Methane (CH4) emissions currently contribute ~20% to the planet’s greenhouse effect, with a large, but poorly defined, fraction derived from freshwater ecosystems1,2,3,4. Accurately placing freshwaters in the global CH4 budget requires a better understanding of the controls and contributions of CH4 from different sources2,3,4,5. In this regard, the surface waters of lakes and rivers are systematically supersaturated with CH4, and it has been traditionally assumed that this CH4 is derived from anoxic environments, via vertical and lateral transport from profundal and littoral sediments3,6,7,8. This assumption certainly holds for small, shallow ecosystems where the surface layers are in relatively close contact with sediments. Yet CH4 supersaturation is also prevalent in large, deep lakes3,5,7,8,9,10,11,12 and oceanic surface waters13,14,15,16,17, where deeper water columns result in significantly reduced surface water exchange with anoxic sediments. Studies of CH4 dynamics in surface waters of oceans and large lakes have indeed concluded that pelagic CH4 supersaturation cannot be sustained either by lateral inputs from the littoral or from benthic inputs alone9,10,13,14,15,17,18,19.

A pelagic source of CH4 would thus be required to sustain the observed CH4 supersaturation, and multiple lines of evidence suggest that CH4 is produced in oxic water columns of freshwater and marine systems. In vitro experiments point to CH4 production in both microanoxic habitats in metazoan guts and on particles14,17,18,19,20,21,22,23, and in particle-free oxic water15,16,18,19,23,24,25, via multiple biochemical pathways. The presence19,22,26,27,28 and activity19,22,27 of hydrogenotrophic and acetoclastic methanogenic archaea have been confirmed at the molecular level in diverse oxic environments. At the same time, marine studies suggest that the microbial decomposition of methylated compounds such as methanethiol15 and methylphosphonate16,23,24,25 could be an important source of CH4 in surface waters of the open ocean. Further, metalimnetic peaks of CH4 are a recurrent feature in many ecosystems, which correlate positively with dissolved O2 (DO), algal biomass and production8,9,10,14,15,18,19. However, despite past efforts, the ecological and biogeochemical significance of oxic water column methanogenesis remain speculative, and we have yet to determine the dominant biochemical pathway, its controls, and its contribution to diffusive CH4 emissions from aquatic ecosystems.

In this study, we experimentally assess pelagic CH4 production in the oxygenated water column of Lac Cromwell, a typical Canadian temperate Shield lake, using floating mesocosms open to the atmosphere, but closed at the bottom and therefore uncoupled from non-pelagic sources of CH4. The results presented herein represent the first direct, ecosystem-scale demonstration of oxic water CH4 production, its drivers and source pathway, and the potential widespread importance of this poorly considered process.


CH4 dynamics in experimental enclosures

The initial CH4 concentrations in the enclosures were four- to tenfold lower than surrounding lake waters, because of degassing during the initial filtration and filling process, but the enclosures were nevertheless supersaturated relative to the atmosphere at the onset of the experiment (Fig. 1a). Vertical profiles of the enclosures (Supplementary Fig. 1) showed a range in DO between 45.6 and 128.6% saturation, indicating that the entire water column remained oxic in all treatments for the duration of the experiment. Despite oxic conditions, CH4 concentrations ranged between 0.10 and 0.53 μM throughout the entire experiment, across all treatments (Fig. 1a), which represents ~50- to 265-fold super-saturation relative to the atmosphere.

Figure 1: Surface water CH4 dynamics in experimental enclosures.
figure 1

Treatments include ambient untreated controls (C), addition of dissolved organic carbon (DOC), nitrogen and phosphorus (NP) or a combination of both (DOC-NP). (a) CH4 concentrations (error bars ±1 s.e.m., n=3) generally increased through time, but differed between treatments (RM-ANOVA, treatment: P=0.004, F=10.71; time: P=0.007, F=7.67; treatment* time: P=0.04, F=3.04; Tukey’s HSD post-hoc grouping=NP, DOC-NP, C>C, DOC). Dashed lines show the expected declines in dissolved CH4 because of diffusion at the air–water interface, had no methane been produced in situ. Without internal production of CH4, concentrations would have equilibrated with the atmosphere (mean concentration at equilibrium=0.002 μM) in all treatments within 1–2 weeks. (b) Collectively, net CH4 production (evasion plus water column accumulation per enclosure) and evasion rates (second y axis) displayed large differences among treatments (ANOVA, P=0.003; post-hoc groups =NP & DOC-NP & C>C>C & DOC). Thick and thin horizontal lines are mean and median, respectively.

The observed surface CH4 concentrations in the mesocosms represent the net balance between CH4 production, CH4 oxidation (MOX) and CH4 exchange with the atmosphere. In the absence of internal inputs of CH4, the initial CH4 supersaturation in the enclosures would have declined to atmospheric equilibrium within approximately 1 week (Fig. 1a, dotted lines), as estimated on the basis of our own empirical measurements of the gas exchange coefficient (see Methods section). Thus, an internal CH4 source was necessary to sustain the systematic CH4 supersaturation observed in all mesocosms throughout the experiment. Further, there was an overall significant increase in CH4 concentrations through time across treatments (Fig. 1a; repeated measures analysis of variance (RM-ANOVA) time effect: P=0.007), with greater increases in nutrient-amended enclosures (nitrogen and phosphorus (NP), dissolved organic carbon (DOC)-NP), and weaker increases in the DOC and control enclosures (RM-ANOVA treatment × time effect: P=0.04). As a result, all enclosures emitted CH4 to the atmosphere for the duration of the experiment, and the estimated water/air CH4 fluxes (based on our measurements of gas transfer coefficients in the mesocosms) ranged from 0.07 to 0.36 mmol m−2 per day (Fig. 1b). Net CH4 production in the enclosures, calculated as the sum of the observed net increase in concentration and the calculated CH4 fluxes to the atmosphere, were on average highest in the nutrient-enriched enclosures, whereas the DOC-only addition generated the lowest production rates (Fig. 1b; analysis of variance (ANOVA): P=0.003). We assessed the robustness of chamber-based calculations of CH4 emissions by comparing chamber-derived gas exchange coefficients (kCH4) to estimates based on wind speed (detailed in Supplementary Note 1). There was good agreement between the two approaches, and therefore we conclude that chamber-derived results yield reliable estimates of kCH4 and CH4 fluxes for both the enclosures and the lake.

Linking CH4 production to algal dynamics

The differences among treatments in water column CH4 concentrations and fluxes were strongly linked to pelagic gross primary production (GPP; Fig. 2a) and net ecosystem production (NEP=GPP-R; Fig. 2b). There was a 20-fold range in GPP across treatments (Fig. 2a), whereas there was only a 5-fold range in ecosystem respiration (R), therefore nutrient additions resulted in a strong shift in NEP (Fig. 2b). The average CH4 fluxes at day 7 of the experiment were strongly positively related to average GPP rates in the enclosures (Fig. 2a), and to NEP, whereas the DOC addition had no positive influence relative to the control. Increases in CH4 flux appear to have been driven mostly by increases in GPP rather than heterotrophic metabolism.

Figure 2: Linking pelagic ecosystem metabolism and CH4 dynamics.
figure 2

Metabolic estimates from day 7 of the experiment revealed CH4 evasion was strongly positively correlated to (a) ecosystem gross primary production (GPP; P=0.003; r2=0.93; y=215.3[1−e−1.6x]), and (b) shifts to increased net ecosystem production (NEP), which is different between GPP and community respiration (P=0.02; r2=0.99; y=111–106[1−e−2.6x]). Symbols with error bars (±1 s.e.m., n=3) denote the different treatments, including ambient untreated controls (red circles), addition of dissolved organic carbon (blue squares), nitrogen and phosphorus (grey triangles) or a combination of both (yellow diamonds).

Other potential sources of CH4 in the enclosures were also considered. The enclosures contained very little particulate organic C because of the initial filtration of the surrounding lake water before filling, and there was little or no accumulation of particulate organic C in the bottom of enclosures at the end of the experiment. Moreover, we estimate that zooplankton-derived CH4, calculated from the measured zooplankton biomass in the enclosures and published production rates (Supplementary Table 1, Supplementary Note 2), contributed <10% to the estimated gross CH4 production rates in ambient enclosures, a result that is consistent with previous studies10,18,19,29. Membrane permeability and the resulting exchange with the surrounding lake waters were also considered as a potential cause for CH4 supersaturation. However, the estimated contributions of CH4 from the surrounding lake environment were on the order of <1% of gross CH4 production (Supplementary Methods and Supplementary Table 2). Finally, there was no significant development of biofilm on the walls of the enclosures during the experiment, and this is unlikely to have been a significant source of CH4.

Identifying the biochemical source of CH4

Several methanogenic pathways exist, but as acetoclastic methanogenesis is less isotopically discriminatory than CO2 reduction or methylotrophy19,27,28, apparent fractionation factors (αapp; see Methods for calculation details) can be used to qualitatively distinguish whether CH4 is produced via methanogenesis from acetate, or CO2 reduction30. Our measured αapp values were well within the range indicating dominance of the acetoclastic pathway (Fig. 3a), suggesting that acetate was the dominant source material for pelagic methanogenesis in all treatments. As we only had one estimate of water column MOX per treatment (range=0.003–0.008 mmol m−3 h−1), we used a scenario analysis with two extreme MOX rates (0 and 0.05 mmol m−3 h−1; detailed in Methods section) to conservatively constrain the possible αapp values based on variable MOX rates and associated isotopic fractionation. In all cases, αapp values remained well within the acetoclastic range, that is, <1.055 (ref. 30; Fig. 3a). Acetoclastic methanogenesis was likely dominant through time, because δ13C-CO2 and δ13C-CH4 showed little temporal variability (Fig. 3b).

Figure 3: Assessing the biochemical source of pelagic CH4.
figure 3

(a) Here patterns in the apparent fractionation (αapp; see methods for calculation details) during methanogenesis suggest that the acetoclastic pathway strongly dominated in all treatments, and increased in importance with the addition of nutrients. White and black circles represent values estimated using measured CH4 oxidation rates (MOX) and a maximum and minimum associated isotopic fractionation factor, respectively. To constrain the effect of error introduced by MOX estimates, αapp was estimated from a highly conservative scenario analysis, summarized here by bars and error bars (midpoint and upper potential range, respectively). Values of αapp were estimated for day 7 of the experiment, although isotopic signatures of CO2 and ambient CH4 displayed little change through time for each treatment (b). Red, blue, grey and yellow symbols indicate control, DOC, DOC+NP and NP treatments, respectively, whereas circles, squares and triangles, respectively, indicate days 0, 7 and 28 of the experiment. One enclosure was followed for each treatment, and sample points represent one unreplicated measurement.

Rates of oxic water-column methanogenesis

The CH4 accumulation and outgassing that we measured in the enclosures (Fig. 1) represent the net result of pelagic methanogenesis, gas exchange and MOX. In order to derive a first-order estimate of pelagic methanogenesis, we carried out a detailed CH4 mass balance (described in the Methods section; summarized in Supplementary Table 2) for day 7 for the control enclosures. Here we combined the measured CH4 evasion, the change in storage corrected for cross-membrane inputs from surrounding lake water and measured MOX. This exercise yielded rates of apparent pelagic methanogenesis on the order of 0.21–0.24 mmol m−3 per day (Table 1), of which ~60% was apparently oxidized and the remainder evaded to the atmosphere.

Table 1 Greenhouse gas dynamics in Lac Cromwell.

Oxic water methanogenesis in a whole-lake perspective

Mean pelagic CH4 evasion from control (that is, non-manipulated) enclosures (0.15±0.03 mmol m−2 per day; Table 1) represented ~20% of the average diffusive CH4 fluxes measured in the lake over the summer (0.78±0.35 mmol m−2 per day), and at the whole-lake scale, were similar to CH4 ebullitive fluxes (15.2±32.5 mol per day or 0.31±0.66 mmol m−2 per day in lake regions <3 m deep; Table 1). Further, oxic pelagic methanogenesis contributed on the order of 4% to total lake greenhouse gas emissions (CO2+CH4, expressed as CO2 equivalents; Table 1). It was possible to compare fluxes directly between the enclosures and the lake because the physical conditions shaping gas exchange dynamics were quantitatively similar for each. As shown in Supplementary Fig. 2, floating-chamber-derived gas exchange coefficients (kCH4) for the lake and the enclosures varied on a diurnal basis, but averaged 0.65 and 0.69 m per day and ranged between 0.10–1.91 and 0.24–1.44 m per day for the enclosures and the lake, respectively.


Here we present an ecosystem-scale, experimental demonstration of significant CH4 production and evasion from the oxic water column of Lac Cromwell that is closely linked to algal dynamics, and which contributes a baseline flux of CH4 that is likely present in all lakes. The relationship between CH4 and phytoplankton observed here (Fig. 2) has been hypothesized before to explain both the presence of metabolically active methanogens, and the recurrent metalimnetic and near-surface CH4 peaks in oxic lake9,10,18,19 and marine13,14,15,16,17 environments. We confirm this link experimentally, and further show that it generates a significant out flux of CH4 from the mesocosms to the atmosphere (Fig. 1b). These results in turn imply that factors influencing phytoplankton standing stock and GPP, such as grazing, nutrient availability and the physical structure of the water column, will have a strong bearing on pelagic CH4 dynamics and resulting CH4 emissions.

To our knowledge, our results represent the first ecosystem-level estimates of methanogenesis in oxic freshwaters. For all treatments, intense CH4 production was needed to sustain the observed patterns in CH4 concentrations and to offset atmospheric losses. In the absence of methanogenesis, the decreases in enclosure CH4 concentrations due solely to atmospheric evasion would have rapidly depleted the CH4 pool (dashed lines, Fig. 1a). These calculations are potentially conservative, as we did not consider MOX. Considering all potential sources and sinks, we estimate an average rate of methanogenesis of 0.23 mmol m−3 per day (Table 1; Supplementary Table 2). This rate is higher than previous estimates from in vitro incubations for oligotrophic Lake Stechlin, Germany (0.04–0.09 mmol m−3 per day)18,19, but it should be noted that this estimate is very sensitive to the values of MOX used. In-situ MOX can vary dramatically across systems31, with depth, diurnally and with changes in light exposure18,19. Therefore, it is possible that our estimates of MOX and methanogenesis may be biased by the fact that water from only one depth was used, and incubations were run in the dark (see Methods section for details). Consequently, we consider our estimates of net CH4 production (excluding MOX; Fig. 1b) as a lower bound for epilimnetic methanogenesis, as these estimates are analogous to methanogenesis where MOX equals zero. In the ambient enclosures, net production averaged 0.09 and ranged from 0.04 to 0.12 mmol m−3 per day, and is in close agreement with estimates of methanogenesis in Lake Stechlin, where MOX was shown to be extremely low throughout the epilimnion18,19. The fact that our approximation of pelagic methanogenesis based on mass balance (Fig. 1b; Table 1) is in good agreement with in vitro18,19 estimates is very promising, and suggests that both approaches could be incorporated into future detailed studies of CH4 dynamics for other systems.

Our conclusion that acetate consumption supplied most CH4 in the enclosures, based on estimates of αapp, is particularly sensitive to changes in the rate of MOX. Given the potential variability in MOX discussed above, estimates of αapp may be biased by our limited MOX measurement. However, scenario analyses estimating αapp from extreme hypothetical MOX rates (see Methods for calculations) were all well within the range of acetoclastic dominance (Fig. 3a). This supports our conclusion that acetate supplied the majority of CH4 across our treatments, likely for the duration of the experiment (Fig. 3b). Consistent with this conclusion, observations of enriched surface δ13C-CH4 in numerous lakes8,10,18,29,32 have been hypothesized to indicate dominance of acetoclastic methanogenesis in oxic freshwater pelagic zones18, and both known acetoclastic genera (Methanosaeta and Methanosarcina) have recently been detected in oxic freshwater19,26 and terrestrial27,28 environments. Furthermore, acetoclastic dominance was enhanced with the addition of nutrients and increased algal production (Fig. 3a). This pattern parallels that from anoxic environments, where increased abundance of biologically young, high-quality algal material promotes acetoclastic over hydrogenotrophic methanogenesis, the latter instead supported by lower quality, aged terrestrial DOC30. Finally, cyanobacteria were near absent in the mixed layer of our enclosures (<3.5% relative abundance, as biomass, estimated from one enclosure per treatment on day 7 in all treatments), ruling out major contributions of diazotroph-derived water-column H2 for hydrogenotrophic methanogenesis19, and cyanobacterial methanogenesis during DOM demethylation23. Taken together, it appears that acetoclastic methanogenesis linked to in-situ, algal DOC production, potentially plays a central role in supporting water column CH4 production in oxic freshwaters.

The potential importance of acetate for oxic water methanogenesis is intriguing, as acetate concentrations are typically extremely low in oxic surface waters18,33,34,35, and the presence of oxygen may inhibit acetate consumption27. Yet past reports based on radioactive substrate additions have unequivocally demonstrated extremely rapid turnover of acetate in oxic environments across fresh and marine surface waters33,34,35,36, suggesting active production of acetate from multiple metabolic pathways37 in oxic environments, as well as very efficient uptake33,34,35,36. Fermentative metabolism by diverse and widespread facultative anaerobic bacteria in particle-associated microanoxic zones could be the main source of water column acetate38,39,40, and such microhabitats could be directly associated to algae or to particles19, yet in vitro work as well as our own results further suggest that methanogenesis proceeds even when these potential sites are removed18,19. One possibility that should be further explored is that fermentative bacteria themselves create the conditions for anoxic methanogenesis through syntrophic interactions with acetoclastic methanogens in mixed cell aggregates41.

Is oxic water methanogenesis a significant component of the CH4 budget of Lac Cromwell? We have combined our estimated rates of oxic water methanogenesis and CH4 emissions in the ambient enclosures with measurements of diffusive and ebullitive CH4 fluxes from the surrounding lake waters (summarized in Table 1), and conclude that methanogenesis in the oxic pelagic zone may potentially contribute ~20% of mean summertime CH4 diffusive fluxes, approximately the same magnitude as ebullitive fluxes measured for the entire lake. Careful extrapolation of our absolute emissions rates (Fig. 1b, Table 1) to other ecosystems should be made, as previously overlooked, system-specific differences in gas exchange coefficients ( k CH 4 ) and non-Fickian fluxes will introduce error to conventional cross-system flux comparisons (see Supplementary Note 3 for extended coverage of this potential problem). Although oxic water methanogenesis contributed ~4% to Lac Cromwell’s summertime greenhouse gas footprint, it is conceivable that this pathway could have greater relative importance in more nutrient-rich, productive ecosystems, and in hardwater environments where low to negative (that is, net CO2 uptake) CO2 emissions are common42,43,44.

Within Lac Cromwell, it is interesting to note that oxic water methanogenesis contributed a significant component of whole-lake emissions, despite the fact that it is a small (0.1 km2), relatively shallow (~3.5 m mean depth) lake with an extensive littoral area, and a CH4-rich anoxic hypolimnion, all of which are typically major sources of CH4. A corollary is that in large, deep lakes and open ocean sites where surface waters are increasingly decoupled from benthic or littoral sources of CH4, surface CH4 concentrations should be mostly driven by pelagic methanogenesis and therefore a function of algal biomass and metabolism. In this regard, whereas chlorophyll a does not explain the large-scale patterns in surface CH4 across lakes in general45, the relationship we observed between surface water CH4 and chlorophyll a in our isolated enclosures (log10[CH4]=0.46log10[chl a]+0.68, r2=0.68, n=4) generally agrees with observations specifically from large lakes, extending all the way to ultraoligotrophic open ocean regions (Fig. 4; log10[CH4]=0.99log10[chl a]−1.63, r2=0.73). This continuity suggests that algal-linked pelagic CH4 production may explain much of the ambient CH4 dynamics observed in large and deep aquatic ecosystems, regardless of the fact that different mechanisms have been identified as potentially important in marine (that is, methylphosphonate or methanethiol catabolism)15,16,23,24,25 versus freshwater (hydrogenotrophic or acetoclastic methanogenesis; Fig. 3)18,19. There are clearly a number of potential sources of CH4 in oxic pelagic waters, and although it remains unknown how each pathway differs in relative importance across these diverse aquatic environments, our results highlight a consistent relationship between open-water phytoplankton dynamics and the regulation of oxic pelagic CH4.

Figure 4: Linking dissolved CH4 and algal biomass across diverse open-water aquatic ecosystems.
figure 4

Here we combine experimental data from this study with literature-derived measurements from large (>10 km2), deep (>10 m maximum depth) lakes and marine environments where both CH4 and chlorophyll a samples were simultaneously measured. A strong positive relationship between CH4 and chlorophyll a exists across systems: (a) averaged data ([log10x]=0.993[log10y]−1.630, r2=0.73, P<0.0001), where error bars represent ±1 s.e.m.; (b) raw data ([log10x]=0.855[log10y]−1.166, r2=0.72, P<0.0001). Guidelines for meta-analysis are discussed in the Methods section, and raw data are available in Supplementary Table 3.

Taken together, our findings suggest that pelagic CH4 production in oxic environments may be widespread, and likely supports a baseline evasion of CH4 from all oxic water columns. In small and shallow ecosystems, such as Lac Cromwell, this pathway will be quantitatively significant but secondary relative to anoxic littoral and benthic sources. In large and deep aquatic environments, however, this pathway could become the single dominant source fueling CH4 supersaturation and fluxes to the atmosphere. It should be emphasized that consideration of oxic water column methanogenesis does not necessarily lead to higher flux estimates for lakes, as the CH4 derived from this process is already included in routine measurements of surface water pCH4 and CH4 surface fluxes2,3,4,5,6,7,32,46. Incorporating the origin of the CH4 that outfluxes to the atmosphere from these lakes, however, is critical to our understanding of the regulation of these fluxes and our capacity to predict their future change. In this regard, there are potentially large global implications for algal-driven oxic-water methanogenesis. CH4 emissions from surface waters, particularly from freshwaters2,3,4, are a major element of the global atmospheric CH4 budget, and we suggest here that the algal-mediated baseline flux is not only a major contributor to these overall aquatic CH4 emissions, but also one that is particularly sensitive to environmental change. As such, widespread and intensifying human- and climate-driven changes in pelagic nutrient availability47,48,49, terrestrial DOC inputs50,51 and physical structure of the water column47,48,51, which strongly shape aquatic algal dynamics47,51,52,53,54, may have major, but previously unconsidered, consequences for oxic water methanogenesis and aquatic CH4 emissions.


Study site

Lac Cromwell is located at the Station de Biologie des Laurentides, a field research facility of the Université de Montréal. The lake turns over in spring and fall, is relatively shallow and small (mean depth=3.5 m, maximum depth=9.5 m, surface area of 0.11 km2, volume of 3 × 105 m3), and is oligo-mesotrophic, with little phytoplankton biomass (~6.0 μg of chl a per litre)55.

Experimental design

Here, 12, 1 m diameter, 6 m deep, 4712-L polyethylene enclosures were attached to floating wooden frames anchored in the middle of the lake (~8 m depth). On 5 June 2012, bags were filled by pumping epilimnetic water sequentially across two mesh screens of 110 and 45 μm to remove zooplankton, and three mesocosms each received N+P, DOC, N+P+DOC additions, or no addition (control). Both KH2PO4 and NaNO3 additions were made to reach target concentrations of 50 and 700 μg P and N per litre, respectively. For DOC, Superhume brand humic slurry was added to reach a target concentration of 15 mg of DOC per litre. For a full review of Superhume properties and applicability to freshwater experimentation, see Lennon et al.56. The mesocosms were allowed to stabilize for 1 week before initial sampling and re-stocking zooplankton at ambient lake concentrations. Zooplanktons were collected from Lac Cromwell by vertical tows using a 54-μm Nitex net, and immediately released into mesocosms after each haul.

Limnological measurements

All limnological samples were gathered at ~1200 hours on days 0, 7 and 28 of the experiment (13, 19 June 2012 and 10 July 2012, respectively; Supplementary Fig. 3). Dissolved oxygen (DO) and temperature were measured at 0.5 m depth in all bags, whereas water column profiles were taken from one enclosure per treatment using an Yellow Springs Instruments (YSI) Pro Plus multiparameter sonde (Supplementary Fig. 1). On all three dates, water was collected at 0.5 m depth, filtered to 0.45 μm and chlorophyll a was measured spectrophotometrically in ethanol filter extracts. Dissolved CH4 concentrations at 25 cm depth were measured in triplicate for each mesocosm following Prairie and del Giorgio5. Partial pressures were converted to concentrations using Henry’s constant corrected for temperature. On day 7, dark incubations to estimate MOX using unfiltered surface water were conducted in the laboratory at room temperature, lasting 96 h (Supplementary Fig. 4). Incubations began at approximately noon on 18 June, day 6 of the experiment; samples for measurement of MOX were collected from one mesocosm per treatment, at 25 cm depth, into 12 ml precombusted vials. Samples were capped to exclude any air with a gastight, 3-mm-thick, butyl rubber-lined plastic cap, then incubated in the dark at 20 °C. Ambient CH4 measured in the mesocosms was used as a time 0 concentration, and samples were subsequently killed every 24 h by injection of HgCl2, and stored at 4 °C in the dark until analysis. Upon analysis, 5 ml of water was displaced with ambient air, and the remaining sample was equilibrated with the air by physically shaking the vials. Headspace equilibration and gas chromatography analysis were performed as for ambient CH4. Respiration rates were determined by dark in situ, 24 h incubations on experiment days 6 and 7 in all enclosures, using unfiltered water in 4-l cubitainers. Initial and final samples of DO were collected and measured by membrane inlet mass spectrometry57.

CH4 chamber flux estimates

Air–water CH4 fluxes in the mesocosms and the surrounding lake were calculated by combining the measured surface water pCH4 and the CH4 gas exchange coefficient (kCH4), the latter derived from the measured gas exchange coefficient for CO2 (kCO2) in mesocosms supersaturated in CO2. kCO2 was measured in one mesocosm per treatment, 3–5 times daily during days 6 and 7 and 27–28 of the experiment, using the floating chamber method following Vachon et al.58 Briefly, chambers were connected via a closed loop system to an EGM-4 infrared gas monitor (PP Systems). Fluxes were calculated as the rate of change of chamber pCO2 per min during a 15-min interval. Diel sampling periods were spaced to obtain flux estimates from the morning, afternoon and nighttime periods, and mean values were used for daily flux estimates. Dissolved CO2 samples were taken before all chamber measurements using the same headspace technique as for CH4, but with direct injection of the equilibrated gas into the EGM-4 in the field. The gas exchange coefficient for CH4 was then calculated from kCO2 using equation (1), following Jahne et al.59:

where Sc is the Schmidt number of CO2 and CH4, respectively60. To calculate CH4 fluxes, we used these average k CH 4 values, the measured pCH4 and the temperature-dependent solubility of the gas following equation (2):

where F is flux of CH4 (mmol m−2 per day), Kh is the temperature-corrected Henry’s constant and Δgas is the difference in partial pressures between the air and the water partial pressures of CH4. We assumed an atmospheric partial pressure of CH4 of 1.75 p.p.m.

Isotopic analyses

Isotope compositions are reported in delta notation following equation (3):

where R is the ratio of heavy to light isotope, δ13C represents δ13C–CH4 or 13C of dissolved inorganic carbon (DIC), δ18O represents δ18O–O2 or δ18O–H2O and Rstandard is the δ13C or δ18O signature of Vienna Pee Dee Belemnite or standard mean ocean water, respectively. All samples were taken at ~25 cm depth. Samples for δ18O–O2 and δ13C–CH4 were stored in 12-ml precombusted borosilicate vials, preserved with HgCl2, and air-free samples capped with a gastight rubber-lined plastic cap. δ18O–O2 samples were analysed at the University of Ottawa Stable Isotopes Laboratory, and δ13C–CH4 samples were measured at the University of Waterloo, both using standard laboratory methods. A single sample for δ18O–H2O was collected from the lake before filling enclosures, analysed with a Picarro L230i isotopic water analyser, and assumed representative for all enclosures for the duration of the experiment. δ13C-DIC samples were collected in acid-washed, 40 ml vials with both Teflon and rubber-lined gastight plastic caps, and measured at the Colorado Plateau Stable Isotopes Laboratory using standard methods. Day 7 respiration (R), δ18O–O2 and δ18O–H2O were used to estimate GPP and NEP following Quiñones-Rivera et al.61

Numerical methods

We used RM-ANOVA to compare the effects of treatments, and treatment by time interactions, on ambient CH4. One-way ANOVA was used to assess differences between treatments for grouped data, and log10(x+1) transformations applied to maintain the homogeneity of variances, followed by Bonferroni post-hoc pairwise comparisons of treatment means. All regression analyses were performed using Sigmaplot version 12, and ANOVAs were computed with SPSS version 16. We used ordinary least squares regression to quantify the link between GPP or NEP and CH4 evasion. To estimate rates of MOX, the loss of CH4 through time was fitted with a polynomial function, the derivative was taken and the slope at time 0 was calculated (Supplementary Fig. 4). This approach was deemed more accurate in quantifying the slope at time 0, as traditional methods, such as log transformations followed by visual selection of the linear portion of the data set31, can introduce large errors in the slope of the regression line when deciding which points to fit and which to exclude. Although MOX was estimated only once, our estimates are within the range of pelagic MOX (0.170–0.250 mmol m−3 per day) previously reported31,32,62, particularly for boreal lakes of similar trophic condition62.

Apparent isotopic fractionation of methanogenesis

We estimated the apparent fractionation factor (αapp) during methanogenesis following Conrad30. The apparent fractionation factor is defined using equation (4) as

where δ13C–CH4source is the isotopic signature of source CH4, and δ13C–CO2 was calculated from measured δ13C-DIC following the study by Stumm and Morgan63, using fractionation factors from the study by Mook et al.64. As we measured ambient δ13C–CH4 (δ13C–CH4ambient), we estimated δ13C–CH4source by correcting for the isotopic fractionation effects of MOX and evasion to the atmosphere using an open-system, steady-state model for isotope fractionation65, as defined in equation (5):

where f and 1−f are the fractions of CH4 loss, standardized to daily rates for the entire enclosure, estimated, respectively, as evasion and MOX rates divided by the sum of each. Δ is the isotopic effect of each loss pathway. In cases such as ours where Δ<100‰, Δ can be approximated using equation (6):

where α is the fractionation factor of a given reaction. Literature-derived values of α were used for both loss pathways. For evasion, an α value of 0.9992 was used66, and for MOX, the range reported by Bastviken et al.62 (0.9816–0.9792) was used.

To better constrain the potential range of αapp estimated on day 7 of our experiment, we performed a scenario analysis by varying combinations of MOX rates and isotopic fractionation values, to determine the sensitivity of our estimates of αapp to potential variability in MOX. For this analysis, we chose two very conservative, extreme (for oxic pelagic freshwaters) values in MOX rates (0 and 0.05 mmol m−3 h−1) and paired each rate with both minimum and maximum values reported for isotopic fractionation during MOX62 (18.4 and 20.8‰). Because the MOX rate of zero did not have associated isotopic fractionation, we calculated three different combinations of extreme parameters: no oxidation with no fractionation, high oxidation with low fractionation and high oxidation rates with high associated fractionation. We then took the maximum and minimum estimates of αapp, based on this analysis, as upper and lower bounds for αapp estimated using our actual measurements of MOX.

Mesocosm mass balances

To quantify methanogenesis in the oxic mixed layer of ambient enclosures, mass balances from the ambient control treatment enclosures were constructed (Supplementary Table 2) for comparison with other lake-based CH4 fluxes using equation (7):

Here rates of methanogenesis (CH4gross) are estimated by summing the rate of change in storage (ΔCH4), defined as the rate of increase in the ambient CH4 pool, with the rates of CH4 evasion (E) and oxidation (O), minus the horizontal influx of external CH4 via diffusion across the walls of the enclosures (M; detailed in the Supplementary Methods). In all cases, mass balance terms were first standardized to the entire mixed layer or surface of the enclosure (as mmol per enclosure per day) in order to combine aerial (E and M) and volumetric (ΔCH4 and O) processes on each sampling date.

The enclosure mixed layer average depth of 1.68 m (Supplementary Fig. 1) was used to calculate both mixed layer volume, and the surface area of the enclosure wall that contributed to CH4 influx from the surrounding lake environment. The change in storage ( Δ CH 4 ) was calculated as the rate of change in ambient CH4 concentrations in each bag as a function of time using the slope of the regression line as determined by least-squares regression analysis, then averaged and adjusted volumetrically to the enclosure scale (mmol per enclosure per day; Supplementary Table 2). The rate of MOX (O) for day 7 of the experiment was used for all calculations. Finally, estimates of CH4gross were converted to units of mmol m−3 per day for comparison with other lake-based measurements (Table 1).

Whole-lake CH4 dynamics

Enclosure measurements and mass balance results were compared with the surrounding Lac Cromwell environment. To quantify summertime rates of ebullition in Lac Cromwell (Table 1), bubble traps were fixed at five permanent sampling sites along a transect from the littoral to pelagic at ~1, 2, 3, 5 and 7 m depths. Bubble traps were left in place over the entire sampling period and collected monthly in June, July and August 2012. Traps consisted of an inverted funnel (63.5 cm diameter) suspended at 0.5 m below the surface of the water (0.25 m at the shallowest sample site). A graduated 1-l glass bottle was attached to the funnel by gluing the sides of the bottle cap to the neck of the funnel to make a gastight seal. The bottles were covered with aluminum foil to prevent overheating or light exposure of the collected gas. Bottles were filled with water upon deployment, so the gas accumulated by displacing water in the bottle. Bottles were collected once a month or until a volume greater than 250 ml of gas was detected. All gas was carefully removed by displacement with ambient water using a stopcock cap fitted with two syringes. A subset of the gas was injected in 30 ml saline vials, and analysed upon return to the laboratory by gas chromatographic analysis as detailed for enclosure samples. As the majority of samples settled in the bottles for up to a month before collection, we assumed that the composition of CH4 in bubbles was 0.6 atm or 60%, based on a subset of samples analysed that were collected within days of sedimentary release. This assumption is consistent with literature estimates of bubble CH4 concentration32.

Diffusive evasion of lake CH4 was estimated by first measuring concentrations of CH4 on a monthly basis near each bubble trap site along the transect. Concentrations ranged between 0.63 and 2.09 with an average of 1.21 μM, and did not vary across sites (ANOVA, P=0.98, F=0.10). As detailed above for the enclosure mass balance, atmospheric fluxes were estimated by applying the average 24 h gas exchange coefficient measured at 7 m depth along the transect (mean k CH 4 =0.65 m per day, Supplementary Fig. 2).

Whole-lake CH4 and CO2 diffusive fluxes were calculated by applying average areal estimates of diffusion to the entire lake surface area (102,000 m2). As ebullition was only detected at depths <3 m, average rates of ebullition (0.31±0.66 mmol m−2 per day) were applied to the lake surface <3 m in depth (49,166 m2). This area was estimated by determining the area of the lake <3 m in depth from a hypsographic curve for Lac Cromwell67, and existing bathymetric information for Lac Cromwell68.

Data compilation for meta-analysis

To assess the relationship between algal dynamics and CH4 in open water environments, we compared the results from our experimental enclosures with directly obtained results from large (>10 km2), deep (>10 m maximum depth) boreal lakes45 and other freshwater and marine data from the literature (raw data and selection methods tabulated in Supplementary Table 3). Literature data include only studies where CH4 and chlorophyll a were measured simultaneously, except for Lake Baikal, where detailed studies of CH4 and chlorophyll a were measured independently, but overlapped temporally and thus were included. To remain consistent with the sampling approach from our study, and sampling of boreal lakes, only near-surface, open-water results from the centre of the lake or furthest from the coastal environment (in marine studies) were used. Summertime results were taken if seasonal data were presented. If surface water results were given as a range, the midpoint of that range was chosen. If multiple samples were taken at one site, then results were averaged.

Additional information

How to cite this article: Bogard, M. J. et al. Oxic water column methanogenesis as a major component of aquatic CH4 fluxes. Nat. Commun. 5:5350 doi: 10.1038/ncomms6350 (2014).