Climate versus emission drivers of methane lifetime against loss by tropospheric OH from 1860 – 2100

With a more-than-doubling in the atmospheric abundance of the potent greenhouse gas methane (CH 4) since preindustrial times, and indications of renewed growth following a leveling off in recent years, questions arise as to future trends and resulting climate and public health impacts from continued growth without mitigation. Changes in atmospheric methane lifetime are determined by factors which regulate the abundance of OH, the primary methane removal mechanism, including changes in CH 4 itself. We investigate the role of emissions of short-lived species and climate in determining the evolution of methane lifetime against loss by tropospheric OH, ( τCH4 OH), in a suite of historical (1860–2005) and future Representative Concentration Pathway (RCP) simulations (2006–2100), conducted with the Geophysical Fluid Dynamics Laboratory (GFDL) fully coupled chemistry-climate model (CM3). From preindustrial to present, CM3 simulates an overall 5 % increase in τCH4 OH due to a doubling of the methane burden which offsets coincident increases in nitrogen oxide (NO x) emissions. Over the last two decades, however, the τCH4 OH declines steadily, coinciding with the most rapid climate warming and observed slow-down in CH4 growth rates, reflecting a possible negative feedback through the CH 4 sink. Sensitivity simulations with CM3 suggest that the aerosol indirect effect (aerosolcloud interactions) plays a significant role in cooling the CM3 climate. The projected decline in aerosols under all RCPs contributes to climate warming over the 21st century, which influences the future evolution of OH concentration and τCH4 OH. Projected changes in τCH4 OH from 2006 to 2100 range from−13 % to+4 %. The only projected increase occurs in the most extreme warming case (RCP8.5) due to the near-doubling of the CH 4 abundance, reflecting a positive feedback on the climate system. The largest decrease occurs in the RCP4.5 scenario due to changes in short-lived climate forcing agents which reinforce climate warming and enhance OH. This decrease is more-than-halved in a sensitivity simulation in which only well-mixed greenhouse gas radiative forcing changes along the RCP4.5 scenario (5 % vs. 13 %).


Introduction
Atmospheric methane (CH 4 ) is the second most important anthropogenic greenhouse gas after carbon dioxide (CO 2 ), reflecting its stronger heat-trapping efficiency (100-yr global warming potential of 25) and more-than-doubling in abundance since pre-industrial times (Forster et al., 2007).As a precursor to global tropospheric ozone (O 3 ) (e.g., Prather et al., 2003), the third most important greenhouse gas, methane exerts an additional indirect influence on climate (Shindell et al., 2005), as well as on background levels of surface O 3 (Fiore et al., 2002), with corresponding adverse impacts on human health (West et al., 2006).Quantifying climate impacts requires an estimate of the CH 4 lifetime, which is largely determined by its loss via reaction with the hydroxyl radical (OH, Levy et al., 1971) in the troposphere, (τ CH4 OH ).The OH distribution, in turn, is strongly influenced by changes in both climate and emissions.Furthermore, since methane is an important sink for OH, any change in the methane abundance is amplified via a positive feedback on Published by Copernicus Publications on behalf of the European Geosciences Union.
τ CH4 OH and thereby on the climate system (Prather, 1994).We investigate here the relative importance of changes in climate factors (temperature, OH, water vapor (H 2 O), lightning NO x (LNO x ), photolysis rates) versus anthropogenic emissions (CH 4 abundance, carbon monoxide (CO) and nitrous oxide (NO x )) in contributing to changes in the τ CH4 OH from 1860 to 2100, in the context of the new set of historical and future emission scenarios (Lamarque et al., 2010;Meinshausen et al., 2011;van Vuuren et al., 2011) developed for the fifth phase of the Coupled Model Intercomparison Project (CMIP5) (Taylor et al., 2012), in support of the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment (AR5).
The best constraints on τ CH4 OH in the contemporary atmosphere are from top-down estimates on global mean OH inferred from methyl chloroform (CH 3 CCl 3 ) measurements which indicate a τ CH4 OH of 10.2 +0.9 −0.7 for the period 1978-2004 (Prinn et al., 2005).For comparison, recent multi-model estimates for τ CH4 OH fall within this range 10.2 ± 1.7 yr (Fiore et al., 2009), and 9.8 ± 1.6 yr (Voulgarakis et al., 2012), and 9.90 +1.68  −1.76 yr (Stevenson et al., 2006, derived from their reported total lifetime of 8.67 ± 1.32 yr).A multi-year modeling study for 1979-1993 estimates a τ CH4 OH of 9.0 ± 0.13 yr (Dentener et al., 2003).Prather et al. (2012) have recently estimated an observationally-derived present-day τ CH4 OH of 11.2 ± 1.3 yr, which includes updated best estimates, with uncertainties, for the factors contributing to the atmospheric methane budget.This new estimate suggests models may overestimate the loss by tropospheric OH, consistent with the underestimate of CO and overestimate of O 3 in the current generation of chemical transport and chemistry-climate models (Naik et al., 2012b;Young et al., 2012).
Combining τ CH4 OH with estimates for other methane loss processes which include absorption by soils (120 yr, Prather et al., 2012), stratospheric destruction (150 yr, Prather et al., 2012), and possibly oxidation by atomic chlorine in the marine atmospheric boundary layer (200 yr, Allan et al., 2007) yields a present-day total atmospheric lifetime of 9.1 ± 0.9 yr (Prather et al., 2012).The approximate balance of the methane budget, as evidenced by the observed nearconstant methane abundance from 1999 until 2007 (Dlugokencky et al., 2009;Rigby et al., 2008), suggests that these methane sinks balanced the source globally.While the global methane source is thus fairly well constrained, the contributions from individual source sectors are poorly constrained, along with any temporal variability in their relative importance.Methane emission from wetlands dominates the natural source, with smaller contributions from termites, oceans and methane hydrates.Various anthropogenic CH 4 sources (rice paddies, ruminants, coal and natural gas, landfills and biomass burning) together account for over 60 % of the present-day global budget (Table 7.6 in Denman et al., 2007).Overall, annual source and sink strengths are estimated at 582 ± 87 Tg CH 4 and 581 ± 87 Tg CH 4 , respectively (Denman et al., 2007).
Prior studies have established the dominant role of the tropical lower troposphere in oxidation of CH 4 due to high OH abundance and temperatures (as the CH 4 +OH reaction rate constant is strongly sensitive to temperature, ∼ 2 % K −1 ) (Logan et al., 1981;Prather and Spivakovsky, 1990;Crutzen and Zimmermann, 1991;Crutzen et al., 1999, Spivakovsky et al., 2000), with estimates ranging from 75-78 % CH 4 loss in the tropics and 79-90 % CH 4 loss below 500 hPa (Spivakovsky et al., 2000;Lawrence et al., 2001;Fiore et al., 2008).The OH abundance and distribution reflect a complex interplay among various factors including anthropogenic and natural emissions, as well as meteorological processes that may change with climate.Increases in atmospheric CH 4 burden, carbon monoxide, and nonmethane volatile organic compounds (NMVOCs) all act to decrease OH concentrations and increase the methane lifetime.Increases in UV (photolysis frequencies viz.J(O 1 D)), water vapor, and nitrogen oxide sources tend to increase OH, which in turn decreases the lifetime of methane.The non-linearities in OH chemistry complicate definitive attributions to individual processes controlling OH (e.g., Spivakovsky et al., 2000).Furthermore, the relative importance of individual OH sources and sinks varies widely by region (e.g., Lelieveld et al., 2004).
The potential for these various, sometimes offsetting, factors to influence OH poses a substantial challenge to accurate identification of the processes driving OH trends and variability.While observations of CH 3 CCl 3 and the atmospheric methane growth rate could theoretically provide constraints on OH trends over recent decades (Dlugokencky et al., 1998;Spivakovsky et al., 2000;Prinn et al., 2005), attributing small changes (a few percent of the large CH 4 atmospheric abundance) unambiguously to the many individual processes exerting multiple competing influences on OH is complicated.For example, Krol et al. (1998) report a positive OH trend of 0.46 % yr −1 from 1978-1993 due to combination of stratospheric O 3 loss, reduced carbon monoxide concentrations, and increases in water vapor abundance and NO x emissions, while Dentener et al. (2003) identify water vapor as the major driver of increases in OH from 1979 to 1993 (Table 2).Hypotheses to explain the decade-long near-zero growth rate of atmospheric methane range from a steady-state equilibrium of methane sources and sinks (Dlugokencky et al., 1998), to decreases in natural and/or anthropogenic emissions (e.g., Kai et al., 2011;Aydin et al., 2011), to variations in meteorological factors and processes (e.g., Fiore et al., 2006;Hodson et al., 2011;cf. Table 2).Fewer constraints exist for the change in OH since the pre-industrial period, with estimates ranging from a 33 % decrease to a 7 % increase (cf.Table 1), although a recent multi-model study suggests little change (Naik et al., 2012b).
We use a state-of-the-art fully coupled global climatechemistry model, the Geophysical Fluid Dynamics Laboratory Coupled Model version 3 (CM3) to examine the drivers and trends of τ CH4 OH .We focus our analysis  Crutzen and Zimmerman (1991) −10 % to −20 % 3-D model Thompson et al. (1993) −20 % Multi 1-D model Martinerie et al. (1995) +6 % 2-D model Berntsen et al. (1997) +6 on the role of changes in anthropogenic emissions (CH 4 and gas-phase OH precursors) versus climate (temperature, water vapor, photolysis rates, lightning NO x ) on the evolution of τ CH4 OH on decadal-to-century time scales in a suite of historical simulations and future scenarios.This new tool allows investigation of a broader suite of chemistry-climate interactions and is intended to complement multi-model comparisons conducted for specific decadal time periods such as those occurring under the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP) in support of IPCC AR5.

Model description and simulations
The Geophysical Fluid Dynamics Laboratory CM3 Model incorporates an atmospheric chemistry model within the fully interactive framework of the atmosphere, ocean, land, and sea-ice components (Donner et al., 2011;Naik et al., 2012a;Griffies et al., 2011;C. Milly, personal communica-tion, 2012;Shevliakova et al., 2009).We provide here a summary of the features most relevant to the simulation of OH.The CM3 model simulates interactive tropospheric and stratospheric chemistry, achieved by merging the mechanisms from MOZART-2 (Horowitz et al., 2003) and AM-TRAC (Austin and Wilson, 2006), respectively, as described by Naik et al. (2012a) who evaluate global trace gas distributions over recent decades with in situ measurements of O 3 and other OH precursors.The model simulates atmospheric concentrations of 97 chemical species, 12 of which are aerosol species.Photolysis rates are parameterized by a lookup table for clear-sky conditions and are modified to account for clouds, but are not directly affected by changes in aerosols.The model accounts for the radiative effects of aerosols ("direct effect"), as well as the impact of aerosols on liquid cloud droplet formation ("indirect effect") (Donner et al., 2011).In addition, heterogeneous reactions (including those of N 2 O 5 and NO 3 ) on the surface of aerosols are included in the chemical mechanism.Hamilton and Fan (2000) 1980-1999 stratospheric quasi-biennial oscillation contributes 3-D model to CH 4 growth rate Karlsdóttir and Isaksen (2000) Karlsdóttir and Isaksen (2000) 1980-1996 positive OH trend (0.43 % yr −1 ) due to changes 3-D model, in CO, NO x and NMHC emissions fixed annual meteorology Prinn et al. (2001) 1978-2000 negative OH trend: −0.64 ± 0.60 % yr −1 Observations/Analysis Dentener et al. (2003Dentener et al. ( ) 1979Dentener et al. ( -1993 positive OH trend (0.24 ± 0.06 % yr −1 ) due to 3-D model changes in water vapor in tropical troposphere Wang et al. (2004) 1988-1997 positive OH trend (0.64 % yr −1 ) 3-D model due to decrease in column O 3 Bousquet et al. (2005) 1979-2000 negative OH trend after 1980 Inversion (−0.7 ± 0.2 % yr −1 ) Dalsøren and Isaksen (2006) 1990-2001 positive OH trend (0.08 % yr −1 ) 3-D model, from varying CO, NO x , NMVOC emissions fixed annual meteorology Montzka et al. (2011) 1998-2007 OH interannual variability: 2.3 ± 1.5 % CH 3 CCl 3 observations Emissions of CH 4 , nitrous oxide (N 2 O), and ozonedepleting substances (ODS) are not simulated in CM3.For CH 4 and N 2 O, concentrations for chemistry below 800 hPa are restored to historical or RCP values (Meinshausen et al., 2011), with a timestep of 1 day.For ODS, volume mixing ratios are prescribed as a lower boundary condition, with the source of reactive chlorine and bromine in the stratosphere parameterized following Austin and Wilson (2010).The impact of climate change on CH 4 emissions from wetlands (and any other natural sources); from wildfires; and from other biogenic precursors to OH which can influence the CH 4 lifetime (such as NMVOC and soil nitrogen oxide emissions); are not included.The model applies climatological biogenic isoprene and soil NO x emissions as in Horowitz (2006).Lightning NO x (LNO x ) is the only natural source that is tied to the model meteorology and thus can respond to changes in climate.The parameterization of both continental and oceanic lightning NO x is based on empirical relationships between convective cloud top heights and lightning flash frequencies as in Horowitz et al. (2003), Price et al. (1997) and Levy et al. (1996).Table 3 provides a summary of the CM3 model simulations analyzed here.A multi-century 1860 Control integration (CONTROL) using prescribed (static) vegetation and constant 1860 land use fractions was performed initially, and five historical ensembles (HIST) were branched from CON-TROL after achieving a stable, realistic climate.Both vegetation dynamics and land use (defined as land use change plus harvesting) were activated in the land model component of CM3 for all historical simulations and future projections.
In addition to the historical model ensemble where all forcings vary in time as described by Lamarque et al. (2010), additional "single-forcing" experiments were conducted in which the forcing of interest is allowed to vary with time, with other forcings held constant at 1860 levels.The specifications for these 3-member ensemble simulations, viz.AEROSOL, AEROSOL INDIRECT, ANTHRO, NATURAL, and WMGGO3, are described in Table 3.The AEROSOL simulations include both the direct (radiative) effect and warm-cloud aerosol indirect effect.In AEROSOL INDIRECT, aerosol and aerosol precursor emissions vary as in HIST, but an 1860 climatology taken from years 1-20 of CONTROL is used for radiation.Thus, AEROSOL IN-DIRECT includes the warm-cloud aerosol indirect effect on clouds (and the effect of aerosol changes on heterogeneous reaction rates), but does not include direct (radiative) forcing changes by aerosols.The difference between the AEROSOL and AEROSOL INDIRECT simulations provides an estimate of the aerosol direct effect.In ANTHRO, only anthropogenic forcings (viz., aerosols, ozone, greenhouse gases) vary over the historical period while natural forcings are held constant.
In NATURAL, only solar and volcanic forcings are allowed to change.The WMGGO3 simulation isolates the influences from changing only well-mixed greenhouse gases and ozone.Land use fractions are held at 1860 values (i.e., there are no land use transitions) in AEROSOL, AEROSOL INDIRECT, NATURAL and WMGGO3.CO 2 concentrations for radiation and land plant photosynthesis are held constant at the 1860 value of 286.15 ppm in AEROSOL, AEROSOL INDI-RECT and NATURAL.These simulations can be used to investigate key drivers of overall historical changes, to isolate the "forced" response (i.e., to anthropogenic emissions), and to separate climate drivers (temperature, H 2 O, J(O 1 D), lightning NO x ) from anthropogenic emission drivers (CO and NO x emissions, CH 4 concentrations) of methane lifetime.
The 1860 model tropopause is then applied to all simulation years.A comparison of tropospheric methane lifetime computed using surface to 200 hPa annual CH 4 loss rates in HIST and RCP8.5 yielded values that were within 1 % of those derived using the 150 ppbv O 3 tropopause, suggesting that τ CH4 OH as calculated here is insensitive to the tropopause definition.
For all statistics, we first average annual means over all ensemble members to obtain the ensemble mean time series.We then report the mean and standard deviation over the full length of the simulations.For our 1860 CONTROL simulation (Table 3), the mean and standard deviation of τ CH4 OH averaged over 200 yr is 8.06 ± 0.08 yr, which is within the range of preindustrial estimates (6.71-12.8yr) simulated by other models (Martinerie et al., 1995;Unger et al., 2009;Sövde et al., 2011), and 25 % lower than obtained in a recent multi-model study (10.1 ± 1.7 yr, Naik et al., 2012b).Figure 1 shows the temporal evolution of the global annual mean τ CH4 OH in the CM3/CMIP5 historical simulations and future projections, and in Table 4 we report the mean and standard deviation of τ CH4 OH , as well as the correlation coefficient of τ CH4 OH with the global mean lower tropospheric (below 500 hPa) airmass-weighted OH concentration and temperature, over the full length of the model simulations.The standard deviations in Table 4 reflect trends over the time periods considered and are therefore higher than standard deviation of CONTROL which has no long term trend.
Globally, percent changes in the reaction rate constant ( k) and in OH abundances in the lower tropical troposphere ( OH) should sum to the magnitude of the overall percent change in methane lifetime against loss by tropospheric OH ( τ CH4 OH ), as confirmed in Fig. 2 which shows the linearity between these terms, obtained by differencing 20 yr averages from the start and end of the various model simulations (r 2 = 0.99).As methane lifetime is determined by temperature and OH, we further refine our analysis by examining the evolution of the sources and sinks of OH, which can be broadly separated into emission factors (CH 4 , CO, NO x ) and climate factors (rate constant, water vapor, lightning NO x and photolysis).In Table 5 we present the temperature change, along with percent changes of τ CH4 OH , k, OH, CH 4 , CO emissions, surface NO x emissions, LNO x and J(O 1 D) over 20 yr periods at the beginning and end of the model simulations, noting the non-linear interactions of the individual processes in contributing to the net changes in OH.Correlations of OH with CH 4 , CO emissions, surface NO x emissions, the lightning NO x source (which is not included in the total NO x emissions), lower tropospheric O 3 photolysis rates and stratospheric O 3 over the full model simulation periods (146 yr for historical and 95 yr for RCPs) are reported in Table 6.
Below, we examine the overall drivers of simulated changes in methane lifetime.We further select a few time periods as illustrative examples to separate the roles of changes in climate and anthropogenic emissions in determining the lifetime of methane.

Historical period (1860-2005)
We first separate the anthropogenic forcing from natural variability by comparing the HIST and ANTHRO ensembles (Table 3).We then further separate ANTHRO into influences from aerosol versus well-mixed greenhouse gases by   comparing AEROSOL (and AEROSOL INDIRECT) with WMGGO3.

Overall drivers of methane lifetime
For the HIST five-member ensemble, the global mean τ CH4 OH over the period 1860 to 2005 is 8.63 ± 0.24 yr (Table 4), with an overall 5 % increase from 1860-1879 to 1986-2005.τ CH4 OH is strongly anti-correlated with the global mean airmass-weighted OH below 500 hPa (Table 4), which shows an almost 6 % decrease from 1860-1879 to 1986-2005 (Ta-ble 5).In a sensitivity simulation in which only emissions of NO x , CO, NMVOC, and aerosols (or their precursors) increase from 1860 to 2000, Naik et al. (2012a) find a 14 % decrease in methane lifetime from preindustrial to present.We therefore ascribe the increase in τ CH4 OH to the doubling of the global methane burden in HIST, which offsets the decrease in τ CH4 OH due to the quadrupling of surface NO x emissions (Table 5).The warming over the historical period has little impact, as demonstrated by the weak correlation with temperature in Table 4.The "single-forcing" simulations (Table 3) provide additional insights into the relative importance of individual processes in driving the changes in HIST.The lifetime increases in all of the single-forcing simulations, with the exception of WMGGO3 where τ CH4 OH decreases by 4.3 % (Table 5, Fig. 1).The consistency of the overall τ CH4 OH changes be- tween ANTHRO and HIST confirms the dominant role of anthropogenic drivers, as opposed to natural forcings, in shaping the τ CH4 OH evolution over the historical period.The percent change in the methane loss rate due to the temperature dependence of the reaction rate constant from 1860-1879 to 1986-2005 in ANTHRO is almost double that of HIST (1.4 % vs. 0.8 % in Table 5), as there is no cooling effect from volcanoes.Despite this stronger temperature effect on AN-THRO, the percent increase in the lifetime of CH 4 is comparable to that of HIST due to an offsetting larger decrease in OH (Table 5, Fig. 3b).Correlations of OH with CO emissions, CH 4 and J(O 1 D) (Table 6) are stronger in ANTHRO as the "forced" signal in HIST is masked by natural variability (e.g., volcanoes) to some extent.Without the effects of natural forcings, the average τ CH4 OH over the period 1860 to 2005 is 2.2 % lower (compare ANTHRO with HIST in Table 4).
The largest increases in τ CH4 OH from 1860-1879 to 1986-2005 occur in the AEROSOL and AEROSOL INDIRECT simulations (∼ 7 % each in Table 5).These larger increases as compared to HIST are driven entirely by aerosol-induced climate feedbacks on both temperature and OH, with the OH impact on τ CH4 OH almost double that of the temperature effect (Table 5).The overall reduction in OH can be ascribed to climate change in response to aerosol increases over the historical period because the CH 4 burden and emissions of CO and NO x are held constant at 1860 values in these simulations (Table 3).OH concentrations in AEROSOL and AEROSOL INDIRECT strongly correlate with lightning NO x , water vapor and J(O 1 D) (all below 500 hPa) (Table 6).The consistency between the AEROSOL INDIRECT and AEROSOL in the evolution of τ CH4 OH (identical correlations of τ CH4 OH against temperature and OH below 500 hPa in Table 4; similar percentage changes in Table 5 and correlations of OH versus the climate factors in Table 6), imply that aerosol-cloud interactions (indirect effect) dominate the CM3 climate response to aerosols.τ CH4 OH in the model is therefore sensitive to the aerosol indirect effect (Unger et al., 2009).
In ANTHRO, aerosol-induced cooling is offset by increases in temperature and OH associated with rising greenhouse gases (Fig. 3).Indeed, τ CH4 OH is considerably decreased in WMGGO3 due to a large warming (Fig. 3a), which leads the percent change in the CH 4 +OH reaction rate constant to increase by factors of about 6 and 3.3, respectively, compared to HIST and ANTHRO (4.6 % vs. 0.8 % and 1.4 % in Table 5).In this ensemble, the changes in CH 4 burden, CO and NO x emissions are the same as in HIST and ANTHRO (Tables 3, 5), but rising temperatures (∼ 2 K by 2005) and the associated rise in water vapor increase OH and thus drive τ CH4 OH down in WMGGO3 (Table 5), a negative feedback on climate.The single-forcing experiments thus serve to illustrate that the overall changes in HIST and ANTHRO reflect a complex balance among several opposing influences on temperature and OH, both of which determine the overall τ CH4 OH evolution.

Selected periods within the historical simulations
Although we conclude that CH 4 anthropogenic emissions drive the overall changes in τ CH4 OH from 1860 to the present, climate feedbacks are evident for specific time periods in HIST.Following volcanic events such as Krakatoa (1883), Agung (1963), El Chichón (1982) and Pinatubo (1991), methane lifetime peaks (Fig. 1), resulting from strong stratospheric aerosol-induced cooling in the model and the associated decrease in water vapor (Figs. 3a,4b).OH is reduced as a result of the lower atmospheric humidity, as well as decreases in lightning NO x (Fig. 4a); with minima in temperature, OH, water vapor and LNO x coinciding with the peak methane lifetimes occurring in 1884, 1965, 1983 and 1992.In WMGGO3, temperature, together with water vapor, rises gradually before the 1960's and more sharply after.From 1921-1970, increasing CO emissions, CH 4 burden and decreasing photolysis rates drive OH down, counterbalancing the competing influences of rising temperature, water vapor, NO x and LNO x .In the late 20th century, τ CH4 OH falls continuously as temperature and water vapor continue to rise even more rapidly (Figs.3a, 4b).τ CH4 OH is well correlated with OH within time periods (1860-1920, 1921-1970, and 1971-2005), but the different rates of change of temperature and OH within these time periods results in the lack of correlation over the entire length of the simulation (Table 4).As noted in Sect.4.1, τ CH4 OH in WMGGO3 is driven mostly by temperature through changes in the rate constant, and associated water vapor.

Trends in recent decades
Over the last 20 yr of HIST τ CH4 OH declines steadily, coinciding with the period of most rapid warming, suggesting a negative climate feedback since rising temperature and water vapor shorten τ CH4 OH (Bekki and Law, 1997;Johnson et al., 2002;Dentener et al., 2003;Fiore et al., 2006).Since the 1970's, we also note an increase in J(O 1 D), presumably driven by depletion of stratospheric ozone (Fig. 4c, d) as aerosol changes do not directly affect photolysis rates in these simulations, except through aerosol-induced cloud changes.The increase in J(O 1 D) also enhances OH formation and shortens τ CH4 OH (Wang et al., 2004).From 1998-2005, τ CH4 OH continued to fall as stratospheric O 3 and J(O 1 D) leveled off, indicating that these factors were not contributing to the OH changes during this period.By 2005, τ CH4 OH in HIST tracks with ANTHRO, perhaps in part due to the lack of major volcanic eruptions since Pinatubo.
We find a small (∼ 3 %) increase in modeled OH over the 1980-2000 period, which is consistent with the multimodel ensemble value of 3.5 ± 2.2 % reported by Naik et al. (2012b).This small positive OH trend in recent decades is consistent with the current generation of models (Table 2), but disagrees with the negative OH trends inferred from observations of CH 3 CCl 3 and inversion studies (e.g., Prinn et al., 2001;Bousquet et al., 2005).More recently, however, Montzka et al. (2011) suggest that uncertainties in measurements of CH 3 CCl 3 prior to 1998 may have contributed to the wide OH variability reported in previous studies; they find a small observationally-derived OH interannual variability of 2.3 ± 1.5 % from 1998-2007.Similar to HIST, τ CH4 OH in ANTHRO continues to fall over the last 20 yr, likely reflecting some combination of rising water vapor and an increase in tropical NO x emissions, particularly over India and Southeast Asia (not shown; Gupta et al., 1998;Wild and Palmer, 2008).After 1980, increasing temperatures, water vapor and OH concentrations shorten CH 4 lifetime in ANTHRO, while cooling and decreasing OH in AEROSOL have the opposite effect, causing the lifetime to increase.Stratospheric ozone rises slightly in AEROSOL after 1990, as opposed to the strong decrease in ANTHRO; by 2005 τ CH4 OH in AEROSOL exceeds that in ANTHRO by 0.42 yr (Fig. 1).After 1985, increased J(O 1 D) in WMGGO3, correlated with the depletion of stratospheric ozone, along with rising temperatures and water vapor (Figs.3a, 4b) cause  τ CH4 OH to decrease steadily as in HIST, for an overall 4 % decline over the period .In all the single-forcing simulations except for AEROSOL and AEROSOL INDI-RECT, the climate drivers (rapid warming and associated increase in water vapor) dominate in the late 20th century, and the methane lifetime decreases.
Our historical and single-forcing experiments over the 1860-2005 period illustrate the diverse factors influencing the evolution of the methane lifetime, with cooling from aerosol-cloud interactions masking the effects of greenhouse gas warming over the historical period in CM3.Rising temperatures and increases in water vapor and LNO x drive τ CH4 OH down (Tables 4, 5) in WMGGO3.In contrast, the temperature-driven changes in reaction rate constant, water vapor and lightning NO x are of the opposite sign in AEROSOL (Table 5) and τ CH4 OH increases.Both climate and emissions influence the simulated methane lifetime in HIST; while emission-driven changes dominate the overall increase in τ CH4 OH from the pre-industrial to present, climate warming in the late 20th century shortens the lifetime of CH 4 .The aerosol indirect effect plays a major role in CM3 and its impacts on changes in methane lifetime need further investigation.

Future scenarios (2006-2100)
We focus first on the three RCP scenarios (RCP2.6,RCP4.5 and RCP6.0) that have equivalent decreases in anthropogenic NO x emissions (∼ 45 %) by the end of the century (Table 5).The trajectory of τ CH4 OH in RCP2.6 (Fig. 1) initially drops, reaching a minimum value of 7.22 yr in 2063, after which it rebounds, for an overall 9.1 % decrease from 2006-2025 to 2081-2100.In this "low warming" scenario, associated temperature change below 500hPa over the same time periods is ∼ 1.1 K, while lower tropospheric OH increases by 7.3 % (Table 5, Fig. 3), with both factors contributing to the decrease in τ CH4 OH .The increase in OH concentration reflects increases in water vapor, lightning NO x , and J(O 1 D), and is further enhanced by reductions in CH 4 burden and CO emissions (Table 5), which together outweigh the 46 % reduction in total NO x emissions.Prior to 2063, OH rises rapidly, driven by almost linear declines in methane burden and CO emissions (Fig. 5a, b).After 2063, τ CH4 OH increases as the evolution of OH reverses and starts decreasing, driven by declining anthropogenic NO x emissions, which offset the combined impacts on OH from warming and declining anthropogenic CO emissions.
Under the moderate warming scenario, RCP4.5, there is a linear decline in the lifetime of methane, which drops by 13 % from 2006-2025 to 2081-2100, the largest overall decrease in methane lifetime in any of the RCP projections (Fig. 1).Although the decrease in methane burden is only one-third that in RCP2.6 (Table 5), the larger increases in water vapor and lightning NO x (driven by warming which is twice as large in RCP4.5 as compared to RCP2.6) combined with a larger decrease in CO emissions lead to an overall larger OH change by the end of the century than in RCP2.6.The overall warming in RCP4.5 is closer to that in RCP6.0 but the OH change in RCP4.5 is closer to that in RCP2.6, illustrating the complex interplay of climate and emissions in determining methane lifetimes.
We next explore the role of short-lived aerosol and ozone precursors in contributing to the climate-driven changes by comparing RCP4.5 with RCP4.5* (where the short-lived species are held at 2005 levels (Fig. 5), as are CH 4 , N 2 O and ODS for chemistry, but well-mixed radiative greenhouse gases follow the RCP4.5 scenario, Table 3).In RCP4.5*, we again find a decrease in τ CH4 OH , but the percent change (5 %) is not as large as in the RCP4.5 ensemble -here roughly equal percent changes in rate constant and OH contribute to a 5 % reduction in methane lifetime (Table 5).Increases in temper-ature, water vapor and lightning NO x enhance OH concentrations, and reduce τ CH4 OH in the model.We note that the evolution of temperature is very similar in the RCP2.6 and RCP4.5* simulations (Fig. 3a), as more warming from increasing CO 2 concentrations in RCP4.5* offsets the aerosol cooling, which has been reduced dramatically (i.e., a warming effect) as sulfate aerosols are considerably decreased in RCP2.6 (e.g., Fig. 4d).Around 2035, the temperatures in RCP4.5 separate from RCP4.5*, and by 2100 are warmer by 1.1 K, illustrating that decreasing aerosols can accelerate warming (e.g., Levy et al., 2008;Raes and Seinfeld, 2009;Makkonen et al., 2012).In addition, the decreases in CH 4 and CO further reinforce the climate-driven decrease in τ CH4 OH in RCP4.5 (Table 5).
In the RCP6.0 scenario τ CH4 OH also decreases, by 5.2 % from 2006-2025 to 2081-2100 (Table 5).Although the temperature, water vapor, and lightning NO x increases are about 2.5 times greater than in RCP2.6, and both RCP2.6 and RCP6.0 have similar (∼ 45 %) reductions in NO x emissions, there is a smaller decrease in CO emissions in RCP6.0 (18 % vs. 31 %), and the total CH 4 burden increases in RCP6.0 (Table 5), leading to an overall reduction in OH, and a longer average τ CH4 OH than in RCP2.6 (8.44 vs. 7.72 yr in Table 4).OH decreases from 2006 to the mid 2060's and then rises fairly rapidly until 2100 (Fig. 3b).The latter part of this simulation resembles the last quarter century of WMGGO3 in the CM3 model, in so far as rising temperatures and increased OH (from increasing water vapor, lightning NO x and photolysis), contribute to a steady decline in τ CH4 OH .
RCP8.5 is the most extreme warming scenario, with lower tropospheric temperatures warming by 4.5 K from 2006 to 2100 in CM3 (Fig. 3a, Table 5).Despite the increasing temperature and humidity (Figs.3a, 4b), which our previous analysis indicates should shorten the methane lifetime against loss by tropospheric OH, τ CH4 OH increases, from 8.24 yr in 2006-2025 to 8.73 yr in 2081-2100, with a peak value of 8.91 yr in 2062 (Fig. 1).This increase in lifetime is due to a decrease in OH beginning after 2035, with an overall 14.8 % reduction by the end of the century (Fig. 3b, Table 5).The near doubling of CH 4 abundance by 2100 (Table 5) depletes OH and leads to a positive feedback on the climate system, as concentrations of CH 4 increase further (Prather, 1994).The rise in anthropogenic CH 4 outweighs the climate-driven changes which act to shorten the methane lifetime (Table 5,Figs. 3a,4a,b,5a), viz.: rising temperature, water vapor, lightning NO x .In the late 21st century τ CH4 OH declines slightly, presumably reflecting the negative feedback from climate.

Conclusions
Improved understanding of feedbacks between CH 4 abundances and the climate system are important as they impact not only climate, but also the environment and human health.
The relatively short lifetime of methane makes it a suitable candidate for mitigation efforts, as reducing CH 4 emissions decreases radiative forcing and improves global air quality (e.g., Fiore et al., 2002;Shindell et al., 2012).Understanding the processes driving the trends and variability in methane lifetime is a crucial step towards accurately projecting its future evolution.
We used a suite of CM3 model simulations to evaluate the role of climate and emissions in determining methane lifetime against loss by tropospheric OH, (τ CH4 OH ), and to pinpoint key drivers from the period to the end of the 21st century.Over the historical period, the τ CH4 OH increases by 5 % overall due to anthropogenic emission increases in CH 4 offsetting increases in OH from increased NO x emissions and from climate warming.There are also indications of climatically-driven intervals, e.g., after volcanic events when cooling by aerosols causes the methane lifetime to rise (e.g., HIST in Figs. 1,3a).During the late 20th century, τ CH4 OH decreases steadily as rapid warming occurs, and the observed leveling off in atmospheric CH 4 during that time could partially reflect climate-driven influences on the methane sink in addition to increases in tropical anthropogenic NO x emissions.We find a small OH increase (3 %) in recent decades , consistent with previous model studies.While earlier observation-based analyses indicated large decreasing trends in OH over this period (Prinn et al., 2005), Montzka et al. (2011) show that the measurements are much more variable prior to 1998, potentially confounding estimates of trends in OH.
Projected changes in τ CH4 OH for the 21st century are: −9 %, −13 %, −6 %, and +4 % for RCP2.6,RCP4.5, RCP6.0, and RCP8.5, respectively (changes are 2081-2100 average minus 2006-2025 average).Assuming soil, stratospheric and tropospheric chlorine sinks (120, 150, and 200 yr, respectively) remain constant, projected changes in total methane lifetime are −12 % (RCP2.6),−15 % (RCP4.5),−7 % (RCP6.0)and −0.3 % (RCP8.5).The large increases in CH 4 outweigh the extreme warming in the RCP8.5 scenario, and results in the projected increase in τ CH4 OH .We also find positive feedbacks on the climate system; in RCP8.5, the additional CH 4 and other emission changes increase the methane lifetime (despite the tendency of warmer climate to reduce τ CH4 OH ), and amplify climate forcing.Climate and thereby τ CH4 OH is sensitive to the representation of the aerosol indirect effect in the CM3 model and represents a major source of uncertainty.In a simulation in which only greenhouse gas radiative forcing evolves following the RCP4.5 pathway (viz.: RCP4.5*), rising temperatures and enhanced OH shorten τ CH4 OH .Decreases in aerosol, CO and CH 4 in the full RCP4.5 scenario augment these climate-driven changes to further decrease τ CH4 OH .Combining the sensitivities from Table 2 of Holmes et al. (2012) with our percentage changes in Table 5 suggests that water vapor is a stronger influence on τ CH4 OH than LNO x (both are responding to temperature changes), but the large uncertainties surrounding LNO x parameterizations (e.g., Labrador et al., 2004Labrador et al., , 2005) ) and their impact on OH requires further study.
Ongoing efforts such as the Atmospheric Chemistry and Climate Model Intercomparison Project (Lamarque et al., 2012) provide opportunities to further explore and evaluate the role of climate versus emission-driven changes in determining the trends and variability in the lifetime of methane (Naik et al., 2012b;Voulgarakis et al., 2012).Together, these results provide a benchmark for future analyses of methane lifetime in chemistry-climate models.Future efforts should consider the feedbacks to CH 4 from wetlands as well as other emissions from the biosphere (biogenic VOC and NO x ; wildfires), and any changes in the soil and stratospheric sinks.For example, wetland emissions are expected to increase with warming (e.g., O'Connor, 2010) and the net impact of this positive feedback and the negative climate feedbacks (from water vapor, LNO x ) found here should be assessed.Reducing the uncertainty in the contemporary global methane budget would help build confidence in future projections of the CH 4 response to climate change.In particular, process-based approaches, including new global scale satellite observations (e.g., Bloom et al., 2010 for wetlands) should help to elucidate the mechanisms determining atmospheric abundances of methane and OH.

a
Tropospheric lifetime with respect to OH or chemical lifetime.b Total lifetime (includes all loss processes).

Table 1 .
Percent changes in methane lifetime and OH compiled from literature.Note that the definition of methane lifetime is not necessarily the same in references below.

Table 2 .
Summary of major drivers of methane and OH trends, growth rates and interannual variability (IAV) from previous studies.

Table 3 .
Summary of forcings and emissions used in CM3/CMIP5 simulations.All simulations use dynamic vegetation, except for 1860 CONTROL which uses static vegetation.
a Number in parentheses next to experiment names denotes the number of ensemble members.b AEROSOL: only aerosols, SO 2 /BC/OC emissions are time-varying.c NATURAL: GHG for radiation and chemistry fixed at 1860 values.CFCs fixed at 1860 values.d WMGGO3: SO 2 and aerosols held at 1860 values.e RCP4.5*:CH 4 , N 2 O and ODS for chemistry are fixed at 2005 values.Aerosol and O 3 precursor emissions also fixed to 2005 values.

Table 4 .
Mean and standard deviation (σ ) of annual mean methane lifetime with respect to OH loss, τ CH4 OH , (yr), and correlation coefficient (r) of τ CH4 OH with global annual mean airmass-weighted OH and temperature (both below 500 hPa), in CM3/CMIP5 simulations.All values are reported over the length of model simulation periods (146 yr for historical, 95 yr for RCPs), except for τ CH4 OH 1986-2005 , the methane lifetime against loss by tropospheric OH over the last 20 yr

Table 6 .
Correlation coefficients (r) for global annual mean OH versus global annual total methane burden (CH 4 ), CO emissions (COEMIS), surface NO x emissions (NOEMIS), and lightning NO x (LNO x ); global annual average airmass-weighted water vapor (H 2 O) and photolysis rates (J(O 1 D)); and global total stratospheric ozone (STRATO 3 ), over the length of CM3/CMIP5 model simulation periods (146 yr for historical, 95 yr for RCPs).Values below 500 hPa are used for OH, water vapor, LNO x and J(O 1 D).Boxes are shaded when they have the correct sign of r to match the sign of OH change in Table5.Grey shaded boxes denote variables that have correct sign of r to increase OH.Dark grey shaded boxes indicate variables that have correct sign of r to decrease OH.