Atmospheric pressure drives changes in the vertical distribution of biogenic free-phase gas in a northern peatland

Authors


Abstract

[1] Atmospheric pressure (Patm) is known to regulate methane emissions from northern peatlands. However, recent conceptual models differ in how gas production and release occurs in shallow (defined here as less than 1 m depth) versus intermediate to deep peat soils (i.e., more than 1 m depth). We used ground penetrating radar (GPR) measurements to non-invasively estimate the vertical distribution of free-phase gas and the dependence of this distribution on atmospheric pressure in a northern peatland. Variations in the travel time of the electromagnetic wave to three interfaces in the peat column were used in conjunction with deformation rod data and a petrophysical model relating electromagnetic wave velocity to free-phase gas content to model changes in the vertical distribution of free-phase gas over time. We found a negative linear relation between changes in free-phase gas content and changes in Patm for shallow peat soils and a positive linear relation for deeper soils. Our results suggest that (1) free-phase gas content confined in deep peat soils is larger and less variable to changes in Patm than gas in shallow/intermediate peat soils; (2) increases in Patm result in gas release from shallow peat soils into the atmosphere (i.e., rapid ebullition); and (3) decreases in Patm result in upward gas movement from intermediate layers to replenish shallow layers. Our results suggest that changes in Patm drive changes in the vertical distribution of free phase gas in peat soil and regulate methane ebullition from peat soils to the atmosphere. Our data shows a relationship between free phase gas and depth that may be due to changes in peat properties or increasing water pressure with depth.

1. Introduction

[2] During the last two decades there has been a growing interest in the role of biogenically produced free-phase gas (i.e., CH4 and CO2) in northern peatlands and its importance as a component of the global carbon (C) cycle. Current estimates consider northern peatlands as accountable for 5 to 10% of CH4 flux to the atmosphere while acting as a net sink for atmospheric carbon dioxide (CO2) [Charman, 2002]. Recent models suggest that peatlands have had a cooling effect on climate through the Holocene [Frolking et al., 2006], which could change in the coming decades as peatlands respond to climate change. Release of CH4 in peatlands occurs by diffusion, transport through vascular plants, or episodic ebullition events that can release large volumes of gas over short time scales (e.g., 35 g CH4/m2 in minutes or hours [Glaser et al., 2004]). Although most studies in the last few decades have mainly considered only the first two mechanisms, interest in the causes for and significance of ebullition has grown recently. Research over the past decade suggests ebullition is an underestimated mechanism for biogenic gas release to the atmosphere from northern peatlands [Baird et al., 2009]. Previous studies in northern peatlands show biogenic free-phase gas accumulations ranging from 0 to 19% of peat volume (see Rosenberry et al. [2006] for review). However, the spatial variation in free-phase gas production, storage, transport and release to the atmosphere remains very uncertain. Such uncertainties in part arise from an incomplete understanding of both (1) mechanisms controlling the spatial variability in production/accumulation, and (2) temporal releases of biogenic gas from the peat surface to the atmosphere. In fact, how spatial distribution in free-phase gas (both in terms of production and storage) regulates ebullition fluxes from the surface is poorly understood. Such uncertainty in part exists due to a lack of suitable methods for quantifying the spatiotemporal variability in free-phase gas in peatlands in situ.

[3] Given the importance of atmospheric pressure for soil gas exchange [Kimball and Lemon, 1971], it is anticipated that atmospheric pressure will be a major controlling variable on episodic ebullition of free-phase gas in peat soils. Several authors have noted increased ebullition during periods of low hydrostatic pressure that has been explained by bubble volume expansion beyond some critical value and increased buoyancy [Chanton et al., 1989; Shurpali et al., 1993; Fechner-Levy and Hemond, 1996]. Similar to fracturing of confined layers in petroleum and gas reservoirs [Roberts and Nunn, 1995], researchers have described zones of overpressure in peat soils attributed to biogenic gas accumulation beneath hydraulically confining layers that episodically rupture during depressuring events, resulting in large ebullition events [Romanowicz et al., 1995; Rosenberry et al., 2003]. Glaser et al. [2004] hypothesized a breach threshold for these confining layers as roughly equivalent to 0.6 times hydrostatic pressure in shallow peat deposits, and attributed depressuring events to sharp declines in atmospheric pressure and water table in northern peatlands of Minnesota. Similar episodes of low atmospheric pressure have been also proposed as a trigger of large ebullition events in peat soils elsewhere [Moore and Roulet, 1993; Kellner et al., 2005; Tokida et al., 2005]. On the other hand periods of increased atmospheric pressure have been also proposed as triggers for ebullition events, resulting from decreases in bubble size and subsequent enhanced mobility through smaller pore spaces [Chanton and Martens, 1988; Beckwith and Baird, 2001; Rosenberry et al., 2006; Waddington et al., 2009].

[4] In a recent review on methane dynamics in peat soils, Coulthard et al. [2009] stressed the need for further studies on biogenic free-phase gas dynamics in order to better understand how production and accumulation may vary along the peat column and how this regulates releases from the peat surface to the atmosphere. A model proposed by Glaser et al. [2004] describes production and release as enhanced at depth, supported by anaerobic conditions and transport of labile carbon to depth. This model assumes free-phase gas accumulations to be concentrated in overpressured pockets below confining layers (2–3 m deep) that may breach during rapid drops in atmospheric pressure, releasing large volumes of biogenic gas. In contrast, Coulthard et al. [2009] present a model that argues for more production in the near surface (i.e., upper 1 m of peat) where labile carbon is most available, whereby air bubbles entrapped below the water table act as nuclei for biogenic bubble growth. Although both models are partly supported by experimental data, whether biogenic free-phase gas in peat soils accumulates more at deep or shallow depths along the peat column and how atmospheric pressure may regulate vertical transport and emissions of biogenic gas along the peat column still remains unclear.

[5] In this paper we examine how atmospheric pressure regulates the vertical distribution of free-phase gas in a northern peatland in central Maine over a 6 week period containing changes in atmospheric pressure up to 16 mbar. Our main objective is to determine gas content variability within the peat column and how atmospheric pressure influences gas content in the deep versus shallow peat. Correlations between free-phase gas concentration in shallow (<1 m) and deep (>2 m) peat are investigated in order to better understand how atmospheric pressure regulates changes in the vertical distribution of free-phase gas, as well as ultimate release to the atmosphere.

2. Methodology and Experimental Field Design

[6] Ground penetrating radar (GPR) is a geophysical method capable of estimating in situ biogenic gases in peat soils at several scales of measurement (see Comas and Slater [2009] for review). The technique has been used in peatlands to investigate (1) vertical distribution of biogenic gases from GPR borehole measurements [Comas et al., 2005], (2) biogenic gas dynamics at the field [Comas et al., 2007] and laboratory [Comas and Slater, 2007] scales using surface GPR, (3) seasonal changes in biogenic gas dynamics and the effect of a seasonal ice layer on gas build-up [Comas et al., 2008], and (4) changes in overall gas content [Strack and Mierau, 2010] or vertical distribution of gas along the peat column [Parsekian et al., 2010] in relation to peat landform type. The GPR method can be applied non-invasively from the surface to determine the distribution of biogenic gas within the peat profile without disturbance of the peat fabric. Furthermore, it provides free-phase gas content estimates at support volumes much larger than point measurements (e.g., moisture probes) and therefore likely more representative of gas concentrations relevant to understanding field-scale fluxes.

[7] Time-lapse surface GPR measurements in the common offset (CO) acquisition mode (Figure 1a) were conducted over a 2.4 m × 1 m area at Caribou Bog, a multiunit peatland located near Bangor, Maine (4973597 m N; 521495 m E). The site was located close to the easternmost basin edge of the central unit of Caribou Bog, in a shrub-dominated area containing multiple shrubby-lichen hummocks [Comas et al., 2008]. Water level was recorded from a well cluster located less than 100 m south [Comas et al., 2008] and averaged 0.47 m depth from the surface. A platform anchored in mineral soil at depths between 3.1 and 3.4 m was constructed to provide a fixed frame of reference during GPR measurements and to avoid direct contact and therefore disturbance of the peat surface while collecting data. All GPR measurements were conducted with a Mala-RAMAC system equipped with 100 MHz antennas. A total of three 2.4 m CO profiles parallel to each other and spaced 0.5 m apart were collected using a 0.1 m trace spacing. Reflectors in the CO profile were selected to investigate time-lapse changes in travel time associated with biogenic gas build-up and release as a function of depth. As shown in Figure 1a, three reflectors in the CO profile were chosen based on distinctive interfaces detected in a peat core as follows: (a) reflector 1 at approximately 65 ns corresponding to a sharp contrast in humification (transition from H4 to H6 on von Post humification scale); (b) reflector 2 at approximately 100 ns corresponding to a highly humified layer containing large fragments of wood; and (c) reflector 3 at approximately 170 ns corresponding to the peat-mineral soil interface. Figure 1b exemplifies the repeatability of these reflectors for two different dates (August 3 and August 21). Electromagnetic travel times for each surveying day (i.e., t1 and t2 as shown in the inset of Figure 1b) were picked from these reflectors at the peak of the first sidelobe of the reflected wavelet and the average velocity for the entire peat column above each reflector was then determined using the depth to the reflector corrected for (1) raypath geometry as a result of the common offset between antennas; and (2) peat matrix deformation and surface expansion as explained later. Differences in travel time up to 4.2 ns between surveys were recorded for particular layers. For a low-loss (i.e., conduction currents orders of magnitude smaller than displacement currents) non-magnetic medium, electromagnetic wave velocity (v) can be defined as

equation image

where co is the electromagnetic wave velocity in a vacuum (3 × 108 m s−1), and ɛr(b) is the bulk relative dielectric permittivity of the soil.

Figure 1.

(a) GPR common-offset profiles along study site platform in the Central Unit of Caribou Bog. Coring results showing differences in peat type (von Post humification scale, H), deformation rod location with depth, mineral soil interface, and interpreted wood layers (white dot, as per wood debris recovery) are also shown. Reflector location used for layer analysis is also shown in red; and (b) GPR 2.4 m common-offset profiles used in our analysis for August 3 and August 21. Inset shows the variability of picked arrival times for the same reflector.

[8] The Dix equation [Dix, 1955] was then used to estimate interval velocity (vi),:

equation image

where t is the two-way travel time and n refers to the reflector number in the sequence of reflectors shown in Figure 1. An interval velocity was calculated for layers 0–1 m (shallow peat), 1–2 m (intermediate peat) and 2–3.4 m (deep peat) in Figure 1.

[9] Given the elastic nature of the peat matrix, changes in the elevation of the two reflectors above the mineral soil due to peat deformation must be considered. These reflectors were anticipated to show different amounts of deformation, with the humification interval (reflector 1) expected to vary more than the woody layer (reflector 2). Following previous studies [Price, 2003] deformation rods were used to monitor vertical movement as a function of peat depth. Deformation rods were (1) used to measure differences in elevation between the peat surface and a fixed datum; and (2) anchored at 0.5 m, 1.5 m and 2.5 m depth (as shown in Figure 1a) using T-sections twisted in place. The vertical movement of these rods relative to a fixed datum was established with a precision laser level fixed to the platform and recorded for a total of 9 rods equally spaced across the surface of the peat and three rods at 0.5 m, 1.5 m, and 2.5 m depth. Values were used to estimate changes in peat surface and reflector depth over time due to matrix deformation. Estimated maximum measurement error in all cases was <0.0035 m (mean value). We acknowledge the limitation of this approach by assuming that matrix deformation occurs entirely as vertical movement and therefore lateral displacement is not considered. Gas content was estimated from electromagnetic wave velocities using the Complex Refractive Index Model (CRIM) [e.g., Huisman et al., 2003], which is a volumetric three phase-mixing model for the soil [Wharton et al., 1980],

equation image

where ɛr(a), ɛr(w), and ɛr(s) are the relative dielectric permittivity of gas (= 1), water (temperature dependant) and soil particles (= 2; after Comas et al. [2005]) respectively, n is the porosity, θ is the volumetric soil water content and α is a factor accounting for the orientation of the electrical field and the geometrical arrangement of minerals (typically 0.35 for peat soils [Kellner et al., 2005; Parsekian et al., 2010]). Gas content estimated using the CRIM accounted for (1) variation in ɛr(w) between 84 and 86 as a result of the temperature variation between 4.5 and 7.3°C with depth [Comas et al., 2008]; (2) changes in water table elevation for the shallow interval up to 0.024 m between July–August as measured in the study by Comas et al. [2008]; (3) changes in depth to the reflector (i.e., layer thickness) due to vertical matrix deformation as recorded with the elevation rods; and (4) the effect of the unsaturated part or acrotelm region in the shallow interval between the surface and water table, by assuming an average water content of 25% above the water table [e.g., Hayward and Clymo, 1982].

[10] Error in the gas content changes estimated from the CRIM model was based on propagation of measurement errors [Bevington, 1969] for electromagnetic wave velocity, and porosity, ɛr(w) (all up to 0.5%), and ɛr(s) (up to 2%) as per repeatability of measurements accounting for a total of <±0.8%. However, this error does not account for departures from assumptions of our model such as porosity and α values. For instance, our measurements include a correction factor for porosity variation due to matrix deformation but uncertainty in the initial porosity values of our model, assumed to range between 92 and 93% along the peat column, still exists. However, considering the reported range of porosity from 91 to 94% in Caribou Bog [Comas et al., 2005], our total error in free-phase gas content is still <±1.0%. Errors associated with uncertainty in α are potentially larger. Variations in α between 0.35 and 0.5, being the range of values previously assumed for peat soils, could result in up to 2% variations in free-phase gas content. Changes in physical properties of the peat matrix will likely result in changes in the dielectric permittivity of the organic soil that is assumed constant in our model. Given these uncertainties, the relative changes in GPR estimated gas content are likely better estimated than the absolute gas content at the scale of the GPR measurement.

[11] Gas fluxes were concurrently estimated using a static chamber set at the ground surface and attached to a portable combustible gas detector (VRae), factory-calibrated for methane with a resolution of 500 ppm of CH4. Although the instrument does not provide accurate estimates of flux, it does provide a semiquantitative estimate of CH4 flux variations over the duration of the experiment. Given that detection of short-term ebullition events with gas chambers is challenging [Baird et al., 2004], we assume these flux chambers preferentially capture the background flux emitted from plants mostly derived from the uppermost peat (20–30 cm). Gas concentration recorded as a function of time was used to estimate gas flux by applying the ideal gas law (further details of this approach can be found in the study by Comas et al. [2008]. Atmospheric pressure (Patm), water table elevation, and temperature were measured in the field using Solinst LTC Levelogger (Model 3001) data loggers as previously shown in the study by Comas et al. [2007]. A full assessment of the impact of antecedent conditions for this or other environmental variables reported here (such as atmospheric pressure, temperature or water table elevation) is beyond the scope of our study and therefore not included here. Furthermore we recognize that our discrete GPR measurements likely spatially alias the true variations in gas content, and that our restricted sampling interval may result in loss of information on the relation of gas content to environmental variables due to the low frequency filtering of the system dynamics.

3. Results

[12] Figure 2a shows changes in layer deformation (relative to 21-July) with positive and negative values indicating matrix expansion and compaction respectively. Vertical movement of elevation rods (relative to 21-July) indicates total deformation of 0.05 m for the entire peat column by assuming changes to be manifest entirely as vertical movement [Price, 2003]. Most (0.04 m) of this deformation occurred in the shallow peat layer over this time period (Figure 2a), whereas little (0.01 m) deformation was measured in the intermediate and deep peat layers. Deformation changes in the shallow peat layer are larger than the changes that can be attributed to the reduction in volume of the gas bubbles and associated matrix deformation (up to 0.015 m for the largest Patm reduction) with increasing pressure as predicted from the ideal gas law.

Figure 2.

(a) Change in layer deformation (relative to 21-July) from averaged deformation rod location; (b) CRIM estimated volumetric gas content (%) for each layer after Dix equation [Dix, 1955], and Patm (note reversed scale) as a function of time. Solid gray line shows continuous Patm.

[13] Interval velocities for all data within the time frame of this study ranged between 0.0352 and 0.0397 m/ns, resulting in GPR estimated gas contents between 0.6 and 9.5%. Figure 2b shows free-phase gas content estimated for each layer and Patm as a function of time (note the reversed Patm scale for clarity). Continuous data on Patm (shown as a solid gray line in Figure 2b) is also included although only Patm values matching discrete GPR data points (solid dots) are used in our analysis. The time series suggests that free-phase gas content in the 0–1 m (shallow) peat increases with decreases in Patm. For example, decreases in Patm (between 24-July and 3-Aug, and between 18-Aug and 21-Aug) coincide with periods of increased gas content (3.7% and 1.8% respectively), while a 16 mbar increase in Patm (between 3-Aug and 18-Aug) coincides with a 3.0% decrease in this layer. Changes in gas content and changes in atmospheric pressure (ΔPatm) relative to each previous measurement were computed to investigate the relationship between the two variables. Linear regression analysis (Figure 3a) reveals a statistically significant negative linear relationship between changes in gas content and ΔPatm for the shallow peat layer (R2 and P values of 0.70 and 0.005 respectively).

Figure 3.

Least squares regression of % change in gas content versus change in ΔPatm (mbar) for (a) shallow peat (0–1 m) layer; (b) intermediate (1–2 m) depth layer; and (c) deep peat (2–3.4 m) depth layer. R2 and P-values for each data set are also shown.

[14] In contrast, the time series for the 1–2 m layer (intermediate peat) suggests a direct relation between free-phase gas content and Patm. For example, 8 mbar and 11 mbar decreases in Patm (between 21-July to 3-Aug, and between 18-Aug to 21-Aug respectively) now coincide with decreases in free-phase gas of 2.1% and 0.8% respectively, while the 16 mbar increase in Patm (between 3-Aug to 18-Aug) coincides with a 1.8% increase in free-phase gas. Regression analysis (Figure 3b) confirms a statistically significant direct relationship between changes in gas content and ΔPatm for layer 1–2 m (R2 and P values of 0.5 and 0.032 respectively). The deepest 2–3.4 m layer (deep peat) shows almost no variation in changes in gas content with ΔPatm (Figure 2b). Regression analysis shows no significant relationship between changes in gas content and ΔPatm for the lower layer (R2 and P values of 0.29 and 0.12 respectively).

[15] Changes in surface deformation (relative to 21-July) show a statistically significant direct relationship with average gas content of the entire peat column as confirmed by the linear regression analysis shown in Figure 4a resulting in R2 and P values of 0.55 and 0.05 respectively. A similar relationship is also shown between the shallow layer and overall changes in surface deformation resulting in R2 and P values of 0.62 and 0.04 respectively (Figure 4a). Despite the resolution limitations of the gas detector used and the fact that our flux data is mostly used for qualitative purposes only (i.e., periods of high versus low flux), methane gas fluxes ranged between 3.9 to 8.0 g CH4 m−2 day−1 and compare well with other reported values for ebullition fluxes in northern peatlands [i.e., Rosenberry et al., 2006]. Furthermore, there is a statistically significant linear negative relationship between gas flux and Patm (Figure 4b), suggesting an increase of 0.3 g m−2 day−1 in methane gas flux per mbar of Patm decrease.

Figure 4.

Least squares regression of (a) change in surface deformation versus % gas content along the entire peat column and for the shallow (0–1 m) depth; and (b) gas flux (g m−2 day−1) versus ΔPatm (mbar). R2 and P-values for each data set are also shown.

4. Discussion

[16] Our results identify free-phase gas concentrations ranging between 8.0 ± 0.8% to 9.1 ± 0.8% for the deep layer that appear independent of ΔPatm. In contrast shallow, and intermediate layers show free-phase gas concentrations between 6.4 to 12.1 ± 0.8% and 1.2 to 4.8 ± 0.8% respectively, with a greater range of variability than in the deep layer. Furthermore changes in gas content appear related to ΔPatm in these layers. Whereas a direct linear correlation between changes in gas content and ΔPatm exists in the shallow layer, a negative linear relation exists for the intermediate layer.

[17] We attribute the variations in the changes in gas content-ΔPatm relation among all layers to two distinct processes. The direct linear correlation between changes in gas content and ΔPatm for the upper layer is consistent with a reduction in gas bubble size with increasing pressure and resulting higher bubble mobility through the pore spaces [Chanton and Martens, 1988; Beckwith and Baird, 2001; Rosenberry et al., 2006]. On the other hand bubble volume in deeper peat may increase in response to a decrease in Patm and result in increased buoyancy [Moore and Roulet, 1993; Fechner-Levy and Hemond, 1996]. Differences in bubble buoyancy between deep and shallow peat may be explained by changes in peat structure as indicated by the degree of decomposition as discussed below (i.e., sharp interface in von Post scale at 1 m deep between the shallow and intermediate peat layer in Figure 1a). For instance, peat decomposition results in decreased fiber content and a tighter packing of organic particles [Price et al., 2005], resulting in reduced spacing for bubble accumulation and/or coalescence. Little decomposition translates into increased fiber content and increased spacing for bubble accumulation/coalescence. Our interpretation complements other surface-based studies on ebullition in peatlands [Strack et al., 2006] by suggesting that vertical changes in matrix structure may control ebullition in addition to spatial changes across the surface of different peatland forms. The idea of bubble distribution and dynamics being controlled by the structural properties of the peat has been also suggested and modeled by other authors [Kettridge and Binley, 2008; Coulthard et al., 2009]. Furthermore our results are also consistent with the presence of confining layers exerting a control on ebullition dynamics.

[18] Despite the dependence on structural properties, the reason for the relative strengths of the two proposed mechanisms (mobility due to bubble size reduction versus buoyancy due to bubble expansion) being different as a function of depth is not clear. We hypothesize that gas loss due to bubble size reduction and increased mobility dominates the shallow peat due to a less decomposed fiber-rich peat that allows for higher gas volume accumulation. Larger free-phase gas volumes could encourage bubble coalescence, resulting in later collapse of the bubble cluster and subsequent free-phase gas release once a certain threshold is reached [Kellner et al., 2006]. Coulthard et al. [2009] describe a model for upward free-phase gas movement under buoyancy, whereby a single bubble may result in a cascade of collapses and a large ebullition event. Such dynamics are less likely for the intermediate layer, where increased peat decomposition limits availability of space for bubble occupancy and clustering (as reflected in much lower free-phase gas volumes). For that reason, increased Patm for the intermediate layer could result in some individual bubble mobility but without inducing much coalescence and/or collapse. We attribute the 3.0% free-phase gas decrease in the shallow layer and 1.8% increase in the intermediate layer peat between 3-Aug and 18-Aug (Figure 2b) to this mechanism.

[19] Although decreases in Patm may result in bubble expansion and enhanced buoyancy along the peat column, we hypothesize that the upward movement of a larger (i.e., expanded) bubble may be impeded in the shallow peat due to the intricate network of fibers in the matrix of a relatively undecomposed peat. Such bubble expansion will still result in an overall free-phase gas volume increase in the shallow layer. This upward movement then seems more feasible in the intermediate layer, where bubble expansion due to decreased Patm could more easily displace individual (more decomposed) peat particles raising bubbles toward the upper layer and adding to the already increased free-phase gas volume in the shallow peat layer. We attribute the large 3.7% free-phase gas increase in the shallow layer and the 2.1% free-phase gas decrease in the intermediate layer between 24-July and 3-Aug to this mechanism.

[20] The lack of correlation between the deep peat layer and Patm (Figure 3c) is here attributed to the combination of an overlaying confining layer and the fact that increased lithostatic pressure in deep peat minimizes the overall effect that changes in Patm may induce on pore fluid pressures. For example, the increase in Patm of 15.9 mbars recorded between 3 and 13-August is equivalent to a pressure increase of 16.2 cm of water. Since we can assume that hydrostatic pressure due to the water column at 0.25 m and 3 m depth is 25 and 300 cm of water respectively, a pressure increase of 16.2 cm of water results in a 65% pressure increase in the shallow peat while it only increases 5% in the deep peat. This effect results in much smaller changes in bubble size for the deep peat. Furthermore, given the presence of a confining layer at 2 m depth and according to previous studies in peat soils [i.e., Glaser et al., 2004] a fracture threshold of 0.6 times hydrostatic pressure (or 120 cm of water) has been proposed in order to induce a breakout. Since hydrostatic pressure at 2 m depth corresponds to 200 cm of water, the increase in 16.2 cm of water describe above will be insufficient to trigger such breakout. The lack of deformation recorded below 2 m depth (Figure 2a) is consistent with this interpretation. Although not shown in our data, we hypothesize that a larger change in pressure may be needed in order to induce layer breakout and subsequent ebullition for deep peat. Alternatively, bubble clustering and increased buoyancy in deep peat may result in confining forces from the layer to be overcome, inducing a breach in the layer and subsequent free-phase gas upward release.

[21] Changes in gas content along the peat profile have been described here in relation to bubble dynamics resulting from a combination of upward movement from either mobility increase (due to bubble volume decrease) or buoyancy increase (due to bubble volume increase), bubble coalescence and entrapment, clustering collapse and a breach threshold. In order to illustrate this approach, two contrasting events are chosen: (1) the period of Patm decrease from 24-July to 3-Aug; and (2) the period of Patm increase from 3-Aug to 18-Aug. By applying the ideal gas law and assuming a methane fraction for the gas phase up to 52% (A. Parsekian, unpublished data) and temperature variations between 7.3 and 4.5°C with depth, gas content increases or decrease per event and for each layer are expressed as an equivalent methane flux (Table 1, column labeled ‘Total’). Fluxes were corrected for variability in methane solubility to account for changes in gas volume due to temperature fluctuations with depth by following the approach in the study by Kellner et al. [2006] and applying Henry's law and the ideal gas law. Variability of Henry's constant with temperature was considered following Lekvam and Bishnoi [1997], reaching final gas content corrections of 0.27% and 0.09% for the intermediate and deep layer respectively. The effect of Patm on methane solubility was also estimated following the approach in the study by Kellner et al. [2006] obtaining maximum corrections of 0.06% in gas content for the range of Patm change in this study. Average layer thicknesses in each case are 0.9 m, 1.2 m and 1.5 m for shallow, intermediate and deep peat layers, respectively. Methane fluxes are expressed as a proposed carbon budget in Table 1 by assuming that (1) diffusion [Blodau, 2002] and consumption rates [Yavitt et al., 1990] are only significant in the shallow layer; and (2) average steady ebullition rates [Coulthard et al., 2009] in the shallow layer are zero (i.e., inputs from intermediate layer and outputs to the atmosphere cancel each other) and did not occur below the woody layer. Production rates in the intermediate layer for event 1 are extracted from estimates in the same location as per Comas et al. [2008] and used as final input to calculate production rates in the shallow layer for event 1. Estimated production rates for the shallow layer in event 1 are used as input for event 2 in order to estimate episodic ebullition in the shallow layer (Table 1). A Patm decrease of 8.5 mbars during event 1 coincides with a total flux of -1.95 g CH4 m−2 day−1 for the intermediate layer (negative sign indicates gas released). During the same period a total flux of 3.16 g CH4 m−2 day−1 entered the shallow layer (Table 1). This observation could be explained by upward migration of 2.78 g CH4 m−2 day−1 as episodic ebullition from the intermediate to the shallow layer (Table 1), resulting in a total of 0.15 g CH4 m−2 day−1 released into the atmosphere. Although such release does not match our semiquantitative estimates as measured from the gas meter (as anticipated due to the high degree of spatial and temporal variability), it does correspond well with common diffusive fluxes in northern peatlands ranging between 0.005 and 0.2 g CH4 m−2 day−1 as measured with chambers and eddy covariance methods [Blodau, 2002].

Table 1. Summary of Results for Methane Flux (in g CH4 m−2 day−1) per Layer and Total Budget During Two Contrasting Events Assuming 52% Methane Fraction for the Gas Phasea
EventAtm P Change (mbar)LayerDiffusionbEbullitionConsumptiondProductionTotale
SteadycEpisodic
1 (24 July to 3 Aug)−8.5atmosphere0.080.07   0.15
  shallow−0.080.07–0.072.78−0.090.553.16
  intermediate −0.07−2.78 0.90f−1.95
  deep    0.460.46
  total000−0.091.011.82
         
2 (3–18 Aug)15.9atmosphere0.080.072.07  2.22
  shallow−0.080.07–0.07−2.07−0.090.55−1.69
  intermediate −0.07  1.691.62
  deep    0.650.65
  total000−0.092.892.8

[22] During event 2, a Patm increase of 15.9 mbars coincides with a total flux of −1.69 g CH4 m−2 day−1 for the shallow peat layer (Table 1), likely due to upwards migration and rapid release (i.e., ebullition) of 2.07 g CH4 m−2 day−1 (Table 1) to the atmosphere. A total average increase of 1.62 g CH4 m−2 day−1 in the intermediate peat layer is also recorded during event 2. Since downward gas migration from the shallow layer seems physically implausible and upward bubble movement from the deep peat layer was not detected, we assume that this increase reflect a 1.69 g CH4 m−2 day−1 gas production within the layer that is larger than that estimated for event 1. Although a plausible explanation for this increase goes beyond the scope of our data set, such production rate (averaging 1.0 g CH4 m−2 day−1 for the entire peat column) is still consistent with previous estimates for methane production at this site [Comas et al., 2008], where average production rates of 0.9 g CH4 m−2 day−1 for the entire peat column were estimated, with other nearby areas reaching 2.0 g CH4 m−2 day−1. These values are also consistent with ranges for other northern peatlands [Siegel et al., 2001] and wetland soils [Segers, 1998]. The deep layer showed little variation during the two events resulting in consistent increases between 0.46 and 0.65 g CH4 m−2 day−1 as related to production rates (Table 1) lower than the intermediate and shallow layers.

[23] Enhanced production rates and high gas variability for shallow peat soils (as compared to large stable gas accumulations below confining layers and lower production rates in deep peat) match well with current models of methane accumulation and dynamics in northern peatlands [Glaser et al., 2004; Coulthard et al., 2009]. Furthermore, our findings seem consistent with conceptual models related to bog breathing or Mooratmung [Ingram, 1983; Glaser et al., 2004]. Glaser et al. [2004] found a direct relationship between peat volume and Patm, and although they did not offer an explanation for this response, they proposed the existence of a cyclic pattern as related to the breathing and release of free-phase gas to the atmosphere as driven by changes in Patm. In this regard we can hypothesize that although diffusion or steady ebullition fluxes may be larger during low pressure events (as reflected in Figure 4b) free-phase gas is redistributed rapidly as episodic ebullition within the peat column due to buoyancy effect, while high pressure events result in free-phase gas “breathed” out at the surface as a result of rapid increases in gas mobility in near surface soils.

5. Conclusions

[24] Our results show that (1) overall gas contents are larger and less variable for deep peat soils as compared to variable shallow and intermediate soils; (2) decreases in Patm result in upward gas movement from intermediate layers into shallow layers; and (3) increases in Patm result in gas release from shallow peat soils into the atmosphere. Our results also suggest that vertical distribution and dynamics of biogenic free phase gas in peat soils are dependent on Patm as possibly related to changes in the internal nature of the peat matrix. These findings have implications for better understanding the spatial and temporal distribution, production and release of biogenic gases in peat soils.

Acknowledgments

[25] This material is based upon work supported by the National Science Foundation under grants 0609534 and 1045084. Graduate students Jay Nolan and Mike O'Brien from Rutgers University, and Zach Tyczka and Joshua Rhodes from University of Maine provided valuable field support. We thank Editor Baldocchi, one Associate Editor, Paul Glaser, Mike Waddington and one anonymous reviewer for their suggestions to enhance the quality of an earlier version of this paper.