Aircraft measurements of carbon and hydrogen isotopic ratios of atmospheric CH4 (δ13CH4 and δD-CH4), with the respective precisions of 0.08‰ and 2.2‰, as well as CH4 concentration were made at 1 and 2 km altitudes over western Siberia during 2006–2009. δ13CH4 and δD-CH4 were almost always lower at lower altitudes, while the CH4 concentration was higher, implying strong sources on the ground with low isotopic values. δ13CH4 showed a clear seasonal minimum in the late summer, while seasonality of CH4 and δD-CH4 was ambiguous due to the local disturbances. By inspecting the relationships between the CH4 concentration and isotopes, we found that isotopic source signatures in the winter (December–April) are −41.2 ± 1.8 and −187 ± 18‰ for δ13CH4 and δD-CH4, respectively, and the corresponding values in the summer (June–October) are −65.0 ± 2.5 and −282 ± 25‰. These values indicate predominant CH4emissions from fossil fuel facilities in the winter and wetlands in the summer. It was also found that the shorter-term CH4 variations are more influenced by fossil CH4 than that from wetlands. The finding presumably reflects the fact that the former is released from limited areas such as leakage from fossil fuel facilities, while the latter is released from a vast expanse of wetland. By employing a CH4 emission data set used in an atmospheric chemistry transport model, we calculated seasonal isotopic changes of CH4 sources in western Siberia and compared them to the estimates obtained in this study. The results indicated that the seasonal change in the CH4 emission data set is reasonable, at least in terms of a ratio of fossil to biogenic emissions.
 Although methane (CH4) is an important trace gas for atmospheric chemistry and climate, quantitative estimation of its global budget still has large uncertainties [Forster et al., 2007]. CH4 is emitted into the atmosphere from both natural and anthropogenic sources through such processes as biogenic production (wetland, rice paddy, livestock, termites, etc.), ventilation and leakage during drilling and transportation of fossil fuel, and biomass burning (wildfire, biofuel and agricultural burning), and once it is in the atmosphere more than 80% is destroyed by an oxidation reaction with the hydroxyl radical (OH) in the troposphere [Forster et al., 2007]. It has also been reported that terrestrial plants emit CH4 under aerobic environment [Keppler et al., 2006], although its production mechanism and global emission strength are still debated.
 Atmospheric CH4 concentration distribution is closely correlated with spatially and temporally varying CH4 sources and sinks. To understand the variations of atmospheric CH4 over geographically wide area, a great effort has been devoted to establishing global observation network [e.g., Cunnold et al., 2002; Dlugokencky et al., 2011]. Although many of these observational stations have been established at sites relatively remote from areas of strong CH4 sources for monitoring the baseline CH4 variations, we need to establish stations at many unmonitored large regional areas with strong CH4 emissions. In fact, modeling studies using CH4 data obtained from the present observation network have suggested that CH4 sources in regions with no direct CH4 measurements play an important role in global CH4 cycle [Houweling et al., 2006; Bousquet et al., 2011]. Observations of atmospheric CH4using the SCIAMACHY instrument aboard the ENVISAT satellite, as well as model analyses of its observational results, have also suggested that Tropical Africa, Tropical Asia and South America play important roles in global-scale atmospheric CH4 variations [Frankenberg et al., 2005, 2006; Bergamaschi et al., 2007, 2009; Bloom et al., 2010]. Northern high latitude areas, such as the western Siberian Lowland and the Hudson Bay Lowland, are also important for atmospheric CH4, since the boreal wetlands acting as a strong natural CH4 source dominate a significant portion of the mid to high latitude Northern Hemisphere.
 In the light of such an importance of this region for the CH4 cycle, aircraft observations of atmospheric CH4 were made at various locations using a continuous measurement system [Tohjima et al., 1996, 1997] and a grab sampling method [Nakazawa et al., 1997a] in the summers of 1992 to 1994, as a part of the Siberian Terrestrial Ecosystem-Atmosphere-Cryosphere Experiments (STEACE). These observations revealed summertime spatial distributions of atmospheric CH4 over Siberia, reflecting strong CH4 emissions from boreal wetlands and oil/gas fields; this is consistent with the results of simultaneous measurements of atmospheric δ13CH4 [Sugawara et al., 1996]. The measurements of δ13CH4 also indicated that the wetland CH4 emission contributed more than the fossil fuel CH4 emission to the observed high atmospheric CH4 over western Siberia [Sugawara et al., 1996]. More recently, the National Institute for Environmental Studies (NIES), Japan has made systematic observations of CH4 concentration in western Siberia by using towers [Sasakawa et al., 2010] and aircraft [Machida et al., 2001; Xiong et al., 2010].
 The CH4 emission from the boreal wetlands is seasonally dependent with a maximum in the summer [Whalen and Reeburgh, 1992; Walter et al., 2001; Zhuang et al., 2004; Pickett-Heaps et al., 2011]. On the other hand, the CH4 leakage from the fossil fuel facilities is assumed to be seasonally invariable in modeling studies [e.g., Patra et al., 2011]. In this regard, summertime technological emissions such as venting during routine maintenances might increase the CH4 release from the natural gas pipelines in that season [Reshetnikov et al., 2000]. The CH4leakage is also thought to be slightly larger in the winter than in the summer due to higher inner pressures inside the pipelines for the increased fossil fuel consumption in the winter (S. Maksyutov, personal communication). Using a coupled Eulerian-Lagrangian transport model with seasonally changing CH4 fluxes from wetlands and biomass burning, along with an annually constant fossil fuel CH4 flux, Sasakawa et al.  showed that a major contribution to the observed atmospheric CH4 variations at their tower sites came from the wetlands in the summer and the fossil fuel in the winter. δ13CH4 and δD-CH4 originating from wetlands and fossil fuel are significantly different, with the respective values of about −60 and −300‰ for wetlands and about −40 and −180‰ for fossil fuel [Quay et al., 1999]. This will allow us to distinguish between these two major sources of CH4. Indeed, Fisher et al.  recently showed, from their observations at an Arctic background site Zeppelin, Spitzbergen, under substantial influence from Siberian CH4 sources, that δ13CH4 signals are different between winter and summer, since fossil and biogenic sources are dominant in the respective seasons.
 In this paper, we present the measurements of δ13CH4 and δD-CH4, along with the CH4 concentration, from the air samples collected by the NIES aircraft over western Siberia. The study period is from 2005 to 2009. Experimental procedures are described in section 2, data analyses in section 3, and the results and discussion in section 4, with conclusion in section 5.
2. Experimental Procedures
 Under the NIES program, collection of air samples over western Siberia has been carried out vertically at altitudinal levels ranging from 0.5 to 7.0 km upwind of Surgut (61°N, 73°E) within about 100-km distance once a month since July 1993 using a chartered aircraft (AN-24) [Machida et al., 2001; Xiong et al., 2010]. The ground surface is mainly covered by boreal wetlands crisscrossed by many oil and natural gas pipelines (Figure 1). Air samplings were made over areas away from fossil fuel facilities to avoid direct contaminations. The air sample was introduced into the cockpit through an inlet situated in front of the engine exhaust and pressurized into a 550 mL Pyrex glass flask at about +0.2 MPa over cabin pressures by using an electric diaphragm pump (MOA-P101-JH, GAST Manufacturing Inc.). Prior to use, the flasks were washed using an ultrasonic cleaner filled with purified water, and then evacuated for at least 6 h to pressures lower than 0.13 Pa at 100°C. The air samples collected at 1 and 2 km height levels during May 2004–May 2009, April 2006–May 2009 and June 2005–May 2009 were analyzed for CH4 concentration, δ13CH4 and δD-CH4, respectively.
 The air samples collected for this study were sent to Tohoku University and first analyzed for the CH4concentration by using a gas chromatograph (Agilent 6890, Agilent Technologies Inc.) equipped with flame ionization detector (GC-FID), with a precision of 2 ppb. The CH4 concentration values were determined against the working standard gases that were calibrated by the gravimetrically prepared primary standards (Tohoku University 1988 scale) [Aoki et al., 1992; Umezawa, 2009]. The intercomparison of CH4 standard gases, conducted by the World Meteorological Organization (WMO), shows differences of 2.1–2.6 ppb between the Tohoku University and WMO scales in a concentration range of 1750–1840 ppb [Zhou et al., 2009].
 The values of δ13CH4 and δD-CH4were determined using a gas chromatograph combustion/pyrolysis isotope ratio mass spectrometer (GC-C/P-IRMS) installed at Tohoku University. Since the technical aspects of our mass-spectrometric analysis have been described byUmezawa et al. in detail, only a brief description is presented here. A 100-mL aliquot of an air sample was flushed by pure helium into a CH4 preconcentration trap containing HayeSep D maintained at −130°C and then warmed to −83°C to release simultaneously trapped gases such as N2. The trapped CH4was transferred into a cryo-focusing trap (CP-PoraBOND Q) kept at −196°C. The concentrated CH4 was released into a PoraPLOT Q column for separation from the residual gas components. Then CH4 was combusted into CO2 at 940°C or pyrolyzed into H2 at 1450°C for the subsequent continuous flow mass spectrometer measurements of δ13CH4 and δD-CH4, respectively, using ThermoQuest/Finnigan Delta Plus XP. The analytical precision was estimated to be 0.08‰ for δ13CH4 and 2.2‰ for δD-CH4.
3. Data Analysis
 To extract a long-term trend and seasonal cycle from the time series of the CH4 concentration, δ13CH4 and δD-CH4, a digital filtering technique developed by Nakazawa et al. [1997b]was applied. The technique consists of a stepwise calculation process involving linear interpolation, Reinsch-type cubic splines, Fourier harmonics and a Butterworth filter. In this study, a Butterworth filter with a cutoff period of 24 months, at which a signal is attenuated by 50%, was used to obtain long-term trend, while the fundamental and its first harmonics from the Fourier analysis with periods of 12 and 6 months, respectively, yielded average seasonal cycle. Additionally, a Butterworth filter with a cutoff period of 4 months yielded short-term irregular fluctuation with periods 4–24 months. A best fit curve to the observed data was obtained by summing the long-term trend, the average seasonal cycle and the short-term irregular variations.
 The isotopic signatures of CH4 sources can be derived from measured values of the atmospheric CH4 concentration and δ13CH4 or δD-CH4 [e.g., Pataki et al., 2003]. If a constant flux of CH4 is added from a source to the background atmosphere, the corresponding mass conservation equations for CH4 concentration and its isotope (δ13CH4 or δD-CH4) are given by
Here, C and δ represent the CH4 concentration and δ13CH4 (or δD-CH4), respectively, and the respective subscripts of obs, BGD and S denote the observed, background and source values. From equations (1) and (2), we can obtain a linear relationship between δ and 1/C (often referred to as the Keeling plot technique),
This equation tells us that the intercept value of the regression line obtained by plotting the observed δ values against the reciprocal of the observed CH4 concentration is the isotope ratio of the source that exchanges CH4 with the atmosphere. Miller and Tans  also derived a modified form of the Keeling plot technique based on equations (1) and (2),
Equation (4) is equivalent to equation (3), but the slope of the regression line, obtained by plotting the product of the observed C and δ values against the observed C, is interpreted as the isotopic ratio of the source. Hereafter we refer to the plot using equation (4) as the Miller/Tans plot A method. In this connection, Zobitz et al.  found that the results obtained using the Keeling plot and the Miller/Tans plot A methods are essentially consistent with each other. Miller and Tans also derived the following equation by re-arrangingequation (4) for the case that the C and δ values of the background atmosphere vary with time;
The slope of the regression line, obtained by plotting the left-hand side ofequation (5)against the bracketed term on the right-hand side, indicates the isotopic ratio of the source. Hereafter we refer the plot usingequation (5) as the Miller/Tans plot B method.
 To determine the slope of the linear regressions for Miller/Tans plots A and B, Pataki et al.  and Miller and Tans  recommended using a geometric mean regression (GMR). On the other hand, Zobitz et al.  found, by examining the source isotopic signatures from the atmospheric CO2 concentration and its δ13C, that the GMR method yields biases when the data distribute in a narrow range, while an ordinary least squares regression (OLR) does not show such biases. We therefore use the OLR method for the determination of the slope. Uncertainties given in this study for the source isotopic ratio are based on the standard errors of the OLR method.
4. Results and Discussions
4.1. Variations of Atmospheric CH4 Concentration, δ13CH4 and δD-CH4
Figure 2 shows time series of the CH4 concentration, δ13CH4 and δD-CH4observed at 1 and 2 km altitudes over Surgut, together with the best fit curves and their long-term trends (see also theauxiliary material). As seen in Figure 2a, variations in the CH4 concentration exhibit certain noticeable characteristics. First, CH4 shows no clear seasonality. In contrast, the CH4 concentrations observed at northern baseline sites, such as Point Barrow, Alaska (71°N, 157°W) and Ny Ålesund, Svalbard (79°N, 12°E), show clear seasonal cycles, with a prominent minimum and a broad maximum appearing around July and in the winter/spring, respectively [Dlugokencky et al., 1994; Morimoto et al., 2006]. We will discuss the details of the observed CH4 seasonality later. Second, the observed CH4 concentration has large irregular variations at both altitudes. The range of the variation reaches nearly 150 ppb, which exceeds the seasonal CH4 amplitude of about 50 ppb at Ny Ålesund [Morimoto et al., 2006]. Third, the CH4 concentration is generally higher at 1 km than at 2 km, suggesting the presence of strong CH4 sources on the ground. For comparison, the annual mean CH4 concentration observed at Ny Ålesund during this period was about 1860 ppb [S. Morimoto, personal communication]. The CH4 concentration level at 1 km was always higher than that observed at Ny Ålesund, while the concentration at 2 km was comparable in magnitude. This indicates that atmospheric CH4 at 1 km over Surgut was significantly elevated from the baseline level at the northern high latitudes. Such elevated CH4 concentrations with large variations were also observed at tower sites in western Siberia [Sasakawa et al., 2010].
 As seen Figures 2b and 2c, δ13CH4 and δD-CH4 also show large temporal variations at both altitudes, with the range of variation at 1 km generally larger than at 2 km. Similar to the CH4 concentration, the observed δ13CH4 and δD-CH4 variations are much larger than those reported previously at baseline sites in the Northern Hemisphere [Bergamaschi et al., 2000; Morimoto et al., 2006; Tyler et al., 2007]. It is likely that these features of atmospheric CH4 and its isotopes are the results of strong influences from local CH4 sources on the ground.
 No significant increases or decreases of the CH4 concentration trend are apparent at both altitudes during the time before 2007, period during which the global baseline CH4 level had remained nearly constant [Rigby et al., 2008; Dlugokencky et al., 2009]. However, from the last quarter of 2007 to the first half of 2008, we see a general increase due to the sporadic occurrence of high CH4 concentrations. This CH4 increase is coincident with the large increase of the baseline CH4 level in 2007 [Rigby et al., 2008; Dlugokencky et al., 2009], although the limited temporal resolution of our data should be considered. Following this short period of increase, CH4 shows a decrease, keeping in mind the end effect of the curve fitting procedures.
 As far as the isotopes are concerned, measurements showed no significant increase or decrease in the δ13CH4 trend for 2006–2009, while relatively low δD-CH4 values were observed during 2007–2008. If regional CH4 emissions from wetlands were enhanced in 2007, as suggested by previous studies [Rigby et al., 2008; Dlugokencky et al., 2009; Sasakawa et al., 2010; Bousquet et al., 2011], it would have caused a simultaneous decrease in δ13CH4 and δD-CH4, with a corresponding increase in CH4. Our observations show decrease of δD-CH4, but no significant decrease/increase of δ13CH4. It should be also noted that high CH4 concentration with very low δ13CH4 and δD-CH4 values in the late summer 2007 could be arisen from enhanced CH4 emissions from wetlands. Low δ13CH4 values were also observed at a baseline site Alert, Canada, which suggests enhanced CH4 emissions from Arctic wetlands [Dlugokencky et al., 2009]. Temporally higher resolution data are required to better represent the enhancement of CH4 emissions in this region.
Figure 3 shows seasonal cycles of the CH4 concentration, δ13CH4 and δD-CH4at 1 and 2 km over Surgut against calendar months, together with their seasonal components of the best fit curve. All data are shown as deviations from the long-term trends of the best fit curve (i.e., detrended). As seen in this figure, no clear seasonal cycles are apparent in the CH4 concentration and δD-CH4. On the contrary, δ13CH4 shows a clear minimum in the late summer (September) and a broad maximum in the winter/spring season. The seasonal δ13CH4cycle at 1 km is more prominent than at 2 km with a peak-to-peak amplitude of about 0.9 and 0.6‰, respectively. In contrast,Morimoto et al.  showed a clear seasonal CH4cycle with a minimum and a maximum in July and January–February, respectively, with a peak-to-peak amplitude of 48 ppb, and aδ13CH4cycle with a maximum and a minimum in June and October, respectively, with a peak-to-peak amplitude of 0.42‰ at Ny Ålesund. At the same observation site,Umezawa  showed that δD-CH4 also displayed a clear seasonal cycle with a phase nearly opposite to that of the CH4concentration, and its peak-to-peak amplitude was 10.8‰. These seasonal features at Ny Ålesund are generally similar to those at baseline sites in the Northern Hemisphere [Bergamaschi et al., 2000; Miller et al., 2002; Tyler et al., 2007], but different from those observed at Surgut shown in Figure 3. It is likely that the background seasonal cycle is masked by strong influences from the local sources at Surgut.
 It is reasonable to assume that CH4 in this region is predominantly influenced by emissions from fossil fuel facilities in the winter and from both wetlands and fossil fuel facilities in the summer. Contributions from additional sources such as domestic animals and biomass burning are possible but thought to be relatively small [Sasakawa et al., 2010]. Since fossil and wetland sources have distinct δ13CH4 and δD-CH4 signature values, effect of each source on atmospheric δ13CH4 and δD-CH4 would depend on the relative ratio of two emission contributions. If fossil fuel CH4, whose respective δ13CH4 and δD-CH4 values are −40 and −180‰, is added to the atmosphere, the CH4 concentration would increase accompanied by an increase in δ13CH4 and a decrease in δD-CH4. On the other hand, if biogenic CH4, whose respective δ13CH4 and δD-CH4 values are −60 and −300‰, is added, the CH4 concentration would increase accompanied by a decrease in both δ13CH4 and δD-CH4. Note that these two different sources drag atmospheric δ13CH4 in the opposite directions, while simultaneously pushing atmospheric δD-CH4 to lower values.
 Another possible process influencing atmospheric δ13CH4 and δD-CH4 is kinetic isotope effect (KIE) that occurs during the CH4 + OH oxidation reaction that preferentially removes lighter isotopologue (12CH4). As a result, the residual atmosphere becomes enriched in heavier isotopologues (13CH4 and CH3D) that causes δ13CH4 and δD-CH4 to increase. It is expected that CH4 destruction by OH is most active in the summer when the OH density reaches a maximum [Spivakovsky et al., 2000]. However, the observed CH4 concentration does not show a seasonal minimum in the summer (Figure 3a), indicating that the effect of the OH reaction is overwhelmed by the local CH4 emissions from wetlands that also peak in the summer.
 Based on the formulae by Lassey et al. , Morimoto et al.  calculated the sensitivity factors of δ13CH4 to be −0.007, +0.004‰/ppb for wetland and fossil fuel CH4 emissions, respectively. Corresponding value for the CH4 + OH reaction is −0.002‰/ppb with an experimental carbon KIE [Saueressig et al., 2001]. Likewise, corresponding sensitivity factors of δD-CH4 are −0.14 and −0.05‰/ppb for wetland and fossil fuel emissions, respectively, and −0.11‰/ppb for the CH4 + OH reaction with an experimental hydrogen KIE [Saueressig et al., 2001]. Suppose wetland CH4 emission and the CH4 destruction by OH both reach their peaks in the summer over Surgut. Both of these processes will have an opposite effect on δD-CH4, resulting in a relatively small net effect on the δD-CH4 seasonality. On the other hand, the δ13C-CH4 sensitivity to the wetland is about three times larger than to the OH reaction. Consequently, atmospheric δ13CH4 would be significantly more influenced by the wetland emission than by the OH reaction. This would explain why δ13CH4 shows a clear seasonal minimum in the late summer while the seasonal δD-CH4 cycle is masked by the local disturbance (Figures 3b and 3c).
4.2. Vertical Differences in the CH4 Concentration, δ13CH4 and δD-CH4
Figure 4 shows seasonal changes in the vertical difference in the CH4 concentration, δ13CH4 and δD-CH4 between the 1 and 2 km height levels obtained from all the flights. As seen in Figure 4a, most data show positive values (the CH4 concentrations at 1 km are higher than those at 2 km) with a highest value of about 120 ppb, suggesting the presence of a strong CH4 source on the ground throughout the year. On the other hand, the vertical differences in δ13CH4 and δD-CH4 are generally negative (these isotopic values at 1 km are lower than those at 2 km). This indicates that, in general, the CH4 source on the ground has lower δ13CH4 and δD-CH4 values compared with those of the atmospheric CH4. It is noteworthy that these vertical gradients are pronounced in the summer, implying that the CH4 source on the ground is enhanced and/or isotopic signatures of the source change in the warm season.
 Using the vertical differences in the CH4 concentration and isotopes observed by different flights, we estimated δ13CH4 and δD-CH4 values emitted from the ground source with the following equation:
where subscripts S, 1 km and 2 km refer to the source value and the observed altitudes, respectively. Note that equation (6) assumes simple mixing of air between the two altitude levels. Figure 5 displays seasonal changes of δS against calendar months obtained by the equation. Some calculations yielded values outside the vertical ranges shown in the figure, due to a small value in the denominator of equation (6) (the vertical gradient is small) and/or inappropriate assumption of the simple mixing. We therefore excluded calculated values in cases where the vertical CH4 gradient fell less than 10 ppb. In this sense, the vertical CH4 gradient would be an indicator of reliability of the estimated δS values and is indicated in Figure 4 by the size of each circle. As seen in Figure 5a, δ13CH4 of the source (δ13CS) has a dip in the summer (July–September) compared to relatively high values for the rest of the year. Although δD-CH4 of the source (δDS) is more ambiguous, a similar summertime dip could be seen in the relatively scattered values. The respective δ13CS and δDS values of around −70 and −350‰ in the summer agree well with those of biogenic sources, indicating summertime predominance of wetlands as the major source of CH4 in this region. On the other hand, δ13CS and δDS values of the source in the rest of the year mostly range between −40 and −70‰ and between −150 and −350‰, respectively, which correspond to intermediate values of fossil and biogenic CH4 sources, suggesting contributions from both sources.
4.3. Source Identification Using Isotope-Concentration Relationships
 To find out the types of CH4 source at Surgut that would affect the atmospheric CH4 variations, we applied the Miller/Tans plot method to δ13CH4 and δD-CH4. Figure 6 shows Miller/Tans plot A (see equation (4)) using data from 1 and 2 km altitudes. Open and closed circles indicate two different clusters consisting of wintertime (December–April) and summertime (June–October) data, respectively. Particularly, the wintertime δ13CH4 values are generally higher than the summertime values at the same CH4 concentration. We applied OLR to each cluster, yielding respectively slopes of −41.2 ± 1.8 and −187 ± 18‰ in the winter and −65.0 ± 2.5 and −282 ± 25‰ in the summer for δ13CH4 and δD-CH4. Strictly speaking, these slopes correspond to flux-weighted isotopic ratios of CH4 sources and sinks contributing to the atmospheric variations. The wintertime slopes of δ13CH4 and δD-CH4 agree well with those reported for the fossil fuel CH4, supporting the prior suggestion that the CH4 release from fossil fuel facilities is dominant in the winter. On the other hand, the summertime slopes of δ13CH4 and δD-CH4 are much lower than those obtained for the winter. For comparison, CH4 emitted from the wetlands in the northern high latitudes has δ13CH4 and δD-CH4 values ranging from −60 to −80‰ and from −300 to −420‰, respectively [Martens et al., 1992; Nakagawa et al., 2002; Walter et al., 2006, 2008]. The summertime slope of the δ13CH4 regression line in Figure 6 falls well within these values for CH4 emitted from northern wetlands, while the summertime slope of the δD-CH4 regression line falls slightly higher. This comparison result suggests that the fossil fuel emissions also contribute to some extent during the summer.
 Some studies have examined CH4 sources in western Siberia based on isotopic measurements of atmospheric CH4. Sugawara et al.  conducted aircraft observations in the summer 1994 and found that the observed atmospheric CH4variations over Khanty-Mansiysk and Plotnikovo located in wetlands were predominantly affected by CH4 emitted from the wetlands with δ13CH4 values ranging from −67 to −75‰. From the air samples collected over Surgut, they also estimated the source δ13CH4 to be −59.4‰, which suggests a substantial contribution from fossil CH4 in addition to wetland CH4. Observations of atmospheric δ13CH4 and δD-CH4were also made using the Trans-Siberian railway during the TRans-Siberian Observations Into the Chemistry of the Atmosphere (TROICA) program, running through the southern part of western Siberia [Bergamaschi et al., 1998; Tarasova et al., 2006]. They identified air samples influenced by CH4 from wetlands, whose δ13CH4 and δD-CH4 values were estimated to be −62.4 ± 1.8 and −314 ± 19‰, respectively in the summer 1996 [Bergamaschi et al., 1998] and −62.9 ± 0.7 and −355 ± 29‰, respectively in the summer 1999 [Tarasova et al., 2006]. Yamada et al.  collected surface air samples at Plotnikovo and by using aircraft over Surgut in the summer of 1994. They obtained the source isotopic signatures at Plotnikovo to be −69.8 ± 0.6 and −355 ± 29‰ for δ13CH4 and δD-CH4, respectively, which are in good agreement with the values of wetland CH4. Using the data obtained over Surgut, they also estimated δ13CH4 and δD-CH4 signatures of CH4 source in this region to be −34 ± 4 and −428 ± 13‰, respectively. The δD-CH4 signature obtained agrees with the value of wetland CH4, but the δ13CH4 signature is much higher. This inconsistent behavior of the δ13CH4 and δD-CH4 signatures might relate to confusion of fossil and wetland sources, but each contribution was difficult to be quantified [see Yamada et al., 2005]. Tower measurements at Korotchaevo, northern part of western Siberia, in the summer 2004 also showed that the δ13CH4 signatures of CH4 source range from −49.8 to −67.3‰, suggesting contribution of both wetland and fossil fuel sources [Nisbet, 2005]. These observational campaigns were conducted during the summer and the estimated δ13CH4 and δD-CH4 signatures for wetlands compare well with our summertime values obtained using the Miller/Tans plot A, with exceptions of isotopic signatures obtained by Yamada et al. .
 In order to resolve the temporal variations in the relationship between δ13CH4 or δD-CH4 and the CH4concentration revealed by the Miller/Tans plot A method, we introduce a Miller/Tans plot B approach. In this approach, we employ various components from the curve-fitting procedures to represent different time scales of interannual, seasonal and short-term fluctuations. For the background terms inequation (5), we insert, in Case I, the long-term trend components, and in Case II the best fit curve components. In addition, we also examined Case III in which the data obtained at 2 km are assumed to represent background variations.
 Our first choice (Case I) was to assume the long-term trend as background variations.Figures 7a and 7b show respectively Miller/Tans plots B for δ13CH4 and δD-CH4. It is clearly seen in Figure 7a that the wintertime slope of the regression line differs from the summertime one. The slope of δ13CH4 in the winter was estimated to be −44.4 ± 2.9‰ (R2 = 0.91), and −61.4 ± 3.0‰ (R2 = 0.95) for the summer. For δD-CH4 (Figure 7b), the winter-summer difference is less obvious. Wintertime and summertime slopes are estimated to be −197 ± 24 (R2 = 0.91) and −227 ± 24‰ (R2 = 0.73), respectively.
 The second choice (Case II) was to assume the best fit curve as background variations. δ13CH4 and δD-CH4 regressions for this choice are shown in Figures 7c and 7d, respectively. Seasonal separation is clearly found for the δ13CH4 regression, while it is ambiguous for the δD-CH4 regression. The respective slopes in the winter and summer were calculated to be −41.8 ± 2.1 (R2 = 0.94) and −58.9 ± 1.9‰ (R2 = 0.98) for the δ13CH4 regression, and −197 ± 17 (R2 = 0.82) and −211 ± 20‰ (R2 = 0.95) for the δD-CH4 regression.
 The third choice (Case III) was to regard the variations at 2 km as the background. In this case we inspect the data at 1 km in terms of the influence from surface sources. It should be noted that the analysis for this case is conceptually the same as we made for Figure 5, but presentation method is different. Plots of δ13CH4 and δD-CH4 for this case are shown in Figures 7e and 7f, respectively. For δ13CH4, the winter and summer slopes are −46.5 ± 2.6 (R2 = 0.97) and −66.1 ± 1.1‰ (R2 = 0.99), respectively, while the corresponding slopes for δD-CH4 are −152 ± 43 (R2 = 0.49) in the winter and −323 ± 26‰ (R2 = 0.92) in the summer. It is noted that these isotopic signatures for this case show a good agreement with the values in Figure 5. However, Case III would represent the average isotopic signatures of the surface CH4 source over the winter and summer, while the values in Figure 5 are those of CH4 source important for the atmospheric variations observed on each flight.
Table 1 summarizes the estimated δ13CH4 and δD-CH4 values of CH4 source in the winter and summer derived from Miller/Tans plot A method and the three choices of Miller/Tans plot B described above. As shown in this table, the summertime CH4 sources all have lower δ13CH4 and δD-CH4 values. This is attributable to a substantial contribution from the wetland CH4 emission in the summer, since such CH4 has low δ13CH4 and δD-CH4 values. In the winter, CH4 emission from wetlands stops due to low temperature and snow cover on the ground, allowing fossil fuel facilities as the primary CH4 source in that season. In fact, our estimated δ13CH4 and δD-CH4 values of CH4 source in the winter agree relatively well with those of fossil CH4.
Table 1. Source δ13CH4 and δD-CH4 Signatures (‰) Estimated Using Miller/Tans Plots A and B
Long-Term (Case I)
Best Fit (Case II)
2-km Data (Case III)
−41.2 ± 1.8
−187 ± 18
−44.4 ± 2.9
−197 ± 24
−41.8 ± 2.1
−197 ± 17
−46.5 ± 2.6
−152 ± 43
−65.0 ± 2.5
−282 ± 25
−61.4 ± 3.0
−227 ± 24
−58.9 ± 1.9
−211 ± 20
−66.1 ± 1.1
−323 ± 26
 Interestingly, the estimated δ13CH4 and δD-CH4in the summer rise in value from those obtained by Miller/Tans plot A, Miller/Tans plot B (Case I) and Miller/Tans plot B (Case II). Recall that in the Miller/Tans plot A method, all variable components are included in the calculation, while in Case I only the seasonal and shorter-term sporadic variations are included, and in Case II only the short-term fluctuating components are included. Therefore, shorter-period variations become prominent in this comparison. Since biogenic CH4 sources have lower δ13CH4 and δD-CH4, while fossil fuel CH4sources have higher values in the study region, our finding suggests that the longer-term CH4variations are under larger influence from biogenic sources than those from fossil fuel sources, while the shorter-term variations are the opposite. The following scenario might provide an explanation for this. Suppose certain point sources such as oil and natural gas wells and pipeline joints release substantial amount of CH4 randomly. Resulting spikes of high CH4 concentration could be observed by aircraft, depending on its flight route and weather condition; such phenomena were previously observed by Tohjima et al. . In contrast, CH4 emission from vast wetlands would contribute to a general increase in atmospheric CH4in the summer and would be observed by the aircraft on every flight, resulting in overall long-term CH4 variations. It is also expected that CH4 emission from wetlands is highly variable in space and time [Moore et al., 1994; Walter et al., 2006]. If this is the case, wetland CH4 is partly responsible for the fact that the isotopic signatures from irregular variations (Case II) are lower than the fossil source values (see Table 1).
 Finally, we discuss potential contribution of the CH4destruction by OH to the source estimation made using Miller/Tans plots. The source signature estimated using this type of plot represents the flux-weighted isotopic ratio of CH4 sources and sinks, which is expressed as the following [e.g., Miller and Tans, 2003],
where F is the relative strength of flux a or b, and δ is a isotopic ratio of the corresponding variables. Here we assume two “source” components, wetlands with δ13CH4 of −60‰ and CH4 destruction by OH with an effective isotopic signature of −52‰ (atmospheric δ13CH4 plus KIE). If Fa = 5 and Fb= −1 are taken for the respective relative source strength, we obtain the flux-weighted source signatureδS to be −62‰, which is lower than the isotopic value assumed here for wetlands. This may imply that the isotopic “source” signature derived using Miler/Tans plot is underestimated unless the contribution of CH4 destruction by OH is negligibly small. Considering the lifetime of atmospheric CH4 with respect to reaction with OH is still several years at latitudes of interest even in summer, the effect of OH on the irregular variations around the best fit curve (Case II) would be small. However, this is presumably not the case for the Case I in which the seasonal cycles of CH4 and isotopic ratios are considered as part of the signal. Therefore, it could be possible that the differences of the δ13CH4 signatures estimated in Case I and Case II is partly attributable to the OH effect in addition to the relative ratio of biogenic and fossil sources. On the other hand, the OH effect on δD-CH4 is expected to be small, since the large hydrogen KIE (∼220‰) [Saueressig et al., 2001] yields an effective δD-CH4 value of about −310‰ for the CH4 destruction by OH, which is close to the wetland isotopic signature.
4.4. Seasonal Variation in CH4 Emission in Western Siberia
 To further investigate the relative contributions of biogenic and fossil CH4 sources to the observed variations in atmospheric CH4 in our study area, we evaluated the CH4 emission data set used in an atmospheric chemistry transport model [Patra et al., 2009]. This CH4 emission data set was prepared by applying optimal scaling factors to the natural/biogenic emission data from the Goddard Institute for Space Studies (GISS) [Fung et al., 1991; Matthews and Fung, 1987] and the anthropogenic/industrial emission data from the Emission Database for Global Atmospheric Research (EDGER) inventory [Olivier and Berdowski, 2001]. Patra et al.  demonstrated that the observed CH4 variations at surface baseline sites around the world were reproduced relatively well by their model employing this CH4 emission data set. CH4 emissions from an area that enclosed Surgut (Latitude: 51.5–71.5°N, Longitude: 59.5–92.5°E, see Figure 1) were categorized into five major sources. Monthly CH4 emissions from these five source categories are shown in Figure 8a. While the fossil fuel component remains relatively constant throughout the year, seasonality of biogenic emission is controlled by the emission from northern bogs. The biogenic emission contributes 80% and 10% of the total emission in the summer and winter, respectively (see black solid line in Figure 7a). The biogenic category here includes animals, swamps, bogs and tundra (see Patra et al.  for more information).
 Using this CH4emission data set, we calculated flux-weighted averageδ13CH4 and δD-CH4 of sources (abbreviated as δ13CS and δDS) in this region. The results are shown in Figures 8b and 8c for δ13CH4 and δD-CH4, respectively. For the calculation, we assumed that the biogenic and fossil sources have δ13CH4 values of −65 and −40‰ and δD-CH4 values of −300 and −180‰, respectively. As contribution from the biogenic CH4 source grows toward the summer, both δ13CS and δDS decrease due to its much lower isotopic values against fossil fuel source. Since δ13CH4 and δD-CH4 values of each source are thought to have a range of variations, we performed other calculations with slightly different δ13CH4 and δD-CH4 source values (see S2 and S3 in Figure 8).
 We compared these δ13CS and δDS values with the isotopic signatures of CH4 sources estimated from the Miller/Tans plot methods applied to the summer and winter data. The estimated isotopic values are shown in Figures 8b and 8cas bands of different colors. Red and dark green shaded bands were obtained by Miller/Tans plot A, while light-blue, blue and light-green shaded bands were estimated from Miller/Tans plot B Case I, Case II and Case III, respectively. In the winter, the calculated values ofδ13CS and δDS from the CH4 emission data set are about −42‰ and −190‰, reflecting more than 90% contribution by the fossil emission to the total source emissions. The wintertime values of δ13CH4 and δD-CH4 obtained by the Miller/Tans plot A method agree well with these values (see Table 1). This agreement validates the earlier assertion of atmospheric CH4 variation in the winter being predominantly driven by CH4 emitted from fossil fuel sources in this region.
 In the summer, δ13CS decreases and reaches a minimum value of about −60‰ in July due to the enhanced emission from bogs, and agrees relatively well with the summertime isotopic values estimated from the Miller/Tans plot A and B methods (the colored shaded bands in Figure 8b). On the other hand, δDS agrees with the isotopic values obtained by Miller/Tans plot A (the green shaded band in Figure 8b), but lower than those obtained by Miller/Tans plot B Case I and Case II (the light-blue and blue shaded bands inFigure 8b). Similar to behavior in the δ13CH4 case, the enhanced CH4 emission from bogs in the summer drags δDS to lower values, reaching a minimum of about −280‰ in July. Based on the previous discussion about the relative influences from biogenic and fossil fuel sources as a function of timescale, the estimated isotopic values of CH4 source in the summer contain a larger contribution from biogenic and fossil fuel sources for longer and shorter timescale variations, respectively. In this context, δDS in the summer should be compared to the δD-CH4 value estimated using Miller/Tans plot A or Case III of Miller/Tans plot B that contains signals from biogenic CH4 sources. Accordingly, the CH4 emission data set used in the modeling study is thought to be reasonable.
 Finally, using the results from the calculations involving the three cases of isotopic signatures of δ13CS and δDS for biogenic and fossil fuel sources (S1, S2 and S3 in Figures 8b and 8c), we were able to estimate the likely isotopic signatures of CH4 emission in the study area. Wintertime comparisons among these different cases suggest that for the winter season, fossil δ13CH4 values of around −40‰ (S1 and S2) are more likely than the higher value of −35‰ (S3) (Figure 8b). Similarly, fossil δD-CH4 values are more likely to be around −180‰ (S1 and S2) instead of −160‰ (S3) (Figure 8c). These values agree well with those suggested previously as representative values for natural gas [Quay et al., 1999]. For the summertime, our calculated δ13CH4 value of −70‰ (S2) agrees with the estimates from Miller/Tans plot A and Miller/Tans plot B Case III better than with −65‰ from S1 and S3 (Figure 8b), suggesting it to be a more likely signature for biogenic CH4 sources in the region. For δD-CH4 (Figure 8c), the values between −300‰ (S1 and S3) and −350‰ (S2) seem to be likely. These values fall well within the range of variation for CH4 from wetlands in the northern high latitudes [Martens et al., 1992; Nakagawa et al., 2002; Walter et al., 2006, 2008], but the δ13CH4 value of −70‰ is somewhat lower than the previous suggested representative value, while the δD value agrees well with the reported value [Quay et al., 1999]. In this regard, Fisher et al.  indicated that a dominant summertime CH4 source in the Arctic is wetland with δ13CH4 of −68.7 ± 2.4‰. The isotopic signatures of biogenic CH4 source would vary from region to region depending on the processes controlling their emission [e.g., Nakagawa et al., 2002; Chanton et al., 2006; Walter et al., 2008].
 Aircraft observations of CH4 concentration, δ13CH4 and δD-CH4 over western Siberia were simultaneously made during 2006–2009. δ13CH4 obtained at 1 km altitude showed clear seasonal variation with a sharp minimum in the late summer and a broad maximum in the winter and spring, while the seasonality of CH4 and δD-CH4 was masked by large irregular variations. Vertical gradients in the CH4 concentration, δ13CH4 and δD-CH4 between 1 and 2 km altitudes indicated the presence of strong CH4 sources on the ground with low δ13CH4 and δD-CH4 signatures. The vertical gradients were also used to estimate δ13CH4 and δD-CH4 of the ground source of CH4, the analysis of which indicated that emission from the wetlands around Surgut was the major contributor to the vertical CH4 gradients in the summer.
 Using the Miller/Tans plots A and B methods, we examined δ13CH4 and δD-CH4 signatures of different sources influencing the atmospheric CH4 concentration variations of various timescales. The lowering of δ13CH4 and δD-CH4 signatures obtained during the summer season (June–October) suggested that CH4from wetlands has a significant influence on the long-term atmospheric variations while fossil fuel CH4 appears to impact a more shorter timescale variation. On the other hand, δ13CH4 and δD-CH4 signatures obtained in December–April agreed with those previously reported for CH4 originating from fossil fuel. Thus, in the winter season, fossil fuel facilities provide the primary source of CH4.
 The results obtained from Miller/Tans plots illustrate characteristics of CH4 sources in western Siberia. Ventilation or leakage of CH4 from fossil fuel facilities such as oil and natural gas wells and pipeline joints occurs throughout the year, providing relatively constant emission of CH4. On the other hand, since wetland is the dominant vegetation in the region around Surgut, its influence on atmospheric CH4 is seasonal, with major contribution occurring in the summer; it also contributes to the overall background variation in the CH4concentration and acts as a controlling factor in long-term trend. Such seasonal characteristics of the two primary CH4 sources are consistent with the CH4 emission data set presently employed in an atmospheric chemistry transport model.
 We are grateful to Dr. Shamil Maksyutov for his helpful comments and information on Siberia, as well as for providing us a map of Surgut. We thank Dr. Shinji Morimoto for providing CH4 concentration data at Ny Ålesund, Svalbard, and Dr. Prabir K. Patra for providing a CH4 emission data set. We also acknowledge Dr. Kaz Higuchi for his helpful comments on our manuscript. We are thankful to Central Aerological Observatory and Institute of Microbiology, Russia, for the air samplings in this study. We acknowledge Dr. John B. Miller and an anonymous reviewer for their very constructive comments on this paper.