We have reported that the “Big Pond” site was a strong source of atmospheric CH4 with net emissions of CH4 averaging 20 g m−2 yr−1 and active emissions occurring from mid June to early November (Flessa et al., submitted manuscript, 2008). These measurements were carried out with the static chamber technique and included both diffusive and plant mediated transport. Emissions were calculated from a linear concentration increase with R2 > 0.85 and thus also included ebullition that did not lead to violation of this criterion. Emissions started with the disintegration of the ice layer during spring thaw and ended after the formation of a compact ice layer in November. Emissions at the time of pore water sampling in August 2006 were 110–170 mg CH4 m−2 d−1, which is close to average daily emissions from wet ombrotrophic sites at the Stordalen mire in northern Sweden [Christensen et al., 2004], thermokarst wetlands at Bonanza Creek, Alaska [Wickland et al., 2006] and in Manitoba, Canada [Bubier et al., 1995], and thaw lakes in North East Siberia [Walter et al., 2006]. Emissions were thus similar to other thermokarst wetlands and lakes and high compared to CH4 emissions across peatland ecosystems, exceeding median emission up to an order of magnitude or more [Blodau, 2002]. In the Little Grawijka catchment, surrounding bogs that are underlain by permafrost and are “dry” during summer even acted as a weak CH4 sink (Flessa et al., submitted manuscript, 2008).
 In agreement with the large fluxes, we recorded maximum CH4 concentrations of 0.53 to 0.97 mmol L−1 at depths of 20 to 40 cm (Figure 2) at the Big Pond site and similar CH4 concentrations also in other thermokarst depressions (Figure 3). At the “Big Pond” site, CH4 was likely transported by aerenchymatic roots of Eriophorum and other emergent vegetation, which act as effective conduits for CH4 [Frenzel and Rudolph, 1998; Shannon et al., 1996]. Ebullition may have further contributed to emissions as the CH4 partial pressures in the pond exceeded 0.2 atm, equivalent to ca. 0.39 mmol L−1 at 8°C, which is initially sufficient for ebullition to occur [FechnerLevy and Hemond, 1996]. It has been previously documented that accelerated transport by root conduit transport can lead also to a lowering of CH4 pools in the wetland soils. In the study by Shannon et al.  emissions of 200 to 400 mg C m−2 d−2 were reached in a temperate wetland with a dense cover of Scheuchzeria palustris and fairly uniform CH4 concentrations ranging from 300 to 500 μmol L−1, which is lower than in the floating mat at the “Big Pond” site. In the study by Shannon et al.  and studies cited therein, 60 to 90% of the CH4 was emitted through aerenchymatic transport and some additional contribution of ebullition, whereas diffusion was subordinate. In agreement with such findings the estimated diffusive fluxes at the “Big Pond” site only reached 3.5 mg m−2 d−1 or less than 2 % of the total CH4 emission.
Figure 2. Dissolved concentrations of CO2, CH4, H2, sulphate, and chloride in MLPs located approximately 1 m, 3 m, and 6 m from the edge of the “big Pond” site.
Download figure to PowerPoint
 Information about the predominating pathways of CH4 production can be gained from both the isotopic composition of produced CH4, as well as thermodynamic considerations [Conrad, 1999, 2005; Whiticar, 1999]. CH4 produced by acetate cleavage is usually not as depleted in 13C as CH4 produced from CO2 reduction with H2, resulting in apparent fractionation factors of <1.055 versus >1.065 [Conrad, 2005; Whiticar et al., 1986]. Even in experiments with cultured microorganisms, fractionation factors are quite variable and influenced by environmental conditions, such as temperature, substrate concentrations, and Gibbs free energies available for the process [Conrad, 2005]. A higher fractionation factor for hydrogenotrophic methanogenesis in the range of 1.07 to 1.09 can for example be expected when Gibbs free energies for the process are <40 KJ mol−1 CH4 [Penning et al., 2005]. It has been also documented that diffusive transport in aerenchymatic plants leads to preferential emission of 12C. At a field site in Florida this process lead to an increase in δ13C of 10.6 ± 3.7‰ in emitted CH4 compared to pore water sedimentary CH4 [Chanton and Whiting, 1996] which would entail an increase in the derived apparent fractionation factor of methanogenesis. Without further information, the diagnostic power of the apparent fractionation factor is thus limited and can only be used as a coarse indication of the contribution of methanogenic pathways to CH4 production [Conrad, 2005]. In both fens and bogs, CH4 has been found to be increasingly depleted in 13C deeper into the peat [Hornibrook et al., 1997; Popp et al., 1999]. Accordingly it has been suggested that in the biologically active uppermost layers, where fresh substrates are available and fluctuations in temperature and redox conditions are frequent, acetoclastic methanogenesis is generally more important and gives way to hydrogenotrophic methanogenesis at greater depths [Hornibrook et al., 1997].
 We determined δ13C values of −11.8 to −25.7‰ for CO2 and −56.2 to −62.4‰ for CH4 (Figure 4) resulting in apparent fractionation factors of 1.039 to 1.060. Such intermediate values suggest that both acetate cleavage and CO2 reduction contributed to CH4 production at the “Big Pond” and two other sites, which were characterized by fractionation factors of 1.039 to 1.051. In agreement with this interpretation, acetate, which is a ubiquitous intermediate of organic matter degradation [Conrad, 1999], did not accumulate in the pore water and was mostly below LOQ. Conversion of acetate into CH4 thus did not appear to be inhibited as for example reported by Duddleston et al.  for a bog in Alaska. Particularly at the “Big Pond” site, the CO2 was also enriched in 13C compared to organic matter, which is generally characterized by a 13C of about −25 to −30 ‰ (median ∼27‰) when produced by C3 plants [O'Leary, 1988]. This finding supports a reduction of CO2, since during this process the remaining pool of CO2 is left enriched with the heavier isotope. In contrast to previous findings [Hornibrook et al., 1997; Popp et al., 1999] we could not find evidence for a substantial shift in methanogenic pathways from the floating mat, consisting mostly of recently fixed organic matter, into the underlying water body, containing an unknown mixture of “old” and “new” carbon. In the investigated ponds CO2 reduction was apparently more important than in incubated organic matter samples from “dry” bog locations. In a previous study, acetotrophic methanogenesis dominated at summer soil temperatures when bog samples taken from the catchment were incubated after a storage period of 4 months [Metje and Frenzel, 2007]. The exact reason why acetoclastic methanogenesis did not predominate more clearly in the floating mat, despite the availability of recently fixed organic matter, cannot be identified. One may speculate that permanently water saturated conditions and high partial pressures of CO2 and H2, which favor CO2 reduction thermodynamically, are factors that encouraged hydrogenotrophic methanogenesis in the ponds compared to other wetland environments.
Figure 4. Ratios of 13C/12C in δ13C notation determined in the headspace of selected samples of the “Big Pond,” “Sedge Pond,” and “Sphagnum Pond” sites.
Download figure to PowerPoint
 Both pathways were thermodynamically viable processes, as Figure 5 illustrates. Hydrogen concentrations reached between 50 and 900 nmol L−1 (Figure 2) and resulted in a quite large driving force for hydrogenotrophic methanogenesis on the order of 50 to 70 kJ mol−1 (substrate). The driving force for acetotrophic methanogenesis was considerably smaller on the order of 26 to 38 kJ mol−1 (acetate) but still larger than threshold values determined in slurry incubation experiments with rice paddy soil [Chin and Conrad, 1995], and assuming that acetate occurred at half LOQ of 10 μmol L−1. In contrast, the energy gain from homoacetogenesis of CO2 and H2 to acetate was closer to 0 with values of 7 to 17 kJ mol−1. The fact that hydrogen concentrations were high suggests that 1) hydrogenotrophic methanogenesis was not able to lower hydrogen concentrations to thermodynamically controlled steady state levels and 2) that alternative electron acceptors were not able to compete effectively for hydrogen. The reasons for the high H2 levels cannot be clarified. In absence of alternative electron acceptors hydrogenotrophic methanogenesis often operates at lower H2 concentration levels, resulting in a ΔGr −20 to −40 kJ mol−1 (CH4) in experiments with freshwater soils and sediments [Chin and Conrad, 1995; Conrad, 1999; Rothfuss and Conrad, 1993]. In their presence and with rapid utilization, hydrogen concentrations are typically lowered to steady state values of <0.1 to 2 nmol L−1 thus resulting in a positive ΔGr of hydrogenotrophic methanogenesis [Heimann et al., 2007; Lovley and Goodwin, 1988].
Figure 5. Gibbs free energies (ΔGr) of hydrogenotrophic and acetotrophic methanogenesis and homoacetogenesis calculated from gaseous and dissolved concentrations in MLPs installed in “Big Pond.”
Download figure to PowerPoint
 In agreement with the high H2 concentrations and a subordinate importance of alternative terminal electron accepting processes, nitrate was generally not detectable and sulfate concentrations in the Big Pond site were <40 μmol L−1 in the surface layer of all three wells (Figure 2). A concentration minimum of <10 μmol L−1 was reached at a depth of 30 to 40 cm. This suggests that some sulfate was reduced in the floating mat itself and either provided from the surface (“1 m” and “3 m”), e.g., from precipitation and reoxidation of reduced sulfur [Blodau et al., 2007b], and additionally from the water body and deeper layers below (“6 m”). The latter is also confirmed by the chloride profiles which indicated a deeper source of more mineralized water particularly at the “6 m” site (Figure 2).
 The fairly constant δ13C values of CH4 in the profiles of the ponds (Figure 4) also suggests that CH4 was not oxidized to a significant degree by aerenchymatic oxygen transport in the rhizosphere, which has been described and analyzed elsewhere [Arah and Stephen, 1998]. Methane oxidation is associated with increasing δ13C values due to enrichment of the heavier isotope in the remaining CH4 pool, which generally leads to increasing δ13C CH4 values where oxygen is available [Whiticar, 1999], e.g., near the water table. We did not obtain δ13C values near the surface of the floating mat but up to 20 cm below surface δ13C values did not increase. Significant CH4 oxidation was thus likely restricted to the aerated organic uppermost layers, possibly Sphagnum mosses [Basiliko et al., 2004] and root conduits, although in Eriophorum oxidation was shown to be very slow [Frenzel and Rudolph, 1998].
3.2. Carbon Dioxide
 Concentrations of CO2 were about an order of magnitude higher than concentrations of CH4 and reached a maximum of 6.9 mmol L−1 in Big Pond and between 5 and 12 mmol L−1 in the other sampled thermokarst depressions (Figures 2 and 3). Similar CO2 concentrations have been reported from thermokarst wetlands at the Bonanza Creek field site using pore water peepers [Wickland et al., 2006]. The resulting diffusive flux of CO2 determined by Fick's law and the pore water profiles in “Big Pond” was 20.5 mg C m−2 d−2 and thus only about 15 % of the chamber CH4 efflux in August 2006. This finding raises the question whether anaerobic decomposition in the floating mat was capable of producing the emitted CH4, given that decomposition of organic matter with an oxidation state of 0, such as polysaccharides, requires the release of equivalent amounts of CO2. To gain further insight into this question we first examined the shape of the CO2 concentration profiles. Both CH4 and CO2 concentrations peaked in “Big Pond” wells “2 m” and “3 m” in the floating mat and at intermediate depths also at the other sites. This first of all indicates a stronger respiration activity that was likely caused by higher “soil” temperatures near the surface, the input of labile organic matter from roots and possibly root respiration. Secondly, such concentration profiles are also an indication of nonsteady state between production and concentrations, as in the long run production coupled to a zero-flow boundary at the bottom must lead to monotonously increasing concentrations with depth [Berner, 1980]. The accelerated production during summer did apparently not equilibrate with the pool size present in the floating mat and underlying water and the calculated diffusive flux underestimated the true CO2 production. To infer the magnitude of nonsteady state CO2 production in the floating mat we thus simulated the CO2 depth profiles.
 To estimate CO2 production required to generate the CO2 maximum in the floating mat we ran two simulations, 1) to create the profile in the lower part of the pond (<60 cm depth) assuming an arbitrary accumulation period of 10 years and 2) based on this profile, to create the concentration maximum in the floating mat within a three month period (May to August). The rationale behind this approach was the assumption that absence of autotrophic respiration and slowness of heterotrophic production in the cold season as well as gradual freezing of the floating mat (<60 cm depth) would result in low CO2 concentrations in spring so that the pool in the floating mat had to build up with the beginning of summer. A reasonable fit was obtained using a long-term production rate of 1 nmol cm−3 in the water body below and a short-term production rate of 30 to 80 nmol cm−3 d−1 in the floating mat (Figure 6). The resulting cumulative 3 month production of CO2 in the floating mat was 23 g C m−2, which was on the order of annual CH4 emissions from the surface. Such an ad hoc simulation of a single date of concentration profiles is of course subject to considerable uncertainty; but the order of magnitude of respiration rates may be adequately reflected. If so, nonsteady state anaerobic respiration in the recently formed floating mat was capable of sustaining the determined CH4 fluxes from the pond.
Figure 6. Model fits of the mixed reservoir model used to estimate nonsteady state anaerobic CO2 production in the floating mat during summer. “Scenario 1” was run using a production of 1 nmol cm−3 d−1 in the deeper layer for 10 years. “Scenario 2” was run for a time period of three months with production rates of 30 to 80 nmol cm−3 d−1 in the floating mat using the “Scenario1” concentration profile as starting condition.
Download figure to PowerPoint