Main

Atmospheric measurements of carbon-13 in methane (δ13CH4) have been used to estimate the fossil fraction of the contemporaneous CH4 budget3. This approach relies on having accurate estimates of the δ13C signatures of the major CH4 source categories (fossil, microbial and biomass burning) and the strength of the biomass burning source. Large uncertainties in these parameters in the past preclude accurate δ13CH4-based estimates of preindustrial-era fossil CH4 emissions8,11,12. Radiocarbon (14C) is an ideal tracer for quantifying the fossil component of the atmospheric CH4 budget because all 14C in fossil CH4 has decayed. By contrast, biogenic CH4 sources (wetlands, biomass burning) have a 14C activity similar to that of contemporaneous atmospheric CO2 (ref. 4,8). Interpretation of atmospheric 14CH4 measurements from 1987–2000 suggests that the fossil fraction of the contemporary CH4 budget is 30 ± 2.3% (ref. 13; 1σ). However, the interpretation of atmospheric 14CH4 in recent decades has been complicated by (1) rapidly changing atmospheric 14CO2 (from above-ground nuclear testing and fossil fuel emissions) that propagates into biospheric CH4 emissions13, and (2) direct 14CH4 emissions from nuclear power plants4,5. By contrast, palaeoatmospheric 14CH4 measurements from ice cores offer a direct constraint on natural geological CH4 emissions without these complications. Whereas geological CH4 emissions have the potential to change on tectonic- and glacial-cycle timescales14, they have very probably been constant over the past few centuries. The preindustrial-era emission estimates can therefore be applied to the modern CH4 budget with confidence.

Ice core 14CH4 analysis is challenging owing to both the very large sample requirement (~1,000 kg of ice) and interference from in situ cosmogenic 14C production within the ice crystal lattice15. We address the former by using a large-diameter ice drill and a large-volume ice-melting apparatus (Supplementary Information section 1) to obtain sufficient CH4 (~20 μg C) for 14C analysis by accelerator mass spectrometry. To address the latter, we follow the established8,16 approach of analysing 14C of carbon monoxide (CO) in parallel with 14CH4. 14CO is very sensitive to in situ cosmogenic 14C production15 and can be used to precisely establish the effective cosmic ray exposure history of each sample. We then correct the 14CH4 data using the known in situ cosmogenic 14CH4/14CO production ratio in ice15 (Supplementary Information sections 5, 6). The in situ cosmogenic 14CH4 component in the samples used in this study is much lower (<2% of total 14CH4) than in ablation-zone ice used in previous palaeoatmospheric 14CH4 studies (~30% of total 14CH4)8,16. We present new 14CH4 data from large-volume ice core samples and firn air sampling from Summit, Greenland, which we combine with prior firn air 14CH4 measurements from Law Dome DSSW20K4 and Megadunes17, Antarctica. Our combined record spans from about 1750 to 2013 and captures the evolution of atmospheric 14CH4 since the preindustrial era (Fig. 1). The movement of gases within the firn and closure into bubbles is characterized using a firn air transport model18, and the time series of atmospheric 14CH4 is reconstructed using a matrix inversion technique19,20 (Supplementary Information section 9).

Fig. 1: Reconstruction of atmospheric 14CH4 from firn air and ice core data.
figure 1

a, Global CH4 mole fraction, [CH4], reconstructed from ice core, firn air and atmospheric measurements1. ppb, parts per billion. b, Reconstructed history of atmospheric Δ14CH4 from firn air and ice core samples (this study). Dotted lines represent the 95% confidence range based on all calculated 14CH4 histories using three different inversion methods (Supplementary Information section 9). Ice core and firn air Δ14CH4 measurements are shown at the mean age of the modelled air age distribution. Vertical error bars on the Δ14CH4 data from each site represent the 2σ uncertainty for each sample after corrections (Supplementary Information Tables 2, 6), and horizontal error bars represent ±2Δ, where Δ is the spectral width of the sample-air age distribution20. We also plot the 14CH4 signature of the contemporaneous biogenic source (blue; Supplementary Information section 10). Our time series begins in 1850 because the age distributions of the collected ice core samples have poor coverage of air from ~1780 to 1850 (Supplementary Information section 10, Supplementary Fig. 3B).

Full size image

Our atmospheric 14CH4 reconstruction (Fig. 1) is indistinguishable from the 14CO2-derived contemporaneous biogenic 14CH4 signature (blue curve, Supplementary Information section 10) before 1880, suggesting very low natural geological CH4 emissions. Atmospheric 14CH4 began to decrease around 1880, coincident with substantial increases in the use of coal, oil and natural gas (Fig. 2)21. The precise timing of the 14CH4 minimum (in the 1940s in our reconstruction) is difficult to establish owing to the broad age distributions of individual firn air and ice core samples, as well as the smoothing applied by the matrix inversion technique to address the non-uniqueness of the solution19. Beyond this fossil 14C minimum, our samples are affected by the propagation of 14C from atmospheric nuclear testing into the carbon cycle22 and by emissions from nuclear power plants (starting in the 1970s), which drove a sustained 14CH4 increase despite decreasing 14CO24,5. We calculate the fossil CH4 fraction and develop a time series of fossil CH4 emissions (Fig. 2) using a one-box atmospheric model (Supplementary Information section 10). The broad age distributions of our air samples (Supplementary Fig. 3) result in a smoothed representation of the atmospheric 14CH4 history that cannot capture the abrupt increase of bomb 14CO2 (and subsequently 14CH4) starting in 1955. Therefore, we interpret the fossil CH4 fraction only before the 1940s. We find an increase in the total (geological plus anthropogenic) fossil emissions from negligible CH4 emissions in the mid-19th century to 64.8 teragrams CH4 per year (Tg CH4 yr−1) in 1940.

Fig. 2: Growth in fossil CH4 emissions and fossil fuel consumption.
figure 2

a, Historical fossil fuel energy consumption21. b, Calculated total fossil CH4 emission history (solid line) from the one-box model (Supplementary Information section 10). The dashed lines show the 95% confidence interval.

Full size image

Assuming that the oldest ice core 14CH4 sample in our reconstruction (mean age 1756 ad; Fig. 1) is devoid of anthropogenic fossil CH4 contributions, we use the contemporaneous biogenic 14CH4 source signature to calculate the natural geological CH4 emissions during the preindustrial era: 1.6 Tg CH4 yr−1 with a 95% confidence interval (CI) maximum of 5.4 Tg CH4 yr−1 (Supplementary Information section 10, Supplementary Fig. 5). Our 95% confidence limit of 5.4 Tg CH4 yr−1 agrees well with, and provides a tighter constraint than, the only other published 14CH4-based estimate of natural geological CH4 emissions from ice cores, which sampled air from the most recent deglaciation (0 to 15.4 Tg CH4 yr−1, 95% CI range)8.

Our result is much lower than estimates from recent source inventory (‘bottom-up’) studies typically used in global CH4 budgets2, which suggest natural geological emissions of ~40–60 Tg CH4 yr−1 (ref. 6). A recent study7 aimed at developing gridded maps of geological CH4 emissions revised this estimate downwards to 37 Tg CH4 yr−1 on the basis of data and modelling specifically targeted for gridding; however, the CH4 emissions increased to 43–50 Tg CH4 yr−1 when extrapolated to account for temporal variability in mud volcano eruptions and onshore or submarine geological seeps that lack location-specific measurements. Natural fossil CH4 emissions of about 40 Tg CH4 yr−1 (out of total preindustrial-era CH4 emissions of 215 Tg CH4 yr−1; Supplementary Fig. 5) would result in a preindustrial-era ∆14CH4 of around −185‰, which is in clear disagreement with our data (1.5‰ ± 21.2‰, 2σ; Fig. 1). Bringing our 14C results into agreement with the bottom-up estimates of natural fossil CH4 emissions would require an order-of-magnitude larger correction for in situ cosmogenic 14CH4. This would in turn require either an order-of-magnitude higher 14CO content in the sampled ice or an order-of-magnitude higher in situ 14CH4/14CO production ratio; both of these possibilities are well outside the respective uncertainties. The added uncertainties arising from the in situ and procedural corrections to the measured 14CH4 are also too small to explain the disagreement (Supplementary Information section 10, Supplementary Information Table 8).

Diffuse microseepage (24 Tg CH4 yr−1), macro-seeps and mud volcanoes (8.1 Tg CH4 yr−1), submarine seepage (>7 Tg CH4 yr−1) and geothermal manifestations (5.7 Tg CH4 yr−1) represent the main categories of natural geological CH4 emissions in the latest comprehensive bottom-up analysis7. Each of these four categories is nearly equivalent to, or exceeds, our upper bound (at 95% confidence) on the total preindustrial-era geological CH4 emissions (5.4 Tg CH4 yr−1). Emission estimates for diffuse microseepage are based on limited flux-chamber measurements in regions of known gas seepage (for example, ref. 23), which are scaled up to a global flux estimate based on the total dryland area situated above hydrocarbon reservoirs (~10% of Earth’s total land surface area), the percentage of measurements that show a positive flux, and emission rates chosen on the basis of several geological factors7. It is possible that the uncertainties associated with such global upscaling are much larger than reported, resulting in an overestimation by an order of magnitude or more. Similarly, emission estimates from macro-seeps, mud volcanoes and geothermal manifestations are derived from limited observations, which are scaled up to a global total7. To provide a sense of scale for the extrapolation in the case of mud volcanoes, ~0.0026 Tg CH4 yr−1 of measured CH4 emissions (table S2 in the supplement of ref. 7) are scaled up to 6.1 Tg CH4 yr−1 (table 2 in ref. 7).

With regard to submarine seepage, recent studies suggest that CH4 emissions to the atmosphere are probably very low owing to rapid dissolution of rising bubbles24 and rapid oxidation of dissolved CH4 (ref. 25). Furthermore, 14CH4 measurements in surface waters indicate minimal quantities of fossil CH4 even in shallow waters over areas of active seeps or methane hydrate degradation26. Our atmospheric 14CH4 measurements for the preindustrial era indicate that either (1) the uncertainties associated with global upscaling of geological emissions from discrete measurements result in overestimation by an order of magnitude, or (2) geological CH4 emissions quantified by these measurements were not present in the preindustrial era and may have been triggered by fossil fuel extraction from hydrocarbon reservoirs or other anthropogenic activity such as groundwater aquifer depletion. If the latter is true, such emissions cannot be considered natural.

A recent study3 used ice core δ13CH4 measurements to arrive at a natural geological CH4 emission estimate that is on par with what is indicated by bottom-up methods (~50 Tg CH4 yr−1). However, ref. 8 showed that ice core δ13CH4 data do not provide a strong constraint on preindustrial-era geological emissions and are also compatible with a minimal geological source. Measurements of ethane27 in ice cores have also been used to suggest considerable emissions of fossil CH4 during the preindustrial era. However, this is also an ambiguous constraint because ice core measurements of ethane mole fraction cannot discriminate between contributions from biomass burning (a major source) and natural geological emissions11. Our preindustrial-era 14CH4 measurements, by contrast, place an unambiguous constraint on natural fossil CH4 emissions by directly recording the 14C-free fraction of atmospheric CH4.

Our 14CH4 reconstruction does not allow accurate quantification of the post-1950 fossil CH4 budget, owing to relatively poor constraints on the interfering nuclear 14CH4 sources. Previous work used atmospheric δ13CH4 measurements to quantify the global fossil CH4 source in recent decades3, but relied on inventory-based assessments to constrain the natural geological component. We combine our 14CH4-based constraint on natural geological emissions (1.6 Tg CH4 yr−1) with δ13CH4-based constraints on the total fossil source (following the same one-box model approach as ref. 3; Supplementary Information section 11) to estimate recent anthropogenic fossil CH4 emissions. This approach yields 177 ± 37 Tg CH4 yr−1 (1σ) for anthropogenic fossil CH4 emissions during 2003–2012. Our estimate is 22% higher than the previous estimate of 145 ± 23 Tg CH4 yr−1 (1σ) over the same interval3, and 33–55% higher than the range of bottom-up estimates (114–133 Tg CH4 yr−1; ref. 2). We note that our δ13CH4-based calculation uses an updated value for the CH4 sink isotopic fractionation (Supplementary Information section 11); if we use the same value as ref. 3, the anthropogenic fossil source estimate is 194 ± 34 Tg CH4 yr−1 for the same time period.

Our results indicate that bottom-up inventories strongly underestimate CH4 emissions from fossil fuel extraction, distribution and use. A study using both ground-based facility-scale measurements and verification from aircraft sampling found that US oil and natural-gas CH4 emissions (largely from the production and gathering industry segments) are ~60% higher than those reported by the US Environmental Protection Agency28, one of the primary data sources used in bottom-up inventories2. If we consider a scenario in which the global bottom-up emissions of fossil CH4 from the oil and natural-gas industries (79 Tg CH4 yr−1; ref. 2) are similarly underreported by 60%, this would amount to unreported emissions of ~47 Tg CH4 yr−1, which is in agreement with the fossil CH4 emission shortfall that we identify in the current generation of bottom-up inventories (44–63 Tg CH4 yr−1). Our results imply that anthropogenic fossil CH4 emissions now account for about 30% of the global CH4 source and for nearly half of anthropogenic emissions, highlighting the critical role of emission reductions in mitigating climate change9,10.