Sustained methane emissions from China after 2012 de-

spite declining coal production and rice-cultivated area

Jianxiong Sheng1 , Rachel Tunnicliffe2,3 , Anita L. Ganesan3 , Joannes D. Maasakkers4 , Lu Shen5 ,

Ronald G. Prinn1 , Shaojie Song5 , Yuzhong Zhang6,7 , Tia Scarpelli5 , A. Anthony Bloom8 , Matthew

Rigby2 , Alistair. J. Manning9 , Robert J. Parker10,11 , Hartmut Boesch10,11 , Xin Lan12,13 , Bo Zhang14 ,

Minghao Zhuang15 , Xi Lu16

1
Center for Global Change Science, Massachusetts Institute of Technology, Cambridge, MA, USA

2
School of Chemistry, University of Bristol, Bristol, UK

3
School of Geographical Sciences, University of Bristol, Bristol, UK

4
SRON Netherlands Institute for Space Research, Utrecht, The Netherlands

5
School of Engineering and Applied Science, Harvard University, Cambridge, MA, USA

6
School of Engineering, Westlake University, Hangzhou, Zhejiang, China

7
Institute of Advanced Technology, Westlake Institute for Advanced Study, Hangzhou, Zhejiang,

China

8
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA

9
Hadley Centre, Met Office, Exeter, UK

10
National Centre for Earth Observation, University of Leicester, UK

11
Earth Observation Science, School of Physics and Astronomy, University of Leicester, UK

12
Earth System Research Laboratory, National Oceanic and Atmospheric Administration, Boulder,

CO, USA

13
Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder,

1
 CO, USA

14
State Key Laboratory of Coal Resources and Safe Mining, China University of Mining & Tech-

nology (Beijing), Beijing, PR China

15
College of Resources and Environmental Sciences, China Agricultural University, Beijing, PR

China

16
School of Environment, Tsinghua University, Beijing, PR China

1 China’s anthropogenic methane emissions are the largest of any country in the world. A re-

2 cent study using atmospheric observations suggested that recent policies aimed at reducing

3 emissions of methane due to coal production in China after 2010 had been largely ineffec-

4 tive. Here, based on a longer observational record and an updated modelling approach, we

5 find that China’s growth in methane emissions did decline (0.3±0.1 (±1σ) Tg CH4 yr−2 for

6 2012-2017, as compared to 0.77±0.2 Tg CH4 yr−2 for 2010-2012). We find that the decrease

7 in growth rate after 2012 can in part be attributed to a decline in China’s coal production.

8 However, coal mine methane emissions have not declined as rapidly as production, imply-

9 ing that there may be substantial fugitive emissions from abandoned coal mines that have

10 previously been overlooked. We also find that emissions over rice-growing regions do not

11 show a negative trend (0.13±0.05 Tg CH4 yr−2 for 2010-2017) despite reports of shrinking

12 rice paddy areas, implying potentially significant emissions from new aquaculture activities,

13 which are thought to be primarily located on converted rice paddies.

2
14 Methane is the second most important anthropogenic greenhouse gas after carbon dioxide

15 and accounts for nearly 25% of radiative forcing since the pre-industrial era 1 . It has been high-

16 lighted as an important target for meeting climate policies such as the Paris Agreement2 . Global

17 methane concentrations stabilized in the atmosphere from 1999 to 2006 but growth resumed in

18 20073, 4 and still continues at some of the highest rates in the recent measurement record5 . The

19 causes of these recent changes in the atmosphere remain controversial 6–12 .

20 China is the largest anthropogenic methane emitting country in the world according to United

13
21 Nations Framework Convention on Climate Change (UNFCCC) reports . Coal mining, rice

22 cultivation, ruminant livestock, and waste management are thought to account for about 90% of

23 the country’s total methane emissions 14–16 . Previous inverse analyses (or “top-down”, atmospheric

24 data-based estimates) of satellite and surface network observations suggested that China’s annual

25 emissions grew by ∼1 Tg CH4 yr−2 from 2000 to 201017, 18 , and that this trend continued for

26 2010-201519 , primarily due to increased emissions from coal mining. However, recent bottom-

27 up inventory estimates using localized emission factors and information on coal production from

16, 20
28 China show that China’s coal mine methane (CMM) emissions have instead stabilized or

29 decreased since 2012, with coal production declining by about 10% by 2017 from the peak levels

30 in 2012/2013. There are also emerging sources such as abandoned coal mines and freshwater

31 aquaculture 21, 22 , which have not been assessed by previous inverse modeling studies.

32 Here we use eight years (2010-2017) of Greenhouse Gases Observing Satellite (GOSAT)

33 column methane observations23 and high frequency data from the National Institute for Environ-

3
34 mental Studies Japan (NIES) surface network at Cape Ochiishi and Hateruma in Japan24 to es-

35 timate methane emissions and trends in China through a regional Bayesian inverse analysis. Our

36 regional inverse approach has the benefit over previous global studies19, 25 that estimates of methane

37 emissions can be made at higher spatial resolution (crucial for accurate source attribution) and in-

38 dependent of the large uncertainties in the main sink8, 9 , atmospheric oxidation by the hydroxyl

39 radical. Source emission attribution often relies on knowledge of relative fractions of sectoral

40 emissions within model grid cells. We use state-of-the-art bottom-up inventories as the prior for

41 the inversion, which include accurate geo-coded locations of coal mines in China16 , known to

42 better than 20 km, spatially finer than the model resolution. Most previous inverse analyses for

43 China used the EDGAR v4.2 or EDGAR v4.3.2 gridded inventories15 as their prior estimates for

44 anthropogenic emissions. However, incorrect source locations for coal mining, oil, and natural gas

45 sectors have been found in the EDGAR inventories, which can bias inversion results and lead to

16, 26–30
46 erroneous source attribution in top-down estimates . The improved inventory information

47 in our inversion allows us to more accurately quantify emissions and attribute contributions from

48 different source sectors. Uncertainties in source attribution due to prior fractional information are

49 assessed by an ensemble of 1000 inversions using perturbations of the prior inventories (referred

50 to as SENSP rior , see Methods for details). The details of the prior inventories are summarized in

51 Table S1 in Supplementary Information (SI). Throughout this text, estimates are presented as the

52 mean of the “SENSP rior ” ensemble with uncertainties represented by ±1σ of the ensemble.

4
Figure 1: China’s national methane emissions over 2010-2017. The posterior estimate from this study (blue) is the

mean of the inversion ensemble (SENSP rior ) with shading corresponding to (1σ) of the ensemble. Two additional

inversions using the EDGAR v4.2 (orange) and scaled EDGAR v4.2 (light blue) inventories as the prior are also shown.

Bottom-up estimates from China’s 2014 UNFCCC report 13 , and previous top-down estimates with (1σ) uncertainties

from Miller et al.19 and Wang et al.31 are provided.

5
Figure 2: Spatial distribution of China’s methane emissions. (a) Prior methane emissions in China from invento-

ries; (b) Mean posterior emissions for 2010-2017 from this study; (c) Absolute difference between mean 2010-2017

posterior emissions and the prior; and (d) Major source sectors for grid cells with high emissions (>5 metric ton per

day) and dominated (>50%) by a single source as identified by the prior.“Other” denotes high emitting grid cells but

with no source comprising >50% of a grid cell.

6
53 China’s national emission estimates and trends

54 We estimate China’s mean annual total methane emissions over 2010-2017 to be 57.6 ±2.4 Tg CH4

55 yr−1 (of which 55.2±2.3 Tg CH4 yr−1 are anthropogenic emissions) (Fig. 1). Our posterior an-

56 thropogenic estimates are consistent with the China’s 2014 UNFCCC report 13 (55.3 Tg CH4 yr−1 )

19, 25, 31
57 and with the previous top-down estimates over similar time periods . While our estimates

58 are only 6% lower than the prior (61.2 Tg CH4 yr−1 , Fig. 1), there are large positive and negative

59 differences on a sub-national/provincial scale that tend to cancel each other on the national scale

60 (Fig. 2).

61 We find that China’s emissions exhibit an increase during 2010-2012 (inclusive, 0.77±0.2

62 Tg CH4 yr−2 ), but the trend slows after 2012 to 0.3±0.1 Tg CH4 yr−2 for 2012-2017. Our trend

63 estimate for 2010-2015 (0.5±0.2 Tg CH4 yr−2 ) is smaller than 1.1±0.4 Tg CH4 yr−2 suggested by

64 a previous study19 for the same time period. This is discussed further below.

65 Emissions and their trends for major source sectors

66 Fractions of each source in the prior emission inventories can be used to apportion emissions and

67 emission trends to source sectors in countries such as China where the different sources have dis-

68 tinct spatial or temporal distributions29, 32 . Coal mining, rice cultivation, livestock, and waste are

69 dominant sources for China’s anthropogenic emissions. CMM emissions are mainly in Shanxi

70 province and Southwest China, whereas emissions from rice paddies dominate in Southeast China.

71 Livestock emissions are mainly concentrated in a number of provinces in north-central and South-

7
72 west China. Waste emissions are localized in urban areas. Wetland emissions are small in China

73 and only dominate a few places in Southeast China and North China. Emissions from other an-

74 thropogenic sources and natural sources are too small to dominate any grid cells. Regions having

75 mixed sources (i.e., no source comprising > 50% of a grid cell) are mainly in Sichuan and Shan-

76 dong provinces. Overall, the locations dominated by these major source sectors are well separated

77 at the 0.352◦ × 0.234◦ model spatial resolution and account for 70% of the total emissions from

78 China (Fig. 2d).

79 Coal mining is the largest contributor to the total methane emissions in China with mean

80 2010-2017 emissions of 14.5 ±1.4 Tg CH4 yr−1 (Table S1). The second largest source in China

81 is rice cultivation with mean emissions of 13.2 ±1.0 Tg CH4 yr−1 , followed by livestock at 11.2

82 ±1.1 Tg CH4 yr−1 and waste at 10.2 ±0.8 Tg CH4 yr−1 (Table S1). Trends for these sectors are

83 discussed below, with other minor source sectors showing no significant trends (see Fig. S5 in SI).

84 CMM emissions increased by 0.7±0.3 Tg CH4 yr−2 from 2010 to 2012, driving the national

85 trend, but flattened afterwards with a smaller growth of 0.1 ± 0.06 Tg CH4 yr−2 for 2012-2017

86 (Fig. 3a). According to activity data, coal production peaked in 2013, and in 2016/2017 returned

87 to levels similar to those of 201035 (Fig. 3b). The derived emission trend is consistent with coal

88 production activities for 2010-2012 (Fig. 3b), but deviates after that. This inconsistency between

89 top-down estimates and production may be due to emissions from abandoned coal mines. Since

90 2010/2011, China has consolidated its coal industry to concentrate production in the existing larger

91 and more efficient coal mines (typically state-run mines), and to gradually close a large number of

8
Figure 3: China’s methane emissions and emitting activities since 2010 for 4 major source sectors. (a,b) coal

methane emissions and mining activities (production and number of active mines); (c,d) rice-aquaculture methane

emissions and activities (areas for rice paddies and aquaculture ponds); (e,f) waste methane emissions and activities

(the amount of solid waste at landfills and the number of waste water treatment plants); (g,h) livestock methane

emissions and activities (cattle population). Shaded areas denote ±1σ uncertainties from the posterior ensemble

(“SENSprior ”). Our inversion results using EDGAR v4.2 as the prior are also shown in addition to previous results

from Miller et al. 19 . Freshwater aquaculture areas are from Bureau of Fisheries China33 . Activity data for waste

water treatment plants are from Zhao et al.34 . Other activity data are from the National Bureau of Statistics of China
9
35 .
92 small, village or town-owned coal mines16, 34 . Our CMM prior inventory includes 11,000 operating

93 coal mines in 2010/2011, of which about 4000 mines have been abandoned over 2010-201716 (Fig.

94 3b) but are still allocated in the prior inventory for the inversions. Therefore the derived trend

95 over coal mine fields comprises the contributions from abandoned mines since 2011. An actively

96 venting abandoned mine can emit methane up to 40-90% of its initial rate in the first 3-4 years,

36
97 shrinking to 10% after 30 years . Increasing emission factors for active coal mines may also

98 explain the inconsistency, but reported recovery rates of CMM have been increasing 16, 20, 37 . Thus

99 emission factors are more likely to have decreased or remained stable, which would suggest about

100 1.7 Tg CH4 yr−1 coming from abandoned coal mines in 2017 assuming emissions in 2017 from

101 active coal mining are similar to those in 2010.

102 Our CMM emission trend differs from that derived in the previous inverse analysis19 , which

103 shows a continuous increase after 2012 1.0±0.3 Tg CH4 yr−2 (Fig. 3a). We propose that this dif-

104 ference is due to two factors related to the spatial pattern of emissions in the prior. First, the trend

105 in total emissions is higher after 2013. When we use the EDGAR v4.2 as the prior in our inversion

106 rather than the China-specific coal inventory 16 , we derive a similar larger trend after 2012 in the

107 total emissions (Fig. 1). The magnitude of prior CMM emissions do not significantly influence the

108 posterior trend as indicated by our inversion results using scaled-down EDGAR v4.2 coal emis-

109 sions (Fig. 1). This finding indicates that inaccuracies in the spatial distribution of the coal sector

110 in EDGAR can to some extent lead to errors in the derived trend in total emissions. Second, the

111 EDGAR v4.2 inventory at coarse resolution (2◦ ×2.5◦ ) that were used by the previous global inver-

112 sion 19 (Fig. S6 in SI), show that grid cells dominated by coal or mixed sources account for about

10
113 85% of the total emissions, in contrast to the high-resolution China-specific inventory used here

114 which has 28% of total emissions from coal dominated grid cells. The strongly differing emission

115 patterns between the two inventories remain even at higher spatial resolution (Fig. S7 in SI). This

116 would lead to a higher apportionment of the total trend to the CMM sector and consequently affect

117 the trends of other source sectors (Figs. 3c,e,g). These two factors together may explain the larger

118 derived trend in CMM emissions in the previous work, compared to our estimates.

119 The emission estimates from rice paddies have increased by 0.13±0.05 Tg CH4 yr−2 from

120 2010 to 2017. Rice emissions are proportional to paddy area38 , but the trend in derived emissions

121 is opposite to that of reported paddy area after 2013 (2% decline from 2013 to 201735 ). The decline

122 of rice paddy areas is largely related to ongoing conversion of paddy fields towards industrial-scale

123 aquaculture 21 . Freshwater aquaculture in China has emission factors about 4 times larger than rice

124 cultivation and is estimated to emit 3.5 Tg CH4 yr−1 based on the total area of aquaculture ponds

21
125 , but has been overlooked in previous gridded bottom-up inventories (such as EDGAR v4.3 we

126 used here for rice) and inverse analyses. The freshwater aquaculture industry is primarily located

127 in the southeast of China 39 and more than half of aquaculture facilities been converted from rice

128 paddies21 . Therefore the estimated emissions from rice-growing regions informed by the prior

129 include contributions from rice paddies, aquaculture ponds converted from rice paddies, and to a

130 large extent existing aquaculture ponds. China’s total area of freshwater aquaculture ponds has

131 increased more than 20% from 2010 to 201735 (Fig. 3d). This could explain the opposite trends

132 between emissions and rice paddy areas.

11
133 Emissions from waste management inferred by the inversion exhibit a positive trend of 0.06±

134 0.04 Tg CH4 yr−2 between 2010 and 2017 (Fig. 3e). Landfill methane emissions are partly related

135 to the amount and type of solid waste and management at landfills40 . Due to increased urbanization,

136 the amount to solid waste added per year in China has increased by 50% from 2010 to 2017, and

137 its trend accelerates after 201435 (Fig. 3f). In addition, the number of wastewater facilities has

138 increased by 40% from 2010 to 2014 based on the most recent available data 34 (Fig. 3f), which

139 is potentially a significant source contributing to the waste trend. However we are not able to

140 distinguish contributions from landfills and wastewater because they are both in urban areas and

141 largely co-located at ∼30 km spatial resolution.

142 China’s livestock emissions do not exhibit a significant trend (0.01±0.03 Tg CH4 yr−2 ) over

143 2010–2017 (Fig. 3g). Livestock emissions include enteric fermentation and manure management.

144 The former accounts for more than 90% of China’s total livestock emissions and is linearly cor-

145 related to cattle population, which does not show a significant trend over 2010-2017 but suggests

146 a small decrease between 2010-2012 (Fig. 3h). The emission trend from livestock is expected to

147 remain stable in the future because the growing meat consumption in China is due to be met by

148 increasing imports 35 .

149 Effective climate policies aimed at limiting these emissions will rely on accurate estimates

150 and robust source attribution. In conclusion, our inverse analysis suggests that China’s annual

151 methane emissions continue to increase from 2010 to 2017 but we find that the rate of increase has

152 slowed to 0.3±0.1 Tg CH4 yr−2 for 2012-2017, as compared to 0.77±0.2 Tg CH4 yr−2 from 2010-

12
153 2012. We show that emissions from China’s coal mining and rice-growing regions have remained

154 positive after 2012 despite a decrease in coal production and in the area harvested for rice. We

155 propose that this may be due to sources such as fugitive emissions from abandoned coal mines and

156 the growth in aquaculture systems in previous rice-growing areas. These sources have not been

157 widely considered in previous national-scale studies of emissions from China. However, our work

158 suggests that they should be carefully considered in any future emission mitigation efforts, as they

159 may have had, and will likely continue to have, a substantial influence on China’s overall methane

160 emissions trends.

161 Methods

162 Observations. We use the version 7.2 proxy nadir retrievals of GOSAT methane column data from

163 the University of Leicester 23 in our inverse analysis. GOSAT retrieves the atmospheric methane

164 column by nadir measurements of solar backscatter (1.65 µm absorption band) 41 . Observations are

165 made at three circular pixels of 10 km diameter across the orbit track 260 km apart, separated by

166 260 km along the track. The same locations are sampled every 3 days. In China, GOSAT retrieves

167 more data in the west, where there are less clouds. The number of GOSAT retrievals over a given

168 location is similar for different seasons, though it is slightly larger during October - December in

169 China (see Fig. S8 in SI).

170 We also include ground-based hourly measurements at (43.2◦ N, 145.5◦ E, 96 m above sea

171 level) and Hateruma (24.1◦ N, 123.8◦ E, 46.5 m above sea level) stations in Japan operated by the

172 Institute for Environmental Studies (NIES), which have been described in detail by Tohjima et

13
 24
173 al. . The two stations are relatively insensitive to China’s emissions, but help to improve the

174 constraints to boundary conditions.

175 Inversion framework. We use the UK Met Office NAME (Numerical Atmospheric dispersion

176 Modelling Environment) model, a Lagrangian particle dispersion model42 , as the atmospheric

177 transport model used to provide the relationship between emissions and concentrations in the at-

178 mosphere. The model domain is 54◦ E - 170◦ W, 5◦ S - 84◦ N. We derive an optimized estimate

179 of spatially resolved methane emissions in China using Bayesian inverse analysis. The inversion

180 minimizes the cost function J(x) by solving ∇x J(x) = 0, with J(x) defined as follows:

1 1
J(x) = (x − xprior )T P−1 (x − xprior ) + (y − Hx)T R−1 (y − Hx) (1)
2 2

181 Here y is the vector of observations, H is the Jacobian matrix representing the sensitivities of

182 observations to changes in the state vector x and xprior is the the prior value of x. P is the prior

183 error covariance matrix and R is the observational error covariance matrix.

184 The state vector contains 250 elements for aggregated methane emissions (200 elements for

185 China) within the model domain using a Gaussian Mixture Model (GMM) with radial basis func-

186 tions based on spatial proximity and source type patterns43 . The use of the GMM enables us to

187 retain high resolution for major sources up to native model resolution while coarsening resolution

188 for weak or broadly distributed sources. The state vector also includes additional 4 elements for

189 boundary concentrations at the 4 edges of the model domain and 1 element for an offset parameter

190 between satellite and surface data. The prior boundary conditions are from a 4◦ × 5◦ global GEOS-

191 Chem simulation using methane emissions optimized with GOSAT satellite data by Maasakkers

14
192 et al.25 . The offset parameter (20 ppb as prior) accounts for any systematic differences between

193 the GOSAT and the NIES ground-based measurements and their representations by the model 32 .

194 We assume 100% prior error for each emission state vector element without correlation (i.e. off-

195 diagonal elements in P are zero), 2% prior error for boundary conditions, and 50% prior error for

196 the offset parameter. Observational errors include GOSAT instrument, model transport, and repre-

197 sentation errors. We use 10 ppb as the mean observational error (model error + instrument error)

198 standard deviation derived by previous NAME inverse modeling against ground-based observa-

44 23
199 tions . GOSAT mean instrument error standard deviation is 11 ppb , indicating that satellite

200 instrument error dominates the observational error. For a given satellite observation, we take the

201 maximum of the reported instrument error and NAME model error as the observational error.

202 The sensitivity matrix H is computed by NAME. We release particles at a rate of 2000

203 particles per hour for each GOSAT vertical level, over a one minute period centered around the

204 retrieval time, and trace them back in time for 30 days, as done by previous studies 32, 45 . For surface

205 sites, we release 20000 particles at the release height of the stations (100 m and 50 m above sea

206 level for Cape Ochiishi and Hateruma, respectively). The model is driven by the Unified Model’s

207 Model meteorology 46 with horizontal resolution spanning from 0.352◦ ×0.234◦ to 0.141◦ ×0.094◦

208 over 2010-2017. The model output is set to be 0.352◦ × 0.234◦ for the inversion to be consistent

209 with the lowest meteorological resolution used. NAME has been used extensively to calculate

210 sensitivity matrices for inverses analyses of long-lived greenhouse gases and ozone-depleting gases

32, 44, 45, 47–49

212 We perform the inversion at monthly resolution. We optimize the logarithms of the emis-

15
213 sion to ensure positivity. Under the lognormal assumption, the inverse problem is non-linear and

214 can be solved numerically by the Levenberg-Marquardt iterative algorithm 25 with a convergence

215 tolerance of 0.05% for maximum relative error:

xn+1 = xn + (R−1 + HTn P−1 Hn )−1 (HTn P−1 Hn )−1 (y − Hxn ) + R−1 (xn − xprior ) (2)

216 where Hn = ∂y/∂(ln x) is the Jacobian matrix at the n-th iteration. Each individual element

217 ∂yi /∂ ln xj of Hn can be recalculated in the iteration by ∂yi /∂ ln xj = xj ∂yi /∂xj with ∂yi /∂xj

218 being individual elements of H for linear problems computed by the forward model.

219 Sensitivity tests. To test the inversion sensitivity to the prior boundary conditions, we performed

220 100 inversions by randomly perturbing the boundary conditions on each grid cell at four domain

221 edges (referred to as “SENSBC ”, Fig. S3). The perturbations followed a uniform distribution

222 within ±2%. We also performed a sensitivity test on the model XCO2 by using a different GOSAT

223 proxy product from the RemoTeC v2.3.9, which used the CarbonTracker Model to model XCO2 50 ,

224 while the University of Leicester GOSAT proxy product used in our inversion is based on an

225 ensemble of model XCO2 data23 (Fig. S4). These sensitivity tests show that inversion results

226 are generally insensitive to different GOSAT proxy product and random perturbations on prior

227 boundary conditions.

228 To evaluate the inversion sensitivity to the prior fractional information, we perform an en-

229 semble of inversions using 1000 different prior estimates generated by perturbing the baseline prior

230 (this ensemble is referred to as “SENSP rior ”). For each ensemble member, we first randomly select

231 a source from coal, rice, livestock, waste, oil/gas, wetlands, or other sectors within a given grid

16
232 cell. We then perturb the relative fraction of the selected sector in that grid cell following a uni-

233 form distribution (±20%). This is done by decreasing or increasing the selected sectoral emissions

234 accordingly in the given grid cell. We neglect grid cells that have small emissions less than 0.5

235 ton CH4 per day. We do not perturb source sectors that have a fraction of 0 or 1 in a grid cell (i.e.

236 have no source emissions or dominate the grid cell completely). An example of perturbed sector

237 fractions is shown in Fig. S9 in SI. This sensitivity test allows us to assess the effect on source

238 attribution due to uncertainties in relative sector fractions within grid cells.

239 We also perform two additional inversions using the EDGAR v4.2 inventory and a scaled

240 EDGAR v4.2 inventory as the prior. The latter scales coal mining emissions in EDGAR v4.2

241 from 28 Tg CH4 yr−1 to 16.7 Tg CH4 yr−1 to match the total emissions of coal mining in the

242 China-specific coal inventory 16 .

243 Acknowledgments

244 This work was supported by NASA Grant NNX16AC98G to the Massachusetts Institute of Tech-

245 nology. The authors thank Michael B. McElroy, J. William Munger, and Chris Nielsen for helpful

246 discussions. The authors also thank Yasunori Tohjima for providing the data at Cape Ochiishi

247 and Hateruma. Part of this research was carried out at the Jet Propulsion Laboratory, California

248 Institute of Technology, under a contract with the National Aeronautics and Space Administration.

249 RT was funded by the Newton Fund through the Met Office Climate Science for Service Partner-

250 ship Brazil (CSSP Brazil) and through UK Natural Environment Reseach Council (NERC) grant

251 NE/N016548/1. ALG was funded by NERC Independent Research Fellowship NE/L010992/1.

17
252 RJP and HB are funded via the UK National Centre for Earth Observation (NE/R016518/1 and

253 NE/N018079/1). We thank the Japanese Aerospace Exploration Agency, National Institute for En-

254 vironmental Studies, and the Ministry of Environment for the GOSAT data and their continuous

255 support as part of the Joint Research Agreement. This research used the ALICE High Performance

256 Computing Facility at the University of Leicester for the GOSAT retrievals.

258
257 1. Myhre, G. et al. Anthropogenic and natural radiative forcing. Climate change 423 (2013).

259 2. Ganesan, A. L. et al. Advancing Scientific Understanding of the Global Methane Bud-

260 get in Support of the Paris Agreement. Global Biogeochemical Cycles 33, 1475–1512

261 (2019). URL https://agupubs.onlinelibrary.wiley.com/doi/abs/10.

262 1029/2018GB006065.

263 3. Rigby, M. et al. Renewed growth of atmospheric methane. Geophys. Res. Lett.

264 35, L22805 (2008). URL http://onlinelibrary.wiley.com/doi/10.1029/

265 2008GL036037/abstract.

266 4. Dlugokencky, E. J. et al. Observational constraints on recent increases in the atmospheric

267 CH4 burden. Geophys. Res. Lett. 36, L18803 (2009). URL http://onlinelibrary.

268 wiley.com/doi/10.1029/2009GL039780/abstract.

269 5. Nisbet, E. G. et al. Very Strong Atmospheric Methane Growth in the 4 Years 2014–2017:

270 Implications for the Paris Agreement. Global Biogeochemical Cycles 33, 318–342

271 (2019). URL https://agupubs.onlinelibrary.wiley.com/doi/abs/10.

272 1029/2018GB006009.

18
273 6. Helmig, D. et al. Reversal of global atmospheric ethane and propane trends largely

274 due to US oil and natural gas production. Nature Geosci advance online publica-

275 tion (2016). URL http://www.nature.com/ngeo/journal/vaop/ncurrent/

276 full/ngeo2721.html.

277 7. Schaefer, H. et al. A 21st-century shift from fossil-fuel to biogenic methane emissions indi-

278 cated by 13CH4. Science 352, 80–84 (2016). URL http://science.sciencemag.

279 org/content/352/6281/80.

280 8. Rigby, M. et al. Role of atmospheric oxidation in recent methane growth. PNAS 114(21),

281 5373–5377 (2017). URL http://www.pnas.org/content/early/2017/04/11/

282 1616426114.

283 9. Turner, A. J., Frankenberg, C., Wennberg, P. O. & Jacob, D. J. Ambiguity in the causes for

284 decadal trends in atmospheric methane and hydroxyl. PNAS 114, 5367–5372 (2017). URL

285 http://www.pnas.org/content/early/2017/04/18/1616020114.

286 10. Worden, J. R. et al. Reduced biomass burning emissions reconcile conflicting estimates of

287 the post-2006 atmospheric methane budget. Nature Communications 8, 2227 (2017). URL

288 https://www.nature.com/articles/s41467-017-02246-0.

289 11. Turner, A. J., Frankenberg, C. & Kort, E. A. Interpreting contemporary trends in atmospheric

290 methane. PNAS 116, 2805–2813 (2019). URL https://www.pnas.org/content/

291 116/8/2805.

19
292 12. Saunois, M. et al. The Global Methane Budget 2000–2017. Earth System Science Data

293 Discussions 1–136 (2019). URL https://www.earth-syst-sci-data-discuss.

294 net/essd-2019-128/. Publisher: Copernicus GmbH.

295 13. UNFCCC. Greenhouse Gas Inventory Data Interface, available at:

296 http://di.unfccc.int/detailed data by party (2020).

297 14. Peng, S. et al. Inventory of anthropogenic methane emissions in mainland China from

298 1980 to 2010. Atmos. Chem. Phys. 16, 14545–14562 (2016). URL http://www.

299 atmos-chem-phys.net/16/14545/2016/.

300 15. Janssens-Maenhout, G. et al. EDGAR v4.3.2 Global Atlas of the three major Green-

301 house Gas Emissions for the period 1970–2012. Earth System Science Data Discus-

302 sions 1–52 (2019). URL https://www.earth-syst-sci-data-discuss.net/

303 essd-2018-164/.

304 16. Sheng, J., Song, S., Zhang, Y., Prinn, R. G. & Janssens-Maenhout, G. Bottom-Up Estimates

305 of Coal Mine Methane Emissions in China: A Gridded Inventory, Emission Factors, and

306 Trends. Environ. Sci. Technol. Lett. 6, 473–478 (2019). URL https://doi.org/10.

307 1021/acs.estlett.9b00294. Publisher: American Chemical Society.

308 17. Bergamaschi, P. et al. Atmospheric CH4 in the first decade of the 21st century: Inverse mod-

309 eling analysis using SCIAMACHY satellite retrievals and NOAA surface measurements. J.

310 Geophys. Res. Atmos. 118, 7350–7369 (2013). URL http://onlinelibrary.wiley.

311 com/doi/10.1002/jgrd.50480/abstract.

20
312 18. Thompson, R. L. et al. Methane emissions in East Asia for 2000–2011 estimated using an

313 atmospheric Bayesian inversion. Journal of Geophysical Research: Atmospheres 120, 4352–

314 4369 (2015). URL https://agupubs.onlinelibrary.wiley.com/doi/abs/

315 10.1002/2014JD022394.

316 19. Miller, S. M. et al. China’s coal mine methane regulations have not curbed growing emis-

317 sions. Nature Communications 10, 303 (2019). URL https://www.nature.com/

318 articles/s41467-018-07891-7.

319 20. Zhu, T., Bian, W., Zhang, S., Di, P. & Nie, B. An Improved Approach to Estimate Methane

320 Emissions from Coal Mining in China. Environ. Sci. Technol. 51, 12072–12080 (2017). URL

321 https://doi.org/10.1021/acs.est.7b01857.

322 21. Yuan, J. et al. Rapid growth in greenhouse gas emissions from the adoption of industrial-scale

323 aquaculture. Nature Climate Change 9, 318–322 (2019). URL https://www.nature.

324 com/articles/s41558-019-0425-9.

325 22. Gao, J., Guan, C. & Zhang, B. China’s CH4 emissions from coal mining: A review of current

326 bottom-up inventories. Science of The Total Environment 725, 138295 (2020). URL http:

327 //www.sciencedirect.com/science/article/pii/S0048969720318088.

328 23. Parker, R. J. et al. Assessing 5 years of GOSAT Proxy XCH4 data and associated uncertain-

329 ties. Atmos. Meas. Tech. 8, 4785–4801 (2015). URL http://www.atmos-meas-tech.

330 net/8/4785/2015/.

21
331 24. Tohjima, Y. et al. Temporal changes in the emissions of CH4 and CO from China estimated

332 from CH4 / CO2 and CO / CO2 correlations observed at Hateruma Island. Atmospheric Chem-

333 istry and Physics 14, 1663–1677 (2014). URL https://www.atmos-chem-phys.

334 net/14/1663/2014/. Publisher: Copernicus GmbH.

335 25. Maasakkers, J. D. et al. Global distribution of methane emissions, emission trends, and

336 OH concentrations and trends inferred from an inversion of GOSAT satellite data for

337 2010–2015. Atmospheric Chemistry and Physics 19, 7859–7881 (2019). URL https:

338 //www.atmos-chem-phys.net/19/7859/2019/. Publisher: Copernicus GmbH.

339 26. Jacob, D. J. et al. Satellite observations of atmospheric methane and their value for quantifying

340 methane emissions. Atmos. Chem. Phys. 16, 14371–14396 (2016). URL http://www.

341 atmos-chem-phys.net/16/14371/2016/.

342 27. Maasakkers, J. D. et al. Gridded National Inventory of U.S. Methane Emissions. Environ. Sci.

343 Technol. 50, 13123–13133 (2016). URL http://dx.doi.org/10.1021/acs.est.

344 6b02878.

345 28. Sheng, J.-X. et al. A high-resolution (0.1x0.1) inventory of methane emissions from

346 Canadian and Mexican oil and gas systems. Atmospheric Environment 158, 211–

347 215 (2017). URL http://www.sciencedirect.com/science/article/pii/

348 S1352231017301164.

349 29. Sheng, J.-X. et al. High-resolution inversion of methane emissions in the Southeast US using

350 SEAC4RS aircraft observations of atmospheric methane: anthropogenic and wetland sources.

22
351 Atmos. Chem. Phys. 18, 6483–6491 (2018). URL https://www.atmos-chem-phys.

352 net/18/6483/2018/.

353 30. Scarpelli, T. R. et al. A global gridded (0.1 0.1) inventory of methane emissions from oil, gas,

354 and coal exploitation based on national reports to the United Nations Framework Convention

355 on Climate Change. Earth System Science Data 12, 563–575 (2020). URL https://www.

356 earth-syst-sci-data.net/12/563/2020/.

357 31. Wang, F. et al. Methane Emission Estimates by the Global High-Resolution Inverse Model

358 Using National Inventories. Remote Sensing 11, 2489 (2019). URL https://www.mdpi.

359 com/2072-4292/11/21/2489.

360 32. Tunnicliffe, R. L. et al. Quantifying sources of Brazil’s CH4 emissions between 2010 and

361 2018 from satellite data. Atmospheric Chemistry and Physics Discussions 1–40 (2020).

362 URL https://www.atmos-chem-phys-discuss.net/acp-2020-438/. Pub-

363 lisher: Copernicus GmbH.

364 33. Bureau of Fisheries - Ministry of Agriculture, China. China Fishery Statistics Yearbooks

365 (2011-2017) (Agriculture Express, Beijing, 2018).

366 34. Zhao, X. et al. China’s Urban Methane Emissions From Municipal Wastewater

367 Treatment Plant. Earth’s Future 7, 480–490 (2019). URL https://agupubs.

368 onlinelibrary.wiley.com/doi/abs/10.1029/2018EF001113. eprint:

369 https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2018EF001113.

23
370 35. National Bureau of Statistics of China. China Statistical Yearbook 2019. Tech. Rep., China

371 Statistics Press (2020).

372 36. US EPA. Methane Emissions from Abandoned Coal Mines in the United States: Emis-

373 sion Inventory Methodology and 1990-2002 Emissions Estimates 90 (2004). URL

374 https://www.epa.gov/sites/production/files/2016-03/documents/

375 amm_final_report.pdf.

376 37. Gao, J., Guan, C. & Zhang, B. China’s CH4 emissions from coal mining: A review of current

377 bottom-up inventories. Science of The Total Environment 725, 138295 (2020). URL http:

378 //www.sciencedirect.com/science/article/pii/S0048969720318088.

379 38. Eggleston, H. S., Buendia, L., Miwa, K., Ngara, T. & Tanabe, K. 2006 IPCC Guide-

380 lines for National Greenhouse Gas Inventories (2006). URL https://www.osti.gov/

381 etdeweb/biblio/20880391.

382 39. Wang, J., Beusen, A. H. W., Liu, X. & Bouwman, A. F. Aquaculture Production is a Large,

383 Spatially Concentrated Source of Nutrients in Chinese Freshwater and Coastal Seas. Environ.

384 Sci. Technol. 54, 1464–1474 (2020). URL https://doi.org/10.1021/acs.est.

385 9b03340. Publisher: American Chemical Society.

386 40. Cai, B. et al. CH4 mitigation potentials from China landfills and related environmen-

387 tal co-benefits. Science Advances 4, eaar8400 (2018). URL https://advances.

388 sciencemag.org/content/4/7/eaar8400. Publisher: American Association for

389 the Advancement of Science Section: Research Article.

24
390 41. Kuze, A. et al. Update on GOSAT TANSO-FTS performance, operations, and data products

391 after more than 6 years in space. Atmos. Meas. Tech. 9, 2445–2461 (2016). URL http:

392 //www.atmos-meas-tech.net/9/2445/2016/.

393 42. Jones, A., Thomson, D., Hort, M. & Devenish, B. The U.K. Met Office’s Next-Generation At-

394 mospheric Dispersion Model, NAME III. In Borrego, C. & Norman, A.-L. (eds.) Air Pollution

395 Modeling and Its Application XVII, 580–589 (Springer US, Boston, MA, 2007).

396 43. Turner, A. J. & Jacob, D. J. Balancing aggregation and smoothing errors in inverse models.

397 Atmos. Chem. Phys. 15, 7039–7048 (2015). URL http://www.atmos-chem-phys.

398 net/15/7039/2015/.

399 44. Manning, A. J., O’Doherty, S., Jones, A. R., Simmonds, P. G. & Derwent, R. G. Estimating

400 UK methane and nitrous oxide emissions from 1990 to 2007 using an inversion modeling

401 approach. Journal of Geophysical Research: Atmospheres 116 (2011). URL https://

402 agupubs.onlinelibrary.wiley.com/doi/abs/10.1029/2010JD014763.

403 eprint: https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2010JD014763.

404 45. Ganesan, A. L. et al. Atmospheric observations show accurate reporting and little growth in

405 India’s methane emissions. Nature Communications 8, 836 (2017). URL https://www.

406 nature.com/articles/s41467-017-00994-7.

407 46. Walters, D. N. et al. The Met Office Unified Model Global Atmosphere 4.0 and

408 JULES Global Land 4.0 configurations. Geoscientific Model Development 7, 361–

25
409 386 (2014). URL https://gmd.copernicus.org/articles/7/361/2014/

410 gmd-7-361-2014.html. Publisher: Copernicus GmbH.

411 47. Rigby, M., Manning, A. J. & Prinn, R. G. Inversion of long-lived trace gas emissions us-

412 ing combined Eulerian and Lagrangian chemical transport models. Atmospheric Chemistry

413 and Physics 11, 9887–9898 (2011). URL https://www.atmos-chem-phys.net/

414 11/9887/2011/.

415 48. Ganesan, A. L. et al. Characterization of uncertainties in atmospheric trace gas inversions

416 using hierarchical Bayesian methods. Atmos. Chem. Phys. 14, 3855–3864 (2014). URL

417 http://www.atmos-chem-phys.net/14/3855/2014/.

418 49. Rigby, M. et al. Increase in CFC-11 emissions from eastern China based on atmospheric obser-

419 vations. Nature 569, 546–550 (2019). URL https://www.nature.com/articles/

420 s41586-019-1193-4.

421 50. Schepers, D. et al. Methane retrievals from Greenhouse Gases Observing Satellite (GOSAT)

422 shortwave infrared measurements: Performance comparison of proxy and physics retrieval

423 algorithms. J. Geophys. Res. 117, D10307 (2012). URL http://onlinelibrary.

424 wiley.com/doi/10.1029/2012JD017549/abstract.

26
Supplementary Information for “Sustained methane emis-
sions from China after 2012 despite declining coal pro-
duction and rice-cultivated area”

Jianxiong Sheng1 , Rachel Tunnicliffe2,3 , Anita L. Ganesan3 , Joannes D. Maasakkers4 , Lu Shen5 ,

Ronald G. Prinn1 , Shaojie Song5 , Yuzhong Zhang6,7 , Tia Scarpelli5 , A. Anthony Bloom8 , Matthew

Rigby2 , Alistair. J. Manning9 , Robert J. Parker10,11 , Hartmut Boesch10,11 , Xin Lan12,13 , Bo Zhang14 ,

Minghao Zhuang15 , Xi Lu16

1
Center for Global Change Science, Massachusetts Institute of Technology, Cambridge, MA, USA

2
School of Chemistry, University of Bristol, Bristol, UK

3
School of Geographical Sciences, University of Bristol, Bristol, UK

4
SRON Netherlands Institute for Space Research, Utrecht, The Netherlands

5
School of Engineering and Applied Science, Harvard University, Cambridge, MA, USA

6
School of Engineering, Westlake University, Hangzhou, Zhejiang, China

7
Institute of Advanced Technology, Westlake Institute for Advanced Study, Hangzhou, Zhejiang,

China

8
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA

9
Hadley Centre, Met Office, Exeter, UK

10
National Centre for Earth Observation, University of Leicester, UK

11
Earth Observation Science, School of Physics and Astronomy, University of Leicester, UK

12
Earth System Research Laboratory, National Oceanic and Atmospheric Administration, Boulder,

CO, USA

1
13
Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder,

CO, USA

14
State Key Laboratory of Coal Resources and Safe Mining, China University of Mining & Tech-

nology (Beijing), Beijing, PR China

15
College of Resources and Environmental Sciences, China Agricultural University, Beijing, PR

China

16
School of Environment, Tsinghua University, Beijing, PR China

2
Validation of inversion results

To assess inversion performance, we simulate methane concentrations using posterior estimates

from the inversion. The posterior estimates reduce the root mean squared errors (RMSE) with

GOSAT and surface observations at Hateruma and Cape Orchiishi from 29-69 ppb using the prior

to 6-26 ppb, and present no significant trends in the residuals (Figs. S1 and S2 in SI). The mean

posterior offset parameter between satellite and surface measurements is 19.8 ppb (see Methods.

This term includes contributions from both model and measurement biases). We also verify the

inversion results by comparing modeled methane concentrations with independent measurements

(not used in the inversion) from the NOAA’s Cooperative Global Air Sampling Network8 . The

posterior estimates improve the RMSEs at all five NOAA flask stations in or close to China and

show no significant trends in the residuals (see Figs. S1 and S2 in SI). Sensitivity tests to the prior

boundary conditions and different proxy XCH4 data also verify the robustness of our inversion

results (see Figs. S3 and S4 in SI).

3
 Table S1: China’s prior and posterior methane emissions with 1σ uncertainties.

Source Sector Prior (Tg CH4 yr−1 ) Posterior (Tg CH4 yr−1 )

Fossil Fuels 17.8 15.6 ±1.5

Coal mining 16.7 14.5 ±1.4

Oil and natural gas 1.1 1.1±0.4

Rice Cultivation 14.1 13.2 ±1.0

Livestock 11.7 11.2 ±1.1

Waste Management 10.7 10.2 ±0.8

Wetlands 2.4 2.3±0.4

Other 4.6 5.2±0.6

Total 61.2 57.6±2.4

Coal mining is from Sheng et al 1 for the year 2010/2011. Oil and natural gas are from Scarpelli et al. 2

for 2016. Rice, livestock, and waste management emissions are from EDGAR v4.3.2 for 2012 3 . Wetland

emissions are mean values for 2010-2017 from the WetCHARTS ensemble 4 . “Other” including combustion,

industrial processes, agricultural field burning, composting from EDGAR v4.3.2 for 2012 3 as well as open

fires emissions from the Quick Fire Emissions Dataset (QFED5 ). Methane seasonality due to rice cultivation

and manure management in China is taken into account by applying seasonal scaling factors from Yan et

al. 6 and Maasakker et al. 7 , respectively.

4
Figure S1: Differences between observations and modeled XCH4 or CH4 using the prior and posterior estimates.

Also shown are the locations of Hateruma, Cape Ochiishi, and the five independent NOAA flask stations (bottom-

right panel). GOSAT and modeled XCH4 are averaged every 3 days for clarity in the figure. The posterior offset

parameter between satellite and surface measurements is 19.8±14.1 (±1σ) ppb.

5
Figure S2: Observations and modeled XCH4 or CH4 using the prior and posterior estimates. GOSAT and modeled

XCH4 are averaged every 3 days for clarity in the figure.

6
Figure S3: China’s methane emissions from major source sectors inferred from the inversions using the University of

Leicester v7.2 GOSAT proxy data (UoLv72) and RemoTec v2.3.9 proxy data. XCH4 in UoLv72 proxy data is based on

an ensemble of model XCO2 9 , while XCO2 used in “RemoTec v2.3.9” proxy data is derived from the CarbonTracker

Model10 . The two data products also differ in their inversion methods and treatments of aerosols in their forward

models.

7
Figure S4: China’s methane emissions from major source sectors inferred from the inversion using prior boundary

conditions calculated by the GEOS-Chem global simulation (baseline) and the ensemble mean (SENSBC ) of 100

inversions using perturbed boundary conditions. Perturbation follows a uniform distribution of ±2% of the baseline

value at the four NAME domain edges. Verticals bars are the ±2σ range of of the ensemble.

8
Figure S5: China’s methane emissions from oil/gas, wetlands, and “other” inferred from the mean of the inversion

ensemble (SENSP rior ) (see Methods in the main text). “Other” includes combustion, industrial processes, agricultural

field burning, composting, and open fires. Verticals bars are ±1σ of the ensemble.

9
Figure S6: Source grid cells at 2◦ × 2.5◦ resolution dominated (> 50%) by a particular sector as identified by the

EDGAR v4.2 inventory. This coarse distribution was used in previous studies to apportion emissions estimates to

different sources 11 . “Other” denotes source grid cells with no dominant source (> 50% within a grid cell). Here

regions dominated by coal mining or having mixed sources (“Other”) account for about 85% of total emissions.

10
Figure S7: Spatial distributions of China’s coal mine methane emissions at 0.25◦ × 0.25◦ resolution from Sheng et

al.1 (used in this study) and EDGAR v4.2. The figure is adapted from Sheng et al.1 .

11
Figure S8: Number of GOSAT observations over China for different seasons during 2010-2017 (Jan-Mar, Apr-Jun,

Jul-Sep, Oct-Dec).

12
Figure S9: An example of coal mining fraction differences between perturbed (one ensemble member and ensemble

mean) and baseline prior inventories.

13
References.

1. Sheng, J., Song, S., Zhang, Y., Prinn, R. G. & Janssens-Maenhout, G. Bottom-Up Estimates

of Coal Mine Methane Emissions in China: A Gridded Inventory, Emission Factors, and

Trends. Environ. Sci. Technol. Lett. 6, 473–478 (2019). URL https://doi.org/10.

1021/acs.estlett.9b00294. Publisher: American Chemical Society.

2. Scarpelli, T. R. et al. A global gridded (0.1 0.1) inventory of methane emissions from oil, gas,

and coal exploitation based on national reports to the United Nations Framework Convention

on Climate Change. Earth System Science Data 12, 563–575 (2020). URL https://www.

earth-syst-sci-data.net/12/563/2020/.

3. Janssens-Maenhout, G. et al. EDGAR v4.3.2 Global Atlas of the three major Green-

house Gas Emissions for the period 1970–2012. Earth System Science Data Discus-

sions 1–52 (2019). URL https://www.earth-syst-sci-data-discuss.net/

essd-2018-164/.

4. Bloom, A. A. et al. A global wetland methane emissions and uncertainty dataset for atmo-

spheric chemical transport models (WetCHARTs version 1.0). Geosci. Model Dev. 10, 2141–

2156 (2017). URL https://www.geosci-model-dev.net/10/2141/2017/.

5. Darmenov, A. & da Silva, A. The quick fire emissions dataset (QFED)–documentation of

versions 2.1, 2.2 and 2.4. NASA Technical Report Series on Global Modeling and Data As-

similation, NASA TM-2013-104606 32, 183 (2013).

14
 6. Yan, X., Cai, Z., Ohara, T. & Akimoto, H. Methane emission from rice fields in mainland

China: Amount and seasonal and spatial distribution. Journal of Geophysical Research: Atmo-

spheres 108 (2003). URL https://agupubs.onlinelibrary.wiley.com/doi/

abs/10.1029/2002JD003182.

7. Maasakkers, J. D. et al. Global distribution of methane emissions, emission trends, and

OH concentrations and trends inferred from an inversion of GOSAT satellite data for

2010–2015. Atmospheric Chemistry and Physics 19, 7859–7881 (2019). URL https:

//www.atmos-chem-phys.net/19/7859/2019/. Publisher: Copernicus GmbH.

8. Dlugokencky, E. J., Crotwell, A. M., Mund, J., Crotwell, M. J. & Thoning, K. Atmospheric

Methane Dry Air Mole Fractions from the NOAA GML Carbon Cycle Cooperative Global

Air Sampling Network, 1983-2019, Version: 2020-07 (2020). URL https://doi.org/

10.15138/VNCZ-M766.

9. Parker, R. J. et al. Assessing 5 years of GOSAT Proxy XCH4 data and associated uncertain-

ties. Atmos. Meas. Tech. 8, 4785–4801 (2015). URL http://www.atmos-meas-tech.

net/8/4785/2015/.

10. Schepers, D. et al. Methane retrievals from Greenhouse Gases Observing Satellite (GOSAT)

shortwave infrared measurements: Performance comparison of proxy and physics retrieval

algorithms. J. Geophys. Res. 117, D10307 (2012). URL http://onlinelibrary.

wiley.com/doi/10.1029/2012JD017549/abstract.

15
11. Miller, S. M. et al. China’s coal mine methane regulations have not curbed growing emis-

sions. Nature Communications 10, 303 (2019). URL https://www.nature.com/

articles/s41467-018-07891-7.

16