Interpreting contemporary trends in atmospheric methane.

Atmospheric methane plays a major role in controlling climate, yet contemporary methane trends (1982-2017) have defied explanation with numerous, often conflicting, hypotheses proposed in the literature. Specifically, atmospheric observations of methane from 1982 to 2017 have exhibited periods of both increasing concentrations (from 1982 to 2000 and from 2007 to 2017) and stabilization (from 2000 to 2007). Explanations for the increases and stabilization have invoked changes in tropical wetlands, livestock, fossil fuels, biomass burning, and the methane sink. Contradictions in these hypotheses arise because our current observational network cannot unambiguously link recent methane variations to specific sources. This raises some fundamental questions: (i) What do we know about sources, sinks, and underlying processes driving observed trends in atmospheric methane? (ii) How will global methane respond to changes in anthropogenic emissions? And (iii), What future observations could help resolve changes in the methane budget? To address these questions, we discuss potential drivers of atmospheric methane abundances over the last four decades in light of various observational constraints as well as process-based knowledge. While uncertainties in the methane budget exist, they should not detract from the potential of methane emissions mitigation strategies. We show that net-zero cost emission reductions can lead to a declining atmospheric burden, but can take three decades to stabilize. Moving forward, we make recommendations for observations to better constrain contemporary trends in atmospheric methane and to provide mitigation support.

Methane accounts for more than one-quarter of the anthropogenic radiative imbalance since the preindustrial age (1). Its largest sources include both natural and human-mediated pathways: wetlands, fossil fuels (oil/gas and coal), agriculture (livestock and rice cultivation), landfills, and fires (2, 3). The dominant loss of methane is through oxidation in the atmosphere via the hydroxyl radical (OH). Apart from its radiative effects, methane impacts background tropospheric ozone levels, the oxidative capacity of the atmosphere, and stratospheric water vapor. As such, changes in the abundance of atmospheric methane can have profound impacts on the future state of our climate. Understanding the sources and sinks of atmospheric methane is critical to assessing future climate and also global tropospheric background ozone, which can impact air quality.

From ice core records, we know that atmospheric methane levels have nearly tripled since 1800 (4). Blake et al. (5) made the first accurate in situ measurements in 1978 and measurements from the National Oceanic and Atmospheric Administration (NOAA) (6) and Advanced Global Atmospheric Gases Experiment (AGAGE) (7) reached global coverage in 1983. These measurements showed a continued increase (with fluctuations) until 2000 when the globally averaged concentration stabilized at 1,750 parts per billion (ppb) (8). In 2007 atmospheric levels began increasing again (9, 10), with this rise continuing today. There has been much speculation about the cause of these recent trends, with numerous seemingly contradictory explanations (2, 3, 8–31). Attribution of these trends has proved to be a difficult task because () this period of renewed growth is characterized by a source–sink imbalance of only 3% and () there are a myriad of diverse processes with large uncertainties that could potentially emit methane. Here we leverage the extensive work conducted by the methane community over the last decades to clarify the current state of the science, specifically addressing the following: (i) What do we know about sources, sinks, and underlying processes driving observed trends in atmospheric methane? (ii) How will global methane respond to changes in anthropogenic emissions? And (iii), What future observations could help resolve changes in the methane budget?

Recent History of Atmospheric Methane

Preindustrial atmospheric methane levels were stable over the last millenium at 600–700 ppb, as inferred from ice core measurements in Antarctica (Fig. 1). Methane concentrations have been altered by humans even before industrialization (32) but began increasing more rapidly in the 1900s (4) due to both human agricultural activities and expanded use of fossil fuels. This rapid rise closely mirrors that of other greenhouse gases that are driven by industrialization and agriculture (e.g., ) (1). There is no debate about the cause of the bulk of this rise in atmospheric methane from preindustrial times to the present: human activities.

Fig. 1.

Observations of atmospheric methane over the past 2,000 y. Shown are Law Dome ice core record (blue) (4) and direct atmospheric observations from the South Pole (black, deseasonalized in gray) (6). Red line illustrates if the 7-y stabilization period is removed.

It is likely that natural sources of methane changed during this period as well; for example, Arora et al. (33) found an increase in simulated wetland emissions from 1850 to 2000 due to changes in temperature and Dean et al. (34) discuss how natural methane emissions may change in response to climatic changes. However, these changes in natural sources are small relative to the more than 300 Tg/y increase in anthropogenic sources from preindustrial times to the present (1, 3, 35). This rise in atmospheric methane from preindustrial levels continued unabated until the 1990s, at which point the methane record diverged from and O (which both showed continued growth).

Methane concentrations stabilized in 2000 (8) and then growth resumed in 2007 (9, 10) that continues today (6, 7). This period from 2000 to 2007 is referred to as the “stabilization” and the increase from 2007 to present is referred to as the “renewed growth.” Both stabilization and renewed growth have seen conflicting explanations in the literature. Dlugokencky et al. (8) suggested that this stabilization may be a new steady state for atmospheric methane and, as such, many analyses have viewed the period of renewed growth as anomalous. This view of the renewed growth as a departure from steady state has led to a search for methane sources that increased in 2007. However, if the stabilization period is removed from the contemporary methane record, then the long-term trend becomes a continuous rise (Fig. 1, Inset) with little change in the growth rate. One may wonder which period (if any) is anomalous in the contemporary methane record: If one expects steady state, then the renewed growth appears anomalous; conversely if one expects a long-term rise, then the stabilization appears anomalous. These two views may result in different research foci. For example, the former view may lead one to search for an increasing source while the latter may lead one to look for a decline in sources or increasing sink. The renewed growth has now continued for more than a decade, underlining that the 7-y stabilization period could be considered as anomalous. This perspective does not necessarily require a new, sustained emissions increase in 2007 as many papers have sought. The gaps that need explanation become the anomalous stabilization period and the evolving combination of emissions that contribute to the continued rise.

Atmospheric Clues and Inventory/Process Understanding of Atmospheric Methane

Explanations of recent atmospheric methane trends can be broadly grouped based on the types of proxy measurements used. Measurements of C- (the 13C/12C ratio in atmospheric methane) provide information about the fraction of methane coming from biotic (i.e., microbial) and abiotic sources, as biotic methane is produced enzymatically and tends to be depleted in 13C, making it isotopically lighter. Atmospheric ethane () can be coemitted with methane from oil/gas activity and, as such, has been used as a tracer for fossil methane emissions (11, 15, 18–20). Similarly, carbon monoxide can be coemitted with methane from biomass burning. Methyl chloroform () is a banned industrial solvent that has been used to infer the abundance of the dominant methane sink (the hydroxyl radical, OH) (38, 42–46). These four measurements (C-, , CO, and ) have been used in conjunction with atmospheric methane measurements. However, studies generally reached differing conclusions regarding the recent methane trends.

Fig. 2, Left shows the observations of atmospheric methane and the proxies used to explain the stabilization and renewed growth. Studies using ethane have argued that decreases in fossil fuel sources led to the stabilization of atmospheric methane in the 2000s (e.g., refs. 11 and 15) and that increases in fossil fuel sources contributed to the growth since 2007 (e.g., refs. 18–20). Studies using isotope measurements tend to find that decreases in microbial sources led to the stabilization (e.g., ref. 12) and increases in microbial sources are responsible for the renewed growth (e.g., refs. 17, 24, and 25). Studies that include methyl chloroform measurements tend to find that changes in the methane sink played a role in both the stabilization and renewed growth (e.g., refs. 22, 27, 28, and 47). Finally, Worden et al. (31) included measurements of carbon monoxide and inferred a decrease in biomass burning emissions, an isotopically heavy methane source, that helps reconcile a potential increase in both fossil fuel and microbial emissions.

Fig. 2.

Constraints on atmospheric methane over the past 40 y. Left column illustrates atmospheric constraints: methane (6), ethane (18), C- (ftp://aftp.cmdl.noaa.gov/data/ and www.iup.uni-heidelberg.de/institut/forschung/groups/kk/en/) (36, 37), and OH sink inferred from methyl chloroform (27, 28, 38), assuming a global methane source of 550 Tg/y. Black lines in the ethane panel are taken directly from Hausmann et al. (18). Right column illustrates deseasonalized process and inventory representations for the same time period: total anthropogenic (35), anthropogenic disaggregated to three most important anthropogenic sectors, wetland models (30, 39, 40), and fire emission estimates (41). The stabilization period is indicated in both columns by the vertical gray area.

The problem of inferring processes responsible for the stabilization and renewed growth is often underconstrained when framed in a global or hemispherically integrated manner. From a globally integrated perspective, we have three observables (, C-, ) and attempt to infer changes in methane emissions, the partitioning between methane source sectors, emissions, and OH concentrations. Solving this requires additional constraints, which can also have large uncertainties. Adding ethane or carbon monoxide helps only if we can assume that their emission ratios (/ or /CO) and their variation in time are well known and well characterized. Many studies have assumed that OH is unchanging in the atmosphere (e.g., refs. 17, 24, and 25) because it is well buffered (38, 48), thus making the problem well posed, leading to stronger conclusions regarding the processes driving the stabilization and renewed growth. However, changes of a few percent in OH are sufficient to perturb the global budget (27, 28), with a 4% decrease in global mean OH being roughly equivalent to a 22 Tg/y increase in methane emissions.

Fig. 2, Right shows our current inventory- and process-based understanding of global methane sources. Based on this, the only sources that show a multidecadal trend are anthropogenic (waste, agriculture, and fugitives from fossil fuels). Natural sources and sinks (e.g., wetlands, fires, and OH) exhibit substantial variability on subdecadal scales but we do not have a process/inventory-based explanation for a long-term trend. For example, Poulter et al. (30) were unable to explain the renewed growth with changes in wetland emissions. Some individual wetland models do find increases in emissions [e.g., McNorton et al. (49)], but the increases are small (2 Tg/y) relative to the source–sink imbalance (20 Tg/y). Variations in many of these natural sources and sinks have been found to be driven in part by the El Niño–Southern Oscillation (ENSO) (e.g., refs. 31 and 50–53). The long-term growth trend in atmospheric methane is best explained by the continued rise in anthropogenic emissions—even though the most uncertain sectors are predominantly natural (wetlands and OH)—and as long as anthropogenic emissions continue to rise we expect a concurrent rise in atmospheric methane with variability superimposed due to fluctuations in natural sources and sinks. There is significant uncertainty in anthropogenic emissions, as evidenced when two different versions of the same inventory produce different expected emissions (Fig. 2, Top Right), but anthropogenic sources remain alone as able to explain the long-term rise in methane emissions over the past 40 y.

As mentioned above, there are large uncertainties in many aspects of the methane budget relative to the changes needed to reconcile the contemporary trends. Specifically, a 20 Tg/y imbalance (or 3.5% change) in the source–sink budget is sufficient to explain observed changes in methane. Current uncertainties in individual components of the methane budget greatly exceed this threshold. Namely, uncertainties in OH are on the order of 7% [1- from Rigby et al. (28), corresponding to 38 Tg/y]; differences in tropical wetlands can be as large as 80 Tg/y [max–min from Saunois et al. (3)]; and the uncertainties in the C- source signatures for fossil fuel and microbial sources are 10.7‰ and 6.2‰, respectively [1- from Sherwood et al. (54)], which are large enough to attribute the entire source–sink imbalance to either fossil or nonfossil sources [supplemental section 1 in Turner et al. (27)].

Can all of the various lines of evidence be consistently explained? If we focus on the perspective that the stabilization period is anomalous, it can be identified as a time of elevated OH relative to preceding and succeeding years. This shift alone could explain the stabilization period as well as the renewed increase. It is likely a decrease in anthropogenic emissions in the late 1990s (masked at first by the large fire emissions from El Niño) also contributed. There has been a long-term decline in atmospheric ethane [Simpson et al. (15)] that can be seen in the Southern Hemispheric ethane record in Fig. 2; however, the Northern Hemispheric measurements have been more variable and Hausmann et al. (18) suggest an increase since 2007 due to an increase in fossil fuel emissions. Inventories also predict increased fossil fuel emissions, but estimated resumption starting a few years earlier, in the middle of the stabilization period. While there may be a timing offset in the inventory, the more recent increase in atmospheric ethane could also be largely driven by expanded production of gas in wet oil fields where : ratios are very large (55). These proposed source/sink changes would require concomitant changes in the partitioning between isotopically heavy and light sources to satisfy the constraints from C-. It is tempting to conclude the isotopic shift in atmospheric methane must prove the growth is driven by an increase in microbial emissions; however, the problem is underconstrained in a globally integrated framework and one can find scenarios that are consistent with the C- measurements that include increasing fugitive fossil fuel emissions [e.g., Worden et al. (31)].

All studies that include measurements of methyl chloroform find changes in OH that resemble those shown in Fig. 2, Bottom Left (e.g., refs. 22, 27, 28, 38, and 47) while studies that do not include methyl chloroform find that changes in sources alone drive contemporary trends and that OH changes are negligible (e.g., refs. 17, 24, and 56). This implies that either () there are latent issues in how methyl chloroform observations are being used to estimate OH or () future work on methane trends should include measurements of methyl chloroform to jointly infer OH. Studies that attributed methane trends to OH (e.g., refs. 22, 27, and 28) did not identify a physical mechanism for the OH changes and the lack of a mechanism remains a valid criticism (e.g., ref. 57). Holmes et al. (58) discuss the processes that impact global mean OH (and methane lifetime) and found temperature, water vapor, stratospheric ozone column, biomass burning, lightning , and methane abundance to be important drivers. Gaubert et al. (59) found that decreases in CO emissions may have increased OH from 2002 to 2013, opposite to what has been inferred via methyl chloroform. Recently, Turner et al. (52) found ENSO to be the dominant mode of OH variability in the absence of external forcing, acting primarily through changes in deep convection and lightning . However, as mentioned above, ENSO would likely contribute to the variability but not long-term trends.

Further, papers that inferred OH changes from the available observational constraints (e.g., refs. 27 and 28) did not explicitly simulate the feedbacks with or CO as suggested by Prather and Holmes (60). In summary, we currently lack independent evidence to confirm or refute OH changes. At the same time, we need to consider that mechanistic global atmospheric chemistry transport models fail to even simulate the partitioning of OH between the Northern and Southern Hemispheres (e.g., refs. 44 and 61), which alone warrants further OH studies. It should also be stressed that a similar discrepancy between what mechanistic models predict and what is inferred from observations holds for wetlands, where an ensemble of wetland models is inconsistent with the hypothesis of a large shift in tropical emissions [Poulter et al. (30) and wetland emissions in Fig. 2, Right]. This stresses the need to reconcile process-based models with observations because findings of either large changes in OH or wetland emissions are not particularly enlightening if we fail to understand the causes of these variations.

Isotopic and ethane observations provide valuable clues to the relative balance of sources and sinks of methane. One of the most critical gaps in isotopic- and/or ethane-based global observations is the underlying assumption that source/sink signatures and their variation in time are well known. That is, we a priori know the isotopic (ethane) characteristic of every source (sink) and how it varies in time. However, this assumption generally does not hold. For example, a recent update to our understanding of isotopic characteristics of sources from Sherwood et al. (54) shifted the expected recent historical balance of biotic/abiotic emissions (25, 54). However, this new inventory still has little information on tropical wetlands’ microbial signature (only 50 samples from tropical wetlands). A further update to the inventory would likely shift the interpretation of the trends and budget. Furthermore, the assumption of temporally invariant signatures is likely false, as the C- signal from a wetland is the balance of production (methanogenesis) and loss (oxidation by methanotrophs)—if that wetland exhibits changing fluxes in response to changing water/temperature, the relative production/loss terms will shift and the isotopic signal will change (e.g., refs. 62 and 63). McCalley et al. (64) demonstrated this for microbial communities in permafrost thaw and Dean et al. (34) highlighted the importance of quantifying whether consumption by microbes will balance production in the future. A similar problem holds for ethane, where oil/gas fields have drastically different : ratios, and within a single field this ratio can change over the history of production of a field. In addition, different amounts of ethane are extracted from natural gas, depending on the economic value of ethane as petrochemical feedstock. These confounding factors are more tractable at higher spatial resolution (e.g., the isotopic source signatures and : ratios are well characterized for individual sources or basins) than at the global or hemispheric scale.

Spatial gradients in observed methane concentration have also been used to infer emissions at a variety of scales. This is typically done via “atmospheric inversions,” using models to account for atmospheric transport. Houweling et al. (65) provide an extensive review of work on atmospheric inversions over the past 25 years that was started by Fung et al. (66) in 1991. Briefly, these atmospheric inversions have leveraged existing surface, aircraft, and satellite observations to infer our best understanding of methane fluxes for specific time frames (e.g., refs. 51 and 67–79). The Global Carbon Project (GCP) published a synthesis of the methane budget in 2013 [Kirschke et al. (2)] that was recently updated by Saunois et al. (3) based on an ensemble of inversions. The GCP highlighted the importance of reducing the uncertainty on wetland emissions and reducing “double counting” of sources. It did not address changes in the methane sink but reported a climatological range for the sink based primarily on the work of Naik et al. (61). Atmospheric inversions are limited by the spatiotemporal coverage of the observations and our ability to accurately simulate atmospheric transport. As such, increases in the spatiotemporal coverage of traceable, calibrated, and validated observations (from surface, aircraft, or satellite) and improvements in atmospheric transport models would help this approach in constraining the methane budget.

Space-borne observations of methane and proxies related to specific sectors represent an attractive constraint on the methane budget [e.g., Sellers et al. (80)], as they provide a unique spatial coverage. Jacob et al. (81) provide a detailed review of the role of satellite observations. Briefly, satellite observations have proved to be be useful in constraining methane sources at local-to-regional scales (e.g., refs. 16, 74, 75, 77, 78, and 82–84) but have thus far played a relatively limited role in the discussion of global methane trends because the record is short compared with in situ measurements. For example, the first total column measurements of methane were made by Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) in 2003 (73, 85) and Greenhouse Gases Observing Satellite (GOSAT) (86) is the longest-running satellite that measures total column methane with 9 y of data (measurements started in April 2009) (87, 88). Networks like the Total Carbon Column Observing Network (TCCON) (89) and AirCore [Karion et al. (90)] are crucial to identify biases in satellite measurements, evaluate their uncertainties, and facilitate intercomparisons between different satellite instruments. Satellite observations will likely play a growing role in the discussion of future methane trends as the record length increases and new missions like the recently launched Tropospheric Monitoring Instrument (TROPOMI) (91) and recently funded Geostationary Carbon Cycle Observatory (GeoCARB) instrument (geostationary orbit) (92) emerge. TROPOMI launched in October 2017 and reported encouraging observations of CO (93) and methane (94). For satellites to provide their full potential value added, rigorous validation and traceability are necessary. Atmospheric inversions should also attempt to cope with potential biases in satellite data by jointly inferring bias terms.

The role of specific regions such as the United States and the Arctic in recent methane trends is also debated. For example, Turner et al. (16) inferred an increase in US emissions but Bruhwiler et al. (95) find that this increase is inconsistent with a model ensemble from the GCP. This topic (US methane emissions) was the focus of a review paper by Miller and Michalak (96) and a recent National Academy of Sciences Report (97); however, the role of US methane emissions is still under debate [Sheng et al. (98)]. It underlines the sobering fact that even for the data-rich United States, we still cannot conclusively determine whether there has been a long-term trend in methane emissions. The role of methane emissions from the East Siberian Arctic Shelf (ESAS) is another topic that has been heavily debated in the recent literature. Work from Shakhova et al. (99) extrapolated ship-based measurements to estimate ESAS methane emissions; however, more recent work from Berchet et al. (100), Thornton et al. (101), and Warwick et al. (102) found emissions that were a factor of 4–30 lower. Widespread emissions of methane hydrates are unlikely [Ruppel et al. (103)] as methane sources in waters deeper than 100 m have negligible contributions to the atmosphere (104, 105) and recent work from Sparrow et al. (106) uses radiocarbon measurements from the Beaufort Shelf in the Arctic Ocean to infer that less than 10% of methane in surface water is from sources deeper than 30 m. More broadly, there has been a lot of interest in understanding how methane emissions from the Arctic may change in the future because of the temperature dependence of microbial methane sources and enhanced warming due to Arctic amplification (1). While it is important to understand these regional emissions, current uncertainties in the tropics greatly exceed the absolute magnitude of Arctic sources. Further, Sweeney et al. (107) suggest Arctic emission changes would have little impact on global budgets if the temperature sensitivity is similar to what has been observed in the present. This is not to discount the potential importance of future Arctic methane emissions, but the prime uncertainties in the current global methane budget lie in the tropics for a number of reasons: () tropical wetlands are the largest natural source, () methane oxidation through OH is largest in the tropics, and () the ground-based observational network is least dense and frequent cloud cover reduces satellite data densities.

Impact of Changing Anthropogenic Methane Emissions on Global Methane

Despite the uncertainty of the current relative balance of different controls on atmospheric methane, there is no debate that the large increase from preindustrial times is driven by anthropogenic emissions and that reducing anthropogenic emissions can lead to direct, near-term decreases in atmospheric methane. However, changing methane emissions will alter the methane lifetime via chemical feedbacks with OH [Prather (108, 109)] and, as such, atmospheric abundances can exhibit longer timescales than one may assume. We illustrate this in Fig. 3 by using a simple box model [adapted from Turner et al. (27)] to evaluate four scenarios to bound the future methane abundances: continued growth in anthropogenic methane emissions (case A), a stabilization of methane emissions in 2012 (case B), and an emission decrease over 10 y (case C) or instantaneously (case D). The emissions decrease in the latter two scenarios is based on a recent report from the International Energy Agency (110) that estimates current methane emissions from oil and gas could be reduced by 40–50% with zero net cost. For all scenarios, we consider how these changes in methane abundances will impact OH using a simplified -CO-OH system [Prather (108, 109)] and cases where methane does not feed back on OH. The latter case with constant OH is meant to account for factors that might buffer methane-induced OH changes [e.g., changes in the ozone photolysis rate or changes in emissions; see Murray et al. (111) and Holmes et al. (58) for a discussion of some of these factors]. The methane emissions and OH anomalies for these four scenarios are shown in Fig. 3, Left and Right, respectively.

Fig. 3.

Projections of atmospheric methane over the next 30 y. (Left) The methane emissions from 1980 to 2050 under four different emission scenarios: continued growth in anthropogenic emissions (case A, red), stabilization of emissions in 2012 (case B, blue), and an emission decrease over 10 y (case C, orange) or instantaneously (case D, green). Anthropogenic emissions from 1980 to 2012 are from Emission Database for Global Atmospheric Research (EDGAR) v4.3 (black). (Center) The simulated methane concentrations under the four emission scenarios with interactive OH (solid line) and a constant OH concentration (dashed line; no OH or CO feedback). Colored vertical lines to the right of the panel show the range of the concentrations in 2050 for interactive and constant OH. (Right) The OH anomaly due to changes in methane and CO. The stabilization period in all panels is indicated by the vertical gray shading.

In Fig. 3, Center we see the range of possible methane responses. With increasing emissions, atmospheric levels increase unabated. Important subtleties remain: If OH dynamically responds to methane, atmospheric levels would be 180 ppb higher in the case of continued increasing emissions. Even if emissions stabilized in 2012, atmospheric levels are still increasing in 2050 with interactive OH. This highlights a subtle but important point relevant for understanding recent atmospheric methane behavior: with emissions stabilization atmospheric methane can still increase for more than three decades [see Prather (108, 109) for a detailed discussion of these feedbacks and their relation to the eigenvalues of the chemical system]. In the scenarios of net-zero cost emission reductions, we do see the atmosphere exhibits decreases in atmospheric concentrations, but depending on the time frame of emission reductions, the atmospheric decrease can take a decade to detect it, and if OH responds dynamically, atmospheric abundances of methane will remain significantly higher (50 ppb). Also worth noting, in all of the dynamic OH cases a significant perturbation is projected. In the case of continued rising emissions, this could impact global mean OH by 10%—a large shift that could have profound impacts on the oxidative capacity of the global atmosphere (e.g., ref. 112).

How Can We Do Better Moving Forward?

Long-term in situ observations provide the backbone upon which our current understanding of atmospheric methane is founded. Continuation of these observations is paramount to observing and understanding future methane changes. However, it is now abundantly clear that these in situ observations alone are not sufficient for unequivocally partitioning contemporary variations in atmospheric methane (from 1980 to the present) to specific source/sink pathways. This is, in part, because contemporary methane trends are driven by a source–sink imbalance of 20 Tg/y (or 3.5%) yet uncertainties in regional and sectoral components of the methane budget greatly exceed this threshold. In particular, methane emissions from wetlands have an uncertainty of 40 Tg/y [range from Saunois et al. (3) is 80 Tg/y] and methane loss due to reaction with OH has an uncertainty of 7% [or 38 Tg/y; e.g., Rigby et al. (28)]. These two sectors represent the largest sources of uncertainty in the methane budget and reconciling the contemporary trends will require observations that can (i) provide better constraints on these uncertain sectors and (ii) improve our process-level understanding and representation at regional scales. Expansion of the current observational network of methane (and coemitted species) from surface or space will provide valuable information. However, no single program is likely to settle the debate; addressing the major uncertainties in the contemporary methane budget will require a concerted effort in multiple areas. Here we highlight a few potential pathways toward better constraining future methane emissions and their drivers.

i) Expand Measurement Networks to Include More Proxies for Methane Source Partitioning.

Radiocarbon (14C; e.g., ref. 113), deuterium (i.e., D), and “clumped” isotopologue measurements (molecules multiply substituted with rarer isotopes, such as 13CH3D or 12CH2D2; ref. 114) could provide additional leverage on partitioning the global budget because they would help isolate changes due to the most uncertain sectors (e.g., wetlands and OH). Specifically, radiocarbon measurements would help to separate fossil and nonfossil methane emissions [Petrenko et al. (113)] while clumped isotopologue measurements can constrain biogenic/thermogenic emissions [Stolper et al. (114)] or the loss via reaction with OH [Haghnegahdar et al. (115)]. However, both of these measurements will require advances in the analytical techniques before they could be used in ambient conditions. D measurements, on the other hand, are less useful than radiocarbon or clumped isotopologues but the measurements are substantially easier to make. All of these isotopic measurements could help to constrain the most uncertain sectors in the methane budget, but there is a trade-off between added value and cost. Expanded studies of source signatures would be required for these isotope-driven approaches to provide maximum value. Radiocarbon shows potential with a less extensive source signature study requirement, as this tracer provides a cleaner delineation between fossil and contemporary methane sources. We encourage more observing-system simulation studies that quantify the added values of different proxies as well as redundancies, at the local to regional and global scales. In the interim, archiving of air samples [such as those at Commonwealth Scientific and Industrial Research Organisation (CSIRO), ref. 116] would provide an affordable strategic approach for enabling future measurements of attributive tracers that are infeasible with current technology or have not yet been recognized. As such, expansion of the air archive would enable the community to work backward in future years and address the most uncertain aspects of the methane budget.

ii) Targeted Measurement and Modeling Programs Focused on Tropical Wetlands and Global OH.

These sectors are currently the largest uncertainties in interpreting trends in methane and moving forward will continue to present a challenge unless we can improve the observational constraints and our ability to represent emissions/uptake with process-driven models. Development of high-resolution inventories that resolve, for example, wetland and lake emissions without double counting [e.g., Thornton et al. (117)] and spatially resolved isotopic source signatures [e.g., Ganesan et al. (118)] will be crucial to help reduce uncertainties in the use of isotopologue measurements. Dense observations (ground, airborne or space-borne, campaign or sustained) coupled with methane wetland model development for multiple tropical regions could provide a pathway toward more accurate representation and understanding of emissions from this sector. A similar observational approach was applied to the US oil and gas sector [Alvarez et al. (119)] that led to substantial improvements in the representation of methane sources [Zavala-Araiza et al. (120)]. Such a campaign could help improve the dynamics of methane emissions in wetland models (including regionally relevant isotopic source signatures) and their sensitivity to changes in temperature and inundation. Global OH presents a different challenge, as point measurements of OH are unlikely to adequately sample the variability in OH to the precision needed for methane trends (better than 3%). Further, the methyl chloroform constraints on OH are degrading with time as the ambient concentrations of methyl chloroform are now 2 parts per trillion (a 50-fold decrease from the 1990s) (27); alternate strategies need to be developed [Liang et al. (46)]. Recent work from Zhang et al. (53) suggests that satellite observations of midtropospheric methane could be used for this purpose. Additional work using existing measurements, such as those from AirCore [Karion et al. (90)] or the Atmospheric Tomography Experiment (ATom) (https://espo.nasa.gov/atom/content/ATom), and future campaigns should further investigate the possibility of inferring OH with midtropospheric measurements from satellites.

Implications for Emissions Mitigation

While uncertainties in the methane budget exist, they should not detract from the key points discussed here. Namely, reducing anthropogenic methane emissions will slow or reverse the rise in atmospheric concentrations; however, depending on the timescale and magnitude of reduction, it may take decades before atmospheric levels decline. When considering recent decades, the stabilization period is emerging as anomalous due in part to fluctuations in natural sources/sinks, whereas the last decade of growth continues the long-term, increasing trend that is due to human activities.

Even with present uncertainties on global methane trends, there have been a a number of recent advances in measurement technology that have tremendous potential for opportunistic mitigation (i.e., reducing emissions at no net cost). A few notable examples include identifying large fugitive leaks in oil and gas infrastructure and changing the diet of livestock. Specifically, remote sensing has demonstrated the ability to identify anomalous, large emitters and focused programs to use aircraft- or space-based observations to identify and mitigate emissions could prove cost efficient and effective (82, 121–125). Recent advances in frequency-comb spectrometers (126, 127) and affordable, small ground-based sensors may also provide a mitigation opportunity for superemitters in oil/gas basins (128). Changes in the diet of livestock could reduce the production of methane in dairy cattle without reducing milk production and, as such, could be an opportunity to reduce methane emissions from livestock (129, 130). Implementation of these or other mitigation strategies could help to curb future increases in atmospheric methane and provide detectable changes in the global methane burden within decades.