We measured the escape of methane-containing gas bubbles to the water table in two microhabitats (muddy hollows and Sphagnum-plus-sedge lawns) in a raised bog in West Wales, typical of many northern peatlands. Our study was unusual in its degree of replication (14 gas traps in each microhabitat). Seasonally integrated bubble loss of CH4 to the water table did not differ significantly between microhabitats. After applying an oxidation correction to give CH4 fluxes to the atmosphere, the microhabitats still did not differ. Our results suggest that ebullition is an important mechanism of CH4 loss to the atmosphere, with mean summer rates of 11.7 mg CH4 m−2 d−1 (muddy hollows) and 6.8 mg CH4 m−2 d−1 (Sphagnum-plus-sedge lawns). Our data show that the process is spatially and temporally very variable, and that the small sample sizes of many studies (e.g., n = 5) may lead to considerable errors in flux estimation.
 Northern peatlands are important sources of atmospheric methane (CH4) [e.g., Frolking et al., 2011]. CH4 emissions from peatlands occur via three mechanisms: (i) diffusion through both water- and gas-filled soil pores, (ii) diffusion and advection through plant tissue, particularly aerenchyma, and (iii) as bubbles moving through the peat pore network, a process called ebullition. There is considerable uncertainty on the importance of ebullition, with some authors suggesting it is a negligible component of total fluxes [e.g., Green and Baird, 2012, 2013] and others suggesting it is important (> 10%) or even dominant (> 50%) under some conditions [e.g., Ström et al., 2005; Tokida et al., 2005]. Most previous studies of the process have been based on laboratory incubations of peat samples, but problems with these include (i) few peat samples or a lack of replication [e.g., Ström et al., 2005; Tokida et al., 2005]; (ii) short experimental runs of just a few days [e.g., Tokida et al., 2005]; (iii) removal of the growing surface of the peat [e.g., Baird et al., 2004], which may lead to unnatural patterns of gas build up and release; and (iv) unrealistically high incubation temperatures [e.g., Christensen et al., 2003]. Most of the (few) field studies that have separated fluxes due to ebullition from those due to other processes have had similar problems. For example, ebullition was monitored for only a few days and in just two flux chambers by Tokida et al. , while Strack et al.  used inverted funnel traps at five depths in a single location to look at bubble movement upward through the peat profile. Most recently, Comas and Wright  used inverted plastic boxes placed on the flooded surface of a subtropical peatland to investigate ebullition with time lapse cameras over a 60 day period. While their study was useful for indicating temporal patterns in ebullition, they used only four boxes on two sites covering a total area of 0.35 m2, so they were unable to investigate spatial variability in any detail.
 Therefore, despite the cited studies, the overall importance of ebullition as a CH4-transport pathway compared to diffusion and plant-mediated transport is still not known, and it is not clear whether ebullition should be represented in more detail in peatland CH4 models such as that of Walter et al. . We sought to address this knowledge gap by taking time-integrated measurements of ebullition in a typical northern peatland over a whole summer season (107 days) using a large number of measurement locations.
 Measurements were taken at a raised bog in W Wales from two types of microhabitat associated with high rates of CH4 emission: mixed Sphagnum and sedge lawns and mud-bottomed hollows (henceforth S + SL and MBH). Our main purpose was to quantify rates of CH4 release via ebullition. However, we also hypothesized that ebullition losses of CH4 would be greater from MBH than that from the S + SL because the hollow muds are unable to trap bubbles as effectively as peat, allowing bubbles to be released more readily, and because the hollows contain few vascular plants. In terms of the latter, it has been suggested that, because they act as CH4 transport pathways, vascular plants like sedges may cause reductions in pore water CH4 concentrations, rates of bubble formation, and rates of ebullition [Chanton, 2005].
2 Field Site and Methods
 Ebullition fluxes were measured over a ~15 week period between 28 May and 12 September 2009 from sites at Cors Fochno, a patterned raised bog in W Wales (4°1′W 52°30′N) typical of many northern peatlands in terms of its vegetation. The S + SL were dominated by Sphagnum pulchrum (Lindb. ex Braithw.) Warnst. and Rhynchospora alba (L.) Vahl., while those parts of the MBH that were sampled were mostly devoid of vegetation. Time-integrated measurements of ebullition were made using stoppered, inverted funnels placed in shallow pits on the peatland surface, such that the rim of the funnel was below the water table. To take measurements, the funnel is initially filled with water. Rising bubbles displace the water and are trapped, allowing the volumetric rate of ebullition to be estimated. The trapped bubbles may be extracted periodically for measurement of their CH4 concentration—denoted [CH4]—so that the mass rate of CH4 flux due to ebullition may be calculated.
 The fluxes were measured mostly every 7 days (for 11 of the 15 weeks; intervals of 5, 6, 9, and 10 days for the remainder) using 28 funnel traps, 14 in each microhabitat type. A total of 414 readings were obtained across the monitoring period, termed here the season (fewer than the possible 420 readings because of gas leaks in two funnels). We used 25 cm diameter glass funnels, the glass eliminating permeation of CH4 to the atmosphere. The funnel spouts were replaced with 3.6 × 10 cm lengths of glass tubing. Rubber stoppers sealed the top of the measurement tube, and each stopper was drilled and fitted with: (i) a 50 cm sampling tube (Tygon, 3.2 mm internal diameter) with three-way valves at each end to allow funnel gas to be sampled using syringes and (ii) a water tube (Tygon, 6.4 mm i.d.) for use when the funnel trap was being filled/refilled. The funnels were wrapped with a silvered cover to minimize solar heating of the entrapped gas and funnel water, leaving a north facing strip uncovered so that readings could be taken.
 Each funnel was installed in a shallow cylindrical pit cut into the peat surface [~30 cm × 5–10 cm (diameter × depth)]. Any vascular plants that were present were clipped rather than pulled-out, thereby minimizing the formation of artificial conduits for bubble flux. The peat beneath the S + SL funnels contained active roots of R. alba growing around the periphery of the funnel. After filling with water, the water level (and therefore bubble volume) in the funnel was read every 5–10 days between 11:00 and 12:00 GMT. Measurements were taken using binoculars from a distance of c. 2 m to minimize observer-induced ebullition, and water levels read within ± 2 mm. To measure [CH4] and estimate mass fluxes, trapped gas was sampled on 8 June (half), 11 June (half), 4 July, 9 August, and 13 September. [CH4] was analyzed using a gas chromatograph fitted with a flame ionization detector as described by Baird et al. .
 The data from the funnels give CH4 fluxes to the water table and not the atmosphere. To estimate the latter, we assumed that some of the CH4 arriving at the water table as bubbles was subsequently oxidized to CO2 in the zone above the water table. The rate of oxidation or CH4 consumption was estimated using the data published by Hornibrook et al.  for a 30 km distant ombrotrophic raised bog—Cors Caron—that has a similar climate to Cors Fochno. Hornibrook et al.  estimated the oxidation potential (potential methanotrophy) of the upper 3 cm of near-surface Sphagnum peat at their site. We assumed that bubbles arriving at the water table did not move through the oxic zone as rapid mass flow, which can result in CH4 bypassing methanotrophic bacteria [Rosenberry et al., 2006]. Therefore, our estimate of how much CH4 from bubbles reached the atmosphere is low (conservative). Rapid bypassing flow may occur when large bubbles or groups of bubbles are lost episodically [Rosenberry et al., 2006]. For the S + SL, we directly applied the values estimated by Hornibrook et al. . For the MBH, which have water tables more or less at the surface, we assumed an oxic zone of 1 cm and adjusted the values of Hornibrook et al.  accordingly.
 We compared ebullition fluxes to both the water table and the atmosphere using t tests on the square root-transformed data and Mann-Whitney U tests on the untransformed data. Where we calculated the median and interquartile range, the SPSS method was used.
 The results, on a whole-season basis, are summarized in Table 1, with more detail given below.
Table 1. Summary of the Flux (Bubble and CH4) and [CH4] Data Collected
Seasonal Bubble Flux to Water Table (mL)
Seasonal CH4 Flux to Water Table (mg)
Seasonal CH4 Flux to Atmosphere (mg)
Volume-Weighted Seasonal Mean [CH4] (ppm)
3.1 Bubble Flux
 The “weekly” (see previous section) bubble fluxes, expressed as a mean daily rate, ranged from −0.4 to 85.7 mL d−1 for MBH (n = 206) and from −0.3 to 227.9 mL d−1 for S + SL (n = 208). The frequency distribution of bubble fluxes was similar for both microform-types (Figure 1); both were non-normal with a strong positive skew. There was considerable temporal and spatial variability within the data. For example, for funnel 15 (S + SL), fluxes varied between 0.0 and 220.2 mL d−1 (n = 15 periods). For the same funnel, the flux during the last 9 days (1981.7 mL) represented 91% of the total (2178.7 mL). Spatially, seasonal bubble volumes ranged between 85.0 and 2393.1 mL for the MBH (n = 14 funnels) and 24.0 and 3641.3 mL for the S + SL (n = 14 funnels).
3.2 [CH4] of Bubble Samples
 [CH4] in bubble samples taken from the funnels ranged from ~0% to ~51%. Bubble samples from the MBH funnels had a lower average volume-weighted [CH4] (~17%) than those from the S + SL (~20%), but this difference was not significant (two-sample t test, untransformed data, p(2) = 0.70). Overall, CH4 concentrations increased between June and August and decreased between August and September.
3.3 CH4 Flux to the Water Table in Individual Bubble Traps
 Using the bubble flux and [CH4] data, and by applying the Ideal Gas Equation, we calculated the mass of CH4 entering each funnel, which is equivalent to the mass flux to the water table. Based on the ~ weekly readings, and expressed as an average daily rate per unit area for each time period, fluxes to the water table ranged between ~ −1 and 573 mg CH4 m−2 d−1 (n = 414). The median flux was 4.0 mg CH4 m−2 d−1, the lower quartile 0.4 mg CH4 m−2 d−1, and the upper quartile 14.6 mg CH4 m−2 d−1. All 28 funnel traps captured CH4 ebullition in at least 1 week during the season. However, most were dormant (i.e., showed no evidence of ebullition) for at least 1 week, and several were dormant for longer periods of up to 4 weeks.
 There was considerable spatial variation too in the mass of CH4 reaching each funnel, as shown in Figure 2. Cumulative seasonal CH4 ebullition fluxes for all 28 traps ranged from 0.8 to 360.7 mg per funnel (16.0–7515.2 mg m−2), with a median of 49.1 mg (1022.9 mg m−2), lower quartile of 12.1 mg (252.9 mg m−2), and upper quartile of 118.2 mg (2463.2 mg m−2). There is a strong positive skew in the spatial data, with nine funnel traps accounting for ~76% of the summed seasonal flux from the total of 28, and two traps (numbers 15 [S + SL] and 22 [MBH]) accounting for ~30% of the total seasonal flux. Over the monitoring season, MBH traps between them (n = 14) captured a total of 1124.8 mg (1673.8 mg m−2), and S + SL a total of 1016.6 mg (1512.8 mg m−2). There was no significant difference between the two groups [p(2) = 0.87, t test; p(2) = 0.80, Mann-Whitney U test].
3.4 CH4 Flux to the Atmosphere
 After application of the oxidation potentials published by Hornibrook et al. , only 34 of the 208 ~ weekly CH4 fluxes measured for the 14 funnels in the S + SL exceeded the potential rates of CH4 oxidation. For the MBH funnels, the equivalent figure was 81 out of 206. The non-zero fluxes are shown in Figure 3, from which it can be seen that there is little obvious pattern in the data, except a period at the beginning of the study period when there were very few fluxes to the atmosphere.
 Cumulative seasonal CH4 fluxes to the atmosphere were calculated for each of the 28 funnels. From the 14 MBH funnels, the total seasonal CH4 flux to the atmosphere was 838.0 mg, equivalent to 1247.0 mg CH4 m−2 or 11.7 mg CH4 m−2 d−1 (~25% less than the total seasonal ebullition flux of CH4 to the water table). From the 14S + SL funnels, the total seasonal ebullition flux of CH4 was 486.1 or 723.4 mg CH4 m−2 or 6.8 mg CH4 d−1 (~52% less than the total seasonal ebullition flux to the water table). As with fluxes to the water table, and despite the smaller oxic zone in the MBH, these differences were not significant [p(2) = 0.55, t test; p(2) = 0.87, Mann-Whitney U test].
4 Analysis and Conclusions
 Ebullition appears to be an important component of total fluxes at the site. Mean summer rates were 11.7 mg CH4 m−2 d−1 from the MBH and 6.8 mg CH4 m−2 d−1 from the S + SL. Unpublished chamber flux data from the site from the summer of 2008 averaged 75.3 mg CH4 m−2 d−1 (MBH) and 49.2 mg CH4 m−2 d−1 (S + SL). These averages are based on five chambers, each with an area of 0.11 m2. Chambers record an undefined mixture of fluxes including steady ebullition but usually excluding episodic ebullition, whereas the funnels measure steady and episodic ebullition. Bearing this in mind, and notwithstanding the different year and small number of flux chambers, our data suggest that ebullition is an important component of total peatland CH4 emissions (~14%–16%) at Cors Fochno.
 As well as suggesting some areas of the peatland are hot spots for emission, while others are “cold spots,” our data show that ebullition is temporally very variable, thus highlighting the need for continuous monitoring. A similar conclusion was reached by Comas and Wright , although our much larger sample size suggests their finding is not a sampling artefact.
 Perhaps, the most novel part of our study was the level of spatial replication. In previous studies, n ≤ 5 is typical. With n = 5, the probability of getting a mean half of that obtained using all 14 funnels is 22.6% (MBH) and 20.9% (S + SL), while a value double that obtained from all 14 funnels has a probability of 4.5% (MBH) and 9.2% (S + SL). These values were obtained by calculating the mean flux for every combination (n = 2002) of five funnels drawn from the 14 that were used and suggest that sample sizes of n ≤ 5 may lead to considerable error in ebullition estimates. Nevertheless, bootstrapping analysis (999 bootstrap replicates) of our data suggest sample sizes should be bigger still than n = 14 because the 95% confidence interval for seasonal fluxes of CH4 to the atmosphere are 151.1–2125.8 mg m−2 (MBH) and −27.6 to 1217.8 mg m−2 (S + SL). We recommend that future studies use much larger sample sizes than has been standard practice; meta-analysis of existing data may also be helpful in further constraining ebullition estimates.
 The work reported herein was done while IS was in receipt of a studentship funded by Queen Mary University of London. The Countryside Council for Wales (now part of Natural Resources Wales) and their Senior Reserves Manager, Mike Bailey, are thanked for allowing access to the research site and for logistical help. Dr. Graeme Swindles from the University of Leeds is thanked for help with the sample size and bootstrapping analysis.
 The Editor thanks Christian Blodau and an anonymous reviewer for their assistance in evaluating this paper.