Investigation of the global methane budget over 1980-2017 using GFDL-AM4.1

. Changes in atmospheric methane abundance have implications for both chemistry and climate as methane is both a strong greenhouse gas and an important precursor for tropospheric ozone. A better understanding of the drivers of trends and variability in methane abundance over the recent past is therefore critical for building confidence in projections of future 10 methane levels. In this work, the representation of methane in the atmospheric chemistry model AM4.1 is improved by optimizing total methane emissions (to an annual mean of 576  32 Tg yr -1 ) to match surface observations over 1980-2017. The simulations with optimized global emissions are in general able to capture the observed global trend, variability, seasonal cycle, and latitudinal gradient of methane. Simulations with different emission adjustments suggest that increases in methane sources (mainly from energy and waste sectors) balanced by increases in methane sinks (mainly due to increases in 15 OH levels) lead to methane stabilization (with an imbalance of 5 Tg yr -1 ) during 1999-2006, and that increases in methane sources combined with little change in sinks (despite small decreases in OH levels) during 2007-2012 lead to renewed methane growth (with an imbalance of 14 Tg yr -1 for 2007-2017). Compared to 1999-2006, both methane emissions and sinks are greater (by 31 Tg yr -1 and 22 Tg yr -1 , respectively) during 2007-2017. Our results also indicate that the energy sector is more likely a major contributor to the methane renewed growth after 2006 than wetland, as increases in wetland 20 emissions alone are not able to explain the renewed methane growth with constant anthropogenic emissions. In addition, a significant increase in wetland emissions would be required starting in 2006, if anthropogenic emissions declined, for wetland emissions to drive


Introduction
Atmospheric methane (CH4) is the second most important anthropogenic greenhouse gas with a global warming potential [28][29][30] 34 times that of carbon dioxide (CO2) over a 100-year time horizon (Myhre et al., 2013). Methane is also a precursor for tropospheric ozone (O3) -both an air pollutant and greenhouse gas -influencing ozone background levels (Fiore et al., 2002).
Controlling methane has been shown to be a win-win for both climate and air quality (Shindell et al., 2012). From a preindustrial level of 722±25 ppb (Etheridge et al., 1998;Dlugokencky et al., 2005), methane has increased by a factor of ~2.5 to a value of 1850±1 ppb in 2017 , mostly due to anthropogenic activities (Dlugokencky et 35 al., 2011). The global network of surface observations over the past 3-4 decades indicates that methane went through a period of rapid growth from the 1980s to 1990s, nearly stabilized from 1999 to 2006, and then renewed its rapid growth.
Studies of the drivers of observed changes in methane trends and variability have focused on the contributions from changes in methane sources and sinks. Here, we apply a prototype of the new generation NOAA Geophysical Fluid Dynamics Laboratory chemistry-climate model, GFDL-AM4.1 (Zhao et al., 2018a, b;Horowitz et al., manuscript in preparation) with 40 an improved representation of methane, to explore the contributions of methane sources and sinks to its observed trends and variability.
Methane is emitted into the atmosphere from both anthropogenic activities (e.g., agriculture, energy, industry, transportation, waste management, and biomass burning) and natural processes (e.g., wetland, termites, oceanic and geological processes, and volcanoes), and is removed from the atmosphere mainly by reaction with hydroxyl radical (OH) in the troposphere, with 45 less contributions to destruction by reactions with excited atomic oxygen (O( 1 D)) and atomic chlorine (Cl) in the stratosphere and uptake by soils. Measurements of the global distribution of surface methane beginning in 1983 have revealed that atmospheric methane approached steady state during 1983-2006 and renewed its growth since then. During 1983During -2006 methane growth rates decreased from 12 ppb yr -1 during 1984-1991 to 5 ppb yr -1 during 1992-1998 (Nisbet et al., 2014;Dlugokencky et al., 2018) and to 0.7±0.6 ppb yr -1 during 1999-2006 . After 2006, renewed 50 methane growth started with a growth rate of 5.7±1.2 ppb yr -1 in 2007-2013 and reached 12.6 ± 0.5 ppb yr -1 in 2014 and 10.0 ± 0.7 ppb yr -1 in 2015 (Nisbet et al., 2016;Dlugokencky et al., 2018). Investigations of the drivers of the observed methane trend and interannual variability have mainly focused on the changes in the global methane budget. While anthropogenic activities are widely considered responsible for the long-term methane increase since pre-industrial times (Dlugokencky et al., 2011), there is no consensus on the drivers for the methane stabilization during 1999-2006 and renewed growth since 55 2007. Previous studies have attributed the stabilization during 1999-2006 to the combined effects of alternatively increased anthropogenic emissions with decreased wetland emissions (Bousquet et al., 2006), decreased fossil fuel emissions (Dlugokencky et al., 2003;Simpson et al., 2012;Schaefer et al., 2016) or rice paddies emissions (Kai et al., 2011), stable emissions from microbial and fossil fuel sources (Levin et al., 2012), or variations of methane sinks (Rigby et al., 2008;Montzka et al., 2011;Schaefer et al., 2016). The observed renewed growth since 2007 has been explained alternatively 60 through increases in wetland emissions (Dlugokencky et al., 2009;Bousquet et al., 2011;Nisbet et al., 2016), increases in https://doi.org/10.5194/acp-2019-529 Preprint. Discussion started: 12 July 2019 c Author(s) 2019. CC BY 4.0 License.

Model description and initialization
We use a prototype version of the new generation NOAA Geophysical Fluid Dynamics Laboratory chemistry-climate model, 95 GFDL-AM4.1 (Zhao et al., 2018a, b;Horowitz et al., manuscript in preparation). A detailed description of the physics and dynamics in AM4.1 is provided by Zhao et al. (2018a, b). The version of AM4.1 with full interactive chemistry used in this work is described by Schnell et al. (2018). In its standard form, this model setup consists of a cubed sphere finite-volume dynamical core with a horizontal resolution of ~100 km with 49 vertical levels extending from the surface up to ~80 km. The model's lowermost level is approximately 30 m thick. The chemistry and aerosol physics in this model have been updated 100 from the previous version (GFDL-AM3; Naik et al., 2013), as described by Mao et al. (2013a, b) and Paulot et al. (2016).
There are a total of 102 advected gas tracers and 18 aerosol tracers, 44 photolysis reactions, and 205 kinetic reactions included in the chemical mechanism in this version of AM4.1 to represent tropospheric and stratospheric chemistry.
The standard AM4.1 configuration uses global annual-mean methane concentrations as a lower boundary condition to simulate the atmospheric distribution of methane. Although the model simulates reasonable global-mean methane 105 abundances, large biases exist in the simulated latitudinal distribution and seasonal cycle. This modeling framework also does not allow for the full characterization of the drivers of methane trends and variability. To overcome this issue, we updated AM4.1 to be driven by methane emissions. Table 1  over seven ecosystem-dependent altitude levels between the surface and 6 km above the surface, following the methodology of Dentener et al. (2006). Anthropogenic and biomass burning emissions are represented by monthly gridded emissions including seasonal and interannual variability. Natural emissions include wetland (WET) emissions from the WetCHARTs version 1.0 inventory , ocean (OCN) emissions from Brasseur et al. (1998) with near-shore methane fluxes from  and Patra et al. (2011), termites (TMI) from Fung et al. (1991), and mud volcanoes 120 (VOL) from Etiope and Milkov (2004) and Patra et al. (2011). Wetland emissions and ocean emissions are climatological monthly means without interannual variability. The remaining natural emissions are based on a climatological annual mean (repeated every month without seasonal variability). Trends in the total emissions and emissions from major sectors over other short-lived species also follow CEDS and SSP2-4.5 inventories. Natural emissions of other short-lived species are from Naik et al. (2013). Biogenic isoprene emissions are calculated interactively following Guenther et al. (2006).
The methane sinks considered in AM4.1 include oxidation by OH radicals, Cl, and O( 1 D), and dry deposition. Since the model does not represent tropospheric halogen chemistry, it does not consider removal of methane by Cl in the troposphere, which has been shown to be extremely minor (Gromov et al., 2018). The dry deposition flux of methane is calculated based 130 on a monthly climatology of deposition velocities (Horowitz et al, 2003) calculated by a resistance-in-series scheme (Wesely, 1989;Hess et al., 2000) and used to mimic methane loss by soil uptake, which accounts for about 5% of the total methane sink (Kirschke et al., 2013;Saunois et al., 2016).
In this work, we included 12 additional methane tracers tagged by source sector to attribute methane from agriculture (CH4AGR), energy (CH4ENE), industry (CH4IND), transportation (CH4TRA), residents (CH4RCO), waste (CH4WST), 135 shipping (CH4SHP), biomass burning (CH4BMB), ocean (CH4OCN), wetland (CH4WET), termites (CH4TMI), and mud volcanoes (CH4VOL). The tracers are emitted from corresponding sources, and undergo the same chemical pathways and dynamics as the full CH4 tracer. For analysis, we combine CH4IND, CH4TRA, CH4RCO, and CH4SHP as other anthropogenic tracers (i.e., CH4OAT), and combine CH4OCN, CH4TMI, and CH4VOL as other natural tracers (i.e., CH4ONA). 140 Initially the model was spun up in a 50-year run with repetitive 1979 emissions until stable atmospheric burdens of methane and tagged tracers were obtained. After the spin-up, several sets of simulations were conducted for 1980-2017 to quantify the methane budget and investigate the impacts of changes in methane sources and sinks (see Section 2.2). All model simulations are forced with interannually-varying sea surface temperatures and sea ice from Taylor et al. (2000), prepared in support of the CMIP6 Atmospheric Model Intercomparison Project (AMIP) simulations. Horizontal winds are nudged to the 145 National Centers for Environmental Prediction (NCEP) reanalysis (Kalnay et al., 1996) using a pressure-dependent nudging technique (Lin et al., 2012).

Simulation design
We conduct several sets of hindcast simulations for 1980-2017, as listed in Table 2, to quantify the methane budget and investigate the contributions of sources and sinks to the trend and variability of methane. The model simulation using the 150 initial methane emissions inventory (Einit) described in Section 2.1 was found to largely underestimate the methane burden.
Assuming that this mismatch is due to a bias in the simulated methane budget, we can either increase methane sources or decrease methane sinks to match the observations. We perform several optimization simulations that explore the sensitivity of methane to uncertainties in emissions of methane and levels of OH, the dominant sink for methane. Because OH trends and variability depend on a number of factors, including temperature, water vapor, O3, and emissions of nitrogen oxide 155 (NO x ), carbon monoxide (CO), and volatile organic compounds (VOCs), it is not straightforward to perturb OH. Previous work has shown that interannual variability of global OH is highly correlated with NOx from lightning (Fiore et al., 2006;Murray et al., 2013). Therefore, we apply a scaling factor to lightning NOx (LNOx) emissions to indirectly adjust OH levels https://doi.org/10.5194/acp-2019-529 Preprint. Discussion started: 12 July 2019 c Author(s) 2019. CC BY 4.0 License.
without influencing its variability. The LNOx emissions are calculated interactively based on Horowitz et al. (2003) as a function of subgrid convection parameterized in the model. The climatological global mean LNOx emission simulated by 160 AM4.1 is about 3.6 TgN yr -1 , within the range of 2-8 TgN yr -1 estimated by previous studies (e.g., Schumann and Huntrieser, 2007).
Here, we test the sensitivity of simulated methane to three assumptions: 1) standard OH levels simulated by AM4.1 (referred as "S0"); 2) low OH levels via applying a scaling factor of 0.5 to the default LNOx emission calculations (referred as "S1"); 3) high OH levels via applying a factor of 2 to the default LNOx emission calculation (referred as "S2"). For each OH option, 165 we begin with initial methane emissions and then optimize global total emissions as described below to match simulated methane with surface observations. Different OH levels could lead to different estimations of the optimized total emissions, which provides a measure of uncertainties in our optimized total methane emissions.
We apply a simple mass balance approach to optimize global total methane emissions, following the methodology of Ghosh et al. (2015). We calculate an increment ΔE by which global emissions need to be modified for each year. Unlike inverse 170 modeling studies such as , we do not optimize emissions for each grid cell. Instead, we uniformly scale emissions for particular sectors (as described below) globally for each year by the rate of the optimized emission total (Eopt = Einit + ΔE) to the initial emissions (Einit). We assume that the spatial distribution of methane emissions from the initial emission inventories are the best available information we have. Considering the large uncertainties in the anthropogenic and wetland emissions, we perform two simulations in which we achieve the optimized emission totals by scaling either 175 anthropogenic sources, including biomass burning sector only (referred to as "Aopt") or the wetland sector only (referred to as "Wopt") for the standard (S0) LNOx scenario. The purpose of conducting these simulations is to investigate the impact of optimizing emissions from different sectors on methane predictions. For the Aopt case, eight anthropogenic sectors (i.e., AGR, ENE, IND, TRA, RCO, WST, SHP, and BMB) are uniformly scaled by the ratio of ΔE to total anthropogenic emissions, keeping the fractions of individual sources unchanged. For the Wopt case, wetland emissions are rescaled to 180 increase this source by ΔE. For S1 and S2 scenarios, we apply ΔE to wetland sector only. The total Eopt emissions are the same for both Aopt and Wopt cases. Time series of methane optimized total emissions and emissions from major sectors from S0Aopt and S0Wopt over the 1980 to 2017 period are shown in Figure S2 in the Supplement.

Observations
We evaluate the simulated methane dry-air mole fraction (DMF) against a suite of ground-based and aircraft observations 185 and satellite retrievals of column-averaged CH4 to thoroughly evaluate the model simulated spatial and temporal distribution of methane. To evaluate surface CH4, we use measurements from a globally distributed network of air sampling sites  Table S1 in the Supplement. A function fit consisting of yearly harmonics and a polynomial trend, with fast fourier transform and low pass filtering of the residuals are applied to the monthly mean methane DMF to approximate the longterm trend and average seasonal cycle at each MBL site (Thoning et al., 1989;Thoning, 2019). A meridional curve (Tans et al., 1989) was fitted through these site values to get the latitudinal distribution of methane. The same sampling and 195 processing approach (Thoning et al., 1989;Tans et al., 1989) is applied to the simulated monthly mean methane DMF to calculate global and zonal averages to facilitate consistent model-observation comparison. Besides the comparison with global estimates from MBL sites, we also evaluate the model performance at various GMD sites to investigate the contributions from local sources. For site specific evaluation, we sample the model grid cell at the location of the corresponding site and at the model layer with height closest to the altitude of the corresponding site. 200 Due to the sparseness of the ground-based observational sites, especially over continental regions, we also evaluate simulated methane against satellite retrievals to reveal information on regional characteristics. Total column-averaged methane DMFs are evaluated against satellite retrievals from the Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) instrument on board the European Space Agency's environmental research satellite ENVISAT (Frankenberg et al., 2011) for January 2003 to April 2012 and the Thermal And Near Infrared Sensor for 205 carbon Observations -Fourier Transform Spectrometer (TANSO-FTS) instrument onboard the Japanese Greenhouse gases Observing SATellite (GOSAT) (Kuze et al., 2016) for April 2009 to December 2016. We compare monthly mean satellite retrievals with simulated monthly mean methane. Retrieval-specific averaging kernels are also applied to simulated monthly mean methane to calculate simulated column-averaged methane DMF.
To investigate background tropospheric methane variability, we compare the simulated vertical profiles with aircraft 210 measurements from the High-performance Instrumented Airborne Platform for Environmental Research (HIAPER) Pole-to-Pole observation (HIPPO) campaigns from January 2009 to September 2011 (Wofsy et al., 2012). A total of 787 profiles were flown during the 5 campaigns with continuous profiling between approximately 150 m and 8500 m altitudes, but also including many profiles up to 14 km altitude. For each HIPPO mission, we spatially sample the model consistent with the observations and average the model for the months of the campaign to create climatological monthly means. 215

Observations
The detailed model evaluation for S0Aopt and S0Wopt are discussed below. We first evaluate the mean climatological spatial distribution and seasonal variability simulated by the model and then evaluate the trends and variability. GMD sites is shown in Figure S3 in the Supplement. GMD sites with at least 20 years of observational records are selected for model climatological evaluation. The information of these sites is shown in Table S2  S0Wopt and S0Aopt, likely due to a model sampling bias, with model grid box overlapping land while samples are collected with onshore winds. Over middle and high latitudes of Northern Hemisphere, the simulated surface methane DMF shows low and high biases at individual sites, possibly due in part to uncertainties in the local emissions. As shown in Figure 2b, both S0Aopt and S0Wopt are able to capture the methane seasonal cycle at most sites (with a correlation coefficient (R) larger than 0.5 for about 80% of sites). Both S0Aopt and S0Wopt are able to reproduce the methane seasonal cycle over the 235 Southern Hemisphere. However, both S0Aopt and S0Wopt show poor performance in the seasonal cycle over the Southern tropical Pacific Ocean, with R < 0.5 (e.g., POCS10 and POCS15 in Figure S3 in the Supplement), but show good performance in the seasonal cycle over the Northern tropical Pacific Ocean, with R = 0.9 (e.g., POCN05, POCN10, and POCN15 in Figure S3 in the Supplement). Poor performance also exists at a few sites in middle and high northern latitudes (e.g., AZR, UUM, LEF, MHD, and ICE shown in Figure S3 in the supplement), mainly due to overestimates of methane 240 during summer. Uncertainties in the seasonality of methane emissions, OH abundances, and long-range transport could lead to biases in the seasonal cycle. In general, both S0Aopt and S0Wopt are able to capture the methane latitudinal gradient (e.g., R = 0.9). This suggests that the spatial distribution of methane in emissions is reasonable on the large scale despite uncertainties in representing local sources.

Climatological evaluation 220
To investigate background tropospheric methane variability, Figure 3 shows the bias in the simulated vertical distribution of 245 methane with respect to HIPPO observations for the S0Aopt and S0Wopt simulations. S0Aopt and S0Wopt simulations produce very similar methane profiles. Both S0Aopt and S0Wopt match observed methane profiles very well over Southern Hemisphere. Compared to HIPPO measurements, methane in both simulations is consistently high over the tropical Pacific Ocean (by up to ~ 50 ppb) from the surface to 700 mb during all HIPPO campaigns. These biases decrease with altitude and decrease with latitude except for summer. In the Northern Hemisphere, both S0Wopt and S0Aopt simulations capture the 250 observed methane from near the surface to 700 mb, but are generally biased low, except in summer when they are biased high, especially at mid-latitudes. Mid-latitude background methane is affected by both high-latitude and low-latitude air masses on synoptic scales. Biases over these regions could result from many processes (e.g., overestimation of the summer emissions, insufficient OH levels, and model transport). In general, the relative differences between the simulated methane profiles and HIPPO measurements are within 2% over most regions, demonstrating the capability of the improved GFDL-255 AM4.1 for simulating tropospheric methane.

Trend evaluation
As described in Section 2.3, we applied a function fit consisting of yearly harmonics and a polynomial trend, with fast fourier transform and low pass filtering of the residuals to the monthly mean methane DMF (Thoning et al., 1989;Thoning, 2019) to estimate the long-term trend and growth rates discussed below. . The simulated growth rates during 1984-1991 are slightly higher than the NOAA-GMD estimates (11.6±1.3 ppb yr -1 ), while the simulated growth rates during 1992-1998, 1999-2006, and 2007-2017 are within the ranges of NOAA-GMD 280 estimates (5.6±3.5 ppb yr -1 , 0.7±3.1 ppb yr -1 , 6.9±2.6 ppb yr -1 , respectively). Over the tropics, both S0Aopt and S0Wopt overestimate methane growth rates (by about 5-10 ppb yr -1 ) during 1984-1990, but are able to reproduce methane growth rates relatively well afterwards. Agreement of the methane growth rate is worse in the Northern Hemisphere than in the Southern Hemisphere, especially at the high northern latitudes, where R is smaller than 0.5. Over 30-90 o N, neither S0Aopt nor S0Wopt is able to reproduce methane growth rates during 1984-1989 and there is a slight mismatch (~1-2 years) in 285 methane growth rates afterwards. These biases indicate larger uncertainties in the methane emissions at high Northern Hemisphere than in other regions.
Comparisons of simulated surface methane DMF to NOAA-GMD observations at individual sites are shown in Figure S4  S0Aopt are able to capture monthly variations in methane at most of the sites except at LEF, where R = 0.4 for S0Wopt and 0.5 for S0Aopt, respectively. In general, both S0Wopt and S0Aopt are able to reproduce the surface methane DMF and capture the trend at most sites (e.g., with R greater than 0.5 at 98% of total sites and with RMSE less than 30 ppb at 74% of total sites). 305 Unlike the evaluation of global mean surface methane DMF, which is based on observations from a number of sites with well-mixed MBLs, the evaluation of global mean column-averaged methane DMF against satellite retrievals mainly covers continents, considering the impacts from polluted areas and the contributions from the troposphere and the stratosphere.
Simulated monthly mean column-averaged methane DMF are compared with satellite retrievals (e.g., SCIAMACHY and GOSAT) in Figure 6. The averaging kernels of SCIAMACHY and GOSAT are individually applied to the model to calculate 310 column-averaged methane abundances. Both simulations are able to capture the monthly variation of methane with R greater than 0.9, but underestimate column-averaged methane, with RMSE of about 21 ppb and 29 ppb when compared to SCIAMACHY and GOSAT retrievals, respectively. The biases increase poleward in both SCIAMACHY and GOSAT comparisons; large uncertainty exists in the satellite retrievals over high latitudes (e.g., > 70 o ) due to large solar zenith angles and potential high cloud cover. The underestimates are mainly due to biases in the middle/upper troposphere and 315 stratosphere (as shown in Figure 3). The differences in the column-averaged methane abundances between satellite retrievals and model simulations are mostly within 2% except in polar regions where there are large uncertainties in the satellite retrievals. Both simulations are also able to capture the latitudinal distribution of the column-averaged methane DMF with R close to 1.  Table 3 provides a summary of decadal mean methane budget for 1980-2017. Compared to Kirschke et al. (2013) and 350 Saunois et al. (2016), the total natural sources from the initial emission inventories (203 Tg yr -1 ) are at the lower range of top-down estimates during this period, except for the 1990s, when they are slightly higher than top-down estimates but still much lower than the bottom-up estimates. This mainly results from wetland emissions in the initial emission inventories that are slightly higher than top-down estimates but much lower than bottom-up estimates during 1990s. The total anthropogenic sources from the initial emission inventories are overall within the range of top-down or bottom-up estimates, except for 355 1980-1989, when  1980s result mainly from low estimated sources from agriculture and waste sectors in the CEDS inventory. With the optimized global total emissions, the total sources used in this work and the total sinks estimated by AM4.1 are either in the range of top-down or bottom-up estimates by previous studies. As a result, the imbalance between total sources and total sinks estimated in this work are overall within the range of estimates by previous studies although we find a smaller 360 imbalance than previous estimates for the 2000s and afterwards. The atmospheric growth rates simulated by the model (sampled identically as for observations) are also comparable to the observed atmospheric growth rates.

Spatial distribution
As described in Section 2.2, the emission optimization is conducted for anthropogenic sectors (i.e., S0Aopt) and wetland 365 sector (i.e., S0Wopt). Although global total methane emissions are the same for S0Aopt and S0Wopt, they have different allocations for anthropogenic and wetland sectors and different spatial distributions as well. Here we analyze the sensitivity of sector optimization on the spatial distribution of simulated methane concentrations. Figures 9 and 10 show the spatial distributions of the differences in the methane emissions and surface methane abundance between S0Aopt and S0Wopt during the four periods (i.e., 1980-1989, 1990-1999, 2000-2006, and 2007-2017). Surface methane is always lower in Aopt 370 than Wopt in the tropics (e.g., 15 o S-10 o N) during the four periods. This is mainly due to much lower wetland emissions in S0Aopt than in S0Wopt (Figure 10), which dominates total emissions over these regions (e.g., tropical South America and Central Africa). There is not much difference in surface methane over low and high southern latitudes (e.g., 15-90 o S) between the two simulations. This agreement is mainly because larger anthropogenic emissions in S0Aopt compensate smaller wetland emissions, producing only small differences in the total emissions, within 0.1 Tg yr -1 (Figure 10). Unlike the 375 Southern Hemisphere, surface methane concentrations are in general higher in S0Aopt than S0Wopt in the Northern Hemisphere, especially over the Eastern U.S. and Eurasia, due to much higher anthropogenic emissions in S0Aopt. The lower surface methane values in S0Aopt over northern Canada are due to much lower wetland emissions in S0Aopt. Figure 11 shows the methane growth rates simulated by Aopt and Wopt during the four time periods. Global mean methane growth rates simulated by Aopt and Wopt are very consistent during the four periods, with growth rates decreasing from 380 1980s to 1990s, stabilizing during 2000-2006, and increasing after 2007. During the 1980s and 1990s, methane growth rates in both S0Aopt and S0Wopt increase over most of the globe except a decrease over Russia, due to significant decreases in anthropogenic emissions (mainly from the energy sector) in the former Soviet Union, consistent with previous studies (Dlugokencky et al., 2011). During 2000-2006, methane growth rates increase significantly over East Asia in both S0Aopt and S0Wopt while they decrease over tropical South America and Central Africa in S0Wopt but not in Aopt. This is mainly 385 due to decreases in wetland emissions in the S0Wopt case, while wetland emissions are constant for each year in Aopt case.
After 2007, both Aopt and Wopt suggest large increases in methane growth rates over East Asia (mainly due to increases in anthropogenic emissions) by up to ~38 ppb yr -1 with smaller increases elsewhere (< 7 ppb yr -1 ), with noticeable increases https://doi.org/10.5194/acp-2019-529 Preprint. Discussion started: 12 July 2019 c Author(s) 2019. CC BY 4.0 License. also over the Arctic ( > 7 ppb yr -1 ). The relatively large methane growth over the Arctic is mainly due to increases in anthropogenic methane from lower latitudes. 390 As discussed in Sections 3.1 and 3.2, the similarity in S0Aopt and S0Wopt simulation results suggests that for 3-dimensional chemistry transport models, reasonable estimates of total global methane emissions are critical for global methane predictions, despite the uncertainties in the spatial distribution of the emissions and in the estimates of individual sources, which are more important for regional methane predictions. At the same time, accurate estimates of individual sources are necessary to attribute the methane trend and variability into individual sources. 395

Source tagged tracers
In this section, we apply Mann-Kendall (M-K) test to estimate the linear trend (different from long-term trend discussed in Section 3.1.2) of global mean source tagged tracers and total methane for 1983-1998, 1999-2006, and 2007-2017 to investigate possible drivers in total methane trends. Figure 12 compares the trends of source tagged tracers and total methane from S0Aopt and S0Wopt during 1983-1998, 1999-2006, and 2007-2017. As shown in Figure 12, both S0Aopt and S0Wopt 400 are in general able to capture the methane trends during different time periods. For S0Aopt, globally, total methane shows an increasing trend of 10.5 ppb yr -1 during 1983-1998, slightly greater than observations (8.8 ppb yr -1 ), but correlated very well with the observations (R = 1.0). The tagged anthropogenic tracers all show increasing trends during 1983-1998 despite the increases in OH levels, with larger increasing trends by AGR (3.6 ppb yr -1 ) and WST (3.6 ppb yr -1 ) consistent with emission trends. Major anthropogenic tracers (e.g., CH4AGR, CH4ENE, and CH4WST) correlate very well with total methane, with 405 R varying between 0.9 to 1.0 over this time period. Since wetland emissions and other natural emissions are constant every year in S0Aopt, with the increases in OH levels during 1983-1998, both CH4WET and CH4ONA decrease by -0.5 ppb yr -1 and -0.1 ppb yr -1 , respectively, over this period. During 1999-2006, total methane has a small increasing trend of 1.3 ppb yr -1 , still slightly greater than observations (0.6 ppb yr -1 ), but correlated relatively well with the observations (R = 0.8). During this time period, there are increasing trends in CH4ENE (2.6 ppb yr -1 ) and CH4WST (2.3 ppb yr -1 ) with slightly decreasing 410 trends in CH4AGR (-0.1 ppb yr -1 ), CH4BMB (-0.9 ppb yr -1 ) and CH4OAT (-0.5 ppb yr -1 ). Anthropogenic tracers such as CH4ENE and CH4WST correlate well with total methane, whereas CH4AGR shows a poor correlation with total methane, and CH4BMB and CH4OAT show an anticorrelation with total methane over this time period. Similarly, with the increases in OH levels during 1999-2006, both CH4WET and CH4ONA decrease, with a linear decreasing trend of -1.8 ppb yr -1 and -0.4 ppb yr -1 . During 2007-2017, total methane shows a renewed increasing trend of 5.3 ppb yr -1 , slightly below observations 415 (6.0 ppb yr -1 ) but correlated relatively well with the observations (R = 1.0). During this time, CH4ENE shows a large increasing trend (5.8 ppb yr -1 ), dominating the total methane trend. Interestingly, although there is a slight decrease in OH show significant increasing trends (2.3 ppb yr -1 , 6.9 ppb yr -1 , and 1.6 ppb yr -1 , respectively) and correlate quite well with total methane (R = 1.0) whereas all other tracers except CH4OAT show decreasing trends and poor correlations with total 435 methane. On the other hand, CH4WET shows a significant decreasing trend during this period (-4.6 ppb yr -1 ) and an anticorrelation with total methane. The decreasing trend of CH4WET is due to higher CH4WET sinks (217 Tg yr -1 ) than sources (206 Tg yr -1 ) during this period. During 1983During -1998, wetland emission growth is larger than anthropogenic emission growth due to emission optimization in S0Wopt, leading to the dominancy of wetland to drive global methane growth.
During 1999-2006, when methane stabilizes, increases in methane emissions from energy and waste sectors dominate the 440 increases in total methane sources as well as their tagged tracers (i.e., CH4ENE and CH4WST), whereas increases in methane sinks dominate all other tracers. Therefore, the imbalance between total methane sinks and sources dominate the total methane trend, which is also the case in S0Aopt during this time period. During 2007-2017, the energy is the major contributor to the renewed methane growth similar to that in S0Aopt. Figures 7 and 8 Based on evidence from isotopic composition (δ 13 CH4), recent studies suggest increasing wetland emissions may be responsible for the renewed growth of methane (Dlugokencky et al., 2009;Nisbet et al., 2016). To test this hypothesis in our modeling framework, we conducted another sensitivity simulation for 2006-2014, by repeating 2006 anthropogenic emissions for all the years but adjusting wetland emissions to ensure that the total methane emissions are the same as in 460

As shown in
S0Wopt (or S0Aopt), which would imply that the increases in methane emissions are only due to the increases in wetland emissions. This sensitivity simulation is referred to as "S0A06" and the trends for source tagged tracers and total methane are shown in Figure S5 in the supplement. Interestingly, in S0A06, anthropogenic tracers still an increasing trend during 2007-2014, with the trend in CH4ENE dominating (trend = 3.6 ppb yr -1 and R = 1.0), whereas CH4WET shows a small decreasing trend (trend = -1.0 ppb yr -1 and R = -0.8) despite rising emissions. As OH levels slightly decrease during this time 465 period, with constant emissions except wetland, one might expect possible increasing trends in all tagged tracers except CH4WET. In fact, CH4AGR, CH4ENE, CH4WST, and CH4BMB increase over 2007-2014 in S0A06, but at a slower rate than in S0Wopt (and S0Aopt) due to no emission growth for these tracers. On the other hand, the decreasing OH levels ( Figure 8) would lead to less methane sink and therefore higher methane concentrations. Since methane loss is proportional to the product of OH levels and methane concentrations and concentrations of CH4WET are much greater than other source 470 tagged tracers, the loss of CH4WET is also much higher than other tracers. Higher CH4WET loss (224 Tg yr -1 ) than CH4WET sources (207 Tg yr -1 ) leads to a decreasing trend in CH4WET. Nevertheless, S0A06 results still suggest that the renewed growth during 2007-2014 is dominated by the increases of CH4ENE, which means OH trends play an important role in determining the increasing trend of total methane since emissions of the energy sector are kept constant in this sensitivity simulation. In addition, increases in wetland emissions alone are not able to drive increases in CH4WET over this 475 period, as CH4WET sinks are equally important for determining the trend in CH4WET under constant anthropogenic emissions condition. Our analysis also suggests that increases in other microbial sources (e.g., agriculture and waste) would be needed to match the observed negative trend in δ 13 CH4 since 2007 (Nisbet et al., 2019).
We perform an additional sensitivity simulation to test the possibility of wetland emissions driving the methane trend during the period of renewed methane growth by combining the emissions of S0Aopt and S0Wopt as follows: S0Aopt emissions for 480 1980-2005 and S0Wopt emissions for 2006-2014. This simulation is referred to as "S0Comb"; the trends for source tagged tracers and total methane are shown in Figure S6 in the supplement. For 2007-2014, all anthropogenic tracers show decreasing trends except CH4ENE (2.7 ppb yr -1 ), whereas CH4WET shows a significant increasing trend (6.3 ppb yr -1 ) and dominates the total methane trend. This is mainly due to lower anthropogenic emissions during this period than previous periods, allowing sinks of anthropogenic methane tracers to start to take over their trends except for CH4ENE. At the same 485 time, significantly higher wetland emissions during this period than previous periods dominate the increasing trend of than in S0Wopt. Therefore, CH4WET loss is much lower in S0Comb (190 Tg yr -1 ) compared to S0Wopt (220 Tg yr -1 ) over this time period, leading to an increasing CH4WET trend in S0Comb, but a decreasing trend in S0Wopt. S0Comb results 490 suggest the need for a significant increase in wetland emissions along with decreases in anthropogenic emissions starting in 2006, compared to the stabilization period, for wetland emissions to drive renewed growth in methane. However, this is a less likely scenario as both top-down and bottom-up inventories indicate anthropogenic emissions increasing over 2007-2014. A more likely scenario is that both anthropogenic and wetland emissions increase (i.e., higher during 2007-2014 than [1999][2000][2001][2002][2003][2004][2005][2006]. However, in that case, the dominance of wetland emissions in driving the total methane trend would decrease 495 based on our analysis.

Sensitivity to OH levels
As described in Section 2.2, we perform two additional simulations for low and high OH levels (i.e., S1 and S2) for 1980-2017 to investigate the sensitivity of methane predictions to different OH levels. For both OH cases, the interannual variations in OH levels are the same as in S0 because the simulations are driven by the same meteorology. Figures 13(a) and 500 (b) show global tropospheric OH concentrations, methane OH loss, and methane tropospheric lifetime for the three cases (i.e., S0, S1, and S2) in which wetland emissions are optimized (Wopt; Aopt shows a very similar global OH trend as Wopt).
Compared to S0, scaling LNOx production in the model by a factor of 0.5 leads to a reduction in simulated annual global mean OH levels by -6.4 % in S1 over 1980-2017; scaling by a factor of 2 leads to an increase in simulated annual global mean OH by +9.1% in S2 . The global mean OH levels increase from 1980 to 2008 (by 3.6%, with respect to 1980 level) 505 with a linear rate of increase of 4.110 3 molecule cm -3 yr -1 , a decrease from 2008 to 2015 (by 2.3%, with respect to 2008 level) with a mean rate of -7.110 3 molecule cm -3 yr -1 , and an increase from 2015 to 2017 (by 4.6%, with respect to 2015 level) with a mean rate of 3.210 4 molecule cm -3 yr -1 . However, compared to the 1998-2007, OH levels during 2008-2015 and 2015-2017 are still higher by 2.5% and 1.3%, respectively. Changes in OH levels depend on a number of factors (e.g., temperature, water vapor, O3, NOx, CO, and VOCs). Therefore, OH is influenced by the specific chemistry and forcing data 510 used in the model. Since emission optimization is also based on methane sinks, the total optimized emissions in S1 are lower than those in S0 by about 4.1% (with an annual mean of -23.7 Tg yr -1 ), and the total optimized emissions in S2 are higher than those in S0 by about 5.8% (or 33.4 Tg yr -1 ). This indicates that a 1% change in OH levels could lead to about 4 Tg yr -1 difference in the optimized emissions. Increasing methane loss due to OH is simulated for 1980 to 2007 in the three cases due to increases in OH and methane concentrations (except over the stabilization period when methane was not increasing 515 but OH was increasing). , which is about 0.5 year lower than S1Wopt (10.40.5 years), and about 0.7 year higher than S2Wopt (9.2 0.3 years), due to different OH levels and therefore methane sinks, but with similar methane burdens. This indicates that a 1% change in OH levels could lead to about 0.08 year 525 difference in the tropospheric methane lifetime. The mean tropospheric methane lifetimes simulated by the three simulations are within the uncertainty range of observation-derived estimates for the 2000s (Prather et al., 2012) and model estimates (Voulgarakis et al., 2013;Naik et al., 2013). All simulations show an increase in methane lifetime during 2011-2015, which could be a signal of the methane feedback on its lifetime (Holmes, 2018)  Essentially, the global atmospheric methane trend is driven by the competition between its emissions and sinks. Emissions dominate sinks leading to an increasing trend while sinks dominate emissions leading to a decreasing trend. Our model results suggest that the methane stabilization during 1999-2006 is mainly due to increasing emissions balanced by increasing sinks, whereas the methane renewed growth during 2007-2013 is mainly due to increasing sources combined with little change in sinks despite small decreases in OH levels. The significant increases in methane growth during 2014-2015 are 575 mainly due to increasing sources combined with decreasing sinks. Most of the model simulations conducted here suggest that increases in energy sources drive the renewed methane growth, in agreement with previous studies (e.g., Rice et al., 2016;Hausmann et al., 2016;Worden et al., 2017), but in disagreement with other studies that consider emissions from microbial sources as the major contributor (e.g., Nisbet et al., 2016;Schaefer et al., 2016). However, optimization of emissions from anthropogenic sources depends on the "shares" of individual anthropogenic sectors in the initial emission 580 inventories. Uncertainties in these shares could lead to uncertainties in the emission adjustment for each anthropogenic sector. Recent studies using methane isotopic composition suggest that renewed growth in methane is more likely due to the increases in biogenic sources (e.g., Schaefer et al., 2016) as the ratio  13 C is shifting to more negative values since 2007.
However, it also implies increases in isotopically lighter fossil fuel emissions, or decreases in isotopically heavy sources (e.g., biomass burning), or increases in both microbial and fossil fuel emissions but with increases in microbial emissions 585 stronger than those from fossil fuel sources (Nisbet et al., 2019). It is quite possible that, rather than the energy sector, the https://doi.org/10.5194/acp-2019-529 Preprint. Discussion started: 12 July 2019 c Author(s) 2019. CC BY 4.0 License. increases in the agriculture and waste sectors may drive the renewed methane growth. In that case, it is possible that the growth of agriculture and waste emissions could be underestimated in the optimized emissions, while the growth of energy emissions could be overestimated.
The optimized emission totals estimated in this work represent temporal and spatial distribution of methane total sources 590 reasonably well. However, the emission adjustments are either applied to anthropogenic sectors only or wetland sector only.
Uncertainties therefore exist on the distribution of the emission adjustments to individual sectors. Without accurate estimates of emissions from individual sources, it would be difficult to attribute the methane trend and variability to specific sectors.
The application of methane isotopes and additional observational constraints (e.g., ethane and δ 13 CH4) could potentially help better partition the emission adjustments to different sectors. In addition, the spatial distribution of optimized emissions 595 depends on the spatial information in the initial emission inventories. Uncertainties in the spatial distribution from the initial emission inventories may remain in the optimized emissions. Our model evaluation suggests that the optimized inventory may overestimate tropical emissions. A process-based emission model (e.g., wetland emissions) coupled with AM4.1 may better represent the spatial and temporal patterns of the emissions than was possible in the present work.  Standard AM4.1 configuration, but with optimized anthropogenic emissions S0Wopt Standard AM4.1 configuration, but with optimized wetland emissions S1Wopt AM4.1 configuration with low OH levels (LNOx emissions scaled by a factor of 0.5), and optimized wetland emissions S2Wopt AM4.1 configuration with high OH levels (LNOx emissions scaled by a factor of 2), and optimized wetland emissions