Introduction
Yedoma regions store large amounts of soil organic carbon (SOC) and are
highly vulnerable to permafrost thaw under climate warming (Zimov et al.,
2006; Strauss et al., 2013). Organic C, freeze-locked for millennia in these
permafrost deposits, may become available for increased microbial activity
and be released in the form of the greenhouse gases CO2 and CH4 after
thaw (Gruber et al., 2004; Tarnocai, 2006; Dutta et al., 2006; Schuur et al.,
2008, 2015). Carbon-rich yedoma regions therefore have the potential to
contribute significantly to greenhouse gas emissions in a rapidly warming
Arctic (Schneider von Deimling et al., 2015).
The term yedoma describes late Pleistocene, ice-rich, silty deposits covering
vast areas in the northern permafrost regions that were not glaciated during
the last ice age (Schirrmeister et al., 2011a, 2013). These deposits can
reach a thickness of up to 50 m, store large amounts of C and are still
present today in an area of approximately 625 000 km2 in the Siberian
and North American permafrost region (Strauss et al., 2017). Since deglacial
warming, large areas of the former yedoma surfaces, expanding across several
million km2 of northern and eastern Siberia, Alaska, north-western Canada, and
the adjacent shelf regions, were eroded by thermokarst processes. These
processes affected the landscape by permafrost thawing and ice wedge melting
leading to surface lowering, lake formation, and lake drainage. Resulting
landscape features include thermokarst mounds (also called baydzherakhs),
thermo-erosional gullies, thermokarst lakes, and thermokarst depressions
(drained thaw lake basins or alas) which now are dominant features in the
yedoma terrain (Jorgenson and Shur, 2007; Veremeeva and Gubin, 2009; Kokelj
and Jorgenson, 2013; Morgenstern et al., 2013; Ulrich et al., 2014).
Ice-rich yedoma and thermokarst deposits are prone to thaw after disturbances
that change the surface thermal regime (e.g. from changing air temperatures,
fire, or anthropogenic disturbances). Since yedoma as well as thermokarst
deposits contain large, potentially vulnerable C pools in permafrost regions
(Zimov et al., 2006; Strauss et al., 2013; Walter Anthony et al., 2014) they
are of global importance for SOC accumulation, degradation, and release.
Whereas degrading yedoma surfaces and thermokarst features lead to
thaw-induced remobilization of organic C, the drainage of thermokarst lakes
and the following permafrost aggradation lead to SOC accumulation and
stabilization (Hinkel et al., 2003; Jorgenson and Shur, 2007; Jones et al.,
2012; Grosse et al., 2013). Olefeldt et al. (2016) estimated that landscapes
potentially vulnerable to thermokarst contain up to 50 % of the total SOC
stored in the permafrost region. In the yedoma region, 60 % of the C is
stored in drained and refrozen thermokarst basins (Strauss et al., 2013).
However, these estimates are based on a few data points from the yedoma region
and it remains important to enhance our understanding of the C storage,
distribution, and vulnerability in these environments with detailed
landscape-scale assessments.
Several permafrost C pool estimates have been carried out in different areas
within the Arctic region (e.g. Michaelson et al., 1996; Kuhry et al., 2002;
Ping et al., 2008; Tarnocai et al., 2009; Horwath Burnham and Sletten, 2010;
Hugelius et al., 2010, 2011; Fuchs et al., 2015; Siewert et al., 2015; 2016;
Palmtag et al., 2015). A synthesis for the circum-Arctic by Hugelius et
al. (2014) merged a large number of data sets into the Northern Circum-polar
Soil C Database (NCSCD) resulting in an estimate of 1035 ± 150 Pg of
organic C for 0–3 m for soils in the permafrost region. Several additional
estimates for SOC storage in the yedoma region exist, some of which take into
account slightly different components of the yedoma region. Zimov et
al. (2006) calculated 450 Pg C for the C pool for the entire yedoma region
(including both yedoma and thermokarst deposits). Walter Anthony et
al. (2014) estimated the total Holocene and Pleistocene soil C pools of the
yedoma region with 429 ± 101 Pg C, while Hugelius et al. (2014)
calculated 181 ± 54 Pg C for all deposits in the yedoma region below
3 m depth and Strauss et al. (2013) calculated 211 + 160/-153 Pg C
for the entire yedoma deposits including the top 3 m. Despite the variation
in these estimates they all suggest a very large C pool of several
hundred Pg for this region and confirm that these ice-rich deep deposits are
a globally important C pool in the northern circumpolar permafrost region.
Detailed local studies for particular parts of the yedoma region are scarce
so far but suggest significant landscape-scale and interregional variation in
SOC stocks that warrant further local studies and regional syntheses
(Schirrmeister et al., 2011b, c; Strauss et al., 2012; Siewert et al., 2015,
2016; Shmelev et al., 2017; Webb et al., 2017).
However, the yedoma and thermokarst landscapes not only host important C
pools but potentially store a significant amount of N. Even though there are
several permafrost soil C studies, only very few report N stocks (Ping et
al., 2011; Harden et al., 2012; Michaelson et al., 2013; Zubrzycki et al.,
2013; Obu et al., 2017; Palmtag et al., 2016). In tundra environments, N
often is the limiting factor for plant growth (Shaver et al., 1986; Chapin et
al., 1995; Mack et al., 2004; Beermann et al., 2015). As a result, thawing
permafrost does not only have the potential to release large amounts of
organic C but can also increase the availability of N which may increase
primary production and partly offset increased greenhouse gas emissions from
permafrost soils (e.g. Keuper et al., 2012, 2017; Natali et al., 2012; Salmon
et al., 2016).
A potential indicator of the potential C loss upon thaw and the
decomposability of C stored in permafrost soils is the carbon to nitrogen (C / N) ratio
(Schädel et al., 2014). Generally, a higher
C / N ratio suggests less degraded organic C, while a lower C / N
ratio points towards already partly degraded C before incorporation into
permafrost (Weiss et al., 2016). Even though the C / N ratio cannot be
taken as a single variable to estimate the recalcitrance of the soil organic
matter, it provides a first indication of the potential degradability of C
stored in permafrost soils (Kuhry and Vitt, 1996; Hugelius and Kuhry, 2009;
Hugelius et al., 2012).
In general, ice-rich landscapes are underrepresented in pan-Arctic
permafrost C pool inventories and the variability of these environments is
poorly accounted for. In particular, the deep C and N stocks below 1 m depth
remain associated with high uncertainties due to low sample numbers. In
addition, thermokarst areas are temporally dynamic and heterogeneous
environments with characteristics that are difficult to generalize and
characterize. A wide range of local case studies are needed to capture
spatial variability and to improve C pool estimations. This study presents
new data on C and N storage in two yedoma regions in the north of eastern
Siberia and analyses the variability of C and N contents along landscape
gradients extending from yedoma uplands into adjacent drained thaw lake
basins (DTLBs) investigating whether C and N storage is significantly higher
in DTLBs than in yedoma soils as proposed in regional studies by Strauss et
al. (2013) or Walter Anthony et al. (2014).
In this study, sediment cores up to 3 m depth from typical
thermokarst-affected yedoma regions were analysed (1) to quantify the organic
C and N variability with depth and along geomorphologic gradients in
thermokarst-affected terrain; (2) to understand depositional chronologies as
well as C accumulation rates; and (3) in combination with a satellite-image-based terrain classification to quantify landscape-scale SOC and N
inventories for the study areas Bykovsky Peninsula and Sobo-Sise Island.
(a) Location of the two study areas in the Lena River delta
region, Sobo-Sise Island, and Bykovsky Peninsula (Landsat 5 satellite image,
acquisition date: 19 September 2009). (b) The Lena Delta region with
a digital elevation model (ESA DUE-Permafrost DEM, Santoro and Strozzi,
2012), and (c) overview map including the permafrost zonation in
Siberia (after Brown et al., 1997).
![]()
Material and methods
Study area
The two study areas are situated in the north of eastern Siberia in the Lena
River delta region (Fig. 1). The first study site, Sobo-Sise Island
(72∘29′ N, 128∘16′ E), is a yedoma remnant within the
eastern Lena Delta. These remnants are commonly referred to as the third
geomorphological river delta terrace (Schwamborn et al., 2002). According to
Morgenstern et al. (2011), Sobo-Sise Island accounts for almost 16 % of
the entire yedoma coverage in the Lena Delta. It is characterized by yedoma
uplands but also features permafrost degradation landforms such as
thermokarst lakes, drained thaw lake basin, and thermo-erosional gullies.
Sobo-Sise is located between two of the main channels of the Lena Delta
(Sardakhskaya and Bykovskaya channels), elongated in east–west direction
(45 km) and has an area of 336 km2. Very few studies have been
conducted on the depositional characteristics of Sobo-Sise. For example,
Grigoriev (1993) dated a mammoth bone from a yedoma cliff on Sobo-Sise and
Grigoriev (2007) investigated the shore erosion on Sobo-Sise's coastline.
Morgenstern et al. (2011) as well as Nitze and Grosse (2016) included
Sobo-Sise in regional remote sensing studies aiming at the quantification of
thermokarst lakes and basins and surface landscape changes, respectively.
The second study site, Bykovsky Peninsula (71∘51′ N,
129∘19′ E), is similar to Sobo-Sise and also dominated by yedoma
uplands that are intersected by thermokarst lakes, drained thermokarst lake
basins and thermo-erosional gullies. The Bykovsky Peninsula is located to the
south-east of the Lena River delta at the Laptev Sea coast and represents an
erosional remnant of a late Pleistocene yedoma accumulation plain blanketing
the foreland of low mountain ridges (Grosse et al., 2007). The peninsula is a
narrow tongue of land with a width between 1 and 4 km and an area of
167 km2. In contrast to Sobo-Sise Island in the Lena Delta, the
Bykovsky Peninsula is strongly affected by rapid coastal erosion from all
sides and by coastal inundation, manifested in several thermokarst lagoons.
Studies on the Bykovsky Peninsula focused mostly on palaeoenvironmental
reconstructions using late Quaternary deposits at the exposure Mamontovy
Khayata (Kunitsky, 1989; Slagoda 1991, 1993; Grigoriev, 1993; Siegert et al.,
2002; Meyer et al., 2002; Schirrmeister et al., 2002a). Kholodov et
al. (2006) described organic matter characteristics in Alas and yedoma
deposits, Grigoriev et al. (1996) and Lantuit et al. (2011) determined
coastal erosion rates on the coasts of Bykovsky Peninsula, and Grosse et
al. (2005) mapped the periglacial geomorphology based on remote sensing data.
The climate of Sobo-Sise Island and Bykovksy Peninsula is continental,
despite their proximity to the coast. Both sites are dominated by polar
tundra climate (Peel et al., 2007). The mean annual air temperature (MAAT;
period 1998–2011) on Samoylov Island (the closest weather station to
Sobo-Sise, around 50 km to the west) is -12.5 ∘C with an
approximate mean annual precipitation of 180–200 mm (including mean summer
rainfall and mean snow water equivalent, period 1998–2011) (Boike et al.,
2013). On Bykovsky, the climate is similar with a MAAT of -12.9 ∘C
(period 1933–2013) at the closest weather station in Tiksi (Günther et
al., 2015), but a higher mean annual precipitation with around 240–260 mm
(Grigoriev, 1993). The higher precipitation can be explained by the influence
of the Laptev Sea and the mountain ridge nearby in the hinterland (Boike et
al., 2013). At both study sites, permafrost is continuous and is assumed to
have a thickness of about 500–650 m (Griogriev, 1993). Boike et al. (2013)
reported a mean annual permafrost temperature at Samoylov Island to be
-8.6 ∘C in 10.7 m depth.
At both study sites, the tundra vegetation is sparse and is limited by the
short growing season (June to September). The vegetation composition is
dominated by sedges, grasses, mosses, lichen, and sporadic dwarf shrubs
(Salix species). According to Schneider et al. (2009) most of
Sobo-Sise Island is classified as moist grass- and moss-dominated tundra and
wet sedge- and moss-dominated tundra. In the Circum Arctic Vegetation Map
(Walker et al., 2005), large parts of the Lena River delta including
Sobo-Sise Island are classified as sedge, moss, dwarf shrub wetland and
Bykovksy Peninsula is classified as non-tussock sedge, dwarf-shrub, moss
tundra.
Following the USDA Soil Taxonomy (Soil Survey Staff, 2014) the soil types at
both study sites belong to the Gelisol order with its suborders Turbels and
Orthels. Only two soil profiles sampled in thermokarst deposits in this
study contain thicker organic layers and can be considered as Histels.
Permafrost occurs at almost all sampled sites within the first metre. Active
layer thickness ranges from a minimum of 18 cm in thermokarst deposits to a
maximum of 84 cm in sandy fluvial deposits with a mean of all sampled sites
of 40 cm (median 41 cm). Only at two sites in non-vegetated, sandy,
temporally flooded Lena River floodplain deposits of Sobo-Sise, permafrost
was not present near the surface and these soils can be classified as
Entisols.
Field work
Soil samples were collected in August 2014 along four transects of 500
(BYK14-T3) and 600 m (BYK14-T2, SOB14-T1, SOB14-T2) length. We chose
transects to cover key geomorphologic gradients of the thermokarst-affected
landscapes starting with yedoma uplands and passing through different stages of
yedoma degradation (including yedoma slopes) and stages of the thermokarst
lake cycle. Similar transect-based approaches were used in other regions of
Siberia (Siewert et al., 2015; Palmtag et al., 2015), Canada (Hugelius et
al., 2010), and Alaska (Jorgenson, 2000; Kanevskiy et al., 2014). We
determined the starting points as well as the direction of each transect. To
include randomness, the sampling points along the transects were then chosen
at equidistant intervals in alignment with the scale of landscape features at
a site. A 150 m distance between sampling points was chosen on the first transect (SOB14-T1) and a 100 m distance on the second transect
(SOB14-T2). The first
transect on Sobo-Sise extended from a yedoma upland into the adjacent DTLB.
The second transect extended from a yedoma upland, crossing a DTLB and ending
in fluvial deposits of the floodplain. Additional non-random sample points
were collected on a baydzherakh (erosional remnants of polygon centres
forming thermokarst hills) and two DTLBs. No cores were taken from extant
thermokarst lakes or thermo-erosional gullies. On Bykovsky Peninsula,
transects were chosen with the same approach with a 150 m distance on the
first transect (BYK14-T2) and 100 m on the second transect (BYK14-T3)
between the sampling points. BYK14-T2 runs from one DTLB over a yedoma
remnant covered with baydzherakhs into another adjacent DTLB. BYK14-T3
was again similar to the Sobo-Sise transects, running from the top of a yedoma upland
into an adjacent DTLB.
Along the transects, soil pits were excavated down to the bottom of the
active layer. A soil profile description was made and fixed-volume samples
were collected with a metal cylinder of known volume. After sampling the
active layer, permafrost cores and samples were collected with a SIPRE (Snow,
Ice and Permafrost Research Establishment) auger barrel drill (Jon's Machine
Shop, Fairbanks, USA) with a diameter of 7.62 cm (3′′). Total sampling
depths ranged from 45 to 318 cm. As a result, at 23 sites the whole soil
profile was sampled and characterized including active, transient, and
permafrost layers and at five additional sites only the active layer was
sampled. A permafrost core description was made and the cores were subsampled
in the field at 5 to 10 cm intervals depending on facies
horizons. The visual core description included cryostratigraphy according to
French and Shur (2010) as well as a description of the lithology and plant
macrofossils.
Laboratory analysis
In total 455 samples were processed for total carbon (TC), total nitrogen
(TN), and total organic carbon (TOC). Samples were freeze-dried, homogenized
by grinding prior to a measurement of 5–8 mg of the samples with a Vario EL
III elemental analyser for TC and TN. Both %C and N were measured in one
run. We measured two replicates of each sample where we accept a
< 5 % deviation for the two measurements. The sensitivity of the
elemental analyser is < 0.1 %. Afterwards, samples were measured
for TOC (15–100 mg, depending on the TC content) with a Vario Max C
elemental analyser. Again, we allow a deviation of < 5 % for the
double measurements. Total inorganic carbon (TIC) is then calculated as the
difference between TC and TOC. In addition, the C / N ratio was
calculated as quotient between TOC and TN contents.
Based on Hugelius et al. (2010), the (SOC) storage was calculated for each
sample using Eq. (1):
SOC[kgm-2]=TOC[%]×BD[gcm-3]×(1-CF)×length[cm]×10.
Here, TOC is the total organic C content derived from the elemental analysis,
BD is the dry bulk density of the sample, CF is the coarse fragment fraction
(∅ > 2 mm) (which was zero because of the absence
of coarse fragments in the sampled soils), and length is the actual sample
length. The sample-specific SOC contents were added up to the reference
depths of 0–30, 0–100, 0–200 cm. If the cores were not recovered
completely, missing core intervals or missing SOC contents
were interpolated between adjacent samples or from samples with the same
characteristics following the field notes. The ice content in a sample is
reflected in the dry bulk density and therefore included in the calculations.
The SOC contents for the different reference depths, however, do not include
ice wedge volumes. Ice wedge volume was included in a later step when scaling
site-specific SOC stocks to the landscape level. Likewise the N storage was
calculated for individual samples and for the different reference depths.
In addition, 25 subsamples from various depths of seven different sampling
sites were chosen for radiocarbon dating of organic macrofossils. All
samples were wet-sieved with a 2 µm sieve and plant macro remains
(mostly moss leaves or sedge stems) were hand-picked under a microscope. In a
few cases with insufficient macro remains, bulk samples were selected.
Samples were then submitted to the Radiocarbon Laboratory in Poznan, Poland,
where the samples were analysed and dated with the accelerated mass
spectrometer (AMS) dating method (Goslar et al., 2004). The obtained
radiocarbon ages were eventually calibrated with the Calib 7.1 software to
calibrated radiocarbon years before present (cal. a BP) (Stuiver and Reimer,
1993; Stuiver et al., 2017).
Landform classification and upscaling C and N pools
All geospatial analyses were performed in the ESRI ArcGIS 10.1 and ENVI 5.3
software. For both Sobo-Sise Island and Bykovsky Peninsula, multispectral
RapidEye Images (pixel resolution 5 m) in combination with high-resolution
digital elevation models (DEMs) were used to classify the landscape into the
dominant landscape features. For Sobo-Sise Island two scenes from the same
date (27 July 2014) were orthorectified and mosaicked together to cover the
entire island. To improve the data basis, the RapidEye mosaic from the
27 July 2014 was stacked with another RapidEye scene (acquisition:
30 June 2014) which covers the entire island. The same scenes (27 July and
30 June 2014 images) were also used for Bykovsky Peninsula, stacked together
with an additional scene acquired on 9 September 2014 since the first scene
does not cover the entire peninsula. This procedure ensures that additional
landform variability from the phenological stages of the tundra vegetation at
both study sites is captured entirely by at least two RapidEye scenes.
A maximum likelihood supervised classification relying on training areas was
used to classify the landscape into the predominant landscape features.
Identification of training areas has also been facilitated through
near-simultaneous acquisition of RapidEye imagery to our field work. Training
areas were chosen based on field notes and field knowledge. For Sobo-Sise
Island the different RapidEye images were stacked with a DEM (pixel
resolution 2 m) that had been derived from photogrammetric processing of
three overlapping GeoEye stereo image pairs (acquisition dates range from
27 July to 15 August 2014). For Bykovsky Peninsula, a DEM was derived from
overlapping WorldView-1 and WorldView-2 stereo image pairs (acquisition
period: 25 to 29 August 2015) and again combined with the 15-band deep
multispectral RapidEye image stack to run the maximum likelihood
classification. Adding the DEM allowed enhanced classification of yedoma
uplands, since plant communities on yedoma uplands cannot be entirely
distinguished from those in DTLBs based only on spectral signatures. This
advantage was already demonstrated by Grosse et al. (2006) and Siewert et
al. (2016), who showed that, by including a DEM, non-degraded yedoma uplands and
partly degraded yedoma uplands could be better discriminated compared to
image classification only.
The two landform classifications for Sobo-Sise and Bykovsky Peninsula
initially included the main geomorphological units yedoma upland,
partly degraded yedoma (yedoma slope), and DTLBs of
different generations. However, due to the small amount of collected sampling
sites in DTLBs of different generations, the landform classes of DTLBs were
merged to a single-class thermokarst for the upscaling. The final
landform classification used for upscaling included the following classes:
yedoma uplands, degraded yedoma/yedoma slope,
thermokarst, and lakes. The areas of lakes were excluded
from upscaling since this study focuses on terrestrial soil C storage and no
lake cores were collected.
The landform classification accuracy was based on field-based ground truthing
points complemented with data points extracted from high-resolution imagery.
The high-resolution imagery include a GeoEye (Sobo-Sise Island), WorldView2
(Bykovsky Peninsula) and aerial photography to assess the correctness of
classification and overlap with the sampled field sites. For each study area,
300 randomly selected points and an additional 20 points for each landform
class were manually classified and then compared with the landform
classification based on the RapidEye satellite imagery.
The total SOC and N storage for Sobo-Sise and Bykovsky Peninsula was based on
mean C stocks of the collected sampling sites for the reference depths 0–30,
0–100, and 0–200 cm of each class. The mean stocks were upscaled based on
the areal extent of the corresponding landform class. This approach allows a
first estimate of the potential C and N storage in the study areas for the
first 2 m of soil. Confidence intervals for the mean SOC and N
landscape stocks were calculated according to Hugelius (2012). However, these
confidence intervals do not include uncertainties evolving from the landform
classification-based upscaling.
Laboratory results of soil C, N, and ground ice characteristics for
the different geomorphological landform units on Sobo-Sise Island and
Bykovsky Peninsula. All values are mean values ± standard deviation.
Median values for all parameters are provided in Table S2.
TOC (%)
TIC (%)
TN (%)
C / N (-)
Volumetric ice
SOC (kg m-3)
N (kg m-3)
n
content*
(%)
Sobo-Sise
3.5 ± 3.8
0.2 ± 0.4
0.3 ± 0.1
10.9 ± 6.3
61.8 ± 14.8
19.5 ± 12.6
1.7 ± 0.9
279
Sobo-Sise yedoma upland
4.3 ± 4.2
0.3 ± 0.4
0.3 ± 0.1
12.2 ± 6.8
61.5 ± 15.0
25.7 ± 16.1
2.2 ± 1.1
85
Sobo-Sise yedoma slope
3.1 ± 1.8
0.1 ± 0.1
0.3 ± 0.1
10.7 ± 2.9
61.6 ± 16.8
19.6 ± 10.0
1.8 ± 0.7
33
Sobo-Sise thermokarst
3.6 ± 3.9
0.1 ± 0.4
0.3 ± 0.1
11.7 ± 5.7
63.1 ± 13.7
17.6 ± 9.4
1.6 ± 0.6
135
Sobo-Sise fluvial deposits
0.8 ± 0.7
0.0 ± 0.1
0.1 ± 0.0
10.8 ± 6.0
45.0 ± 9.4
10.0 ± 7.2
0.9 ± 0.5
26
Bykovsky
6.6 ± 7.4
0.7 ± 0.4
0.5 ± 0.3
11.9 ± 6.5
62.8 ± 16.4
28.9 ± 21.2
2.4 ± 1.2
176
Bykovsky yedoma upland
5.1 ± 4.3
0.6 ± 0.4
0.4 ± 0.2
10.8 ± 3.5
62.1 ± 13.6
28.1 ± 18.4
2.5 ± 1.2
80
Bykovsky thermokarst
7.9 ± 9.0
0.8 ± 1.0
0.5 ± 0.4
13.2 ± 7.8
66.4 ± 14.3
29.5 ± 23.3
2.2 ± 1.3
96
* Intrasedimentary ice (excluding wedge ice volume).
To avoid overestimating the C and N stocks in the upscaling, we accounted for the
ice wedge volume in the landscape carbon and nitrogen calculations. Ice wedge
contents for thermokarst deposits were adopted from the study of Ulrich et
al. (2014) as a mean value derived from the maximum ice wedge contents from
their three north-eastern Siberian study areas Ebe-Basyn-Sise, Cape Mamontov Klyk
and Buor Khaya Peninsula. This resulted in a mean ice wedge volume of
9 ± 3 % (standard deviation) for thermokarst deposits which is
similar to what Kanevskiy et al. (2013) found for drained lake basins in
Arctic Alaska with 8 %. For the estimation of yedoma upland ice wedge
volume, the GIS-based approach from Günther et al. (2015) was applied to
Sobo-Sise. We orthorectified the GeoEye image using our high-resolution DEM
in order to ensure consistent mapping in an image free of geometrical
distortions. We mapped more than 1500 baydzherakhs that appeared in clusters
on slopes around thermokarst lakes and along thermo-erosional valleys and
river banks to determine the spatial dimensions of ice wedge polygons
(Voronoi diagram). The largest possible circle within each polygon served as
a proxy for the sediment fraction of the polygon and was put in relation to
the remaining size of the polygon, representing the ice wedge fraction. This
resulted in a mean ice wedge volume of 40 ± 8 % (standard
deviation). For yedoma uplands on Bykovsky Peninsula an ice wedge volume of
44 % was applied, which is the mean macro ground ice content on the nearby
Muostakh Island (Günther et al., 2015). Both Muostakh Island and
Bykovsky Peninsula have formerly been connected with each other (Grigoriev,
1993) and are now separated by a 15 km wide sound. For the active layer, we
assumed no ice wedge volume for the calculations. Mean active layer depths
for the different classes were derived from the collected soil sites. Mean
active layer depths for the sampled sites are presented in the Supplement
Table S1.
Results
Sedimentological results
Table 1 presents the laboratory results, indicating that samples from the
Bykovsky Peninsula (6.6 % ± 7.4) have a higher TOC content than
samples from Sobo-Sise Island (3.5 % ± 3.8). Being differentiated into
the various landform types, samples from thermokarst on Sobo-Sise store less
TOC (3.6 % ± 3.9) than samples from yedoma uplands
(4.3 % ± 4.2) which contrasts with the Bykovsky site
(7.9 % ± 9.0 for thermokarst and 5.1 % ± 4.3 for yedoma
upland). However, the standard deviations and therefore the variance are
higher on the Bykovsky Peninsula thermokarst samples.
The total nitrogen (TN) contents show a similar pattern to the TOC values.
There is less TN in the samples from Sobo-Sise. There are, however, only very
small differences in TN between the different landform types for both study
areas. Whereas there is slightly more TN in yedoma upland samples than in
thermokarst samples on Sobo-Sise, there is less in yedoma uplands compared to
thermokarst on Bykovsky Peninsula. However, when looking at the median, the
sample series (yedoma upland on Sobo, yedoma upland on Bykovsky and
thermokarst on Bykovsky) tend to be similar (see Table S2 with the median
values).
Volumetric ice contents are very similar for all the sampled cores. For both
study sites as well as for yedoma and thermokarst the mean values range
between 61 and 67 %. The only exceptions are the samples from fluvial
sandy deposits on Sobo-Sise Island with a mean value of 45 % (see
Table 1). Active layer samples were excluded from this analysis.
C / N ratio for yedoma upland and thermokarst samples on
Sobo-Sise and Bykovsky Peninsula. Blue lines indicate the running mean for
the entire sample set (including active layer and permafrost layer samples).
The correlation (Corr.) (Pearson correlation) between C / N ratio and
depth indicates a decreasing C / N ratio with increasing depth (p<0.01).
![]()
Comparison of mean sampling site SOC storage on Sobo-Sise Island and
on Bykovsky Peninsula. Solid bars: yedoma sites, striped bars: thermokarst
sites. Black T-lines show the standard deviation, and number in brackets
indicates the number of sampled sites. Profiles shorter than 200 cm were
extrapolated to the next reference depth. When an ice wedge was reached at a
collected site, this was included in the extrapolation as well, assuming no
carbon for ice wedge layers. SOC data for this graph are presented in the
supplementary material, Tables S1 and S4.
![]()
C / N ratios also show a similar pattern across all classes, with mean
C / N ratios ranging from 10.7 to 13.2 for the different geomorphological
units. However when separating C / N ratios into active layer and
permafrost layer samples, active layer samples show a higher C / N ratio
at both study sites and in all classes. Mean C / N ratios for active
layer samples (samples from cores of same landform units combined) are 15.8
for Sobo-Sise yedoma uplands, 20.1 for Sobo-Sise thermokarst, 12.5 for
Bykovsky yedoma uplands, and 17.0 for Bykovsky thermokarst. For permafrost
samples mean C / N ratios are 11.1 for Sobo-Sise yedoma uplands, 10.3 for
Sobo-Sise thermokarst, 10.1 for Bykovsky yedoma upland and 11.7 for Bykovsky
thermokarst. In all cases, active layer samples have higher C / N ratios
and there is a moderate correlation of decreasing C / N ratio with
increasing depth for three classes, and one weak correlation (Bykovsky
thermokarst) of decreasing C / N ratio with increasing depth (Fig. 2).
Sampling site SOC and N stocks
The TOC and bulk density values were used to estimate the site SOC stocks
which were averaged for the different landform types. Mean sampling site SOC
stocks (excluding ice wedge volume) were higher for yedoma upland sites than
for thermokarst sites for the reference depths of 0–30 and 0–100 cm for
both study sites (Fig. 3). SOC storages for 0–100 cm for Sobo-Sise are
25.3 ± 8.0 kg C m-2 for yedoma upland and
19.2 ± 5.9 kg C m-2 for thermokarst sites. For fluvial deposits
only one profile down to 1 m depth was collected with an SOC stock of
11.2 kg C m-2. Also for 0–200 cm, more SOC is stored in yedoma upland
soils. The findings for Bykovsky Peninsula are similar with more C stored at
yedoma uplands sites in the first metre of soil than at thermokarst sites,
despite the fact that a higher mean TOC content was found in the thermokarst
samples. For 0–100 cm yedoma upland sites store
29.7 ± 12.9 kg C m-2 and thermokarst sites store
23.9 ± 9.7 kg C m-2. For 0–200 cm there is more organic C
stored in thermokarst than in yedoma upland soils. However, this estimation
is only based on one relatively C-rich core (BYK14-T3-3), since this is the
only core reaching a depth of 2 m for thermokarst on Bykovsky
Peninsula. Therefore the carbon estimation of thermokarst on Bykovsky
Peninsula for the soil interval 0–200 cm has to be interpreted carefully.
Mean sampling site N storage for the different geomorphic units in
both study areas. Solid bars: yedoma sites, striped bars: thermokarst sites.
Black T-lines show the standard deviation and number in brackets indicate the
number of sampled sites. Profiles shorter than 200 cm were extrapolated to
the next reference depth. When
reaching an ice wedge in a collected site, this was included in the
extrapolation as well, assuming no N for ice wedge layers. Soil N data for
this graph are presented in the supplementary material, Tables S1 and S5.
![]()
Landform classification of Sobo-Sise Island and Bykovsky Peninsula.
For upscaling, the three classes of DTLB generations were merged into one
single thermokarst class.
![]()
The mean sampling site soil N stock (excluding ice wedge volume) for yedoma
upland sites on Sobo-Sise Island is 2.3 ± 1.1 kg N m-2
(0–100 cm) and for thermokarst sites it is 1.4 ± 0.4 kg N m-2
(Fig. 4). The mean profiles of N storage for the first metre of soil on
Bykovsky Peninsula are 2.6 kg N m-2 ± 0.9 for yedoma upland
sites and 1.9 kg N m-2 ± 0.6 for thermokarst sites.
Upscaling: landscape SOC and N stocks
The landform classification (Fig. 5) shows that 43 and 51 % of the
landscape on Sobo-Sise and Bykovsky, respectively, are yedoma or partly
degraded yedoma. Thermokarst depressions cover approximately 43 % on
Sobo-Sise and 38 % on Bykovsky (excluding lakes and lagoons which cover
14 and 11 % of the landscapes, respectively).
Based on the landform classification and the sampling site C contents the
total C storage for the two study regions was calculated. In total, 5.81 Tg
of organic C are stored in the first metre of soil on Sobo-Sise
(288 km2) of which around 57 % are stored within the active layer.
These calculations include a landscape-wide ice wedge volume of 40 % for
yedoma and 9 % for thermokarst, which is always applied to deposits below the
active layer. This results in an average SOC storage for non-lake areas on
Sobo-Sise of 20.2 ± 2.9 (95 % confidence interval) kg C m-2
for 0–100 cm. For Bykovsky Peninsula (154 km2) the results are
similar. In the first metre of soil, 3.98 Tg of organic C are stored of
which 58 % are stored in the active layer, including an ice wedge volume
of 44 % for yedoma and 9 % for thermokarst. This results in a
landscape average of 25.9 ± 9.3 kg C m-2 (excluding lakes) for
0–100 cm.
Mean landscape soil organic C and N storages in
kg m-2 ± 95 % confidence interval.
Study site
SOC
SOC
SOC
SOC
N
N
N
N
active layer
0–30 cm
0–100 cm
0–200 cm
active layer
0–30 cm
0–100 cm
0–200 cm
Sobo-Sise Island
11.6 ± 1.94
9.0 ± 1.23
20.2 ± 2.91
31.3 ± 3.56
1.0 ± 0.23
0.7 ± 0.09
1.8 ± 0.16
3.0 ± 0.23
Bykovsky Peninsula
15.0 ± 3.10
10.8 ± 2.25
25.9 ± 9.33
48.4 ± 9.97
1.3 ± 0.21
0.9 ± 0.13
2.2 ± 0.45
4.0 ± 0.37
Besides the organic C, there is a considerable amount of N stored in the
soils of Sobo-Sise Island and Bykovsky Peninsula. About 0.53 Tg of N is
stored on Sobo-Sise, resulting in a mean N storage of
1.8 ± 0.2 kg N m-2 (0–100 cm, excluding lakes). On Bykovsky,
a total of 0.34 Tg of N is stored in the first metre of soil. This results
in a mean N stock of 2.2 ± 0.5 kg N m-2 for 0–100 cm. Mean N
and organic C storage for the reference depths and the two study areas are
summarized in Table 2 and the total landscape stocks are presented in
Table S3.
Radiocarbon dates for selected soil cores.
Sample ID
Depth
Lab no.
AMS 14C age
Calibrated age
Dated
Weight
Coordinates
(cm)
(yr BP)
(yr BP)*
material
(mg)
Latitude (∘)
Longitude (∘)
Yedoma upland
SOB14-T1-1-3
20–24
Poz-74518
970 ± 30
842 ± 46
Sedge stems
13
72.50442
128.03915
DTLB
SOB14-T1-5-2
10–11
Poz-74451
112.88 ± 0.32 pMC
modern
Moss leaves/stems
12
72.50964
128.03435
SOB14-T1-5-15
148–150
Poz-74452
4460 ± 35
5058 ± 91
Bulk organic
12
72.50964
128.03435
SOB14-T1-5-19
187–200
Poz-74454
6605 ± 30
7481 ± 44
Moss leaves/stems
10
72.50964
128.03435
Yedoma upland
SOB14-T2-2-7
55–65
Poz-74455
1420 ± 30
1329 ± 39
Bark of a twig
12
72.52853
127.97281
SOB14-T2-2-16
119–122
Poz-74519
3065 ± 35
3272 ± 92
Sedge stems
17
72.52853
127.97281
SOB14-T2-2-24
173–179
Poz-74538
6200 ± 50
7114 ± 137
Wood with bark
22
72.52853
127.97281
SOB14-T2-2-30
218–223
Poz-74522
8800 ± 50
9807 ± 154
Bulk organic
20
72.52853
127.97281
DTLB
SOB14-T2-5-2
5–6
Poz-74523
111.54 ± 0.33 pMC
modern
Moss leaves/stems
28
72.52852
127.98176
SOB14-T2-5-10
67–74
Poz-74524
135 ± 30
168 ± 110
Bulk organic
25
72.52852
127.98176
SOB14-T2-5-19
145–156
Poz-74525
350 ± 30
364 ± 49
Sedge stems
13
72.52852
127.98176
SOB14-T2-5-31
273–278
Poz-74526
4735 ± 40
5517 ± 70
Bulk organic
20
72.52852
127.98176
SOB14-T2-5-34
299–303
Poz-74857
970 ± 30
842 ± 46
Deciduous leaves
13
72.52852
127.98176
Baydzherakh
BYK14-T2-3-2b
14–16
Poz-74732
595 ± 30
615 ± 37
Sedge stems/leaves
15
71.86050
129.29276
BYK14-T2-3-4
40–45
Poz-74733
1155 ± 30
1109 ± 67
Sedge stems
21
71.86050
129.29276
BYK14-T2-3-8
68–75
Poz-74734
1670 ± 30
1576 ± 53
Sedge stems
40
71.86050
129.29276
BYK14-T2-3-19
159–167
Poz-74735
2715 ± 30
2812 ± 52
Sedge stems
31
71.86050
129.29276
BYK14-T2-3-20
178–179
Poz-74737
41 600 ± 1400
45 203 ± 2512
Large wood piece
118
71.86050
129.29276
DTLB
BYK14-T2-4-4
23–24
Poz-74738
600 ± 30
615 ± 38
Sedge stems
15
71.86143
129.29530
BYK14-T2-4-10
45–48
Poz-74739
1250 ± 30
1222 ± 51
Sedge stems
16
71.86143
129.29530
BYK14-T2-4-16
76–79
Poz-74740
1545 ± 30
1449 ± 78
Sedge stems
22
71.86143
129.29530
BYK14-T2-4-22
117–125
Poz-74741
8350 ± 50
9368 ± 119
Plant remains
60
71.86143
129.29530
Yedoma upland
BYK14-T3-6B-14
110–116
Poz-89712
12 990 ± 70
15 533 ± 256
Sedge stems
20
71.82236
129.31537
BYK14-T3-6B-18
142–148
Poz-89713
13 350 ± 70
16 048 ± 225
Bulk organic
16
71.82236
129.31537
BYK14-T3-6B-23
185–191
Poz-89714
14 770 ± 70
17 970 ± 215
Bulk organic
20
71.82236
129.31537
* Calibrated with Calib 7.1 software (Stuiver et al., 2017).
Relationship between cumulative SOC storage and age. Cal yr BP for
radiocarbon dated samples (blue dots) for each core. Lines indicate the best
fit correlation of the points excluding outliers (red dots).
![]()
Radiocarbon dates
In general, the radiocarbon dates (Table 3) indicate that organic matter in
the first 2 m (and at one thermokarst site down to 3 m) in
both study areas is predominantly of Holocene age. Only one yedoma upland
site (BYK14-T3-6B) clearly has late Pleistocene ages around 1 m depth.
The two other dated yedoma upland sites BYK14-T2-3 and SOB14-T2-2 indicate
the presence of a thick Holocene cover layer exceeding 1.67 and 2.23 m,
respectively. In general, only one age-depth inversion was found
(SOB14-T2-5). At all other sites, organic matter age increased with depth.
Sediment and organic carbon accumulation rates (OC accumulation
rate). Sediment accumulation rates are based on the depth of the sample and
the calibrated radiocarbon date. Organic carbon accumulation rates are based
on cumulative soil organic carbon (cumulative SOC) storage at a specific
depth and the calibrated radiocarbon date at the corresponding depth. Mean
sediment and organic carbon accumulation rates are calculated, always
referring to the soil surface (depth = 0 cm and cumulative SOC = 0 kg C m-2). Relative sediment and
organic carbon accumulation rates are calculated always referring to the
sample above a particular sample.
Sample
Age*
Cumulative SOC
Depth
Mean
Relative
Mean OC
Relative OC
(cal yr BP)
(kg C m-2)
(cm)
sediment
sediment
accumulation
accumulation
accumulation
accumulation
rate
rate
(mm yr-1)
(mm yr-1)
(g C m-2 yr-1)
(g C m-2 yr-1)
SOB14-T1-1-3
842
10.29
19.5–23.5
0.26
0.26
12.2
12.2
SOB14-T1-5-2
modern
1.21
10–11
n/a
n/a
n/a
n/a
SOB14-T1-5-15
5058
13.13
148–150
0.29
0.29
2.6
2.6
SOB14-T1-5-19
7481
20.34
187–200
0.26
0.18
2.7
3.0
SOB14-T2-2-7
1329
18.30
55–65
0.45
0.45
13.8
13.8
SOB14-T2-2-16
3272
32.47
119–122
0.37
0.31
9.9
7.3
SOB14-T2-2-24
7113.5
43.13
173–179
0.25
0.14
6.1
2.8
SOB14-T2-2-30
9807
50.31
218–223
0.22
0.17
5.1
2.7
SOB14-T2-5-2
modern
0.54
5–6
n/a
n/a
n/a
n/a
SOB14-T2-5-10
168
12.43
67–74
4.20
4.20
74.0
74.0
SOB14-T2-5-19
364
23.97
145–156
4.13
4.08
65.9
58.9
SOB14-T2-5-31
5517
38.78
273–278
0.50
0.24
7.0
2.9
SOB14-T2-5-34
842
41.83
298.5–302.5
3.57
3.14
49.7
37.4
BYK14-T2-3-2b
615
10.76
14–16
0.24
0.24
17.5
17.5
BYK14-T2-3-4
1109
26.07
40–45
0.38
0.56
23.5
31.0
BYK14-T2-3-8
1576
43.14
68–75
0.45
0.62
27.4
36.6
BYK14-T2-3-19
2812
66.24
159–167
0.58
0.74
23.6
18.7
BYK14-T2-3-20
45 203
71.34
178–179
0.04
0.004
1.6
0.1
BYK14-T2-4-4
615
10.87
23–24
0.38
0.38
17.7
17.7
BYK14-T2-4-10
1222
17.63
45–48
0.38
0.38
14.4
11.1
BYK14-T2-4-16
1449
29.10
76–79
0.53
1.37
20.1
50.5
BYK14-T2-4-22
9368
34.23
117–125
0.13
0.05
3.7
0.6
BYK14-T3-6B-14
15 533
34.69
110–116
0.07
0.07
2.2
2.2
BYK14-T3-6B-18
16 048
38.96
142–148
0.09
0.62
2.4
8.3
BYK14-T3-6B-23
17 970
42.89
185–191
0.10
0.22
2.4
2.0
* Radiocarbon dates were calibrated with the Calib 7.1 software (Stuiver
et al., 2017). n/a = not applicable.
In addition, cumulative SOC storage and radiocarbon dates were combined to
calculate organic C accumulation rates (Table 4). Figure 6 shows the
radiocarbon ages plotted against the cumulative SOC for each sampling site,
indicating the C accumulation rates. The plots show that the C accumulation
rate was fairly linear in all of the cores, especially when removing one
age-inversion from a core (SOB14-T2-5 with 5517 cal yr BP) and one outlier
(the exceptional old date from BYK14-T2-3 with 45 203 cal yr BP); however
the accumulation rates vary for both the two classes thermokarst and yedoma
upland soils as well as for single sampling sites. The highest mean SOC
accumulation rate is found at the thermokarst site SOB14-T2-5 with
49.7 g C m-2 yr-1 (and 300 cm of sediment accumulation), which
is almost 20 times higher than SOB14-T1-5, another thermokarst site from
Sobo-Sise which has a mean organic C accumulation rate of
2.7 g C m-2 yr-1 (and 200 cm of sediment accumulation). This is
also reflected in the different sediment accumulation rates (Table 4). This
high variability in sediment and C accumulation rates reveals that even
within a small area (the sites are located within 3 km) high heterogeneity
exists in soil forming and C accumulation processes.
Discussion
Site specific soil organic C and N stock characteristics
We found that particularly DTLBs contain less C than estimates
from other studies (Strauss et al., 2013; Walter Anthony et al., 2014).
However, Strauss et al. (2013) and Walter Anthony et al. (2014) also included
samples from greater soil depths and partially included sites from boreal
regions with higher net primary productivity, whereas our study focused
on the first 3 m of the soils in a high-latitude tundra region.
Strauss et al. (2013) did a C inventory for the entire yedoma region
resulting in a SOC storage of 10 + 17/-6 kg m-3 for yedoma and
31 + 23/-18 m-3 for thermokarst deposits.
Thermokarst sites on Sobo-Sise Island are especially more depleted in SOC and
store less N than the yedoma upland sites. In contrast to several previous
studies investigating drained thermokarst lake basin peat accumulation in
Alaska, (Bockheim et al., 2004; Hinkel et al., 2003; Jones et al., 2012) and
the Kolyma region in Siberia (Walter Anthony et al., 2014), it is clear that
the investigated DTLB soils in Sobo-Sise and Bykovsky do not show signs of
increased peat formation and contain only thin organic layers. Organic layer
depths of the studied DTLBs on Sobo-Sise and Bykovsky Peninsula are largely
less than 10 cm with only two sites having thicker organic-rich peaty layers
at the top. This indicates that the conditions for peat accumulation in these
DTLBs were not favourable. Also for yedoma upland soils, organic layers are
relatively shallow (< 10 cm). This is especially important when
considering that the organic layer which insulates the ice-rich yedoma
deposits from warming and thawing is only thin, rendering the yedoma in this
region vulnerable to active layer deepening and permafrost degradation. Due
to the lack of thick organic or peaty layers, most of the profiles were
classified as mineral-dominated Orthels or Turbels. This is consistent with
the argumentation in Hugelius et al. (2016) which emphasized that DTLBs do
not always contain peaty C-rich deposits (Histels).
A key reason for the rather low SOC content in DTLBs is the low primary
productivity of the study sites at ∼ 72∘ N latitude. In
addition, the sampled DTLBs represent only a fraction of all the basins in
the study areas and may not be representative of the full range of basin
ages. Previous studies from Alaska indicated that older basins contained
thicker organic layers than younger basins (Hinkel et al., 2003; Jones et
al., 2012). Other reasons may include the topographic gradient that impacts
how well the DTLBs are drained and whether these basins remain waterlogged
peat-forming landscapes or become dry environments that are not favourable for peat formation. Several of the
studied DTLBs were eroded by the Lena River (Sobo-Sise) or the sea (Bykovsky
Peninsula) and some have deeply incised drainage channels, all of which
caused rather strong drainage gradients and enhanced landscape drying.
Our mean landscape SOC stocks for Sobo-Sise Island and Bykovsky Peninsula
are, however, in the same range as previous studies in similar settings. For
example, Siewert et al. (2016) found a mean of 19.2 kg C m-2
(0–100 cm) in another yedoma-dominated landscape in the central Lena River
delta. Zubrzycki et al. (2013) investigated the SOC characteristics of the
Holocene river terrace and the active floodplain in the Lena River delta and
found mean SOC stocks of 29.5 and 13.6 kg C m-2, respectively. The C
storage of the active floodplain is therefore very similar to the
11.2 kg C m-2 for fluvial deposits in our study area. Ping et
al. (2011) also investigated C storage along the Alaska Beaufort Sea
coastline and found a normalized mean landscape storage of
38 kg C m-2. However, this number is based on profiles only and not on
a landscape-based upscaling. Shmelev et al. (2017) investigated sites in the
Kolyma yedoma region and found a C storage of
17.0 ± 51.1 kg C m-3 (note that it is kg C m-3) for the
Holocene cover layer, 16.2 ± 31.3 kg C m-3 for the Alas
(thermokarst) deposits and 14.0 ± 23.5 kg C m-3 for yedoma
deposits. These values, however, refer not only to the first 2 m of soil
covering Pleistocene Ice Complex deposits but to the upper 25 m of Yedoma
Ice Complex. Also, Webb et al. (2017) investigated deep (15 m) C stocks in a
larch-dominated yedoma area in the Kolyma River basin and found more organic
C at the Alas site than at the yedoma site. Based on a landscape upscaling,
Siewert et al. (2015) calculated the SOC storage for the Kytalik region, a
yedoma- and thermokarst-dominated tundra landscape in the Yana–Indigirka
Lowland, to 25.8 ± 9.9 kg C m-2 for the first metre of soils
which falls in the range spanned by the Sobo-Sise Island and Bykovsky
Peninsula calculations.
Aside from organic C, a significant amount of N is stored in the soils of
Bykovsky Peninsula and Sobo-Sise Island, almost twice as much as what has
been found in the study by Zubrzycki et al. (2013) for the Holocene river
terrace (1.2 kg m-2) and the active floodplain (0.9 kg m-2) of
the Lena Delta. Obu et al. (2017) reported higher N storages from the western
Canadian Arctic, where 3.4 kg N m-2 is stored in cryoturbated or
recently disturbed type soils or sediments and Michaelson et al. (2013)
calculated a mean N storage of 2.7 kg m-2 for Arctic Alaska pedons.
The normalized average N storage from the Alaskan Beaufort Sea coast is in
the same range with 1.9 kg N m-2 (Ping et al., 2011) and N storages
from two study sites on Taymir Peninsula are also in the same range, with 1.0
and 1.3 kg N m-2 (Palmtag et al., 2016).
Even though the N storages are an order of magnitude lower than the organic C
storages, a large amount of N is present in these soils. Since N is the
limiting factor for plant growth in Arctic environments (Shaver et al., 1986;
Chapin et al., 1995; Mack et al., 2004; Beermann et al., 2015), permafrost
thawing will affect the N stocks in the soils. The N could partially become
available to plants upon permafrost thawing. However, the role of N and
whether it can offset an increased organic carbon release through increased
plant growth needs further exploration. In a recent study, Keuper et
al. (2017) found that plant-available N from thawing permafrost is an
additional source of deep-rooting subarctic plants and can increase their
biomass production. Also, Salmon et al. (2016) reported that increased N from
thawed permafrost enhances plant growth and biomass, but that it might
not offset C from deep deposits. An increase in N has, however, the potential
to change the plant productivity and the species composition (Keuper et al.,
2012). Keuper et al. (2012) show that not only dissolved plant-available N
becomes available with permafrost thawing but also organically bound N can be
mineralized at faster rates in thawed near permafrost soils.
However, an increase in N availability might both stimulate vegetation growth
and increase microbial activity (Nowinski et al., 2008). Also, Koven et
al. (2015) report that near-surface N released from thawing permafrost might
reduce nutrient limitations, even though the same study shows limited
importance of the deep soil N to offset deep C release. In an expert
assessment Abbott et al. (2016) stated that even an increase in Arctic and
boreal biomass might not offset permafrost carbon release. Since we only
focused on N stocks, we cannot derive conclusions on potential plant-available N in the soils of Sobo-Sise Island and Bykovsky Peninsula and its
effect on the primary production or C release. Nevertheless, this first
estimation of total N in the soils of Sobo-Sise Island and Bykovsky Peninsula
will be relevant for future climate models.
Upscaling of C and N pools
The total C stocks of our study sites are in the range of other permafrost C
studies and confirm previous high C stock estimates from northern permafrost
regions (e.g. Hugelius et al., 2014). The landform classification proved to
be adequate for the upscaling. The overall accuracies for the classification
are 71.5 and 71.1 % for Sobo-Sise and Bykovsky, respectively. Additional
field data in a greater variety of subclasses and with more diversified
classification would further increase the precision of the upscaling.
Nevertheless, by including a high-resolution DEM, classifying remotely sensed
images into geomorphological landform types results in an accurate map of a
first estimation of SOC and N stocks on the landscape level. In this context,
of particular note is the areal fraction of yedoma uplands we found on
Sobo-Sise Island and the Bykovsky Peninsula of 43 and 51 %, respectively.
While this is within the range Morgenstern et al. (2011) described for the
third terrace of the Lena Delta, on a larger regional level yedoma coverage is
generally lower. For example Veremeeva and Glushkova (2016) calculated
16 % of yedoma coverage for the entire Kolyma Lowland. However, our
higher values are a combination of yedoma uplands and partly degraded yedoma
slopes. Excluding areas with slopes, 19 % of Sobo-Sise and 22 % of
Bykovsky Peninsula are covered by intact yedoma uplands not yet significantly
affected by thermokarst or erosion.
Potential C thaw out for different active layer deepening
scenarios.
Active layer
Active layer
Active layer
Active layer
Area
+ 10 cm
+ 20 cm
+ 50 cm
+ 100 cm
Sobo-Sise Island
0.45 Tg
0.80 Tg
1.62 Tg
3.40 Tg
287.7 km2
Bykovsky Peninsula
0.25 Tg
0.48 Tg
1.32 Tg
2.44 Tg
154.0 km2
Total
0.70 Tg
1.28 Tg
2.94 Tg
5.83 Tg
441.7 km2
For upscaling deep C and N pools, the determination of ice wedge volumes is
important. In our study we assessed ice wedge volume using a combination of
literature values and own values derived form a GIS-based analysis of high-resolution satellite data. This analysis was based on more than 1500 mapped
baydzherakhs and resulted in a mean ice wedge volume of 40 % for Sobo-Sise
Island with a standard deviation of ±8 %. Conducting additional
calculations with an ±8 % ice wedge volume for yedoma uplands and
±3 % for thermokarst areas would lead to a ±4 % higher
respective lower landscape SOC stock (±5 % for N) for 0–100 cm and
±7 % of SOC and total N for 0–200 cm for Sobo-Sise Island. While
these results are still in the same range, this analysis reveals the
importance of assessing ice wedge volumes correctly within a landscape for SOC
and N upscaling, especially for deep deposit calculations. For a more
detailed upscaling and to capture the entire variability of these
heterogeneous environments, additional and deeper soil cores are needed as
well as a more detailed estimation of the landscape ice wedge contents to
further improve SOC and soil N pool estimations.
Sediment and organic C accumulation rates
Most of the analysed soil C was of Holocene age. Even for yedoma upland
soils,
radiocarbon dates indicated a large number of Holocene ages. Hence, the
yedoma uplands appear to be blanketed by Holocene cover material sometimes
exceeding a thickness of 2 m, which needs to be accounted for in
carbon pool inventories, because these 2 m cover deposits cannot be
considered as yedoma. In DTLBs, the Holocene age of soil C fits well with the
findings from previous studies, suggesting that the accumulation of lacustrine
sediments, drainage of thermokarst lakes, and accumulation of soils and
organic layers in the basins occurred mostly during the Holocene
(Kaplina, 2009; Grosse et al., 2013; Walter Anthony et al., 2014). Based on
the radiocarbon dates and the cumulative SOC storages, the accumulation rates
for the soil cores were calculated. The mean (linear) C accumulation rate of
SOB14-T2-5 with 49.7 g C m-2 yr-1 is very high, even higher than
what has been found by Jones et al. (2012) for palaeo peat accumulation rates
(9–35.2 g C m-2 yr-1) in thermokarst basins on the Seward
Peninsula. Nevertheless, it has to be considered that this site location is
close to the Lena River with only approximately 5 m above river water level.
It is likely that this location is affected by spring flood events which can deposit
large amounts of sediments. On the other side, the site SOB14-T1-5 with
2.7 g C m-2 yr-1 has very low accumulation rates and is most
certainly not affected by the Lena River flood. For comparison, Kurganova et
al. (2014) find that modern C accumulation on arable land in Russia was on
average 9.6 g C m-2 yr-1 over a 20-year period after
abandonment. Hicks Pries et al. (2012) found a mean Holocene C accumulation
rate of 25.8 g C m-2 yr-1 for surface soils and
2.3 g C m-2 yr-1 for deep soils in subarctic tundra in central
Alaska. Bockheim et al. (2004) found a mean long-term accumulation rate
of 13 g C m-2 yr-1 for the last 5500 years in DTLBs in Arctic
Alaska, but referring to the organic layer only.
Sediment accumulation rates are all of the same order of magnitude, excluding
SOB14-T2-5, which might be affected by the Lena River, with mean accumulation
rates for the different thermokarst cores ranging between 0.13 and
0.26 mm yr-1 and mean accumulation rates for yedoma upland cores
between 0.10 and 0.57 mm yr-1 (Table 4). For comparison, Murton et
al. (2015) found sediment accumulation rates between 0.75 and
2.00 mm yr-1 for yedoma silt at Duvanny Yar in the Kolyma Lowland.
However, these rates refer to a time period between 38 700 and 23 600 years
BP. Also, Schirrmeister et al. (2002a, b) found a similar accumulation rate
for a yedoma deposit (Mamontovy Khayata) on Bykovsky Peninsula of about
0.75 mm yr-1 for the time period between 60 000 and 6000 years BP.
These rates are slightly higher than the rates calculated in our study. In
addition, all presented sediment accumulation rates (Table 4) will be lower
when taking into account the ice content within the deposits. Sediment
accumulation rates corrected for soil core ice content are presented in the
Table S6.
Characterizing soil organic carbon
The rather low C / N ratio in our study is common to all sampled soils.
Only individual samples showed higher C / N ratios and in general there
is a trend of decreasing C / N ratios with increasing soil depth. Even
though the permafrost organic matter is already partly degraded, these
finding suggests that organic matter in the top permafrost layer may be
remobilized and decomposed when thawed out as the result of fluctuation of
the permafrost table due to climate change. Also, Strauss et al. (2015) found
relatively small C / N ratios for yedoma and thermokarst samples (median
values below 8 and 10), although they looked at samples from deeper deposits.
The mean values from our study might be lower too when incorporating samples
from greater depths; this is indicated by the decreasing C / N ratio with
increasing depth. Higher C / N values in the upper metre of soil were
found by Zubrzycki et al. (2013) with mean values between 20 and 42 from the
Holocene river terrace and mean values between 13 and 21 for the active
floodplain level of the Lena Delta. This indicates fresher material in these
deposits compared to yedoma uplands and DTLB deposits.
The fate of organic carbon in thermokarst-affected yedoma in
Siberia
Permafrost soil layers beneath the active layer and below 1 m depth are
important for future C remobilization, because models suggest permafrost
degradation and thaw well beyond 1 m depth by end of the 21st century
(Lawrence et al., 2012; Koven et al., 2013). The cores and the high sample
resolution in this study provide detailed information on the C stored in the
soil beneath the active layer in the study areas which will be thawed first
by future warming. This study provides additional soil C and N data for
multiple cores deeper than 100 cm for thermokarst-affected yedoma
landscapes. Studies with such deep cores are rare and even the NCSCD
contains three times more profiles for the 0–100 cm (1778 profiles)
interval than for estimations exceeding 100 cm depth (Hugelius et al.,
2014).
Our upscaling suggests that the study sites contain significantly more C than
soils in temperate climate zones (e.g. Wiesmeier et al., 2012). Both study
areas could become sources of organic C and N if permafrost thaw continues in
a warming Arctic. An estimation based on the sampled cores and the landform
classification shows that with an overall deepening of the active layer of
10 cm 700 000 t C (1.6 kg C m-2) will thaw out in both study areas
combined (Table 5). A regional study of Siberian permafrost dynamics
(Sazonova et al., 2004) includes scenarios where the active layer deepens by
more than 100 cm in north-eastern Siberia at the end of the 21st century.
This would result in an additional pool of available SOC of 5 830 000 t
(13.2 kg C m-2) in the two study areas combined. In addition,
different forms of permafrost thaw other than active layer deepening would further
increase the amounts of SOC thawed and mobilized. For example, lake shore
erosion rapidly degrades permafrost around lakes and releases organic C to
the aquatic environment. Shore erosion not only affects the active layer
soils from the top but does affect deeper permafrost soil layers (Walter
Anthony et al., 2016). This vulnerability of large currently frozen C pools
to thaw highlights the importance of deep permafrost organic C to be
considered in future C cycle models.