three different groups based on their solubility in water: fulvic acid (FA, soluble at all pH), humic acid (HA, soluble at pH < 2) and humin (insoluble at all pH). The spectral characteristics of HAs have been used to describe humification indices in soils (including peat) [Kumada, 1987; Ikeya and Watanabe, 2003; Šire et al., 2008; Zaccone et al., 2007, 2011]. The humification analogues reflect molecular stability and the proportion of non- or less-colored functional groups, which have been combined in a classification system to describe the degree of humification in soil horizons [Kumada, 1987; Ikeya and Watanabe, 2003]. The elemental composition of HAs also changes after humification. The H/C ratio in HAs decreases with increased humification, whereas the C/N ratio increases during the later stages of decomposition [Kuwatsuka et al., 1978; Kumada, 1987].

[5] Because soil warming and/or permafrost thaw can lead to elevated rates of C respiration from high latitude soils [Dorrepaal et al., 2009; Schuur et al., 2009], an improved knowledge regarding the lability of SOM pools is crucial for predicting the effect of soil warming and permafrost thawing in periglacial landscapes. In this study, we characterize the SOM stored in different soil horizons (including the active layer and permafrost) of mineral and organic soils from nine sites representative of the different land-cover and soil types in an area of discontinuous permafrost in the European Russian Arctic. This area is currently experiencing permafrost warming and thaw [Oberman, 2008], and hence it is of great interest for assessment of climate change impacts. To assess the degree of SOM decomposition in this region, we have performed relatively simple and inexpensive chemical analyses in the bulk material involving elemental composition, stable isotopes and humification characteristics. The overall emphasis is on utilizing rapid and low-cost geochemical methods as a reconnaissance tool for widespread mapping of SOM pools with different developmental histories. We have combined the SOM characteristics from this periglacial landscape with GIS based high-resolution maps of SOC storage to quantify the SOM pools with low degree of humification, and hence, considered as potentially labile. To exemplify the potential use of such landscape scale classifications, spatial analyses were used to investigate the remobilization potential of different SOM pools under varying scenarios of permafrost thaw, and comparison of active layer deepening to thermokarst expansion.

2. Study Area and Methods

2.1. Study Area

[6] The Seida study site is located in the lowlands of the Usa River Basin, west of the Ural Mountains in the forest-tundra ecoclimatic zone [Kozubov et al., 1999], northeastern European Russia (Figure 1). The area is in the catchment of the small Sedbyaga River, which drains into the Usa River. The Usa River flows in a southwesterly direction and constitutes the largest tributary of the Pechora River. The Usa Basin lowlands are underlain by thick Quaternary deposits that are intersected by river valleys. Seida is underlain by undulating deposits of glaciofluvial sandy loams of uncertain age [Oberman and Mazhitova, 2003], and was last glaciated during the Late Saalian (160–140 kyrs) under the Barents-Kara ice sheet [Svendsen et al., 2004].

Figure 1.

Map showing the location of the Seida study area, generalized vegetation cover [from Hugelius et al., 2011] and sampling sites. The extent of the uplifted permafrost peat plateau is outlined in black. Map projection UTM 40N, WGS84.

[7] At the closest climate station, Vorkuta, 75 km northeast of Seida, the mean annual air temperature in the period 1961–90 was −6.1°C (mean January temperature: −21.2°C, mean July temperature +13.0°C) and the mean annual precipitation was 538 mm (ca. 40% of this falls as snow from November to April). The landscape is underlain by discontinuous permafrost (70–90% permafrost coverage) or island permafrost (20–50% permafrost coverage), with mean annual ground temperatures (MAGT) at the level of zero amplitude between −0.5°C and −2.1°C [Rivkin et al., 2008]. The Seida peat plateau is underlain by continuous permafrost (except in the interspersed thermokarst lakes and collapse fens). Permafrost free areas are generally found along river/stream valleys, and on a bedrock ridge in the eastern part of the study area.

[8] Spruce (Picea obovata) is the dominant tree line species, but downy birch (Betula pubescens) is also common. Only scattered stands of trees are present in the southeastern corner of the study area (Figure 1). Open tundra vegetation on mineral soils is either shrub tundra dominated by prostrate dwarf shrubs (e.g., Vaccinium spp., Empetrum hermaphroditum), or dwarfbirch tundra dominated by Betula nana, with lichen or moss. Dense willow (Salix spp.) stands are common along streams and in paludified lowland areas.

[9] Peat plateau / thermokarst complexes consisting of a mosaic of permafrost bogs, fens and thermokarst lakes are a dominant landscape component. The plateaus rise above the surrounding landscape due to high ground ice content. The thickness of peat is highly variable, but may reach depths of up to >4 m. Outside the peat plateau complexes, fens are also found along the margins of peat plateaus (marginal fens, typically >1 m peat), and in narrow patches surrounded by mineral tundra soils (tundra fens, typically <1 m peat). Fen vegetation varies depending on soil type, water level and nutrient composition. Graminoids (e.g., Carex spp. and Eriophorum spp.) and mosses (including Sphagnum) are common. The raised permafrost (palsa and peat plateau) bogs have a dry surface dominated by prostate dwarf shrubs, mosses (including Sphagnum) and lichens.

2.2. Sampling

[10] The nine sites used for this study were selected to be representative of major land cover and soil types, but with a focus on peatlands because they hold the majority (>70%) of the landscape soil C pool in the region [Hugelius and Kuhry, 2009]. Although forests are a minor landscape component in the study area, a mineral forest soil was included to allow comparison with the mineral tundra soils. For a more detailed description of the larger sampling program from which these nine sites were selected, see Hugelius et al. [2011]. Active layer horizons and permafrost-free upland soils were sampled from open pits with fixed volume samplers. Peat plateaus were sampled near thermally eroding edges on the banks of thermokarst lakes. The erosion edge was cleared to expose fresh material prior to sampling. Permafrost soils (including some peat plateau sites) were cored using steel pipes that were hammered into the ground in 5–10 cm increments, retrieving intact frozen cores in between hammering. Permafrost free fens were sampled using a fixed volume peat corer. General descriptions of the soil horizons and peat stratigraphy are based on field-observations. All soils were classified according theWorld Reference Base of Soil Terminology (WRB) [IUSS Working Group WRB, 2007].

2.3. Geochemical Analyses

[11] Radiocarbon (14C) analysis of bulk SOM samples (after removing the living roots under a microscope) was performed on 17 soil (including peat) samples at the Poznań Radiocarbon Laboratory, Poland. Radiocarbon age was calibrated to calendar years BP (here defined as the sampling year 2008) using OxCal 4 [Bronk Ramsey, 2001]; calibrated ages are the median of the highest probability interval expressed as calendar years BP.

[12] The analyses of other geochemical variables were made at 5 to 10 cm depth intervals (depending on the soil horizon thickness or gross stratigraphy). Individual soil samples were analyzed for dry bulk density (DBD, g/cm3) after drying in oven at 95°C (for 24 h). Loss-on-ignition (LOI, weight %) at 550°C (for 6 h) was used to estimate organic content (data not shown), prior to heating at 950°C (for 2 h) to determine the carbonate content [Dean, 1974; Heiri et al., 2001].

[13] For elemental content and stable C and N isotope analyses, samples were freeze-dried, and kept in a desiccator with 12 M HCl (48 h) to remove carbonates [Hedges and Stern, 1984]. Total C, N and H contents were measured using a Carlo Erba 2500 elemental analyzer. The uncertainty in analyses of duplicates samples was <±10%. The elemental ratios ratios used in this paper are based on weight. The isotopic composition was analyzed using a continuous flow system consisting of an elemental analyzer coupled to a Finnigan MAT Delta Plus mass spectrometer. Data are reported in the conventional delta (δ) notation versus Vienna PeeDee Belmenite (V-PDB) for C and atmospheric N2 for total N. The precision for C and N isotope analyses were ±0.18‰ and ±0.06‰, respectively. The analytical error in duplicate samples was <±0.1‰.

[14] Fulvic acids (FA) and humic acids (HA) were extracted using the IHSS method [Swift, 1996]. In short, this involved treating 1 g of freeze-dried sample with 0.1 M NaOH under N2at 20°C (at 1:20 and 1:4 soil to NaOH ratios for organic and mineral soils, respectively). The supernatant was removed and acidified with 6 M HCl (pH < 1). The precipitate was re-dissolved in 0.1 M NaOH and was quantified using an UV spectrophotometer (Hitachi 1101). We used the variables: ΔlogK = (log(A400/A600)), E4/E6 = (A465/A665) and A600/C = A600/TOC, where A400, A465, A600 and A665were the absorbance for HA at 400, 465, 600 and 665 nm in 0.1 M NaOH, and TOC is the total organic C concentration of HA. Total organic carbon in the FA and HA extracts was measured with a Shimadzu TOC 5000 analyzer using the TOC-IC mode; the analytical error in duplicate samples was <±5%. Recovery of FA and HA was calculated as the proportion of C recovered from the original homogenized bulk sample used during extraction.

2.4. Peat Age-Depth Models, Linear Correlation and PCA

[15] For peat sections with several 14C dates available, age-depth models describing the relationship between bulk SOM age and depth in the profile were made based on linear regressions between dated samples. For fens,14C dates were not available for the top of peat deposits, but since living fen vegetation occurs on the surface of all sites, it implies active peat formation. We assumed the age of the surface as 0 cal yrs BP in these sites. For the peat plateau sites SE7, SE8 and SE9, 14C dates for surface samples were lacking, and a 14C sample from a nearby rootlet peat (peat dominated by vascular plant remains with a large fraction of dark roots) sample at 5–10 cm depth was used. The results of these age-depth models are used in correlation analyses to relate the SOM age to decomposition, and to calculate the net accumulation of peat plateau SOM, and its residence time in permafrost-free environments. From14C dating, macrofossil analyses and investigating the gross-stratigraphy of different sections J. Routh et al. (Soil organic matter characteristics in permafrost terrain, European Russian Arctic: Lability, storage, and impact of thawing, manuscript in preparation, 2012) infer that in the Seida peat plateau epigenetic permafrost first aggraded sometime after 2200 cal yrs BP in SE8 and sometime before 820 cal yrs BP in SE9. There is also evidence for local permafrost collapse (thermokarst) at ca. 800 cal yrs BP and permafrost re-aggradation in the Little Ice Age. In the nearby Rogovaya peat plateau,Oksanen et al. [2001] describe: a first short period of localized permafrost aggradation at ca. 3200 cal yrs cal BP (followed by rapid collapse), extensive permafrost aggraded at ca. 2200 cal yrs BP, and a final period of peat plateau development occurred during the Little Ice Age. Becher [2011]describes intercalated thermokarst deposits in the Rogovaya peat plateau deposits, indicating collapse stages prior to the latest period of aggradation during the Little Ice Age. At the nearby Lek-Vorkuta peat plateau,Andersson et al. [2011] dated permafrost aggradation to ca. 2400 cal yrs BP. In the current study, we use these dates to discuss the relative importance of past permafrost conditions in the role of SOM protection.

[16] In order to describe linear relationships between different variables, Pearson's product-moment correlation analyses were performed using the open source software R (R Development Core Team [2007], library: Hmisc). Adjusted p-values were calculated usingHolm's [1979]correction. To enable comparison of trends between different cores all observations (except SOM age) were standardized to zero mean and unit variance within the separate cores before pooling into a single large data set. Correlations were calculated for (1) all samples, (2) peat/O-horizon samples, and (3) mineral soil samples. The data set for bulk H/C ratio is incomplete for mineral-soil horizons (data available for 14 out of 33 samples because H concentration was below the detection limit). Therefore, the H/C ratio is only included in the regression for organic soil horizons. In samples where data were missing (mainly C/N and H/C of HA when elemental concentrations were below the detection limit), the missing data points were replaced with the mean for that core. The age of the SOM is included for the correlation matrix on peat/O-horizon samples (based on age-depth models described above). Because SOM age is an absolute variable that is independent of botanical origin it has been correlated without normalization, and therefore reflects trends in the whole data set without bias.

[17] Principal Component Analysis (PCA) (software CANOCO 4.5 [Ter Braak and Smilauer, 2002]) was performed on the geochemical variables, which were considered as proxies for degradation (variables were standardized to zero mean and unit variance within separate cores). A data set of supplementary variables (not affecting the ordination) were included in the analyses; these were TOC%, depth (cm) and permafrost/thaw (the last parameter is a nominal variable). The gradient length of the data set was calculated in detrended correspondence analysis to 1.07, confirming that the data showed linear responses suitable for PCA analyses [Jongman et al., 1995].

2.5. Classification of HAs, SOM Humification Maps and Quantification of SOC Remobilization Potential

[18] The spectral characteristics of HAs have previously been used in other studies as humification indices in mineral and organic soil (including peat) [Kumada, 1987; Ikeya and Watanabe, 2003; Šire et al., 2008]. The variable Δlog K and the E4/E6 ratio decrease with the development of conjugation systems in HA (conjugation favors molecular stability), while A600/C increases by the proportion of non- or less-colored moieties (showing less presence of colored functional groups in HA) [Ikeya and Watanabe, 2003]. Because the variables A600/C and Δlog K reflect different aspects of humification, they were combined in a classification system describing the HA types Rp, B and A. The relative degree of humification in these HA types increases following the order Rp < B < A [Kumada, 1987; Ikeya and Watanabe, 2003]. The Rp-type HA fraction was further sub-divided into Rp1 and Rp2 HAs. For classifying HAs into the most labile sub-group (Rp1), the HAs needed a Δlog K value >1 and A600/C ratio < 1 [Ikeya and Watanabe, 2003]. Kumada's [1987]classification system also includes a P-type HA containing a green pigment (Pg), which was argued to interfere with the ΔlogK variable. This pigment probably originates from the sclerotia of the organism Cenococcum graniforme, which has a circum-Arctic distribution. The organism has been found in mycorrhizal association with trees, shrubs and herbs and has also been reported in peat [Kumada, 1987; Oksanen et al., 2001]. An empirical mathematical correction (Δlog K = 0.135 * E4/E6) makes the variables Δlog K and E4/E6 interchangeable [Ikeya and Watanabe, 2003]. Because determination of E4/E6 does not appear to be sensitive to this pigment, Figures S1–S2 show both Δlog K and the mathematically corrected E4/E6 ratio for the sites where P type HAs were found (all sites except SE7–9).

[19] Maps were generated from a high-resolution land cover map using ArcGIS 9.3. The maps show total SOC storage in kg C m−2 following Hugelius et al. [2011] (calculated for a depth of 1 m in mineral soils and full depth of peat deposits) as well as FA/HA recovery and HA types (Kumada's [1987] method), estimated from the nine land cover types. Hugelius et al. [2011]also simulated SOC remobilization from active layer deepening (based on simulations with the GIPL2 permafrost model) and thermokarst expansion (based on GIS-modeling of thermokarst expansion) in the Seida area. These simulations were re-applied with SOC partitioned following the HA classification of bulk SOM derived from this study. For these calculations the upper 3 m of soil are included for both mineral and organic soils. In the active layer deepening scenario, we calculated the percentage of total landscape SOC (calculated to a reference depth of 3 m) that will be stored in the active layer of different soil types for progressive times-slices extending to the end of this century. These fractions are partitioned following the type of HA that is most prevalent in bulk SOM. For the thermokarst expansion scenario, we calculated the percentage of total landscape SOC (calculated to a reference depth of 3 m) that will be affected if current thermokarst lakes would expand laterally.

3. Results

[20] The auxiliary material contains a summary of the results from 14C dating including sample properties, un-calibrated and calibrated14C ages (Table S1 in Text S1). The results from the analyses of individual sites are shown in Figures S1–S3 in the auxiliary material which also includes a detailed description of the observed trends at each site. The overall trends in the upland mineral soils, fens and permafrost plateaus are summarized in section 3.1 below.

3.1. Trends in Geochemical Degradation/Humification Proxies

3.1.1. Upland Mineral Soils (SE1–3)

[21] The upland mineral soil sites include: SE1 with shrub tundra vegetation developed on a Folic Stagnic Cambisol (Gelic), SE2 with Dwarf birch tundra heath developed on a Gelistagnic Cambisol and SE3 with Spruce forest growing on a Haplic Podsol. For these permafrost-free, upland mineral soils, a steady decrease in bulk C/N ratios reflects increased degradation of SOM with depth (Figure S1). This is also supported by the stable isotope composition at all three sites. Both 13C and 15N show enrichment from the O- and A-horizons into B/C horizons. The analogs for humification (A600/C, Δlog K and E4/E6) do not display clear down-core trends. In the upper 20–30 cm of soils, the ΔlogK and E4/E6ratio suggest increased molecular stability with depth, but in the B/C-horizons this trend is reversed. In the shrub tundra (SE1) and spruce forest (SE3) sites, A600/C follows the same pattern as above. In contrast, the birch tundra site (SE2) shows a very low degree of coloration in HAs throughout the core. High rates of humification occur in SE3, which can be expected in this well-drained forest soil. There are no clear trends in the C/N ratio of HAs. The H/C ratios in HAs increase with depth down to ≈50 cm. The FA/HA recovery increases significantly below O-horizon, signifying that humic substances constitute a major component of SOM stored in the mineral horizons of these soils.

3.1.2. Fen Sites (SE4–6)

[22] The fen sites include (1) SE4 a shallow tundra fen classified as a Cryic Fibric Histosol, (2) SE5 a deep graminoid dominated fen that developed on the margins of a peat plateau and is classified as a Fibric Histosol, and (3) SE6 a fen that developed within the uplifted peat plateau classified as a Fibric Histosol (Gelic). In all these sites, Sphagnum dominated peat layers at the surface have high bulk C/N ratio, and there is little or no evidence of decomposition with depth in the underlying fen peat deposits (Figure S2). Likewise, the stable isotope compositions display little indication of degradation. 15N appears to be slightly enriched with depth in the shallow tundra fen (SE4). In the plateau margin (SE5) and peat plateau (SE6) fens, there is high variability without any clear trends in the different geochemical proxies. This variability is likely connected to botanical shifts in peat composition rather than decomposition [Diochon and Kellman, 2008]. δ13C is relatively stable between −30‰ and −25‰ in all three cores. In SE4, Δlog K and the E4/E6ratio decrease in the buried permafrost layers implying a higher degree of humification in the older strata, arising as a result of pre-existing mineral soil that was buried due to paludification. In SE5, ΔlogK and the E4/E6 ratio indicate a gradual increase in molecular complexity with depth in the peat. In contrast these proxies have very low values and vary synchronously in SE6. The low Δlog K value in SE6 suggests possible interference by the Pg pigment [Kumada, 1987]. The result is consistent with a low E4/E6 ratio and high FA/HA recovery, which suggests that the observed trend is not an artifact, but rather indicative of humification process. The A600/C shows that coloration in HAs is low and relatively stable down-core in all fen peat deposits. The only exception to this trend is the upperSphagnum dominated peat layer in SE5 where the humification proxies and recovery of FA/HA indicate more humification than in the older peat layers below (surface peat in SE6 also has high FA/HA recovery).

[23] The C/N and H/C ratios in HAs are relatively constant indicating little change in the composition of HAs with depth. The recovery of FA/HA is generally low (<5%) in peats. However, marked increases in FA/HA recovery at the peat/mineral interface show that humic substances constitute a major component of the organic matter stored in mineral horizons underlying the fen peat deposits.

3.1.3. Permafrost Plateau Sites (SE7–9)

[24] The permafrost plateau sites include (1) SE7 a site with shrub tundra vegetation growing over shallow peat in the uplifted peat plateau that developed on a Folic Fibric Cryosol and (2) sites SE8 and SE9 which are bog peat plateaus with deep peat deposits (3.5–4 m), classified as Folic Cryic Histosols. Bulk elemental ratios and stable isotopes suggest a lack of SOM degradation with depth in the permafrost peat deposits (Figure S3). Bulk C/N ratios are elevated in the Sphagnum dominated peat layers, but remain stable (around 20) with depth in rootlet peats (all sites), and in the thick underlying fen peat deposits at SE8 and SE9. The bulk H/C ratio remains constant around 0.15 throughout the peat deposits, but tends to increase in mineral soil horizons below the peat deposits. Besides some limited enrichment in the surface peat deposits in SE8 and SE9, δ13C and δ15N remain constant in the peat deposits. In mineral soil-horizons underlying the peat deposits, enrichment of15N indicates more degraded SOM in the mineral sub-soil at all three sites. In SE8, A600/C and Δlog K indicate a gradual increase in humification below 200 cm, but there is absence of increased humification with depth in the peat deposits in SE7 or SE9. Δlog K indicates increased humification at the peat/mineral interface at all sites. In the peat plateau tundra site (SE7), A600/C indicates less coloration of HAs in the permafrost than the overlying mineral and organic soil horizons. In SE8 and SE9, low recovery of FA/HA with only slightly higher values in the surface peat and in the mineral sub-soil, indicates that humic compounds constitute a minor component of the peat deposits. In SE7, however, the recovery is consistently above 20%.

3.2. Correlation Matrices and PCA

[25] Correlation matrices summarize the linear relationships between the standardized variables TOC%, DBD, bulk C/N, δ 15N, δ13C, A600/C, Δlog K, E4/E6, HA C/N, HA H/C, FA/HA% recovery and depth in all samples, organic soil horizon samples (also includes the variables bulk H/C and SOM age) and mineral soil samples (Tables 1 to 3, respectively).

Table 1. Matrix Showing Pearson Moment-Product Correlation Coefficient (R) for SOM Characteristics of All Soil Horizons (n = 99)a
 TOC %DBD (g/cm3)C/N (Bulk)δ 15Nδ 13CA600/CΔ log KE4/E6C/N (HA)H/C (HA)FA/HA % rec.Depth (cm)
  • a

    All observations are standardized to zero mean and unit variance within the separate cores before pooling into a single large data set. Asterisks denote different levels of statistical significance: *, p < 0.05; **, p < 0.01; ***, p < 0.001. Adjusted p-values were also calculated usingHolm's [1979] method and correlations that were near significant (p < 0.10) are in italic and correlations that were significant (p < 0.05) are in bold.

TOC %1−0.79***0.35***0.34***−0.22*−0.49***0.47***0.43***−0.38***−0.06−0.59***−0.58***
DBD (g/cm3)−0.79***1−0.44***0.27**0.21*0.37***−0.33*−0.26**′0.23*0.010.62***0.5***
C/N (Bulk)0.35***−0.44***1−0.190.13−0.090.3**0.26**−0.3*−0.01−0.130.53***
δ 15N0.34***0.27**−0.1910.39***0.17−0.36***−0.27**−0.110.160.29**0.01
δ 13C−0.22*0.21*0.130.39***1−0.09−0.21*−0.2*−0.150.32**0.4***−0.05
A600/C0.49***0.37***−0.090.17−0.091−0.56***−0.46***0.05−0.070.34**0.19
Δ log K0.47***−0.33*0.3*−0.36**−0.21*−0.56***10.95***−0.09−0.24*−0.31**−0.37***
E4/E60.43***−0.26**′0.26**−0.27**−0.2*−0.46***0.95***1−0.11−0.24*−0.29**−0.37***
C/N (HA)−0.38***0.23*−0.3−0.11−0.150.05−0.09−0.111−0.2*0.020.58***
H/C (HA)−0.060.01−0.010.160.32**−0.07−0.24*−0.24*−0.2*10.180.06
FA/HA % rec.−0.59***0.62***−0.130.29**0.4***0.34**−0.31**−0.29**0.020.1810.16
Depth (cm)−0.58***0.5***−0.53***0.01−0.050.19−0.37***−0.37***0.58***0.060.161
Table 2. Matrix Showing Pearson Moment-Product Correlation Coefficient (R) for SOM Characteristics of Peat/O-Horizons (n = 66)a
 TOC %DBD (g/cm3)C/N (Bulk)H/C (Bulk)δ 15Nδ 13CA600/CΔ log KE4/E6C/N (HA)H/C (HA)FA/HA % Rec.Depth (cm)SOM Age
  • a

    Observations are standardized to zero mean and unit variance within the separate cores before pooling into a single large data set; the only exception from this is the variable SOM age which is related to unmodified variables from the whole data set. Asterisks denote different levels of statistical significance: *, p < 0.05; **, p < 0.01; ***, p < 0.001. Adjusted p-values were also calculated usingHolm's [1979] method and correlations that were near significant (p < 0.10) are in italic and correlations that were significant (p < 0.05) are in bold.

TOC %1−0.48***0.14−0.22*−0.1−0.06−0.230.150.09−0.34−0.03−0.14−0.34**−0.15
DBD (g/cm3)−0.48***1−0.36**−0.080.10.010.030.070.170.12−0.210.090.140.04
C/N (Bulk)0.14−0.36**10.06−0.060.39**−0.050.180.1−0.26*0.130.3*−0.47***−0.44***
H/C (Bulk)−0.22−0.080.0610.36**0.31*−0.220.070.09−0.220.26*0.36**−0.18−0.31*
δ 15N−0.10.1−0.060.36**10.39**−0.04−0.19−0.03−0.44***0.210.37**−0.28*−0.43***
δ 13C−0.060.010.39**0.31*0.39**1−0.130.020.02−0.43***0.26*0.6***−0.52***−0.58***
A600/C−0.230.03−0.05−0.22−0.04−0.131−0.6***−0.46***0.25−0.110.170.060.08
Δ log K0.150.070.180.07−0.190.02−0.6***10.93***−0.11−0.19−0.22−0.12−0.07
E4/E60.090.170.10.09−0.030.02−0.46***0.93***1−0.14−0.18−0.18−0.16−0.14
C/N (HA)−0.34**0.12−0.26*−0.22−0.44***−0.43***0.25−0.11−0.141−0.19−0.35**0.7***0.70***
H/C (HA)−0.03−0.210.130.260.21*0.26*−0.11−0.19−0.18−0.1910.22−0.12−0.27*
FA/HA % rec.−0.140.090.3*0.36**0.37***0.6***0.17−0.22−0.18−0.35**0.221−0.42***−0.56***
Depth (cm)−0.34**0.14−0.47***−0.18−0.28*−0.52***0.06−0.12−0.160.7***−0.12−0.42***10.90***
SOM age−0.150.04−0.44***−0.31*−0.43***−0.58***0.08−0.07−0.140.70***−0.27*−0.56***0.90***1
Table 3. Matrix Showing Pearson Moment-Product Correlation Coefficient (R) for SOM Characteristics of Mineral Soil Horizons (n = 33)a
 TOC %DBD (g/cm3)C/N (Bulk)δ 15Nδ 13CA600/CΔ log KE4/E6C/N (HA)H/C (HA)FA/HA % Rec.Depth (cm)
  • a

    All observations are standardized to zero mean and unit variance within the separate cores before pooling into a single large data set. Asterisks denote different levels of statistical significance: *, p < 0.05; **, p < 0.01; ***, p < 0.001. Adjusted p-values were also calculated usingHolm's [1979] method and correlations that were near significant (p < 0.10) are in italic and correlations that were significant (p < 0.05) are in bold.

TOC %1−0.79***0.11−0.2−0.08−0.54**0.41*0.33−0.290.14−0.63***−0.58***
DBD (g/cm3)−0.79***1−0.270.040.170.37*−0.28−0.170.120.010.76***0.6***
C/N (Bulk)0.11−0.271−0.05−0.280.20.180.16−0.29−0.15−0.49**−0.39*
δ 15N−0.20.04−0.0510.220.2−0.35*−0.310.16−0.04−0.130.06
δ 13C−0.080.17−0.280.221−0.26−0.38*−0.37*0.180.39*−0.060.67***
A600/C−0.54**0.37*0.20.2−0.261−0.45**−0.38*−0.24−0.120.280.08
Δ log K0.41*−0.280.18−0.35*−0.38*−0.45**10.97***0.15−0.21−0.04−0.41*
E4/E60.33−0.170.16−0.31−0.37*−0.38*0.97***10.14−0.23−0.01−0.38*
C/N (HA)−0.290.12−0.290.160.18−0.240.150.141−0.290.150.31
H/C (HA)0.140.01−0.15−0.040.39*−0.12−0.21−0.23−0.2910.030.24
FA/HA % rec.−0.63***0.76***−0.49**−0.13−0.060.28−0.04−0.010.150.0310.44*
Depth (cm)−0.58***0.6***−0.390.060.67***0.08−0.41*−0.38*0.310.240.44*1

[26] The correlation matrix including all samples shows that TOC% content correlates positively with Δlog K and E4/E6, whereas TOC correlates negatively with DBD, A600/C and FA/HA recovery (Table 1). The bulk C/N correlates negatively with depth and DBD. The stable isotopes 15N and 13C correlate weakly with each other. δ13C correlates positively with FA/HA recovery. The different analogues for humification (A600/C, Δlog K and E4/E6) correlate well with each other. These proxies however correlate weakly with the FA/HA recovery, and indicate negative correlation between depth and Δlog K and E4/E6. The C/N in HAs increases with high bulk TOC (and low DBD). In contrast, the H/C ratio in HAs shows an absence of significant trends.

[27] A correlation matrix for organic soil horizon samples (also includes SOM age from age-depth models) shows no relationship between TOC and humification, FA/HA recovery or depth (Table 2). The bulk C/N ratio correlates negatively with depth and SOM age. Although significant correlation between δ15N and δ13C is absent, both correlate negatively with C/N of HA and SOM age; δ13C correlates positively with FA/HA recovery. The humification analogues (A600/C, ΔlogK and E4/E6) correlate well with each other, but do not correlate significantly with FA/HA recovery or SOM age. There is a strong positive correlation between C/N of HA and SOM age, whereas there is a negative correlation between FA/HA recovery and SOM age.

[28] A correlation matrix for samples from mineral soil horizons shows that increased TOC content correlates negatively with FA/HA recovery and depth (Table 3). Further, there is a strong positive correlation between δ13C and depth. The humification analogues, Δlog K and E4/E6, correlate strongly with each other, but not with A600/C.

[29] In the PCA, the first four loading scores explain 75% of the variance in the complete data set (Table 4). The first principal component (PC) explaining 33% of all variance, is dominated by the influence of humification analogs. The loadings are very high for Δlog K and E4/E6 (≈0.9), whereas the influence of A600/C and FA/HA recovery is smaller (below 0.6). There is significant regression (r = −0.62), between the supplementary variable TOC content and PC1. PC 2 (19% of all variance) is most strongly influenced by δ13C (0.69) and C/N of HAs (−0.62); PC2 has a significant regression coefficient of −0.47 to depth. The strongest loading for PC 3 (12% of variance) is bulk C/N (−0.59) where there is a regression coefficient of 0.33 with depth. PC4 (11% of variance) is most influenced by the variability in H/C of HA (−0.58) and FA/HA recovery (0.50) and has a regression coefficient of −0.57 with TOC. Permafrost condition shows a weak negative regression with PC2 (−0.16).

Table 4. Summary of PCA Resultsa
 Explained Variance
PC1PC2PC3PC4
33.1%18.8%12.0%10.7%
  • a

    Shown are explained variance loading scores for geochemical variables and regression coefficients for environmental variables for the first four PCs. Loading scores >0.8 are bold and scores 0.8 > 0.5 are italic. For the regression coefficients for environmental variables, near significant regressions (p 0.10 > 0.05) are in italic and asterisks denote different levels of statistical significance: *, p < 0.05; **, p < 0.01; *** p < 0.001.

Loading Scores for Geochemical Variables
C/N (Bulk)0.350.47−0.590.17
δ 15N0.550.310.270.23
δ 13C0.420.690.250.16
A600/C0.590.35−0.500.33
Δ log K−0.910.200.180.21
E4/E6−0.860.200.180.28
C/N (HA)0.06−0.620.500.09
H/C (HA)0.360.520.09−0.58
FA/HA % rec.0.580.210.150.52
 
Regression Coefficient for Environmental Variables
Permafrost0.01−0.16*0.07−0.05
Depth (cm)−0.00−0.47***0.33**−0.43***
TOC %−0.62***−0.10−0.01−0.57***

3.3. Classification of HAs

[30] Classifying samples according to Kumada [1987] reveals clear differences between mineral and organic soil horizons in Seida (Figure 2). The HA samples from the shrub tundra sites are of Rp-type (organic horizons), or weakly humified B-type acids (mineral horizons). Samples from the dwarf birch site have low A600/C and Δlog Kthat decreases with depth. In the spruce forest, O-horizon samples are Rp-type HA, whereas HAs at the base of the A-horizon are highly humified P-type HA. A600/C decreases with depth, whereas Δlog K (and E4/E6) remains high. The tundra fen samples are Rp-type HAs. In the marginal plateau fen, all samples except the top mostSphagnumpeat are Rp-type HAs. The plateau fen shows large variability in humification in the surface peat and at the peat/mineral contact. While most samples are Rp-type HAs, two samples are B-type, and some are P-type HAs. In the peat plateaus all samples are Rp-type HAs, whereas some mineral sub-soils are type-B HAs.

Figure 2.

HAs from all nine sites classified according to Kumada's [1987] method into A, B Rp1, Rp2 and P type HAs. Samples from organic soil horizons are shown as solid dark symbols while mineral soil horizons are shown as light gray symbols. The degree of humification in HAs increases following the order Rp < B < A [Kumada, 1987; Ikeya and Watanabe, 2003].

3.4. Maps of SOM Humification, Peat Plateau SOM Age and SOC Remobilization Potential

[31] The maps in Figure 3 show soil C storage (kg C m−2), FA/HA recovery (%) and dominant HA classes for organic and mineral soil horizons separately. Soil C storage in the landscape is high in peat plateaus. There is little variability in soil C storage in mineral soil horizons compared to organic soil horizons (9.0 to11.8 kg C m−2 versus 2.9 to141.4 kg C m−2, respectively). In general, the recovery of FA/HA is very low in most O-horizons, whereas it is considerably higher in the mineral soil-horizons. Organic soil horizons contain Rp1-type or Rp2-type HAs, whereas mineral soil horizons contain Rp2-type or B-type HAs.

Figure 3.

Maps show SOM storage (kg C/m2), HA/FA recovery (%) and HA class (following Kumada's [1987] method) for organic and mineral soil horizons separately.

[32] Analyses of peat age-depth models and inferences on timing of local permafrost aggradation [Oksanen et al., 2001; Andersson et al., 2011; Becher, 2011; Routh et al., manuscript in preparation, 2012] suggest that the uplifted permafrost peat plateau deposits (sites SE8 and SE9) have been stored in permafrost for only a limited period of their development history, and the bulk of the currently frozen peat accumulated in a permafrost-free environment (Figure 4). The permafrost environment has been dynamic with periods of permafrost aggradation at 3100 and 2200–800 followed by total or partial collapse before the establishment of the current peat plateau extent. Calculating the proportional residence times when SOM has been in/out of permafrost (equivalent to the shaded areas in Figure 4), shows that 72% of the SOM residence time has been permafrost-free, 5% has been in the active layer of permafrost deposits and 23% has been in permafrost. One out of the ten peat cores that this reconstruction is based on is currently permafrost free (a collapse scar fen), which gives an estimate that 10% of present peat plateau SOC storage is currently permafrost-free (Figure 4). This can be compared with upscaled landscape mapping by Hugelius et al. [2011]which estimate that 11% of the peat-deposit SOC within the Seida peat plateau extent is currently stored in thermokarst or permafrost-free fens.

Figure 4.

Estimated net-accumulation age for SOC in the present-day peat plateau sites (SE8 and SE9). The different shaded surfaces show: SOC accumulated/stored in a permafrost free environment (including the active layer after permafrost aggradation), SOC accumulated/stored in the active layer of an environment when permafrost had aggraded and SOC accumulated/stored in permafrost.

[33] Calculating the storage of soil C divided between organic and mineral soil horizons reveals that more than 70% of the total landscape SOC pool (calculated to 1 m depth in mineral soils and full depth of peatlands) is in organic soil horizons where FA/HA recovery is <10%, and HA are of type Rp1 or mixed type Rp1/Rp2 (Figure 5a). In contrast, soil C stored in mineral soil horizons has 20–30% FA/HA recovery and Rp2-type HA (Figure 5b). The two different scenarios of SOM remobilization following permafrost thawing shows that active layer deepening and thermokarst expansion will largely affect the different SOM pools (Figures 5c and 5d). Active layer deepening in the 21st century is expected to mainly thaw permafrost SOM with B type HA in mineral soil horizons (equivalent to 37% of all landscape SOC to 3 m depth by the end of this century Figure 5c), and to a lesser degree SOM with Rp2 type HA in mineral soil horizons and Rp1/Rp2 type HAs in organic soil horizons (mainly in peat plateaus). Lateral expansion of thermokarst would mainly affect the peat plateaus resulting in remobilization of SOM with Rp1/Rp2 type HA in peat deposits (equivalent of up to ca. 15% of the total landscape SOC pool if present thermokarst expands 30 m; Figure 5d). Additionally, the deeper mineral soil horizons could also be affected (equivalent of up to ca. 10% of the total landscape SOC pool; Figure 5d).

Figure 5.

(a and b) Partitioning of SOC storage in the landscape between organic and mineral soil horizons and (c and d) remobilization potential of SOC following permafrost thaw. Graph A shows SOC storage divided according to the FA/HA recovery and graph B shows SOC storage divided according to the HA type of the bulk SOM. Graph C shows active layer SOC storage (expressed as percentage of total landscape SOC to a depth of 3 m calculated following Hugelius et al. [2011]) divided according to the HA type of the bulk SOM (analyses does not include fens) for different time periods of the 19th and 21st century (active layer depth from GIPL2 model following Hugelius et al. [2011]) and graph D shows SOC affected by thermokarst (expressed as percentage of total landscape SOC to a depth of 3 m) following expansion of current thermokarst by 1 to 30 m (simulated in a GIS following Hugelius et al. [2011]).

4. Discussion

4.1. Humification as a Proxy for Degree of Decomposition

[34] Structural changes during humification have been measured by different methods over the last several decades [Schnitzer and Desjardins, 1965; Raitio and Huttunen, 1976; Stanek and Silc, 1977; Lévesque and Mathur, 1979; Tolonen, 1982; Kuhry and Vitt, 1996; Caseldine et al., 2000]. In Alaskan tundra soils, HA and FA fractions are more decomposed; they contain more carboxyl-carbonyl C and are less bioavailable than the low-molecular weight SOM fractions [Dai et al., 2001]. Humification is a rather complex process, and therefore no single method is all-inclusive or perfect to be used as a standard accepted protocol to trace these biochemical changes. Hence, researchers advocate the use of multiple proxies including geochemical, spectroscopic and petrological methods to trace humification changes [seeZaccone et al., 2011]. In this study, we have used a combination of low-cost geochemical and spectroscopic methods to trace the humification and decomposition trends, and its relation to SOM formation and storage. This has allowed us to investigate and compare the soil horizons in many sites with different soil and land cover types. The results of the SOM characterization have been combined with existing high-resolution maps of soil C storage to quantify the SOM pools with low, medium or high degrees of decomposition in the landscape. The humification index in soil appears to be a robust method for assessing its degree of decomposition. Emerging knowledge suggests that the molecular structure of SOM may not be a key control of SOM stability, and environmental and biological factors may be more important [Schmidt et al., 2011]. However, we believe that at landscape scale, the degree of decomposition is a useful proxy for SOM lability and is therefore useful for estimating the C remobilization potential in periglacial landscapes under climate warming.

4.2. Overall Trends in Geochemical Proxies

[35] While some consistent trends can be observed from the individual pedons/peat cores (Figures S1–S3 and section 3.1), robust overall trends and patterns in the whole data set emerge more clearly from correlation and PCA analyses. There is an absence of strong regressions between any of the PCs and permafrost (Table 4). Notably, factors such as TOC% content (reflecting differences between organic and mineral soil horizons), botanical composition and storage history (including past shifts in hydrology and permafrost dynamics) are more important controls on SOM lability than the present-day permafrost. Correlations between TOC and humification proxies and FA/HA recovery indicate that C-rich O-horizons are associated with a lower degree of humification and FA/HA recovery than mineral horizons (Table 1). The first PC of the PCA is dominated by the spectral humification analogs and FA/HA recovery (Table 4), which shows that extraction and spectral characterization of HAs produces the most significant gradient in our data set.

[36] The bulk C/N ratio correlates negatively with DBD, depth and SOM age (in O-horizons) indicating increased decomposition in the deeper soil horizons (Tables 1 and 2). For non-peatland soils, these correlations match the observed trends in sites SE1–3 and SE7 (Figures S1a–S1c and S3a). In peatlands, the correlations may be influenced by the fact that all horizons of Sphagnum peat (very high C/N ratio due to its botanical composition) are in the upper and younger peat horizons. In general, the observed bulk C/N ratios in this study are consistent with Sphagnum and fen peat deposits from the Arctic ecoclimatic region [Vardy et al., 2000].

[37] Correlations between δ13C and FA/HA recovery (and depth in mineral soil horizons) indicates enrichment of δ13C during the formation of humic compounds with SOM aging (Tables 1 and 3). However, in organic soil horizons there are negative correlations between stable isotope enrichment and SOM age (Table 2). Higher isotopic ratios in surface peats and lack of enrichment with age/depth occur in cores SE5–6 and SE8–9. These patterns may be linked to the differences in botanical composition of Sphagnum spp. and rootlet peat in surface layers, and graminoid dominated fen peats occurring at greater depths.

[38] The individual peat cores show a declining recovery of FA/HA with depth (Figures S2 and S3), this is also evident from the negative correlation of FA/HA recovery with depth and SOM age in the organic soil horizons (Table 2). Higher FA/HA recovery in surface peats may be caused by changes in permafrost dynamics and/or surface hydrology over time, or removal/leaching of FA/HA by flowing water in sub-surface peat layers (applies to the sites SE4 and SE5). In organic soil horizons, the humification analogs (A600/C, Δlog K and E4/E6) correlate well with each other (Table 2), but they do not correlate with FA/HA recovery or SOM age. This is consistent with observations of nonlinear relationships and absence of increased decomposition with depth in individual peat deposits (Figures S2 and S3). In mineral soil horizons the humification analogs Δlog K and E4/E6 correlate significantly with each other, but not with A600/C indicating that SOM coloration is not as closely linked to molecular complexity in mineral horizons as it is in the organic soil horizons (Table 3).

[39] The C/N ratios in HAs increase linearly with age, and with enrichment of stable isotopes in peat deposits (Figures S2 and S3 and Table 2). This interpretation is further corroborated by the PCA where PC 2 is most strongly influenced by δ13C and C/N of HAs, and has a negative regression with depth suggesting depletion of 13C, increased C/N in HAs and decreased H/C in HAs. Kumada [1987] indicated that H/C ratio in HAs decrease with humification. There is lack of significant correlation supporting this trend in our data (Tables 13), but PC4 of the PCA shows that the H/C ratio decreases with increased FA/HA recovery (Table 4).

4.3. Pedogenesis, Permafrost Aggradation and SOM Decomposition

[40] The presence of well-developed soil horizons and clear trends in some geochemical proxies (e.g., bulk C/N,δ13C and δ15N) implies continuation of soil formation in the permafrost-free mineral soils (SE1–3). In general, SOM in mineral soil horizons is associated with a higher degree of humification than SOM in organic soil horizons. However, humification in the Seida soil horizons is low compared to boreal soils [Kumada, 1987; Ikeya and Watanabe, 2003]. This may be due to the relatively young age of SOM (Table S1). Moreover, protection by soil aggregates and in situ low temperatures reduce SOM decomposition rates. Peat formation in the present-day fens commenced around 3800 and 3300 cal yrs BP (no data available for SE5). Decomposition in the fen peats is generally low with some exceptions in SE6 (Figure S2c). There is considerable variability in SE6, which is possibly connected to the botanical composition of the peat. Consistent with the bulk SOM trends presented in this study, biomarker analyses in the total lipid fractions indicate that SOM in the peat is immature and has undergone little decomposition [Routh et al., 2011]. Likewise, the high acid/aldehyde ratio in lignin phenols suggest that SOM degradation mainly occurs in the upper horizons, which are more aerated [Routh et al., 2011].

[41] The SOM stored in peat plateaus ranges in age from 8 kyrs cal BP to modern, and shows remarkably little variability in decomposition or humification rates. Peat deposits in this region first developed as permafrost-free fens in which epigenetic permafrost aggradation (permafrost that formed after the deposition of the soil material in which it occurs) led to the formation of peat plateaus between 3200 and 1950 cal yrs BP and as recent as the Little Ice Age [Oksanen et al., 2001, 2003; Kultti et al., 2004; Andersson et al., 2011; Becher, 2011]. In line with these previous studies, Routh et al. (manuscript in preparation, 2012), through a combination of radiocarbon dating, macrofossil analyses and gross-stratigraphy, found that epigenetic permafrost first aggraded between 2200 and 820 in the SE8 and SE9 peat plateaus, followed by a period of partial collapse until the most recent peat plateau expansion phase. Thus, the peat deposits investigated in this study have been stored in permafrost for only a limited period of their development history (ca. 22%), and the bulk of the currently frozen peat accumulated in a permafrost-free environment (Figure 4). Despite this, the degree of decomposition is similar throughout the peat deposits, and this has an important implication. We infer that anoxia during earlier fen stages of peatland succession has been equally (or more) important than permafrost in inhibiting SOM decomposition. Because sub-zero temperatures inhibits decomposition, SOM in permafrost is often assumed to be relatively undecomposed leading to substantial post-thaw remobilization within decadal timescales [e.g.,Schuur et al., 2011]. SOM stored in syngenetic permafrost deposits (permafrost that formed more or less simultaneously with the deposition of the soil material in which it occurs) may have undergone little decomposition prior to permafrost aggradation because of rapid accumulation rates in e.g., Yedoma or peat deposits [Zimov et al., 2006; Sannel and Kuhry, 2009]. However, in epigenetic permafrost deposits SOM may have been exposed to substantial degradation prior to permafrost stabilization [Schuur et al., 2008]. This study provides an example of epigenetic permafrost deposits that have mainly developed in permafrost-free environments where anoxia has limited decomposition rates.

[42] There is a marked increase in humification and FA/HA recovery in mineral soil horizons underlying peat deposits which may be caused by the presence of paleosols prior to peat formation. This is especially relevant for the present-day fen peats, which did not develop until after the Holocene Climate Optimum. During the Holocene Climate Optimum the study area was forested [Kultti et al., 2004], and it is likely that well-developed soil horizons were present prior to the development of fen peats. This suggests that the B/C horizons in sites SE1–3 are contemporary soil horizons, whereas in the other sites there may be buried horizons.

4.4. SOM Storage and Remobilization Potential at a Landscape Scale

[43] To the best of our knowledge, there are no integrative multiproxy landscape scale studies of the degree of SOM decomposition in periglacial landscapes. Kumada's method of HA classification [Kumada, 1987] (Figure 2) shows that compared to more comprehensive investigations from boreal soils [Kumada, 1987; Ikeya and Watanabe, 2003], humification is generally low in the investigated soils. Lowe and Kumada [1984]characterized HA in the active layer of a Turbic Cryosol in the Canadian Arctic and found that the bulk SOM was composed of Rp and B-type HAs, which is consistent with the findings in this study. Combining the results from extraction and classification of humic compounds with a GIS based SOC storage map for the Seida region (Figure 3) shows clear differences between the mineral soils versus peat deposits (especially in the uplifted peat plateaus). The maps show a large variability in SOC storage in organic soil horizons, and are characterized by very high values in peat plateaus. In contrast, there is less variability in SOC storage in mineral soil horizons. The recovery of FA/HAs is low in organic horizons, but is relatively high in mineral horizons of non-peatland, permafrost free soils. Organic soil horizons contain weakly humified HAs (Rp1-types or Rp2-type), whereas mineral soil horizons contain Rp2-type or B-type HAs. SOM storage (>70%) in the landscape is primarily concentrated in the organic soil horizons with low humification ratios (Figures 5a and 5b). This corresponds to previous studies which have shown that permafrost peat plateaus are the main stores of soil C in the region (≈60% [Hugelius and Kuhry, 2009]).

[44] By the end of this century, processes of top-down thawing (active later layer deepening) and lateral thawing (thermokarst expansion) are expected to remobilize previously frozen SOC in Seida [Hugelius et al., 2011]. However, due to the insulating properties of dry peat, active layer deepening in peat plateaus is expected to be limited during this century. Peat plateaus typically have an ice-rich core sensitive to permafrost thaw and settlement of upper soil/peat horizons. We have applied the SOC remobilization scenarios described inHugelius et al. [2011]which show that active layer deepening in the 21st century will remobilize SOM in mineral soil horizons with B and Rp2-type HAs (Figure 5c), whereas thermokarst expansion would mainly affect SOM in organic soil horizons with Rp1/Rp2 type HA (Figure 5d). This implies that thermokarst will affect SOM that has undergone less decomposition than that affected by active layer deepening. However, further decay of SOM affected by active layer deepening or thermokarst will also depend on the post-thaw environment. There is field based evidence of relatively rapid mineralization of SOM following permafrost thaw [Schuur et al., 2009; Lee et al., 2010], but there are also potential post-thaw mechanisms that may stabilize SOM including anoxia due to saturation (in thermokarst lakes or at the base of the active layer in soils) or long-term formation of pedogenic reactive minerals or soil structures [Schmidt et al., 2011].

5. Conclusions

[45] There is good agreement between different indicators of decomposition of SOM in the investigated area. At a landscape scale, the combination of FA/HA% recovery and spectral characterization of HAs is a robust indicator of SOM decomposition. We suggest that bulk characterization of SOM involving such simple inexpensive methods could be routinely applied for reconnaissance studies. Kumada's [1987]method of HA classification suggests that humification is low in this periglacial landscape. There is low decomposition of organic matter in the O-horizons of mineral soils (young material) and in peat deposits (from modern to 8 kyrs cal BP). In permafrost-free non-peatland soils, enrichment of stable isotopes and decrease in bulk C/N ratio indicates greater decomposition of SOM with depth. The recovery HAs and FAs is higher in well-developed permafrost free soils, but lower in peatlands. Spectral characterization of HAs indicates higher degree of humification in the deeper soil horizons of non-peatland soils (A, B and B/C-horizons), and in mineral soils underlying the peat deposits.

[46] Combining maps of landscape SOC storage with maps of SOM decomposition shows that the bulk (>70%) of SOC in this discontinuous periglacial landscape is stored in SOM with a low degree of decomposition in the upper O- and A-horizons.

[47] The bulk of permafrost SOM in the study area is stored in peat plateaus that accumulated during the development of the fen peatland stages prior to initial permafrost aggradation during the late Holocene, or recent times (≥72% of SOM residence time has been permafrost-free). During these non-frozen fen stages, SOM was protected from decomposition by anoxic conditions in these deposits. Notably, presence/absence of permafrost is not a key factor controlling the degree of decomposition in SOM.

[48] Simulations of SOM remobilization following scenarios of permafrost thawing show that active layer deepening affects the mineral soil horizons with a relatively high degree of decomposition. Thermokarst expansion mainly affects organic soil horizons with low degree of decomposition stored in peat plateaus. Peat plateaus commonly occur across Arctic landscapes in sporadic, discontinuous and southern continuous permafrost zones, and may become highly susceptible to permafrost degradation [Sannel and Kuhry, 2011]. Thus, we conclude that permafrost has not been the key environmental factor controlling SOM decomposition in this landscape. However, the presence/absence of ice-rich permafrost is nonetheless the key control for susceptibility of undecomposed SOM to rapid remobilization through thermokarst.

Acknowledgments

[49] We wish to thank the late Galina Mazhitova for support in the planning and execution of field sampling and soil mapping. We thank Tarmo Virtanen and Sanna-Maija Susiluoto for providing us the data on land cover in Seida. We are grateful to Nathalie Pluchon and Martin Wik for help with soil sampling. The study was funded through the EU 6th Framework CARBO-North project (contract 036993) and a grant of the Swedish Research Council to P. Kuhry.

Ancillary