Global Biogeochemical Cycles

Potential glacial-interglacial changes in stable carbon isotope ratios of methane sources and sink fractionation

Authors


Abstract

[1] Past atmospheric methane emissions can be constrained by δ13CH4 records from ice cores only if changes to source δ13CH4 signatures and sink isotope effects with varying environmental and climatic conditions are accurately known. We present reconstructions of such changes based on paleodata and recent systems observations. The results are specific for budget scenarios and are reported here for two alternative types of budgets, one including aerobic methane emissions (AMP) from plants and the other type without AMP. Shifting atmospheric δ13CO2 potentially led to 13CH4 enrichment by 0.8‰ in the preindustrial Holocene (PIH) (∼150–11,000 years (a) B.P.) and ∼0.3–0.6‰ at the Last Glacial Maximum (LGM) (∼18,000 a B.P.) relative to today. Differing distribution of C3 and C4 plant precursor material may account for 13CH4 enrichment of ∼0.4‰ (PIH) and ∼0.6–1.1‰ (LGM). Temperature-dependent fractionation and varying methanogenic pathways in wetlands may lead to atmospheric 13CH4 depletion by ∼0.1–1.2‰. Sink fractionation today (7.4‰) is higher than during the PIH (∼7.0‰) and the LGM (∼5.7‰). The cumulative effect of all processes is ∼0.8‰ 13CH4 enrichment in the PIH and ∼1–1.2‰ 13CH4 depletion at the LGM. Budget reconstructions will be inaccurate if these changes are not included.

1. Introduction

[2] The reconstruction of methane emissions in preindustrial and glacial times is critical to understand natural climate variability [Wunsch, 2006] and the significance of the current concentration rise of this radiatively active gas [Prather et al., 2001]. Contributions from different CH4 sources and sinks types can be constrained through isotope-based mass balances using emission rates and the 13C/12C ratio (δ13CH4) of source types, as well as the fractionation caused by sinks, and comparing them to the isotope ratio of the atmosphere [Whiticar, 1990; Fung et al., 1991; Hein et al., 1997]. The first extended preindustrial δ13CH4 records now become available from air occlusions of polar ice, allowing for constraints of past CH4 budgets [Ferretti et al., 2005; Sowers et al., 2005; Schaefer et al., 2006; Schaefer and Whiticar, 2007]. These studies show unexpectedly 13C-rich atmospheric δ13CH4 values, which were attributed to changes in emission rates of individual source types. Schaefer et al. [2006] and Whiticar and Schaefer [2007] considered the possibility that the δ13CH4 of source types and sink fractionation in the past could have been different from today's values. The underlying calculations are presented in detail here. Recent advances in the field lead to slight adjustments of the values reported by Schaefer et al. [2006], but the main findings remain valid. The Ferretti et al. [2005] and Sowers et al. [2005] studies of the late preindustrial Holocene (PIH), assumed that there were no past changes in source δ13CH4 and sink fractionation. We test this assumption by reviewing the influence of environmental parameters on each step of the CH4 cycle (Figure 1), using knowledge of CH4 production in modern systems in conjunction with paleorecords of climate and vegetation.

Figure 1.

Methane isotope signal path. (right) Stages of the CH4 cycle (sink processes produce CO2 again) are shown. (center) Processes that are associated with isotopic fractionation are listed. (left) Environmental parameters influencing those processes are shown.

[3] The δ13CH4 of methane sources is determined largely by physical and chemical parameters, e.g., reaction rates. As CH4 production occurs in biological systems that are adapted to certain environmental conditions, it is possible that climate change rather affects geographical distribution and emission rate of a source type than its isotopic signature. For example, in a colder climate a certain wetland plant community might be reestablished further south in similar environmental conditions as before, but still produce methane with the same, or almost the same, δ13CH4.

[4] In section 2 of this study, we analyze the factors that may affect the δ13CH4 of natural sources under different climatic conditions. As shown in Figure 1, there are three steps in which source δ13CH4 may be altered. First, we determine how changes in atmospheric δ13C of CO2, as well as vegetation patterns influence the δ13C of organic CH4 precursor material. Second, we study the impact of temperature change on CH4 production itself, specifically in wetlands as the largest source, which is very sensitive to climate [Kaplan, 2002]. The main factors are the biochemical pathways of methanogenesis and their attendant isotope effects. Third, we look at partial oxidation before emission to the atmosphere. Throughout these steps, we estimate the resulting changes in atmospheric δ13CH4 using isotope mass balances.

[5] In section 3, we investigate the various processes that remove CH4 from the atmosphere, which have characteristic kinetic isotope effects (KIE). We study how the resulting fractionation varies between modern and preindustrial conditions, as well as between glacial and interglacial stages in dependence of the changing relative contributions of individual sink processes and physical changes to their fractionation coefficients.

[6] Finally in section 4, we use the above estimates to determine whether environmental influences on individual source δ13CH4 and sink fractionation can indeed be neglected or if they account for the unexpected 13C enrichment in the past. Any conclusions must take the large degree of uncertainty in the derived estimates into account. Uncertainty results from the multitude of factors in play, the scaling of case studies to global estimates, uncertainties in past environmental conditions and even gaps in the understanding of present-day systems. Some findings should be regarded as rather qualitative and are presented as such to stimulate further field and modeling studies.

2. The δ13CH4 of Natural Methane Sources

[7] Methane is generally produced during the remineralization of organic matter under anaerobic conditions. Archaeal CH4 production occurs in fresh water and marine sediments, soils, and in anaerobic microenvironments, e.g., animal guts. The produced δ13CH4 depends mostly on the pathway of the breakdown [Whiticar et al., 1986]. Other major influences are the δ13C of the precursor material and the fraction of utilizable organic material consumed [Games and Hayes, 1976; Whiticar, 1996]. If the produced gas is partially oxidized during transport to the atmosphere, the residual emitted CH4 will be more 13C rich [Whiticar and Faber, 1986], which is here treated separately from sink processes that remove atmospheric CH4. For any source type the above parameters will be somewhat variable, resulting in a typical δ13CH4 range for this source rather than one fixed value. Nevertheless, isotope signatures can distinguish different source types [e.g., Whiticar, 1990].

[8] The impact of individual sources on the isotope signature of the global (or integrated) source flux (δ13CE) is given by the isotope mass balance:

equation image

where E is global emission rate in Tg/a (1 teragram = 1012 g); E1 through En are emission rates of source types with their respective isotope values δ13CEi. δ13CHEi values for the present are based on studies reviewed by Bréas et al. [2001] and given in Table 1. Modern E values are based on the IPCC TAR findings [Prather et al., 2001], corrected to the NOAA04 scale [Dlugokencky et al., 2005] and constrained by both δ13CH4 and deuterium/hydrogen ratios [Whiticar and Schaefer, 2007]. The choice of emission scenario is of lesser importance for the considerations in this study. Although the source budget will influence any changes in δ13CE between different time periods, these result from a combination of changes in relative emission rates and changes in the isotopic signature of individual sources. Quantifying the latter is the purpose of this study and we use example budgets in order to delineate, and correct for, the contribution of relative emission rate changes. This will provide an estimate of the global impact of changes in δ13CEi of a given source, which must be accounted for in any budget derived from paleoreconstructions. Such budget reconstructions, e.g., from ice core δ13CH4 data, must therefore follow iterative steps of assigning a suitable budget scenario and correcting the associated δ13CE for δ13CHEi changes until both the changes in emission rates and of δ13CE are consistent with the atmospheric δ13CH4 record over a climatic transition.

Table 1. Isotope Mass Balance for the Modern Systema
 Total ModernNatural Modern With AMPNatural Modern Without AMP
Flux, Tg/aδ13CEi,bFlux, Tg/aδ13CEib, ‰Flux, Tg/aδ13CEib, ‰
Anthropogenic sources      
   Rice paddies111−63    
   Ruminants (cattle)66−61    
   Natural gas20−44    
   Coal25−37    
   Biomass burning41−25    
   Landfills40−55    
Natural sources      
   Boreal wetlands38−6238−6251−62
   Tropical wetlands78−5978−59104−55
   Wild animals15−6015−6015−60
   Termites16−6316−6316−63
   Ocean10−5810−5810−58
   Fresh water4−544−544−54
   Gas hydrates4−634−634−63
   Geologic35−4235−4235−42
   AMP40−5140−51--
Bottom-up δCE −54.2 −55.9 −55.7
Top-down δCE −54.5 −57.1 −57.1
ɛwt 7.4 7.4 7.4
Calculated atmospheric545−46.8241−48.5241−48.3
Measured atmospheric545−47.1c −49.7d −49.7d

[9] One budget item must be discussed in more detail, namely, aerobic methane production (AMP) in plant material, which has been reported from a laboratory study by Keppler et al. [2006]. Some field measurements [Crutzen et al., 2006; Sanhueza and Donoso, 2006] and satellite data [Bergamaschi et al., 2007] seem to support this process, but another laboratory study [Dueck et al., 2007] failed to reproduce the results and offered an alternative explanation for the findings of Keppler et al. [2006]. Also, the global magnitude of AMP emissions is controversial. The original estimate by Keppler et al. [2006] of 150 ± 90 Tg/a, which is as high as wetland emissions (∼140 Tg/a [Kaplan, 2002]), has been revised downward based on stable carbon isotope studies [Ferretti et al., 2005] and scaling to leaf area (∼53 Tg/a; ∼36 Tg/a) or photosynthetic rates (∼10 Tg/a) [Parsons et al., 2006; Kirschbaum et al., 2006], as opposed to NPP in the original work. The value we present for AMP for the modern budget in Table 1 is within the revised estimates. As the exact process that drives AMP remains poorly understood and there is a high degree of uncertainty in the global annual production rate (even whether it exists at all), we present two alternative budget scenarios for every case study throughout this paper. The first includes AMP at a rate comparable to natural wetland emissions following the reconstructions of Houweling et al. [2006], which must be seen as an upper estimate in light of the work by Parsons et al. [2006] and Kirschbaum et al. [2006]. In the AMP scenario, wetland emissions are reduced accordingly to keep the total source strength unchanged. The alternative scenario includes no AMP at all. Together, the alternatives provide a conservative estimate of the range in which individual source δ13CE changes can affect the total budget. One constraint on the modern budget scenarios is the necessity that the natural sources alone produce a δ13CE value that is compatible with ice core data from the period immediately preceding the industrial revolution. The modern AMP scenario gives an atmospheric δ13CH4 of −48.5‰ (all δ13CH4 values are reported in standard δ notation relative to VPDB standard), while the non-AMP version gives −48.3‰, if tropical wetland δ13CEi is revised to −55‰ as discussed by Schaefer et al. [2006]. Although these values are 13C depleted by ∼1‰ compared to the ice data [Ferretti et al., 2005; Schaefer and Whiticar, 2007] the agreement is reasonable given the uncertainty regarding early anthropogenic activity in the late PIH [Ruddiman and Thomson, 2001] and the fact that the modern natural sources are influenced by land use change.

[10] Ei values for PIH and Last Glacial Maximum (LGM) are derived from vegetation reconstructions by Chappellaz et al. [1993] with two additional source types. First, ice core δ13CH4 measurements from the PIH and the late glacial [Ferretti et al., 2005; Schaefer et al., 2006] point to a missing 13C-rich source. Among several possible sources or sinks, we consider geologic, i.e., natural thermogenic emissions, a strong candidate. We chose the lower range of geologic emission estimates (35–45 Tg/a) [Etiope and Milkov, 2004, Kvenvolden and Rogers, 2005] because it is more compatible with total source strength in PIH and LGM [Chappellaz et al., 1993; Brook et al., 2000]. Other emissions were reduced proportionally. Second, we present scenarios including AMP as discussed above. For the PIH the assumption of equal emissions from AMP and wetlands is in agreement with the top-down calculations of Houweling et al. [2006] (85 Tg/a from AMP and 80 Tg/a from wetlands). However, the assumptions made here demonstrate the need for better estimates and reconstructions of emission rates. This is also seen in the fact that the combined bottom-up estimates for geologic [Etiope and Milkov, 2004], wetland [Kaplan, 2002] and AMP emissions [Keppler et al., 2006] exceed total source estimates for the PIH by Chappellaz et al. [1993] twofold.

[11] The choice of specific, sometimes poorly constrained, emission rates will influence the reconstructed δ13CE. Changes between different time periods with varying emission scenarios can therefore not be calculated directly. To isolate these effects, we quantify all potential changes in δ13CH4 relative to scenarios with identical emission rates but present-day values for the tested environmental parameter.

2.1. Changes in δ13C of Methane Precursor Material

[12] The δ13C of organic compounds that serve as precursors in CH4 formation influences δ13CEi. In turn, δ13C of plant matter depends directly on the stable carbon isotope ratio of atmospheric CO2, as well as on the photosynthetic pathway and various environmental parameters. In the following sections we estimate the impact of δ13CO2 and C3–C4 vegetation changes on δ13CEi of individual and total sources. We show that changes in δ13CO2 lead to measurable 13C enrichment in the past. δ13CH4 changes caused by C3–C4 vegetation shifts are complex and seem to depend mostly on emission rates in sensitive areas. We show in the following that the estimated result is a ∼0.4–0.7‰ 13C depletion between glacial and interglacial conditions (in addition to further depletion caused by changes in relative emission rates).

2.1.1. Changes in Atmospheric δ13CO2

[13] Marino et al. [1992] use plant fossils to identify a 1.1‰ 13C enrichment of atmospheric δ13CO2 in the PIH, and a 0.4‰ 13C enrichment in glacial conditions, relative to today. The PIH result is in good agreement with δ13CO2 reconstructions from ice cores [Francey et al., 1999; Smith et al., 1999]. However, the ice data show higher 13C enrichment in the LGM (0.9‰) and a minimum (0.4‰ 13C enrichment versus today) during the Younger Dryas cold period (YD) [Smith et al., 1999]. It is likely that these shifts translate into changing isotopic signatures of plant material and the CH4 derived from it. Therefore we assume that δ13CEi of archaeal and pyrogenic sources change proportionately to δ13CO2. The impact on atmospheric δ13CH4 follows from equation (1). We calculate atmospheric 13C enrichment by ∼0.8‰ in the PIH and by ∼0.3–0.6‰ in the LGM (for the range of reported δ13CO2 changes) compared to scenarios where source δ13CEi is not adjusted to past variations in δ13CO2. These results are the same for scenarios with and without an AMP source.

2.1.2. Changes in C3 and C4 Vegetation

[14] Plants can be classified as following one of two major photosynthetic pathways, i.e., Calvin-Benson cycle (C3) and Hatch-Slack cycle (C4). For C3 plants, CO2 transport is achieved by diffusion, which causes isotopic fractionation. In contrast, C4 plants actively transport CO2 into the cells and consequently discriminate less against 13C. Therefore C4 plants (δ13C of ∼−15‰) are isotopically distinct from C3 vegetation (∼−27‰) [Ehleringer et al., 1997]. The C4 pathway is only dominant in grasslands and generally has a lower yield of carbon fixed per unit of light, but it outcompetes C3 in warm and dry conditions, and under decreased carbon dioxide concentrations [CO2] [Ehleringer et al., 1997]. Accordingly, Collatz et al. [1998] model the global distribution of C3 and C4 plants for present, PIH and LGM. Grassland area dominated by C4 plants increases from ∼70% at the LGM to ∼74% in the PIH. In modern times this proportion drops to ∼57%, because of anthropogenic land conversion and rising [CO2] [Collatz et al., 1998]. One would therefore expect all sources that use C4 material as CH4 precursor to be more 13C rich in the PIH than in the LGM than today. To determine whether this influences δ13CE one must account for the changing percentage of grasslands in global vegetation [Prentice et al., 1993] and for climate-dependent CH4 production rates in different ecosystems. The influence of C3–C4 vegetation change on each CH4 source type is treated in the following sections.

2.1.2.1. Effect on Emissions From Ruminants

[15] Methane produced by ruminants is isotopically dependent on the ratio of C3 and C4 in their diet as shown by Rust [1981], Metges et al. [1990], Levin et al. [1993], Schulze et al. [1998] and Bilek et al. [2001] (Table 2). We omit data by Metges et al. [1990], in which δ13CH4 is calculated from δ13CO2, because they are inconsistent with all other data. The average value of the above studies for δ13CEi from cows feeding on C3 plants is −69‰, whereas cows eating C4 plants produce −54‰. This 15‰ difference is almost the same as between the C3 and C4 feed (12‰). Significant species-dependent differences in δ13CEi between cows, sheep, goats and camels [Levin et al., 1993; Schulze et al., 1998] are smaller than the isotopic offset between C3 and C4 derived CH4 observed in all species. In addition, the difference between feed and CH4 (∼41‰) is independent of species and diet in all studies. Therefore we assume that cows are representative for all ruminants and that δ13CEi produced by wild animals is directly and exclusively dependent on the C3–C4 proportion of their diet.

Table 2. The δ13CEi Produced by Ruminants Depending on Dieta
 SourceC4 in Diet, %δ13C Diet, ‰δ13CEi, ‰Offset Between Diet and CH4, ‰
CowsL0−27−65.138.1
CowsB0−28.9−7243.1
CowsB0−28.9−72.143.2
CowsM0−27.1−70.243.1
CowsR0−27−63.736.7
CowsS0−27−68.241.2
CowsB3−27.6−6638.4
CowsB8−25.6−65.239.6
CowsB8−25.6−65.940.3
CowsB13−23.6−62.639
CowsB13−23.7−61.838.1
CowsL30−22.8−63.340.5
CowsL70−17.2−55.638.4
CowsL100−13−5441
CowsB100−13−54.541.46
CowsM100−12−56.244.2
CowsR100−13−50.337.3
Average C3   −68.6 ±3.5 
Average C4   −53.7 ±2.5 
SheepR0−27.9−68.640.7
SheepL0−27−70.643.6
SheepS0−27−74.447.4
CamelsS0−27−73.546.5
GoatsL0−27−65.238.2
Average    40.9 ±2.9

[16] Today's global average diet of domestic cattle consists of C3 and C4 plants at a ratio of 2.5: 1 [Stevens and Engelkemeir, 1988], resulting in expected ruminant δ13CEi of ∼−64‰. Wild animals exhibit no preference for ingesting C3 versus C4 plants [Ehleringer and Monson, 1993], so that the diet should reflect the natural abundance of the two plant types. Using the distribution of grassland types from Collatz et al. [1998] and the δ13CH4 reported from cows, we calculate the δ13CEi emitted by wild animals as ∼−60.5‰ today, ∼−57.9‰ in the PIH and ∼−58.5‰ at the LGM (Table 3). We neglect differences between species, the effect of diet on the amount of CH4 production [Bilek et al., 2001], differing population density in various ecosystems and the low number of forest living animals (C3 diet only). These are not expected to affect the isotopic differences between different climatic stages. There is a remarkable difference between the δ13CEi of present-day wild (∼−60.5‰) and domestic animals (∼−64‰). The latter value should not be used for ruminant emissions in the past. Under unchanged climatic conditions, we estimate wild animal emissions in the PIH to be ∼2.6‰ more 13C rich than today, which results from more C4 plants at lower atmospheric [CO2] and changes in land use. Owing to colder and drier conditions with lower [CO2] at the LGM, we calculate animal δ13CEi to be ∼0.6‰ more 13C depleted than in the PIH.

Table 3. Influence of C3/C4 Vegetation on δ13CEi of Natural Methane Sourcesa
 PresentPIHLGM
Flux, Tg/aC4 Derived, %δ13CEi, ‰Flux,Tg/aC4 Derived, %δ13CEi, ‰δ, ‰Flux, Tg/aC4 Derived, %δ13CEi, ‰Δδ, ‰
  • a

    Source δ13CEi adjusted for changing proportions of C3 and C4 precursor material [after Collatz et al., 1998; Chappellaz et al., 1993]. Δδ for individual sources is the offset between adjusted and present-day δ13CEi. Total Δδ is the offset between atmospheric δ13CH4 values in the adjusted mass balance and one using present-day source δ13CEi.

Budget Scenarios Including AMP Emissions
AMP4221−48.94232−48.50.43430−48.60.3
Tropical wetland3226−59.14434−58.30.82346−56.72.4
Geologic35 −41.835 −41.80.035 −41.80
Boreal wetland164−61.8264−61.80.029−61.30.5
Termites1631−63.01620−63.0 (−62.2)0.0 (0.8)1318−63.0 (−62.7)0 (0.3)
Wild animals357−60.51274−57.92.61370−58.52
Ocean8 −58.28 −58.20.08 −58.20
Wildfire448−24.6420−26.7−2.1421−26.6−2
Clathrate4 −62.54 −62.50.00 −62.50
 
Total16616−52.519222−53.30.413225−50.70.6
 
Budget Scenarios Without AMP Emissions
Tropical wetland6026−59.17134−58.30.85446−56.72.4
Geologic35 −48.835 −41.80.035 −41.80
Boreal wetland36461.8424−61.80.059−61.30.5
Termites1631−63.01620−63.0 (−62.2)0.0 (0.8)1318−63.0 (−62.7)0 (0.3)
Wild animals35760.51274−57.92.61370−58.52
Ocean8 −58.28 −58.20.08 −58.20
Wildfire448−24.6420−26.7−2.1421−26.6−2
Clathrate4 −62.54 −62.50.00 −62.50
 
Total16615−55.619220−55.90.413228−52.91.1
2.1.2.2. Effect on Wildfire CH4

[17] Chanton et al. [2000] report values between −17 and −26‰ for grass fire δ13CEi (C4) that are distinct from the range of −26 to −30‰ for forest fire δ13CEi (C3), where the respective ranges depend on combustion efficiency. Changing C3–C4 abundance in the past is therefore expected to influence wildfire δ13CEi. We calculate the percentage of C3 and C4 plants burned annually in wildfires in PIH and LGM using (1) paleovegetation data from Prentice et al. [1993], (2) natural fire frequencies for different ecosystems [Wright and Bailey, 1982; DeBano et al., 1998], (3) CH4 production rates in fires from Hao and Ward [1993], and (4) the C3–C4 distribution from Collatz et al. [1998]. The amount of wildfire CH4 that we calculate for present day with this approach is substantially higher than estimates by Hao and Ward [1993], most likely due to overestimated fire frequencies. However, downscaling the results would not affect calculated δ13CH4 shifts, which are the focus of this study. We present different scenarios covering the quoted ranges in fire frequency and δ13CEi (Table 4). Despite these uncertainties and the resulting range of δ13CEi for the PIH (∼−24.2 to −29.4‰, average −26.6‰) and LGM (∼−24.1 to −29.4‰, average −26.7‰), the calculated differences between the two time periods are< 0.3‰ for all scenarios. Using data from Hao and Ward [1993] today's δ13CEi from biomass burning is calculated to be ∼−24.6‰, indicating a ∼2‰ 13C enrichment between PIH and modern conditions.

Table 4. Methane Emissions From Wildfires and Its δ13CEi During LGM and PIH Emission Rates for Each Ecosystem Calculated From Literature Estimatesa
 Biomass1/Fire FrequencyFraction BurnedCH4/Biomass, g/kgCH4 Emissions
LGM, Pg CPIH, Pg CLGM, Tg/aLGM Average, Tg/aPIH, Tg/aPIH Average, Tg/a
Grassland         
   Warm grass and shrubland10.3130.1–0.50.811.51.2–6.22.11.6–7.92.6
   Cool grass and shrubland7.25.70–0.10.811.50.3–4.40.60.3–3.50.5
   Wooded tundra4.25.200.811.50000
   Tundra6.485.700.811.50000
   Tropical dry forest and savanna84.679.10.1–0.30.811.510.3–33.917.19.6–31.716.0
Tropical forest         
   Tropical rain forest17616400.459.30000
   Tropical seasonal forest105102.200.459.30000
   Warm mixed forest87620–0.30.459.314.6–121.426.010.4–86.518.5
Temperate and boreal forest         
   Temperate deciduous forest35580.10.456.19.6–13.711.315.9–22.718.7
   Cool mixed forest19490–0.20.456.10.1–8.70.30.4–22.40.7
   Cool conifer forest10.150.40.0–0.30.456.10.9–9.21.64.3–46.17.9
   Taiga22.688.700.456.10000
   Northern taiga6.511.500.456.10000
   Cold mixed forest5700.456.10000
   Cold deciduous forest8.722.60.010.456.10.20.20.60.6
   Northern cold deciduous forest2.610.40–0.10.456.10–0.40.00.1–1.70.1
   Xerophytic wood, shrublands55.446.30–0.10.656.15.0–11.06.94.2–9.25.8
Global C4     8–31149–3214
Global C3     34–1785239–20157
δ13CEi, ‰     −24.1 to –29.4−26.6−24.2 to −29.4−26.7
2.1.2.3. Effect on Wetland Emissions

[18] δ13CEi of wetland methane is determined by several factors, as discussed in section 2.2, but initially by the δ13C of the decomposing material. Using a compilation of CH4 emissions from different wetland types in PIH and LGM [Chappellaz et al., 1993], as well as the modeled C3–C4 distribution in grasslands by Collatz et al. [1998], we estimate the δ13CEi values of wetland CH4 (Table 5). We use δ13CEi = −62‰ for CH4 derived from C3 plants and −50‰ from C4 plants [Bréas et al., 2001]. We find that wetland CH4 is 13C-enriched by ∼2.8‰ at the LGM and by ∼0.4‰ in the PIH compared to modern values.

Table 5. Methane Emissions From Different Wetlands and Their δ13CEi in Dependence of C3 and C4 Vegetationa
 Emissions
LGM, Tg/aPIH, Tg/aPresent, Tg/a
TotalC3C4TotalC3C4TotalC3C4
Temperate and boreal         
   Boreal/conifer forest11 1717 1515 
   Tundra00 88 88 
   Temperate broad-leaved forest55 2020 99 
   Open conifer woodlands21.30.763.782.2264.381.62
   Total8715149238362
   δ13CEi−61.3‰  −61.8‰  −61.8‰  
Tropical         
   Tropical moist forest1616 3333 2727 
   Tropical scrub/woodlands149.14.92717.019.992518.256.75
   Grassland339.923.1215.4615.54219.6611.34
   Semidesert/steppe20.61.420.521.4820.921.08
   Desert30.92.120.521.4820.921.08
   Total683732855728775720
   δ13CEi−56.7‰  −58.3‰  −59.1‰  
 
Total764432136105311159322
Percentage 58%42% 77%23% 81%19%
δ13CEi−57.2‰  −59.6‰  −60.0‰  

[19] The total contribution of C4-derived CH4 in the PIH is only half that in the LGM (despite higher percentage of C4 plants in PIH grasslands), which explains the strong glacial-interglacial difference. δ13CEi from C3 dominated temperate and boreal wetlands is calculated as ∼−61.3‰ for the LGM and ∼−61.8‰ for both PIH and present (Table 5). For tropical wetlands (mixed C3 and C4), the estimated results are ∼−56.7‰ for the LGM, ∼−58.3‰ for PIH and ∼−59.1‰ at present.

2.1.2.4. Effect on Aerobic Methane Production

[20] The isotopic offset between AMP δ13CEi in living C3 versus C4 plants is ∼6‰, i.e., lower than between the respective plant tissues [Keppler et al., 2006]. Methane from leaf litter shows a higher offset (∼9‰), but is of minor importance for the global budget. Keppler et al. [2006] estimate the present-day δ13CEi of total AMP as ∼−50‰ (based on emission rates in various ecosystems, which are scaled to NPP, and an assumed C3: C4 ratio of 60: 40). It has since been argued that AMP emissions have to be scaled to either photosynthetic rates or leaf area in order to get a realistic estimate. We therefore recalculate the present-day δ13CEi of total AMP using scaling to the leaf area index (LAI) as described by Kirschbaum et al. [2006], as well as the ecosystem classification of François et al. [1998]. The revised estimate is ∼−48.9‰ (Table 6), i.e., ∼1‰ more 13C rich than the number calculated after Keppler et al. [2006]. The conversion between slightly different ecosystem classifications in the two studies introduces very little uncertainty. For example, a recalculation of the present-day value in the classification used by Keppler et al. [2006] and Kirschbaum et al. [2006] gives δ13CEi ∼−49.1‰, and is therefore indistinguishable from the value calculated with the François et al. [1998] classification. Estimates for the PIH and LGM are subject to more uncertainty because LAI for tropical forests [Harrison and Prentice, 2003] as well as season length and hours of sunshine, which control emission rates [Keppler et al., 2006; Kirschbaum et al., 2006], may have been different because of changes in cloud cover and latitudinal shifts of sources. For simplicity, we use the modern values throughout. We calculate a δ13CEi of ∼−48.5‰ for the PIH and ∼−48.6‰ for the LGM. These results suggest that a higher proportion of C4 precursor material enriches AMP in 13C by ∼0.4‰ in the PIH and by ∼0.3‰ at the LGM, compared to today.

Table 6. The δ13CEi of AMP in LGM, PIH, and Present in Dependence of C3 and C4 Vegetationa
 PresentPIHLGM
Area, 109 haEmission, Tg/aC3 Derived, Tg/aC4 Derived, Tg/aArea, 109 haEmission, Tg/aC3 Derived, Tg/aC4 Derived, Tg/aArea, 109 haEmission, Tg/aC3 Derived, Tg/aC4 Derived, Tg/a
Semidesert4.84.23.01.11.721.50.90.63.082.71.70.9
Grassland3.257.93.64.35.6613.83.610.24.4710.93.37.6
Needleleaf forest1.372.92.9 1.914.04.0 1.132.42.4 
Broadleaf evergreen forest1.7518.816.22.51.4715.812.82.90.9510.28.41.8
Broadleaf deciduous forest1.043.43.4 2.187.27.2 2.397.97.9 
Crops1.414.63.90.8        
             
 
Total13.641.733.18.712.9442.228.513.712.0234.023.610.3
Percentage  79.2%20.8%  67.6%32.4%  69.6%30.4%
δ13CEi −48.9   −48.5   −48.6  
2.1.2.5. Effect on Termite Emissions

[21] Tyler et al. [1988] compare the δ13CEi produced by termites in various habitats. They find no correlation between diet, specifically C3 and C4 plants, and produced δ13CEi, not even within a single species. This surprising result suggests that past C3–C4 changes do not influence δ13CEi of termite emissions. Considering that all other CH4 sources display such a dependence on plant species, we explore the potential variation in termite δ13CEi if there were a direct correlation to the precursor material. We use paleovegetation data from Prentice et al. [1993] and ecosystem-specific CH4 production rates from Zimmerman et al. [1982]. We calculate ∼−62.2‰ for the PIH and ∼−62.7‰ for the LGM, compared to the measured modern value of −63.0‰ (Table 3). We show below that such postulated changes have insignificant impact on δ13CE.

2.1.2.6. Effect on δ13CE

[22] The δ13CEi of marine and thermogenic sources is not affected by vegetation change at all, because they do not utilize terrestrial plant material as a precursor. Also, termite emissions have been shown not to reflect the isotopic signature of their precursor material [Tyler et al., 1988]. In contrast, we have demonstrated that the other terrestrial archaeal sources and biomass burning experience significant δ13CEi changes between LGM, PIH and present, because of C3–C4 shifts (Figure 2 and Tables 3 and 7). The largest estimated changes for individual source types occur from PIH to present in animal (∼2.6‰ 13C depletion) and wildfire emissions (∼2.1‰ 13C enrichment), while tropical wetlands and AMP become more 13C depleted by ∼0.8‰ and ∼0.4‰, respectively. From LGM to PIH boreal wetlands become moderately 13C depleted (∼0.5–0.6‰) and animal emissions become more 13C rich by a similar amount. Tropical wetlands become 13C depleted by ∼1.6‰. AMP becomes enriched by ∼0.1‰ and wildfires depleted by the same amount. These calculated results show that the δ13CEi of source types cannot be assumed to be constant over climatic transitions and that the same environmental changes cause 13C enrichment in some source types but depletion in others.

Figure 2.

Changes in δ13CEi of methane sources between LGM, PIH, and today. Estimates are derived from published emission rates (Table 3) and paleodata for individual sources. Negative values denote 13C depletion (top) in PIH relative to LGM and (bottom) for the present relative to PIH. Estimates for total emission changes (δ13CE) are given for scenarios with, and without, AMP emissions. See text for discussion of uncertainties.

Table 7. Influence of Individual Sources on the Isotopic Offset Between LGM and PIHa
 Change δ13CEi to LGM ValuesChange Source Contributions to LGM Values
δ13CEi Offset (LGM-PIH), ‰Δδ13CE, ‰Change in Relative Emission Rate, %Δδ13CE, ‰
  • a

    PIH isotope mass balance recalculated using LGM values for individual sources in order to evaluate the impact of reconstructed changes in source δ13CEi and relative emission rates on δ13CE, compared to the unaltered reconstructed PIH balance (Δδ13CE).

Budget Scenarios Including AMP Emissions
AMP−0.10.0−40.2
Tropical wetland1.60.460.3
Geologic00−80.9
Boreal wetland0.50.1+121.2
Termites00−1−0.1
Wild animals−0.60−4−0.2
Ocean00−2−0.1
Wildfires0.10−10.2
Clathrates00+20.2
All sources 0.4 2.5
Basic LGM scenario 2.6 2.6
     
Budget Scenarios Without AMP Emissions
Tropical wetland1.60.6−4−0.1
Geologic00−81.1
Boreal wetland0.50.1+181.3
Termites00−2−0.1
Wild animals−0.60−4−0.1
Ocean00−20
Wildfires0.10−10.3
Clathrates00+20.1
All sources 0.7 2.3
Basic LGM scenario 3.0 3.0

[23] We calculate the total effect of C3–C4–dependent changes according to equation (1), relative to scenarios with the same flux rates but unchanged, present-day δ13CEi values. We quote all estimates for a budget that includes AMP emissions, followed by the equivalent estimate in a non-AMP scenario in parentheses. We estimate that in the PIH changes in C3–C4 vegetation lead to moderate ∼0.4‰ (∼0.4‰) 13C enrichment of atmospheric δ13CH4. In the LGM, the enrichment is stronger at ∼0.6‰ (∼1.1‰).

[24] The reconstructed atmospheric δ13CH4 values depend on the emission scenarios (Figure 3). In addition, the calculated values carry a large degree of uncertainty because of poorly constrained data and assumptions. However, our main interest is to quantify the relative changes in δ13CE between LGM, PIH, and present. These are better constrained than the absolute values because errors resulting from various assumptions probably are consistent for all periods.

Figure 3.

Changes in emission rates of CH4 sources between LGM, PIH, and today. Emission rates are calculated after Chappellaz et al. [1993] (no uncertainty range reported) and other literature data (see text). Sources are listed according to total emission strength. Estimates are given for budgets with, and without, AMP emissions. (top) Between LGM and PIH, C3-derived sources increased strongly because of higher NPP especially in high latitudes. (bottom) Between PIH and modern conditions most terrestrial natural methane sources decrease because of land conversion.

[25] The most interesting finding is a lack of correlation between global abundance of C3 versus C4 vegetation and δ13CE. Between PIH and present, when C4 plants decrease drastically relative to C3 vegetation (−17%), our study predicts a fairly small impact on δ13CE, i.e., ∼−53.3‰ in the PIH versus ∼−52.5‰ at present (∼−55.9‰ versus ∼−55.6‰; here and in the following values in parentheses are for non-AMP scenarios) (Table 3 and Figure 2). In contrast, between LGM with δ13CE ∼−50.7‰ (∼−52.9‰) and PIH, when the C3–C4 distribution changes by only 4%, we calculate 13C depletion of ∼2.6‰, (∼3.0‰), which would strongly affect ice core δ13CH4 records. The depletion is unexpected, because the PIH environment hosts more13C-rich C4 plants [Collatz et al., 1998]. However, the contribution of C4-derived CH4 to the total source in the PIH is only ∼22% (∼20%), i.e., lower than during the LGM with 25% (28%) (Table 3). Therefore δ13CE does not merely reflect the global abundance of C3 and C4 plants, but is controlled by their distribution in specific, climate sensitive ecosystems with high methane emission rates.

[26] To identify the sources that cause the remarkable LGM-PIH difference, we recalculate the PIH mass balance using LGM values of δ13CEi for individual source types (Table 7). The results show that the change in δ13CEi of tropical wetlands accounts for ∼0.4‰ (∼0.6‰) of the total ∼2.6‰ (∼3.0‰) depletion, and boreal wetlands for ∼0.1‰ (∼0.1‰). The effects of all other sources (including postulated changes to termite δ13CEi) are negligible. In total, the glacial-interglacial changes in δ13CEi of source types due to C3–C4 shifts cause 13C depletion of δ13CE of only ∼0.4‰ (∼0.7‰) (Table 7).

[27] To explain the remainder of the LGM-PIH δ13CE offset that cannot be attributed to changes in δ13CEi of individual sources, we also recalculate the PIH mass balance with LGM emission rates (as percent of total emissions) for each source type (Table 7). Two sources have a strong impact, namely, the ∼8% (∼8%) relative decrease of geologic emissions that causes ∼0.9‰ (∼1.1‰) 13C depletion, compared to the total of ∼2.6‰ (∼3.0‰), as well as the ∼12% (∼18%) increase of boreal wetlands (including temperate regions) that causes ∼1.2‰ (∼1.3‰) 13C depletion. It is notable that a change of only 1% in relative wildfire emission rates causes ∼0.2‰ (∼0.3‰) depletion, underlining the strong impact this source with its extreme δ13CEi has on the isotope budget. Changing all source percentages accounts for ∼2.5‰ (∼2.3‰) of the total offset (Table 7). On one hand, these findings fit the conventional interpretation that changing emission rates determine atmospheric δ13CH4 (Figure 3). On the other hand, the impact of boreal wetland CH4, which is just one of several microbial sources, is larger than might be expected, because of the glacial shutdown of C3-derived CH4 from boreal wetlands. The subsequent strong increase of this source causes the percentage of C4-derived CH4 to drop in the PIH (Tables 3 and 5). Therefore the total contribution of C3 and C4 derived CH4 does play a significant role, but it is driven by changes in emission rates of specific sources (Figure 3). These findings suggest that wetlands with different vegetation types and in different latitudinal belts must be considered separately in isotope budgets in order to avoid significant errors.

[28] The uncertainty of the estimates given above is difficult to quantify. A major factor that affects the results is the choice of source budget. We have presented alternative estimates for budgets with and without an AMP source and see that the two differ substantially in the reconstructed changes of δ13CE between climatic changes. For example, the AMP budget results in a δ13CE difference between LGM and PIH of ∼2.6‰, i.e., 0.4‰ smaller than for the non-AMP budget (∼3.0‰). In a source scenario with AMP scaled to NPP instead of LAI, the LGM-PIH difference due to C3–C4 changes is only ∼1.7‰ (calculation not shown). In a source scenario that includes neither AMP nor geologic emissions the LGM-PIH difference is ∼1.9‰ [Schaefer, 2005]. In the latter case, it is noteworthy that the vegetation model of Kaplan et al. [2004] predicts a similar isotope offset between LGM and PIH for a comparable source budget. Despite the discrepancies, it should be noted that all four of these case studies predict 13C depletion between LGM and PIH as the result of C3–C4 changes at a magnitude that is measurable in ice cores.

[29] A second and more serious source for errors in reconstructed δ13CE is uncertainties in the calculated changes of δ13CEi for individual sources (Table 3). The calculations are based on the vegetation reconstructions by Chappellaz et al. [1993] and the modeled C3–C4 changes of Collatz et al. [1998]. Neither of these studies quantifies the associated errors. Consequently, we cannot present a formal error analysis for our derived values. Instead, we investigate the general uncertainty associated with vegetation models. Then we conduct a sensitivity study based on the findings.

[30] Collatz et al. [1998] use a very simple model to determine the predominance of C3 or C4 vegetation, which is purely based on the specific crossover temperature at which one of the pathways starts to outcompete the other. The area of C4 vegetation is either reported as global potential, i.e., all C4 favorable regions including those covered by woody vegetation, or as percentage of grasslands, which are assumed to cover 25% of global land area. A later study by the same group [Still et al., 2003] used a more sophisticated approach that accounts for croplands and multistoried vegetation. They arrive at a higher global surface area covered by C4 plants that would, applying the Collatz et al. [1998] calculation for C4 grassland area, result in a 24% increase of modern C4 grassland area compared to the original study. This would suggest a larger role of C4 vegetation in the global methane cycle and therefore a higher impact of vegetation changes between climatic stages. However, the increase is mostly due to improved C4 quantification in wooded areas, where the Still et al. [2003] model allows for mixed C3 and C4 vegetation. The two studies may in fact be quite comparable in predicted C4 grassland area, which are the basis for estimates presented here. In addition, the relative changes between climatic stages can be expected to be insensitive to the absolute values.

[31] Heimann et al. [1998] and Cramer et al. [1999] investigated the general range of uncertainty between different vegetation models. For the same input they found that modeled NPP deviated by ±11% (1 SD, six models) and ±19% (sixteen models), respectively. Furthermore, Hallgren and Pitman [2000] studied the sensitivity of a single vegetation model (BIOME 3) to the range of reported values for input parameters. They found that the photosynthesis parameterization, which is essential for determining C3 or C4 predominance, is particularly sensitive when the extremes of reported literature values were used instead of an initially adopted intermediate value. The difference between the generated vegetation maps is quantified by the global κ statistic, where 1 means perfect agreement and 0 that agreement is no better than random. For single parameters κ values reach up to 0.42. The average change in κ for 12 parameters is 0.85 and 0.89 for minimum and maximum literature values.

[32] Finally, the climatic boundary conditions influence the magnitude of the modeled vegetation change. Harrison and Prentice [2003] quantified the differences in model output of the BIOME 4 vegetation model for 17 climate model reconstructions of the LGM. They found that total grassland area in the Northern hemisphere and intertropics varied by ±10% (1 SD) (±15% for LGM climate simulations with modern CO2 levels). Modeled NPP in global forested areas varied only by ±2%. Comparisons of the vegetation maps derived from the model runs with paleovegetation proxies indicate that in general the models underestimate the importance of climate change [Harrison and Prentice, 2003].

[33] The uncertainty associated with modeled paleovegetation changes is on the order of ±10–20% for each of (1) vegetation algorithms, (2) knowledge of physiological parameters and (3) climatic forcing. Consequently, it is necessary to test the sensitivity of our reconstructions to large errors in atmospheric δ13CH4 introduced by the vegetation models. We have already presented calculations of δ13CE for our LGM and PIH scenarios with modern source δ13CEi values (Table 3). These constitute a sensitivity test for the C3–C4 reconstructions at an error margin of minus 100%, i.e., assuming that the reconstructed changes in δ13CEi are 100% too high and correcting for the whole difference to the original (modern) value. The effect on the calculated δ13CE value ranges from 0.4‰ in the PIH up to 1.1‰ at the LGM (without AMP emissions). Similarly, we have presented a minus 100% error estimate for the glacial-interglacial difference by recalculating the PIH budget with glacial δ13CEi values (Table 7) and found that the impact on δ13CE is 0.4‰ in an AMP scenario and 0.7‰ without AMP. Furthermore, to test the model sensitivity, we calculate the difference in δ13CE of the LGM assuming the glacial-interglacial δ13CEi difference of C3–C4–dependent sources varied by (1) ±25%, (2) ±50% and (3) ±100%. In the budget scenario including AMP the resulting changes in δ13CE are (1) ±0.05‰, (2) ±0.1‰, and (3) ±0.2‰, respectively. While such error margins for δ13CE may be negligible given the precision of ice core analyses, the errors in a non-AMP budget scenario are larger and amount to (1) ±0.15‰, (2) ±0.3‰ and (3) ±0.6‰, respectively.

2.2. Isotopic Fractionation During CH4 Production

[34] In contrast to wetland emissions, whose δ13CEi is subject to a complex variety of environmental influences, emissions from sources such as wildfires or intestinal tracts of ruminants come from special microenvironments, where atmospheric conditions do not influence their emitted δ13CEi. Other sources such as AMP are not associated with isotope effects [Keppler et al., 2006]. Wetlands are the largest natural CH4 source and are strongly climate-dependent [Kaplan, 2002; van Huissteden, 2004]. We therefore focus our study of changing isotopic fractionation under different climatic conditions on wetlands. In the following, we investigate the influence of temperature changes on fractionation coefficients, the dominance of methanogenic pathways, and the fraction of precursor material utilized. Although quantitatively many of these effects are poorly known, we show that the glacial wetland source is expected to be more 13C depleted with potentially measurable impact on atmospheric δ13CH4.

[35] In wetlands, CH4 is produced either by methyl fermentation (MF) or carbonate reduction (CR) [Whiticar et al., 1986]. Both are associated with KIE, which leads to 13C depletion of the produced CH4 relative to the precursor plant material. According to Blair et al. [1993], for the narrow range defined by glacial-interglacial changes, the temperature dependence of the fractionation coefficient αCR (defined as the ratio of the reaction rates for 12CH4 and 13CH4) associated with CR is

equation image

where T is absolute temperature. For the current global mean temperature of ∼15°C, equation (2) gives αCR = 1.060. In comparison, the LGM value is 1.062; assuming that mean terrestrial temperature is 7°C lower [Collatz et al., 1998]. Accordingly, CR derived CH4 would be 13C depleted by ∼1.8‰ at the LGM (assuming otherwise identical conditions). This is in general agreement with the findings of Whiticar [1996, 1999] and the CR temperature sensitivity reported by Botz et al. [1996], which results in slightly lower depletion of ∼1.5‰. The equilibrium curve for the offset between CO2 and CH4 predicts a larger depletion (∼2.6‰), as do empirical data from marine and freshwater sediments (∼2.8‰ and ∼6‰, respectively [Whiticar et al., 1986; Whiticar, 1999]). This may indicate that these empirical relationships between T and δ13CEi include not only the enzymatic temperature sensitivity but also factors like substrate availability or microbial community structure. This would make them a better analogy for the LGM scenario than the theoretical Tα correlation [Blair et al., 1993; Botz et al., 1996].

[36] We are not aware of studies that derive a temperature relationship for αMF, but Fey et al. [2004] observe strong 13C depletion in MF-derived CH4 at 10°C (−28‰) compared to 37°C (−21‰) in rice paddy soils. The authors attribute this to an increase in αMF and higher acetate availability (a major CH4 precursor in MF) in colder conditions. It is speculative to use these findings as indicative for (1) natural wetlands and (2) glacial-interglacial climate change, but taken at face value ∼2.9‰ 13C depletion at the LGM would be expected by analogy. Several additional influences must be taken into account. First, Fey et al. [2004] find that the ratio of CH4 produced by CR versus MF increases at lower temperatures. The higher fractionation during CR would lead to additional 13C depletion of wetland CH4 at the LGM. Second, the temperature sensitive KIE of microbial CH4 oxidation [Tyler et al., 1994] could lead to 13C enrichment of emitted wetland CH4 under LGM conditions. Third, substrate availability, the fraction of precursor material consumed and the structure of microbial communities also influence δ13CEi. The impact of the latter three factors on the produced δ13CEi may be captured by present-day field studies on temperature effects. However, a different correlation to temperature for past conditions cannot be ruled out.

[37] Bellisario et al. [1999] observe correlation between 13C enrichment and flux rates from wetlands in Manitoba and conclude that a higher proportion of MF at high-nutrient sites increases CH4 production. It has since been suggested [MacDonald et al., 2006] that boreal peatlands that developed since the LGM started as minerotrophic systems that emit 13C-rich CH4, thus contributing to the 13C enrichment observed in ice samples from the late glacial period [Schaefer et al., 2006]. The following evolution of the peatlands to ombrotrophic systems would have been associated with gradual 13C depletion of this source to modern values. The hypothesis remains speculative but highlights the problem that during periods of climate and vegetation change processes may occur that are not captured by field and modeling studies of the modern system.

[38] With current knowledge, CH4 production in wetlands is too complex to permit a robust estimate of past δ13CEi variability. Nevertheless, the evidence discussed above suggests a significant difference between glacial and interglacial wetland δ13CEi. For further quantification, we note that wetland production in the LGM occurs to ∼70% in the tropics and ∼30% in temperate zones [Chappellaz et al., 1993]. We assume temperature differences between LGM and today of 4°C in the tropics, 7°C in midlatitudes and 10°C in northern latitudes. Using the associated changes in α values for methanogenesis and oxidation, we estimate that LGM wetland δ13CEi is more 13C depleted than in the PIH by ∼0.3–1.1‰, depending on the MF derived fraction (fMF) relative to CR production. A good estimate may be ∼0.9‰ for fMF = 0.7. If fMF was lower, as suggested by findings of Fey et al. [2004], then wetland δ13CH4 depletion would be significantly higher, e.g., an additional ∼1.5‰ at fMF = 0.65.

[39] The effect of changing wetland δ13CEi on the methane budget can be calculated according to equation (1). Again, we are considering two global budget scenarios with and without AMP and compare these to the same emission budgets with unchanged wetland δ13CEi values. We find that temperature-dependent fractionation leads to very minor 13C depletion of δ13CE in the LGM of ∼0.1–0.2‰ for the AMP scenario and somewhat higher depletion (∼0.1–0.5‰) in the non-AMP scenario. With the additional contribution from CR the change would be ∼0.3–0.5‰ with AMP and ∼0.8–1.2‰ without AMP. This result suggests that glacial-interglacial temperature changes have an impact on the CH4 isotope mass balance that is detectable in ice core measurements.

2.3. Net to Gross CH4 Production Ratio

[40] In modern wetlands the net to gross (N/G) production ratio is ∼0.6, i.e., ∼40% of the generated CH4 is oxidized during transport [Walter and Heimann, 2000] while the remaining portion, which is emitted to the atmosphere, becomes enriched in 13C. However, oxidation rates (Q10 values) are about half as sensitive to temperature change as rates of methanogenesis [Segers, 1998]. Accordingly, we calculate that the N/G ratio in the LGM could have been ∼0.53. The Rayleigh equation then predicts ∼7‰ 13C enrichment of wetland CH4 compared to today. However, this was likely not the case because CH4 oxidation is much more strongly controlled by transport mechanism [Chanton, 2005], microbial community structure, previous CH4 exposure of soil layers and actual CH4 levels than it is by temperature [Segers, 1998]. Most importantly, no relative change of oxidation versus production was observed in a field experiment with warming plots [Moosavi and Crill, 1998]. This is consistent with the finding that CH4 production and consumption potentials do not control ecosystem CH4 flux [Bellisario et al., 1999]. We therefore do not consider an effect of changing N/G ratios on wetland δ13CEi. In contrast, temperature-dependent changes in αox partly counter those of CH4 formation (see above).

3. Sink Fractionation

[41] Methane removal from the atmosphere is commonly grouped in three sinks: abstraction with hydroxyl radicals (OH·) in the troposphere [e.g., Saueressig et al., 2001], soil uptake [e.g., Ridgwell et al., 1999], and stratospheric removal [e.g., Rice et al., 2003]. Recently, a fourth sink has been inferred from time series of atmospheric isotope measurements and attributed to removal of CH4 in the marine boundary layer (MBL) through reaction with chlorine [Allan et al., 2005]. Each sink removes 12CH4 preferentially and enriches the atmosphere in 13CH4. The combined isotopic fractionation of all sources, αWT, is calculated using the proportion of CH4 removed by each sink (F) and their experimentally determined α values:

equation image

with

equation image

It is useful to describe the isotopic offset (in ‰) between δ13CE and atmospheric δ13CH4 caused by αWT as the isotope separation factor ɛWT:

equation image

The individual values for F are not well constrained. They are commonly tuned to match the offset between measured atmospheric δ13CH4 and bottom-up source scenarios. The greatest uncertainty concerns the MBL sink that has been estimated to account for ∼5% of the total sink, i.e., FMBL = 0.05 [Allan et al., 2005]. However, because no direct evidence for the sink has been presented to date we will present alternative estimates for scenarios with and without an MBL sink. For the modern non-MBL estimate we adopt the values used by Sowers et al. [2005] with one modification. In order to compensate for lower stratospheric fractionation [Wang et al., 2002], we increase soil uptake by 5 Tg/a to a total of 38 Tg/a, consistent with the preferred estimate of Ridgwell et al. [1999]. The resulting values are FOH = 0.86, Fsoil = 0.076, and Fstrat = 0.07 (Figure 4). Accordingly, ɛWT indicates a 13C enrichment of 7.4‰, which is also the value used by Ferretti et al. [2005] and Sowers et al. [2005]. The MBL scenario includes a 5% contribution from this sink with proportionally lower percentages for the other removal processes, resulting in FOH = 0.817, Fsoil = 0.072, Fstrat = 0.067, and FMBL = 0.05. In this case, ɛWT indicates a 13C enrichment of 10.2‰.

Figure 4.

Relative contribution of sink processes from LGM to present. For scenarios without MBL sink, Fsoil and Fstrat remain as shown (i.e., the difference is negligible), while an increase in FOH compensates for the missing sink (note scale break).

[42] Changes in sink fractionation and magnitude in the past result from climate change and absence of anthropogenic influences. In the following, we present for each sink process estimates of F and α values. We can show that α values are unlikely to have changed significantly in the past, while the changes in F values had considerable impact on ɛWT. For reconstructions of the PIH we neglect climatic differences to the present. We try to quantify the effects of anthropogenic activity since the industrial revolution on atmospheric chemistry, vegetation, and soil properties and the impact on sink processes. Differences between LGM and PIH include climate, changes in ice-free land surface, and atmospheric chemistry. Climate models predict that the LGM was cooler by ∼4.3°C on the global average [Bush and Philander, 1999]. For terrestrial environments, we use an average temperature difference of 7°C to be consistent with the vegetation reconstructions of Collatz et al. [1998]. Reconstructions for the YD are complicated by the fact that there are few specific studies of the CH4 cycle and environmental parameters for this time. Here we assume that atmospheric methane concentration [CH4] is an indicator for conditions that prevailed in regions relevant for the CH4 cycle. Following this argument, environmental parameters were close to the intermediate values between PIH and LGM [Brook et al., 2000] and we calculate ɛWT for the YD accordingly.

3.1. Hydroxyl Sink

[43] The photochemical abstraction reaction with OH· is the primary sink of atmospheric CH4 and occurs mostly in the troposphere, where O(1D) and chlorine radicals (Cl·) contribute to a minor extent. The fractionation coefficient of the OH· reaction αOH has been measured as 1.0054 ± 0.0009, independent of temperature between 0° and 80°C [Cantrell et al., 1990]. Saueressig et al. [2001] determined αOH with higher precision to be 1.0039 ± 0.0004. The loss rate of the OH· sink depends on temperature and [CH4] and [OH·] [Fung et al., 1991]. The fact that these three variables changed in the past did not necessarily affect FOH. Instead, variations in [OH·] likely changed the atmospheric residence time of CH4, which has no effect on δ13CH4 under steady state conditions [Tans, 1997]. In this study, FOH is set to balance variations in the other sinks, relative to total emissions of a given steady state scenario given by Chappellaz et al. [1993]. This is justified by the lack of unequivocal evidence for changes in [OH·] [e.g., Staffelbach et al., 1991; Martinerie et al., 1995], but could be simplistic in light of recent indications for a substantial [OH·] increase between LGM and PIH [Valdes et al., 2005; Kaplan et al., 2006].

3.2. Soil Uptake

[44] Soil uptake is based on diffusive flux of CH4 into the soil and microbial consumption of CH4. The associated αsoil has been experimentally determined as 1.017 [Snover and Quay, 2000], 1.022 [King et al., 1989; Tyler et al., 1994] and 1.025 [Reeburgh et al., 1997]. Case studies show that land conversion, agriculture and forestry affect CH4 consumption of soils [Steudler et al., 1996; Del Grosso et al., 2000]. On a global scale, land use change has decreased soil uptake by ∼10% since the PIH [Ridgwell et al., 1999]. Soil uptake of CH4 is controlled by diffusion [Striegl, 1993; Tyler et al., 1994] and microbial oxidation rates [Ridgwell et al., 1999]. Both processes are sensitive to [CH4] and the response of oxidation is nonlinear below soil concentrations of 300 ppbv [Ridgwell et al., 1999]. While this barely affects present soil uptake, the impact increases with lower atmospheric [CH4]. Using a process-based model, soil uptake in the PIH has been calculated as ∼14 Tg/a, or 37% of the modern estimate [Kaplan, 2002].

[45] For the LGM we calculate a higher αsoil of 1.0272, according to the experimentally determined temperature dependence by Tyler et al. [1994]. The soil sink magnitude decreased because of (1) lower temperature (15% for a temperature drop of 7°C [Yonemura and Yokozawa, 2000]), (2) generally drier glacial conditions [Yonemura and Yokozawa, 2000] and (3) changes in ice-free surface area available for CH4 uptake. The major restriction of the LGM soil sink is that [CH4] is only 50 ppbv higher than the linearity threshold for microbial kinetics [Brook et al., 2000; Ridgwell et al., 1999], so that most microbial communities are starved for CH4. For this study we adopt the modeled LGM soil sink of ∼0.6 Tg/a by Kaplan [2002].

[46] For the YD we infer the magnitude of the YD soil sink by scaling linearly to the findings of Kaplan [2002] for LGM and PIH, respectively. This gives a very conservative range from 0.9 to 9.3 Tg/a. Our preferred value of 6 Tg/a is derived from a power function fit for a crossplot of sink estimates [Kaplan, 2002] and [CH4] [Brook et al., 2000] from LGM to present. We calculate αsoil = 1.0256 for a temperature decrease of 3.5°C relative to the PIH [Tyler et al., 1994].

3.3. Stratospheric Sink

[47] Stratospheric removal occurs through a combination of OH·, O(1D) and Cl· and at present leads to tropospheric 13C enrichment of CH4 by 0.5–0.6‰ [McCarthy et al., 2001]. At stratospheric temperatures around 225 K, αCl is between 1.073 [Saueressig et al., 1995] and 1.070 [Tyler et al., 2000] and αO(1D) is 1.013 [Saueressig et al., 2001]. The fraction of CH4 removed by each reaction type, and therefore αstrat, varies with altitude and latitude. At higher elevation with lower [CH4] the higher fractionating sinks become more important. The observed correlation results in apparent αstrat between 1.0108 and 1.0204 [Rice et al., 2003]. Within this range we adopt a value of 1.0133 that matches the magnitude of the stratospheric sink as modeled by Wang et al. [2002].

[48] Since the PIH, anthropogenic emissions of chlorocarbons increased [Cl·] strongly. Wang et al. [2002] model the effect on tropospheric δ13CH4 and find 13C depletion of ∼0.4‰ for preindustrial conditions. We include this result by simply lowering αstrat for the PIH to 1.0082 although the variation in CH4 removal largely occurred in the troposphere. However, the impact on ɛWT is captured accurately. The stratosphere in the PIH is up to 11°C warmer than today because of heat absorption caused by changes in photochemistry [Martinerie et al., 1995]. This means a significant increase in reaction rate [Michelsen and Simpson, 2001] but does not necessarily translate into a higher sink magnitude, e.g., if Cl availability becomes limiting. The temperature increase would reduce αCl by 0.001 [Tyler et al., 2000], which has a negligible effect on ɛWT. No δ13CH4 changes caused by variations in the O(1D) reaction with CH4 are considered throughout this study because its total modern effect on δ13CH4 is only ∼0.1‰ [McCarthy et al., 2001] and any changes to this value can be neglected. We do not account for feedbacks in atmospheric chemistry, such as the depression of [Cl·] through CH4 [Rice et al., 2003], because even an unrealistic complete shutdown of the Cl sink would lower δ13CH4 only by ∼0.3‰. The magnitude of the stratospheric sink is scaled linearly to the total sink, which includes several assumptions such as an equal impact of [CH4] on all sinks in troposphere and stratosphere.

[49] There are no available data concerning changes in [Cl·] between LGM and PIH. Varying production of the precursor methyl chloride in ocean environments and biomass burning [Wayne, 1991], as well as higher ocean-atmosphere exchange at different wind patterns could have affected the natural source of Cl. However, these processes are hard to quantify, therefore we follow Martinerie et al. [1995] and assume no changes in [Cl·] between PIH and LGM. The total effect of the preindustrial stratospheric sink on tropospheric δ13CH4 amounts only to ∼0.3‰ so that the following processes can be neglected. First, a 6°C higher stratospheric temperature in LGM versus PIH [Martinerie et al., 1995] with reduction of αCl by 0.0005 [Tyler et al., 2000]. Second, a change in reaction rates with opposite sign in a colder troposphere and warmer stratosphere. Third, potential lowering of the apparent αstrat with increasing transport. Fourth, dependence of αstrat on [CH4] [Rice et al., 2003]. The only change we make for the LGM and YD is to adjust the magnitude of the stratospheric sink proportionally to the total sink rate.

3.4. Chlorine Sink in the Marine Boundary Layer

[50] Recent studies suggest that at present ɛWT is ∼2‰ higher than commonly assumed [Allan et al., 2005]. The likely cause is the reaction of CH4 and Cl· in the marine boundary layer (MBL). The amount of CH4 removal seems to vary interannually in an estimated range between 13 and 38 Tg/a. Enhanced ɛWT due to 13C enrichment in the MBL could explain discrepancies between atmospheric δ13CH4 [Ferretti et al., 2005] and bottom-up budgets for the PIH [Chappellaz et al., 1993]. For reconstructions of the MBL sink we assume that there is no anthropogenic influence and hence no difference between present day and PIH other than linear scaling to [CH4]. Potential climate induced changes of the MBL source are difficult to assess. Cl formation by sea spray [Mayewski et al., 1997] and solar radiation on the southern ocean [Berger, 1978], both of which increase photolytic Cl· formation, are higher at the LGM. There is also evidence for locally, but not necessarily globally, increased marine productivity that provides organic Cl precursors [Bender, 2003]. Proxies for formation of dimethylsulfide (also involved in Cl· formation) show opposite trends in records from Greenland and Antarctica [Legrand et al., 1991; Saltzman et al., 1997]. Currently, there is not enough evidence from ice cores or other sources to quantify past [Cl·] and consequently we do not consider climate related changes of the MBL sink magnitude, but scale the sink magnitude proportionally to [CH4]. We adjust its αCl for lower T in YD and LGM with negligible impact on ɛWT.

3.5. Resulting Changes in Total Sink Fractionation

[51] We present two estimates for ɛWT in each climatic period because of uncertainties regarding the MBL sink. First, we list estimates that do not include an MBL sink. According to the information above, the fractions of the sink processes in the PIH were FOH ∼ −0.857, Fsoil ∼ −0.073, and Fstrat ∼ −0.070 (Figure 4). Using the same αOH and αsoil for the PIH as at present, but lower αstrat to account for anthropogenic changes in [Cl·], we calculate ɛWT ∼ −7.0‰ (Figure 5). The ∼0.4‰ decrease relative to modern conditions is mostly due to the lower Cl· sink, whereas reduced soil uptake has little impact. For the LGM we calculate FOH ∼ −0.925, Fsoil ∼ −0.005 and Fstrat ∼ −0.070. αOH and αstrat are the same as in the PIH, while αsoil is ∼1.0272, because of lower temperature. ɛWT is then estimated at ∼5.7‰. The significant difference to PIH and present is caused by the almost complete shutdown of soil uptake at low [CH4]. For the YD we find a preferred solution (based on our best estimate of soil uptake) of FOH ∼ −0.892, Fsoil ∼ −0.038, and Fstrat ∼ −0.070. αsoil is calculated as ∼1.0256, while αOH and αstrat remain unchanged. The resulting ɛWT is ∼6.4‰.

Figure 5.

(left scale) Sink fractionation coefficients and (right scale) ɛWT from LGM to present. See text for discussion of uncertainties.

[52] Including the MBL sink we estimate that it contributed ∼5% to the total source in all studied periods with proportionally lower F values for the other sinks. For α values as discussed above, the resulting ɛWT values are ∼10.2‰ (present), ∼9.9‰ (PIH), ∼9.3‰ (YD) and ∼8.6‰ (LGM). It is worth noting that the differences between the respective time periods are almost identical in the MBL and non-MBL scenarios.

[53] The reconstructions for ɛWT carry a large degree of uncertainty because they rely on various assumptions. The fractionation coefficients of individual sinks are reasonably well known, whereas their magnitudes are poorly constrained. For a comparison we use three scenarios with different sink estimates for the PIH from Houweling et al. [2000] (without early anthropogenic emissions considered in that study) and derive the associated F values. Together with our estimates for α values we calculate ɛWT values between ∼6.5 and ∼8.1‰. The “Upper Limit” scenario of Houweling et al. [2000] that produces a higher ɛWT than today requires an unrealistically large soil sink. The most likely scenario results in total sink fractionation of ∼6.6‰, supporting that ɛWT was lower in the PIH than today. The difference to the PIH value derived above (∼0.4‰) is similar to analytical precision for ice core δ13CH4 measurements.

[54] We also test the sensitivity of ɛWT to the magnitude of individual sinks in the YD scenario. Varying the MBL sink by ±50%, which is equivalent to presently observed interannual variations [Allan et al., 2005], changes ɛWT by ±1.5‰. Without better knowledge of the MBL sink ɛWT will remain poorly constrained. However, relative changes in ɛWT between different stages are expected to be less sensitive to these uncertainties if the MBL sink varies proportionately to [CH4]. The soil sink magnitude for the YD is uncertain within 1 order of magnitude and the associated changes in sink fractionation are −0.7‰ to +0.4‰. Uncertainties in the magnitude of the stratospheric sink can be neglected, as they influence ɛWT by less than ±0.3‰ for a ±50% magnitude change.

[55] The major controls of past variations in ɛWT are the strong increase in αstrat from the PIH to present (Figure 5) and the magnitude of the soil sink (Figure 4). The stratospheric sink and its anthropogenic contribution is fairly well constrained by measurements [Rice et al., 2003] and atmospheric chemistry models [Wang et al., 2002]. Any natural, climate related changes of this sink are small by comparison and likely had a very minor impact on ɛWT. The soil sink is very sensitive to low [CH4] in glacial times. This nonlinear response is the main control of ɛWT in preindustrial times and must be quantified using process-based models of soil uptake. The main unknown for the reconstruction of CH4 sinks in the past is a possible Cl sink in the MBL. This process is inferred rather than directly measured for the modern system and seems to undergo significant short-term fluctuations [Allan et al., 2005]. This fact, together with the potential for change induced by meteorology and biology in a glacial environment, does not currently allow for reliable reconstructions of the MBL sink.

[56] The combined changes lead to ɛWT for the LGM that is estimated to be ∼1.7‰ lower that at present, which must be accounted for in budget reconstructions.

4. Conclusions

[57] The isotopic signatures of sources and sink fractionation change between modern and preindustrial times and over glacial cycles. Changes in the atmospheric CO2 budget and shifting patterns of C3 and C4 vegetation between LGM, PIH, and present significantly change the δ13C of the CH4 precursor material of several source types. C3–C4–dependent sources are generally more 13C rich in earlier time periods, as are total emissions. The contribution of these changes to glacial-interglacial shifts in atmospheric δ13CH4 is minor. The magnitude of the change is specific for the emission budget and must be assessed in an iterative process when methane budgets are derived from δ13CH4 paleodata.

[58] Our results regarding C3–C4 shifts generally support the assumption that emission rates, rather than δ13CH4 of individual sources, are the main control for the stable carbon isotope ratio of atmospheric CH4. It is important, however, that source scenarios have a high degree of sophistication as shown by the following example. With our presented bottom-up model we calculate ∼2.6–3.0‰ 13C depletion between LGM and PIH, which is largely caused by changing emissions in different wetland systems. In a top-down budget that distinguishes only a microbial, a thermogenic and a pyrogenic source, such a change must be interpreted as a strong relative decrease of geologic and wildfire emissions, overestimating their actual decline in our scenario at least threefold. This demonstrates that these simplistic budgets cannot describe variations in δ13CH4 adequately. Even a generic wetland source with uniform δ13CH4 for all latitudes would not allow for resolution of the changes, so that latitudinal wetland belts must be identified in budget scenarios.

[59] The potential impact of temperature on pathways of wetland CH4 formation, their fractionation coefficients and consequently on atmospheric δ13CH4 is difficult to quantify, but is likely large enough to affect ice core records between glacial and interglacial stages and must be accounted for in their interpretation. In contrast, field studies suggest that temperature induced change of the net to gross CH4 production ratio in wetlands with large impact on δ13CH4 is unlikely.

[60] This study also shows that total sink fractionation changes significantly in the past. The difference between the PIH and the present likely results in modest 13C depletion of atmospheric CH4 due to anthropogenic Cl emissions. For time periods with low [CH4], like the YD and LGM, ɛWT is smaller because the soil sink is decreased. This effect is large enough to affect ice core δ13CH4. Differences in ɛWT between the various time periods are the same whether a MBL sink is considered or not.

[61] The individual and cumulative changes in atmospheric δ13CH4 associated with various steps in the CH4 cycle are shown in Figure 6. 13C enrichment of precursor material and higher 13C enrichment during methanogenesis and CH4 loss partly cancel each other. The total estimated effect of environmental changes in the YD (∼0.4‰ 13C depletion) is of similar magnitude as current analytical precision of ice core δ13CH4 measurements [Ferretti et al., 2005]. The total change that we derive for the PIH (∼0.8‰ 13C enrichment) and LGM (∼1–1.2‰ 13C depletion), however, would be detectable in ice records. Our results suggest that the atmospheric 13C enrichment observed in the late PIH is mostly caused by reorganization of the source-sink configuration while the environmental changes studied here make only a minor contribution. For full glacial conditions a substantial 13C depletion of atmospheric CH4 is expected, which is in contrast with the first available ice core δ13CH4 data [Sowers, 2006; V. V. Petrenko, unpublished data, 2006]. The latter show 13C enrichment by several permil relative to today, suggesting that the processes leading to depletion have either been overestimated in this study or are overcompensated by changes in relative emission rates. The conventional approach to interpret changes in δ13CH4 as variations in source flux or atmospheric removal is largely justified, but budget reconstructions will be less accurate if changes in isotopic signatures of sources and sink fractionation are neglected.

Figure 6.

Changes of δ13CH4 associated with various steps of the CH4 cycle. All reconstructed values are relative to present-day numbers and represent best estimates. Estimates are given for each time period and budget scenarios with, and without, AMP emissions. Areas that are not hatched indicate the range between low and high estimates where applicable. MF stands for methyl fermentation; CR stands for carbonate reduction. See text for discussion of uncertainties.

Acknowledgments

[62] We thank Ed Brook for helpful discussions and Todd Sowers, as well as an anonymous reviewer, for valuable comments that strongly improved the manuscript. This work was supported by a fellowship from DAAD (German Academic Exchange Service) (H.S.), Petroleum Research Fund of the American Chemical Society (H.S.), Canadian Foundation for Climate and Atmospheric Sciences CFCAS MAMMOTH Grant (M.J.W.), and NSERC Discovery Grant (M.J.W.).

Ancillary