Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 Using measured rates of bubble release and diffusive gas transport (also termed surface aeration), we address the role of these transport mechanisms in emissions of nitrous oxide and methane from four streams. While ebullition in streams and rivers has received little study, we found that ebullition was an important mode of methane emissions, contributing 20%–67% of methane emissions (among streams). Nitrous oxide emissions via ebullition were negligible (<0.1% of diffusive emissions). Total methane emissions (ebullition + diffusive transport) were over ten times greater than N2O emissions in terms of CO2 equivalents. Rates of bubble release were highly variable, ranging from 20 mL m−2 d−1 to 170 mL m−2 d−1 (seasonal average among streams, with volumes reflecting ambient temperature and pressure). Methane was the most abundant of the bubble gases that were measured (26% by volume on average among streams), followed by carbon dioxide (1% on average) then nitrous oxide. Average bubble nitrous oxide concentrations were below atmospheric mixing ratios for the majority of streams; however, one stream showed concentrations as high as 3600 ppbv. Sediment characteristics were strong predictors of bubble composition. Concentrations of methane and nitrous oxide were positively related to the proportion of fine sediments. High methane concentrations in bubbles were related to high sediment organic carbon.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
1.1. Aquatic Ecosystems as Sources of Greenhouse Gases
 On a global basis, lakes, reservoirs and wetlands have a significant role in global greenhouse gas balances. Lake emissions have been estimated at 70–150 Tg y−1 of carbon in the form of carbon dioxide (CO2-C) [Cole et al., 2007] although a recent reassessment places this number as high as 530 Tg CO2-C y−1 [Tranvik et al., 2009]. Methane (CH4) emissions from lakes are on the order of 8–48 Tg y−1 [Bastviken et al., 2004], but this number may also be an underestimate [Walter et al., 2006]. Wetland CH4 emissions are globally more significant than lakes at between 145 and 231 Tg of CH4 per year [Denman et al., 2007]; however, many wetlands appear to be net sinks of CO2 [Bridgham et al., 2006]. Reservoirs are important not only for their large emissions, but also because they are considered a form of land-use change, and are factored into anthropogenic greenhouse gas budgets. Emissions from reservoirs are estimated at 280 Tg of CO2-C [Cole et al., 2007], and 70 Tg of CH4 per year globally [St. Louis et al., 2000].
 Fluvial systems have received less study in terms of their role in CH4 emissions [Bastviken et al., 2011]; however, existing data suggest that they function as CH4 sources to the atmosphere [Jones and Mulholland, 1998]. The greenhouse gas receiving significant study in fluvial ecosystems is nitrous oxide (N2O), due to an early model suggesting that agricultural and sewage N inputs into streams, rivers and estuaries could lead to 20% of the world's anthropogenic N2O emissions [Seitzinger and Kroeze, 1998]. Based on accumulating evidence that river and stream emissions are overestimated, the IPCC has recently downscaled N2O emissions estimates from fluvial ecosystems [Eggleston et al., 2006]. However, new data suggests IPCC methods may now significantly underestimate riverine N2O fluxes [Beaulieu et al., 2011]. Carbon dioxide emissions from rivers are estimated as 230 Tg CO2-C per year [Cole et al., 2007]. Estimates of stream fluxes (based on stream heterotrophy, which likely represent a lower bound for emissions) suggest that stream emissions are greater than river emissions at 320 Tg CO2-C y−1 [Battin et al., 2008]. These high fluxes are supported by measurements from boreal streams where streams function as strong CO2 sources in the landscape [Teodoru et al., 2009]. In sum, inland waters may emit as much as 1400 Tg CO2-C per year [Tranvik et al., 2009].
1.2. Production and Emission of Greenhouse Gases
 Numerous pathways contribute to greenhouse gas production. Classical denitrification can lead to N2O production, as do the pathways nitrification and nitrifier-denitrification. The two dominant CH4 production pathways are acetate fermentation (aceticlastic methanogenesis) and H2-dependent (hydrogenotrophic) methanogenesis. Aerobic respiration, denitrification, aceticlastic methanogenesis and other pathways contribute to CO2 production.
 Gases are emitted from aquatic ecosystems by three major mechanisms. The first and most frequently studied is diffusive transport (also referred to as air-water gas exchange or surface aeration). However, plant and bubble-mediated (ebullitive) transport can be important, particularly for CH4 [Dacey and Klug, 1979; Walter et al., 2006]. Gas bubbles can be formed in sediments when the partial pressure of the gases exceeds the sum of the pressure on the sediment and the surface tension of water. Hence, bubble production depends on rates of gas production as well as water temperature, water depth, and barometric pressure. Bubble emissions tend to be episodic, triggered by shear stress [Joyce and Jewell, 2003], lowering of the water table [Chanton et al., 1989], or periods of low atmospheric pressure [Mattson and Likens, 1990].
 A major limitation of our understanding of the importance of streams and rivers in greenhouse gas emissions is in our focus on diffusive gas transport across the air-water interface. The importance of bubble-mediated emissions, especially of N2O, has largely been overlooked with the assumption that diffusive transport is the dominant emission mechanism. We assess this assumption in a series of streams in southern Ontario, Canada by contrasting rates of gas transport via these two mechanisms. We also assess rates of CH4 emissions and the role of ebullition in CH4 transport – a topic that has received relatively little study in small streams. Finally, we identify predictors of gas concentrations and fluxes among streams.
2. Site Description
 We studied four 2nd–5th order (Strahler) streams in southern Ontario, Canada. These are north temperate, hard water streams, with significant agricultural land use in their catchments. Monoculture (continuous cultivation of annual crops) constitutes 13%–22% of land area, mixed cropping (rotations of annual and forage crops) constitutes 10%–31% of land area, rural land uses (forage crops, idle lands, hay, pasture and marginal lands) include 12%–44% of catchment areas, and wetlands constitute 9%–25% of land area. Urban areas are <3% of catchment areas in all streams [Ontario Ministry of Natural Resources, 2006]. The highest proportion of wetland area is observed in Layton Creek (25% of catchment) and Jackson Creek (21% of catchment), while the greatest proportion of wooded area is in the Black River (24%) and lowest proportion in Mariposa Brook (9%). Mariposa Brook has the greatest proportion of land in agricultural land use (monoculture, mixed or rural). We sampled two sites (separated by 3.9–8.2 km; Table 1) in each stream. Land use data reflect the most downstream site sampled [Baulch, 2009]. Study streams and sampling sites were selected to represent small streams of this region, while ensuring easy access, and minimizing the number of inflows between sites. All streams were shallow (<50 cm mean depth), with relatively low slope (0.25 to 1.43°; Table 1). A map of study sites, and additional site characterization data are available elsewhere [Baulch et al., 2011a].
Table 1. Site Coordinates, Measured Gas Transfer Coefficients and Discharge at Time of Gas Transfer Measurements, as Well as Discharge Over the Study Period, Slope, and Distance Between Study Sites
K600 (day−1) at 20°C
Discharge (L s−1)
Mean Discharge (L s−1) During Study Period ± Standard Error
Slope (degrees), Distance Between Sites (km)
N44° 09.476, W079° 21.673
SF6, O2 isotope
N44° 11.815, W079° 20.150
N44° 15.976, W078° 28.383
N44° 16.799, W078° 24.850
N44° 16.071, W078° 59.075
SF6, O2 isotope
N44° 14.501, W078° 58.876
N44° 23.148, W078° 52.545
N44° 20.915, W078° 52.084
3. Materials and Methods
 At each site, we measured stream discharge and water chemistry, the rate of bubble release, bubble gas concentrations, and the rate of gas emissions driven by diffusive transport (determined using measurement of piston velocity and dissolved gas concentrations). Sediment analyses were also performed at each site.
3.1. Water Chemistry, Stream Discharge, Temperature, Pressure
 Water samples were obtained in ∼500 mL PET bottles from midstream. A total organic carbon analyzer (Shimadzu TOC-V analyzer) was used to analyze concentrations of dissolved organic carbon (GFF filtered). pH was measured using electrodes (Mantech PC titrator; Mantech Inc, Guelph Canada). Nitrogen species were GFF filtered then analyzed colorimetrically. Nitrate was analyzed using the red azo dye method, following reduction to nitrite, then corrected for measured nitrite [Ministry of the Environment, 2001]. Ammonium was analyzed using the phenate-hypochlorite method [Ministry of the Environment, 2001]. Samples for total phosphorus analysis were obtained directly in glass vials, then analyzed using the ammonium-molybdate-stannous chloride method [Ministry of the Environment, 1994].
 Depth, velocity, and discharge were measured at each site using a Swoffer 2100 velocity meter and applying the velocity-area method. Air and water temperatures were measured using a Fisherbrand Traceable Thermometer. Barometric pressure was measured using a handheld barometer (Kestrel).
3.2. Rate of Bubble Release
 We deployed transparent inverted-funnel style bubble traps [Molongoski and Klug, 1980] from late May to 9 October 2007 to determine the volume of bubbles emitted from the benthic substrate. The initial sampling design involved only three traps being installed at each site. After the first sampling, we installed additional traps (bringing totals to 5–10 traps per site), to account for observed high spatial variability. Our sampling design (including the number of streams, sites, and traps per site) represents a tradeoff between more spatially intensive sampling, resulting in more constrained emissions for a single stream, and more extensive sampling across streams.
 Bubble traps had a bottom area of ∼530 cm2 which narrowed into an upper trap portion with a diameter of ∼2.2 cm. Traps were affixed to a galvanized metal pole (which was inserted into the substrate) and installed 10–30 cm below the water surface and at least 10 cm above the benthic substrate, allowing water flow below the trap. The minimum water depth which we could install bubble traps was 25 cm. The depth of traps at the end of the sampling period is reported in Table 2. Data for the first two weeks following trap installation was omitted to limit effects of initial sediment disturbance on ebullition rates. Although the use of this apparatus in flowing water may also contribute to sediment disturbance, very slow stream velocities during the study period (Table 2) suggest these effects were small. Slow water flow also ensured bubbles emitted from below the trap were transported upwards into the vessel (rather than moving downstream with water flow).
Table 2. Mean Stream Velocity, Sediment Characteristics and Water Depth at Bubble Trap Locations at the End of Sampling Perioda
Mean values are indicated with standard error in parentheses.
During the period for which emissions via ebullition and diffusive transport are contrasted (11 June–11 October 2007).
 Traps were filled with water (no headspace) at the start of each sampling period. Gases emitted from below the trap surface accumulated in the upper (narrow) portion of the funnel and were sampled to determine bubble volume using a luer-lock fitting at the top of the trap attached to a 60 mL syringe (approximately every 2 weeks). Where very high emissions occurred, an inverted water-filled 1000 mL graduated cylinder was used to determine the volume of bubbles emitted. Care was taken to avoid disturbing the benthic substrate when sampling the traps, including maintaining a minimum (lateral) distance of 1 m from the traps, and having one person continuously monitor the traps to ensure no bubble plumes were released from the bottom substrate as a consequence of our sampling. When there was evidence of muskrats tampering with the traps, or evidence that sampling induced substrate disturbance (affecting the volume of gas collected), affected data were excluded. This resulted in the exclusion of 4% of bubble flux measurements.
3.3. Gas Sampling
 Water samples were obtained for dissolved gas analysis from midchannel, approximately 20 cm below the water surface. Sampling was performed at each study site on 9–10 occasions (between 15 May to 30 October 2007). Samples were taken in 125 mL borosilicate glass serum bottles with no headspace, capped with pre-baked butyl rubber stoppers and field preserved with saturated mercuric chloride (0.4% final v/v).
 Following sampling for dissolved gases, trapped bubble volume and water chemistry, we obtained bubble samples from the benthic substrate for analysis of bubble gas concentrations. Samples were obtained by walking transects of the stream channel (perpendicular to streamflow at each site) and periodically disturbing the benthic substrate with a funnel attached to the body of a 60 mL syringe and luer-transfer port. Multiple transects were walked until at least 30 mL of gas was obtained, which was then transferred into a pre-evacuated vial. This process was then repeated 3 times to obtain triplicate samples which were analyzed using gas chromatography (GC). Bubble gas sampling was performed on 9 or 10 dates at each study site (see results for exact timing).
 We opted to analyze fresh bubbles from the sediments rather than those trapped within the headspace of bubble traps because bubbles within the traps will undergo equilibration with stream water. However, fresh bubbles may have undergone further equilibration with surrounding pore waters prior to natural emission. In addition, bubbles undergo equilibration as they travel through the water column to the surface. Because the length of the apparatus used in fresh bubble sampling (∼40 cm) was sometimes less than the stream depth (Table 2), the extent of equilibration may have differed between naturally emitted samples and the values we report. This effect is most important for highly soluble or reactive gases such as H2S and CO2. In contrast, effects on N2O and CH4 are expected to be minimal [Chanton and Whiting, 1995]. However, our CO2 measurements could represent a slight overestimate of the CO2concentration the same bubbles would attain at the water-air interface.
3.4. Gas Analyses
 Dissolved gas samples were analyzed using a Varian CP-3800 gas chromatograph following headspace equilibration. CH4 was analyzed by FID following separation using a Hayesep D column (80/100 mesh, 0.5 m, 1/8” stainless steel). N2O was analyzed by ECD using the same column. Equilibrium concentrations for dissolved gases were calculated using measured temperature and barometric pressure [Yamamoto et al., 1976; Weiss and Price, 1980], assuming atmospheric concentrations of 320 ppbv N2O and 1.85 ppmv CH4. Given the long atmospheric residence time of N2O, it is expected to be well mixed in the atmosphere [Stein and Yung, 2003], which supports our assumption of atmospheric mixing ratios. However, regional variation in CH4 concentrations is likely, and our assumption of constant atmospheric concentrations may bias our flux estimates. Where dissolved CH4 is highly supersaturated, this will have only small effects on CH4 flux estimates (i.e., see equation (1)).
 All samples were obtained during the daytime. This may result in a slight underestimate of diffusive CH4 fluxes. Diel sampling of these streams indicated that sampling near solar noon underestimated diffusive CH4fluxes by 9% in these streams, when compared to time-weighted samples over a 24-h sampling period [Baulch, 2009]. Daytime sampling underestimates diffusive N2O flux from these streams by ∼5% [Baulch, 2009]. Lateral variation in dissolved gas concentrations across the stream channel was observed (average of 6% for N2O, 11% for CH4), but systematic differences between samples obtained from the middle of the stream channel and samples obtained near stream banks were not observed.
 Bubble samples were analyzed on two instruments. Approximately 60% of samples were analyzed on the Varian gas chromatograph as described previously. The remainder of samples were analyzed using a Shimadzu GC-2014 with a Tekmar 7050 autosampler and 9 mL vials with thick butyl rubber septa. Gases were separated on a Poropaq Q column (80/100 mesh size). Samples for CH4 analyses were diluted with UHP N2 or helium. Duplicate bubble samples run on both gas chromatographs showed agreement within 3% (CH4), 2% (CO2), and 7% (N2O) between instruments.
3.5. Gas Emissions via Ebullition
 The masses of N2O, CO2 and CH4released by ebullition were calculated by determining the number of moles of total bubble gas released (using measured volume, air temperature and atmospheric pressure), and multiplying this value by measured gas concentrations in the fresh bubbles (at the start and end of the sampling period) for each site. Average, time-weighted emissions were then calculated, accounting for the duration of each sampling interval. Error in ebullitive flux estimates was estimated using Monte Carlo uncertainty analysis [Beck, 1987]. This method allows us to constrain uncertainty associated with both the volume of bubbles, and bubble gas concentrations. The volume of gases emitted and the concentration of gases were represented as a normal distribution using measured means and standard deviations, and truncating distributions to omit negative values [Decisioneering, Inc., 2001]. Ebullitive emissions of both N2O and CH4 were estimated by subsampling these distributions 1000 times using stratified random (Latin Hypercube) sampling in the software package Crystal Ball 2000.2 (Decisioneering, Inc., Denver, Colo.), resulting in a probability distribution for ebullitive gas fluxes.
3.6. Gas Emissions via Diffusive Transport
 Gas fluxes due to diffusive transport were estimated using the two-layer model of diffusive gas exchange [Liss and Slater, 1974]:
where k is the piston velocity (m h−1). The piston velocity is sometimes termed the gas transfer velocity, and is equal to the gas transfer coefficient K times the mean depth; CS is saturation concentration (μmol m−3); CL is the measured concentration (μmol m−3).
 In this model, there are two interfacial layers, one in the water at the water surface, and one in air, immediately overlying the water surface. For CH4 and N2O, air-phase resistance is negligible [Liss and Slater, 1974], and the main resistance to gas transfer across the air-water interface is molecular diffusion through the surface layer of the water. The depth of this surface layer, which dictates the length of the diffusion pathway, is related to turbulence. Within shallow fluvial ecosystems, benthic turbulence is the primary process affecting rates of gas exchange [Raymond and Cole, 2001].
 Our measurements of rates of air-water gas transfer (Table 1) are described in detail elsewhere [Baulch et al., 2011a]. Briefly, we measured rates of air-water gas transfer via addition of a gas tracer (sulphur hexafluoride; analysis by GC-ECD as for N2O) and a conservative tracer at six of our eight study sites [Kilpatrick et al., 1987; Baulch, 2009]. We also used diel O2 dynamics to constrain the piston velocity [Venkiteswaran et al., 2007]. This method uses an O2-mass balance model, iteratively fit using constrained range of possible rates for respiration (0–5000 mg O2 m−2 h−1), photosynthesis (0–5000 mg O2 m−2 h−1) and piston velocity (0.01 to 0.50 m h−1) [Venkiteswaran et al., 2007]. The model was run with different parameter combinations (i.e., rates of respiration, photosynthesis and piston velocity) to obtain the lowest sums of squares between measured and modeled data. All model runs where an r2 of >0.8 were averaged to obtain a mean estimated piston velocity.
 In one stream (Black River) where substrate, depth, and flow varied little between sites (mean differences: proportion substrate >0.25 mm, 11%; mean depth 16%; mean discharge 9%), we used upstream gas transfer measurements to estimate gas transfer at our downstream site. At all other sites, one or two direct measurements of gas transfer were made at each study site (Table 1). Measured rates of gas transfer were converted to the stream temperature [Thomann and Mueller, 1987] and gas of interest (from SF6 or O2 to CH4 or N2O) using Schmidt number scaling [Wanninkhof, 1992]. We assumed a Schmidt number exponent of 2/3 which reflects smooth surfaces (in contrast to rough surfaces, where a Schmidt number exponent of 1/2 is typically assumed [Jähne et al., 1987]). We report rates of air-water gas transfer in customary units, as the gas transfer coefficient for CO2 at 20°C (K600; Table 1).
 Depth, velocity, slope and substrate type are important factors affecting benthic turbulence, and depth and velocity can vary over time (Table 1). To account for this temporal variation which is likely to affect rates of gas transfer, we use two approaches. First, in 5 of 8 study sites we use two measurements of gas transfer coefficient (Table 1) and report the range of estimates resulting from these measured values (in the remaining 3 sites, only a single measurement was made). This assumes that gas transfer coefficient was constant through the study period. Second, we applied a model-based approach to extrapolate measured values to different flow conditions (varying depth and velocity) [Moog and Jirka, 1998]. We calibrated three empirical models [O'Connor and Dobbins, 1958; Bennett and Rathburn, 1972; Owens et al., 1964] to individual sites, using the ratio of measured K to modeled K under the flow conditions when K was determined [see equation (2); Moog and Jirka, 1998], then applied these calibrated empirical models to predict gas transfer coefficient, using measurements of stream depth and velocity for each sampling occasion. The use of three models is intended to reflect error associated with model assumptions.
where KPC is the calibrated gas transfer coefficient, Km is the measured gas transfer coefficient, KP is the predicted gas transfer coefficient under measurement conditions, and n reflects the number of measurements. This equation is presented in its original form, which uses the gas transfer coefficient, a value equal to the piston velocity (k) divided by depth.
 We used KPCvalues to determine whether dissolved gases at upstream sampling sites were likely to affect gas composition at the downstream location. In all cases, the distance between sampling sites was great enough that dissolved gases at upstream sites had minimal effects (<5%) on concentrations at downstream sites based on first-order loss equations [Chapra and Di Toro, 1991]:
where Dist5% is the distance at which 5% of a gas will remain in solution, v is stream velocity and KPC is the calibrated gas transfer coefficient.
 Dissolved gas concentrations and bubble gas concentrations are reported for the period 15 May to 30 October 2007. Emissions via diffusive transport and ebullition are calculated from 11 June to 11 October 2007 for the Black River, Jackson Creek and Layton Creek. However, in Mariposa Brook, where changes in discharge led to changes in piston velocity and added considerable uncertainly to our diffusive transport estimates, we restrict the time series to 3 July to 12 September 2007 (upstream site) and 11 June to 12 September 2007 (downstream site). Mean (time-weighted) diffusive flux estimates for the study period were determined using mean fluxes for each sampling interval (mean of start and end of each period), then multiplying by the duration of that interval, and dividing by the length of the study period.
3.7. Sediment Analyses
 Duplicate sediment cores were obtained from under each bubble trap at the end of the study (October 2007). A PVC core tube (10 cm diameter, 7 cm depth) was pressed into the stream substrate. The base was sealed from below using a plastic sheet or hand, then the core tube was removed, and its contents were transferred to pre-labeled bags. At two sites (Mariposa Brook, downstream site, Jackson Creek, downstream site), we could not obtain cores from under all traps, because the rocky substrate prevented use of our core tube. Samples were split to allow analysis of dry weight percentage (comparing wet weight to weight following drying at 60°C) and sediment size fractions. Samples for size fractionation were wet sieved following treatment with a dispersant to prevent clumping (sodium hexametaphosphate) [Poppe et al., 2000]. Sample retained on the largest sieve (2 mm) was discarded. Material passing through the smallest sieve (0.0625 mm) was poured into evaporating vessels, dried, weighed, and corrected for the addition of sodium hexametaphosphate. Material collected on intermediate sieves (1, 0.5, 0.25, 0.125, 0.0625 mm) was dried and weighed.
 Samples for carbon analysis were sieved through the 1 mm sieve, dried at 40°C, and ground using a mortar and pestle. Duplicate subsamples were acidified (excess of 6% sulphurous acid at 70°C) prior to carbon analysis. Each subsample was dried, ground again, and split into two additional subsamples prior to analysis using an Elementar CNS analyzer [Skkjemstad and Baldock, 2008].
 Correlation analyses were used to assess relationships between sediment characteristics and rates of bubble emissions, bubble gas concentrations and gas flux driven by diffusive transport (mean of all estimates), and water chemistry variables and total fluxes of CH4 and N2O [JMP, 2007]. In addition, we assessed correlations between bubble N2O and CH4 concentrations and dissolved N2O and CH4 concentrations [JMP, 2007]. To meet the assumptions of the analyses, some variables were log-transformed. The nonparametric Spearmansρwas used for assessing relationships with sediment carbon (variables were not log-transformed for this analysis) [JMP, 2007].
 Three of the four streams (Black River, Layton Creek, Mariposa Brook) were supersaturated with N2O, indicating that they functioned as N2O sources, despite relatively low NO3− concentrations (mean [NO3−] among sites was 10.2–145 μmol L−1 (Table 3 and Figure 1). Jackson Creek was frequently undersaturated in N2O (in 60% of samples (Figure 1)). Mariposa Brook showed the highest degree of N2O supersaturation, followed by Layton Creek (Figure 1). N2O emissions were highest in Layton Creek and Mariposa Brook, although there is uncertainty in diffusive emissions estimates associated with the assumption either of constant rates of gas transfer, or the use of different models in scaling changes in gas transfer (Figure 2). N2O emissions via diffusive transport were related to the proportion of clay/silt in the sediments (r = 0.74 (Table 4)). Nitrate is the only water chemistry variable which is a significant predictor of N2O fluxes (total or diffusive transport fluxes-nitrate concentrationsr = 0.83, p= 0.01). Bubble-mediated N2O fluxes were extremely low (Table 5).
Significant correlations (p< 0.05) are shown. NM indicates that relationship was not assessed. No significant relationships were found with (log-transformed) diffusive CH4 flux. Sediment size fractions of 0.25–0.5 mm and 0.125–0.25 mm were not correlated with reported variables.
Variable is log-transformed.
Note that reported correlations for sediment carbon are related by Spearman ρ, and variables are not log-transformed for this analysis.
 CH4 was always well above saturation even in the oxic stream water (Figure 1). Layton Creek showed the highest degree of supersaturation, followed by Mariposa Brook. The proportion of CH4 transport via ebullition was greatest in the two streams (Mariposa Brook, Layton Creek) with the highest proportion of silt and clay (Tables 2, 5). If we assume bubble fluxes are representative of the stream or reach scale, ebullition was responsible for 10%–80% of CH4 transport (Table 5). Interestingly, relationships between N2O and CH4 are apparent (Table 4). Rates of diffusive N2O transport were related to dissolved CH4 concentrations (r= 0.76). Total (log-transformed) CH4 emissions were not significantly related to any measured water chemistry parameter. The global warming potential of stream methane emissions was greater than that of N2O emissions, if both diffusive transport and ebullition are considered (Table 6).
Table 6. Total Emissions (Bubble + Surface Aeration) in CO2 Equivalents
Calculations based on global warming potentials for a 100-year time period (GWP100). GWP100 is the warming potential associated with 1 kg of a trace gas relative to 1 kg of the reference gas CO2. The GWP100 is 298 for N2O and 25 for CH4 [Forster et al., 2007]. Calculations use mean of data from Figure 2.
A negative sign indicates that net uptake is occurring.
 Spatial variability in bubble fluxes was extremely high both within sites (as indicated by error bars, Figure 3), among sites (within a stream), and among streams. Seasonal average bubble release rates ranged widely among streams and between sites (52–73 mL m−2 d−1 at the two Black River sites, 14–24 mL m−2 d−1 in Jackson Creek, 72–257 mL m−2 d−1 in Layton Creek and 111–232 mL m−2 d−1 in Mariposa Brook). No clear seasonal trend in release rates was apparent across sites (Figure 3).
 In our study streams, CH4 concentrations in bubbles ranged from <5% by volume to approximately 70% (Figure 4), but average values were relatively similar among sites and streams, near 20%–30% (Table 7). CH4 concentrations in bubbles were related to dissolved CH4 (r = 0.74). CO2 concentrations were much lower than CH4 concentrations, averaging 0.4%–1.8% (Table 7). N2O concentrations were consistently the smallest measured proportion of bubble content, often below atmospheric mixing ratios (Figure 4). Concentrations of N2O were greatest in bubble samples obtained from the downstream site in Mariposa Brook (Figure 4). On average, 70% of bubble volume was not accounted for in CH4, CO2 or N2O analyses.
Table 7. Gas Concentrations in Bubbles in Freshwater Ecosystems (This Study and Published Data)
CH4 (% by volume)
CO2 (% by volume)
This study: Streams (seasonal means)
Mariposa Brook (includes additional data from 3rd site)
Past studies: Streams, rivers and estuaries
Stouffville Creek (seasonal mean, 2 sites, H. Baulch, unpublished data)
 Sediment characteristics were correlated with bubble gas concentrations. A significant positive relationship was shown between the volume of bubbles emitted, and the proportion of clay and silt (r = 0.75, Table 4). Bubble N2O concentrations were also significantly related to the proportion of clay/silt (r = 0.73, Table 4). Percentage of carbon in sediment was significantly related to bubble CH4 content (ρ = 0.97; Table 4). Bubble CH4 concentrations were also positively related to the proportion of fine (0.0625 mm–0.25 mm) sediments (r = 0.76, Table 4) and negatively related to the dry weight (r = −0.92, Table 4). Bubble CO2 concentrations were positively correlated to the proportion of coarse sediment fractions (r = 0.85, Table 4).
5.1. Gas Emissions
 With the addition of these data, ebullitive and surface-aeration-mediated CH4 fluxes from rivers have now been compared in 7 streams and rivers, with results indicating ebullitive fluxes may be responsible for 10%–80% of CH4 transport. Broader ranges may be observed across the full suite of freshwater aquatic ecosystems (∼0%–96% (Table 5)). When we contrast the emissions of CH4 and N2O in terms of global warming potential (GWP100), we find that despite greater research effort characterizing N2O emissions from streams, CH4 emissions exceeded N2O emissions in all of our study streams (Table 6). This result has also been shown in New Zealand streams, although not uniformly across all study sites [Wilcock and Sorrell, 2008]. Although this type of comparison has been made for only a small number of streams, we suggest that CH4 emissions from streams may merit greater attention. In addition, CH4 transport through macrophytes might further increase our emissions estimates [Yavitt and Knapp, 1995]. Plant-mediated N2O fluxes can also be significant [Rückauf et al., 2004].
 It is important to note that our data likely reflect the maximal importance of ebullition in these study systems, as the bulk of our work was done during summertime months, which tend to be associated with periods of maximum bubble production [Martens and Klump, 1980b]. As well, we note several important error sources. Error associated with an assumption of constant rates of gas transfer is significant in some cases (Figure 2) and use of the model-corrected values we present may be more appropriate. Spatial variation in bubble gas concentrations and volumes emitted was considerable, leading to significant uncertainty in estimates of bubble gas fluxes (Figure 2). However, the extent of error was not great enough to affect our key findings that ebullition is an important mode of CH4 transport from streams but is not important to N2O transport. In addition, our results indicate that summertime CH4 emissions are more important than N2O emissions from these systems. Beyond the error associated with variation in our measurements is the potential for our sampling techniques to affect the stream environment. Sampling methods were designed to help minimize potential impacts, nonetheless, further developments in hydroacoustic methods may ultimately provide a better method for estimation of bubble fluxes. Hydroacoustic methods have proven useful to constraining bubble emissions in lakes [Ostrovsky et al., 2008; Vagle et al., 2010]. These methods would have the advantage of not requiring sediment disturbance, although, can be affected by the presence of fish in the water column [Ostrovsky et al., 2008], and have not yet been applied in shallow systems, in part due to the size and weight of required equipment.
 In addition to this study, which indicates that nitrate is a predictor of nitrous oxide fluxes, longer-term study and study across a larger range of streams has also demonstrated a strong relationship between nitrate concentrations and annual nitrous oxide emissions [Baulch, 2009; Baulch et al., 2011b]. Shorter-term relationships between N2O and NO3− have also been reported across a wide range of streams [Beaulieu et al., 2011; Baulch et al., 2011b]. Bubble-mediated N2O transport is not significant. This suggests that exclusion of bubble-mediated transport from ongoing efforts to estimate N2O fluxes is justified (Table 5). However, there is some evidence that ebullition of N2O could be important in localized areas affected by springs [Clough et al., 2006].
 Bubble CH4 and stream water CH4 were related (r = 0.74, Table 4). Because the water column in these streams is oxygenated [Baulch, 2009], pelagic CH4 production is likely to be low. Instead CH4 production is expected to be a primarily benthic process, with dissolved CH4 in streams resulting from diffusion of methane from the sediments or groundwater. Dissolution of CH4 in bubbles may contribute to dissolved CH4 in stream water. However, rapid bubble rise through these very shallow water columns is expected to allow only limited dissolution of bubble CH4 [McGinnis et al., 2006]. Not surprisingly, concentrations of dissolved CH4 in stream water were much lower (<6%) than would be predicted if stream water were in equilibrium with the benthic bubbles. This reflects dilution of CH4 from benthic production, as well as loss to gas exchange and methanotrophy. Temporal variation in CH4 production and transport may also affect relative concentration of CH4 in bubbles and dissolved in stream water.
5.2. Volume of Bubbles Emitted and Their Composition
 The volumetric rate of bubble release has not been previously reported for streams; however, our emissions fall below those of thawing thermokarst lakes and within the range shown by most lakes (Table 8). While this study focuses on the importance of bubbles in mediating gas release, bubble release can have important effects upon sediment structure, and transport processes including rates of solute transport [Haeckel et al., 2007; Martens and Klump, 1980b], particle transport [Klein, 2006], internal phosphorus loading [Saarijärvi and Lappalainen, 2002], and the release of sediment contaminants [Fendinger et al., 1992].
Table 8. Rates of Bubble Release Reported From Lakes and Estuaries
 Both CO2 and CH4 concentrations in bubbles fall within ranges reported in the literature (Table 7); however, observed N2O concentrations in trapped bubbles were much lower than shown in two previous studies [Clough et al., 2006; Higgins et al., 2008]. By applying the Henry's law coefficient for N2O to measured bubble concentrations, we inferred the concentration of dissolved N2O that would be in equilibrium with bubbles in our study streams. On average, measured stream water N2O concentrations were 2.5 times higher than concentrations that would be expected based on bubble concentrations. This suggests that areas of bubble production are not near areas of high N2O production and bubbles are separated from significant influence from overlying stream water.
5.3. Sediment Structure
 Sediment characteristics were related to gas emissions, and the concentration of gases in bubbles. There are many reasons these relationships may exist. The relationship between the volume of bubbles emitted and proportion of fine sediments may be due to either more tortuous (i.e., long, indirect) pathways, or lower permeability [Ullman and Aller, 1982; Iversen and Jørgensen, 1993] in fine sediments contributing to more limited oxygen penetration (from oxygenated surface water). As a result, fine sediments may be more likely to develop anoxic conditions suitable for methanogenesis and denitrification. This may help explain the relationship between clay/silt content and emissions of N2O via diffusive transport. Rates of both nitrification and denitrification are affected by oxygen concentrations, as are the N2O yields of these processes [Kemp and Dodds, 2002; Khalil et al., 2004; Silvennoinen et al., 2008]. However, other factors such as rates of mineralization or transport of nitrogenous substrates may also vary with sediment texture. Bubble CH4 concentrations were related to percentage of carbon in sediment and the proportion of fine sediments, which tend to be associated with high oxygen demand within sediments and low oxygen penetration into sediments respectively. Bubble CO2 concentrations were also related to sediment porosity, in this case, related to coarse sediment fractions, which could reflect respiratory production of CO2 or oxidation of CH4 to CO2 under oxic conditions.
 Beyond effects on oxygen penetration, sediment characteristics may also limit transport of gases (e.g., CH4 and N2) from areas with high rates of production, contributing to bubble nucleation and growth. Physical characteristics of fine sediments such as yield strength can affect rates of bubble growth and the ability of bubbles to fracture sediments [Algar and Boudreau, 2009]. Although prior to this study, a relationship between fine sediments and bubble production has not been reported in streams, fine sediments have been associated with a high degree of CH4 entrapment in soils and rice paddies [Wang et al., 1993]. Future study of a wider suite of streams and rivers would help assess the generality of these relationships between gas fluxes and sediment character, and aid in identification of sediment types associated with hot spots of ebullition.
6. Conclusions and Implications
 The importance of inland waters in continental methane fluxes is a topic only beginning to receive attention, but existing evidence suggests that freshwater CH4 emissions may offset a large proportion of the terrestrial carbon sink [Bastviken et al., 2011]. Our results support this assertion, showing high CH4 fluxes from small streams. In fact, despite the much greater effort expended constraining nitrous oxide emissions from streams, results from this study suggest that if ebullition is considered, methane emissions far exceed N2O emissions (in terms of GWP) from streams in summer months. Ebullition was consistently an important means of CH4 transport, as was diffusive transport. N2O emissions via diffusive transport were far in excess of bubble-mediated N2O emissions. More work is required to constrain annual CH4emissions from all emissions pathways (bubbles, diffusive transport, and plant-mediated transport), understand controls on emissions across streams, and assess whether human alteration of streams (e.g., changes in temperatures, nutrients, oxygen concentrations, flow dynamics or substrate composition), particularly in urban and agricultural landscapes, may have affected not only the emissions of N2O, but also of CH4. Beyond the role of ebullition in CH4 transport, high rates of bubble release suggest the role of ebullition in transport of nutrients and contaminants from sediments merits further consideration.
 Funding was provided by an NSERC Discovery grant to P.J.D. and scholarship funding to H.M.B. We are grateful to Jason Venkiteswaran, who assisted with oxygen modeling; to numerous others who assisted in the lab and the field; and to anonymous reviewers for helpful comments on the manuscript.