The large stocks of soil organic carbon (SOC) in soils
and deposits of the northern permafrost region are sensitive to global
warming and permafrost thawing. The potential release of this carbon (C) as
greenhouse gases to the atmosphere does not only depend on the total
quantity of soil organic matter (SOM) affected by warming and thawing, but it
also depends on its lability (i.e., the rate at which it will decay). In this study
we develop a simple and robust classification scheme of SOM lability for the
main types of soils and deposits in the northern permafrost region. The
classification is based on widely available soil geochemical parameters and
landscape unit classes, which makes it useful for upscaling to the entire
northern permafrost region. We have analyzed the relationship between C
content and C-
Permafrost has been recognized as one of the vulnerable carbon (C) pools in
the Earth system (Gruber et al., 2004). A recent report of the International
Panel on Climate Change (IPCC, 2018) identifies the thawing permafrost
carbon-climate feedback as one of the key uncertainties when assessing
global emission targets to keep global warming under 1.5 (2)
In the last decade there has been a surge in papers dealing with the permafrost carbon feedback on climate change (e.g., Schuur et al., 2008; Kuhry et al., 2010). This increased interest was fueled by a new and high estimate of the total soil organic carbon (SOC) storage in the northern permafrost region (Tarnocai et al., 2009), which was received with great interest by the Earth system science community (e.g., Ciais, 2009). Since this first new estimate was published, a multitude of new SOC inventories at the landscape level have been conducted across the circumpolar north (e.g., Hugelius and Kuhry, 2009; Hugelius et al., 2010; Horwath Burnham and Sletten, 2010; Palmtag et al., 2015; Gentsch et al., 2015a; Siewert et al., 2016). Recent studies have also focused on re-evaluating the spatial extent and SOC storage of the Yedoma “Ice Complex” and alas deposits (Strauss et al., 2013; Walter-Anthony et al., 2014; Hugelius et al., 2016; Shmelev et al., 2017).
These new data have prompted an update of the total SOC storage in the northern permafrost region, its vertical partitioning and its broad (continental-scale) distribution (Hugelius et al., 2014). The new estimate amounts to ca. 1400 PgC for the top 3 m of soils and deeper deposits, including permafrost and non-permafrost organic soils (Histels and Histosols, 302 PgC), cryoturbated permafrost mineral soils (Turbels, 476 PgC), non-cryoturbated permafrost mineral soils (Orthels) and non-permafrost mineral soils (256 PgC), and deeper yedoma (301 PgC, > 300 cm) and delta (91 PgC, > 300 cm) deposits. The spatial distribution of SOC stocks according to the major permafrost soil (Gelisol) suborders, non-permafrost mineral soils and Histosols (Soil Survey Staff, 2014) is graphically represented in the updated version of the Northern Circumpolar Soil Carbon Database (NCSCDv2, 2014).
The importance of an accurate estimate of total SOC storage in the northern
permafrost region is illustrated by a recent review of the permafrost carbon
feedback (Schuur et al., 2015), which included a comparison of future C
release in a total of eight Earth system models. The magnitude of the
projected cumulative C loss from the permafrost region by 2100, largely
based on the RCP8.5 (Representative Concentration Pathway) scenario (IPCC, 2013), varied greatly between models
from 37 to 174 PgC. However, by normalizing for the initial C pool size in
the different models, the proportional C loss from the permafrost zone was
constrained to a much narrower range of
The magnitude of the permafrost carbon feedback, however, will not only depend on the rate of future global warming (and its polar amplification), its effect on gradual and abrupt permafrost thawing (Grosse et al., 2011), or the total size (and vertical distribution) of the permafrost SOC pool. As shown by Burke et al. (2012), based on simulations with the Hadley Centre climate model, quality (decomposability) parameters need also to be considered. Thus, in terms of C pool parameters, the potential C release from the northern permafrost region will depend not only on SOC quantity but also on soil organic matter (SOM) lability (i.e., the rate at which soil organic matter will decay following warming and thawing). Laboratory incubation experiments that consider both different types of substrates (e.g., Schädel et al., 2014) and time of incubation (e.g., Elberling et al., 2013) are an important tool to assess potential C release from permafrost soils and deposits.
The aim of this study is to add a measure of SOM lability to the current estimates of SOC quantity, in order to define vulnerable C pools across the northern circumpolar region. We focus on the relationship between solid-phase geochemical parameters (particularly C content) and C release rates in laboratory incubations of active-layer and thawed permafrost samples from the main types of soils and deposits found in the northern permafrost region. Our objective is to develop a SOM lability classification scheme based on widely reported soil geochemical parameters in field SOC inventories and general landscape classes that can be linked to existing spatial SOC databases such as the NCSCD (Tarnocai et al., 2009; Harden et al., 2012; Hugelius et al., 2014). We test the robustness of our SOM lability classification by comparing two very different types of incubation experiments, both in the setup as well as timing of C release measurements.
The samples used in the incubation experiments were collected as part of landscape-level inventories carried out in the context of the European Union Changing Permafrost in the Arctic and its Global Effects in the 21st Century (EU PAGE21) and European Science Foundation Long-term Carbon Storage in Cryoturbated Arctic Soils (ESF CryoCarb) projects to assess total storage, landscape partitioning and vertical distribution of SOC stocks in study areas across the northern permafrost region. SOC storage data from these areas are presented in Weiss et al. (2017) for Svalbard, Siewert et al. (2016) for Lena Delta, Palmtag et al. (2016) for Taymyr Peninsula, Palmtag et al. (2015) for Lower Kolyma, Hugelius et al. (2011) for Seida and Siewert (2018) for Stordalen Mire. The location of all study areas is shown in Fig. 1. The Lower Kolyma experiment includes samples from two nearby located study areas (Shalaurovo and Cherskiy); the Taymyr Peninsula experiment also includes samples from two nearby located study areas (Ary-Mas and Logata). Metadata for each of these areas, including the geographic coordinates, permafrost and vegetation zones, climate parameters, number of soil profiles and incubated samples, type of incubation experiment, and time of field collection, are presented in Table S1 (Supplement).
Location of study areas in northern Eurasia. PAGE21 experiment (Ny-Ålesund, Adventdalen, Stordalen Mire, Lena Delta); CryoCarb 1-Kolyma experiment (Shalaurovo, Cherskiy); CryoCarb 2-Taymyr experiment (Ary-Mas, Logata); CryoCarb 3-Seida experiment. Permafrost zones according to Brown et al. (1997, 2002).
The sampling strategy applied for SOC field inventories was aimed at capturing all major landscape units in each of the study areas, while at the same time it ensured an unbiased selection of soil profile location. This semi-random sampling approach consisted of deciding on the positioning of generally 1 or 2 km long transects that crossed all major landscape units, with a strictly equidistant sampling interval at normally 100 or 200 m that eliminated any subjective criteria for the exact location of each soil profile. For SOC storage calculations, the mean storage in each landscape unit class was weighed by its proportional representation in the study area based on remote sensing land cover classifications.
At each soil profile site, the topsoil organic layer was collected by cutting out blocks of a known volume in three random replicates to account for spatial variability. These samples do not always strictly adhere to the definition of an `O' (organic) soil genetic horizon, because in areas with thin topsoil organics (like in floodplains and mountain terrain) there can be a large admixture of minerogenic material resulting in C contents of less than 12 %. Active-layer samples were collected from excavated pits by horizontally inserting fixed-volume cylinders. The permafrost layer was sampled by hammering a steel pipe of a known diameter incrementally into the ground, retrieving intact samples for each depth interval. Depths intervals are normally 5 to 10 cm or less (e.g., when the topsoil organic layer was very thin). The standard sampling depth was down to 1 m below the soil surface; at some sites it was not possible to reach this depth due to large stones in the soil matrix or thin soil overlying bedrock (often in mountainous settings).
The PAGE21 incubation experiment was carried out at the University of Copenhagen (Denmark). This experiment included one sample from the topsoil organics, one sample from the middle of the active layer and one sample from the upper permafrost layer (normally 10–15 cm below the upper permafrost table) from all mineral soil profiles collected in three of the PAGE21 study areas (Ny-Ålesund, Adventdalen and Lena Delta). Samples were selected based on depth criteria and not any specific soil characteristic (e.g., presence of C-enriched cryoturbated material or absence of excess ground ice). In some cases, upper permafrost samples could not be collected due to very deep active layers and/or thin soils (particularly in mountain settings). Peat samples are available from a fourth PAGE21 study area (Stordalen Mire). In total ca. 240 soil samples from four study areas across the northern permafrost region (Ny-Ålesund and Adventdalen, Svalbard; Stordalen Mire, northern Sweden; Lena Delta, northern Siberia) were incubated in one and the same experiment (Faucherre et al., 2018).
The dry bulk density (DBD) of samples used for incubation was measured at Stockholm University (Sweden). The %C and %N of dry weight of the incubated samples were measured in an elemental analyzer (EA Flash 2000, Thermo Scientific, Bremen, Germany) at the University of Copenhagen (Denmark).
Samples were kept in frozen condition from collection until the start of the
laboratory incubation experiment. Samples were incubated at 5
The CryoCarb incubations were carried out at the University of South Bohemia (České Budějovice, Czech Republic). These experiments included all samples from all profiles collected in each of three study areas (CryoCarb 1-Kolyma in northeastern Siberia, CryoCarb 2-Taymyr in northern Siberia and CryoCarb 3-Seida in northeastern European Russia). In total ca. 1000 samples were incubated.
The dry bulk density of samples used for incubation was measured at Stockholm University (Sweden). The %C and %N of dry weight were measured in an EA 1110 Elemental Analyzer (CE Instruments, Milan, Italy) at Stockholm University (Seida samples) and the University of Vienna (Kolyma and Taymyr samples).
The CryoCarb 1-Kolyma and CryoCarb-2 Taymyr samples were stored in a ground pit dug into the active layer for up to 2 weeks, before further processing. Active-layer samples would be little impacted by this storage under “natural” conditions, but (some of) the gradually thawing permafrost layer samples might have experienced initial decay. CryoCarb 3-Seida samples collected in 2009 were kept in frozen storage for ca. 10 years (see Table S1), before further processing.
In the laboratory, soil samples were dried at 40–50
The study area and layer-specific composite soil inoculi were prepared from
fresh soil taken separately from the topsoil organic-layer, mineral active-layer, peat active-layer, mineral permafrost layer and peat permafrost
layer from multiple soil profiles collected in each study area. Fresh soil
was kept in a cold room (at 4
The short term C-
As potential explanatory geochemical parameters we have considered dry bulk
density, carbon content (%C of dry weight) and carbon-to-nitrogen
weight ratios (
We have investigated C-
In addition, we have applied a further subdivision of landscape unit classes in the PAGE21 experiment to allow a more detailed statistical analysis of the dataset and assess the role of minerogenic inputs, cryoturbation and peat accumulation in SOM lability. For this purpose, the eolian class is separated into actively accumulating deposits (Adventdalen) and Holocene soils formed into Pleistocene yedoma parent materials (Lena Delta). Alluvial deposits are subdivided into profiles from active and pre-recent floodplains (multiple study areas). Mineral soils are separated into active colluviation sheets (mountain slopes on Svalbard) and other mineral soils (multiple study areas). Finally, for wetland deposits we discriminate between peat deposits (fens and bogs in Stordalen Mire; > 40 cm peat) and peaty wetland profiles (multiple study areas; < 40 cm peat). It should be stressed that these subclasses are not specifically recognized in any circumpolar SOC database and are therefore of limited use for further upscaling. In all cases, SOM lability in samples of deeper C-enriched buried layers and cryoturbated pockets is shown for comparative purposes.
Relationships between C-
To alleviate the issue on non-normal distributions, C-
We first assessed the relationship between C release rates in incubation
experiments and widely available physicochemical parameters in samples from
soil carbon inventories carried out throughout the northern permafrost
region. The latter include dry bulk density, C content as a percentage
of dry sample weight, and carbon-to-nitrogen weight ratios. In
recent studies dealing with the incubation of soil samples from the northern
permafrost region, %C and
All three considered geochemical parameters are significantly
(anti-)correlated with measured C release rates in the four different
incubation experiments. Lower DBD, higher %C and higher
Our results show a significant relationship between
Figure 2 shows the relationships between C release rates and %C in the samples for the data grouped according to major landscape unit classes in the PAGE21 (measured on day 343 of incubation) and CryoCarb 1-Kolyma (measured over the first 4 d of incubation) experiments. For the sake of simplicity, we apply linear regressions with intercept zero to all classes. These are identified by different colors and symbols that have been consistently used in Figs. 2–3 and S3–S5. The regression for the full dataset is provided as a reference (dotted lines), but it should be noted that its slope is partly determined by the number of samples in each of the recognized landscape units.
A first observation is that C release rates per gdw are ca. 15 times lower in
the longer-term PAGE21 experiment compared to the short-term CryoCarb
1-Kolyma experiment. In the PAGE21 dataset (Fig. 2a), the soils developed
into alluvial and eolian deposits, and in peaty wetlands they all show similar and
relatively high SOM lability. Mineral soils show intermediate values,
whereas the peat deposits display low SOM lability (when considering %C
values). All regressions are significant, except for “peat deposits” due to
very high variability in three surface peat samples (but see Fig. 3d). In
the CryoCarb 1-Kolyma dataset (Fig. 2b), alluvial and eolian soils and deposits
show the highest SOM lability, followed by mineral soils. In this case,
peaty wetlands show a slightly lower lability than mineral soils and deposits,
but they still show considerably higher lability than peatlands. This clear dichotomy in the
SOM lability of mineral soils (including peaty wetlands) and peat deposits
is also apparent from the CryoCarb 2-Taymyr and CryoCarb 3-Seida results
even though not all landscape classes are represented in those experiments
(Fig. S3). The explanatory power of the regressions (
Linear regression analyses between C-
The PAGE21 dataset with C-
Figure 3a–c present SOM lability in a further subdivision of the mineral
soil and deposit landscape classes in the PAGE21 dataset. We have compared
profiles with active accumulation or movement in eolian, alluvial and colluvial
settings, with Holocene soils developed into older eolian, alluvial or other
mineral parent materials, respectively. In each of these comparisons, we
specifically identify samples from deeper C-enriched buried layers and
cryoturbated pockets. Generally speaking, a second-order polynomial
(intercept zero) provides the best fit and has been applied for the sake of
uniformity to all described subclasses. All these datasets have in common that the subclasses with active surface accumulation or movement have topsoil
samples that show relatively low C content due to the continuous admixture
of minerogenic materials. At the same time, these all show the highest
C-
Figure 3d compares the SOM lability in fen and bog deposits (Stordalen Mire)
and peaty wetland profiles (multiple study areas), adding for comparison the
results from the previously described deeper C-enriched buried layers and
cryoturbated pockets in mineral soils (see Fig. 3a–c). In this case,
exponential functions best describe observed trends and indicate very high
lability of surface peat(y) samples. The thin peat layers in peaty wetlands
have relatively low %C values pointing to admixture of minerogenic
materials. The SOM in these profiles show relatively high C-
In the PAGE21 incubation each collected profile included only one sample
from the mineral soil in the middle of the active layer and one sample from
the upper permafrost layer. Thus, the selection of samples was based on
depth-specific criteria. As a result, the number of samples from deeper
C-enriched buried layers and cryoturbated pockets is limited (
The PAGE21 experiment does not include any samples from Pleistocene yedoma
deposits. In contrast, the CryoCarb 1-Kolyma dataset has samples from two
yedoma exposures along river and thermokarst lake margins. The material was
collected from perennially frozen yedoma deposit as well as from thawed-out
sections of the exposures. C release from these samples are presented in
Fig. 4b, which for comparison also shows samples from Holocene lowland
soils, mineral subsoil samples beneath peat deposits and deeper C-enriched
cryoturbated samples. The C-
Table 3a shows the slopes of the linear regressions (intercept zero) between
C-
We also tested SOM lability in samples grouped according to soil horizon criteria (PAGE21 and CryoCarb 1-Kolyma experiments), with special attention to those horizon classes that can be linked to the specific landscape classes showing low relative SOM lability (C-enriched pockets for Turbels, peat samples for Histels and loess samples for Pleistocene yedoma). This approach yielded classes with data distributions that are much better constrained in terms of %C values.
In this analysis we focus on C-
Figure 5 shows C-
C release rates in topsoil organic samples from actively accumulating Holocene loess soils are significantly higher than those in topsoil organic samples from the remaining PAGE21 mineral soils (Fig. 5a and Table 4a). Both topsoil organic classes show significantly higher rates than all mineral soil and peat classes. Peat samples have the lowest mean and median C release rates from all these classes, but only the rates from permafrost layer mineral soil and C-enriched pocket samples are significantly higher. Both mean and median C release rates from active-layer and permafrost layer C-enriched pockets are somewhat lower (but not significantly different) than those from adjacent, non-C-enriched mineral soil samples.
C release rates in the soil horizon classes from the CryoCarb 1-Kolyma experiment show similarities, but there are also some differences compared to those observed in the PAGE21 experiment. Absolute C release rates per gC are more than an order of magnitude higher in the CryoCarb 1-Kolyma experiment (measured as a mean release over the first 4 d of incubation) compared to those in the PAGE21 experiment (measured on day 343 of incubation). Another important difference is that C release rates per gC in the short-term CryoCarb 1-Kolyma incubation do not differ significantly between the topsoil organic class and the active-layer and permafrost layer mineral soil classes, which we ascribe to the presence of a highly labile C pool (e.g., DOC and plant roots) in the mineral soil layers that is quickly decomposed (see Weiss et al., 2016; Faucherre et al., 2018). However, rates from active-layer and permafrost layer C-enriched pockets are significantly lower than those from adjacent, non-C-enriched mineral soil samples. Both active-layer and permafrost layer peat samples show significantly lower C release rates than all other classes, with active layer peat samples having significantly higher rates than permafrost layer peat samples. Samples from the Pleistocene yedoma loess “frozen” and “thawed” classes display significantly lower C release rates per gC than those in the topsoil organic-layer, active-layer and permafrost layer mineral soil classes, but they are significantly higher than those in the peat classes. The two yedoma classes do not differ significantly from each other, the active-layer and permafrost layer C-enriched pocket classes, or the peaty wetland class.
The analysis and comparison of results in the PAGE21 and CryoCarb 1-Kolyma
incubations show consistent trends in C-
In quantitative terms, C-
Nonetheless, a comparison of C-
A further subdivision of landscape classes and more careful analysis of
incubation results in the PAGE21 experiment provide additional useful
insights. For example, the separation of eolian deposits into actively
accumulating deposits during the Late Holocene (Adventdalen) and Holocene
soils formed into Pleistocene yedoma parent materials (Lena Delta) showed
clear differences in C release rates per gdw (when considering %C), with
the former displaying a higher SOM lability in topsoil organic samples (see
Fig. 3a). The topsoil organic samples from the actively accumulating eolian
deposits in Adventdalen also displayed significantly higher C release rates
per gC than all other topsoil organic, mineral layer and peat(y) horizon
classes (see Table 4a). The separation of alluvial deposits into active
floodplain deposits and Holocene soils formed in pre-recent river terraces
and of mineral soils into active colluviation sheets (mountain slopes on
Svalbard) and other mineral soils showed similar trends in SOM lability (see
Fig. 3b–c). These results suggest that the admixture of minerogenic material in
topsoil organics of actively accumulating eolian, alluvial and colluvial
deposits promotes SOM decomposition. Peaty wetland deposits display much
higher C release rates per gdw (when considering %C) than peatland
deposits (see Fig. 3d). These two landscape classes are poorly represented
in the PAGE21 experiment, but a statistical test of C release rates per gC
in these peat(y) soil horizon classes of the CryoCarb 1-Kolyma incubation
confirms this difference (see Table 4b). This is interesting because even
though wetlands with a thin peat layer do not have particularly high C
stocks, they can be important sources of methane (
The implementation of landscape classes (and their subdivisions) in the PAGE21 and CryoCarb incubation experiments have greatly constrained variation in C release rates compared to the full datasets. However, much within-class variability remains, and there is a need to further investigate the sources of this variability. Important additional soil and environmental factors such as microbial community, moisture, texture, pH, redox potential, etc. were not available for the (full) PAGE21 and CryoCarb datasets and could, therefore, not be tested. We conclude that additional research is needed to further constrain observed SOM lability across the northern permafrost region and within the classes proposed here.
The relatively low lability in the peatland class is surprising. The low
DBD, high %C and high
In the case of peat deposits, it should be considered if this low
decomposability is an evolved “biochemical trait” in peat-forming species
that maintains their favored habitat, similar to the role of
Our results on the low lability of peat deposits can be compared to the findings of Schädel et al. (2014) in their assessment of SOM decomposability in the northern permafrost region. That study recognized a group of organic soil samples (> 20 % initial C), ranging in depth between 0 and 120 cm. We consider that this group will include both topsoil organic samples as well as deeper peat deposits. In the Schädel et al. (2014) study, this group showed the largest range in decomposability, with some samples showing high potential C losses, whereas deeper organic samples were less likely to respire large amounts of C. We suggest, therefore, that both studies might show the same trends.
In our incubation experiments, SOM from deeper C-enriched buried layers and cryoturbated pockets show relatively low lability when compared to organic-rich topsoil samples. These results are corroborated by Čapek et al. (2015) and Gentsch et al. (2015b), who report low bioavailability of SOM in subducted horizons of Lower Kolyma soils (northeastern Siberia). The reason why this relatively undecomposed material displays low lability remains unclear. One reason could be that the decomposer community needs time to adapt to the new environmental conditions following thawing and warming; another one is that there is a simple mismatch between the microbial community adapted to decompose relatively undecomposed organic material and the physicochemical environment (e.g., higher bulk density) prevailing in (thawed-out) deeper soil horizons (Gittel et al., 2013; Schnecker et al., 2014). Kaiser et al. (2007) and Čapek et al. (2015) reported low microbial biomass in deeper C-enriched soil samples.
These results pose interesting questions regarding the role of organic aggregates and organomineral associations for SOM lability (e.g., Gentsch et al., 2018). On the one hand, our samples from topsoil organic horizons with active minerogenic inputs in eolian, alluvial and colluvial settings display (very) high C release rates, whereas deeper C-enriched soil materials show low decomposability. The underlying soil physicochemical and microbial processes require urgent attention in order to better constrain C release rates from soils and deposits in the northern permafrost region.
Pleistocene yedoma deposits, represented in the CryoCarb 1-Kolyma incubation experiment, also display low relative SOM lability, despite the incorporation of relatively fresh plant root material caused by syngenetic permafrost aggradation. These results are corroborated by results from Schädel et al. (2014) for their group of deep mineral samples (with yedoma provenance).
An important consideration is whether the consistent differences in relative SOM lability of landscape and soil horizon classes observed in our incubation experiments will be maintained over periods of decades to centuries of projected warming and thawing. Very short-term incubations, such as in the CryoCarb setup (4 d) might register the initial decomposition of highly labile SOM components, such as microbial necromass, simple molecules (e.g., sugars or amino acids), low molecular-weight DOC, etc., or it might not provide enough time for an adaptation of the microbial decomposer community to new environmental settings (Weiss et al., 2016; Weiss and Kaal, 2018). On the other hand, in longer incubation experiments such as in the PAGE21 experiment (1 year), the conditions in the incubated samples become gradually more artificial compared to field conditions. Specifically, microbes in long-term incubations become increasingly C limited, as no new C input by plants occur, whereas inorganic nutrients, such as nitrate or ammonium, accumulate to unphysiological levels. Care, therefore, should be taken when extrapolating our results over longer time frames if no corroborating field evidence for longer term decay rates can be obtained (e.g., Kuhry and Vitt, 1996; Schuur et al., 2009).
The PAGE21 and CryoCarb incubation experiments confirm results from previous
studies that simple geochemical parameters such as DBD, %C and
When considering the full datasets of the four experiments, our regressions
of C release as a function of %C were statistically significant but
explained less than 50 % of the observed variability. Subsequently, we
investigated whether a further division of samples into predefined landscape
unit classes would better constrain the observed relationships. In defining
these classes, we applied a scheme that could easily be used for spatial
upscaling to northern circumpolar levels. We adopted the main Gelisol
suborders (Histels, Turbels and Orthels), non-permafrost Histosols and
mineral soils, and types of deeper Quaternary (deltaic and floodplain,
eolian and yedoma) deposits that have been used in the NCSCD and related
products to estimate the total SOC pool in the northern permafrost region
(Tarnocai et al., 2009; Strauss et al., 2013; Hugelius et al., 2014). We
conclude that these landscape classes better constrain observed variability
in the relationships and that the relative SOM lability rankings of these
classes were consistent among all four incubation experiments, for both
regressions against %C and
An important conclusion from these results is that purportedly more
undecomposed SOM, such as in peat deposits (Histels and Histosols),
C-enriched cryoturbated samples (Turbels) and Pleistocene yedoma deposits,
does not seem to imply higher SOM lability. These three SOC pools, which
together represent
The soil geochemical data and incubation results presented in this paper are available upon request from Peter Kuhry (peter.kuhry@natgeo.su.se). For the full PAGE21 incubation dataset, please contact Bo Elberling (be@ign.ku.dk). For the full CryoCarb incubation dataset, please contact Jiří Bárta (jiri.barta@prf.jcu.cz).
The supplement related to this article is available online at:
PK developed the initial concept for the study. All authors contributed with the collection of soil profiles at various sites. The PAGE21 incubation experiment was planned and conducted at CENPERM (University of Copenhagen) by SF, CJJ and BE, whereas the CryoCarb incubation experiments were carried out at the University of South Bohemia (České Budějovice) under guidance of HS and JB. PK performed all statistical analyses, in cooperation with GH. All co-authors contributed to the writing of the paper, including its discussion section.
The authors declare that they have no conflict of interest.
The collection and laboratory analyses for Svalbard (Adventdalen and Ny-Ålesund), Stordalen Mire and Lena Delta samples were supported by the EU-FP7 PAGE21 project (grant agreement no. 282700). Lower Kolyma and Taymyr Peninsula samples were collected and incubated in the framework of the ESF CryoCarb project, with support from the Swedish Research Council (VR support to Kuhry), the Austrian Science Fund (FWF; grant no. I370-B17 to Richter), the Czech Science Foundation (project no. 16-18453S to Barta) and the Czech Soil & Water Research Infrastructure (MEYS CZ; grant nos. LM2015075 and EF16-013/0001782 to Šantrůčková). Seida samples were originally collected in the framework of the EU FP6 CARBO-North project (contract no. 036993). Gustaf Hugelius acknowledges a Swedish Research Council Marie Skłodowska Curie International Career Grant. Kateřina Diaková is acknowledged for the collection of the soil inoculi in Seida. The Seida samples were subsequently incubated at the University of South Bohemia. We are most grateful to Nikolai Lashchinskiy (Siberian Branch of the Russian Academy of Sciences, Novosibirsk, Russia) and Nikolaos Lampiris, Juri Palmtag, Nathalie Pluchon, Justine Ramage, Matthias Siewert and Martin Wik (all of Stockholm University), for help in sample collection. We would also like to thank Magarethe Watzka (University of Vienna) for elemental analyses of soil samples. Zhanna Kuhrij is acknowledged for the preparation of Figs. 5 and S6. We thank two anonymous reviewers for their constructive comments on the paper.
This research has been supported by the EU FP7 PAGE21 (grant no. 282700), the Swedish Research Council (no. 90735701), the Austrian Science Fund (no. FWF I370-B17), the Czech Science Foundation (no. 16-18453S), the Czech Soil & Water Research Infrastructure grants (MEYS CZ; nos. LM2015075 and EF16-013/0001782), the EU FP6 Carbo-North grant (no. 036993), and the Swedish Research Council Marie Skłodowska Curie International Career grant (no. 330-2014-6417).The article processing charges for this open-access publication were covered by Stockholm University.
This paper was edited by Lutz Merbold and reviewed by two anonymous referees.