Methane excess in Arctic surface water- triggered by sea ice formation and melting

Arctic amplification of global warming has led to increased summer sea ice retreat, which influences gas exchange between the Arctic Ocean and the atmosphere where sea ice previously acted as a physical barrier. Indeed, recently observed enhanced atmospheric methane concentrations in Arctic regions with fractional sea-ice cover point to unexpected feedbacks in cycling of methane. We report on methane excess in sea ice-influenced water masses in the interior Arctic Ocean and provide evidence that sea ice is a potential source. We show that methane release from sea ice into the ocean occurs via brine drainage during freezing and melting i.e. in winter and spring. In summer under a fractional sea ice cover, reduced turbulence restricts gas transfer, then seawater acts as buffer in which methane remains entrained. However, in autumn and winter surface convection initiates pronounced efflux of methane from the ice covered ocean to the atmosphere. Our results demonstrate that sea ice-sourced methane cycles seasonally between sea ice, sea-ice-influenced seawater and the atmosphere, while the deeper ocean remains decoupled. Freshening due to summer sea ice retreat will enhance this decoupling, which restricts the capacity of the deeper Arctic Ocean to act as a sink for this greenhouse gas.

Scientific RepoRts | 5:16179 | DOi: 10.1038/srep16179 concerning the methane cycle in the Arctic e.g. the buffering capacity of the winter mixed layer in summer and the isolation of the deeper Arctic Ocean. The permanent isolation influences the sink capacity of the Arctic Ocean which becomes relevant for methane released by terrestrial sources in the Arctic. Hence aiming on the interaction between sea ice and sea water is fundamental and has up to now not yet been considered when it comes to understanding the balancing of trace gas fluxes in the ice covered ocean. Concerning the pronounced seasonality in trace gas fluxes our observations highlight the differences between trace gas exchanges in ice-covered Polar Regions relative to open water areas in lower latitudes.

Results
The current study is based on the analysis of concentrations, stable carbon isotope ratios and aerobic oxidation rates of methane in surface water and sea ice brine coupled with the hydrographic data along two transects in the central Arctic Ocean occupied by RV Polarstern between early August and late September 2011 ( Fig. 1 and methods).
Water masses. Sea ice influenced surface water encompasses the winter mixed layer (WML) and the melt water layer (MWL). The MWL comprises the upper 20 m and is formed by melting summer sea ice (Fig. 2a,b). The WML is formed by convective overturning due to brine rejection during freezing notably with temperatures at the freezing point. This temperature minimum is even sustained in summer as warming does at many places not reach the depth of this winter mixing thus leaving a local temperature minimum and indicating regions where the WML is formed [13][14][15] (Fig. 2a). Isolated by a strong halocline, sea water below the WML consists of upper halocline water (UHW, > 60 m). The Arctic Ocean upper halocline is being created and advected from the Chukchi shelf and/or from the Bering Sea inflow and is not affected by the local surface processes 16 . The UHW is clearly detectable by the pronounced decrease in oxygen concentration (Fig. 3).
Methane super-saturation. In the interior Arctic methane is super-saturated in the WML and MWL and under-saturated below the halocline in UHW (> 60 m) (Fig. 4). We suggest that local and seasonal processes interacting with sea ice create the heterogeneous levels of methane excess. Along transect 1, methane is incorporated at the depth of the pronounced temperature minimum between 40-60 m. By comparison at transect 2 the WML contains less methane and also the temperature minimum is clearly less pronounced, indicative of lower brine rejection during winter (Figs 2b and 4). These observations indicate that sea ice may act as a methane source and brine release as the pathway to transport methane into the WML. The brine-released methane remains stored in the WML as haline stratification restricts downward mixing. Hence the methane excess in the WML is detected even in the summer while the  The red dotted rectangle shows the range of the temperature minimum, i.e. the depth of winter mixing on the stations with methane excess at this depth. The temperature minimum is formed during winter when freezing induces brine rejection and haline convection but is even sustained in summer [13][14][15] . The grey dotted rectangle at the salinity profiles show the thickness of the layer, which becomes fresher by seasonal ice melt and homogenised by wind mixing (Transect 2 see UHW remains isolated and methane under-saturated due to the decoupling (Fig. 4). The MWL (< 20 m) is less methane super-saturated than the WML. This perplexing circumstance points to the different scenarios, which create the excess in both water masses. Micro bubbles included in the brine may be entrained and then re-dissolved in the less saline seawater while the dilution with fresh water doesn't occur in winter. In comparison methane in the MWL is released when basal melting starts but the top of the sea ice still remains impermeable inhibiting direct methane release from sea ice to the atmosphere. Then seawater flushes the brine channels and gas bubbles included in the brines will be re-dissolved. Brine super saturated with methane will then be incorporated into the MWL. However, the MWL is also diluted by fresh water from melting sea ice, thus reducing the final methane concentration therein. δ 13 C signature from methane. The stable carbon isotopic signature of methane shows that the atmosphere and surface water in the central Arctic Ocean is not equilibrated. Deviations in the δ 13 C signature from the atmospheric background illustrate the strong evidence for the presence of additional sources of methane in surface water. We compare the measured data with calculated curves in order to establish whether the methane pool in seawater is modified by oxidation and mixing (Fig. 5). A Rayleigh distillation model. i.e., δ 13 C CH4 = 1000* (1/α − 1)* ln f + (δ 13 C CH4 ) 0 is used to calculate the δ 13 C signature of methane which results when methane is being microbially consumed (Fig. 5a). (α , the isotope fractionation factor, is 1.008 10 , f is the fraction of methane remaining, (1-f) is the methane consumed by oxidation and (δ 13 C CH4 ) 0 is the initial isotopic composition). RC-1 (Fig. 5a, dashed line) shows the δ 13 C signature of methane that results when consumption of initial atmospheric methane, which has equilibrated in surface water, occurs. In that case (δ 13 C CH4 ) 0 is − 45‰ corresponding to the atmospheric δ 13 C CH4 value (− 47‰) 17 corrected by the kinetic isotopic fractionation (KIE) effect 18 . Methane residing in UHW (Fig. 5a, open circles) fits to RC-1, and indicates that the isotopic signature of this methane corresponds to that of original atmospheric methane. In comparison, methane in the winter mixed layer and melt water layer (WML and MWL, 0-60 m, blue open squares) is more depleted in 13 C and deviates clearly from the RC-1 curve. These values are localized between RC-1 and a second Rayleigh curve (RC-2, black line, Fig. 5a), which results from consumption of methane with (δ 13 C CH4 ) 0 − 75‰, i.e. the end member of methane detected in brines (blue squares). A Rayleigh curve describes ideal conditions, assuming that only one source and only one sink (oxidation) exists. The measured data have been compared with calculated curves in order to check if the methane pool in seawater is modified by oxidation. Deviations of the measured values from the calculated Rayleigh curves reflect the influence of additional processes, i.e. mixing with sources with different isotope signatures and sea-air flux. Since not only oxidation but a combination of several processes modifies the isotopic signature of methane in sea water we also displayed our data using a Keeling type plot.
A Keeling type plot is used (Fig. 5b) to calculate the mixing between two assumed end-members. The calculated mixing trend is compared with the measured values to illustrate potential mixing of methane of different isotopic signatures. Mixing end points are: atmospheric methane and residual methane. Residual methane is the pool of initial atmospheric methane, which has been oxidised up to the threshold value where microbial consumption finally ceases because of substrate limitation. Methane residing in UHW scatters around the mixing line between atmospheric and residual methane as end members and exerts an important control on the distribution of the isotopic ratio of methane from these precursors. In comparison methane residing in the WML and MWL does not fit the mixing line, which is evidence for the presence of additional sources.

Discussion
Along both transects we observed distinct lateral and vertical variations both in concentration and in the δ 13 C signature of methane and suggest that the surface Arctic Ocean is not equilibrated with the atmosphere. Methane is commonly super-saturated down to the WML. Being isolated by a strong salinity gradient the underlying water masses are homogeneously under-saturated with respect to the atmospheric equilibrium concentration (Fig. 4). They mainly consist of upper halocline water (UHW) as indicated by the low oxygen saturation 16 (Fig. 3). This signature is formed in the Chukchi Sea. Consequently, a lateral transport of methane in sea water from the shelf areas into the central Arctic can be ruled out.
Compared to the atmospheric background 17 , methane in sea water is enriched in 13 C. The preferred microbial use of 12 C makes methane consumption the most likely cause of this isotopic shift 19 . In addition mixing of methane of different isotopic signatures from other sources may contribute to deviations from the background signal. Hence, to trace the isotopic fractionation, which potentially occurs when methane with different initial δ 13 C signatures is being consumed or mixed, we calculated a Rayleigh distillation and Keeling type plot (Fig. 5). Differences between methane residing in UHW and methane residing in sea ice-influenced water (i.e. WML and MWL) become obvious by comparing the measured values with calculated curves. Methane residing in UHW scatters around an oxidation line and a mixing line between atmospheric and residual methane (i.e. threshold value) as end members and exerts an important control on the distribution of methane from these precursors. Decoupling from the atmosphere and simultaneous long-term methane consumption creates this pool of under-saturated and 13 C enriched methane. In comparison, methane from sea ice-influenced water (i.e. WML and MWL) is more depleted in 13 C than methane in UHW and values scatter around an oxidation line with methane from brine as end member (Figs 4 and 5).
The large potential of sea ice to discharge methane is evident in methane super-saturation of up to 8000% in brine (Fig. 1). Furthermore, considerable alterations in the δ 13 C signature (− 36‰ to − 75‰) of methane in brine are indicative of extensive methane cycling within sea ice i.e. production and consumption, as both processes induce isotopic fractionation. Oxygen depletion in brine channels 20 and microbial production of up to 300 mg C mg 2 d −1 21 favour anaerobic degradation of organic matter and are the most likely preconditions for methane production in sea ice.
Methane in sea ice may be present in two phases; dissolved in brine or most likely in small gas bubbles formed during freezing. Temperature and salinity changes lead to a sharp decrease in methane solubility, in which case a much larger amount of methane will accumulate as gas in the ice 12,22 . In winter, sea ice is to a large extent impermeable and the ice-air gas transfer to the atmosphere becomes severely restricted 12,23 . Then the discharge of methane dissolved in brine becomes relevant. Also micro-bubbles included in the brine may be entrained and then re-dissolved in the less saline sea water. The depth of winter mixing by brine rejection is clearly indicated by the temperature minimum even in the following summer [13][14][15] (Fig. 2). Hence we conclude that methane ejected with brine during winter becomes dispersed in the WML and may remain stored until the following summer (Fig. 4).
In spring, methane in the ice may be released to the MWL, when basal melting starts but the top of the sea ice still remains impermeable, inhibiting direct methane release from sea ice to the atmosphere. Subsequently, sea water or melt water will flush the brine channels and gas bubbles included in the brines will re-dissolve. Several freezing and melting events may contribute to locally different levels of super saturation in the MWL (Figs 4 and 6).
In summer, the methane excess remains preserved, as on the one hand, haline stratification restricts downward mixing and on the other, the reduced turbulence and restricted gas transfer in the presence of sea ice 8,24 impedes the efflux (Fig. 6). Excess methane is also conserved by slow methane consumption. Seawater sampled directly under the ice shows methane oxidation rates at the lower end of reported data (0-820 nM/d) 25 , i.e. rates with a median 0.013 nM/d using a 3 H-CH 4 tracer and with a median of 0.0009 nM/d with a 14 C-CH 4 tracer, respectively. Although both tracers increase ambient methane concentrations by 1.7 nM and 457 nM, respectively, an increase in uptake rate due to an increase in substrate was not observed over three days incubation. These pulse chase experiments illustrate that the abundance and activity of methanotrophic bacteria is low in the water investigated. Methanotrophs might be limited due to unfavourable growth conditions (e.g., low temperature or lack of copper 26,27 ) since only bacteria with a low uptake velocity inhabit surface water. A weakly developed methanotrophic community due to restricted mixing with other water masses could also prevent the seeding with methanotrophic bacteria 28 .
In autumn, cooling and freezing generate convective mixing down to the halocline that will both homogenize methane excess in surface sea water and initiate methane efflux allowing the cold water to equilibrate with the atmosphere. After a new sea ice cover has formed and methane super saturation is still present, methane efflux may occur from leads or fractional sea ice cover during the entire winter (Figs 4 and 6).
The methane surplus in sea ice-influenced water (i. e. WML and MWL) ranges between 1.5 and 1.9 nM relative to the methane equilibrium concentration (see methods). By extrapolating the methane surplus in the top 10 m to the area of seasonal sea-ice melting (6.4 million km 2 ), we calculate that between 7.5 and 10 Gg of methane may be stored in sea ice during summer. This estimate should be considered as being on the low side, since only the surplus methane near the surface is taken into account. Methane stored in sea ice will exceed this amount considerably but to confirm this, more measurements are needed. Summer sea ice retreat may change the amount of excess methane eventually making estimates of future atmospheric emissions more uncertain. In addition, increasing summer sea ice retreat strengthens the density stratification in the upper ocean and consequently the decoupling of sea iceinfluenced water masses from the UHW. The permanent isolation of the UHW from the atmosphere becomes evident by the homogeneous methane under-saturation (Figs 4 and 6). Several years are needed for UHW to cross the Arctic Basin 29 . During that long journey on-going microbial methane consumption, albeit slow, creates isotopic fractionation, which eventually enriches the residual methane in 13 C. Thus not only does the concentration steadily decrease during the long-term water mass separation, but methane also becomes heavier (Fig. 4). We found that methane is consumed down to a concentration, which corresponds to less than 50% of the saturation level. A depletion of this magnitude results Figure 6. Seasonal differences in methane cycling between sea ice, surface water and atmosphere. In winter deep convection transports brine, charged with methane, into the WML. Methane efflux may occur by convection in leads, while efflux through impermeable sea ice is restricted. In spring, basal melting starts and transports methane (dissolved or re-dissolved gas bubbles) into the MWL. In summer, stabilized by thermal stratification and less turbulence, methane remains entrained in both layers. In autumn, sea surface temperatures drop, enabling methane efflux by surface water convection. Water masses below the WML are not affected by this methane cycling.
in threshold values, i.e. values where consumption finally ceases because of substrate limitation 30 . The depth, at which threshold values is reached differs slightly along transects, i.e. between Atlantic-derived and Pacific-derived water and depends on and corresponds to the depth of haline winter convection. The permanent isolation of the UHW reduces the capacity of the interior Arctic Ocean to act as a sink for increasing atmospheric methane. This decoupling from the deeper ocean reinforces the near surface cycling of methane (i.e. between sea ice-influenced water, sea ice and the atmosphere). Summer sea ice retreat triggers this pathway with hitherto unknown consequences for the source as well as sink capacity for this greenhouse gas in the Arctic.

Methods
Sampling. We measured concentrations, stable carbon isotope ratios and aerobic oxidation rates of methane in surface water and sea ice brine along two hydrographic transects in the central Arctic Ocean occupied by RV Polarstern between early August and late September 2011. The first transect runs from 82.5°N, 50 o E via the North Pole to the Alpha Ridge, the second transect runs along 84°N over the Makarov Basin into the Amundsen Basin (Fig. 1). Covering both Atlantic and Pacific derived water, our sampling was carried out in regions with different ice coverage. We use hydrographic data (salinity, temperature and oxygen) and water samples down to 100 m depth obtained with a ship-borne CTD (Seabird) and a mounted Rosette. Salinity and temperature were measured with a Seabird SBE 911 plus CTD. Oxygen was measured with the SBE 43 dissolved oxygen sensor SN 743 and sensor calibration was done on water samples using Winkler titration. Water samples from up to six different depths were collected during the upcast at each CTD station with 10 L Niskin bottles mounted on a rosette sampler. Sea ice brine was collected by extracting ice cores from sea ice to depths of up to 50 cm and not more than half way through the ice depth, using a KOVACS Mark II ice corer. The remaining, so called "sack holes" were covered and left to allow brine to drain into the holes for a few minutes, after which brine was removed with a syringe and treated the same way as water samples. Methane concentrations were analysed within a few hours after sampling. The dissolved gas was extracted from the water or brine by vacuum-ultrasonic treatment and subsequently measured with a gas chromatograph (Chrompack 9003 (GC) with a flame ionization detector (FID). For gas chromatographic separation we used a packed column (Porapac Q 80/100 mesh). The GC oven was operated isothermally (60 °C) and the FID was held at 250 °C. Two sets of standard gas mixtures were used for calibration. The standard deviation of duplicate analyses was 5%. This high overall error is almost exclusively due to the gas extraction procedure and not to GC precision, which had an error of only 1%. After GC analyses an aliquot of the extracted gas was transferred into pre-evacuated glass containers for analysis of the stable carbon isotopic signature on shore.
The δ 13 C CH4 values were determined using a Delta XP plus Finnigan mass spectrometer. The extracted gas was purged and trapped with PreCon equipment (Finnigan) to pre-concentrate the sample. All isotopic ratios were given in a δ -notation relative to the Vienna Pee Dee Belemnite (VPDB) standard using conventional delta notation. The reproducibility derived from duplicates was 1-1.5‰.
A significant correlation between 1/methane concentration vs. the δ 13 C values (R 2 = − 0.84) is given for methane in UHW when y-intercept is − 47‰. In sea ice-influenced water (i.e. WML and MWL) a correlation is missing when y-intercept is − 47‰ (R 2 = 0.0554) but given when the y-intercept is − 55‰. This different correlation refers to deviating methane sources in both water masses.
Methane consumption. Water was sampled directly from the ice floes through ice core holes for ex situ tracer incubations in order to quantify the methane consumption. Methane oxidation rates were determined from ex situ incubations of water samples in 100 ml serum vials. Samples were incubated with 50 μ l gas mixture comprised of 3 H-labelled CH 4 (200-300 Bq) and a second set was incubated with 10 μ l of 14 C-labelled CH 4 (12000-15000 Bq). The samples were subsequently shaken to facilitate tracer dissolution and then incubated in the dark at 0 °C. Tracer consumption was measured after 48 h and 72 h. Incubations with 14 C-CH 4 were terminated by adding a 5 ml headspace and injecting 0.5 ml of 10 M NaOH so that the remaining 14 C-CH 4 accumulated in the headspace and the produced 14 C-CO 3 2and 14 C-biomass was trapped in the aqueous NaOH solution. Separation and activity measurement of 14 C-CH 4 and 14 C-CO 3 2− was carried out analogous to previous measurements of CH 4 turnover in sediments 25 (and references therein) . In short, 14 C-CH 4 in the headspace was combusted to 14 C-CO 2 , while 14 C-CO 3 2− was converted to 14 C-CO 2 through acidification with HCl. In either case, 14 C-CO 2 was then trapped and the radioactivity was measured by wet scintillation counting. Incubations with 3 H-CH 4 : Total activity ( 3 H-CH 4 + 3 H-H 2 O) was measured in 1 ml of sample aliquot by wet scintillation counting and activity of 3 H-H 2 O was measured after sparging the sample for ≥ 30 min with nitrogen gas to remove remaining 3 H-CH 4 . CH 4 oxidation rates (r OX ) were calculated assuming first order kinetics: r OX = k' [CH 4 ] where k' is the effective first order rate constant calculated as the fraction of labelled CH 4 oxidized per unit time and [CH 4 ] is the ambient CH 4 concentration. Killed controls resulted in methane oxidation rates near zero.