Controls on the hydrogen isotopic composition of biogenic methane from high-latitude terrestrial wetlands



[1] To investigate the controls on the δD isotopic composition of biogenic CH4 in terrestrial wetlands, we collected a series of samples for δD-H2O, δD-CH4, δ13C-CH4, and δ13C-DIC (dissolved ΣCO2) along a N-S transect across Alaska from 60°N to 70°N latitude from 7 to 15 August 2001. The δD-H2O and δD-CH4 values varied from −108‰ to −161‰ and from −308‰ to −394‰, respectively, from south to north and were significantly correlated, indicating the significant influence of the δD-H2O on the δD of terrestrial CH4 collected along a latitudinal spatial gradient. Additionally, the apparent fractionation factors (α, αA–B = RA/RB, where R = 13C/12C or D/H) for H2O → CH4 (αD) and DIC → CH4 (αC) varied from 1.26 to 1.42 and from 1.035 to 1.084, respectively, and were inversely correlated. We conclude that while δD-H2O is a critical factor controlling δD-CH4 across latitudes, the CH4 production mechanism is also responsible for variation in the δD-CH4. If the isotopic values of the precursors of methane, H2O, DIC, and organic matter are relatively constant, CH4 produced by acetate fermentation will be enriched in δ13C and depleted in δD relative to CH4 produced via the CO2 reduction pathway. Production mechanism variation will be particularly important in controlling variations in CH4 isotopic composition with depth. The strong dependence of δD-CH4 on the δD of environmental H2O indicates that specific fields on plots of δD-CH4 versus δ13C-CH4 may not always accurately represent the isotopic composition of CH4 produced by CO2 reduction in northern wetlands. Because of this variation at high latitude, we assert that δD-CH4 values cannot be associated with production mechanism in an absolute sense. The production mechanism effect is in addition to the effect of the δD-H2O.

1. Introduction

[2] Methane is the final end product of the anaerobic decomposition of organic matter. It is formed by two main pathways, acetate fermentation and reduction of CO2 with hydrogen. Variation in these two main mechanisms of CH4 production has been hypothesized to be associated with inverse or antipathetic shifts in δD and δ13C of CH4 [Whiticar et al., 1986; Schoell, 1988; Burke et al., 1988a, 1988b; Whiticar, 1999; Hornibrook et al., 1997; Chanton et al., 2005]. It is often stated that fermentation of acetate will result in CH4 that is 13C enriched and D depleted relative to CH4 produced via the CO2 reduction pathway. Whiticar et al. [1986] defined somewhat distinct isotope compositional fields on a plot of δD-CH4 versus δ13C-CH4 resulting from the different biogenic pathways and from thermogenic and geothermal origin. However, a portion of the δD variation is due to variation in the precursor for CH4 hydrogen, which is in part H2O. As discussed by Waldron et al. [1999], part of the reason for the δD distinction between acetate-formed CH4 versus CO2 reduction CH4 is because samples that were representative of the CO2 reduction pathway were often from the marine environment where the δD-H2O is close to 0‰. Methane representative of acetate fermentation was more likely sampled from terrestrial environments where the δD-H2O was somewhat depleted relative to marine H2O [Waldron et al., 1999]. For terrestrial environments at high latitudes, the effect will be more acute [Waldron et al., 1999].

[3] While the usefulness of δ13C-CH4 as an indicator of methanogenic pathway has been recently demonstrated [e.g., Conrad et al., 2002; Sugimoto and Wada, 1993], the variation of δD-CH4 with shifts in the acetate fermentation and CO2 reduction pathways has been called into question [Waldron et al., 1999, 1998b; Sugimoto and Wada, 1995]. In several elegant experiments, investigators who have incubated organic matter of differing reactivity with H2O of differing δD isotopic composition [Sugimoto and Wada, 1995; Waldron et al., 1998a] have failed to observe shifts in the δD of CH4 with changing production mechanism. What does appear to control the δD of CH4 in these incubation studies is the δD of H2O. The δD-CH4 values fall very close to a single line (r = 0.986) when plotted against the δD-H2O used in the experiment [Waldron et al., 1999].

[4] Waldron et al. [1999] also plotted 51 cogenetic pairs of δD-H2O and δD-CH4 from environmental samples from a variety of environments across latitude. They obtained a significant relationship between δD-H2O and δD-CH4 and concluded that the isotopic composition of formation H2O was the primary control on the δD-CH4. They observed, however, that there was more variation for δD-CH4 than for δD-H2O, which might be caused by other processes. One of the purposes of this paper is to determine if variations in CH4 production pathways might be one of these processes. Furthermore, the data of Waldron et al. [1999] was weighted toward more tropical and temperate areas. The authors suggested a need for additional data at higher latitudes.

[5] Chanton et al. [2005] presented results from field studies ranging from tropical to boreal regions and provided strong evidence that CH4 produced from acetate fermentation is relatively 13C enriched and δD depleted relative to CH4 produced from reduction of CO2. However, they pointed out that the relationship only holds if precursor isotopic values (δ13C-DIC and δD-H2O) are relatively constant across the spatial or temporal gradients across which the measurements are made.

[6] In this paper we will examine field data to provide evidence that although precursor H2O isotopic composition is a primary factor controlling the δD-CH4, variations in CH4 production mechanisms are a second controlling factor. Therefore along a N-S transect across Alaska we measured δD and δ18O of H2O, δ13C and δD of CH4 and δ13C of DIC. Our first objective was to determine if fractionation of δD-CH4 varies with CH4 production mechanism by examining the variation of apparent fractionation factors for the pairs δ13C-DIC and δ13CH4 and δD-H2O and δD-CH4. To accomplish this we sampled semisynoptically across a horizontal gradient at different scales and synoptically along a depth gradient. Methane production mechanisms have been observed to vary as a function of depth in northern wetlands [Hornibrook et al., 1997; Chasar et al., 2000a, 2000b; Chanton et al., 2005]. Our second objective was to extend the observations of Waldron et al. [1999] by collecting additional data from a transect between 60 and 70 degrees north in Anchorage, Fairbanks and along the pipe line road in Alaska up to Deadhorse.

2. Methods

[7] Samples were collected from bogs and fens in Alaska along a latitudinal gradient, from boreal to polar regions (60.0 N to 69.5 N) and across a depth gradient at a single site. Sampling occurred from 7 to 15 August 2001 and represents a synoptic data set, not a seasonal study. Replicate samples of pore water were collected from depths ranging from 10 to 50 cm with a “swamp sucker,” which is a quarter inch stainless steel tube with holes drilled at the base combined with a 50 mL syringe. At Turnagain Bog, near Anchorage Alaska (60°N latitude) in addition to shallow swamp sucker samples, additional samples were collected at depths ranging from 1 to 2 meters using a piezometer and a peristaltic pump. DIC (Dissolved Inorganic Carbon or ∑CO2), samples were filtered through glass fiber filters and 10 mL injected into 25 mL evacuated serum vials fitted with butyl rubber septa. DIC samples were stored frozen and upside down prior to analysis and then acidified with 0.3 mL of degassed 30% H3PO4 and brought to ambient pressure through an open split tube with nitrogen for testing. Small subsamples (20 μl) of the headspace from the DIC samples were run via direct injection on the GC-IRMS (HP-Finnigan Delta S) to determine δ13C-DIC. For δ13C-CH4 and δD-CH4 analysis, pore water was injected into evacuated 500 ml Qorpak brown glass bottles, containing 1.5 g KOH and stored upside down. Storage tests indicate that samples keep for over a year in this manner. Samples of H2O for isotopic analysis were placed within hypovials and stoppered with a needle fitted through the septae to allow a sample to be collected with no headspace. The needle was removed after the stopper was inserted.

[8] The headspaces of the CH4 samples were pressurized to one atmosphere with an open split tube and equilibrated with the dissolved phase. Then a small portion of each headspace was analyzed via direct injection into the GC-IRMS (Finnigan Delta S) to determine δ13C-CH4 and concentration. Methane from the rest of the headspace was removed with a helium stream and combusted over copper oxide at 800°C and the resultant H2O and CO2 cryogenically trapped. The H2O from the CH4 combustion was isolated using an ethanol slush trap and transferred to glass tubes containing zinc shavings from Indiana University, which were then flame sealed. The tubes were heated to 500°C for 30 minutes to react the H2O with the zinc and the resultant H2 was analyzed for δD-CH4 using the duel inlet system of the Finnigan MAT IRMS [Coleman et al., 1982]. δD-CH4 was determined relative to Standard Mean Ocean Water (SMOW) using gases that were calibrated against standard waters (V-SMOW, SLAP). Precision for δD measurements on the IRMS is 0.3‰ while overall precision of replicate field samples is 6 ‰ [Lombardi, 1999]. Samples of surface and pore water were sent to Waterloo University for analysis to determine the isotopic signatures of the pore water, δD-H2O and δ18O-H2O. The δ13C values were determined relative to VPDB in ‰. Lab standards were calibrated against NIST standards. Precision for δ13C measurements was ±0.2‰ on the basis of repeated measurements of a laboratory working standard.

3. Data Analysis and Calculations

[9] Replicate samples were analyzed to discern the mean δ13C-CH4, δ13C-DIC, δD-CH4, and δD-H2O for each sample location and then an overall mean for each parameter for each site. From these data, the apparent fractionation factors [Chasar et al., 2000a; Hornibrook et al., 1997; Whiticar et al., 1986] for 13C/12C and D/H (αC and αH), can be determined for each location and site. The apparent fractionation factors for CO2 → CH4 (αC) and H2O → CH4 (αD) [Chasar et al., 2000a; Hornibrook et al., 1997; Whiticar et al., 1986] are calculated as

display math
display math

[10] Changes in the apparent αs have been hypothesized to represent changes in relative importance of the production mechanism (see discussion by Whiticar [1999]); larger αC and smaller αD are typical of CO2 reduction while smaller αC and larger αD values are typical of acetate fermentation [Whiticar et al., 1986; Whiticar, 1999; Sugimoto and Wada, 1993; Conrad et al., 2002]. We refer to these factors as apparent, because while CO2 and H2O are the precursors for CH4 formed from CO2 reduction, they are not the immediate precursors for CH4 formed from acetate fermentation. Nonetheless, variations in apparent fractionation factors are interpreted to represent variations in CH4 production mechanism [Conrad et al., 2002; Chasar et al., 2000a; Hornibrook et al., 1997, 2000a, 2000b].

4. Results

[11] The isotopic signatures of the water (δD-H2O and δ18O-H2O) are reported in Table 1. Replicate analytical results are reported for every δD measurement and at intervals for δ18O. The surface and pore water samples fall along a line (Figure 1) with a slope of 6.9 (±0.5) and an intercept of −16.6 (±7.7). The global meteoric water line has a slope of 8 with an intercept of +9 [Hoefs, 1997]. The δD-H2O decreases with increasing latitude, but shows a slight increase approaching 70°N and the Arctic Ocean (Figure 2). Pore water samples were depleted in δD and δ18O relative to surface water samples by an average of 11.2 ± 9.8‰ and 1.6 ± 1.5‰ (error represents 1 standard deviation), presumably due to surface enrichment by evaporation.

Figure 1.

A plot of δD-H2O versus δ18O-H2O indicates that both isotopes are affected by the same processes, i.e., evaporation, transpiration, and precipitation. The correlation is significant, p < 0.005. Filled squares represent pore water samples while open diamonds are surface water samples.

Figure 2.

A plot of the average isotopic signatures, δD-CH4 (diamonds) and δD-H2O (squares), of the pore water for each Alaska site/depth versus the latitude of the site. Note the increase in δD of both water and methane approaching 70 degrees north latitude and the coast of the Arctic Ocean. Correlation of both factors with latitude from 60 to 68 degrees is significant, p < 0.005.

Table 1. Isotopic Composition of Water Samples Versus Latitudea
SampleLatitudeδ18O‰±Mean δD ‰±
  • a

    Sites at Turnagain Bog were roughly 10 meters apart. The±represents 1/2 of the range of analytical replication of a single sample.

Turnagain Bog Site 1, 20cm60.16−14.50 −115.20.2
Turnagain Bog Site 2, 35cm60.16−13.55 −112.00.1
Turnagain Bog Site 3, 35cm60.16−16.30 −124.70.4
Turnagain Bog Site 4, 20cm60.16−16.23 −122.60.3
Turnagain Bog Site 5, 35cm60.16−15.47 −119.80.2
Turnagain Bog Site A, 35 cm60.16−14.190.02−112.30.2
Turnagain Bog Site B, surface water60.16−13.50 −108.30.5
Turnagain Bog Site B, 50 cm60.16−15.76 −124.50.2
Turnagain Bog Site B, 100 cm60.16−16.03 −124.90.1
Turnagain Bog Site B, 150 cm60.16−16.23 −126.40.2
Turnagain Bog Site B, 200 cm60.16−16.27 −123.10.0
Fen 61, surface water61.00−15.130.10−123.50.0
Fen 62, surface water62.81−14.300.05−115.00.4
Fen 62, Sphagnum, 20 cm62.81−17.65 −138.40.3
Fen 62, Carex, 35 cm62.81−15.37 −125.80.1
Fen 63, surface water63.00−15.92 −129.00.5
Bonanza Creek Mix, surface water64.70−18.06 −139.90.2
Bonanza Creek Mix, 27 cm64.70−18.710.06−145.00.6
Beaver Sports Hummock, 10cm64.86−18.380.20−141.90.3
Beaver Sports Hollow, 10 cm64.86−18.41 −144.90.2
Smith Lake surface water64.87−13.41 −128.80.7
Smith Lake Carex, 40cm64.87−14.59 −136.40.0
UAF Bog, surface water64.87−18.53 −143.70.2
UAF bog, 25cm64.87−18.650.03−145.00.0
Discontuous Permafrost, Surface water66.09−15.46 −131.60.2
Discontuous Permafrost Sphagnum, 30cm66.09−18.77 −154.20.4
Discontuous Permafrost Carex, 30cm66.09−19.79 −158.80.4
Fen 67, surface67.48−20.51 −160.70.4
Fen 67, 35cm67.48−20.51 −160.70.4
Galbraith Lake Slough, 35cm68.48−17.77 −145.10.7
Galbraith Lake Hummock, 35cm68.48−17.83 −146.60.1
Immnaviat Creek Carex, 35 cm68.62−20.25 −154.40.3
Imnvaiat Creek Sphagnum, 30cm68.62−19.76 −149.80.2
Pipeline 70 Fen Carex, 25 cm69.33−19.11 −147.60.0
Franklin Bluff Fen, 45cm69.80−16.48 −134.80.2

[12] The δ13C and δD mean values for duplicate sample analysis of the dissolved gases CH4 and DIC are reported in Table 2. The δ13C of DIC ranged from −19 to +3‰ driven by the importance of CH4 production and possibly acetogenesis. The most enriched DIC-δ13C values were found at depth in Turnagain Bog and at the discontinuous permafrost sites. Other than depth trends, there was no consistent pattern in DIC data, and there was no trend with latitude. Methane δ13C ranged from −50‰ to −86‰. Interestingly both extremes of the range were obtained from Carex sites (Table 2). The δD-CH4 values for these two sites were −362 ± 5‰ and −346 ± 12‰, respectively, so CH4 oxidation is not immediately obvious as an explanation for the δ13C variation observed. Had CH4 oxidation driven the differences in δ13C between these two sites, the δD values would likely have covaried [Coleman et al., 1981]. The mean δD-CH4 varied from −394‰ to −308‰. As observed for the δD-H2O, δD-CH4 decreased with latitude and similarly exhibited a small increase approaching 70°N and the Arctic Ocean (Figure 2).

Table 2. δ13C-CH4, δD-CH4, δ13C-DIC and δD-H2O Data for Each Sitea
SiteLatitudedepth, cmδ13C CH4±δ13C DIC±δD CH4±δD H2O
  • a

    Asterisk indicates that δD-H2O data was used from a pore water sample from the same site but a different depth for regression analysis and alpha calculations. The ± represents 1/2 of the range for replication of duplicate samples.

Turnagain Bog Site 160.1620−64.00.4−7.10.3−350.72−115.2
Turnagain Bog Site 360.1635−71.90.6−2.30.2−331.42−124.7
Turnagain Bog Site 460.1620−75.00.1−8.90.8−328.91−122.6
Turnagain Bog Site 560.1635−63.90.5−7.20.1−343.010−119.8
Turnagain Bog Site A60.1635−73.01.0−2.90.1−338.84−112.3
Turnagain Bog Site B60.1650−70.10.5−4.30.4−335.65−124.5
Turnagain Bog Site B60.16100−69.30.4−2.20.9−314.77−124.9
Turnagain Bog Site B60.16150−−317.71−126.4
Turnagain Bog Site B60.16200−−307.910−123.1
Fen 62 Carex62.8135−82.20.2−7.61.2−329.93−125.8
Fen 62 Spagnum62.8120−86.00.2−13.30.1−345.812−138.4
Bonanza Mix 64.7027−55.10.1−16.40.7−389.52−145.0
Beaver Sports Hollow64.8610−56.60.1−18.00.1−394.17−144.9
Beaver sports Hollow64.8640−56.91.0−14.00.3−365.015*
Smith Lake carex64.8740−50.50.3−19.10.1−362.45−136.4
UAF Bog64.8725−52.30.4−15.10.2−377.110−145.0
Discont. Perm Sphagnum66.0930−62.80.1−0.20.5−384.15−154.2
Discont. Perm Carex66.0930−−392.73−158.8
Fen 6767.4835−66.21.3−8.20.1−394.03−160.7
Galbraith Lake Hummock68.4835−67.20.2−7.30.3−371.75−146.6
Galbraith Lake Slough68.4835−62.90.2−5.90.1−388.2 −145.1
Immnaviat Creek Carex68.6235−62.30.3−15.21.3−367.013−154.4
Imnvaiat Creek Sphagnum68.6230−65.90.9−19.12.7−379.02−149.8
Pipeline 70 Fen69.3325−57.80.1−12.40.1−363.314−147.6
Franklin Bluff fen69.8025−58.10.1−9.00.5−369.04*
Franklin Bluff fen69.8045−62.80.8−5.50.0−366.05−134.8

[13] A plot (Figure 3a) of δD-CH4 versus δD-H2O for the Alaska data yields a significant correlation at the 99% confidence level indicating that δD of H2O controls a large portion of the variation in δD-CH4. The Alaska transect data extend the Waldron et al. [1999] data set as shown in Figure 3b. The overall fit to the entire data set has a slope of 0.54 and an intercept of −290. The linear regression fit to the data of Waldron et al. has a slope of 0.74 and an intercept of −284.3. The Alaska data have a steeper slope (1.55) and a more positive intercept (−145.4) relative to the other relationships. Waldron et al. [1999] obtained a slope of 0.68 and an intercept of −284 but used a weighted linear regression using the standard error of the mean δD-CH4 for each site, whereas we used the actual values.

Figure 3.

(a) Trend of δD-CH4‰ versus δD-H2O‰ in samples from the Alaska transect, standard error of slope, 0.46, intercept 29.91, p < 0.005. (b) Trend of δD-CH4‰ versus δD-H2O‰, Alaska data, as in Figure 3a (squares), with data from Waldron et al. [1999] (diamonds) included. Our line fit to the Waldron data is y = 0.74 (±0.11)x − 284.3 (±6.3); r2 = 0.50, n = 51, p < 0.005. The line fit to the entire data set is y = 0.54 (±0.05)× − 290.1 (±5.3), r2 = 0.53, n = 77, p < 0.005.

[14] One objective of this paper is to extend the observations of Waldron et al. [1999] to include more northern latitude data in the examination of the relationship between δD-CH4 and δD-H2O across the globe. Clearly, a large portion of the variation of the δD-CH4 data is explained by the variation of the δD-H2O results (Figure 3a). To the extent that the proportion of the variability in the y values is explained by a straight line fit to the x values, the data indicate that roughly 70% of the variability of δD-CH4 can be explained by δD-H2O. We will also examine the data to discern if a portion of the variation of δD-CH4 about the line fit to the δD-H2O is due to variation in CH4 production mechanism.

5. Discussion

[15] Generally, the slope of the line of δD-CH4 fit to δD-H2O has been interpreted as being indicative of the CH4 production mechanism, with CH4 generated purely from CO2 reduction having a slope of 1 because all of the CH4 hydrogen is thought to be derived from coexisting H2O. Acetate fermentation is thought to have a slope of 0.25 as only 1 hydrogen has been considered to be derived from environmental H2O while the other 3 have been thought to be transferred intact from the acetate methyl group [Whiticar, 1999; Whiticar et al., 1986]. Strict interpretation of the overall results (Figure 3b) in light of this paradigm would suggest a 60–40 mix of acetate fermentation and CO2 reduction in the generation of global terrestrial CH4 as included in the data set. However, the picture is probably not so simple as it initially appeared to be [Waldron et al., 1999]. Recent evidence has suggested that there may be an exchange of hydrogen between the acetate methyl group and H2O in the final stages of methanogenic acetate metabolism [de Graaf et al., 1996]. Presumably this isotopic exchange is accompanied by an isotopic fractionation but the magnitude of this effect has not been quantified, as pointed out by Hornibrook et al. [1998].

[16] Hydrogen concentration has also been suggested as a factor that may influence the isotopic composition of H in CH4 [Burke, 1993; Sugimoto and Wada, 1995], and this factor may be particularly responsible for the lack of agreement between incubation studies and surveys of natural wetlands [Chanton et al., 2005]. Burke [1993] proposed that hydrogen concentration could affect the fractionation of H between H2O and CH4, resulting in greater fractionation at higher H2 concentrations but not affecting the δ13C of CH4. Recently, Valentine et al. [2004] observed that microbial growth rate can have a profound effect on fractionation (αD) between H2O and CH4 during CO2 reduction, thus influencing the δD of CH4. Furthermore environmental H2O δD variation affects the δD of organic matter and thus should affect the δD of the acetate methyl group. Given these uncertainties, it is best not to interpret the slope value too strictly. It can be suggested on the basis of Figures 3a and 3b that 50 to 70% of the variation in δD-CH4 is directly related to variation in δD of environmental H2O.

[17] It is interesting that the Alaska data set indicated a slope of 1.5 (Figure 3a), greater than what would have been expected for CO2 reduction alone. This finding could be due to variation with production mechanism with latitude in our data set. Lower-latitude (60°N) data were dominated by samples from Turnagain Bog where acetate is known to accumulate in interstitial water and not form CH4 [Hines et al., 2001; Duddelston et al., 2002]. In general studies have been interpreted to indicate that CH4 production via the acetate pathway is less consequential in high-latitude Sphagnum-dominated bogs [Lansdown et al., 1992; Chasar et al., 2000a, 2000b; Chanton et al., 2005]. Variation in radiocarbon distributions between DOC, CH4, and DIC further support the idea that respiration modes fundamentally differ in bogs and fens [Chasar et al., 2000b]. Turnagain bog samples varied from −115‰ to −123‰ in δD-H2O isotopic composition, and thus represented the right side of the graph in Figure 3a. Nonetheless the Turnagain data fit the line quite well. As most of the sites we sampled were represented by only 1 or 2 data, we investigated reducing the number of surface Turnagain data points to only two randomly chosen surface samples, samples 1 and 3, which are fairly representative of the other surface samples. This reduction changed the line fit to the data little, to y = 1.27x − 188.45, r2 = 0.59. Thus the possible overrepresentation of Turnagain Bog as a lower-latitude (60°) sampling site is not responsible for the observed pattern.

[18] Addition of the Alaska data to the Waldron et al. [1999] data set has the effect of reducing the slope of the line of global water versus global CH4, converging the field data with the laboratory data. The new slope is more similar to the slope of the plot of δD-CH4 versus δD-H2O obtained from the incubation results of Waldron et al. [1998a], Sugimoto and Wada [1995], and Schoell [1980] as plotted by Waldron et al. [1999], which had a slope of 0.44 and an intercept of −321.

[19] The intercept of the line fit to the δD-CH4 fit to δD-H2O has been interpreted as being indicative of the fractionation between water H and the resulting methane H. The line fit to marine data, which is thought to be indicative solely of CO2 reduction, has an intercept of −160‰, while the intercept of various lines fit to acetate fermentation vary from −300‰ to −370‰ [Whiticar, 1999]. Our Alaska data had a more positive intercept than the overall data set, which might be interpreted as due to a greater influence of CO2 reduction at higher latitudes because of the inhibition of the acetate fermentation pathway in Sphagnum-dominated bogs. However, the steeper slope of the line clearly influenced the intercept so this interpretation is somewhat suspect.

[20] We next evaluate the hypothesis that departures from the linear fit observed in Figure 3a are related to variations in the CH4 production mechanism. We will assume that variation in the signature of δ13C-CH4 indicates variation in CH4 production mechanism, as discussed by Conrad et al. [2002] [see also Sugimoto and Wada, 1993; Waldron et al., 1998a], since samples were collected from depths that were for the most part 25 cm below the surface and thus should be relatively free of CH4 oxidation as was observed for northern wetlands by Popp et al. [1999]. When δD-CH4 is plotted versus δ13C-CH4, an inverse correlation results (Figure 4), consistent with this assumption. If δ13C-CH4 was controlled by CH4 oxidation, a positive correlation between δ13C and δD of CH4 would have been observed [Coleman et al., 1981; Chanton et al., 2005]. Additionally, the finding of a negative correlation in Figure 4 is consistent with the hypothesis that some of the variation in δD that we observed across our transect was due to variation in CH4 production mechanism. In Figure 4, there appear to be two clusters of data, those on the left where CO2 reduction is more important, and those on the right where acetate fermentation is more dominant. The left cluster data are from latitudes below 63°N and include data from Turnagain bog, where incubation studies have clearly demonstrated that CO2 reduction is the dominant production mechanism. The right-side data are from above 64°N. We suspect that this segregation of the data with latitude is serendipitous, and does not imply that there are variations in CH4 production pathways with latitude on this scale. On larger latitudinal scales, however, for example in Sphagnum-dominated wetlands which are more often found in northern latitudes, one is likely to observe that CO2 reduction is the dominate methane production mechanism. This finding has been observed in both isotopic and incubation studies [Landsdown et al., 1992; Chasar et al., 2000a, 2000b; Hines et al., 2001; Duddleston et al., 2002].

Figure 4.

Plot of δD-CH4 versus δ13C-CH4 for Alaska transect samples. While there is a significant inverse correlation between the values, a portion of this variation is due to variation in the precursors, δ13C-DIC and δD-H2O. Standard error on slope is 0.52 and on intercept is 34.2; p < 0.005.

[21] However, some portion of the variation in δD-CH4 in Figure 4 is caused by the variation in δD-H2O with latitude (Figure 3a). To remove the effect of precursor variation, we can determine if the apparent fractionation factors (αC and αD) for δ13CH4 and δD-CH4 vary inversely with respect to δ13C-DIC and δD-H2O. Plots of δ13C-CH4 and δ13C-DIC and δD-CH4 and δD-H2O are shown in Figures 5a and 5b, respectively. A wide range of αC values (equation (1)) was observed, from 1.035 to 1.084. The αDs (equation (2)) ranged from 1.26 to 1.42. While the αCs plot perpendicular to the lines of constant α, the αDs are at more of an angle, possibly because of the variation of δD-H2O with latitude.

Figure 5.

(a) Cross plot of δ13C-DIC‰ versus δ13C-CH4‰ in pore waters. Lines depict varying values of αC. (b) Cross plot of δD-H2O‰ versus δD-CH4‰ in pore waters. Lines depict varying values of αD.

[22] The αDs are plotted versus αCs in Figure 6, and a significant inverse relationship is obtained (p > 0.025). This means that as αC increases αD decreases and vise versa. Assuming that the variation in δ13C-CH4 is due to variation in CH4 production mechanism, then this finding is consistent with the hypothesis that δD-CH4 varies with production mechanism as well, in an opposite manner to δ13C-CH4. Thus it appears that variation in production mechanism is a second factor controlling the δD-CH4. As with any correlation, we cannot address the causes of the variation, we can only note that our data are consistent with its occurrence. Again, it may be argued that the data overrepresent Turnagain Bog. Omitting all but two of the surface samples and refitting the linear regression did not change the slope or the correlation coefficient (p > 0.025).

Figure 6.

Plots of the αD and αC of methane produced in peatland sites along a transect in Alaska. Plot shows a significant antipathetic relationship (p < 0.005) between αD and αC presumably resulting from shifts in production mechanism. Standard error on slope, 0.43; intercept, 0.45.

[23] Our results indicate that δD-H2O is a primary factor controlling δD-CH4 across latitudes, but that CH4 production mechanism also affects δD-CH4. If precursor isotopic values are relatively constant, CH4 produced by acetate fermentation may be enriched in 13C and depleted in D relative to CH4 produced via the CO2 reduction pathway. For depth profiles at one site, variation in production mechanism may be the dominant feature controlling δD and δ13C variation in CH4 [Hornibrook et al., 1997].

[24] If CH4 precursor isotope values vary, then variation in production mechanisms can be discerned by variation in apparent α values. Given the strong dependence of δD-CH4 on the δD of environmental water, it seems to us that specific fields on plots of δD-CH4 versus δ13C-CH4 that represent CH4 produced by CO2 reduction [Whiticar et al., 1986; Whiticar, 1999] may not accurately represent CH4 from northern terrestrial wetlands. As pointed out by Waldron et al. [1999] the CO2 reduction field in these plots was primarily determined from marine samples where the δD-H2O is near 0‰. Methane produced by CO2 reduction in terrestrial northern wetlands where the δD-H2O is ≪0‰ will clearly not fall in the CO2 reduction fields as defined by Whiticar et al. [1986, their Figure 12] and Whiticar [1999, his Figure 4]. The δD-CH4 from the CO2 reduction field in the Whiticar figures ranged from −150‰ to −250‰ for δD-CH4 and −60‰ to −110‰ δ13CH4 (our Figure 7). However, the δD and δ13C values for CH4 from Turnagain Bog ranged from −308‰ to −350‰ and from −64 to −73‰ respectively (Figure 7). At Kings Lake Bog in Washington, USA, δD-CH4 isotopic values ranged from −276‰ to −331‰ [Lansdown et al., 1992]. In both of these systems, incubation studies using field samples have shown that CH4 production is dominated by CO2 reduction and these incubation studies have been interpreted to represent field conditions [Lansdown et al., 1992; Hines et al., 2001; Duddleston et al., 2002]. Methane isotopic values from several other systems where CO2 reduction has been inferred to be the dominant mechanism of production yield similar results. Methane from raised bogs in the Glacial Lake Agassiz peatland in northern Minnesota had CH4 isotopic values of from −260‰ to −290‰ and −65 to −70‰ [Chasar et al., 2000a]. Bleak Lake Bog in Alberta CH4 values were from −300‰ −330‰ and −70 to −78‰ and are also clearly outside of the range of δD-CH4 values ascribed to CH4 from CO2 reduction (Figure 7a), although the δ13C data fit the model.

Figure 7.

(a) Diagram for classification of CH4 by a combination of δ13C-CH4‰ and δD-CH4‰. Data symbols were placed from northern North American Sphagnum-dominated bogs where CO2 reduction dominates CH4 production. Glacial Lake Agassiz raised bog in northern Minnesota, United States [Chasar et al., 2000a] (open diamonds); Bleak Lake Bog in northern Alberta, Canada [Chanton et al., 2005] (open squares); Turnagain Bog, Alaska, United States (open triangles); Kings Lake Bog in Washington State, United States [Landsdown et al., 1992] (filled squares). While the samples fall toward the bacterial CO2 reduction field, they are generally δD depleted. This depletion is due to the difference in δD-H2O between terrestrial water at northern latitudes and marine waters. Methane derived from CO2 reduction in marine samples was used to define the CO2 reduction field in this figure. Note that Glacial Lake Agassiz Fen samples (+ symbols, Chasar et al. [2000a]) do fall within the bacterial methyl fermentation field and are to the lower right (13C enriched and δD depleted) relative to Agassiz bog samples [see Chanton et al., 2005]. (b) Similar to Figure 7a except that all δD-CH4 data have been adjusted for the δD-H2O of the water at the location where the CH4 was formed (equation (3)). Glacial Lake Agassiz raised bog (open diamonds) +82‰ [Chanton et al., 2005]; Bleak Lake Bog in northern Alberta, (open squares) + 134‰ unpublished data; Turnagain Bog, Alaska (open triangles) + 120‰; Kings Lake Bog in Washington State [Lansdown et al., 1992](filled squares) + 64‰ [Waldron et al., 1999]; Glacial Lake Agassiz Fen samples (+ symbols, plus 82‰). This adjustment effectively normalizes the formation water to be 0‰, similar to marine waters. Adapted from Whiticar [1999] and Whiticar et al. [1986] with permission from Elsevier.

[25] We assert that δD-CH4 values cannot be associated with CH4 production mechanism in an absolute sense. The production mechanism effect is combined with the effect of the δD of H2O. Note that Glacial Lake Agassiz Fen surficial samples [+ symbols, Chasar et al., 2000a, depths above 1 m] do fall within the bacterial methyl fermentation field and are located to the lower right of (13C enriched and δD depleted) Agassiz bog samples as discussed by Chanton et al. [2005]. Acetate fermentation is thought to be greater in the surface soils of fens where the organic matter is more labile [Chanton et al., 2005].

[26] However, if we shift the δD values of these northern wetland CH4 samples by the value of the environmental water where they were collected (Figure 7b),

display math

then the values fall within the CO2 reduction field. Essentially we have corrected them to values they might have had if they had been produced from δD-H2O = 0‰.

[27] In conclusion, the results of this study and results of field studies compiled by Chanton et al. [2005] have provided strong evidence that CH4 produced from acetate fermentation is relatively 13C enriched and δD depleted relative to CH4 produced from reduction of CO2, if precursor isotopic values are relatively constant. Laboratory incubation studies have produced similar results for δ13C; CH4 produced from acetate is 13C enriched relative to CH4 produced from CO2 reduction [Sugimoto and Wada, 1993; Waldron et al., 1998a]. The laboratory studies have not produced the apparent δD variation with production mechanism observed in field studies. However, the time span represented by the vertical profiles [Hornibrook et al., 1997; Chasar et al., 2000a, 2000b] where antipathetic variation of δ13C-CH4 and δD-CH4 have been observed, represent much longer timescales (hundreds of years) than can be represented in laboratory incubations which simulate changes in organic matter quality. One must consider that it may take 500 years to generate a pathway shift through increased recalcitrance of organic matter in a natural system (E. Hornibrook, personal communication, 2006). So in that respect, incubation studies do not totally rule out the notion that CH4 production mechanism could be caused by long slow changes in organic matter reactivity.

[28] Burke [1993] noted that CH4 produced in laboratory incubations and in the rumen with elevated H2 concentrations exhibit depleted δD values relative to values found in wetlands and sediments. He hypothesized that fractionation between the δD-H2O and the δD-CH4 was related to the partial pressure of H2. It is possible that variation of δD-CH4 associated with variations in H2 concentration may be an additional factor complicating comparisons of field samples with laboratory incubations. Valentine et al. [2004] clearly demonstrated that variations in microbial growth rate can have a huge effect on δD-CH4 and postulated that this factor might explain the results of Sugimoto and Wada [1995]. Carrying this idea further, it is interesting to speculate that the variation in δD that has been observed in field studies [e.g., Chanton et al., 2005; Hornibrook et al., 1997; this study] and attributed to production mechanism variations may have been due to covariation of H2 concentration or microbial growth rate with production mechanism, or some other factor that covaries with production mechanism [c.f. Waldron et al., 1998b]. Perhaps H2 concentration is greater in soils and sediments with more labile organic matter where acetate fermentation is of more relative importance [Burke, 1993]. Overall, there is a clear need for additional investigation of H isotope fractionation between organic matter, acetate, CH4, H2 and H2O.


[29] We thank Kate Chanton, Juliette Rooney-Varga, and Khrys Duddleston for assistance with field work. Cheryl Kelley read and commented on an earlier version of this paper. This project was supported by the National Science Foundations, Office of Polar Programs, grants 0093677 and 0095034. Reviews by E. Hornibrook, D. Valentine, and S. Waldron helped us improve the manuscript.