Increasing anthropogenic methane emissions arise equally from agricultural and fossil fuel sources

R B Jackson, M Saunois, P Bousquet, J G Canadell, B Poulter, A R Stavert, P Bergamaschi, Y Niwa, A Segers and A Tsuruta 1 Department of Earth System Science, Woods Institute for the Environment, and Precourt Institute for Energy, , Stanford University, Stanford, CA 94305-2210, United States of America 2 Laboratoire des Sciences du Climat et de l’Environnement, LSCE-IPSL (CEA-CNRS-UVSQ), Universit́e Paris-Saclay, 91191 Gif-sur-Yvette, France 3 Global Carbon Project, CSIRO Oceans and Atmosphere, Canberra, ACT 2601, Australia 4 NASA Goddard Space Flight Center, Biospheric Sciences Laboratory, Greenbelt, MD 20771, United States of America 5 Global Carbon Project, CSIRO Oceans and Atmosphere, Aspendale, VIC 3195, Australia 6 European Commission Joint Research Centre, 21027 Ispra (Va), Italy 7 Center for Global Environmental Research, National Institute for Environmental Studies, Tsukuba, Ibaraki 305-8506, Japan 8 Meteorological Research Institute, Tsukuba, Ibaraki 305-0052, Japan 9 TNO, Dept. of Climate Air & Sustainability, NL-3508-TA Utrecht, The Netherlands 10 Finnish Meteorological Institute, FI-00101 Helsinki, Finland

Climate stabilization remains elusive, with increased greenhouse gas concentrations already increasing global average surface temperatures 1.1 • C above pre-industrial levels (World Meteorological Organization 2019). Carbon dioxide (CO 2 ) emissions from fossil fuel use, deforestation, and other anthropogenic sources reached~43 billion metric tonnes in 2019 , Jackson et al 2019. Storms, floods, and other extreme weather events displaced a record 7 million people in the first half of 2019 (IDMC 2019). When global mean surface temperature four million years ago was 2 • C-3 • C warmer than today (a likely temperature increase before the end of the century), ice sheets in Greenland and West Antarctica melted and parts of East Antarctica's ice retreated, causing sea levels to rise 10-20 m (World Meteorological Organization 2019).
Methane (CH 4 ) emissions have contributed almost one quarter of the cumulative radiative forcings for CO 2 , CH 4 , and N 2 O (nitrous oxide) combined since 1750 (Etminan et al 2016). Although methane is far less abundant in the atmosphere than CO 2 , it absorbs thermal infrared radiation much more efficiently and, in consequence, has a global warming potential (GWP)~86 times stronger per unit mass than CO 2 on a 20-year timescale and 28times more powerful on a 100-year time scale (IPCC 2014).
Global average methane concentrations in the atmosphere reached~1875 parts per billion (ppb) at the end of 2019, more than two-and-a-half times preindustrial levels (Dlugokencky 2020). The largest methane sources include anthropogenic emissions from agriculture, waste, and the extraction and use of fossil fuels as well as natural emissions from wetlands, freshwater systems, and geological sources (Kirschke et al 2013, Saunois et al 2016a, Ganesan et al 2019. Here, we summarize new estimates of the global methane budget based on the analysis of Saunois et al (2020) for the year 2017, the last year of the new Global Methane Budget and the most recent year data are fully available. We compare these estimates to mean values for the reference 'stabilization' period of 2000-2006 when atmospheric CH 4 concentrations were relatively stable. We present data for sources and sinks and provide insights for the geographical regions and economic sectors where emissions have changed the most over recent decades.

Methods
We use the same data and approaches to estimate CH 4 emissions as in Saunois et al (2020). One approach we use is a top-down (TD) ensemble of 11 inversions using atmospheric CH 4 concentrations to constrain total possible emissions and partition them to primary sources. The TD inversions were constrained by surface observations for the period 2000-2006, and by surface and/or satellite observations in 2017. Prior fluxes, treatment of observations, and optimization configurations varied somewhat across the 11 inversions as described in the supplementary material of Saunois et al (2020). Most of the inversions considered the same OH field, constant over time, attributing changes in methane atmospheric concentrations to altered emissions rather than to atmospheric oxidative capacity. Consequently, the inferred changes in methane emissions would be higher if OH is increasing in the atmosphere, as suggested by chemistry climate models (e.g. Zhao et al 2020) or lower if OH is decreasing in the atmosphere, as suggested by some methyl chloroformbased studies (e.g. Rigby et al 2017). Uncertainties in regional and sectoral partitioning vary across models based on transport errors, prior flux ratios, and inversion baselines. Our TD ensemble derived an estimated uncertainty of ±5% on total global emissions, a range larger than for transport model errors alone of ±2%-3% attributable to different inversion systems (Locatelli et al 2013). We were unable to include uncertainties in TD total emissions attributable to uncertainties in the methane chemical sink; uncertainty on the global burden of OH is about 10%-15% and translates to an uncertainty of approximately ±9% on total global emissions (Zhao et al 2020).
The second approach is a detailed bottom-up (BU) accounting method that uses global inventories and biogeochemical modeling that provides a more detailed attribution to sources but lacks the total atmospheric growth rate constraint that accompanies TD approaches. BU trends in methane emissions are available for anthropogenic emissions using four global inventories (EDGARv4.3.2, CEDS, GAINS and EPA2012), for biomass burning using three fire products (GFEDv4.1s, QFED, and FINN) and for wetlands calculated by 13 biogeochemical models (see Saunois et al 2020). However, estimates for other natural sources such as geological, termites, permafrost, rivers, lakes, and reservoirs available in the literature do not provide any temporal changes in methane emissions and trends cannot be calculated for these sources. Uncertainties in 'natural emissions' for wetlands plus all other inland waters arise from factors that include wetland flux density, seasonal to interannual variability in wetland extent, and some double-counting of wetland and small inland waters, contributing to higher BU estimates for natural sources than in the TD inventory. For the 2000-2017 methane budget (Saunois et al 2020), the Wetland Area Dynamics for Methane Modeling dataset (WAD2 M) was developed to avoid some double counting by removing inland waters from surface inundation data to estimate these fluxes separately, combining Landsat-based (Pekel et al 2016) and radar-based observations (Jensen and Mcdonald 2019).

Global and latitudinal sources and sinks of methane
Average estimated global methane emissions for 2017 were 596 Tg CH 4 yr −1 (figure 1, table 1) based on 11 top-down atmospheric inversions, with an ensemble max.-min. range of 572-614 Tg CH 4 yr −1 . This value is 9% (50 Tg CH 4 yr −1 ) higher than the average for the period 2000-2006 (546 Tg CH 4 yr −1 , range 538-555), with the increase attributable primarily to greater anthropogenic emission sources (table 1). Anthropogenic sources also contributed 61% of total TD global methane emissions in 2017. The estimate from the BU approach yielded an increase of 51 Tg CH 4 yr −1 , from 696 (560-834) Tg CH 4 yr −1 in 2000-2006 to 747 (602-896) Tg CH 4 yr −1 in 2017 (table 1). Anthropogenic sources contributed an estimated 51% of total global BU emissions in 2017. The difference of~150 Tg CH 4 yr −1 in total global emissions between TD and BU methods arises primarily from a divergence in estimates of natural sources, particularly from freshwater and geological ones (table 1) and from the absence of TD atmospheric constraints for BU approaches (see below).
The latitudinal attribution of methane emissions highlights the role of tropical and temperate sources relative to boreal and Arctic systems The average global atmospheric and soil methane sink estimated for 2017 increased to 571 (540-585) Tg CH 4 yr −1 from 546 (531-555) Tg CH 4 yr −1 for the 2000-2006 average based on the TD approaches. Partitioning the global methane sink into components in the atmosphere (CH 4 destruction from tropospheric OH and Cl and total stratospheric losses) and soil (microbial consumption) for 2017 yields an average TD atmospheric sink of 531 (502-540) Tg CH 4 yr −1 and an average soil sink of 40 (37-47) Tg CH 4 yr −1 (table 1).  Mean annual methane emissions rose sharply in some sectors from 2000-2006 to 2017 (figure 4). Increased agricultural emissions predominated in South Asia/Oceania, Africa, and South America, with increases of 9-10 Tg CH 4 yr −1 in South Asia/Oceania and 7-9 Tg CH 4 yr −1 in Africa (figure 4). By comparison, Europe's agricultural methane emissions decreased −1.4 to −2.8 Tg CH 4 yr −1 for TD and BU methods, respectively. Increased emissions from the fossil fuel sector were the largest in China (5.3 and 12.2 Tg CH 4 yr −1 for TD and BU, respectively) and North America, Africa, and South Asia and Oceania (4 to 6 Tg CH 4 yr −1 in all three regions and using both approaches). Fossil fuel-related methane emissions in the United States increased 3.4 to 4.0 Tg CH 4 yr −1 for TD and BU estimates, respectively, approximately    (table 1); mean bottom-up estimates were substantially higher: 367 (243-489) and 368 (245-482) Tg CH 4 yr −1 for the two periods, respectively. Natural sources remain more poorly constrained than anthropogenic ones, with divergent estimates for the bottom-up and topdown emissions. Vegetated wetlands contributed 194 (155-217) Tg CH 4 yr −1 of the total, about 83% of natural sources based on TD methods (table 1). In contrast, BU methods estimate vegetated wetland emissions to be 145 (100-183) Tg CH 4 yr −1 in 2017, a value unchanged from the 2000-2006 average but only three-quarters of the TD estimate (that also includes inland water emissions).

Natural methane sources
Wetlands and freshwater systems more broadly are the largest source of methane but also the greatest  source of uncertainty to the global methane budget. Their inclusion in BU methodologies leads to a difference of roughly 150 Tg CH 4 yr −1 when compared to the atmospheric constraint. Wetland definitions and challenges in properly understanding the location of wetlands have led to 'double counting' of inland waters and vegetated wetlands in previous studies. Our use of the Wetland Area Dynamics for Methane Modeling (WAD2M) dataset reduced the effect of double counting by~35 Tg CH 4 yr −1 compared to the previous budget in Saunois et al (2016), with vegetated wetlands accounting for 101-179 Tg CH 4 yr −1 . However, the inland waters estimate is revised to a range of 117-212 Tg CH 4 yr −1 , higher than in Saunois et al (2016) due to newer studies that measured higher emission factors in freshwater systems (Delsontro et al 2018, Saunois et al 2020. Reconciling the wetland methane emissions flux requires continued attention and the use of independent lines of data from isotopes, flux towers, and satellite observations (e.g. Knox et al 2019).
Another source of uncertainty is the amount of methane released from natural geological sources, particularly seeps and mud volcanoes. The new global BU estimate for natural geological sources (terrestrial and marine) of 45 (range of 18-65) Tg CH 4 (Etiope et al 2019, Saunois et al 2020 is 7 Tg CH 4 smaller than the value in Saunois et al (2016a). However, recent studies analyzing radiocarbon methane ( 14 CH 4 ) in ice cores have concluded that pre-industrial emissions of thermogenic (i.e. ancient or 'fossil') methane were close to zero (~0-5.4 Tg CH 4 ; Hmiel et al 2020)-substantially less even than the 15.4 Tg CH 4 estimated for the abrupt warming event that occurred between the Younger Dryas and Preboreal intervals 11 600 years ago (Petrenko et al 2017). Hmiel et al (2020) also conclude that current estimates of CH 4 emissions from the fossil fuel industry are therefore too low by 30 to 40 Tg CH 4 (Lassey et al 2007). However, the uncertainties in isotopic budget studies remain substantial due to the uncertainties in the isotopic signature of the sources.
Unlike top-down approaches, bottom-up inventories estimate activity and emissions factors separately. In contrast to the results of Hmiel et al (2020), a new annual estimate of natural methane emissions from the East Siberian Arctic Shelf alone is 3.0 Tg CH 4 , most of it thermogenic methane (Thornton et al 2020). A number as small as 5 Tg CH 4 per year for all natural geologic emissions (Hmiel et al 2020) seems difficult to reconcile with the results of Thornton et al (2020), the work of other researchers more broadly, and with BU approaches generally. Research is needed to constrain geologic sources fully.
Additional focus and monitoring is also needed to track the potential for rapid methane release from the Arctic (e.g. Post et al 2019, Zhang et al 2019. Average surface temperatures in the Arctic have risen twice as fast as the global average of 1.1 • C over the past two decades (compared to the period 1850-1900; WMO 2019). As a result of permafrost thaw and other changes in peatland ecosystems, many investigators and models predict a substantial increase in Arctic methane emissions this century. However, our latitudinal estimates from TD methods shows no evidence for the start of such a transition through year 2017 (figure 4; see also Saunois et al 2020).

Conclusions
Methane emissions have continued to rise over the past decade and are tracking concentrations most consistent with the warmest marker scenario of the Intergovernmental Panel on Climate Change (RCP8.5, a representative concentration pathway) that yields an estimated global warming of 4.3 • C by year 2100 (Saunois et al 2016b, Nisbet et al 2019. Current trajectories in socioeconomic development also suggest the world is likely to follow IPCC Shared Socioeconomic pathways (SSP) leading to relatively higher emission trajectories over the next decade (Saunois et al 2020). Estimates for 2018 and 2019 show increases in atmospheric methane of 8.5 and 10.7 ppb, respectively, two of the four highest annual growth rates since 2000 (Dlugokencky 2020).
Increased emissions from both the agriculture and waste sector and the fossil fuel sector are likely the dominant cause of this global increase (figures 1 and 4), highlighting the need for stronger mitigation in both areas. Our analysis also highlights emission increases in agriculture, waste, and fossil fuel sectors from southern and southeastern Asia, including China, as well as increases in the fossil fuel sector in the United States (figure 4). In contrast, Europe is the only continent in which methane emissions appear to be decreasing. While changes in the sink of methane from atmospheric or soil uptake remains possible (Turner et al 2019), atmospheric chemistry and land-surface models suggest the timescales for sink responses are too slow to explain most of the increased methane in the atmosphere in recent years. Climate policies overall, where present for methane mitigation, have yet to alter substantially the global emissions trajectory to date. Methane Budget activity of the Global Carbon Project (globalcarbonproject.org). The authors acknowledge additional support from the Center for Advanced Study in the Behavioral Sciences at Stanford University (RBJ), the Australian Government's National Environmental Science Programme's Earth Systems and Climate Change Hub (JGC), and Future Earth. The data that support the findings of this study are openly available at globalcarbonproject.org.

Data Availability Statement
The data that support the findings of this study are openly available at globalcarbonproject.org.