Ratio of In Situ CO2 to CH4 Production and Its Environmental Controls in Polygonal Tundra Soils of Samoylov Island, Northeastern Siberia

Arctic warming causes permafrost thaw and accelerates microbial decomposition of soil organic matter (SOM) to carbon dioxide (CO2) and methane (CH4). The determining factors for the ratio between CO2 and CH4 formation are still not well understood due to scarce in situ measurements, particularly in remote Arctic regions. We quantified the CO2:CH4 ratios of SOM decomposition in wet and dry tundra soils by using CO2 fluxes from clipped plots and in situ CH4 fluxes from vegetated plots. At the water‐saturated site, CO2:CH4 ratios decreased sharply from 95 at beginning of July to about 10 in August and September with a median of 12.2 (7.70–17.1; 25%–75% quartiles) over the whole vegetation period. When considering CH4 oxidation, estimated to reduce in situ CH4 fluxes by 10%–31%, even lower CO2:CH4 ratios were calculated (median 10.9–8.41). Active layer depth and soil temperature were the main factors controlling these ratios. Methane production was associated with subsoil (40 cm) temperature, while heterotrophic respiration was related to topsoil (5 cm) temperatures. As expected, CO2:CH4 ratios were substantially higher at the dry site (median 373, 292–500, 25%–75% quartiles). Both tundra types lost carbon preferentially in form of CO2, and CH4‐C represented only 0.27% of the dry tundra total carbon loss and 6.91% of the wet tundra total carbon loss. The current study demonstrates the dynamic of in situ CO2:CH4 ratios from SOM decomposition and will help improve simulations of future CO2 and CH4 fluxes from thawing tundra soils.

might be caused by differences in incubation conditions and duration, and differences in the composition of OM and microbial communities (Treat et al., 2015). Since the energy gain of methanogenesis is low in comparison to microbial processes using electron acceptors such as oxygen (O 2 ), nitrate, iron, and sulfate, CH 4 production is low as long as these electron acceptors are available (Bodegom & Stams, 1999). Hence, the CO 2 :CH 4 ratio in incubations is not constant and may decrease with incubation time, owing to the depletion of alternative electron acceptors but also due to the establishment of an active CH 4 -producing community (Knoblauch et al., 2018;Philben et al., 2020). Finally, soil incubations are generally done either under anoxic or oxic conditions, while soils, even under water-saturated conditions, are characterized by redox gradients, which may last from completely anoxic to fully oxic. Thus, incubation experiments give only limited information on CO 2 and CH 4 production under in situ conditions. One approach used to measure reliable in situ CO 2 and CH 4 production rates is to derive them from soil concentration depth profiles (Clymo & Bryant, 2008;Clymo et al., 1995), but attempts to deduce the CO 2 to CH 4 production partitioning from gas measurements resulting from soil organic matter decomposition have not been done yet.
Microbial CH 4 oxidation in soil layers that contain oxygen, such as the rhizosphere, where oxygen is leaking from the roots, or the surface soil, is further modifying the in situ CO 2 :CH 4 ratio. Aerobic CH 4 oxidizing bacteria may oxidize up to 99% of produced CH 4 to CO 2 in water-saturated permafrost soils, with particularly high importance of CH 4 oxidation at sites without vascular plants (Knoblauch et al., 2015;Popp et al., 2000). The relevance of CH 4 oxidation strongly depends on CH 4 transport pathways. Methane transported by molecular diffusion through the water phase is >10 4 times slower than ebullition or plant-mediated transport through the gas phase in air-filled tissue. Hence CH 4 oxidation is most relevant when CH 4 moves slowly through the soil (molecular diffusion in the water phase) and lowest when rapidly transported from its production zone into the atmosphere (ebullition, plant-mediated transport) (Bastviken et al., 2008;Knoblauch et al., 2015;Whalen, 2005). Hence, the release of carbon as CO 2 or CH 4 is determined by complex interactions of the factors influencing not just their production, but also their transport.
The extensive Siberian tundra is currently underrepresented in international CH 4 emissions databases (Saunois et al., 2020). There are large uncertainties regarding the controls of CH 4 production and emission in the vast Russian Arctic tundra, and consequently the response of CH 4 emission to climate change. Permafrost acts as a barrier to drainage resulting in lakes, ponds, and wetlands, characterized by anoxic conditions and accumulation of OM. The seasonal freezing and thawing of the upper soil layer (active layer) promote the creation of patterned ground, such as polygonal structures, where the depressed centers are often water saturated while the elevated rims are drained and characterized by oxic conditions (van Huissteden, 2020). Differences in vegetation type may account for substantial variation in CH 4 and CO 2 emissions from the Arctic tundra (Cannone et al., 2016;Knoblauch et al., 2015). Due to the complexity of environmental and microbial parameters affecting CH 4 production and turnover in permafrost soils, there are still large uncertainties on how thaw-induced changes in the soil hydrology of permafrost landscapes will impact in situ CO 2 and CH 4 production from thawing permafrost OM and related GHG emissions (Euskirchen et al., 2017). A quantitative understanding of the regulation of the CO 2 :CH 4 ratio is needed to improve process-based models simulating the feedback between thawing permafrost and global change. Most global models still calculate CH 4 production or emission as a fixed ratio of soil organic matter (SOM) decomposition (Kleinen et al., 2021;Melton et al., 2013;Wania et al., 2010).
The objective of this study is to improve our knowledge of the regulation of the ratio between microbial CO 2 and CH 4 production from OM decomposition under in situ conditions of a typical polygonal tundra. For this, small-scale approaches, such as chamber measurements are more appropriate than eddy-covariance systems (Krauss et al., 2016). Therefore, we used an experimental setup based on the chamber approach and measured for a whole summer season (July-September) in situ greenhouse gas fluxes at two soils of the Siberian polygonal tundra characterized by different hydrological regimes. Methane production was estimated using fluxes from plots vegetated by vascular plants, and CO 2 fluxes representing heterotrophic respiration (R h ) were quantified from clipped plots of a root-trenching experiment (Eckhardt et al., 2019). By this approach, we identified the relative contribution of CO 2 and CH 4 production during OM decomposition in permafrost-affected soils under in situ conditions, identified the most important environmental parameters regulating the in situ ratio between CO 2 and CH 4 production, and estimated the contribution of plant-mediated CH 4 transport to total CH 4 emissions. Finally, we used the obtained CO 2 and CH 4 data to calculate the production of CO 2 and CH 4 in the polygonal tundra of Samoylov Island for one summer thaw season.

Study Site
The study was conducted on Samoylov Island in the southern central Lena River Delta, Northeastern Siberia (72°22′N, 126°28′E) in the continuous permafrost zone, with permafrost depths of 300-500 m (Yershov, 1998) ( Figure 1). The annual mean permafrost temperature is −8.6°C at about 11 m depth, while temperatures at the surface soil can vary from 20°C to −35°C throughout the year. The island has an arctic continental climate with an annual mean temperature of −12.5°C and annual mean precipitation of 321 mm (Boike et al., 2013). Snowmelt starts in June and the growing season lasts from mid-June to mid-September. Polar days, which are days when the sun remains above the horizon for 24 hr, span from 7 May to 8 August, and polar nights, which are days when the sun remains below the horizon for 24 hr, spanning from 15 November to 28 January. The study has been executed at the eastern part of Samoylov Island, which is covered by ice-wedge polygonal tundra on a Late Holocene river terrace, characterized by ice-rich alluvial deposits. The polygonal tundra is formed by depressed polygon centers surrounded by elevated polygon rims with an elevation difference of about 0.5 m. Water-saturated soils are often found in the polygon centers due to the underlying permafrost that prevents drainage while soils at the elevated polygon rims are well-drained. Mean carbon pools in the uppermost 1 m are higher in the polygon centers (33 kg SOC m −2 ) than in the polygon rims (19 kg SOC m −2 ) (Zubrzycki et al., 2013).
The study was carried out in a polygon (72°22′26″N, 126°29′49″E) with one sub-site in the water-saturated, depressed polygon center (wet tundra) and another sub-site in its surrounding polygon rim (dry tundra). The soil at the polygon center was classified as Reductaquic Cryosol (WRB, 2015) with the water table varying from 7 cm below to 7 cm above the soil surface during the measurement period, a maximum yearly active layer depth (ALD) of 40 cm and vegetation dominated by mosses (Drepanocladus revolvens, Meesia triqueta, Scorpidium scorpioides) and the hydrophilic sedge Carex aquatilis. Total organic carbon (TOC) concentrations in the polygon center soil ranged between 10.1% and 19.6%, with the highest concentrations in the uppermost 6 cm ( Figure S1 in Supporting Information S1). The soil at the polygon rim was classified as Turbic Glacic Cryosol (WRB, 2015) with the water table a few centimeters above the permafrost table, maximum ALD of 30 cm, and vegetation dominated by mosses (Hylocomium splendens, Polytrichum spp., Rhytidium rugosum), some small vascular plants (Dryas punctata and Astragalus frigidus) and lichens (Peltigera spp.). The polygon rim soil contained 12.3% of TOC in the uppermost 15 cm depth and cryoturbated horizons bellow 15 cm ( Figure S2 in Supporting Information S1).

Soils and Meteorological Data
The soil temperatures (SoilT) at 2, 5, 10, 15, and 40 cm and air temperature (AirT) at 2 m height, precipitation, and incoming and outgoing components of shortwave and longwave radiation were recorded by a nearby  (Boike et al., 2019). The surface temperature (SurfT) was estimated by the following equation: where L↑ B is the upward infrared radiation (W m −2 ), ε is the dimensionless emissivity with a value of 0.98 and σ is the Stefan-Boltzmann constant (W m −2 K −4 ) (Wilber et al., 1999). The SurfT was calculated in Kelvin with Equation 1 and converted to degrees Celsius.
The description of the water table depth (WT) measurements in the center, the volumetric soil water content (VWC) at the rim, and the ALD in both sub-sites can be found in Eckhardt et al. (2019).

Chamber Measurements and Plant-Mediated CH 4 Transport
At each sub-site, 10 measurement plots made of PVC frames (50 × 50 cm) were established. From these, four had their original vegetation intact (hereafter called "vegetated"), while six had the surface vegetation removed (hereafter called "clipped"). The experiment has been described in detail by Eckhardt et al. (2019). The frames were inserted 20 cm deep into the soil, below the main rooting zone. The opaque acrylic chamber was equipped with a fan for the mixing of the air. The chambers were 50 cm high and enclosed a volume between 124 and 143 L, depending on the terrain inside the chamber frames. Two holes with 3 cm of diameter were left open at the top of the chamber while placing it slowly on the frames to avoid pressure-induced gas release from the soil (Christiansen et al., 2011;Eckhardt et al., 2019). These holes were closed before the measurement. A boardwalk was installed at the polygon to avoid disturbances. We performed a clipping experiment by using the root-trenching method in four frames at each sub-site on the polygon (center and rim) in 2014. In addition to the cutting of the lateral roots through the insertion of the PVC frames, all living plant biomass, including mosses, was carefully removed from within the "clipped" frames. During the measurement period in 2015, this procedure was repeated periodically to prevent plant regrowth. Additionally, two frames were installed in 2015 and clipped at each sub-site to test if CO 2 fluxes were biased due to the additional decomposition of residual roots, which might be a possible artifact of the root-trenching method. However, no significant difference between the CO 2 fluxes in the clipped plots of 2014 and those clipped in 2015 was observed (Eckhardt et al., 2019).
The CH 4 concentrations in the headspace of the chambers were measured with a portable gas analyzer (UGGA 30-p; Los Gatos Research, USA) and recorded with a data logger (CR800 series; Campbell Scientific Ltd., USA). The precision of the gas analyzer for CH 4 is better than 0.005 ppm. Chamber closure time was 120 s, during which chamber headspace air was pumped in a closed loop through the analyzer at a rate of 200 mL min −1 . The chamber closure of 120 s was chosen to prevent the warming inside the chamber in relation to the exterior temperature. This amount of time was sufficient for the detection of small CH 4 concentration variations in the headspace while averting its temperature increase. Measurements were conducted between 11 July and 22 September, 2015, except in the periods between 2 and 9 August, when a shift change between researchers took longer than expected, and the 17 and 24 August, when the measurements were impossible due to a heavy storm event. During the measurement period, measurements were taken at least every third day. The R h fluxes used in this study are from Eckhardt et al. (2019). The CH 4 fluxes were calculated using MATLAB (R2019a. In The MathWorks Inc.) with a routine combining different regression models, such as linear, exponential, and increasing polynomial degrees, and statistical analysis for model selection. Details can be found in Eckhardt and Kutzbach (2016). Methane concentrations inside the chamber headspace increased linearly over time ( Figure S3 in Supporting Information S1), thus linear regressions have been used for the flux calculation. The first 30 s of each 120 s measurement period were discarded to eliminate possible perturbations when placing the chamber on the frame. The fraction of plant-mediated CH 4 fluxes was calculated as the difference between the daily mean CH 4 fluxes from the vegetated plots and the clipped plots (Table 1).

Calculation of CO 2 :CH 4 Ratios and Uncertainty Range
The CO 2 :CH 4 ratios were calculated on a molar basis, using the daily mean CH 4 fluxes from the vegetated plots, and the daily mean R h fluxes, which were measured with dark chambers in the clipped plots (see Eckhardt et al., 2019). The CH 4 fluxes above vegetated plots in the water-saturated polygon center are the best estimate of in situ CH 4 production. These fluxes might underestimate the CH 4 production since no CH 4 oxidation is considered. However, we have estimated the range of CH 4 oxidation that could be affecting our measured fluxes using the fraction of oxidized CH 4 and plant-mediated CH 4 transport data from studies in sites similar to this study, as is described later in this section. Hence, the calculated CO 2 :CH 4 ratios of the polygon center are considered the CO 2 :CH 4 production ratios. However, this is not the case for CH 4 fluxes from the polygon rim due to high CH 4 oxidation in the unsaturated soil. This procedure resulted in one single CO 2 :CH 4 ratio per day in each sub-site on the polygon center and rim. Thus, to obtain a variation measure for these daily ratios, we propagated the standard deviation of the CH 4 and CO 2 daily fluxes using the following equation (Singh & Chaturvedi, 2021): (2) where σ CO2:CH4 is the estimated standard deviation of the CO 2 :CH 4 ratio; CO 2 :CH 4 is the mean CO 2 :CH 4 ratio; σ CO2 is the standard deviation of the CO 2 fluxes; f CO2 is the mean CO 2 flux; σ CH4 is the standard deviation of the CH 4 fluxes and f CH4 is the mean CH 4 flux.
Our approach to estimating CH 4 production in the soil from CH 4 fluxes above vegetated plots is neglecting CH 4 oxidation in the soil, which causes an underestimation of in situ CH 4 production. To account for this potential error, we used published data on CH 4 oxidation (Knoblauch et al., 2015;Preuss et al., 2013) at water-saturated polygon centers on Samoylov that were vegetated by the same vascular plant (Carex aquatilis) as our site ( Table 2). The polygon center daily average CH 4 fluxes (f CH4(corrected) ) were recalculated as follows: where MOR is the fraction of produced CH 4 oxidized in the soil, D CH4 is the fraction of CH 4 transported through the soil, P CH4 is the fraction of CH 4 transported through plants and f CH4 is the mean CH 4 flux. The MOR values were retrieved from the literature, and the D CH4 values were either CH 4 fluxes from the clipped plots of this study or additional values retrieved from the literature. The use of these values in the definition of the upper and lower boundaries of the uncertainty range is shown in Table 2. The lower boundary uses the lowest CH 4 oxidation from Table 1 Median (

in Bold), First (Q1), and Third (Q3) Quartiles of CH 4 Fluxes During the Growing Season of 2015 From Vegetated and Clipped Plots in a Polygon Center and a Polygon Rim on Samoylov Island, and Contribution of Plant-Mediated CH 4 Transport
Polygon sub-site  Preuss et al. (2013), and the highest P CH4 from Knoblauch et al. (2015). The upper boundary uses the highest CH 4 oxidation from Knoblauch et al. (2015) and the lowest P CH4, which was found in this study. This approach assumes that only CH 4 diffusing slowly through the soil is affected by CH 4 oxidation but not that fraction of CH 4 , which is released rapidly by plant-mediated transport. These new daily mean CH 4 fluxes were then used to recalculate the center CO 2 :CH 4 production ratios as described above.

Calculation of the Seasonal Heterotrophic Respiration and CH 4 fluxes
The seasonal R h and CH 4 fluxes were calculated from the CO 2 of clipped plots and CH 4 fluxes of vegetated plots of the wet (polygon center) and dry tundra (polygon rim) in kilograms per hectare per day, during the measurement period.  Kwon et al., 2017;Natchimuthu et al., 2017;Schrier-Uijl et al., 2010;Wickland et al., 2006). For the direct comparison of R h and CH 4 fluxes (Table 3), the median R h fluxes and the median CH 4 fluxes were normalized to carbon (kg CO 2 -C/CH 4 -C ha −1 d −1 ) to consider the mass differences between the two molecules. For the comparison based on the whole area of Samoylov (Table 3), we used the fraction of wet and dry tundra (19% and 65%, respectively) of the polygonal tundra mapped by Muster et al. (2012) on Samoylov Island, resulting in 54 ha of wet tundra and 185 ha of dry tundra. The active floodplains and open water bodies have not been included in our study. The CO 2 -C and CH 4 -C fluxes were multiplied by the area of each land cover (dry and wet tundra), resulting in daily absolute CO 2 -C and CH 4 -C production (kg CO 2 -C/CH 4 -C d −1 ) of each tundra type.

Statistics
The CO 2 :CH 4 ratios of the polygon center and rim were analyzed separately through linear regression models. First, we have tested the relationship between CO 2 :CH 4 ratios and each of the daily mean environmental variables, namely soil temperatures at 2 cm, 5 cm (SoilT5), 10, 15, and 40 cm depth (SoilT40), SurfT, ALD, WT (for the polygon center) and VWC (for the polygon rim) and selected the best predictors. The same was made with the CH 4 and R h daily mean fluxes, to test the relationships between the predictors and the CO 2 :CH 4 ratios. Natural logarithm transformation was applied for non-normally distributed data.
Then, a multivariate regression model relating the CO 2 :CH 4 ratios with the best predictors, chosen during the abovementioned linear regressions between CO 2 :CH 4 ratios and individual environmental parameters, was set including the possible interaction between predictors. If the interaction between predictors was not significant (p > 0.05), the sum of their effects was considered. The possibility of logarithmic and exponential relationships between response variables and predictors was also verified throughout the previous steps.
We chose the model with the highest R 2 , smallest root mean squared error (RMSE), lowest p-value from the model F-test (α = 0.05), and the lowest Akaike Information Criterion (AIC) that gives the relative quality of statistical models for the same data set. The best-adjusted regressions are shown in the results section. The statistical analyses were performed in R (R, 2020).

CH 4 Fluxes and Plant-Mediated Transport
The CH 4 fluxes from the vegetated plots at the polygon center varied throughout the measurement period, showing a clear seasonality (Figure 2). There was a fast increase in July, at the beginning of the growing season, followed by a slower increase in the first weeks of August. The fluxes stabilized at the peak rates, representing six times the flux measured at the beginning of July, from mid-August until the beginning of September, when the fluxes started to decrease at the end of the growing season. The variation of CH 4 fluxes at the polygon rim showed no seasonality but a sudden increase on the 22nd of September, the last day of measurements.
The median CH 4 flux of the vegetated plots at the polygon center was 26.4 mg m −2 d −1 (21.1-32.5 mg m −2 d −1 ; 25%-75% quartiles). The median CH 4 flux at the polygon rim was 1.85 mg m −2 d −1 (1.58-2.22 mg m −2 d −1 ; 25%-75% quartiles). The daily CH 4 fluxes means at the center were between 3.9 and 20.2 times higher than at the rim throughout the season. The log-transformed CH 4 fluxes from vegetated plots at the center had significant relationships with the ALD (R 2 = 0.73; p < 0.001), with the SoilT40 (R 2 = 0.68; p < 0.001) and had no relationship with the SurfT (p = 0.172). The log-transformed CH 4 fluxes at the rim also had a significant relationship with the ALD but with a lower R 2 (R 2 = 0.44; p < 0.001) and with the SoilT40 (R 2 = 0.42; p < 0.001), but no relationship with shallower soil temperatures like the SoilT5 (p = 0.404) and SurfT (p = 0.824). The mean daily CH 4 fluxes of the clipped plots at the polygon center were lower than those from the vegetated plots. The differences between vegetated and clipped plots are smaller at the beginning of the growing season compared to later periods. Based on the differences in CH 4 fluxes between clipped and vegetated plots, the median plant-mediated CH 4 transport at the polygon center is 79% (65%-89%, 25%-75% quartiles). At the polygon rim, there was only a very small difference between clipped and vegetated plots of 3% (Table 1).
Based on stable isotope signatures of CH 4 dissolved in the water-saturated soil of polygon centers on Samoylov, it has been estimated that a fraction between 0.45 and 0.68 of the CH 4 diffusing through the bulk soil is oxidized to CO 2 before emitted into the atmosphere (Knoblauch et al., 2015;Preuss et al., 2013) (Table 2). Using this range of CH 4 oxidation and the fraction of plant-mediated CH 4 transport (P CH4 ) and CH 4 transport through the bulk soil (D CH4 ) determined in this and previous studies (Knoblauch et al., 2015;Preuss et al., 2013) we calculated that a fraction between 0.10 and 0.31 of produced CH 4 might have been oxidized before emitted into the atmosphere (assumptions 2 and 3, Table 2). Using these CH 4 oxidation estimates would increase the median of CH 4 production rates from 26.4 mg m −2 d −1 (no oxidation assumed) to 29.5 mg m −2 d −1 (fraction of 0.10 oxidized) and 38.2 mg m −2 d −1 (fraction of 0.31 oxidized) ( Table 2).

CO 2 :CH 4 Ratios, Their Environmental Controls, and Uncertainty Range
The CO 2 :CH 4 ratios, calculated from in situ CH 4 fluxes from vegetated plots and R h , varied between 3 and 95 at the polygon center (Figure 3), while they varied at the rim between 80 and 1,074. The CO 2 :CH 4 ratios at the polygon center were around 95 at the beginning of July and decreased rapidly until mid-July. From 21 July on, the ratios ranged between 7 and 15. There was a high variability of CO 2 :CH 4 ratios in the polygon rim. The amplitude of this variation and the ratios themselves are higher in July and steadily decrease over time. From 12 September on, this downward trend is seen more clearly ( Figure S7 in Supporting Information S1). The median CO 2 :CH 4 ratio was 12.2 (7.70-17.1; 25%-75% quartiles) at the polygon center and 373 (292-500, 25%-75% quartiles) at the polygon rim.
The best predictor of the CO 2 :CH 4 ratios at the polygon center was the ALD, with a linear relationship (R 2 = 0.72; p < 0.001) (Figure 3). Both the CO 2 :CH 4 ratio and the ALD stabilize around a constant value from the end of July until the 22 September. The second-best predictor for the CO 2 :CH 4 ratios at the polygon center was the SoilT40 with a logarithmic relationship (R 2 = 0.65; p < 0.001). The center CO 2 :CH 4 ratios have exponential relationships with AirT (R 2 = 0.36; p < 0.001) and SurfT (R 2 = 0.35; p < 0.001).
The CO 2 :CH 4 ratios at the polygon rim had higher variability and a more gradual decreasing trend along the season than at the center. The best predictors of the CO 2 :CH 4 ratios at the polygon rim were the SoilT5 (lin; R 2 = 0.55; p < 0.001) and the SurfT (lin; R 2 = 0.45; p < 0.001). However, the inclusion of the SurfT as a predictor along with the SoilT5 into the multivariate regression model did not add explanatory power to it due to collinearity, which was shown by a higher AIC. For that reason, the ALD, which was the third best predictor (lin; R 2 = 0.30; p < 0.001), was included instead, with a significant increase in the quality of the model (p = 0.024). The statistical models that showed the best explanatory power for the CO 2 :CH 4 ratios at the polygon center and rim are shown in Table 4.
In addition to the relationships between R h and environmental variables reported by Eckhardt et al. (2019), we present here some additional relationships that improve our understanding of the CO 2 :CH 4 ratios. The R h at the polygon center showed relationships with the SurfT (lin; R 2 = 0.59; p < 0.001), with the SoilT40 (log; R 2 = 0.24; p < 0.001) and no relationship with the ALD (lin; p = 0.062). The R h at the polygon rim had a linear relationship with the SoilT5 (R 2 = 0.60; p < 0.001) but none with the ALD (lin; p = 0.076). Furthermore, there was a weak linear relationship between the SurfT and the SoilT40 at the polygon center (R 2 = 0.1305; p = 0.016).
Using CH 4 fluxes corrected for CH 4 oxidation (f CH4(corrected) ) to calculate ratios between CO 2 and CH 4 production results in a decrease in CO 2 :CH 4 ratios ( Figure 3) with median values of 10.9 (fraction of 0.10 oxidized) and 8.41 (fraction of 0.31 oxidized) in comparison to 12.2 (no oxidation assumed) ( Table 2).

Seasonal Heterotrophic Respiration and CH 4 fluxes
The median R h , which was measured with the dark chamber in the clipped plots, of the wet tundra during the thaw season was lower than the one of the dry tundra's, while the median CH 4 flux, which was measured with the transparent chamber in the vegetated plots, of the wet tundra, was higher than of the dry tundra (Table 3). The R h and CH 4 fluxes were converted to kg-C fluxes to allow a direct comparison between them. The wet tundra, although releasing more CH 4 than the dry tundra, emitted around half the total kg-C of the dry tundra. When the size of each land cover is considered, the dry tundra emits more than six times the total kg-C, as R h and CH 4 , of the wet tundra. Due to the small area of wet tundra on Samoylov Island and the low contribution of CH 4 to the dry tundra GHG fluxes, CH 4 contributed only 1.17% (or 13.4 kg-C d −1 ) to the total daily flux of carbon from the polygonal tundra on Samoylov (1,142 kg-C d −1 , Table 3). Note. "Interaction" states the interaction between predictors.

CH 4 Fluxes in Comparison to Other Arctic Sites
The median CH 4 fluxes from the vegetated plots during the growing season from the polygon center in this study (26.4 mg m −2 d −1 ) are at the lower end of CH 4 fluxes reported from the water-saturated polygonal tundra on Samoylov vegetated by vascular plants of 28.0-100.0 mg m −2 d −1 (Knoblauch et al., 2015;Kutzbach et al., 2004;Sachs et al., 2010;D. Wagner et al., 2003). These differences can be related to the high spatial and temporal variability of CH 4 fluxes in the Siberian tundra (Skeeter et al., 2020), but also to differences in CH 4 flux calculation. The highest CH 4 flux from water-saturated polygon centers on Samoylov (77.9-100.0 mg m −2 d −1 [Sachs et al., 2010]) was calculated from an initial non-linear CH 4 concentration increase with an exponential model, resulting in higher fluxes than linear models. We could only observe an initial non-linear rise of CH 4 concentrations after inducing a pressure increase inside the chamber when placing it on the frames. Under standard measurement conditions, CH 4 concentrations increased linearly inside the chambers ( Figure S3 in Supporting Information S1). Hence, we used a linear fit to calculate CH 4 fluxes. Also, median CH 4 fluxes from the dry polygon rim (1.85 mg m −2 d −1 ) were lower than from previous measurements ( . The soils of both polygon center and rim in our study are mineral soils with a sandy texture (Eckhardt et al., 2019), and lower CH 4 fluxes from mineral soils than from organic soils have been reported in previous studies (Christiansen et al., 2015). Moreover, the Lena River Delta is one of the coldest permafrost regions on Earth (Romanovsky et al., 2010) and methanogenesis is largely controlled by temperature (Yvon-Durocher et al., 2014).
The unusually high CH 4 fluxes measured on 9 August, coincided with the second hottest day of the season, only after 13 July. The average AirT was 17.1°C, and the average SurfT was 21.3°C. The temperature of the methane-producing layer in the soil at around 40 cm depth on 9 August was −0.1°C on average, which was the last day of negative temperatures measured at this depth during the season. This "zero curtain" period of this specific soil layer could explain the high fluxes since CH 4 trapped in the soil could have been released during thawing. Although 13 July was hotter (average air temperature of 18°C), CH 4 fluxes were much lower because the active layer was not as deep as on 9 August, and the soil temperature at 40 cm depth was −0.7°C, the second lowest temperature at this depth after 12 July.
As expected, the CH 4 fluxes at the drained polygon rim in our study were very low and less variable than at the wet polygon center sub-site, however, it did not act as a CH 4 sink, as reported in other studies Kwon et al., 2017;Skeeter et al., 2020;St Pierre et al., 2019). The lack of water saturation and consequently oxic conditions demonstrated by the VWC, which varied from 28.1% to 31.8% within the study period, is probably the main reason for the lower CH 4 fluxes detected at the rim compared to the center. In contrast to the polygon center, water saturation was never achieved at the polygon rim, reducing the habitat for methanogens to produce CH 4 , and setting the conditions for methanotrophs to oxidize CH 4 . The CH 4 fluxes of the polygon center showed a gradual increase over the first part of the growing season until a peak and posterior decrease, while constant CH 4 fluxes were measured at the polygon rim. These patterns were very similar to those reported by D. Wagner et al. (2003), although, differently than here, the center CH 4 fluxes reached the peak earlier in mid-July, when they started to decrease continuously. D. Wagner et al. (2003) related the seasonal fluctuations to the microbiological processes of CH 4 production and oxidation, with higher CH 4 production and lower CH 4 oxidation in the early summer and the opposite by the end of the season. The same seasonality trends of polygon center and rim CH 4 fluxes can be found in other regions in Northeastern Siberia (Kwon et al., 2017), as well as in other Arctic environments (Ström et al., 2012). Physical environmental parameters such as surface temperature (Pickett-Heaps et al., 2011) and soil temperature (Delwiche et al., 2021) have been pointed out as strong predictors of seasonal CH 4 variability. Seasonal vegetation patterns are also playing a crucial role in CH 4 seasonality. Evidence in the literature shows a strong correlation between peak season CH 4 fluxes and maximum gross ecosystem production (Nielsen, Michelsen, Strobel, et al., 2017) and between the development of the vegetation root system along the growing season and CH 4 fluxes (Joabsson & Christensen, 2001). Yet, inter-annual CH 4 fluxes variation can be significant, with largely different seasonal patterns found between years (Mastepanov et al., 2013).
Despite the absence of seasonality in the variation of CH 4 fluxes at the polygon rim, anomalously high CH 4 fluxes were measured there on the 22 September, the last day of measurements. The 22 September was also the only day of the campaign in which negative soil temperatures were recorded at 2 cm depth, indicating that the soil profile was becoming completely frozen since negative temperatures were already being recorded at and below 5 cm. We hypothesize that CH 4 stored at soil and vegetation cavities was being squeezed out to the atmosphere during the zero curtain period, in which the soil layer between the freezing top and permafrost remains unfrozen at around 0°C for months. Previous studies showed similar results with a strong soil gas pressure increase  and CH 4 fluxes increase (Mastepanov et al., 2008;Pirk et al., 2015;Zona et al., 2016) as the soil started to freeze in fall.

Impact of Plant-Mediated CH 4 Transport
The importance of plant-mediated CH 4 transport in wetlands has been documented in previous studies and mainly depends on plant type and density (Knoblauch et al., 2015;Popp et al., 2000). Plants enhance CH 4 release by facilitating CH 4 transport from the soil into the atmosphere through air-filled plant tissue (aerenchyma) and also support CH 4 production by the release of root exudates, which fuel methanogenesis in the anoxic soil (Girkin et al., 2018). The seasonality of the difference between CH 4 fluxes from clipped and vegetated plots in our study is characterized by a smaller difference between treatments at the beginning of the growing season and a gradual increase of the difference until mid-September when the difference starts to decrease again. Notwithstanding, Noyce et al. (2014) showed, in a similar clipping experiment, that the largest difference between clipped and vegetated sites occurred during fall when sedges were senescing. The authors hypothesized that a higher water table in fall kept the rooting zone saturated, intensifying the root's influence on CH 4 production and transport. This was not the case in our measurements in September, as the water table depths were at the lowest level since the start of the measurements in July (Eckhardt et al., 2019), therefore keeping a smaller volume of the rooting zone saturated.
The plant-mediated CH 4 transport in the polygon center (79%) is within the range of 66%-98% reported for wet sub-Arctic and Arctic tundra sites dominated by sedges and grasses (Knoblauch et al., 2015;Kutzbach et al., 2004;Morrissey & Livingston, 1992;Nielsen, Michelsen, Strobel, et al., 2017). The growth of aerenchymatous plants in Arctic sedge-dominated sites is a key factor linking CH 4 production with CH 4 fluxes to the atmosphere. Nielsen, Michelsen, Strobel, et al. (2017) identified a decoupling between peak dissolved CH 4 in the soil solution and peak CH 4 fluxes, suggesting that substrate availability of CH 4 was not the only factor controlling CH 4 fluxes. Moreover, they identified a connection between the peak gross ecosystem production and peak CH 4 fluxes indicating the importance of aerenchymatous plants in the transport of CH 4 to the atmosphere.
However, it is important to notice that the contribution of plant-mediated CH 4 transport to total CH 4 fluxes calculated from the clipping experiment is a conservative estimate. At the vegetated plots, the produced CH 4 is released rapidly through the plants into the atmosphere, thereby by-passing the oxidative layer of the soil, while some produced CH 4 is transported by soil diffusion. After clipping, the CH 4 that was once transported via the vegetation is diffusing through the water-saturated soil. Likely, a part of the produced CH 4 that would have been transported via plants at the vegetated plots is now oxidized in the soil, since now it has to slowly diffuse through the water phase. Thus, the fluxes measured at the clipped plots might not represent only the CH 4 that is transported via soil diffusion in the intact environment but account also for a part of CH 4 that is transported through the vegetation under natural conditions. On the other hand, the removal of the vegetation also ends the support for CH 4 production by root exudates, limiting CH 4 production. These two contrary processes, whose effects on CH 4 fluxes were not quantified, introduce further uncertainties to our estimates.
We consider the difference between clipped and vegetated sites as a measure of the plant-mediated transport of CH 4 because there was no evidence of significant ebullitive transport at our sites. There was no abrupt increase in CH 4 concentrations during chamber closure times in any of our chamber measurements throughout the growing season (e.g., Figure S3 in Supporting Information S1), which would indicate ebullitive CH 4 fluxes. We presume that the dense root system of the vegetation at the polygon center prevented CH 4 to move through the water as bubbles. Although we cannot exclude the possibility of a small contribution of ebullitive CH 4 fluxes into the total CH 4 transported to the atmosphere, we consider that the linearity of CH 4 concentrations during chamber deployment in all our CH 4 chamber measurements demonstrate that the transport of CH 4 was overall comprised by plant-mediated and diffusive transport. Additionally, Knoblauch et al. (2015) have found negligible ebullitive fluxes in most of the studied polygon ponds on Samoylov Island, with similar soils as the one in this study.
At the polygon rim, we estimated a negligible plant-mediated transport of 3% similar to D. Wagner et al. (2003), while Kutzbach et al. (2004) detected 27%. However, the latter study also reported higher CH 4 fluxes at the rim and used a different approach to quantify plant-mediated transport that could be one of the causes for the distinct results: Kutzbach et al. (2004) used closed chambers that enclosed single Carex aquatilis culms, excluding CH 4 being transported by the soil. Ebullitive CH 4 fluxes were not observed at the polygon rim due to non-water-saturated conditions. The soil at the polygon rim in our study contained a deep oxic, unsaturated layer and released less CH 4 regardless of the removal of the vegetation. Presumably, both low CH 4 production and high CH 4 oxidation (Liebner & Wagner, 2007) affected the CH 4 release at the polygon rim simultaneously (Vaughn et al., 2016). Moreover, the polygon rim was mostly vegetated by mosses and small vascular plants (Dryas punctata and Astragalus frigidus) that do not possess aerenchyma like sedges. Hence, well-drained soil conditions, resulting in low CH 4 production and low presence of aerenchymatous plants are the likely reasons for only a small difference between CH 4 fluxes at the clipped and vegetated plots.

In situ CO 2 :CH 4 Ratios and Their Environmental Controls
The rapid decrease of CO 2 :CH 4 ratios at the polygon center from 94.6 to 12.5 between 11 July and 1 August shows the increasing relevance of the CH 4 production and decreased impact of R h during the progression of the growing season. Methane fluxes increased from 5.8 to 21.5 mg m −2 d −1 in this period, following the deepening of the active layer. The negative relationship of the ALD with the CO 2 :CH 4 ratios has also been demonstrated in previous studies (McCalley et al., 2014;van Huissteden et al., 2005). The reason for this relationship is an increase in the water-saturated, and hence anoxic, unfrozen soil volume (Lagomarsino & Agnelli, 2020;Rößger et al., 2022) but also the growth of vegetation, a consequence of increased substrate and nutrient availability with active layer thawing (Andresen et al., 2017).
In addition to ALD, soil temperature had a strong effect on CH 4 fluxes and CO 2 :CH 4 ratios. Methane production is, as with any other microbial process, temperature-dependent (Y. Chen et al., 2022;Kelly & Chynoweth, 1981;Schädel et al., 2016;Treat et al., 2015;Zeikus & Winfrey, 1976;Zinder et al., 1984). Kolton et al. (2019) detected a negative relationship between temperature and CO 2 :CH 4 ratios, indicating a stronger increase in CH 4 production with rising temperatures than CO 2 production. The production of CH 4 was directly correlated to rising soil temperatures at the depth of 40 cm, but also indirectly by promoting the deepening of the active layer. Similar results were reported by Rößger et al. (2022) in Samoylov Island, showing that polygon center soil temperature at 30 cm, thaw depth, and growing degree days were the variables with the highest CH 4 fluxes predictive power. In contrast, R h at the center had only a weak relationship with the soil temperature at 40 cm depth. However, Eckhardt et al. (2019) identified air temperature followed by SurfT as the best predictor for R h at the polygon center. These observations indicate that CH 4 fluxes are affected by processes and changes happening at deeper soil layers (D. Wagner et al., 2003), while R h is affected by processes in the top-soil (Ferréa et al., 2012). The deepest soil layer of our site was not affected by water table variations and maintained its anoxic conditions, enabling the establishment and active methanogenic community (Knoblauch et al., 2015;Liebner et al., 2015), while the soil surface contains the highest O 2 concentrations, staying frequently above the water table. An incubation experiment using samples from Samoylov Island presented evidence for the markedly higher methanogenesis potential of the bottom active layer compared to the surface (Walz et al., 2017). Therefore, the warming of the soil surface stimulates R h , causing elevated CO 2 :CH 4 ratios, while the warming of deeper anoxic soil layers, stimulates methanogenesis and decreases the CO 2 :CH 4 ratios. It might be expected that the surface and deep soil temperatures should be highly related, but just a weak relationship between SurfT and SoilT40 at the polygon center was detected in this study. The soil at the depth of 40 cm is in constant interaction with the underlying permafrost and remained frozen during about one third of the measurement period. The thermal regime at this depth is less controlled by solar radiation and more by the underlying permafrost temperatures than at the soil surface. The CO 2 :CH 4 ratios at the polygon rim was governed mainly by the R h variability since the CH 4 production is low throughout the measurement period. This is also demonstrated by the relationship between CO 2 :CH 4 ratios at the rim and the SurfT, which was also the case for R h .
The theoretical CO 2 :CH 4 production ratio during anaerobic organic matter decomposition via methanogenesis is about 1 when the oxidation number of the organic matter is zero (Symons & Buswell, 1933). However, under natural conditions, CO 2 :CH 4 ratios above one are generally reported even under anoxic conditions, since alternative electron acceptors such as nitrate, ferric iron, sulfate, or even organic matter may be used for SOM decomposition (Dettling et al., 2006). Anoxic incubations show a decrease in the CO 2 :CH 4 ratios throughout the experiment, reaching values of about one, most likely because alternative electron acceptors get depleted and methanogenic communities become active (Knoblauch et al., 2018;Philben et al., 2020). The increase of CH 4 production with decreasing availability of alternative electron acceptors has been observed at other Arctic sites (Rissanen et al., 2017) and could have also played a role in the change of CO 2 :CH 4 ratios in the current study. A shallow ALD at the beginning of the growing season results in a dominance of oxic organic matter decomposition in the oxic surface soil and a reoxidation of alternative electron acceptors. Later on, as the active layer keeps deepening, the anoxic bottom soil thaws and anoxic microbial organic matter decomposition becomes increasingly important, and with the depletion of the alternative electron acceptors, methanogenesis may increase. However, a ratio between CO 2 and CH 4 production above one is expected even under optimum conditions for methanogenesis in the anoxic soil at the polygon center, since CO 2 from heterotrophic respiration was predominantly formed in the oxic surface soil (see discussion above), where no CH 4 is produced.

In situ CO 2 :CH 4 Production Ratios Uncertainty Range
While several methods are established to quantify microbial CO 2 production from SOM decomposition (R h ), the quantification of in situ CH 4 production is still challenging. The current study uses CO 2 fluxes from clipped plots to determine R h fluxes (Eckhardt et al., 2019). To estimate in situ CH 4 production, the CH 4 fluxes above intact vegetation have been used before (Cooper et al., 2017). These CH 4 fluxes are the sum of CH 4 transported by aerenchymatous plants and by molecular diffusion through the soil and ebullition. This approach is likely underestimating in situ CH 4 production since a part of the produced CH 4 is oxidized in the soil before being emitted.
By using the fraction of oxidized CH 4 and the fraction of CH 4 diffusing through the soil from Knoblauch et al. (2015) and Preuss et al. (2013) we estimated that a fraction between 0.10 and 0.31 of produced CH 4 might have been oxidized in the soil before emitted into the atmosphere. Knoblauch et al. (2015) and Preuss et al. (2013) determined the fraction of oxidized CH 4 by using the δ 13 C-values of CH 4 from rhizospheric and emitted CH 4 . This method is based on the fact that isotopic fractionation occurs when CH 4 is oxidized during transport. Methanotrophs oxidize preferentially the lighter 12 CH 4 , leaving the heavier 13 CH 4 behind. Our estimated fractions of 0.10-0.31 oxidized CH 4 are within in the range (0.01-0.40) reported by previous studies carried out in similar environments, including sites dominated by sedges of Carex sp. (Nielsen, Michelsen, Ambus, et al., 2017;Popp et al., 1999;Ström et al., 2005). This large range is both due to differences in the methodology to determine CH 4 oxidation in the soil but also due to differences in soil conditions and vegetation composition.
Methane oxidation is closely related to the efficiency of oxygen transport by the plant's roots, which is a species-specific characteristic (Nielsen, Michelsen, Ambus, et al., 2017), and seemingly a more important factor determining CH 4 oxidation than plant biomass (Ström et al., 2005). The contribution of CH 4 oxidation varies throughout ecosystems and vegetation composition. A study in a wetland in South Sweden identified that 20%-40% of produced CH 4 was oxidized in a Carex rostrata-dominated peat monolith, and more than 90% in monoliths with Eriophorum vaginatum (Ström et al., 2005), while in a southern Estonian bog, no significant CH 4 oxidation was found in an Eriophorum spp. dominated site (Frenzel & Rudolph, 1998). Thus, rhizospheric CH 4 oxidation regulated by the efficiency of oxygen transport by the plant's roots is apparently a site-specific characteristic in addition to being species-specific. Sedge species have been related to higher CH 4 fluxes than other plant species, not because they support CH 4 production but because they facilitate CH 4 transport through their aerenchyma, bypassing the oxidative soil layer and avoiding oxidation (Green & Baird, 2012). Nielsen, Michelsen, Ambus, et al. (2017) showed in a wet tundra ecosystem in Greenland, that the radial oxygen loss of Carex aquatilis has a minor impact on CH 4 oxidation. In their study, the fraction of CH 4 oxidized was less than 2%, most likely because CH 4 diffusion is faster than it would be required for CH 4 oxidation and because CH 4 diffuses into plant roots at a lower depth than where oxidation takes place. Isotopic signatures of CH 4 from a bog in Stordalen Mire, Sweden, show a higher fraction of oxidized CH 4 occurring in periodically inundated shallower soil layers instead of permanently inundated deeper soil layers (Singleton et al., 2018). This is evidence of the latter hypothesis conceived by Nielsen, Michelsen, Ambus, et al. (2017) about the depth separation between the soil layers where CH 4 diffuses into plant roots and the region where CH 4 oxidation takes place.
Data from Samoylov Island, from a study on a similar polygon center, report no significant differences between the δ 13 C signatures of dissolved CH 4 in the anoxic soil and of the emitted CH 4 from plots vegetated by Carex aquatilis, and hence give no evidence for rhizospheric CH 4 oxidation (Knoblauch et al., 2015). Since the vascular plants in our studied polygon were dominated by C. aquatilis, which was shown to only weakly support CH 4 oxidation in the rhizosphere, we consider that the calculated range of rhizospheric CH 4 oxidation (0.10-0.31) is rather at the higher end of CH 4 oxidation under in situ conditions and that the measured CH 4 fluxes in the polygon center are not severely affected by CH 4 oxidation. Due to the lack of estimates of the fraction of CH 4 oxidized at the polygon rim, low fluxes measured, low presence of aerenchymatous plants and oxic conditions, we could not calculate such a range of rhizospheric CH 4 oxidation as we did for the polygon center. Thus, we cannot consider the CO 2 :CH 4 ratios as production ratios in the polygon rim.

Seasonal Heterotrophic Respiration and CH 4 fluxes
As expected, the wet tundra, which is water saturated, showed lower R h fluxes, which were measured with the dark chambers in the clipped plots, while higher R h fluxes were found in the dry tundra, and the opposite was the case for CH 4 , which was measured with the transparent chambers in the vegetated plots. The dry tundra dominates total carbon fluxes, both due to high R h and due to the larger area, since 185 ha is occupied by dry tundra and only 54.1 ha of the island is occupied by wet tundra (Muster et al., 2012). While both tundra types lost carbon preferentially in the form of R h , CH 4 -C represented 0.27% of the dry tundra total carbon flux and 6.91% of the wet tundra total carbon flux. The contribution of CH 4 fluxes to the seasonal R h and CH 4 fluxes on Samoylov might change in the future. The ALD at Samoylov Island is predicted to increase (Beermann et al., 2017), as well as at several Arctic sites, related to increasing temperatures affecting the extension of the thawing period (Andresen et al., 2017;Euskirchen et al., 2006;Strand et al., 2021). This trend can be detected at Arctic sites in the past and today (Andresen et al., 2017;Strand et al., 2021). As shown in this study, the importance of CH 4 fluxes increases with increasing ALD depth. Our results are consistent with the study of Rößger et al. (2022) who also found a significant relationship between ALD and CH 4 fluxes and evidence for an increasing trend of early summer CH 4 fluxes from wet tundra linked to atmospheric warming. Moreover, to predict the future response of CH 4 fluxes from thawing permafrost landscapes it will be crucial to understand if the ALD increase will result in wetter soils (Krogh & Pomeroy, 2019), thereby increasing CH 4 fluxes  or drier soils (Jin et al., 2021;Natali et al., 2015) resulting in lower CH 4 fluxes (Kim, 2015).

Conclusions
Methane fluxes presented here are at the lower end compared to other Arctic sites, and the differences found between these fluxes and the ones from other measurements in Samoylov Island show the high temporal and spatial variability found in GHG fluxes in the Siberian tundra. Understanding the mechanisms that control the CO 2 :CH 4 production ratios is especially important for improving Earth System models. To estimate the CO 2 :CH 4 production ratios under in situ conditions we used the R h fluxes from clipped plots and the CH 4 fluxes from vegetated plots, which were corrected for potential CH 4 oxidation in the soil. The CO 2 :CH 4 production ratio is associated with active-layer depth (ALD) due to a direct effect of ALD on methane production. The effect of air temperature seasonality on the CO 2 :CH 4 ratio is complicated. Topsoil (5 cm) warming stimulates heterotrophic respiration under more oxic conditions, hence increasing the CO 2 :CH 4 ratio. In contrast, warming of anoxic subsoil (40 cm) layers leads to enhanced CH 4 production, hence a lowering of the CO 2 :CH 4 ratio.
Arctic warming will lead to a warming of the active layer and its deepening. Our study indicates that for wet tundra this warming and deepening can result in a pronounced rise in CH 4 production. Changing vegetation patterns and functions, however, will further convolute the net response of Arctic CH 4 fluxes to global warming. Further studies are needed focusing on the complexities of in situ CH 4 and CO 2 production, especially in regions where there is yet scarcity of data, like the vast Siberian tundra. There is still a high uncertainty on the response of CH 4 production and fluxes to thawing permafrost and the related feedback mechanism.