Balancing the global methane budget: constraints imposed by isotopes and anthropogenic emission inventories

Abstract

The doubling of atmospheric methane (CH₄) during the twentieth century due largely to growth in anthropogenic emissions has made CH₄ the second largest contributor behind carbon dioxide to anthropogenic forcing of climate change. However, the global CH₄ budget and its decadal evolution remain poorly quantified despite re-evaluations that include the IPCC Fourth Assessment Report. Potentially, the aggregation of national anthropogenic emission inventories as reported to the United Nations Framework Convention on Climate Change could document the changing anthropogenic emission since 1990, as could other “bottom–up” inventories such as EDGAR. As an examination of the recent CH₄ budget evolution, we compare two constructions of CH₄ source history, one based on an aggregation of national emission inventories, the other version 4 of EDGAR, each in combination with alternative natural CH₄ emissions, for consistency with observed atmospheric mixing ratio and carbon isotope content (δ¹³C(CH₄)). We conclude that despite the utility of isotopic constraints on budget evolution, the level of uncertainty in sink strengths and their isotopic fractionation limits the confidence in constructing anthropogenic emission histories over recent decades.

Keywords:

• methane
• carbon-13
• national emission inventories
• global budget

1. Introduction

The global methane (CH₄) budget remains ill-quantified, notwithstanding several decades of research. According to the recent IPCC Assessment hereafter abbreviated AR4 (Denman et al. 2007), the global strengths of the various processes that remove tropospheric CH₄ are uncertain to within about 20%, which confers a similar uncertainty on the “top–down” estimate of global CH₄ emission. However, the emission contribution from individual source categories such as wetlands, coal-mining, etc, is in general poorly determined, so that source partitioning (i.e. the global source inventory) is quite poorly constrained. This has enabled significant new source categories to be postulated or discovered and still be consistent with the top–down aggregate.

A case in point is the plant source of CH₄ which its discoverers assessed to be as large as 236 Tg yr⁻¹ (Keppler et al. 2006), fully 40% of the global source. However, further research has revised that upper limit downward markedly (e.g. Kirschbaum et al. 2006; Parsons et al. 2006), and led some to question that plants produce CH₄ at all other than as a degradation product (e.g. Beerling et al. 2008; Nisbet et al. 2009). We here ignore “plant CH₄” as a significant CH₄ source that is independent of that of its host ecosystem.

Another case in point is hitherto-ignored terrestrial seepages of CH₄ from geologic reservoirs which Etiope et al. (2008) assess at 53 ± 15 Tg yr⁻¹.

Stable isotope information, especially ¹³CH₄, can be used to constrain the budget. Primarily, these constraints arise because the “isotope signature”, δ¹³C (defined below), of an emission depends strongly on how CH₄ production is mediated. Specifically, microbially mediated (“biogenic”) CH₄ is relatively depleted in ¹³CH₄ (δ¹³C ≈ −60‰), CH₄ as a byproduct of biomass combustion (“pyrogenic CH₄”) is relatively enriched (δ¹³C ≈ −25‰ or −12‰ for C3 or C4 biomass), while “fossil CH₄” has an intermediate signature (δ¹³C ≈ −40‰). While the contemporary signature in atmospheric CH₄ is well known at about −47‰ through measurements at various sites (e.g. Quay et al. 1999; Tyler et al. 2007; Lassey et al. 2010) and its history can be reconstructed from measurements in air extracted from polar ice (e.g. Ferretti et al. 2005), the isotope fractionation associated with the various sink processes and the relative roles of those processes confer some uncertainty. In addition, the minor radio-isotopologue ¹⁴CH₄ of which fossil CH₄ is uniquely devoid has been used to assess the “fossil fraction” in the CH₄ source (Quay et al. 1999; Lassey et al. 2007b).

The temporal pattern in the global source history can arguably be constructed with at least as much confidence as the budget at a time snapshot. Lassey et al. (2007a) have demonstrated that this pattern together with its companion pattern in δ¹³C can be used to further constrain the budget history. This constraint pre-supposes that any systematic change in isotope fractionation of the combined (first-order) sink is not the principal cause of systematic variation in atmospheric δ¹³C. Applying mass balance to postulated CH₄ and δ¹³C source histories yields a simulated history of atmospheric δ¹³C to within an additive offset. One can then examine for reasonableness both the match between the simulated and historical patterns of atmospheric δ¹³C and the offset between them. Lassey et al. (2007a) argued that the offset should be consistent with a fractionation of magnitude 7.7 ± 1.4‰ (95% confidence interval), revised here to 7.6 ± 1.4‰ (Table 1). This approach has been used to address the reconciliation of constructed source inventories with the atmospheric record (Lassey et al. 2005, 2007a).

Table 1. Representative construction of the global contemporary methane sink and isotope fractionation.^a

Sink	Strength, Tg yr^−1	εsink, ‰
Tropospheric OH	511 ± 103^a	−4.65 ± 0.75^b
Soil	30 ± 15^a	−20 ± 2^b
Stratosphere	40 ± 8^a	−3 ± 3^c
Atomic chlorine	25 ± 12^d	−60 ± 1^b
Total	606 ± 105^e	−7.6 ± 1.4^e
^aSink strengths and 95% confidence intervals for all sinks except chlorine are taken from the IPCC Fourth Assessment Report (Denman et al. 2007). All sink strengths scale with mixing ratio, but are used in the present work only for their role in estimating fractionation εtotal and its uncertainty.
^bThe kinetic isotope shifts, εsink, are taken from the literature, with confidence interval reflecting the spread in measured values: εOH from Cantrell et al. (1990) and Saueressig et al. (2001); εsoil from Tyler et al. (1994) and Snover and Quay (2000); εchlorine from Crowley et al. (1999) and Tyler et al. (2000).
^cThe stratospheric sink is a net cross-tropopause flux with corresponding δ^13C estimated from the δ^13C contrast in the two gross fluxes, themselves poorly determined: see Lassey et al. (2007a) for detail.
^dThe chlorine sink strength is taken from Allan et al. (2007) who cite 25 Tg yr^−1 as a mid-point between bounds assessed at 13 and 37 Tg yr^−1 in the contemporary troposphere.
^eThe confidence intervals for the total sink and corresponding kinetic isotope shift are calculated with the assumption that all individual confidence intervals are uncorrelated.

The aim here is to extend earlier work (Lassey et al. 2005) to examine the compatibility of globally-aggregated “bottom–up” anthropogenic emission inventories with the global CH₄ cycle, constrained by δ¹³C data. Two such aggregations are examined, each in tandem with two alternative constructions of the natural emission inventory with contrasting δ¹³C. Further constraints by ¹⁴CH₄ data are included in the formalism but their implementation is outside the present scope.

2 Modelling strategy

With the modelling strategy described in detail elsewhere (Lassey et al. 2007a), only a brief summary is presented here.

Both the global CH₄ budget and that of each isotopologue separately (viz, ¹²CH₄, ¹³CH₄, ¹⁴CH₄, identified by subscripts “12”, “13”, “14”) satisfy global mass balance:

where C(t) and S(t) are the tropospheric burden and global source in consistent units (mass or molar) and λ(t) the combined removal rate (inverse tropospheric turnover time) at time t. For ¹⁴CH₄, λ(t) is augmented by the radioactive decay rate λ _R  = (8267 yr)⁻¹. The tropospheric CH₄ burden is taken as proportional to the mean surface mixing ratio, 2.767 Tg ppb⁻¹ after Lassey et al. (2007a).

A specification of any two of the time series C(t), S(t) and λ(t) plus an initial state over a time interval fully defines the budget history over that interval. An initial steady state is prescribed for 1700AD prior to the near-monotonic growth in atmospheric CH₄ (e.g. Ferretti et al. 2005). With this long spin-up time, simulations of recent decades when national emission inventories are available are essentially independent of initial state, even allowing for the slowness of δ¹³C(CH₄) to respond to source perturbations (Tans 1997; Lassey et al. 2000). With largely aseasonal input to the model, integration from 1700 is discretised into annual time steps. Indeed, processes within polar ice and firn limit temporal resolution available to several years. The discretised Equation (1) can be solved for C(t) in forward mode or for either S(t) or λ(t) in inverse mode, and we adopt a combination of these to apply isotopic mass balance as summarised in the following subsections.

2.1 Total methane

Of the three time series in Equation (1), C(t) has the highest quality through mixing ratio determinations in air extracted from polar ice (MacFarling Meure et al. 2006) and, since the 1980s, through direct atmospheric monitoring (e.g. Dlugokencky et al. 2009). We apply the MacFarling Meure (2006) dataset for C(t), and construct time series of the global CH₄ source to define S(t) (Section 3). The history of removal rate λ(t) is then “inverse modelled” through mass balance, Equation (1).

2.2 The stable isotopologue ¹³CH₄

With time series C(t), S(t) and λ(t) all defined from 1700, mass balance can be addressed separately for each isotopologue through “forward modelling”. By prescribing a time-independent sink fractionation ε ≡ λ₁₃/λ₁₂−1 both λ₁₃(t) and λ₁₂(t) follow immediately as time series proportional to λ(t). Further assigning δ¹³C to each source component allows S ₁₃(t) and S ₁₂(t) to be constructed. Equation (1) is then solved separately for C ₁₃(t) and C ₁₂(t) whose sum reproduces C(t) but whose ratio allows the evolving tropospheric δ¹³C to be constructed. Here, δ¹³C follows the usual definition:

where R and R _std are the respective ¹³C/¹²C isotope ratios in the CH₄ sample and carbonate standard (Vienna peedee belemnite, VPDB). The effect of ε is to offset the atmospheric δ¹³C pattern relative to the source δ¹³C pattern. The offset that minimises the mismatch between the atmospheric δ¹³C pattern and the measurement record can then be compared with ε_total in Table 1 for reasonableness.

2.3 The radio-isotopologue ¹⁴CH₄

The ¹⁴CH₄ cycle has been studied for its value in constraining the “fossil fraction” in the global contemporary CH₄ source (e.g. Quay et al. 1999). This constraint, which is predicated on fossil CH₄ sources being devoid of ¹⁴C, is severely weakened by the need to construct two time series that are poorly guided by data: the ¹⁴C content in non-fossil CH₄ sources (modified by “bomb carbon”); and the direct atmospheric release of ¹⁴CH₄ by the commercial nuclear industry. Lassey et al. (2007a) proposed an approach to minimise the impact of these weaknesses, while Lassey et al. (2007b) applied a regression analysis that, with certain assumptions, simultaneously constrained both the fossil fraction and strength of the nuclear-industry source. While the approach of Lassey et al. (2007a) can be applied to the present study, it is outside the scope and constraints of this article.

3 Source construction

A natural and an anthropogenic global source component are combined and extrapolated back in annual time steps to a postulated steady state in 1700. These are selected from two alternative natural source inventories (Section 3.1), and two alternative anthropogenic histories (Section 3.2).

3.1 The natural source

One of two alternative natural source inventories, “Nat1” or “Nat2” of Table 2, is applied to every year since 1700. While there will have been systematic emission variations over time, such variations are poorly documented and would have been dwarfed by the growth in anthropogenic emissions during the twentieth century, and likely even by the uncertainty in that growth.

Table 2. Natural global source scenarios, Nat1 and Nat2.

Source category	Nat1^a, Tg yr^−1	Nat2^b, Tg yr^−1	δ^13C^c, ‰
Wetlands	163 (83–240)	129 (83–240)	−60
Termites	20 (10–30)	20 (10–30)	−57
Wild animals	15	15	−62
Geologic (incl. oceans)	19 (9–29)	53 (42–64)	−45
Boreal wildfires	5	5	−25
Total natural	222	222
Fossil fraction^d	8.6%	23.9%
δ^13C in total	−57.8‰	−55.5‰
^aIndividual source strengths and confidence intervals are taken from Houweling et al. (2000), who also estimate the total source on the basis of the assessed pre-industrial sink (“top–down” estimate).
^bNat2 is adjusted from Nat1 by substantially increasing the geologic source and its confidence limits after Etiope et al. (2008) at the expense of some wetland emission, maintaining the same total source.
^cThe δ^13C assignments are from Houweling et al. (2000) except that the estimate by Etiope et al. (2009) is preferred for the geologic source.
^dThe “fossil fraction” is the proportion of geologic CH4, which is distinguished by having negligible ^14CH4.

Scenario Nat1 is the natural source previously adopted by Lassey et al. (2005, 2007a) and taken from Houweling et al. (2000), aside from a small change in the δ¹³C assignment for geologic sources after Etiope et al. (2009), −40‰ to −45‰, that lightens the total natural source by 0.3‰. Nat2 embraces a reappraised geologic CH₄ source that takes account of previously-overlooked terrestrial CH₄ seeps (Etiope et al. 2008). We elect to retain the same total source as for Nat1, 222 Tg yr⁻¹, in order that simulation differences can be related to isotopic contrasts between Nat1 and Nat2 without ambiguity from source-strength differences. The geologic source in Nat2 is therefore enlarged at the partial expense of wetland emission which nevertheless still lies within accepted bounds and is similar to that recently proposed by Kaplan et al. (2006) of ∼110 Tg yr⁻¹.

3.2 The anthropogenic source

Two globally-aggregated anthropogenic source histories are denoted “EPA06” and “Edgar4”.

EPA06 (USEPA 2006) comprises a consistent and comprehensive time series of CH₄ emission based on emission inventories compiled and submitted to UNFCCC by over 90 individual countries. Careful scrutiny has identified data gaps, including missing countries, and filled them from authoritative reports or by applying IPCC estimation methodologies to assure completeness and consistency. While inventory compilation may be perceived as lacking validation and scientific rigour, it can nonetheless accommodate local knowledge that uncritical use of global data could miss. Inventories are constructed for every 5th year, 1990–2020, of which 2005–2020 are projections based on established mitigation programmes, sector agreements, or measures already in place, but excluding planned mitigation activities that have yet to be implemented. Figure 1 illustrates the dataset for eight country groupings, including projections beyond 2005 that are not considered here. Table 3 illustrates EPA06 for 2000 grouped by biogenic, pyrogenic and fossil components.

Figure 1. EPA06 dataset of anthropogenic CH₄ emissions by eight country groupings at 5-yearly increments from 1990 projected to 2020, along with the world total shrunk by a factor of 5.

Table 3. Anthropogenic global source EPA06 for 2000.^a

Source component	Strength, Tg yr^−1	δ^13C, ‰
Biogenic sources
Ruminants	86	−62
Rice agriculture	30	−64
Landfills	35	−55
Waste water	25	−55
Animal waste	11	−55
Total biogenic	187	−60.1
Pyrogenic sources
C3 biomass	18	−25
C4 biomass	12	−12
Total pyrogenic	30	−19.8
Fossil sources
Coal mining	18	−35
Oil, gas, industry	52	−40
Total fossil	70	−38.7
Total source	287	−50.7
Fossil fraction	24.3%
^aSources: USEPA (2006) for source strengths, Houweling et al. (2000) for δ^13C values.

Edgar4 is the most recent release of EDGAR (“Emission Database for Global Atmospheric Research”) which is available and documented at http://edgar.jrc.ec.europa.eu/ (J.G.J. Olivier, personal communication). Annual emissions for 1970–2005 are reported, and display inter-annual variations that make it less smooth than EPA06. Since C(t) is inherently smoothed, these variations propogate into a removal rate λ(t) (Section 2.1) with inter-annual variations that are at least partially spurious.

Model integration requires that the source be extrapolated in annual steps back to the initial state in 1700. The detailed realism of the entire extrapolation is less important than the requirement of a plausible transition through the mid-late twentieth century. We follow Lassey et al. (2005, 2007a) by applying Edgar-Hyde (E-H) v1.4, an emission time series for 1890–1995. E-H v1.4, available from http://www.mnp.nl/edgar/, updates v1.3 (van Aardenne et al. 2001) to match EDGAR v3.2 (Olivier and Berdowski 2001). EPA06 and Edgar4 are each extrapolated back to 1890 by scaling the biogenic, fossil and pyrogenic components of E-H v1.4 to match those of EPA06 in 1990 and of Edgar4 in 1970. Each scaled dataset is extrapolated linearly from 1890 back to 30 Tg yr⁻¹ assumed as part of a steady source of 252 Tg yr⁻¹ for 1700 (Houweling et al. 2000; Lassey et al. 2007a).

For application herein, each source history is interpolated onto an annualised series, which ensures a reasonably smoothly changing time series other than for Edgar4 after 1970.

4 Results

Figure 2a and 2b show the model inputs, C(t) and S(t), respectively. The turnover time required by mass balance is shown in Figure 2d, and the simulated atmospheric δ¹³C evolution in Figure 2c. Only the period 1950 to 2005 is highlighted in the absence of unexpected features in the pre-1950 period which in any case could be unduly influenced by pre-1990 (EPA06) or pre-1970 (Edgar4) back-extrapolation of the source. The turnover times in Figure 2c are not fully compatible in magnitude because they depend on the strength of the particular sink process or aggregation of processes being considered, recognising that turnover time is the ratio of atmospheric burden to sink strength.

Figure 2. The upper two panels show the 1950–2005 sub-intervals of the model inputs, and the lower panels the corresponding modelled output, with atmospheric data in the left-hand panels. Figure 2a is the globally-adjusted MacFarling Meure (2006) mixing ratio dataset, C(t), with illustrative annual global-mean data from the NOAA/ESRL network (E.J. Dlugokencky, personal communication), and Southern Hemisphere datasets from Baring Head, New Zealand (Lassey et al. 2010) and from archived samples collected at Cape Grim, Australia (Francey et al. 1999). Figure 2b shows the back-extrapolated EPA06 and Edgar4 constructions of the anthropogenic source history, each of which is augmented by a natural source of 222 Tg yr⁻¹ to give S(t). Figure 2d shows the corresponding turnover times, λ(t)⁻¹, along with modelled estimates of turnover time based on OH chemistry alone (Karlsdóttir and Isaksen 2000; Dentener et al. 2003). Figure 2c shows the modelled atmospheric δ¹³C evolution for four combinations of natural and anthropogenic source histories, each for a value of ε (cited in square brackets in the legend) chosen to provide an adequate fit to δ¹³C data from the Southern Hemisphere (Craig et al. 1988, Francey et al. 1999, Ferretti et al. 2005).

The choice of isotope fractionation, ε, determines the position on the δ¹³C axis in Figure 2c. The value chosen to optimally account for data in recent decades is indicated in the legend. Since those data all originate from the Southern Hemisphere (SH), each choice of ε requires adjustment for an inter-hemispheric gradient in δ¹³C(CH₄) of ∼0.3‰ (heavier in the SH atmosphere) (Lassey et al. 2007a). That adjustment would decrease the magnitude of each ε reported in the legend to Figure 2c by ∼0.15‰. Thus, only for Edgar4 with Nat2 is the adjusted ε value outside the 95% confidence interval for ε_total =−7.6 ± 1.4‰ (Table 1), and even then only by ∼0.05‰.

5 Discussion and conclusions

The mass-balancing sink appears to have slowly weakened through the industrial era until the 1980s (Figure 2d), noting that inter-annual sink variations may be partially an artefact of a smooth C(t). This is qualitatively consistent with model simulations of OH chemistry that suggest such weakening (Lelieveld et al. 1998) followed by 1980s strengthening (Karlsdóttir and Isaksen 2000; Dentener et al. 2003). Recognising the inverse relationship between turnover time and sink strength, the IPCC AR4 with larger sink and source than considered here (581 Tg yr⁻¹ ± 15%, excluding the chlorine sink) also requires a longer turnover time (8.7 ± 1.3 yr).

Of the four combinations of natural and anthropogenic source constructions, all give a reasonable portrayal of the recent atmospheric δ¹³C pattern observed in the SH, once a reconciling ε value (Figure 2c) is chosen. Edgar4 even partially captures the levelling off in δ¹³C since the late 1990s at Baring Head, probably due to the inclusion in Edgar4 of growing biogenic emissions such as from livestock. For all combinations of anthropogenic and natural inventory except Edgar4 with Nat2 the chosen ε value is within the acceptable range defined by Table 1. Given the quite different fossil components of Nat1 and Nat2 (Table 2), simulations of ¹⁴CH₄ could potentially be more discriminating.

The available atmospheric δ¹³C(CH₄) data, dominantly from the SH, are adequately simulated within uncertainty bounds by three of the four combinations of natural and anthropogenic source history. These uncertainties are strongly influenced by poor knowledge of the relative roles of the various sink processes and their isotopic fractionations (Table 1). With the additional uncertainty inherent in constructing inventories of natural emissions, anthropogenic source inventories remain poorly constrained by available data.

Acknowledgements

Funding support from the NZ Foundation for Research, Science and Technology to develop the model, and from the US Environmental Protection Agency to apply it is acknowledged. Useful discussions with Bill Allan (NIWA) and Jos Olivier (PBL, Netherlands) are also acknowledged.

Additional information

Notes on contributors

K.R. Lassey

The views of the authors do not necessarily represent the views of the U.S. Government or the Environmental Protection Agency.