Article https://doi.org/10.

1038/s41467-023-38907-6

Characteristics of methane emissions

from alpine thermokarst lakes on the
Tibetan Plateau
Received: 6 July 2022 Guibiao Yang1, Zhihu Zheng1,2, Benjamin W. Abbott 3, David Olefeldt 4
,
Christian Knoblauch 5, Yutong Song1,2, Luyao Kang1,2, Shuqi Qin1,
Accepted: 18 May 2023
Yunfeng Peng 1 & Yuanhe Yang 1,2

Check for updates Understanding methane (CH4) emission from thermokarst lakes is crucial for
1234567890():,;
1234567890():,;

predicting the impacts of abrupt thaw on the permafrost carbon-climate

feedback. However, observational evidence, especially from high-altitude
permafrost regions, is still scarce. Here, by combining field surveys, radio- and
stable-carbon isotopic analyses, and metagenomic sequencing, we present
multiple characteristics of CH4 emissions from 120 thermokarst lakes in 30
clusters along a 1100 km transect on the Tibetan Plateau. We find that ther-
mokarst lakes have high CH4 emissions during the ice-free period
(13.4 ± 1.5 mmol m−2 d−1; mean ± standard error) across this alpine permafrost
region. Ebullition constitutes 84% of CH4 emissions, which are fueled primarily
by young carbon decomposition through the hydrogenotrophic pathway. The
relative abundances of methanogenic genes correspond to the observed CH4
fluxes. Overall, multiple parameters obtained in this study provide bench-
marks for better predicting the strength of permafrost carbon-climate feed-
back in high-altitude permafrost regions.

Soils in high-latitude and high-altitude permafrost regions store more from high-altitude permafrost regions, hampering our ability to assess
than 50% of world’s soil organic carbon (SOC) in about 15% of the global the impact of abrupt thaw on the permafrost carbon cycle10,12,13.
land area1,2 and play an important role in the global carbon cycle. These Therefore, improved understanding of CH4 emissions from alpine
permafrost regions are experiencing higher rates of climate warming thermokarst lakes is crucial for predicting permafrost carbon-climate
compared to other parts of the world3. This rapid warming is triggering feedback in this climate-sensitive region.
widespread gradual and abrupt permafrost thaw4 and subsequent The Tibetan Plateau is the largest alpine permafrost region in the
release of carbon to the atmosphere in the form of carbon dioxide (CO2) world (Fig. 1a), accounting for approximately 75% of the total alpine
and methane (CH4), potentially acting as a strong positive feedback to permafrost area in the Northern Hemisphere14. Similar to high-latitude
climate warming5–7. Thermokarst lake formation, a common abrupt permafrost regions, this region has experienced fast climate warming
thaw process, occurs due to the melting of excess ground ice in areas of and extensive permafrost thaw3,15, which has triggered the widespread
permafrost degradation8. Such lakes are the most widespread feature expansion of thermokarst lakes (Fig. 1b, c) and other types of abrupt
due to abrupt permafrost thaw, and cover 7% of permafrost regions9. permafrost thaw16. The number of thermokarst lakes in this permafrost
Due to their anaerobic environment, thermokarst lakes can be hot spots region is estimated to be 161,300, with a total area of ∼2800 km2 17.
for CH4 emissions10–12, but in-situ measurements are sparse, especially Most of the lakes (~80%) are located in alpine grasslands which can be

1
State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, Beijing 100093, China. 2University of
Chinese Academy of Sciences, Beijing 100049, China. 3Department of Plant and Wildlife Sciences, Brigham Young University, Provo, UT 84602, USA.
4
Department of Renewable Resources, University of Alberta, Edmonton, Alberta T6G 2H1, Canada. 5Institute of Soil Science, University of Hamburg, 20146
Hamburg, Germany. e-mail: [email protected]

Nature Communications | (2023)14:3121 1

Article https://doi.org/10.1038/s41467-023-38907-6

a b c
Qilian

Budongquan
Madoi

Zadoi Permafrost
Sampling clusters
Nagqu 0 120 240 Sampling lakes Sampling locations
km

Step 1 Selecting sampling clusters: 30 clusters were Step 2 Sampling at each cluster: Four thermokarst lakes were selected to cover
located in three representative permafrost regions different lake sizes at each cluster. In each lake, 4 to 6 sampling locations were distributed
(Madoi, Budongquan-Nagqu-Zadoi and Qilian sections) from the shore to the center.

d 16 May.-15 Jun. e 16 Jun.-15 Jul. f 16 Jul.- 15 Aug. g 16 Aug.-15 Sep. h 16 Sep.-15 Oct.

Step 3 Repeated sampling at each cluster during the ice-free period: Each lake was sampled five times at monthly intervals during the ice-free period
from mid-May to mid-October of 2021 to explore seasonal variations of CH4 and CO2 fluxes.

Fig. 1 | The flow chart of the sampling campaign. Our field sampling consists of explore seasonal variation of CH4 or CO2 fluxes (d–h). In-situ CH4 and CO2 fluxes
the following three key steps. First, we choose 30 clusters of thermokarst lakes are determined using an opaque lightweight floating chamber equipped with a
along a 1100 km transect on the Tibetan Plateau (a). Second, multiple locations closed loop to a near-infrared laser CH4/CO2 analyzer (GLA231-GGA, ABB., Canada).
within multiple lakes are selected at each cluster to eliminate spatial heterogeneity. In (a), the permafrost map of the Northern Hemisphere is obtained from the
In particular, four thermokarst lakes with different sizes are selected at each cluster National Snow & Ice Data Center85. Spatial distribution of permafrost on the Tibetan
(b). In each lake, 4 to 6 sampling locations are distributed from the shore to center Plateau is derived from Zou et al.86. The ellipses indicate the three representative
(c), and at each location flux measurements are taken and averaged to estimate the permafrost regions in our study area, including the Madoi, Budongquan-Nagqu-
CH4 and CO2 flux from the lake. Finally, each lake is sampled five times at monthly Zadoi and Qilian sections. Photos are taken by G.Y.
intervals during the ice-free period from mid-May to mid-October of 2021 to

subdivided into alpine steppe, alpine meadow and swamp meadow17. standard error (SE) unless stated otherwise). Ebullition is the main
Given that this permafrost region stores a large amount of SOC pathway for CH4 release, contributing to 84% of CH4 emissions. Radio-
(15.3–46.2 Pg carbon in the top 3 m of soils; 1 Pg = 1015 g)18–20, perma- and stable-carbon analyses show that old carbon is not the dominant
frost thaw could facilitate the rapid microbial decomposition of source for CH4 production, with CH4 being derived mainly from the
organic matter, leading to substantial carbon emissions20,21. Addition- hydrogenotrophic pathway in most of the sampled lakes. The relative
ally, the unique environmental conditions, characterized by lower air abundances of methanogenic genes correspond to the in-situ CH4
pressure and oxygen concentration due to the high elevation15,22, could fluxes. These findings lay the groundwork for a comprehensive under-
be beneficial for CH4 production. Thermokarst lakes in this permafrost standing of CH4 emissions in high-altitude thermokarst lakes.
region are thus expected to behave as hot spots for CH4 emissions23,24.
However, compared with the Arctic permafrost region, our under- Results and discussion
standing of CH4 emissions from alpine thermokarst lakes is limited. CH4 emissions across sampling clusters
Specifically, relatively little is known about CH4 fluxes, the contribution Across the 30 sampled clusters, CH4 concentrations ranged from 107.1
of old carbon from thawing permafrost, methanogenic pathways (CO2 to 159.4 nmol L−1 with a mean of 136 ± 30 nmol L−1 (n = 30; Supplemen-
reduction versus acetate fermentation), and microbial characteristics tary Fig. 2a). CH4 was supersaturated relative to the local atmosphere in
(methanogenic functional genes and communities). These knowledge all the studied thermokarst lakes across the clusters with a mean value
gaps prevent accurate prediction of the magnitude of carbon-climate of 2921 ± 62% (ranging from 2393 to 3719%; n = 30). CH4 fluxes had high
feedback in this alpine permafrost region. spatial variability within the 30 clusters, with values ranging from 0.1 to
In this context, we conducted a large-scale sampling campaign 39.2 mmol m−2 d−1 (Fig. 2). The lowest values occurred in thermokarst
across 120 thermokarst lakes in 30 clusters (four lakes in each cluster: 4 lakes located in alpine steppe (8.7 ± 3.0 mmol m−2 d−1) and the highest in
lakes/cluster × 30 clusters) along a 1100 km transect on the Tibetan alpine meadow (16.1 ± 1.7 mmol m−2 d−1). The CH4 fluxes also displayed
Plateau (Fig. 1a, b). Each lake was sampled five times at monthly intervals temporal variability during the sampling period (Supplementary
during the ice-free period from mid-May to mid-October of 2021 Fig. 3a). The maximum monthly mean CH4 flux was observed during the
(Fig. 1d–h), with each campaign lasting ~25 days. We measured CH4 period of mid-June to mid-July (30.5 ± 4.9 mmol m−2 d−1), while the
fluxes to the atmosphere using a portable opaque dynamic chamber minimum occurred at the end of ice-free season (7.6 ± 1.5 mmol m−2 d−1),
(Supplementary Fig. 1). Our results show that the thermokarst lakes on possibly due to the low temperature.
the Tibetan Plateau are an important CH4 source (regional CH4 emis- The mean CH4 flux during the ice-free season was 13.4 ±
sion: 76.6 Gg (109 g) CH4 yr−1 with a mean flux of 13.4 ± 1.5 mmol m−2 d−1 1.5 mmol m−2 d−1. This value is at the high end of the range reported from
during the ice-free period; hereafter, values are reported as mean ± Arctic thermokarst water bodies regarded as hot spots for CH4

Nature Communications | (2023)14:3121 2

Article https://doi.org/10.1038/s41467-023-38907-6

0.1 10 20 30
40 and exhibited a large spatiotemporal variability (Fig. 4a). Throughout
CH4 fluxes (mmol m-2 d-1)
the ice-free season, CH4 ebullition fluxes were highest from mid-June
93 oE 96 oE 99 oE 102 oE
to mid-July with a mean value of 29.7 ± 4.7 mmol m−2 d−1 (n = 30; Sup-
plementary Fig. 3a). During the ice-free period, mean CH4 diffusion
across all the sampled clusters varied from 0.1 to 7.9 mmol m−2 d−1
while ebullition varied from 0.03 to 31.4 mmol m−2 d−1. The mean values
were 2.1 ± 0.3 mmol m−2 d−1 for diffusion and 11.2 ± 1.5 mmol m−2 d−1 for
ebullition (n = 30; Supplementary Fig. 3a). The maximum ebullition
flux (200.8 mmol m−2 d−1) was recorded at the shore of one of the
36 oN thermokarst lakes, where micro-bubbles were visible (Supplementary
Movie 1). The contribution of ebullition to lake CH4 fluxes showed no
significant difference among these three grassland types in which the
thermokarst lakes are mainly distributed (p = 0.07; Supplementary
Fig. 5a). Overall, alpine thermokarst lakes on the Tibetan Plateau were
of high ebullition fluxes which constituted ~84% of the total CH4 fluxes
33 oN (diffusion plus ebullition fluxes; Fig. 4a).
The high contribution of ebullition to the total CH4 flux might be
potentially explained by low atmospheric pressure on the Tibetan
Plateau. Due to the high elevation, atmospheric pressure across the
Permafrost Sampling sites 0 120 240 study area has a mean value of 60.9 kPa: roughly three-fifths of that at
km
90 oE 93 oE 96 oE 99 oE 102 oE sea level (Supplementary Table 1). Two ways in which this low atmo-
spheric pressure could result in a larger contribution of ebullition to
Fig. 2 | CH4 fluxes in alpine thermokarst lake on the Tibetan Plateau. Bubble size the total CH4 flux. On the one hand, according to the Henry’s law, the
is proportional to the value of the CH4 flux at each cluster, with a larger size lower atmospheric pressure causes lower CH4 solubility in the water29,
representing a higher value. The background permafrost maps of the Northern
which can be unfavorable for CH4 diffusion in the water column and
Hemisphere and the Tibetan Plateau are derived from the National Snow & Ice Data
thus impels CH4 to be transported from the lakebed to the atmosphere
Center85 and Zou et al.86, respectively. The inset shows the comparison of CH4 fluxes
in the form of bubbles. On the other hand, atmospheric pressure can
in thermokarst lakes located in the three grassland types. AS, AM and SM represent
alpine steppe, alpine meadow and swamp meadow, respectively. Box plots present
directly affect bubble formation. Bubbles containing CH4 occupy
the 25th and the 75th quartile (interquartile range), and whiskers indicate the data vertical tubes within the lake sediments30. The lower air pressure will
range among thermokarst lakes located in AS (n = 5), AM (n = 13) and SM (n = 12), benefit the vertical expansion of these bubbles and promote their
respectively. The notches are the medians with 95% confidence intervals. Observed escape from the lake sediment31, and is thus associated with greater
values are shown as black dots. Significant differences are denoted by different ebullition of CH432. Consistently, the proportion of ebullition to total
letters (one-way ANOVAs with two-sided Tukey’s HSD multiple compar- CH4 fluxes was negatively associated with air pressure (R2 = 0.42,
isons, p = 0.049). p < 0.001; Supplementary Fig. 6a) but positively correlated with ele-
vation across sampling clusters (R2 = 0.42, p < 0.001; Supplementary
Fig. 6b). In addition, the shallowness of alpine thermokarst lakes on the
release13,25. During our field sampling campaign, we also measured CO2 Tibetan Plateau (Supplementary Table 1) could contribute to the high
emissions (see Supplementary Note 1 for detailed information). Our relative ebullition because shallow water creates less hydrostatic
results showed that the contribution of CH4 to total carbon emissions pressure22, increasing the formation and release of gas bubbles from
(CH4 + CO2) from thermokarst lakes was 7.3% (Fig. 3a). If considered in lake sediments31,33.
terms of CO2-equivalent (CO2-e) emissions, mean CH4 flux during the
ice-free season was estimated to be 136.8 CO2-e mmol m−2 d−1 (a 28-fold Radiocarbon age of CH4 emissions
higher global warming potential relative to CO2 over 100 years)26, which To estimate the contribution of old carbon to CH4 emissions, we col-
was of the same order of magnitude as the CO2 emissions (170.4 mmol lected ebullition gas samples with floating plastic bubble traps and
CO2 m−2 d−1; Supplementary Fig. 4) and approximately 44.6% of total determined the CH4 radiocarbon age. The results showed relatively
CO2-e emissions (Fig. 3b). In conjunction with the total area of ther- young CH4 radiocarbon age, ranging from −360 to 3810 years before
mokarst lakes (~2800 km2)17, these findings demonstrate that alpine present (yrs BP; Fig. 4b). The mean CH4 radiocarbon age was only
thermokarst lakes are hot spots of CH4 emission, and also highlight the 325.8 yrs BP, and 46% of thermokarst lakes had modern (defined here
importance of CH4 fluxes in the total carbon emissions from alpine as created after 1950) age of CH4 emissions (Supplementary Table 2).
thermokarst lakes. The high CH4 fluxes can be attributed to two char- Only two lakes had higher CH4 radiocarbon ages than 1000 yrs BP,
acteristics of the Tibetan Plateau. First, atmospheric oxygen con- while radiocarbon ages were between modern and 1000 yrs BP in the
centration across our study area is low due to the high elevation remaining 22 samples. These observations indicated that old carbon
(between 3279 and 5014 m above sea level at our study sites), which can was not the dominant source for CH4 production in most thermokarst
cause low dissolved oxygen concentration in all water bodies, including lakes on the Tibetan Plateau. Potential explanations for the relatively
thermokarst lakes (4.3 ± 0.2 mg L−1; n = 30; Supplementary Table 1). Low low contribution of old carbon to CH4 fluxes observed in this study
dissolved oxygen concentration can stimulate CH4 production, and could be associated with the young permafrost carbon in this study
thus increase the CH4 flux27. Second, the thermokarst lakes across our region. It has been reported that permafrost on the Tibetan Plateau
study area are mostly shallow (depth range: ∼0.2-3.7 m; Supplementary forms relatively recently compared to other permafrost regions34,35,
Table 1)22. The shallow water column allows more rapid exchange with which may lead to the situation that the frozen carbon is also relatively
the atmosphere and less time for CH4 removal by microbial oxidation28, young. This deduction is supported by the measured average radio-
and is thus conducive to the release of CH4. carbon age of surface permafrost below the active layer at 24 sites
across the Tibetan Plateau (6100 ± 880 yrs BP; n = 24; Supplementary
CH4 diffusion and ebullition fluxes Fig. 7; see Supplementary Note 2 for details of radiocarbon age mea-
To isolate the main pathways of CH4 release, we quantified both dif- surements). Therefore, the relatively low contribution of old carbon to
fusion and ebullition fluxes. Ebullition occurred at all sampling clusters CH4 fluxes in alpine thermokarst lakes could be attributed to the young

Nature Communications | (2023)14:3121 3

Article https://doi.org/10.1038/s41467-023-38907-6

a
170.3 (92.7%)
Total C 0.005
0.05
CH4
0.04 13.4 (7.3%)
13.44 (7.3% )

0.004

Density

Density
0.03 0.003

0.02 0.002

0.01 0.001

0.00 0.000
0 100 200 300 400 500 600
Gas emission rates (mmol m-2 d-1)

170.3 (55.4%) 0.005

0.005 CH4
CO2 0.004
0.004

Density
Density

136.84 (44.6% )

136.8 (44.6%) 0.003

0.003

0.002 0.002

0.001 0.001

0.000 0.000
-100 0 100 200 300 400 500 600
CO2-equivalent emission rates (mmol m-2 d-1)

Fig. 3 | Contribution of CH4 to total carbon emissions from the investigated emissions. Mean CH4 flux during the ice-free season and its CO2-equivalent emis-
thermokarst lakes. Panels (a, b) represent the density of carbon fluxes and sions are shown outside parentheses, and the corresponding contribution to total
CO2-equivalent emmissions, respectively. The lines indicate the fluxes from four carbon emissions and CO2-equivalent emissions are presented in parentheses,
thermokarst lakes at each cluster during the measurement period. The pie charts respectively.
show the contribution of CH4 fluxes to total carbon emissions and CO2-equivalent

permafrost carbon in this study region. In addition, thermokarst lakes High αC values may be attributed to the alkaline and saline con-
on the Tibetan Plateau are characterized by the small surface area22, ditions across our study area (Supplementary Table 1). Specifically,
and small lakes have a high perimeter to surface area ratio, which high pH could stimulate the dissociation of acetic acid into its anion
potentially increase terrestrial loading of organic matter from sur- form (CH3COO−), which could then inhibit transmembrane diffusion
rounding plants and surface soils. This terrestrial organic matter is and prevent the transportation of acetate39. Therefore, despite the
dominated by modern carbon36, and can thus stimulate modern car- accumulation of acetic acid, acetoclastic methanogenesis is likely to be
bon emissions from thermokarst lakes37. The last but not the least, less energetically favorable than hydrogenotrophic methanogenesis
some of the studied lakes may have developed following thermokarst under alkaline conditions, leading to the high contribution of the CO2
processes a long time ago or even not be thermokarst lakes. This might reduction pathway to CH4 production40. Beside the alkaline condi-
also be part of the explanation of the modern age of emitted CH4. tions, saline environment-associated methanogenic substrates may be
another potential driver for high αC values. In saline environment,
CH4 production pathway methanol has been reported to be a methanogenic precursor and can
To estimate the relative contribution of two major pathways of serve as substrates for CH4 production41. Moreover, methanol-derived
methanogenesis (CO2 reduction and acetate fermentation), we mea- methanogenesis is usually accompanied by the highly depleted
sured the δ13C of CH4 and CO2 in bubble samples to calculate the δ13C–CH4 values42. Consequently, methanol-dependent methanogen-
apparent carbon fractionation factor (αC); an indicator of the CH4 esis may also be responsible for the high αC values from alpine ther-
production pathway (see “Methods”; αC > 1.055 indicates CO2 reduc- mokarst lakes on the Tibetan Plateau due to their saline environment
tion; αC < 1.04 indicates acetate fermentation)38. The δ13C–CH4 had a (ranging from 0.2 to 2.6 ppt; Supplementary Table 1).
mean value of −72.5 ± 1.1‰ (n = 29), and the δ13C–CO2 ranged between
−0.5‰ and −22.9‰ with a mean of −13.4‰ (Supplementary Table 2). Methanogenic microorganisms
Gas samples exhibited high αC values, ranging from 1.052 to 1.079 To evaluate the potential differences in methanogenic microorgan-
(1.064 ± 0.001, n = 29). Only four samples had αC values between 1.04 isms from thermokarst lakes in the three ecosystem types, we analyzed
and 1.055, while the rest had αC values higher than 1.055 (Fig. 4c), methanogenic functional genes and communities in surface sediment
indicating that CO2 reduction dominated CH4 production in alpine samples (0–15 cm) using metagenomic sequencing. Multiple func-
thermokarst lakes on the Tibetan Plateau. Furthermore, the αC values tional genes of methanogens were more abundant in the thermokarst
showed no significant difference among the three grassland types in lakes located in alpine meadow and swamp meadow (Supplementary
which thermokast lakes are mainly distributed (p = 0.52; Supplemen- Fig. 8). This result suggested that there were higher potentials for CH4
tary Fig. 5b). production in thermokarst lakes located in these two grassland types,

Nature Communications | (2023)14:3121 4

Article https://doi.org/10.1038/s41467-023-38907-6

which matched with higher CH4 fluxes observed in the field at these (Supplementary Fig. 8). This result was confirmed by the validated
locations (Fig. 2). The largest differences in relative abundance were mcrA gene predicted by contigs with length ≥ 1000 bp (Fig. 5a).
observed for the mcr gene, with mean value being 4.1-fold and 3.2-fold With regard to the sediment methanogenic community compo-
higher in thermokarst lakes distributed in alpine meadow and swamp sition of the thermokarst lakes, there were five dominant methano-
meadow than those located in alpine steppe, respectively genic classes: Methanomicrobia, Methanobacteria, Thermoplasmata,
Methanococci, and Methanopyri (Fig. 5b). The genus composition dia-
CH4 fluxes (mmol m-2 d-1) grams showed that Methanosarcina within Methanosarcinaceae order
0 10 20 30 40 was the most abundant methanogenic genus, accounting for about
a 65% of all methanogens. It has been reported that Methanosarcina has
0.3 a high volume-to-surface ratio with a large cell size and spherical form.
Ebullition flux
Density

0.2
Diffusion flux Together with the formation of clusters, this leads to low levels of
diffusion per unit cell mass43. Accordingly, Methanosarcina is relatively
0.1 tolerant of adverse environmental conditions compared to other
4000 methanogens44. Moreover, the Methanosarcina genus contains cyto-
0
b c chromes and methanophenazine (a functional menaquinone analog),
which enable methanogens to conserve energy via membrane-bound
Years before present

3000
-10
electron transport chains so as to maintain high growth yields45.

δ13C-CO2 (‰)
2000 Overall, Methanosarcina can be competitive at low temperature,
-20
alkaline, and saline conditions (Supplementary Table 1), potentially
1000 contributing to high CH4 emission rates from alpine thermokarst lakes
-30 on the Tibetan Plateau.
0
-40 Regional estimates
-90 -80 -70 -60 -50 To upscale our lake-level measurements to regional efflux estimates,
0. 0

0. 6
09
0. 3
0
0
00
00
00

δ13C-CH4 (‰) we conducted a Monte Carlo analysis to randomly sample thermokarst

Density
lake CH4 flux for each grassland type from a normal distribution
Fig. 4 | CH4 ebullition and diffusion fluxes, radiocarbon age and production around the mean. We then weighted the CH4 flux by the corresponding
pathway in alpine thermokarst lakes on the Tibetan Plateau. In (a), the corre- area of thermokarst lakes in each grassland type to calculate the
sponding values of light green and dark green lines represent the ebullition and regional flux. Total CH4 emissions were estimated to be 76.6 Gg CH4
diffusion CH4 fluxes across 30 clusters, respectively. Line shows the mean CH4 flux yr−1. The contribution of CH4 to total carbon emissions was approxi-
for the four thermokarst lakes at each cluster. In (b), line indicates the radiocarbon mately 9.2% (Table 1), which was 1.4–8.4-fold greater than the mean
age of surface permafrost below the active layer at each sampling site (n = 24). contribution from lakes globally28,38–40. The 100-year global warming
In (c), the light blue and light orange lines are the apparent carbon fractionation
potential of CH4 was 2144.7 Gg CO2-e yr−1, which was approximately
factor (αC) values of 1.04 and 1.055. The αC values indicate the pathway of CH4
equivalent to lake CO2 emissions estimated in the same way (2084.7 Gg
production, with αC > 1.055 suggesting that CH4 is mainly produced by CO2
CO2 yr−1; Table 1). CH4 emissions from thermokarst lakes are often
reduction (hydrogenotrophic methanogenesis, HM), and αC < 1.055 suggesting that
CH4 is produced increasingly by acetate fermentation (acetoclastic methanogen-
ignored when evaluating the regional CH4 balance across Tibetan
esis, AM). 14C and δ13C isotopic signatures were measured in the bubble gas from alpine grasslands17. However, our results illustrated that CH4 emissions
only 24 and 29 lakes respectively, due to the limited volume of gas samples that from thermokarst lakes could offset 15.3% of the CH4 uptake from
could be collected in the field. alpine steppe and meadow which cover ~95% of alpine grasslands on

Fig. 5 | Methanogenic micoorganisms of thermokarst lakes on the Tibetan dots. Significant differences are denoted by different letters (one-way ANOVAs with
Plateau. Panel (a) respresents the differences in the key functional gene of mcrA two-sided Tukey’s HSD multiple comparisons, p = 0.038). Panel (b) shows metha-
among thermokarst lakes located in alpine steppe (AS), alpine meadow (AM), and nogenic taxonomic infromation. The colors in the inner ring represent the different
swamp meadow (SM). The relative abundance of mcrA gene was predicted by taxa. The triangles in the first ring indicate relative abundance with ≥ 5 per 10,000
contigs with length ≥ 1000 bp. Box plots present the 25th and the 75th quartile (upper triangle) or <5 per 10,000 (lower triangle). The mean relative abundances
(interquartile range), and whiskers indicate the data range among thermokarst for all samples are shown in the second ring and pillars, where color depth and
lakes located in AS (n = 5), AM (n = 13) and SM (n = 12), respectively. The notches are height are proportional to the cubic root of relative abundance.
the medians with 95% confidence intervals. Observed values are shown as black

Nature Communications | (2023)14:3121 5

Article https://doi.org/10.1038/s41467-023-38907-6

Table 1 | Regional carbon and CO2-equivalent emissions (CH4 may increase substantially. Moreover, this study evaluated the
and CO2) from alpine thermokarst lakes on the Tibetan landscape-level radiocarbon ages of thermokarst lake CH4 emissions in
Plateau this alpine permafrost region. Our results illustrated that old carbon
was not the dominant source for CH4 production from alpine ther-
Target variables CH4 emissions CO2 emissions
mokarst lakes on the Tibetan Plateau. This finding is in contrast to
Carbon emissions (Gg C yr−1) 57.4 (0.8–114.1) 568.6 (17.5–1119.7)
studies from the high-latitude permafrost region where old carbon
CO2-equivalent emissions 2144.7 (28.9–4260.5) 2084.7 (205.8–4105.5)
contributes significantly to CH4 emissions upon permafrost thaw51.
(Gg CO2-e yr−1)
Hence, when modeling permafrost carbon-climate feedback across
5th–95th percentiles are presented in parentheses.
our study area, more efforts should be put into accurately predicting
the effect of permafrost thaw on the spatial extent of thermokarst
lakes, rather than on the rate of anaerobic decomposition of previously
the Tibetan Plateau46,47. Further, thermokarst lake CH4 emissions were frozen soil. Furthermore, our results demonstrated that methanogenic
equivalent to ~50–150% of the CH4 emissions from swamp meadow functional genes corresponded with the in-situ CH4 fluxes, suggesting
which occupies ~5% of Tibetan alpine grasslands18,47. The incorporation that methanogenesis might be the potential driver of CH4 emissions
of CO2 fluxes gives an estimate of the overall carbon emissions from alpine thermokarst lakes. Overall, multiple parameters observed
(CH4 + CO2) in thermokarst lakes of 4.2 Tg (1012 g) CO2-e yr−1 (Table 1), in this study can function as benchmarks for better predicting per-
which could offset 11.9% of net carbon sink in Tibetan alpine grasslands mafrost carbon-climate feedback.
(37.1 Tg CO2-e yr−1)48. Taken together, these results demonstrate that
any assessment of the carbon budget in this climate-sensitive region is Methods
incomplete without considering the significant carbon emissions from Study area
thermokarst lakes. The Tibetan Plateau, the largest alpine permafrost region in the world,
Although this study advances our understanding of CH4 release has a mean elevation of over 4000 m above sea level. Discontinuous
from thermokarst lakes on the Tibetan Plateau, it does have some and sporadic permafrost are widely distributed across the region52.
limitations. First, carbon emissions during the ice-cover period were Approximately 1.06 × 106 km2 of this region is underlain by permafrost,
not considered in this study. To make a rough evaluation of ice-cover accounting for 40% of the overall plateau14. The permafrost mainly
carbon emission, we applied the average percent contribution of ice- formed in the late Pleistocene during the Last Glaciation Maximum and
cover to annual carbon emissions from high-latitude lakes (CO2: 17%; the Neoglaciation period35. The current active layer thickness across
CH4: 27%)49. These increased annual emissions estimate from alpine the plateau is ~1.9 m, but this is deepening at a rate of ~1.3 cm yr−1 based
thermokarst lakes on the Tibetan Plateau to 427.0 Gg CO2 yr−1 and 28.3 on observations from 1981 to 201053,54. The climate is characterized as
Gg CH4 yr−1, respectively. Nevertheless, given the potential differences cold and dry14. Mean annual temperature ranges from −2.9 to 7.0 °C
between high-latitude and high-altitude permafrost regions, such as and mean annual precipitation varies from 129 to 590 mm, with
climate, thermokarst lake depth and ice-cover duration34, the annual approximately 90% of the precipitation falling during the growing
contribution of carbon emissions during the ice-cover period season (late May to early October). The arid and semiarid climate has
observed from high-latitude lakes may not be simply applied to ther- suppressed the development of ground-ice, leading to a relatively low
mokarst lakes in high-altitude permafrost region. More measurements ice content with a percentage by weight of 12.2% in this permafrost
during the ice-cover period are thus needed to further elucidate the region34,55. The dominant vegetation types include alpine steppe,
role of thermokarst lakes in the regional carbon budget. Second, alpine meadow, and swamp meadow, with the corresponding domi-
uncertainties exist in the area of the thermokarst lakes used in the nant species being Stipa purpurea and Carex moorcroftii, Kobresia
regional carbon budget. Particularly, newly formed thermokarst lakes pygmaea and K. humilis, and K. tibetica, respectively. The plateau has
and thermokarst lakes located in desert and barren land were not experienced rapid climate warming, causing the formation and
considered, but they do have potential to contribute to CH4 expansion of thermokarst lakes in various vegetation types (Supple-
emissions50. Meanwhile, theomarkast lake map used in this study suf- mentary Fig. 2c–e)56. The distribution of thermokarst lakes is domi-
fers from uncertainties because it is based on low resolution and single nated by lakes with small surface area (<10,000 m2) and shallow water
satellite images and without consideration of ground-ice content17. It is which account for ~80% of the total number17. The mean ice-free
thus essential to incorporate multi-satellite, higher spatial resolution duration of lakes across the study area is around 200 days57.
remote sensing data (e.g. GF-2 and Planetscope) and ground-ice con-
tent to re-evaluate thermokarst lake area and its temporal dynamics in Flux and environmental measurements
the future. Based on the updated lake distribution and the expansion In this study, we selected 30 clusters of thermokarst lakes for carbon
rate, an improved estimation by considering thermokarst lakes dis- flux measurements along a 1100 km transect on the Tibetan Plateau.
tributed in other ecosystems and fresh thermokarst lakes is needed to The 30 clusters were evenly located in three representative permafrost
obtain more accurate prediction of carbon release from alpine ther- regions (10 sites in the Madoi section on the eastern plateau, 15 sites in
mokarst lakes. Third, the development of thermokarst lake is not taken the Budongquan-Nagqu-Zadoi section, and 5 sites in the Qilian section
into account in this study. The studied lakes may cover the various on the northeastern plateau in the central part of the plateau; Fig. 1a).
thermokarst development stages or even non-thermokarst lakes, At each cluster, four thermokarst lakes were selected to cover different
which can result in the uncertainty of the regional carbon emissions. lake sizes. A total of 120 thermokarst lakes were thus sampled (30
Therefore, additional attention should be paid for the development of clusters × 4 lakes/cluster). In each lake, 4–6 sampling locations were
thermokarst lakes to further advance our understanding of CH4 distributed from the shore to the center of the lake if the size allowed,
emissions in alpine thermokarst lakes on the Tibetan Plateau. and flux measurements were taken at each location and then averaged
In summary, this study offers an insight into the spatial patterns, to estimate CH4 or CO2 fluxes from the respective lakes. This sampling
sources and microbial characteristics of CH4 emissions from alpine at multiple locations within multiple lakes allowed us to consider the
thermokarst lakes on the Tibetan Plateau. Our results indicated that spatial variability of carbon fluxes in each cluster. Each thermokarst
thermokarst lakes were an important but under-quantified CH4 source lake was sampled five times during the ice-free period from mid-May to
in the Tibetan alpine permafrost region. Considering that the expan- mid-October of 2021 (once a month) to explore the seasonal variations
sion of thermokarst lake landscapes will accelerate under future cli- of the CH4 and CO2 fluxes (120 lakes × 5 times). In-situ total CH4 and
mate warming, their potential contribution to the regional CH4 budget CO2 fluxes were determined between 9:00 a.m. and 18:00 p.m. using

Nature Communications | (2023)14:3121 6

Article https://doi.org/10.1038/s41467-023-38907-6

an opaque lightweight floating chamber (diameter: 26 cm, and height: speeds are >3.6 m s−1 or 2/3 if wind speeds are <3.6 m s−1 33. Given that
25 cm; Supplementary Fig. 1) equipped with a closed loop to a near- CO2 ebullition is negligible due to its high solubility, CO2 is exclusively
infrared laser CH4/CO2 analyzer (GLA231-GGA, ABB., Canada; Preci- diffusive59. Due to this point, the CO2 transfer coefficient was calcu-
sion: <0.9 ppb CH4 (1 s) and 0.35 ppm CO2 (1 s); Measurement rates: lated from Eq. 3.
0.01 to 10 Hz.)33,58. Specifically, the floating chamber was flushed with Ebullition CH4 flux was determined from the difference between
ambient air for ~10 s before each measurement to ensure ambient CH4 the measured total CH4 flux and the calculated diffusion CH4 flux. We
and CO2 concentrations inside the chamber. The chamber was then would like to point out that, although the two-layer model is popularly
pulled into its sampling location with a rope to avoid the need to enter used to evaluate CH4 diffusion flux from Arctic thermokarst lakes11,28
the lake and potentially disturb sediment and gas release. CH4 and CO2 and alpine rivers33, uncertainties could be introduced due to the
concentrations in the chamber were continuously recorded for 150 s at adopted theoretical gas transfer k coefficient60. Future studies should
an interval of 1 sec after an equilibration period (20 s) to eliminate any thus attempt to quantify CH4 ebullition and diffusion fluxes from
disturbance of the surface boundary layer induced by chamber alpine thermokarst lakes using other approaches (e.g. bubble traps61).
deployment. At each sampling location, measurements were made for During flux measurement at each cluster (n = 30), we quantified
170 sec. This measurement time (170 s) was adopted to avoid CO2 wind speed and atmospheric temperature with a portable anem-
accumulation in the floating chamber which could inhibit CO2 emis- ometer (Testo 480, Testo SE & Co. KGaA, Lenzkirch, Germany). The air
sion. The 20 s equilibration period was selected based on the time pressure, water temperature, oxidation-reduction potentiality, pH and
needed for CO2 concentration changes in the chamber to become dissolved oxygen concentration were measured in each thermokarst
linear. The 1 s interval was chosen to capture the change of gas con- lake (n = 120) with a portable multiparameter water quality instrument
centration over time within the chamber in real time. Total CH4 and (ProSolo Digital Water Quality Meter, Yellow Springs Instrument,
CO2 fluxes were calculated using the following Eq. 133,58: Brannum Lane, USA). Atmospheric CH4 and CO2 concentrations were
also recorded with the CH4/CO2 analyzer (GLA231-GGA, ABB., Canada)
nt
n0 as the steady values obtained when being flushed through with
F total = ð1Þ
A×t ambient air. All parameters were measured five times—once a month
from mid-May to mid-October, 2021.
where Ftotal is the total CH4 or CO2 flux (mmol m−2 d−1), nt and n0
represent the number of moles of CH4 or CO2 in the chamber at time C and δ13C isotopic analyses
14

170 and 20 s after chamber deployment (mmol), respectively, A is the To evaluate the CH4 radiocarbon age and the pathway of CH4 pro-
base area of the chamber (m2), t is the recording time during flux duction, we quantified the 14C-CH4, δ13C-CH4, and δ13C-CO2 isotopic
measurement (s). ratios. Due to the high cost of isotopic analyses, only one thermokarst
Diffusion and ebullition CH4 fluxes were calculated using a two- lake at each cluster was randomly sampled for 14C and δ13C measure-
layer model to estimate their relative contributions to the total CH4 ments. In addition, due to the limited gas samples, 14C and δ13C isotopic
release. Specifically, during flux measurements, 50 ml of surface water signatures were measured for only 24 and 29 lakes, respectively.
from a depth of 0–10 cm was collected with a 100 ml syringe (1:1 ratio Specifically, a combined bubble sample was collected using 4 sub-
of air-water). Subsequently, 50 ml of pure N2 was injected into the merged plastic traps (diameter 0.7 m) placed along the transect from
syringe to create 50 ml of headspace. The syringe was immediately the shore to the center of each of the selected lakes during mid-July to
shaken for 5 min to equilibrate the headspace in the field. The head- mid-August, 202162. Bubble gas from the traps was collected for about
space sample was then injected into a vacuumed airtight vial and two weeks to enable sufficient volume to accumulate, and then divided
transported to the laboratory for analysis of CH4 and CO2 concentra- into two parts. The first part was injected into 1 L pre-evacuated airtight
tions using a gas chromatograph (Agilent 7890 A, Agilent Technolo- gas-sampling aluminum bags (Dalian Delin Gas Packing Co., Ltd, China)
gies Inc., Santa Clara, Canada). Dissolved CH4 and CO2 concentrations for the determination of radiocarbon isotopic composition63. In par-
were calculated using Henry’s Law adjusted for temperature and ticular, CO2 and H2O in the sample were removed using two traps
atmospheric pressure58. The degree of CH4 and CO2 saturation (S) was which were filled with ethanol-liquid nitrogen and Alkali lime-
calculated by comparing the dissolved CH4 and CO2 concentration Magnesium perchlorate, respectively. CH4 was then combusted with
(CW) with the dissolved CH4 and CO2 concentration at equilibrium with copper oxide to produce CO2 and H2O at 950 °C. Prior to this, the
the local atmosphere corrected for changes in solubility (Ceq) copper oxide was charged with oxygen at 600 °C overnight. Following
according to Eq. 2:33,58 combustion, water was removed through a trap filled with Alkali lime-
Magnesium perchlorate, and the pure CO2 was then locked by a liquid
S = C w =C eq ð2Þ nitrogen trap. Finally, the samples were quantified and catalytically
reduced to graphite (containing ~1 mg C), and the 14C/12C isotopic ratio
Diffusion CH4 flux across the water surface into the chamber was was measured by accelerator mass spectrometry (0.5MV 1.5SDH-1,
estimated from Eq. 3:33,58 NEC, USA) at Third Institute of Oceanography, Ministry of Natural
Resources, Xiamen, China. The measured 14C values were corrected for
F d = kðC w
C eq Þ ð3Þ mass-dependent fractionation by being normalized to a fixed δ13C
value level64 and reported as conventional radiocarbon ages (years
where Fd is the diffusion CH4 flux (mmol m−2 d−1), k is the CH4 transfer before present, yrs BP; where 0 yrs BP = AD 1950; Supplementary
coefficient (m d−1), Cw is the dissolved CH4 concentration (mmol m−3), Table 2).
and Ceq is the CH4 concentration in water (mmol m−3) at equilibrium The second part of the gas sample was stored in 20 ml glass
with the atmosphere in the field corrected for changes in solubility bottles for determining the stable-carbon isotopic composition of CH4
according to the Henry’s law29. The CH4 transfer coefficient (k) was and CO238. Briefly, CH4 in the sample was purified through a trap filled
estimated from Eq. 4:33,58 with liquid nitrogen and then combusted to CO2 and H2O in an oxi-
dation oven. After removing H2O, δ13C–CH4 was measured with an
k = k CO2 ðScCH4 =ScCO2 Þ
n ð4Þ isotopic ratio mass spectrometry (IRMS 20-22, SerCon, Crewe, UK) at
Institute of Botany, Chinese Academy of Sciences, Beijing, China.
where kCO2 is the CO2 transfer coefficient, ScCH4 and ScCO2 are the δ13C–CO2 measurements were made using a similar procedure, but
Schmidt number of CH4 and CO2, respectively, n is 1/2 when wind without the purification and combustion. The apparent fractionation

Nature Communications | (2023)14:3121 7

Article https://doi.org/10.1038/s41467-023-38907-6

factor (αC) was calculated from δ13C of CH4 and CO2 (‰, relative to total CH4 fluxes, αC, the relative abundances of methanogenic func-
Vienna PDB) with Eq. 5:38 tional genes and community from thermokarst lakes located in the
three grassland types (alpine steppe, alpine meadow and swamp
αC = ðδ 13 C
CO2 + 1000Þ=ðδ 13 C
CH4 + 1000Þ ð5Þ meadow). Tukey’s HSD difference test was used for multiple compar-
ison at a significance level of α = 0.05. p value corrections were per-
The αC value indicates the dominant pathway of methanogenesis; formed for multiple comparisons using the Benjamini–Hochberg
small αC values (1.04–1.055) are associated with acetate fermentation, correction factor80. Regression analyses were performed to examine
while large αC values (1.055–1.09) are caused by CO2 reduction38. the relationship of the ebullition proportion to total CH4 fluxes with
elevation and air pressure. Before applying regression analyses, we
Metagenomic sequencing, functional annotation and taxo- conducted outlier analysis for the proportion of ebullition to total CH4
nomic analysis fluxes based on Boxplot Procedures81. Log transformation was con-
To explore the potential role of the microbial community and func- ducted when the continuous variables and their residuals violated
tional genes in mediating CH4 emissions from thermokarst lakes, we assumptions of normality.
collected lake sediment samples (0–15 cm) at equally spaced intervals We upscaled CH4 and CO2 fluxes from lake levels to the regional
along the transect from the shore to the center of each lake during scale using Monte Carlo analysis that ran 1000 iterations for each of
mid-July to mid-August, 2021. The sediment samples were passed the grassland types (including alpine steppe, alpine meadow and
through a 2 mm sieve with visible roots being removed. Due to the high swamp meadow) where thermokarst lakes are mainly located. Each
experimental cost, only one lake at each of the 30 clusters was sampled iteration randomly resampled a CH4 or CO2 flux (for one of the three
for metagenomic analysis. Lake sediment samples were immediately grassland types) based on a normal distribution surrounding the mean
transported to the laboratory and stored at −20 °C until the metage- and standard deviation. Then, to generate the total thermokarst lake
nomic analysis. Total DNA was extracted from thawed sediment sam- CH4 or CO2 flux per unit of time for each grassland type, we multiplied
ples (0.4 g) using the DNeasy PowerSoil kit (Qiagen, Hilden, Germany) the randomly resampled CH4 or CO2 flux values by the area of ther-
according to the manufacturers’ instructions. The ratios of spectro- mokarst lakes for each grassland type which was determined by the
metry absorbance for the extracted DNA were between 1.7 and 1.9 at distribution of thermokarst lakes on the Tibetan Plateau17 and the
260/280 nm and between 1.7 and 2.1 at 260/230 nm. vegetation map was derived from China’s Vegetation Atlas82. Subse-
The metagenomic sequencing was carried out using the DNBSEQ- quently, we multiplied the total CH4 or CO2 flux per unit of time by the
T7 platform (BGI, Shenzhen, China) to obtain 2 × 150 bp paired-end ice-free season duration (~200 d) to obtain the CH4 or CO2 emissions
reads. Raw reads and adapters were first removed using SOAPnuke from each grassland type. Finally, we summed CH4 or CO2 emissions
software (-n 0.001 -l 20 -q 0.4 --adaMR 0.25 --adaMis 3 --outQualSys 1) from each grassland type to estimate total CH4 or CO2 emissions. All
to generate trimmed reads consisting of approximately 15–18 Gb of statistical analyses were performed using R statistical software v3.6.2
sequencing data for each sample. Clean reads for each sample were (http://r-project.org)83.
assembled using Megahit (v.1.2.9)65, and contigs with length > 300 bp
were retained. Then, open reading frames (ORF) were predicted by Reporting summary
Prodigal (v.2.6.3)66, and dereplicated with 95% nucleotide identity Further information on research design is available in the Nature
using CD-HIT-EST (v.4.8.1)67. The translated proteins of non-redundant Portfolio Reporting Summary linked to this article.
gene clusters were annotated by searching against the EggNOG (v.5.0)
database68 using DIAMOND (v.2.0.9)69. The final annotation table was Data availability
obtained from the KEGG database70 which is derived from EggNOG. To All data supporting the findings are available in the Figshare data
estimate the relative abundance of representative non-redundant repository (https://doi.org/10.6084/m9.figshare.22743968)84 and
genes in each sample, Salmon (v.1.5.1)71 was used to calculate the TPM Supplementary Information. The nucleotide sequences generated by
(transcripts per million) of each predicted gene via mapping to clean metagenome sequencing have been deposited in the NCBI database
paired reads of each sample. Taxanomic annotation was evaluated (ncbi.nlm.nih.gov/bioproject/?term=PRJNA942440).
using Kraken2 (v.2.1.2)72 against the mini database (downloaded on
Feb.−2022), and then visualized using GraPhlAn (v.0.9.7)73. References
To further verify the annotations of genes involved in CH4 pro- 1. Gruber, S. Derivation and analysis of a high-resolution estimate of
duction, assembled contigs with length ≥1000 bp were used for global permafrost zonation. Cryosphere 6, 221–233 (2012).
identifying the methyl-CoM reductase alpha subunit (mcrA). Specifi- 2. Hugelius, G. et al. Estimated stocks of circumpolar permafrost
cally, contigs with length ≥1000 bp were predicted using Prodigal carbon with quantified uncertainty ranges and identified data gaps.
(v.2.6.3) and then screened using HMM profiles (PF02249 and Biogeosciences 11, 6573–6593 (2014).
PF02745) derived from the Pfam database74 with hmmsearch (HMMER 3. Biskaborn, B. K. et al. Permafrost is warming at a global scale. Nat.
v.3.3.2)75. Putative mcrA genes were validated by phylogenetic tree with Commun. 10, 264 (2019).
reference sequences derived from Annotree76 based on HMM profiles 4. Brown, J. & Romanovsky, V. E. Report from the international per-
(PF02249 and PF02745 from Pfam) and Zhou et al.77. Putative mcrA mafrost association: state of permafrost in the first decade of the 21st
proteins were aligned with reference sequences using MUSCLE century. Permafr. Periglac. 19, 255–260 (2008).
(v.3.8.31)78 and built for a phylogenetic tree using FastTree (v.2.1.11)79 5. Schuur, E. A. G. et al. Vulnerability of permafrost carbon to climate
with the WAG + GAMMA models. Validated mcrA genes were mapped change: implications for the global carbon cycle. Bioscience 58,
to clean paired reads of each sample by Salmon (v.1.5.1) to obtain 701–714 (2008).
counts, lengths and effective lengths. Counts per gene were normal- 6. Schaefer, K. et al. The impact of the permafrost carbon feedback on
ized to reads per kilobase per million mapped reads (RPKM). global climate. Environ. Res. Lett. 9, 085003 (2014).
7. Schuur, E. A. G. et al. Climate change and the permafrost carbon
Statistical analyses feedback. Nature 520, 171–179 (2015).
We conducted a series of statistical analyses to explore the basic 8. Kokelj, S. V. & Jorgenson, M. T. Advances in thermokarst research.
characteristics of CH4 emissions from alpine thermokarst lakes. Spe- Permafr. Periglac. 24, 108–119 (2013).
cifically, one-way ANOVAs were carried out to evaluate the differences 9. Olefeldt, D. et al. Circumpolar distribution and carbon storage of
in total carbon emissions (CH4 + CO2), the contribution of ebullition to thermokarst landscapes. Nat. Commun. 7, 13043 (2016).

Nature Communications | (2023)14:3121 8

Article https://doi.org/10.1038/s41467-023-38907-6

10. Turetsky, M. R. et al. Carbon release through abrupt permafrost 34. Wang, X. et al. Contrasting characteristics, changes, and linkages of
thaw. Nat. Geosci. 13, 138–143 (2020). permafrost between the Arctic and the Third Pole. Earth Sci. Rev.
11. Serikova, S. et al. High carbon emissions from thermokarst lakes of 230, 104042 (2022).
Western Siberia. Nat. Commun. 10, 1552 (2019). 35. Jin, H. J. et al. Evolution of permafrost on the Qinghai-Xizang (Tibet)
12. Walter Anthony, K. M. et al. 21st-century modeled permafrost car- Plateau since the end of the late Pleistocene. J. Geophys. Res. 112,
bon emissions accelerated by abrupt thaw beneath lakes. Nat. F02S09 (2007).
Commun. 9, 3262 (2018). 36. Dean, J. F. et al. Siberian Arctic inland waters emit mostly con-
13. Wik, M. et al. Climate-sensitive northern lakes and ponds are critical temporary carbon. Nat. Commun. 11, 1627 (2020).
components of methane release. Nat. Geosci. 9, 99–105 (2016). 37. Elder, C. D. et al. Greenhouse gas emissions from diverse Arctic
14. Wang, B. & French, H. M. Permafrost on the Tibet Plateau, China. Alaskan lakes are dominated by young carbon. Nat. Clim. Chang. 8,
Quat. Sci. Rev. 14, 255–274 (1995). 166–171 (2018).
15. Wang, B. et al. Tibetan Plateau warming and precipitation changes 38. Walter Anthony, K. M. et al. Methane production and bubble
in East Asia. Geophys. Res. Lett. 35, L14702 (2008). emissions from arctic lakes: Isotopic implications for source path-
16. Li, L. et al. Thermokarst lake changes over the past 40 years in the ways and ages. J. Geophys. Res. 113, G00A08 (2008).
Qinghai–Tibet Plateau, China. Front. Environ. Sci. 10, 1051086 (2022). 39. Welte, C. et al. Experimental evidence of an acetate transporter
17. Wei, Z. et al. Sentinel-based inventory of thermokarst lakes and protein and characterization of acetate activation in aceticlastic
ponds across permafrost landscapes on the Qinghai-Tibet Plateau. methanogenesis of Methanosarcina mazei. FEMS Microbiol. Lett.
Earth Space Sci. 8, e2021EA001950 (2021). 359, 147–153 (2014).
18. Ding, J. et al. The permafrost carbon inventory on the Tibetan Pla- 40. Wormald, R. M. et al. Hydrogenotrophic methanogenesis under
teau: a new evaluation using deep sediment cores. Glob. Chang. alkaline conditions. Front. Microbiol. 11, 614227 (2020).
Biol. 22, 2688–2701 (2016). 41. Conrad, R. Importance of hydrogenotrophic, aceticlastic and
19. Wang, D. et al. A 1 km resolution soil organic carbon dataset for methylotrophic methanogenesis for methane production in ter-
frozen ground in the Third Pole. Earth Syst. Sci. Data 13, restrial, aquatic and other anoxic environments: a mini review.
3453–3465 (2021). Pedosphere 30, 25–39 (2020).
20. Wang, T. et al. Permafrost thawing puts the frozen carbon at risk 42. Conrad, R. Quantification of methanogenic pathways using stable
over the Tibetan Plateau. Sci. Adv. 6, eaaz3513 (2020). carbon isotopic signatures: a review and a proposal. Org. Geochem.
21. Cheng, F. et al. Alpine permafrost could account for a quarter of 36, 739–752 (2005).
thawed carbon based on Plio-Pleistocene paleoclimate analogue. 43. Calli, B. et al. Methanogenic diversity in anaerobic bioreactors
Nat. Commun. 13, 1329 (2022). under extremely high ammonia levels. Enzym. Microb. Technol. 37,
22. Niu, F. et al. Characteristics of thermokarst lakes and their influence 448–455 (2005).
on permafrost in Qinghai–Tibet Plateau. Geomorphology 132, 44. De Vrieze, J. et al. Methanosarcina: the rediscovered methanogen
222–233 (2011). for heavy duty biomethanation. Bioresour. Technol. 112, 1–9 (2012).
23. Mu, C. C. et al. High carbon emissions from thermokarst lakes and 45. Thauer, R. K. et al. Methanogenic archaea: ecologically relevant
their determinants in the Tibet Plateau. Glob. Chang. Biol. 29, differences in energy conservation. Nat. Rev. Microbiol. 6,
2732–2745 (2023). 579–591 (2008).
24. Wang, L. et al. High methane emissions from thermokarst lakes on 46. Wei, D. et al. Considerable methane uptake by alpine grasslands
the Tibetan Plateau are largely attributed to ebullition fluxes. Sci. despite the cold climate: in situ measurements on the central
Total Environ. 801, 149692 (2021). Tibetan Plateau, 2008-2013. Glob. Change Biol. 21, 777–788 (2015).
25. Kuhn, M. A. et al. BAWLD-CH4: a comprehensive dataset of methane 47. Wei, D. et al. Revisiting the role of CH4 emissions from alpine wet-
fluxes from boreal and arctic ecosystems. Earth Syst. Sci. Data 13, lands on the Tibetan Plateau: evidence from two in situ measure-
5151–5189 (2021). ments at 4758 and 4320m above sea level. J. Geophys. Res.
26. Masson-Delmotte, V. P. Z. et al. IPCC, 2021: Summary for policy- Biogeosci. 120, 1741–1750 (2015).
makers. In Climate Change 2021: The Physical Science Basis. Con- 48. Yan et al. The spatial and temporal dynamics of carbon budget in
tribution of Working Group I to the Sixth Assessment Report of the the alpine grasslands on the Qinghai-Tibetan Plateau using the
Intergovernmental Panel on Climate Change (Cambridge University Terrestrial Ecosystem Model. J. Clean. Prod. 107, 195–201 (2015).
Press, 2021). 49. Denfeld, B. A. et al. A synthesis of carbon dioxide and methane
27. Juutinen, S. et al. Methane dynamics in different boreal lake types. dynamics during the ice‐covered period of northern lakes. Limnol.
Biogeosciences 6, 209–223 (2009). Oceanogr. Lett. 3, 117–131 (2018).
28. Holgerson, M. A. & Raymond, P. A. Large contribution to inland 50. Mu, C. et al. Dissolved organic carbon, CO2, and CH4 concentra-
water CO2 and CH4 emissions from very small ponds. Nat. Geosci. tions and their stable isotope ratios in thermokarst lakes on the
9, 222–226 (2016). Qinghai-Tibetan Plateau. J. Limnol. 75, 313–319 (2016).
29. Wiesenburg, D. A. & Guinasso, N. L. Equilibrium solubilities of 51. Estop-Aragonés, C. et al. Assessing the potential for mobilization of
methane, carbon monoxide, and hydrogen in water and sea water. old soil carbon after permafrost thaw: a synthesis of 14C measure-
J. Chem. Eng. Data 24, 356–360 (1979). ments from the northern permafrost region. Glob. Biogeochem.
30. Martens, C. S. & Klump, J. V. Biogeochemical cycling in an organic- Cycles 34, e2020GB006672 (2020).
rich coastal marine basin—I. Methane sediment-water exchange 52. Cheng, G. & Wu, T. Responses of permafrost to climate change and
processes. Geochim. Cosmochim. Acta 44, 471–490 (1980). their environmental significance, Qinghai-Tibet Plateau. J. Geophys.
31. Mattson, M. D. & Likens, G. E. Air pressure and methane fluxes. Res. 112, F02S03 (2007).
Nature 347, 718–719 (1990). 53. Zhao, L. & Sheng, Y. Permafrost and its Changes on the Qinghai-
32. Johnson, K. M. et al. Bottle-calibration static head space method for Tibetan Plateau (Science Press, 2019).
the determination of methane dissolved in seawater. Anal. Chem. 54. Pang, Q. et al. Active layer thickness calculation over the Qinghai-
62, 2408–2412 (1990). Tibet Plateau. Cold Reg. Sci. Technol. 57, 23–28 (2009).
33. Zhang, L. et al. Significant methane ebullition from alpine perma- 55. Zhao, L. et al. Estimates of the reserves of ground ice in perma-
frost rivers on the East Qinghai–Tibet Plateau. Nat. Geosci. 13, frost regions on the Tibetan Plateau. J. Glaciol. Geocryol. 32,
349–354 (2020). 1–9 (2010).

Nature Communications | (2023)14:3121 9

Article https://doi.org/10.1038/s41467-023-38907-6

56. Gao, T. et al. Accelerating permafrost collapse on the eastern 80. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a
Tibetan Plateau. Environ. Res. Lett. 16, 054023 (2021). practical and powerful approach to multiple testing. J. R. Stat. Soc.
57. Guo, L. et al. Uncertainty and variation of remotely sensed lake ice B 57, 289–300 (1995).
phenology across the Tibetan Plateau. Remote Sens. 10, 1534 (2018). 81. Sim, C. H. et al. Outlier labeling with boxplot procedures. J. Am.
58. Bastviken, D. et al. Methane emissions from Pantanal, South Stat. Assoc. 100, 642–652 (2005).
America, during the low water season: toward more comprehensive 82. Editorial Board of Vegetation Map of China. Vegetation Atlas of
sampling. Environ. Sci. Technol. 44, 5450–5455 (2010). China (Science Press, 2001).
59. Sepulveda-Jauregui, A. et al. Methane and carbon dioxide emis- 83. R Core Team. R: A Language and Environment for Statistical Com-
sions from 40 lakes along a north–south latitudinal transect in puting (R Foundation for Statistical Computing, 2020).
Alaska. Biogeosciences 12, 3197–3223 (2015). 84. Yang, G. B. et al. Characteristics of methane emissions from alpine
60. Musenze, R. S. et al. Assessing the spatial and temporal variability of thermokarst lakes on the Tibetan Plateau. Dataset. Figshare. https://
diffusive methane and nitrous oxide emissions from subtropical doi.org/10.6084/m9.figshare.22743968 (2023).
freshwater reservoirs. Environ. Sci. Technol. 48, 14499–14507 85. Brown, J. et al. Circum-arctic map of permafrost and ground-ice
(2014). conditions, version 2. [Data Set]. Boulder, Colorado USA. NASA
61. Walter Anthony, K. M. et al. Methane bubbling from Siberian thaw National Snow and Ice Data Center Distributed Active Archive
lakes as a positive feedback to climate warming. Nature 443, Center (accessed 5 May 2023). https://doi.org/10.7265/
71–75 (2006). skbgkf16 (2002).
62. Bouchard, F. et al. Modern to millennium-old greenhouse gases 86. Zou, D. et al. A new map of permafrost distribution on the Tibetan
emitted from ponds and lakes of the Eastern Canadian Arctic (Bylot Plateau. Cryosphere 11, 2527–2542 (2017).
Island, Nunavut). Biogeosciences 12, 7279–7298 (2015).
63. Walter Anthony, K. M. et al. Methane emissions proportional to Acknowledgements
permafrost carbon thawed in Arctic lakes since the 1950s. Nat. We thank Dr. Liwei Zhang (College of Urban and Environmental
Geosci. 9, 679–682 (2016). Sciences, Peking University) for valuable advice on flux measurement
64. Garnett, M. H. et al. Radiocarbon analysis of methane emitted from and calculation, Dr. Pengfei Liu (Center for the Pan-Third Pole
the surface of a raised peat bog. Soil Biol. Biochem. 50, 158–163 Environment, Lanzhou University) for helpful suggestions regarding
(2012). bioinformatic analyses and data interpretations, and Dr. Katey
65. Li, D. et al. MEGAHIT v1.0: A fast and scalable metagenome Walter Anthony (Water and Environmental Research Centre,
assembler driven by advanced methodologies and community University of Alaska Fairbanks) for constructive comments on an early
practices. Methods 102, 3–11 (2016). version of the manuscript. Permissions to work and collect gas and
66. Hyatt, D. et al. Prodigal: prokaryotic gene recognition and transla- sediment samples across the study area were granted by the Three-
tion initiation site identification. BMC Bioinformatics 11, 119 (2010). River-Source National Park Management Bureau. This work was sup-
67. Fu, L. et al. CD-HIT: accelerated for clustering the next-generation ported by the National Natural Science Foundation of China
sequencing data. Bioinformatics 28, 3150–3152 (2012). (31988102 and 31825006, Y.Y., and 32201359, G.Y.), National Key
68. Huerta-Cepas, J. et al. EggNOG 5.0: a hierarchical, functionally and Research and Development Program of China (2022YFF0801903,
phylogenetically annotated orthology resource based on 5090 Y.Y.), and China National Postdoctoral Program for Innovative Talents
organisms and 2502 viruses. Nucleic Acids Res. 47, D309–D314 (BX20200363, G.Y.).
(2018).
69. Buchfink, B., Reuter, K. & Drost, H. Sensitive protein alignments at Author contributions
tree-of-life scale using DIAMOND. Nat. Methods 18, 366–368 (2021). Y.Y. and G.Y. designed this research. G.Y. and Z.Z. performed field
70. Kanehisa, M. et al. KEGG as a reference resource for gene and samplings. G.Y., Z.Z., L.K. and S.Q. performed laboratory experiments.
protein annotation. Nucleic Acids Res. 44, D457–D462 (2016). G.Y., Y.S., L.K. and Y.P. analysed data. G.Y., B.A., D.O., C.K. and Y.Y. wrote
71. Patro, R. et al. Salmon provides fast and bias-aware quantification of the manuscript with input from other co-authors.
transcript expression. Nat. Methods 14, 417–419 (2017).
72. Wood, D. E. et al. Improved metagenomic analysis with Kraken 2. Competing interests
Genome Biol. 20, 257 (2019). The authors declare no competing interests.
73. Asnicar, F. et al. Compact graphical representation of phylogenetic
data and metadata with GraPhlAn. PeerJ 3, e1029 (2015). Additional information
74. Finn, R. D. Pfam: the Protein Families Database (Encyclopedia of Supplementary information The online version contains
Genetics, Genomics, Proteomics and Bioinformatics, 2005). supplementary material available at
75. Jaina, M. et al. Challenges in homology search: HMMER3 and con- https://doi.org/10.1038/s41467-023-38907-6.
vergent evolution of coiled-coil regions. Nucleic Acids Res. 41,
e121 (2013). Correspondence and requests for materials should be addressed to
76. Mendler, K. et al. AnnoTree: visualization and exploration of a Yuanhe Yang.
functionally annotated microbial tree of life. Nucleic Acids Res. 47,
4442–4448 (2019). Peer review information Nature Communications thanks Blaize Denfeld
77. Zhou, Z. et al. Non-syntrophic methanogenic hydrocarbon degra- and the other, anonymous, reviewer(s) for their contribution to the peer
dation by an archaeal species. Nature 601, 257–262 (2022). review of this work. A peer review file is available.
78. Edgar, R. C. MUSCLE: multiple sequence alignment with high
accuracy and high throughput. Nucleic Acids Res. 32, Reprints and permissions information is available at
1792–1797 (2004). http://www.nature.com/reprints
79. Price, M. N., Dehal, P. S. & Arkin, A. P. FastTree 2 – approximately
maximum-likelihood trees for large alignments. PLoS ONE 5, Publisher’s note Springer Nature remains neutral with regard to jur-
e9490 (2010). isdictional claims in published maps and institutional affiliations.

Nature Communications | (2023)14:3121 10

Article https://doi.org/10.1038/s41467-023-38907-6

Open Access This article is licensed under a Creative Commons

Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if
changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.

© The Author(s) 2023

Nature Communications | (2023)14:3121 11