The effect of acid rain SO42− deposition on peatland CH4 emissions was examined by manipulating SO42− inputs to a pristine raised peat bog in northern Scotland. Weekly pulses of dissolved Na2SO4 were applied to the bog over two years in doses of 25, 50, and 100 kg S ha−1 yr−1, reflecting the range of pollutant S deposition loads experienced in acid rain-impacted regions of the world. CH4 fluxes were measured at regular intervals using a static chamber/gas chromatographic flame ionization detector method. Total emissions of CH4 were reduced by between 21 and 42% relative to controls, although no significant differences were observed between treatments. Estimated total annual fluxes during the second year of the experiment were 16.6 g m−2 from the controls and (in order of increasing SO42− dose size) 10.7, 13.2, and 9.8 g m−2 from the three SO42− treatments, respectively. The relative extent of CH4 flux suppression varied with changes in both peat temperature and peat water table with the largest suppression during cool periods and episodes of falling water table. Our findings suggest that low doses of SO42− at deposition rates commonly experienced in areas impacted by acid rain, may significantly affect CH4 emissions from wetlands in affected areas. We propose that SO42− from acid rain can stimulate sulfate-reducing bacteria into a population capable of outcompeting methanogens for substrates. We further propose that this microbially mediated interaction may have a significant current and future effect on the contribution of northern peatlands to the global methane budget.
 Methane (CH4), on a molecule for molecule basis, is some 21 times more powerful than CO2 as a greenhouse gas [Intergovernmental Panel on Climate Change (IPCC), 1996], making it responsible for an estimated 22% of the present greenhouse effect [Lelieveld et al., 1998]. Although the concentration of atmospheric CH4 has been increasing since the onset of the industrial revolution, there has, in recent years, been a slowdown in this growth rate [Dlugokencky et al., 1998; Dlugokencky et al., 2001]. This implies either an increase in a CH4 sink or a decrease in a CH4 source, although reasons for this declining trend remain elusive. In this paper we examine the hypothesis that SO42− from acid rain may have contributed to this slowdown by suppressing the contribution of CH4 from wetlands.
 Microbial decomposition in waterlogged soils, as found in natural wetlands and rice paddies, is the largest source of CH4 to the atmosphere [Matthews and Fung, 1987; Aselmann and Crutzen, 1989; Matthews et al., 1991]. In such systems, O2 is rapidly removed by aerobic microorganisms. This is followed by microbial reduction of a suite of oxidized inorganic compounds, such as (in order of decreasing energy yield for microbes) NO3−, Mn (IV) and Mn (III), Fe (III), and SO42− [Van Breemen and Feitjel, 1990]. Once SO42− has been consumed, the lowest energy-yielding terminal step in anaerobic microbial decomposition is the consumption of H2/CO2 and acetate to produce CH4. Consequently, since reduction of SO42− by microorganisms (SO42− reducing bacteria (SRB)) provides a more efficient means by which competitive substrates can be consumed, methane-producing archaea (MP) are placed at a competitive disadvantage [Abram and Nedwell, 1978; Schonheit et al., 1982; Kristjansson et al., 1982]. Suppression of methanogenesis by stimulation of SO42−-reducing populations explains the observation that salt marshes and wetlands overlying SO42− rich deposits emit considerably less CH4 than otherwise comparable freshwater wetlands [Bartlett et al., 1987; Rejmankova and Post, 1996]. Experimental work with large (103 kg SO42−-S) fertilization doses of SO42− on rice paddies [Dernier van der Gon and Neue, 1994; Lindau et al., 1994, 1998] has also shown a clear suppression of CH4 emission. While in SO42−-rich marine sediments, CH4 release is completely inhibited [Martens and Berner, 1977], the interaction has also been documented in environments with SO42− concentrations at freshwater levels [Lovely and Klug, 1983]. This suggests the hypothesis that acid rain, of which a principal component is SO42−, may affect the emission of CH4 from impacted wetlands.
 In eastern Europe and Asia, there is a trend of increased SO42− deposition due to economic growth [Bhatti et al., 1992; Rodhe et al., 1995; Rodhe, 1999]. This trend of enhanced supply of a species that is known to adversely affect CH4 production in anaerobic environments therefore deserves close examination as it presents us with a potential mechanism that may explain the observed decline in the atmospheric CH4 growth rate. Few studies have, until recently, investigated this potentially important link.
 In peat core incubation experiments, Fowler et al.  showed that single doses of SO42− representing annual total deposition rates in acid rain-impacted areas (40 kg SO42− - S ha−1) reduced CH4 fluxes from peat by around 40%. They also found that following an initial 3 week period of suppression, emissions recovered to pretreatment levels, which implies that large, individual inputs of SO42− may create a “boom-bust” cycle in SRB populations as SO42− is either lost gaseously from the system or is converted to more biologically recalcitrant forms.
 In field studies of underlying processes in two peatlands with contrasting SO42− deposition regimes, Watson and Nedwell  showed that SO42− reduction is indeed an important pathway in the degradation of organic matter, suppressing CH4 production, albeit to a varying degree depending on the season. No relationship between SO42− load and CH4 production could, however, be deduced owing to the confounding presence of other factors, such as differences in the degradability of peat (variable C:N ratio) and differences in nitrogen primary production (NPP) at the two sites due to climatic differences.
Dise and Verry  alleviated the problems of intersite heterogeneity encountered by Watson and Nedwell  by manipulating SO42− deposition to a bog over a 12 week period during a single growing season. Although the S deposition rate amounted to a maximum of 145 kg SO42−-S ha−1 yr−1, at the extreme high end of SO42− deposition in both Europe and Asia, individual doses were no higher than 2.7 kg ha−1, far lower than in single dose and fertilization experiments conducted thus far and more reflective of the mode of pollutant S deposition experienced in nature. They found that CH4 emissions were reduced by 30–40%, similar to the level of suppression found in single dose experiments. This work suggests that small, continuous inputs of SO42− may have the same net suppressive effect as single large SO42− doses, as SRB communities may be maintained at competitive levels by the continuous low-level influx of SO42−. They were, however, unable to fully examine the effect that changes in temperature and water table may have on treatment effect since the measurement period was confined to the warm summer months when these variables remained relatively constant. To our knowledge, no study has examined the effect of SO42− on CH4 emission measured over a full year, nor the interactions between this effect and the temperature and hydrology of a wetland.
 The objective of this experiment is to investigate the effect of low, continuous SO42− deposition on CH4 emissions from a peatland by manipulating SO42− deposition levels over two growing seasons within a range experienced in areas of the world that are impacted by acid rain. In addition, with natural changes in temperature and water table over this period we are also able to investigate the degree to which any suppression in CH4 flux is controlled by climate.
2. Materials and Methods
2.1. Experimental Site Description
 The experiment was located on an extensive pristine portion (i.e., unaffected by cutting or drainage) of the Moidach More in Morayshire in northeast Scotland (57.46°N, 3.62°W) at an altitude of 275 m above sea level. The raised mire consists of peat of >0.5 m thick, extending over 760 ha and averaging 2.1 m in depth. The mean annual precipitation at the site is ∼900 mm [Meteorological Office, 1987], and the mean annual temperature is 8°C [Williams et al., 1999]. The vegetation mainly comprises Sphagnum species, which include S. magillanicum and S. capillifollium [Ehrh.] Hedw. and S. revurvum [P. Beauv.]. The dominant sedge is Trichophorum cespitosum [L.] Hartm. (deer grass). Other plants include Erica tetralix L. and, in areas of the bog affected through cutting and burning, Calluna vulgaris [L.]. The site was selected for its low ambient SO42−-S deposition rate of 5 kg ha−1 yr−1 (R. Smith personal communication, 2000). Rainfall data for the area were collected at Grantown on Spey, ∼10 km south of the study site (supplied by the British Atmospheric Data Centre).
2.2. Sulfate-S Applications
 Wooden boardwalks were installed within the sampling area to allow repeated access while minimizing site disturbance during sampling. Twenty experimental plots (2 × 2 m) were established on an area of the peatland that exhibited uniform characteristics in terms of its vegetation, topography, and hydrology. The plots were separated by a 1 m buffer strip and were randomly assigned to one of three treatments or a control. The treatments consisted of additions of 20, 45, and 95 kg SO42−-S per hectare per year applied as Na2SO4. In addition to annual ambient deposition, this amounted to total annual deposition rates of 25, 50, and 100 kg SO42−-S ha−1, respectively. The treatments were applied as weekly doses of between 1.2 and 4.7 mmol SO42−-S in 1 L of deionized water, which amounted to between 0.5 and 1.9 kg SO42−-S ha−1 week−1. Over the winter period (November to March), we dosed once a month rather than weekly, and doses were correspondingly four times stronger. SO42− was added as Na2SO4 (as opposed to H2SO4) to minimize vegetation damage that may have arisen through the monthly application of concentrated acid pulses. Weekly doses of Na2SO4 in 1 L of deionized water were sprayed evenly onto each 2 × 2m plot using a pressurized garden sprayer (Hozelock). The solution added amounted to a hydrologically negligible weekly increase of 0.25 mm of water to the system. Controls received the same volume of deionized water. All experimental additions began on 25 June 1997, following 5 weeks of CH4 flux monitoring at the site.
2.3. Methane Flux Measurements
 CH4 flux was measured using static chambers which were semipermanently (for the duration of the experiment) placed within each experimental plot. The chambers consisted of sections of polypropylene pipe (length of 25 cm, internal diameter of 30 cm). A groove was machine cut into the top edge of each section to accommodate a neoprene O-ring. The bottom edge of each section was beveled to facilitate installation of the chambers to a depth of 2–3 cm into the peat surface. Once in position, the chambers were not moved again for the duration of the experiment.
 The headspace volume was defined by placing a transparent acrylic lid onto the O-ring. Each lid was fitted with a silicone rubber septum (Suba Seal) which allowed repeated sampling of the headspace gas with a needle and syringe, both immediately after enclosure and then again 20 min after the initial sample was taken. Three-way stopcocks allowed samples to be stored in the syringes prior to analysis (within 24 hours) on a gas chromatographic flame ionization detector (GC FID) (Chrompack CP9000) with a 2 m long Poropack Q column connected to a Spectra Physics integrator. At bimonthly intervals the headspace of the chambers was sampled repeatedly (a minimum of three samples) over an hour-long period to test for linearity in the increase of methane concentration. Each plot was sampled weekly (late spring to early autumn) to monthly (late autumn to early spring) starting in May 1997. Samples were taken within an hour of noon on each sampling day. Chambers were sampled for CH4 flux prior to SO42− applications at all times.
2.4. Additional Measurements
 On each sampling day, peat temperature (0, 5, 10, 15, 20, 30, 40, and 50 cm below the peat surface) was measured using a thermocouple probe (ATP Technology) at three locations spanning the experimental area in the bog. Water table was also monitored using “dip wells” (50 cm lengths of 3 cm diameter polycarbonate tubing, which were positioned 0.5–1 m from each static chamber. Since vascular plants provide a major conduit of CH4 release and are in some cases the dominant means by which methane is emitted to the atmosphere [Schimel, 1995], the density of the dominant vascular plants (Trichophorum cespitosum) in the plots was calculated by periodically (once in August 1997 and monthly in 1998) counting the number of individual live shoots within each chamber.
2.5. Pore Water Chemistry
 Pore water samples were collected using “sippers” which were positioned in the peat. These were constructed from 20 mm external diameter polycarbonate tubing containing an inverted 10 mL polypropylene syringe at the base which served as a pore water reservoir. The large syringe plunger openings were sealed off, and 1 mm perforations were drilled into the side to allow lateral inflow of surrounding pore waters. The upward pointing needle end of the syringe was connected to a three-way stopcock valve above the peat surface via a 30 cm length of 1.5 mm internal diameter Teflon tubing. This minimized exposure of pore waters in the syringe reservoir to oxic conditions. Sippers were positioned at three depths in four control and four 50 kg S continuous treatment plots. Samples of pore water were taken in November 1998 and were drawn into syringes fitted with three-way stopcocks while ensuring the sample filled the whole syringe volume (no headspace) to ensure anaerobic conditions during storage for transportation back to the laboratory.
 Pore water [CH4] was measured by introducing 5 mL of each pore water sample into 40 mL boiling tubes (35 mL ambient air headspace) fitted with Suba Seal silicon rubber septa. The tubes were shaken vigorously for 2 min to strip dissolved methane into the headspace. Headspace samples were analyzed for CH4 by GC FID (see above) and were corrected for CH4 concentration in ambient air [Dise, 1993]. Remaining pore water samples were filtered with 0.45 μm membrane filters (Whatman) under vacuum and frozen prior to analysis by ion chromatography.
2.6. Calculations and Statistical Analysis
 Methane fluxes are expressed in mg CH4 m−2 d−1 by calculating the linear change in CH4 concentration over time from within a chamber of known volume enclosing a known area of peat. Temperature 10 cm below the water table (close to the zone of maximum CH4 production [Daulat and Clymo, 1998]) was calculated by interpolating temperatures measured at different depths beneath the peat surface. To evaluate the effects of water table and peat temperature on net CH4 flux from the site (all treatments pooled and averaged), we employed multiple linear regression analysis (MINITAB, release 11, Minitab Inc.). The total mass of CH4 emitted from the different treatments was estimated by integrating flux measurements over time. Treatment effects on CH4 fluxes were determined by repeated measures multivariate analysis of variance (MANOVA) (MINITAB, release 11, Minitab Inc.) with sampling time points included as a variable and sedge shoot density included as a covariate. This method permitted the evaluation of within-treatment variability (time effect and interactions between time and treatment (time times treatment)).
3.1. Methane Emissions
 Methane emissions over the course of the experiment show a distinct seasonal pattern, with peaks in emission during the warm summer months and lower emissions during cooler winter periods (Figure 1). Between 1997 and 1998, distinct interannual differences in emission are evident, with total emissions in 1997 (7.7 g CH4 m−2), 51% lower than over the same measurement period in 1998. These differences correspond to a lower water table in 1997 than in 1998 (Figure 2), which is due to lower than average rainfall in 1997 (670 mm, 26% lower than the 1916–1950 mean) and slightly lower than average rainfall in 1998 (860 mm, 4% lower than the mean). Over the April to September growing season, when most CH4 is emitted, there was 26% less rainfall in 1997 than 1998. The time of maximum emission also differed between the two years, peaking in early July in 1997 (37.4 mg CH4 m−2 d−1) and in late September/early October in 1998 (139.7 mg CH4 m−2 d−1).
 After log transforming the skewed flux data, multiple linear regression analysis of mean weekly data (all treatments combined) from 1997 showed that both water table (WT) and temperature T explained the majority of variability in log CH4 flux (Table 1); however, in 1998, only a third of the variability could be explained by these two variables. This indicates that some other variable (or variables) was driving most of the variability in 1998. Changes in water table were of a smaller magnitude in 1998 than in 1997 (Figure 2), and methane emissions during the warm summer months of 1998 exhibited spikes that were unattributable to measured variables.
Table 1. Relationship Between CH4 Flux (Pooled and Averaged for All Plots Per Sampling Period) and Water Table and Temperature T 10 cm Below the Water Tablea
Flux has units of mg m−2 d−1, water table (WT) is measured in centimeters, and temperature is in °C. Here r2 is explained variance, and n is the number of sampling dates during each period.
 As anticipated and as previously reported by Schimel , fluxes were well correlated with sedge shoot density on individual sampling dates during the pretreatment phase of the monitoring program (r = 0.68, p < 0.001, n = 19). This relationship accounted for the majority of within-treatment variation on any individual sampling date during that period. Since competing microorganisms are known to form intimate relationships with plant roots [Rooney-Varga et al., 1997; Watson and Nedwell, 1998; Kusel et al., 1999; Wind et al., 1999], any effect of treatment on subsequent (posttreatment initiation) changes in the relationship between fluxes and shoots of Trichophorum could not be discounted.
3.3. SO42−-S Deposition Experiments
 Over the course of the first year, differences between daily CH4 fluxes from control versus treatment plots were not significant (Table 2 and Figure 1). However, on adjusting fluxes to sedge density (which explained most of the variability within treatment groups on individual sampling dates), the highest dose treatment (100 kg SO4-S ha−1 yr−1) was of borderline statistical significance (n = 70; p = 0.07) (Table 2).
Table 2. Summary of Average Methane Emissions Over Three Monitoring Periodsa
Here p values indicate significance of treatment (treatment versus control) effect on mean daily flux as evaluated from repeated measures MANOVA (Wilk's lambda test) both without a covariate (indicated by A) and including Trichophorum sedge density (shoots m−2) as a covariate (indicated by B). No significant time times treatment interactions were observed. The total flux of CH4 in 1998 was calculated by integrating the mean daily flux over time.
 In 1998 both the 25 and 100 kg SO42−-S ha−1 yr−1 treatments exhibited significantly smaller fluxes than the controls over the entire year (n = 130 RM-MANOVA, p < 0.01 and p < 0.001, respectively). With mean fluxes over the year amounting to 48.8 and 45.2 mg CH4 m−2 d−1 for the 25 and 100 kg SO42−-S ha−1 yr−1, respectively, compared to 64.8 mg CH4 m−2 d−1 emitted by the control plots, this amounted to a mean annual suppression of 25 and 30% (Table 2). Mean fluxes from the 50 kg SO42−-S ha−1 yr−1 treatment were not significantly lower than controls, which was probably due, by chance, to their having the highest sedge density out of all treatments and controls (40% higher than controls in 1998). On inclusion of sedge densities as a covariate, all treatments showed highly significant suppression in fluxes relative to controls at the p = 0.001 level. This analysis, by repeated measures MANOVA, also showed that there was no time-treatment interaction nor were significant treatment-to-treatment differences observed. Over the 2 year length of the experiment, treatment plots emitted between 22 and 32% less CH4 than controls (mean net fluxes versus 24.5 g m−2, respectively) (Figure 3). However, when considering only 1998 emissions, time-integrated mean daily fluxes show a far stronger suppression of fluxes with respective suppressions of 36, 21, and 42% (in order of increasing SO42− dose rate) from the total control CH4 flux for 1998 of 16.8 g m−2.
 By taking the proportional difference in CH4 flux measured between control and treatment plots before treatments began as our best estimate of background (nontreatment) variability, we are able to estimate the relative extent to which CH4 flux from the treatment plots is lowered over the course of the experiment. This is done by calculating, for every posttreatment flux, the value of ΔCH4 (percentage change in CH4 flux), defined as
where ΔCH4 is the percentage change in methane flux as the result of a treatment effect, x1 and y1 are the control and treatment fluxes during the treatment period, respectively, and xa and yb are the mean control and treatment fluxes prior to the start of treatment applications, respectively. Therefore the more negative the value of ΔCH4, the lower the CH4 flux of treatments relative to controls.
 After an apparent initial stimulation in CH4 flux from treated plots (25 and 100 kg SO42−-S) in July and early August, all treatments show a trend of increasing difference between treatments and controls until the spring of 1998 (Figure 4). Thereafter, the relative difference between treatments and control fluxes varied over time.
 Pore waters collected (10, 20, and 30 cm below the surface) from both the controls and the 50 kg SO42−-S continuous treatment in November 1998 demonstrated that significantly less CH4 (64, 50, and 49% lower, respectively) was present as a dissolved gas in the treated plots (Figure 5a). Pooling data from all three depths, in the control plots, CH4 concentrations ranged from 8 to 112 μM, and they ranged from 1 to 72 μM in the treated plots. Differences were also apparent in SO42− concentrations with consistently (although not significantly) smaller concentrations in the treatment plots (Figure 5b). When pooling results from the three different depths, SO42− concentrations varied between 7 and 107 μM in controls and between 6 and 56 μM in the treated plots. There was a trend of increased concentration of SO42− with increasing depth in the peat.
3.4. Interaction Between Water Table, Temperature, and CH4 Flux in SO42−-S Treatment Plots
 We used multiple nonlinear regression analysis to assess how the response to temperature and water table affected the degree to which CH4 fluxes were reduced in the SO42− treated plots. Only data from 1998 were used in the analysis. The analysis showed that for all three SO42− treatments the degree of suppression on CH4 emissions is strongly linked to changes in temperature and water table. We averaged the ΔCH4 (three-time point moving average) values from the different treatments to give a broad indication of the effect of mean peat temperature and water table position (also three-time point moving averages) on the extent of CH4 flux suppression. Multiple nonlinear regression analysis yielded a highly significant relationship (r2 = 0.56, p = 0.0001, and n =24; Figure 6) between ΔCH4 in 1998 and temperature and water table for all three continuous SO42− treatments considered separately as well as for the three treatments lumped together (Table 3). This analysis implies that the suppressive effect of SO42− on CH4 flux (i.e., more negative ΔCH4) increases with both decreasing temperature and declining water table (within ranges of 0°–15°C and 0–10 cm below the peat surface).
Table 3. Relationship Between Moving Averages (Three Time Points) of ΔCH4 for Different Treatments (Derived From (1)), Water Table and Temperature 10 cm Below the Water Tablea
Treatment,Kg SO42−-S ha−1 yr−1
ΔCH4 has units of percent. Here r2 is explained variance, and n = 24 throughout.
 As detailed by Segers , quantifying the electron balance of wetlands impacted with alternate electron acceptors and then comparison with the mass of suppressed CH4 can help to elucidate whether or not SO42− is recycled in this system. Stoichiometrically, the net equation for suppression of CH4 by SO42− reduction through competitive inhibition is identical to that of anaerobic CH4 oxidation via SO42− reduction (equation (2) [Martens and Berner, 1977]).
As a result, 1 mol of applied SO42− should divert electrons away from the production of 1 mol of CH4. Any deviation from this 1:1 ratio therefore indicates either that SO42− is lost from the system before it can cause a reduction in CH4 emissions (if ratios are <1) or that it is being recycled (if ratios are >1). In comparing the number of moles of suppressed CH4 with the number of moles of applied SO42− for two time periods we found that for all three SO42− treatments, more CH4 was suppressed than the amount of applied SO42− suggests would be possible in the absence of SO42− recycling (i.e., CH4:SO42− > 1; Table 4).
Table 4. Molar Ratios of Suppressed CH4 to Applied SO42− for Two Time Periods a
Treatment kg SO42−-S ha−1 yr−1
Suppression of CH4, mol m−2
Applied SO42−, mol m−2
CH4:SO42− is expressed in moles of suppressed CH4 per mole of applied SO42−.
1997 plus 1998
1997 plus 1998
1997 plus 1998
 For plots receiving 25–100 kg SO42−-S during 1998, suppressed CH4 to SO42− ratios ranged from 4.9:1 to 1.4:1 respectively, indicating that SO42− was being recycled in the peat system (up to 5 times in the 25 kg SO42−-S plots). Even when considering the total amount of SO42− applied throughout the experiment (in 1997 and 1998), CH4 to SO42− ratios ranged from 3.8:1 to 1.1:1 for 25 to 100 kg SO42−-S plots, respectively (Table 4).
4.1. Seasonal and Interannual Variability in CH4 Flux
 Overall, CH4 fluxes exhibited typical seasonal changes which broadly followed changes in temperature (Figures 1 and 2b), i.e., higher in the warm summer months and early autumn and low during cool winter periods [Dise, 1993; Shannon and White, 1994; Saarnio et al., 1997]. There is a major difference in CH4 flux between the two sampling years, which reflects large differences in the water table between the two years, with the field site receiving 26% less rainfall in 1997 than in 1998. With a lower water table, there is a decreased volume of potential CH4 production in the peat column as well as an enlarged oxic layer within which a large proportion of CH4 is oxidized [Daulat and Clymo, 1998].
 A further consideration is when the rainfall occurred. The July 1997 total of 93 mm is misleading as two thirds of the total rainfall fell in the first 2 days of the month, when the water table was already high (1.4 cm below the surface). It is likely that a large proportion of this precipitation ran off of the peat surface. While we cannot eliminate the possibility that this level of high-intensity precipitation may have removed, through surface runoff, some of the SO42− from a treatment given 6 days before the high precipitation event, we believe that the 13 mm of rainfall that fell in the area during the intervening time was sufficient to have “watered in” the applied SO42− and that the low hydraulic conductivity of peat limited lateral redistribution. Furthermore, it has been shown in a U.K. mire that SO42− is immediately assimilated by the peat system at high water tables (<5 cm below the peat surface), and where water table is lower than 5 cm below the surface, uptake by a combination of assimilatory and dissimilatory SO42− reduction occurs after a maximum lag of 4 days [Brown and MacQueen, 1985]. In this system the water table was 1.4 cm below the surface.
 Over both years, fluxes show a pattern that is related to temperature; however, in the warm summer and early autumn months of 1998, emissions from all sites varied on consecutive weekly sampling dates by as much as an order of magnitude (Figure 1, mean fluxes shown). This phenomenon has been observed at sites that are similarly highly productive in terms of their CH4 output [Dise, 1993; Romanowicz et al., 1995], and it has been suggested that this may be the result of changes in atmospheric pressure [Mattson and Likens, 1990], where low pressure may allow the release, as finite pulses, of large stores of dissolved CH4 that have accumulated in the peat. It is likely that the weaker relationship exhibited between CH4 fluxes, water table, and temperature in 1998 (Table 1) is due, at least in part, to such pulsing in emissions.
4.2. Effect of SO42− Treatment on CH4 Emissions
 Our experimental data demonstrate that an enhanced, chronic low-level supply of “acid rain” SO42− suppresses the emission of CH4 from wetland soils. The absence of any significant effect during the first year is likely to have resulted from the very low water table conditions which characterized the summer and autumn of 1997. It is likely that this will have limited the amount of SO42− reaching the anaerobic zone during this initial treatment period, thereby limiting the potential for stimulation of microbial competition.
 Over the duration of the experiment, between 22 and 32% less CH4 was emitted from plots treated with SO42− relative to control plots (Figure 3). This compares with results of a similar SO42− dose experiment where the weekly application rate was 50% higher than the highest dose rate we applied (maximum of 145 kg SO42−-S ha−1 yr−1) [Dise and Verry, 2001]. In addition, the level of suppression of CH4 flux by the different continuous SO42− treatments varied over time (Figure 4). During the spring of 1998, at the highest level of inhibition, treatments fluxes were suppressed by as much as 50–60%. This is within the range of inhibition resulting from single application SO42− treatments of several orders of magnitude larger than those applied here in small regular pulses [Dernier van der Gon and Neue, 1994; Lindau et al., 1994, 1998].
 One interesting and unexpected outcome from these experiments is that there are no significant differences between fluxes from treatments of different amounts of SO42−. It has been shown that at low (or below detection limit) SO42− concentrations, sulfate-reducing bacteria may be sustained by fermentatively degrading higher-chain fatty acids (e.g., propionate [Krylova et al., 1997]) and that levels of resulting hydrogen are regulated through consumption through “interspecies hydrogen transfer” by methanogens [Wolin, 1982; Conrad et al., 1987]. The addition of SO42− may stimulate a change in this mutually beneficial arrangement by switching SRB to more energetically beneficial SO42− reduction [Raskin et al., 1996; Schink, 1997]. This has the dual effect of both depriving H2 utilizing methanogens of a major substrate source (which in itself would reduce CH4 production) and then, in the longer term, enabling SRB to compete with methanogens over available substrates [Raskin et al., 1996]. Through the addition of SO42− and an SRB inhibitor (molybdate), Watson and Nedwell  found evidence to suggest that this “syntrophic” association exists in peatland microbial communities.
 In this study the lowest SO42− application rate may well exceed the threshold for SRB to switch from methanogens to SO42− as electron acceptors, whereupon beyond such a threshold no further suppressive effect will be evident as other factors, such as substrate availability, become limiting. The finding that pore water CH4 concentrations were lower in plots treated with 50 kg SO42− ha−1 yr−1 than in controls (Figure 5a) indicates that suppression of CH4 is indeed occurring at source rather than by the means of CH4 transport to the atmosphere being affected in any way. Shannon and White  found that elevated pore water sulfate corresponds with a zone of depleted CH4. However, our data show that SO42− concentrations tend to be lower (although differences are not significant) in plots treated with 50 kg SO42− ha−1 yr−1 (Figure 5b). One can speculate that SRB in these plots have been “activated” to SO42− reduction by the enhanced SO42− supply and can now outcompete methanogens. In doing so, the population of SRB may have increased such that they are able to reduce SO42− concentrations to a level that is lower in treated plots than is found in controls.
 While we cannot exclude the possibility that CH4 is being oxidized anaerobically by SRB-methanogen consortia (i.e., SO42−-dependent CH4 oxidation, reviewed by Valentine and Reeburgh ), to our knowledge, there remains little evidence of this mechanism occurring in freshwater wetland systems.
4.3. Variability in Extent of Suppressive Treatment Effect
 Not only is there an overall suppression of CH4 flux from Na2SO4 treatment plots but also the level of suppression is significantly related to changes in water table and peat temperature (Table 3 and Figure 6). In May and early June 1998 the suppressive effects of the treatments decreased with increases in temperature, as described in the regression equation effect of temperature (Table 3 and Figure 4). However, in late June and July, while temperatures remained high, fluxes from treatment plots were reduced to a level 45–60% lower than fluxes from controls. Since this enhanced suppression accompanies a lowering in water table (Figure 2), it is also described in the regression equation (effect of water table; Table 3 and Figure 6). While we have no pore water data from this period, we have already shown that later on in the year, CH4 concentrations were significantly smaller in treated plots than in controls. This suggests that with lower CH4 concentrations dissolved in treated pore waters, upon a lowering of the water table, there is less accumulated CH4 available to be released to the atmosphere. An additional effect of water table lowering may be the reoxidation of reduced sulfur compounds in unsaturated surface peat, which in SO42−-treated plots, may provide a temporarily more enriched supply of SO42− to microbial communities than would occur in controls [Freeman et al., 1994].
 The observation that the difference between treatment and control fluxes is greatest when peat temperatures are lowest was also found by Nedwell and Watson , who reported that proportionately less carbon flowed via SO42− reduction than via methanogenesis during warm summer months. They inferred that this was due to SO42− becoming limiting, as the dissolved pore water SO42− pool decreased during the summer months. We have shown that the degree of CH4 suppression is still less in summer than during cooler periods even when maintaining relatively high SO42− inputs during this time (up to 8.3 kg SO42− ha−1 month−1, close to the total amount of SO42− deposited annually on the peatland examined by Nedwell and Watson ).
 We suggest that the lower suppression in summer may have more to do with seasonal changes in substrate supply and methanogenic pathway than the availability of SO42. Several studies have shown that the main methanogenic pathway in peatlands shifts from CO2 reduction (H2 substrate) during cool periods to acetate fermentation during the warmer growing season [e.g., Kelley et al., 1992; Avery et al., 1999]. This may have implications for competitive interactions between methanogens and SRB (when sufficient SO42− is available) as the outcome of competition for acetate is known (in other anoxic soil systems) to be affected by temperature; that is, SO42− reduction is favored at lower temperatures, and methanogenesis is favored during warm episodes [van Bodegom and Stams, 1999]. Alternatively, increased production of noncompetitive substrates during summertime may allow methanogenesis to occur unheeded by the effects of otherwise competitive SRB.
 The extent of flux suppression from the treatment plots in 1998, when water table was high, are close to levels of flux suppression reported from single, large-dose experiments on high water table peat cores [Fowler et al., 1995]. Furthermore, suppression of CH4 flux in the continuous SO42− addition experiments continued (to varying extents) throughout 1998, which implies that small pulses of SO42− are sufficient to maintain a stimulated and possibly enlarged SO42−-reducing population of SRB. This is in contrast to findings by Fowler et al. , which demonstrated a pronounced recovery (after initial suppression) of CH4 fluxes from cores treated with an individual, large dose of SO42−.
 Reoxidation of reduced S compounds (sulfide) to SO42− in the rhizosphere or in upper peat layers during periods of low water table [Freeman et al., 1994] will have facilitated this continued availability of SO42− for further SO42− reduction to take place, thereby enhancing the net effect of such small SO42− pulses on CH4 emissions [Freney et al., 1982]. Indeed, CH4:SO42− ratios, while indicating that recycling of SO42− is taking place and is sustaining CH4 flux suppression, still only represent an integration of SO42− recycling over the entire peat column (Table 4). In reality, rates of SO42− recycling are likely to be far higher within finite zones in the peat column where steep redox gradients exist. Such redox gradient-induced amplification of CH4 flux suppression suggests that the impact of acid rain SO42− deposition on CH4 fluxes may continue long after the problem of acid rain SO42− deposition has been remedied.
 The decrease in CH4:SO42− ratio with increasing SO42−-S application rate suggests that at higher SO42− deposition rates, consumption of O2 (through sulfide oxidation) may limit the rate at which SO42− is recycled (Table 4). This would have the effect of closing the difference in instantaneous SO42− availability to SRB within plots from across the SO42− treatment range. In addition, in peat receiving higher SO42− deposition rates, relative levels of anoxia and competition for O2 between sulfide oxidation and methane oxidation will have increased [Arah and Stephen, 1998]. As SO42− deposition rates increase, these mechanisms will progressively act against further suppressive effect of SO42− on CH4 flux, possibly contributing to the lack of SO42− dose response observed within the experimental range we report.
4.4. Implications for the Global Wetland CH4 Source in a Sulfur-Enriched World
 Globally, anthropogenic emissions of S are forecast to double over the next 50 years from around 75 Tg S yr−1 to around 153 Tg S yr−1 in 2050 [Rodhe et al., 1995]. Areas of Asia, in particular, but also parts of Africa and South America are predicted to receive dramatic increases in S deposition during this time owing to regional economic growth. V. Gauci et al. (manuscript in preparation, 2002) estimate that 1990 levels of SO42− deposition are sufficient to suppress CH4 emissions from high-latitude natural wetlands (>50°N) by at least 5% and as much as 17%. There is, however, a trend of decreasing S deposition in these regions which is likely to continue owing to pollution control legislation [Rodhe et al., 1995]. Gauci et al. (manuscript in preparation, 2001) also suggest that such legislation may have the unforeseen consequence of increasing the total northern wetland CH4 flux as the suppressive effect of SO42− deposition is relieved. The length of time required for recovery in CH4 fluxes from previously SO42− impacted wetlands remains, however, an important uncertainty.
 Many of those regions forecast to become affected by S deposition contain extensive areas of either natural wetlands or are areas of intensive wetland rice agriculture (as in Asia), both of which are large sources of atmospheric CH4. While our data are derived from a high-latitude peatland, the lack of dependency of CH4 flux suppression on SO42− dose size (above our minimum application rate of 25 kg SO42− ha−1 yr−1) coupled with similar levels of suppression to that which has been observed in large-dose rice paddy experiments [Dernier van der Gon and Neue, 1994; Lindau et al., 1994, 1998] suggests that low-dose SO42− manipulation experiments similar to those we report here should be performed in low-latitude wetland systems.
 Results from this experiment clearly demonstrate that low rates of SO42− deposition, at levels commonly experienced in areas impacted by acid rain, significantly suppress the annual emission of CH4 from northern peatlands. We demonstrate that the flux reduction is strongest during cooler time periods as well as during periods where the water table is falling and is weakest during warm periods if the water table is near the surface. It is likely that recycling of applied SO42− facilitated the extent of CH4 flux suppression that was observed.
 With North Atlantic regions having already experienced increased SO42− deposition and subsequent decline and with low-latitude regions (Asia in particular) experiencing a trend of increasing SO42− deposition, the potential for a perturbation in the wetland CH4 source strength through such a mechanism presents us with a possible contributory factor behind recent observed variability in the atmospheric CH4 growth rate.
 The authors would like to thank Ron Smith for statistical advice and Mhairi Coyle for assistance with graphical software. Ute Skiba, Ken Hargreaves, and Reinoud Segers provided helpful comments. Berwyn Williams and Steve Chapman helped in assessing the suitability of Moidach More for this study. Permission to access Moidach More was provided by Moray Estates Development Company and Scottish Natural Heritage. The authors would also like to thank two anonymous reviewers for constructive comments. The work was funded by an Open University Ph.D. studentship for V.G. and by the Centre for Ecology and Hydrology in Edinburgh, Scotland.