Seasonal methane dynamics in three different Siberian water bodies

Arctic regions and their water bodies are being affected by the most rapid climate warming on 20 Earth. Arctic lakes and small ponds are known to act as an important source of atmospheric methane. However, not much is known about other types of water bodies in permafrost regions, which include major rivers and coastal bays as a transition type between freshwater and marine environments. We monitored dissolved methane concentrations in three different water bodies (Lena River, Tiksi Bay and Lake Golzovoye, Siberia, Russia) over a period of 25 two years. Sampling was carried out under ice cover (April) and in open water (July / August). The methane oxidation (MOX) rate in water and melted ice samples from the late winter of 2017 was also investigated. In the Lena River winter methane concentrations were a quarter of the summer concentrations (8 vs 31 nmol L-1) and mean winter MOX rate was low (0.023 nmol L-1 d-1). In contrast, Tiksi 30 Bay winter methane concentrations were 10-times higher than in summer (103 vs 13 nmol L1). Winter MOX rates showed a median of 0.305 nmol L-1 d-1. In Lake Golzovoye, median methane concentrations in winter were 40-times higher than in summer (1957 vs 49 nmol L1). However, MOX was much higher in the lake (2.95 nmol L-1 d-1) than in either the river or https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c © Author(s) 2020. CC BY 4.0 License.

bay. The temperature had a strong influence on the MOX, (Q10 = 2.72 ± 0.69) compared to 35 temperate environments. In the ice cores a median methane concentration of 9 nM was observed, with no gradient between the ice surface and the bottom layer at the ice-waterinterface. MOX in the (melted) ice cores was mostly below the detection limit. Comparing methane concentrations in the ice with the underlaying water column revealed 100 -1000times higher methane concentration in the water column. 40 The winter situation seemed to favor a methane accumulation under ice, especially in the lake with a stagnant water body. While on the other hand, in the Lena River with its flowing water no methane accumulation under ice was observed. Methane oxidation rate was not able to counteract this winter time accumulation.

1 Introduction
Worldwide, the mixing ratio of methane (methane) has been increasing rapidly since 2005 (Hartmann et al., 2013;IPCC, 2013). The reasons for the observed increases in atmospheric methane are unclear. A probable explanation is increased methane emissions from wetlands, both in the tropics (Nisbet et al., 2016) as well as in the Arctic (Fisher et al., 2011), and from 50 other Arctic water bodies (Walter Anthony et al., 2016;Kohnert et al., 2018) or geological methane emissions (Kohnert et al., 2017). In the Arctic, mean atmospheric methane mixing ratio increased by 6 ppb per year from 2001 to 2017 (Platt et al., 2018). Especially the Laptev Sea, in eastern Siberia, is generally a source of methane to the atmosphere, with a great interannual variability of sea-air fluxes of methane (Wåhlström et al., 2016). Also, Saunois 55 and co-authors estimated increased methane emission for freshwater systems and wetlands, but a better quantification of the emissions of different contributor (streams, rivers, lakes and ponds) is needed (Saunois et al., 2016).
Lakes are important sources of atmospheric methane on a regional to global scale (Bastviken et al., 2004;Cole et al., 2007), and their contribution is predicted to increase in 60 response to climate change and rapidly warming waters (O'Reilly et al., 2015;Tan and Zhuang, 2015;Wik et al., 2018). Most of the methane produced in lake sediments enters the atmosphere via ebullition (Bastviken et al., 2004;Walter et al., 2007), a temperature-sensitive transport mode with high spatial and temporal heterogeneity (DelSontro et al., 2015). The role of Arctic rivers as a methane source to the shelf seas is poorly described. Some studies 65 present rivers as strong methane sources (Morozumi et al., 2019), while other studies revealed a complex pattern of riverine methane input (Bussmann et al., 2017). https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License.
One major drawback from most of these studies is that sampling was conducted in the ice-free season, although most of the year Arctic water bodies are ice covered. Thus, the 70 seasonal variation could not be captured within those studies. The ice cover on lakes decouples the water body from the atmosphere and the circulation changes from wind driven to thermohaline. After ice formation a stable winter stratification is set up. As there is no more external oxygen supply, enhanced anaerobic degradation leads to accumulation of methane, H2S and NH3 (Leppäranta, 2015). In several lakes in Alaska and Canada, dissolved methane 75 was highest under the ice-cover, indicating that the spring ice off period is a large source of atmospheric methane (Townsend-Small et al., 2017;Cunada et al., 2018;Serikova et al., 2019). However, the hydrography of a lake is also an important factor to consider since stratification of the water column counteracts an intense gas exchange. Thus, examples are known for lakes with incomplete spring mixing and consequently a maximum gas exchange 80 in autumn during complete mixing (Deshpande et al., 2015).
Rivers of permafrost regions are characterized by an ice season of >100 days duration between autumn freeze-up and spring ice-off; for the Lena River it is >160 days (Shiklomanov and Lammers, 2014). Ice effects and the demobilization of liquid water result in very low discharge during winter freeze-up and runoff is lowest during late winter 85 (Lininger and Wohl, 2019). A decrease in ice thickness of the largest Siberian Rivers during the last 10 to 15 years (Shiklomanov and Lammers, 2014) could enhance the channel connectivity to subchannel and groundwater flow, causing an increase in winter base flow, as suggested by Gurevich (2009). An increased winter flow as well as increasing temperatures in the Lena River are also supported by Tananaev et al., (2016);Yang et al., (2002); Peterson et 90 al., (2002).
An important filter, counter-acting the methane flux into the atmosphere, is microbial methane oxidation. Methane can be oxidized under anoxic conditions close to sediment horizons where it is produced (Martinez-Cruz et al., 2017;Winkel et al., 2018) or during migration through the oxic water column to the atmosphere (Mau et al., 2017a;Bussmann et 95 al., 2017). Under ice cover, it is important to consider methane oxidation below ice as it may reduce the total amount of methane emitted to the atmosphere during ice-off. Active methane oxidation and a methanotrophic community has been shown for permafrost thaw ponds and lakes (Kallistova et al., 2019). Yet, the methane oxidation capacity in such lakes during ice cover with low temperatures and low oxygen concentrations is unknown. In a study covering 100 several boreal lakes, methane oxidation was restricted to three lakes, where the phosphate concentrations were highest (Denfeld et al., 2016). Rates of methane oxidation during the https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License.
winter have been found to be much lower than summer rates, yet, there is no clear consensus on the factors limiting methane oxidation in winter (Ricão Canelhas et al., 2016). Beside oxygen concentration, the geological background (i.e. yedoma type permafrost lakes versus 105 non-yedoma type lakes) also had a significant impact on the methane oxidation rate .
Our study tests the hypothesis that winter ice blocks methane emissions, leading to the accumulation of methane in the underlying water bodies. By studying hydrographically different water bodies (lake, river and sea), we expect insights into the influence of water 110 column dynamics on methane accumulation to result. In addition, we measure methane oxidation rates in the water column and in melted ice to assess oxidation as a potential in situ sink.
2 Study area 115 This study was conducted on the southern coast of Bykovsky Peninsula in northeast Siberia, Russia ( Figure 1). Thermokarst lakes in that area commonly originated in the early Holocene when surficial permafrost started to thaw, leading to accumulation of lake sediments with organic contents of about 5-30 % (Biskaborn et al., 2016;Schleusner et al., 2015).
Offshore of the Bykovsky Peninsula, part of the Yedoma Ice Complex is submerged 120 and subsea permafrost is currently degrading. The coastline erodes at mean rates of between 0.5 and 2 m per year and can intersect inland water bodies, draining them or leading to the formation of thermokarst lagoons (Lantuit et al., 2011). The sea ice season in Tiksi Bay and the Buor-Khaya Bay typically starts in late September / early October and ends in June (Angelopoulos et al., 2019). Due to its isolation behind Muostakh Island and Cape Muostakh, 125 sea ice tends to be preserved longer than in the Central Laptev Sea (Günther et al., 2015.).
The Lena River has a mean annual discharge of 581 km 3 /year. It is the 2nd largest Arctic river by annual discharge and the 6th largest globally. There are no dams on the mainstem of the Lena, but there is a dam on the Viluy River, one of the Lena's main tributaries (Holmes et al., 2013). The Lena's watershed is 2.6 million km 2 , of which 70.5% is underlain 130 by continuous permafrost (Juhls et al., in review). Most of the water is discharged during end of May and beginning of June when the ice in the rivers breaks up, but the Laptev Sea is still covered by sea ice (Holmes et al., 2012). The main Lena River branches enter the Buor-Khaya Bay through the northern and eastern part of the delta and through he Bykovskaya Channel with 20-25% of the Lena River water discharge (Charkin et al., 2011). The further 135 distribution of the river water in Buor-Khaya Bay is mainly driven by the atmospheric https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License. systems of the cyclonic or anticyclonic Arctic circulation (Thibodeau et al., 2014;Wegner et al., 2013).

Study sites
In the course of several expeditions to the Lena Delta (Siberia, Russia), we were able to repeatedly sample the same locations in winter and summer over the years (Table 1).
In September 2016, water samples were taken in the Bykovsky Channel and mouth of 145 the Lena River (Overduin et al., 2017). In April 2017 with ice cover on the water bodies, ice cores were taken at Lake Golzovoye, the Lena River and Tiksi Bay, as well as the water below the ice sampled. Lake Golzovoye is an oval-shaped thermokarst lake about 0.5 km in diameter with a maximal depth of 10 m, surrounded by Yedoma uplands at various stages of degradation and with no ice grounding in its center (Spangenberg, 2018;Strauss et al., 2018). 150 Tiksi Bay is a shallow brackish bay at the southern end of Buor Khaya Bay, but still strongly influenced by the Lena River outflow. The water column is usually stratified, with a colder, more saline water underlying the brackish surface layer . The water of the Lena River was sampled near Samoylov island. In July and August 2017, we sampled the same locations (Lena River near Samoylov, Tiksi Bay and Lake Golzovoye) with open water 155 (Strauss et al., 2018). The transect to the "outer" Tiksi Bay has been investigated repeatedly over the previous years (Bussmann, 2013;Bussmann et al., 2017). In April 2018, again under ice samples were taken from the Lena River (Kruse et al., 2019).

Water sampling for methane analysis
In winter, water samples from below the ice were taken with a 1 L water sampler (Uwitec Austria), transferred to 0.5 L Nalgene bottles. In the field camp, the water was transferred with a 60 ml syringe into glass bottles, closed with butyl stoppers and crimps, poisoned with 170 0.2 ml 25% H2SO4 and stored upside down at 4 °C. In summer the water samples were directly transferred from the water sampler into the glass bottles, and further processed as described above.
In winter 2017, due to problems of freezing, 40 ml of sample water was shaken for 2 min with 20 ml nitrogen in a 60 ml syringe. This headspace was then transferred into glass 175 bottles filled with a saturated NaCl solution. Comparative measurements showed no significant difference between these two sampling strategies (Triputra, 2018). This dataset has been published already https://doi.pangaea.de/10.1594/PANGAEA.905776 .

Ice cores 180
In winter 2017, in addition to water sampling we also investigated ice cores for their methane content and methane oxidation rates. For each station three ice cores were drilled with a Kovacs Mark II ice coring system (9 cm diameter). One core was used to measure the in-situ temperature and back-up, one core was used for methane analysis, the third core was drilled for later molecular analyses. For determining the methane oxidation rates and later methane 185 concentrations the ice cores were processed at the Research Station Samoylov Island. The top 10 cm, a 10 cm mid-section and three 10 cm-sections at the bottom of the core were cut off and transferred to special PVDF gas sampling bags (Keikaventures). The remaining parts of the cores were kept frozen. The bags were evacuated and the cores melted within approx. 5 h in a water bath at 8 °C. The sample was distributed with a 60 ml syringe to 3 glass bottles for 190 determination of MOX and 1 bottle (which was poisoned immediately with H2SO4) for analysis of methane concentration (Strauss et al., 2018).

Methane analysis
Methane concentrations were determined via head-space method by adding 20 ml of N2 as 195 headspace. Headspace methane concentrations were analyzed in the home laboratory with a gas chromatograph (GC 2014, Shimadzu) equipped with a flame ionization detector and a Shinycarbon column (Restek, USA). The temperatures of the oven, the injector, and detector were 100, 120, and 160 °C, respectively. The carrier gas (N2) flow was 20 mL min -1 , with 40 mL min -1 and 400 mL min -1 synthetic air. Gas standards (Air Liquide) with methane 200 concentrations of 10 and 100 ppm were used for calibration. The calculation of the methane concentration was performed according to (Magen et al., 2014). The precision of the calibration line was r 2 = 0.99 and the reproducibility of the samples was < 5%.

Methane oxidation rate (MOX) 205
The MOX rate was determined by adding radioactive, tritiated methane to triplicate samples (Bussmann et al., 2015). The principle of the MOX rate determination is based on the ratio of produced tritiated hydrogen from the added tritiated methane. This ratio corrected for the https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License.
incubation time gives the fractional turnover rate (k' [d -1 ]). To obtain the MOX rate, k' is multiplied with the in-situ methane concentration. Radioactive, tritiated methane (0.1 ml of 210 3 H-methane, 2 kBq ml -1 , American Radiolabeled Chemicals) was added to triplicate samples.
Samples were incubated for 60 h at 1 °C, in the dark. Incubation was stopped by adding 0.2 ml of 25% H2SO4. Abiotic controls were poisoned before adding the tracer. Radioactivity was determined with a liquid scintillation counter (Triathler, Finnland) and Ultima Gold (Perkin Elmer) as a scintillation cocktail. For MOX, the limit of detection was calculated as described 215 in Bussmann et al. (2015) and was determined to be 0.009 nmol L -1 d -1 for this data set.
In a set of experiments, we also assessed the influence of temperature on the MOX rate. Water samples (Main Channel, April 2017) that were incubated at temperatures from 1; 4; 7 and 10 °C show the temperature curve of the MOX reaction. At a previous cruise off Svalbard identical experiments were performed (Mau et al., 2017a) with water from stations 220 HE449-CTD-2, 10, 33 and 37 incubated at temperatures from 0; 4; 8; 13 and 20 °C. We determined the Q10-factor, which indicates the temperature dependence of a biological process according to (Raven and Geider, 1988): where Tis is the in-situ temperature and "m" the slope of the regression line of the Arrhenius plot (the inverse of the absolute temperature vs. the natural logarithm of the MOX rate).

Hydrochemistry and hydrology 230
Profiles of water temperature and conductivity were measured with a Cast-away CTD (SonTek) in summer 2016, and in winter and summer 2017. Water depth measurements were made with an echo-sounding device (Garmin), every 10 m along the profile. Water velocity was measured in three horizons at each vertical profile: 0.2H; 0.6H; 0.8H, with H = total water depth) using a hydrological speed recorder (GR-21). Water discharges were calculated 235 according to the recommendations for Russian hydrometeorological stations (Fedorova et al., 2015). In Lake Golzovoye, in summer 2017, the median methane concentration was 49 nM (n = 5), while in winter concentrations were about 40-times higher with a median of 1957 nM (n = 8, Wilcoxon test, p = 0.002, Figure 2). In winter, the water temperature of the lake was cold at the surface (median 0.3 °C in top 1 m) and warmer (2 °C) at the bottom with a thermocline 270 at 2 meters. In summer, the surface water was heated to 7 -10 °C, while the bottom water only warmed to 4 -5 °C.

285
Ice cores were taken at Lake Golzovoye and in Tiksi Bay, whereas no ice core data are available for the Lena River itself (Strauss et al., 2018). Methane concentrations in the ice cores of Lake Golzovoye and Tiksi Bay were rather low (both with a median of 9 nM). No depth gradients from the ice surface, middles section and the three lower most sections were evident. On a close up for the bottom layers, there was a slight increase of methane towards 290 the ice-water interface for the ice cores from Lake Golzovoye, but not from Tiksi Bay. In higher than in the ice cores, for Tiksi Bay and Lake Golzovoye, respectively, with a median of 102 nM and 985 nM (Bussmann and Fedorova, 2019). However, in one core of Lake 295 Golzovoye (core #24), concentrations were orders of magnitude higher throughout the core

Methane oxidation rates (MOX)
Methane oxidation rates were determined in the melted water from the ice cores and in water from the Lena River. Due to logistic restraints at the field sites, no direct measurements of https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License.
MOX in the waters of Lake Golzovoye and Tiksi Bay was possible. In the first step we 310 determined the fractional turnover rate k'. In the ice cores, k' was very patchy distributed and 73% of all samples were below the detection limit. In all positive samples, the value for k' was rather stable with a median of 0.003 per day (n = 31). Even though very patchy and many samples being below the detection limit, in some samples of the ice cores a low MOX rate could be detected (on average 0.011 nmol L -1 d -1 ). In the water samples below the ice, the 315 fractional turnover rate k' was never below the detection limit. The median k' determined for water samples (Lena River) was also 0.003 (n = 14).
To calculate the MOX for all water samples, we multiplied the median k' of 0.003 from Lena River water samples with the respective in-situ methane concentration of the water columns of Lake Golzovoye, Tiksi Bay and Lena River. Highest MOX rates were found in 320 the water below station #24 in Lake Golzovoye (20.36 nmol L -1 d -1 ), where also very high methane concentrations were observed ( Figure 5). The median MOX in the other water samples of Lake Golzovoye was about one order of magnitude lower (2.95 nmol L -1 d -1 ). In Tiksi Bay and in the Lena River MOX rates decreased by further orders of magnitude (median of 0.305 and 0.023 nmol L -1 d -1 , Figure 5). 325

Temperature influence on MOX
To assess the influence of temperature on the MOX, we incubated water samples at different temperatures and determined their MOX rate. As expected, with increasing temperature the MOX rate also increased ( Figure 6). The Q10 calculated for these water samples from the Lena River was 2.72 ± 0.69. For polar and marine waters off Svalbard a Q10 of 2.99 ± 0.86 335 was calculated. In winter the methane concentrations in the Lena River were 4times lower as in summer ( Table 2). The Lena River displays a reduced but still substantial water flow / discharge under ice cover. In 2017, the discharge in winter (March and April with 2.830 and 2.185 m 3 sec -1 ) was about 10-times lower than in summer (July and August with 29.800 and 26.000 m 3 sec -1 , (Shiklomanov et al., 2018). poor in organic material (Rivera et al., 2006) and we do not expect any in situ methane production. Especially in the Lena River a substantial amount of POC originates from 355 thermokarst-induced, abrupt collapse of Pleistocene Ice Complex deposits. These events mainly occur in late summer, the signal however, is still visible in winter (Wild et al., 2019).
With these collapses also methane could be imported to the river. So, we could only detect low methane concentrations in winter. In contrast, in summer, the active layer soils from the drainage basin allow for several sources of methane and thus increased methane 360 concentrations in summer. Also, methane could be transported from the southern Lena catchment towards our study area, as is suggested for particulate organic matter (Winterfeld et al., 2015). At least during the warm season, methane production from (temperate) river sediments is possible (Bednařík et al., 2019). About the river-ice on the Lena there is no information yet available. 365 In Tiksi Bay, we observed a 7-times increase of dissolved methane under ice cover, compared to open water conditions (Table 2). Tiksi Bay is part of Buor Khaya Bay and via the central Laptev Sea perennially connected to the Arctic Ocean. Not much is known about tidal surge or water movement in Tiksi Bay under ice cover. It is anticipated that the ice cover on Tiksi Bay will lead to a decrease in tidal amplitudes and velocities (Fofonova et al., 2014). 370 The structure of ice formation in Tiksi Bay also suggest that even in winter it is still an open system connected to the outer bay (Spangenberg et al., submitted). Sources of methane could be through diffusion of methane from the underlying sediment (Bussmann et al., 2017) by further degradation of organic material. However, as aerobic methane oxidation in the water column is impaired by low temperatures, methane concentrations in water increase. 375 In Lake Golzovoye, dissolved methane concentrations increased by 40-times from summer to winter (Table 2). Lake Golzovoye is an isolated freshwater lake with presumably only a weak thermohaline circulation (Leppäranta, 2015;Spangenberg et al., submitted). A similar, seasonal pattern of methane among lake waters of the Mackenzie Delta has been observed, ranging from very high concentrations at the end of winter beneath lake-ice (> 380 200.000 nM) to considerably lower concentrations during open water, particularly by late summer (approx. 1000 nM) (Cunada et al., 2018). For Yedoma lakes in Alaska, the same pattern was observed with high methane concentrations in winter, and a decrease of more than one order of magnitude in summer, (even though the actual concentrations > 900,000 nM were much higher than in this study) Sepulveda-Jauregui et al., 385 2015). The same pattern is reported for boreal Swedish lakes (Denfeld et al., 2018). https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License.
The role of water velocity and water column mixing is not clear, but our data suggest that in a stagnant water body such as a lake more methane accumulates under ice than in a water body with running water (river). Water column turbulent diffusivity has a major influence on the methane cycle, where higher turbulence potentially leads to greater 390 proportion of methane being oxidized and lower turbulence leads to a greater proportion being stored (Vachon et al., 2019).

Ice cores
The median methane concentration of all ice cores for Lake Golzovoye and Tiksi Bay was 9 395 nM, which was supersaturated compared to atmospheric concentrations for which equilibrium concentration would be 5 nM. More details on the ice formation in the different water bodies are given in (Spangenberg et al., submitted). Compared to the methane concentrations in the water, the concentrations in the ice were 1 -2 orders of magnitude lower (Table 2). This means that, in terms of methane, a complete separation of the water body from the 400 atmosphere can be assumed. As mentioned earlier in this study and in (Spangenberg et al., submitted), core #24 in Lake Golzovoye had much higher methane concentrations throughout the core and visible inclusions of (methane) bubbles. We assume that core #24 was located above an active ebullition site, which might have slowed ice formation and prolonged direct methane release to the atmosphere. 405 In the ice itself, 28% of the samples showed methane oxidation capability. During ice formation most free-living bacteria are lost from the liquid phase through incorporation into the ice, while bacterial aggregates remain in the water (Santibáñez et al., 2019). In an experimental setup, (Wilson et al., 2012) show that multiple freeze-thaw cycles in water from freshwater lakes reduce the total bacterial cell number at least 100,000-fold. In addition, 410 methanotrophic bacteria are particular sensitive to freezing and thawing (Green and Woodford, 1992;Hoefman et al., 2012). These findings could explain the reduced activity of methanotrophic bacteria within the ice cores.
Also, we did not detect any discolorations or other indications of photosynthesis or other biological processes in the bottom layer of the ice cores. Thus, we conclude that the 415 ecosystem of freshwater ice and its lower margin do not reach the richness observed in polar sea ice (Leppäranta, 2015).

Methane oxidation rates (MOX)
In this study we determined the in situ methane oxidation rate with tritiated methane as tracer. 420 The advantage of the tracer injection method is that natural low concentrations are hardly https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License. altered and thus we assume that our values are close to the actual in situ rates. The fractional turnover rate k' was determined only in the ice cores and in river water; within these k' was evenly distributed. Therefore, this factor was applied to all water samples, i.e. even to the brackish water samples from Tiksi Bay. 425 Our data span three orders of magnitude, ranging from 0.02 nmol L -1 d -1 in the Lena River, 0.31 in Tiksi Bay and 2.95 nmol L -1 d -1 in Lake Golzovoye. Another polar study finds a MOX of 0.004-1.09 mg C m -3 d -1 ( = 0.33 -91 nmol L -1 d -1 ) at the water-ice interface of a Swedish lake (Ricão Canelhas et al., 2016). Also (Bastviken et al., 2002) report a MOX of similar range (0.001-39 mg C m -3 d -1 = 0.08 -3250 nmol L -1 d -1). However, the methods of 430 determining MOX were quite different, and more of a potential rate. Another study in Yedoma lakes in Alaska also reports higher MOX (0.03 -0.28 mg methane L -1 d -1 = 1875 -17500 nmol L -1 d -1 ) with a kinetic approach to determine MOX .
In marine polar waters off Svalbard MOX was determined with the same tracer method and ranged from 1.6 -2.2 nmol L -1 d -1 in summer (Mau et al., 2017b). Thus, our data are within 435 the very low range of reported MOX rates in polar regions, probably due to methodological differences.
In previous years we determined MOX in the study area during summer, applying the same method as in this study. Therefore, we can approach a seasonal comparison (winter vs. summer), assuming interannual variability is negligible. For the Lena river and permafrost 440 lakes we compare our winter data with summer data obtained in July 2012 (Osudar et al., 2016). For the Lena River, with a median MOX of 22.8 nmol L -1 d -1 (n = 8), the summer MOX was about 3 orders of magnitude higher than in winter (0.023 nmol L -1 d -1 ). For MOX in lakes, summer rates from small lakes near Research Station Samoylov Island were 36times higher (median 107 nmol L -1 d -1 , n = 6) than winter rates (2.95 nmol L -1 d -1 ). Also, for 445 Yedoma lakes in Alaska summer values of MOX were about 10-times higher than winter values . For Tiksi Bay there are also summer values of MOX available (Bussmann et al., 2017). However, with a median summer rate of 0.419 nmol L -1 d -1 for surface, riverine water there is little difference when compared to our winter data (0.31 nmol L -1 d -1 , Table 2). 450 https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License. **comparing methane concentration in lowest ice core layer with underlying water *** winter data from this study, summer data from (Osudar et al., 2016) and (Bussmann et al., 2017) There still seems to be no clear consensus on the factors limiting MOX in winter. In 460 several boreal lakes MOX was restricted to lakes, where the phosphate concentrations were highest (Denfeld et al., 2016). Another study reports that in winter MOX is mainly controlled by the dissolved oxygen concentration, while in the summer it was controlled primarily by the methane concentration . The stratification of lakes determining the availability of methane and oxygen for the methanotrophic bacteria also strongly 465 influences MOX (Kankaala et al., 2006;Kankaala et al., 2007).
But temperature is also an important factor affecting winter MOX. MOX is observed at temperatures of 2 °C, mostly by Methylococcaceae (Ricão Canelhas et al., 2016). A recent study with in situ concentrations in a northern temperate lake observes a Q10 of 2.4 ± 1.4 (Thottathil et al., 2019), which is a bit lower than the Q10 of this study with 2.7 and 2.9 for 470 polar, fresh and marine water, respectively. In addition, there are co-correlations between temperature and methane concentration. At substrate (methane) saturation, temperature has a strong influence, while under substrate limiting conditions, temperature has a minor influence (Lofton et al., 2014;King and Adamsen, 1992). In contrast, (Thottathil et al., 2019), observed a strong temperature response for MOX across the entire range of ambient methane 475 concentrations. Measuring MOX with tritiated methane and thus at close to in situ concentrations, we can compare the Q10 from this study with data from temperate environments. The Q10 for polar environments was higher than Q10 values obtained from https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License. temperate waters (Elbe River, Germany and North Sea, (Bussmann et al., 2015).) with 1.52 and 1.75, respectively (Figure 7). Although the substrate concentration in the temperate 480 waters was higher, MOX from polar regions seems to react more sensitively to a temperature increase. One explanation could be that the temperature optimum of psychrotolerant methanotrophs is below 20 °C (Bale et al., 2019). Thus, the polar methanotrophs are further away from their optimum temperature and react with increased activity to temperature increases while methanotrophic bacteria from temperate waters are nearer to their optimum 485 temperature and do not react as sensitively.
Environmental conditions between winter and summer conditions certainly differ and may also affect the population structure of methanotrophs. Some psychrophilic strains adapt to colder temperatures (20 °C versus 4 °C) by modifying their fatty acid composition (Bale et al., 2019), while others may have only limited abilities, resulting in different population 490 structures. This is supported by a recent study revealing that wintertime Arctic bacterial communities and food webs structure change based on carbon availability (Kellogg et al., 2019).  Bussmann et al. (2015), data for the temperate lake are from Thottathil et al. (2019).
In winter when the water bodies are ice covered and methane emissions are stopped, 500 MOX in the water would be the only sink left. However, MOX under winter conditions is reduced, either by low temperatures (this study), low oxygen or phosphate content (Martinez-https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License. Denfeld et al., 2016). The turnover time of MOX (the time needed to consume all of the present methane) is the inverse of our determined median k', i.e. 333 days.
Thus, with an estimated ice coverage of 200 days, MOX will not contribute significantly to a 505 methane reduction, and about 60% of the methane inventory can be consumed. If we consider the winter and summer conditions as two extremes (low concentration and high MOX in summer versus high concentrations and low MOX in winter); we assume that after ice-on, both parameters in/decrease in a linear way. Thus, the duration of ice cover is important and we assume that a shorter ice cover results in less high methane concentrations under ice and 510 subsequently in a reduced pulse of methane emissions at ice-off. In a changing, warming Arctic, a shortened time of ice coverage on lakes and rivers is predicted (Prowse et al., 2011).
In respect to our study this would imply a shortened time for methane accumulation under ice and a shortened time for the less efficient winter-MOX. Thus, an extended time of ice-free conditions could reduce the methane emissions from Arctic water bodies. However, it has to 515 be kept in mind that not much is known about the MOX during water column mixing in spring or autumn. At least one study reports MOX as the dominant methane sink at ice-off (Utsumi et al., 1998b).

6 Conclusions
Our work on an eastern Siberian lake, river and marine bay showed that methane accumulates under ice cover during the winter and is consumed differently in the three water bodies. Our study was restricted to late winter and mid-summer, which represents two extremes of the annual cycle, following the ice-covered and summer periods. Other processes during autumn 525 mixing, ice-on, ice off are not considered.
Two main physical factors affecting the methane cycle in the water bodies under ice cover are the water velocity and the ice cover itself. In most of our ice cores no concentration gradient between the bottom of the ice cores and the top was obvious. As we could hardly detect any MOX within the ice cores, we assume that methane is not integrated into the ice 530 during freeze up. Therefore, the ice cover seems to effectively prevent any methane flux from the highly accumulated methane concentrations in the water towards the atmosphere. In the river with running water under the ice cover, only a minor accumulation of methane was observed. In the bay with a restricted but still present water movement, dilution or mixing with other water bodies, allowed for a moderate accumulation of methane. In the small lake, 535 https://doi.org/10.5194/bg-2020-106 Preprint. Discussion started: 22 April 2020 c Author(s) 2020. CC BY 4.0 License.
we assume a stagnant water body with a subsequent accumulation of high amounts of methane.
The biotic counterpart of the observed methane accumulation is microbial methane oxidation (MOX). In most cases, MOX in summer was much higher than in winter. We observed a strong dependence of MOX on the temperature, and with in-situ temperatures of 540 only 1 °C in winter subsequently low rates were observed. Higher methane concentrations in winter indicate the methanotrophic bacteria were not limited by substrate (methane) concentrations. However, especially in Lake Golzovoye and its stagnant water body oxygen could become a limiting factor. Other factors could be nutrient limitation or shift in the population structure. 545 In a changing, warming Arctic, a shortened time of ice coverage on the water bodies is predicted. In respect to our study this would imply a shortened time for methane to accumulate below the ice and a shorter time for the less efficient winter-MOX. Thus, an extended time of ice-free conditions could reduce the methane flux from the Arctic water bodies. 550 Data availability: Data on methane concentration and MOX will be submitted to the Pangea database.

555
Acknowledgements: This study was part of the Helmholtz program PACES, Topic 1.3. We are thankful to the logistics department of the Alfred Wegener Institute, particularly Waldemar Schneider.