Reducing methane (CH4) emissions is an attractive option for jointly addressing climate and ozone (O3) air quality goals. With multidecadal full-chemistry transient simulations in the MOZART-2 tropospheric chemistry model, we show that tropospheric O3 responds approximately linearly to changes in CH4 emissions over a range of anthropogenic emissions from 0–430 Tg CH4 a−1 (0.11–0.16 Tg tropospheric O3 or ∼11–15 ppt global mean surface O3 decrease per Tg a−1 CH4 reduced). We find that neither the air quality nor climate benefits depend strongly on the location of the CH4 emission reductions, implying that the lowest cost emission controls can be targeted. With a series of future (2005–2030) transient simulations, we demonstrate that cost-effective CH4 controls would offset the positive climate forcing from CH4 and O3 that would otherwise occur (from increases in NOx and CH4 emissions in the baseline scenario) and improve O3 air quality. We estimate that anthropogenic CH4 contributes 0.7 Wm−2 to climate forcing and ∼4 ppb to surface O3 in 2030 under the baseline scenario. Although the response of surface O3 to CH4 is relatively uniform spatially compared to that from other O3 precursors, it is strongest in regions where surface air mixes frequently with the free troposphere and where the local O3 formation regime is NOx-saturated. In the model, CH4 oxidation within the boundary layer (below ∼2.5 km) contributes more to surface O3 than CH4 oxidation in the free troposphere. In NOx-saturated regions, the surface O3 sensitivity to CH4 can be twice that of the global mean, with >70% of this sensitivity resulting from boundary layer oxidation of CH4. Accurately representing the NOx distribution is thus crucial for quantifying the O3 sensitivity to CH4.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Methane (CH4) emission controls are currently receiving attention as a viable low-cost strategy for abating surface ozone (O3) pollution while simultaneously slowing greenhouse warming [Hansen et al., 2000; Fiore et al., 2002a; Dentener et al., 2005; EMEP, 2005; West and Fiore, 2005; West et al., 2006]. In the presence of nitrogen oxides (NOx), tropospheric CH4 oxidation leads to the formation of O3 [Crutzen, 1973]. Over the last century, global background O3 concentrations have risen by at least a factor of two, due mainly to increases in CH4 and NOx emissions [e.g., Marenco et al., 1994; Wang and Jacob, 1998]. Here, we characterize the response of tropospheric O3 to controls on CH4 emissions, analyze the dominant processes determining the distribution of this response, and quantify the resulting benefits to air quality and climate.
 With a lifetime of approximately a decade, CH4 is fairly well-mixed in the atmosphere. Sources of atmospheric CH4 include wetlands, ruminants, energy, rice agriculture, landfills, wastewater, biomass burning, oceans, and termites. Anthropogenic emissions are estimated to contribute at least 60% to total CH4 emissions, with individual studies reporting a range of 500 to 610 Tg a−1 for total CH4 emissions [Denman et al., 2007]. The dominant CH4 sink is reaction with the hydroxyl radical (OH) in the troposphere. If sufficient quantities of NOx are available, CH4 oxidation produces O3 via reactions of peroxy radicals with NOx. In a low-NOx environment, formation of methyl hydroperoxide (CH3OOH) suppresses O3 production and may provide a net O3 sink. In an extremely low-NOx environment, CH4 oxidation may also decrease O3 levels by HO2 reacting preferentially with O3 rather than with NO. Under present-day tropospheric conditions, however, Spivakovsky et al.  show that HO2 + NO is more important than HO2 + O3 globally as a source of OH (their Figure 10), implying that increases in CH4 abundances should yield a net global increase in the tropospheric O3 burden, as has been reported in prior modeling studies [e.g., Prather et al., 2001]. Previously, CH4 and NOx emission reductions have been shown to be the most effective means of lowering tropospheric O3: reductions in anthropogenic NOx emissions decrease surface O3 in polluted source regions by up to four times more than equivalent percentage reductions in anthropogenic CH4 emissions, while CH4 reductions have a stronger impact on the tropospheric O3 burden, and a similar influence to NOx on global average surface O3 concentrations [Fiore et al., 2002a; West et al., 2008].
 To date, most chemical transport model (CTM) studies have applied a uniform CH4 mixing ratio to avoid the computational expense of multidecadal simulations required for CH4 to reach a steady state [e.g., Prather et al., 2001; Stevenson et al., 2006]. We have previously adopted this approach to evaluate the benefits to human health, agriculture, and commercial forests resulting from lower O3 due to CH4 emission reductions [West and Fiore, 2005; West et al., 2006]. In such simulations, termed “steady state” in our analysis below, the uniform CH4 mixing ratio is adjusted to reflect a desired CH4 emission change, accounting for the non-linear feedback of CH4 on its own lifetime through OH [Prather, 1996; Prather et al., 2001]. Since the relationship between CH4 emissions and CH4 concentrations is non-linear, it is important to assess the degree to which this non-linearity affects the accuracy of estimates obtained by scaling results (e.g., changes in O3 concentrations) from one CH4 perturbation to another.
 In Figure 1, we compile estimates of the response of tropospheric O3 to changes in CH4 emissions from several global CTMs in the literature, to investigate whether changes in O3 scale linearly with changes in CH4 emissions. We include results from transient, full-chemistry simulations [Dentener et al., 2005], from “steady state” simulations with a uniform, fixed CH4 concentration [Wang and Jacob, 1998; Prather et al., 2001; Fiore et al., 2002a; West et al., 2006, 2008], and from a hybrid modeling approach [Shindell et al., 2005]. Despite variations in the simulation type, the total CH4 emissions, the anthropogenic fraction of CH4 emissions, and the emissions of other species that affect OH, Figure 1 shows that the tropospheric O3 burden responds roughly linearly to changes in anthropogenic CH4 emissions across the models. Estimates from the individual studies range from 0.12–0.16 Tg tropospheric O3 per Tg a−1 change in CH4 emissions. Although the feedback between CH4 and OH will cause the CH4 concentration to respond in a strongly non-linear manner for sufficiently large increases in CH4 emissions, the relationship is approximately linear for the range of emission perturbations considered in Figure 1, corroborating earlier results derived from theory and applied in a one-box model [Prather, 1996]. We further estimate from the published studies in Figure 1 that anthropogenic CH4 currently contributes ∼50 Tg to the annual mean tropospheric O3 burden, and ∼5 ppb to global mean surface O3 (based on the subset of models reporting changes in surface O3 [Fiore et al., 2002a; Dentener et al., 2005; West et al., 2006]).
 Designing effective CH4 controls to combat O3 air pollution requires knowledge of the magnitude and spatial pattern of the surface O3 response to changes in CH4 emissions. The sensitivity of O3 to changes in CH4 should depend on the emission ratio of NOx to non-methane volatile organic compounds (NMVOC) and carbon monoxide (CO), which affects the abundance of OH [Wang and Jacob, 1998; West et al., 2006]. Here, we apply the global MOZART-2 CTM to characterize the O3 response to CH4 emission changes both with and without changes in emissions of other species (NOx, CO and NMVOC; sections 3 and 4). We then examine the processes contributing to the regional pattern of the O3 response to CH4 (section 5), and identify any dependence of this response on the geographical location of the CH4 source (section 6). Finally, we quantify the global and regional air quality (section 7) and radiative forcing (section 8) impacts that could be attained via CH4 controls from 2005 to 2030.
2. Methane Simulations
 We apply the MOZART-2 global model of tropospheric chemistry [Horowitz et al., 2003] to assess the response of O3 to changes in CH4 emissions. Table 1 provides a summary of the twelve simulations used in our study, which are described in detail below. We first consider sustained CH4 emission reductions in transient simulations in which other emissions are held fixed at present-day values (section 2.1) in order to diagnose the CH4-OH feedback factor in our model and to characterize the tropospheric O3 response to CH4 emission controls. We then apply CH4 controls phased in between 2005 and 2030 along three different trajectories, relative to a baseline future emission scenario in which emissions of CH4 and other O3 precursors change (section 2.2). These future scenarios in which CH4 controls are implemented in a more plausible manner allow us both to quantify the climate and air quality benefits that could be attained via different policy options and to examine the extent to which these benefits can be scaled from one CH4 control trajectory to another. Finally, we employ “steady state” simulations (section 2.3) to examine the relative impact of CH4 oxidation in the free troposphere versus in the boundary layer on surface O3, and to quantify the total contribution of anthropogenic CH4 to tropospheric O3. All simulations are driven by meteorological fields from the NCEP reanalysis [Kalnay et al., 1996] at a horizontal resolution of 1.9° × 1.9° with 28 vertical levels. We update the isoprene nitrate chemistry, from the 8% yield [Carter and Atkinson, 1996] used by Horowitz et al.  to 12% [Sprengnether et al., 2002], and treat isoprene nitrates as a NOx sink [e.g., Chen et al., 1998]; this modification reduces the positive bias in the MOZART-2 surface O3 simulation [Murazaki and Hess, 2006] as discussed further in section 2.4. Our simulations focus exclusively on the role of changes in O3 precursor emissions and do not include any impacts resulting from future changes in climate.
Table 1. Description of MOZART-2 Simulations
CH4 Emissions or Mixing Ratio
Non-CH4 O3 Precursor Emissions
Anthropogenic emissions are as defined in Table 2 but exclude agricultural waste burning. The 97 Tg a−1 reduction in BASE is applied to the anthropogenic emissions as a globally uniform decrease of 39%.
The anthropogenic (industrial plus agricultural) CLE CH4 emissions were scaled to the desired emission reduction according to the spatial pattern of the difference between CH4 emissions in the CLE and “Maximum technologically Feasible Reduction” (MFR) scenarios in 2030 from Dentener et al. .
The RASIA simulation was stopped after 11 years since there was little difference in the O3 response from that in RGLOB.
2.1. Transient Simulations of Sustained CH4 Reductions
 We conduct three full-chemistry transient simulations beginning in 1990, with emissions of all O3 precursors, except for CH4 (and the lightning NOx source which is tied to the meteorology as by Horowitz et al. ), held constant. In the first simulation (BASE), we maintain CH4 emissions at 1990 levels. The BASE CH4 emissions (Table 2) include 308 Tg a−1 from anthropogenic sources [Olivier et al., 1996, 1999] and 25 Tg a−1 from biomass burning [Horowitz et al., 2003]. We uniformly increase the global wetland emissions from Horowitz et al.  by 40% to 204 Tg a−1 on the basis of recent estimates [Wang et al., 2004]. The 1990–2004 winds are recycled to complete 30-year simulations.
Table 2. Methane Emissions (Tg CH4 a−1) Used in This Study
 Since we wish to investigate the sensitivity of O3 to the geographical location of CH4 emissions, we conduct two additional simulations, in both of which global anthropogenic CH4 emissions are decreased by the same magnitude. In one simulation (RASIA), we set Asian (India, East Asia, and Southeast Asia as defined by Naik et al. ) anthropogenic CH4 emissions (97 Tg a−1; excluding agricultural waste burning) to zero. In the other simulation (RGLOB), we obtain the same 97 Tg a−1 reduction by uniformly decreasing CH4 emissions from all anthropogenic sectors (except for agricultural waste burning) by 39%. The decrease of 97 Tg a−1 corresponds to an 18% reduction in total global CH4 emissions.
 The model includes the major CH4 loss mechanism of reaction with tropospheric OH (450–480 Tg a−1 in BASE; range reflects variability over the 15 years), as well as minor losses to soils (20 Tg a−1 in BASE, imposed via a deposition velocity) and in the stratosphere (50–70 Tg a−1 in BASE). Methane losses in the stratosphere by reaction with OH and O(1D) are modeled explicitly, and loss by reaction with chlorine is accounted for by prescribing the CH4 concentration in the upper two model levels (above 14 hPa) to zonally and monthly averaged values from the middle atmosphere model Study of Transport and Chemical Reactions in the Stratosphere (STARS) [Brasseur et al., 1997] as described by Horowitz et al. . For the 30-year RGLOB simulation, these climatological values were decreased by 18% in an effort to account for the decrease in stratospheric concentrations that would result from the reduction in CH4 emissions.
 Emissions of O3 precursors besides CH4 from all sources are as described by Horowitz et al.  except for NOx emissions from ships, which have been removed on the basis that their inclusion likely leads to unrealistically high NOx concentrations in the marine boundary layer in global models that neglect the rapid NOx destruction recently observed to occur inside the ship plume [Kasibhatla et al., 2000; Chen et al., 2005]. Eyring et al. , however, found that the ensemble mean oceanic NOx concentrations from 10 global models that included ship emissions fell within the range of a wider observational data set than that used by Kasibhatla et al. . They point out, however, that the modeled difference from including versus excluding ship emissions is too weak to be accurately evaluated with available measurements [Eyring et al., 2007]. While the impact of ship NOx emissions on the oceanic atmosphere is still uncertain, Eyring et al.  show that the ship NOx emissions in the year 2000 CLE inventory (which we include in our transient future scenarios described below) decrease the CH4 lifetime in the models by 0.13 years (10-model ensemble mean). We further discuss the impact of ship NOx emissions in the context of our results in section 4.
 Here, we conduct transient simulations for 2000–2030 following the CLE scenario, with the period 2000–2004 used for spin-up. We adopt the approach of Dentener et al. , interpolating the CLE emissions provided for the years 2000, 2010, 2020, and 2030 to obtain annual emissions; Table 2 shows the growth of CH4 emissions from 2005 to 2030. Between 2005 and 2030, baseline CLE anthropogenic emissions of CH4, NOx, CO, and NMVOC change by +29% (+96 Tg CH4 a−1), +19% (+5.3 Tg N a−1), −10% (−44 Tg CO a−1), and +3% (+3 Tg C a−1), respectively. Aircraft emissions are assumed to grow linearly, from 0.8 to 1.7 Tg N a−1 (NOx) and 1.7 to 3.7 Tg a−1 CO, as recommended for the ACCENT Photocomp Experiment 2 simulations for 2000 and 2030 [Stevenson et al., 2006], based on the IS92a scenario [Henderson et al., 1999]. Biomass burning emissions are taken from the 1997–2002 GFED v.1 biomass burning climatology [Van der Werf et al., 2003], vertically distributed following the recommendations for the ACCENT Photocomp Experiment 2, and assumed constant into the future. Wetland emissions are based upon the seasonal and spatial distribution from Wang et al.  as described by Fiore et al. , but here we reduce CH4 emissions from swamps by 12 Tg a−1, in an effort to reduce the positive tropical bias as compared to the NOAA GMD observations found in that study. The NCEP meteorology for 2000–2004 is recycled every 5 years to allow for interannual variability in the O3 response to CH4; these years were chosen on the basis of our previous work showing that the meteorology during these years yields a relatively constant CH4 lifetime when emissions are held constant [Fiore et al., 2006], and thus should minimize discontinuous changes in the CH4 sink by tropospheric OH when the winds are recycled. Losses of CH4 transported into the stratosphere are treated as described in section 2.1 with the exception of the prescribed climatology in the upper 2 model levels; we instead relax the model CH4 concentrations in these levels to zero with a six month lifetime to account for CH4 loss by reaction with chlorine. The six month lifetime retains the same present-day stratospheric loss rate as in BASE, while allowing the stratospheric CH4 sink to adjust to changes in the atmospheric burden resulting from changes in emissions. Additional model updates in these simulations include an increase of the O(1D) + N2 rate constant [Ravishankara et al., 2002] and the inclusion of near-infrared photolysis of HO2NO2 [Roehl et al., 2002].
 We conduct three simulations using CH4 reduction scenarios relative to the baseline CLE scenario beginning in 2006 (Figure 2). Compared to the 17% increase in total CH4 emissions in the baseline CLE scenario between 2005 and 2030 (Table 2), emissions increase by only 4% in scenario A and decline by 5% and 15% in scenarios B and C, respectively, over this period. Further details on the development of these scenarios are provided by J. J. West et al. [Management of tropospheric ozone by reducing methane emissions: Comparison of abatement costs and global mortality benefits under future methane abatement scenarios, manuscript in preparation, 2008], along with an estimate of the associated costs and public health benefits. Briefly, scenario A corresponds to an 18% (75 Tg a−1) decrease in global anthropogenic CH4 emissions (defined as the agricultural and industrial sectors provided in the CLE inventory) relative to the projected CLE emissions in 2030. Scenario B involves a 29% (125 Tg a−1) decrease in global anthropogenic CH4 emissions in 2030, slightly less than the reductions achieved in the IIASA Maximum Feasible Reductions (MFR) scenario versus CLE in 2030, and should be cost-effective with available technologies at a marginal cost of approximately $315 per ton CH4 ($15 per ton CO2 equivalent). Scenario C requires development of additional control technologies, likely in the large agricultural sector, to achieve a 42% (180 Tg a−1) reduction of global anthropogenic CH4 emissions by 2030.
2.3. Steady State Simulations
 We conduct four “steady state” simulations to diagnose the relative contribution to surface O3 from CH4 oxidation in the free troposphere versus boundary layer (Table 1). In these simulations, we use the BASE emissions for all species besides CH4, but fix atmospheric CH4 mixing ratios to (1) 1760 ppb everywhere (GLOB1760 simulation), (2) 1460 ppb everywhere (GLOB1460), (3) 1460 ppb in the boundary layer and 1760 ppb elsewhere (PBL1460), and (4) 1760 ppb in the boundary layer and 1460 ppb in the free troposphere (FT1460). The model level centered at 750 hPa (top edge at 724 hPa or ∼2.5 km) is included as the uppermost level within the boundary layer. These four simulations were spun up beginning in May 1999 and results are examined for the year 2000. We conduct an additional simulation in which we use the CLE 2030 emissions for non-CH4 O3 precursors and fix CH4 concentrations uniformly to the 700 ppb pre-industrial level, in order to quantify the total contribution of anthropogenic CH4 to tropospheric O3 in the year 2030.
2.4. Model Evaluation
 The annual mean latitudinal bias in our CH4 simulations (BASE and CLE) is compared to the NOAA GMD observations [Dlugokencky et al., 2005] for the year 2004 in Figure 3. For both simulations, the simulated CH4 concentrations are within 5% of the observations at all locations. The BASE simulation has previously been shown to capture much of the observed CH4 rise in the early 1990s, along with the flattening in the late 1990s [Fiore et al., 2006]. A major shortcoming in BASE is the 50% overestimate of the mean 2004 gradient from the South Pole to Alert (195 ppb versus 127 ppb observed). This overestimate is corrected in the CLE simulation (121 ppb gradient) largely due to the use of the Wang et al.  wetland distribution, which also improves the seasonal cycles in the model at northern hemispheric sites [Fiore et al., 2006]. As we show in section 4, the O3 response to CH4 emission reductions is insensitive to biases in the simulated CH4 distribution.
 Global distributions of O3 and its precursors in a different version of MOZART-2 were evaluated with available observations by Horowitz et al.  who showed that the model generally captures the observed O3 seasonality, as well as horizontal and vertical gradients. On the regional scale, however, Murazaki and Hess  previously showed a >20 ppb mean bias in the MOZART-2 simulation of surface O3 as compared to the EPA AIRS monitoring sites over the eastern United States in summer. Our updated treatment of isoprene nitrates decreases simulated July afternoon surface O3 concentrations by 4–12 ppb over the eastern United States [Fiore et al., 2005]. We note that the surface O3 sensitivity to CH4 does not appear to be strongly influenced by the remaining bias as our results below are consistent with the sensitivity previously diagnosed by the GEOS-Chem model, which exhibits a smaller bias compared to U.S. surface O3 observations [Fiore et al., 2002a, 2002b, 2005].
 Most pertinent to our study is the ability of the model to represent the global distribution of NOx. Horowitz et al.  showed that the model typically fell within the observed range of NOx concentrations throughout most of the troposphere. The largest discrepancies in NOx concentrations occurred in surface air near islands, where the model overestimates measurements of clean marine boundary layer air due to mixing of emissions throughout the coarse-resolution grid cell [Horowitz et al., 2003]. When we compare our CLE and BASE NOx simulations for the meteorological year 2004 with the Horowitz et al.  simulation, we find that the NOx distributions are similar in most regions of the globe (not shown). The largest difference is found in the upper troposphere (beginning about 5–7 km), mainly in the tropical Pacific (e.g., over Christmas Island, Tahiti, Guam, the Philippine Sea) and in the southern Atlantic, where NOx concentrations in our simulations are often lower than those by Horowitz et al.  (and the observations) by a factor of two or more. This result likely reflects differences in the lightning NOx distribution which is driven by the NCEP reanalysis in our simulations but by the NCAR MACCM3 meteorology by Horowitz et al. . In the ACCENT Photocomp Experiment 2, MOZART-2 NO2 columns (in a simulation using year 2000 CLE emissions and CH4 concentrations set to 1760 ppb) were consistently 10–30% higher than the model ensemble mean [van Noije et al., 2006]. The comparison with NO2 columns retrieved from the GOME instrument using three different methods varied widely, however, with MOZART-2 falling below the retrieved range in the Eastern U.S. (−3%), Eastern China (−15%) and South Africa (−48%); within the range in Europe and Northern Africa, and exceeding the range in Central Africa (+7%), South America (+39%), and Southeast Asia (+6%) [van Noije et al., 2006].
2.5. Distribution of CH4 Loss and O3 Production in the BASE Simulation
 We examine the latitudinal and vertical distributions of CH4 and tropospheric O3 production in the BASE simulation, focusing here on the final year of the 30-year simulation. The strong temperature dependence of the CH4-OH reaction largely restricts CH4 oxidation to the lower troposphere. Following the approach recommended by Lawrence et al. , we find that 57% and 90% of the CH4 loss by reaction with OH occurs below 750 and 500 hPa, respectively (Table 3). This estimate is somewhat higher than previous work estimating that CH4 oxidation below 500 hPa accounts for ∼80% of the CH4 loss [Spivakovsky et al., 2000; Lawrence et al., 2001]. While it is possible that OH in MOZART-2 may be larger in the lower troposphere than in previous modeling studies, the CH4 lifetime against tropospheric OH is 10.3 years, within the range of other models (8.2–11.7 years based on Stevenson et al. ). Most of the CH4 loss (75%) occurs in the tropics, consistent with the estimate by Spivakovsky et al.  of 78% of CH4 loss between 32°S and 32°N. Table 3 also shows a hemispheric asymmetry, with nearly twice as much CH4 loss occurring north of 30°N than south of 30°S, and 20% more loss in the northern tropics than in the southern tropics. Since the CH4 burden is evenly distributed (<4% difference between the hemispheres), the asymmetry in the CH4 chemical loss reflects the OH distribution (Figure 4a), which in turn is governed by the larger NOx emissions and resulting NOx abundances in the northern hemisphere (Figure 4b). Ozone production in the model is also larger in the northern hemisphere, where shorter-lived anthropogenic O3 precursors are most abundant (e.g., NOx in Figure 4b), contributing ∼60% to the total global production, consistent with the estimate from Horowitz et al. . Our BASE simulation predicts a slightly larger contribution from lower tropospheric O3 production than by Horowitz et al.  (51% versus 46%).
Table 3. Spatial Distribution of CH4 Loss by Reaction With OH (BASE Simulation) and Change in Gross Tropospheric O3 Production (RGLOB-BASE Simulations)
Methane loss by reaction with OH (%)
Change in gross ozone production (Tg a−1)
3. Linearity of Global Annual Mean O3 Response to CH4 Controls
Figure 5 shows that the decreases in tropospheric CH4 and O3 from sustained CH4 emission reductions (RGLOB-BASE) are approaching a steady state after 30 years of simulation. We use the final year of the simulations to diagnose the model “feedback factor”, the ratio of the perturbation lifetime (PT; the decay time for a perturbation such as a pulse of CH4 emissions) to the total atmospheric lifetime (LT = B/LCH4): PT/LT = 1/(1 − s) where B is the total atmospheric CH4 burden, LCH4 is the total atmospheric CH4 loss, and s = δ ln LT / δ ln B [Prather et al., 2001]. We use the difference in lifetimes and burdens between the RGLOB and BASE simulations to calculate s = 0.23, corresponding to PT/LT = 1.30, somewhat smaller than the model range of 1.33–1.45 reported by Prather et al. . In the BASE simulation, LT = 8.7 years, equal to the mean of the range reported for current CTMs (8.7 ± 1.3 years) [Stevenson et al., 2006]. We thus obtain a CH4 perturbation time of 11.3 years and estimate that the model results in year 30 capture 93% (1-e−30/11.3) of the steady state change, ultimately yielding a 23% steady state decrease in atmospheric CH4 (1090 Tg CH4 or 400 ppb surface mixing ratio) from the 18% decrease in total CH4 emissions in RGLOB.
 We scale the tropospheric O3 burden and surface O3 responses (RGLOB-BASE) in year 30 to obtain ultimate steady state decreases of 10.3 Tg tropospheric O3 (blue filled circle in Figure 1) and 1.0 ppb surface O3 (global annual mean). The sensitivity of 0.11 Tg O3 (or 11 ppt surface O3) decrease per Tg CH4 a−1 emission reduction diagnosed from these simulations is slightly lower than the 0.12–0.16 Tg O3 range of the prior work displayed in Figure 1. In particular, the results of Shindell et al.  (green squares in Figure 1) yield a larger sensitivity of ∼0.16 Tg O3 per Tg CH4 a−1 emission.
 Annual mean results from the transient future scenarios (section 2.2) are shown in Figure 2. With the more aggressive controls in scenarios B and C, CH4 concentrations decrease by 2030 to levels last observed in the 1990s and 1980s, respectively (Figure 2b). The tropospheric O3 burden increases from 2005 to 2030 under all scenarios, by 17.0, 12.3, 7.7, and 2.8 Tg for scenarios CLE, A, B, and C, respectively (Figure 2c), reflecting the 19% growth in anthropogenic NOx emissions. The global annual mean 8-h daily maximum (MDA8) surface O3 increases by 1.8 ppb in the baseline CLE scenario. The CH4 control scenarios counteract some of this increase in surface O3, with the most aggressive CH4 controls (scenario C) almost entirely offsetting this increase (total change of +0.2 ppb from 2005 to 2030; Figure 2d). Figures 2c and 2d illustrate the effect of interannual variability in meteorology on tropospheric O3, as evidenced by the repeating 5-year cycle superimposed upon the longer-term trend. This variability likely reflects year-to-year variations in lightning NOx emissions (which vary from 3.1–3.4 Tg N a−1) and/or exchange with the stratosphere (varies from 850–880 Tg a−1), both of which correlate strongly with annual mean tropospheric O3 burdens in the BASE simulation with 1990–2004 meteorology (r2 = 0.80 and 0.92, respectively, for the first 15 year period).
 Although not at steady state, the results in Figure 2 scale approximately linearly to the emission changes, despite differences in the trajectory shapes of the emission controls over the 25 years. By comparing results for 2030 in scenarios A, B, and C with those from the baseline CLE simulation, we obtain a range of 0.06 (scenario A) to 0.08 (scenario C) Tg O3 decrease per Tg CH4 a−1 reduced. These values are lower than that from the RGLOB simulation since the CLE, A, B, and C simulations have not yet reached a steady state. The larger sensitivity implied by scenario C reflects the deeper CH4 reductions that are imposed earlier in this emission control trajectory (Figure 2a), allowing for a relatively larger O3 burden change to be realized by 2030. In an attempt to remove the influence of the shapes of the trajectories in Figure 2a on the results, we assume that the PT of 11.3 years diagnosed from RGLOB-BASE applies for the future scenarios (it may in fact differ since a different base emission inventory is used, which will influence OH). We then define an “effective CH4 emission control” in the year 2030 (ΔEeff-2030) to represent the change in CH4 emissions that would produce a steady state response equal to the transient response in 2030 in each of our simulations (which are not at steady state). The ΔEeff-2030 is calculated as the sum of the emission controls applied in each year from 2005 to 2030, weighted by the fraction of the steady state response that should be realized by 2030. In this manner, we estimate that ΔEeff-2030 is −35, −76, and −117 Tg CH4 a−1 for scenarios A, B, and C, respectively, corresponding to a realization of 47%, 61%, and 65% of the total CH4 emission reductions by 2030. Using the ΔEeff-2030, we estimate revised sensitivities of 0.14, 0.12, and 0.12 Tg O3 decrease per Tg CH4 a−1 “effective” emission reductions, in good agreement with the 0.12–0.16 range of the published results displayed in Figure 1, but slightly higher than the sensitivities obtained above from RGLOB-BASE. Finally, we use these sensitivities to obtain steady state changes in the O3 burden of 10.1, 15.3, and 21.8 Tg for constant emissions at 2030 levels in scenarios A, B, and C, which we show in Figure 1. An alternative approach to diagnose ΔEeff-2030 is to consider the change in the CH4 loss rate between the perturbation and baseline scenarios. With this approach, we estimate ΔEeff-2030 of −44, −86, and −130 Tg CH4 a−1 for A, B, and C, respectively, which translate to sensitivities of 0.11 Tg O3 decrease per Tg CH4 a−1 reduced, equivalent to that in RGLOB-BASE, and within 11–26% of the results obtained via the former approach.
 We next estimate the total contribution of anthropogenic CH4 emissions to tropospheric and surface O3 by comparing our CH4-700 and CLE baseline simulations in 2030. The CH4 loss by tropospheric OH in the CH4-700 simulation is 235 Tg CH4 a−1. Using the lifetimes from Prather et al.  for CH4 loss to the stratosphere (120 years) and to soils (160 years), we estimate the total CH4 loss (equal to the implied emissions) to be 265 Tg CH4 a−1. The implied emissions are thus within 10% of the 245 Tg CH4 a−1 natural plus biomass burning emissions in the baseline scenario (Table 2), indicating that this simulation can be used to estimate the impacts of setting 2030 anthropogenic CH4 emissions to zero. We note that zero anthropogenic CH4 could not possibly be achieved by 2030 even if anthropogenic emissions were shut off instantly in 2005, since some anthropogenic CH4 would still remain in the 2030 atmosphere (∼11% for a PT of 11.3 years). We find that anthropogenic CH4 contributes 40 Tg to the tropospheric O3 burden and 4 ppb to global annual mean surface concentrations in 2030. For comparison with the other simulations in Figure 1, we calculate the ΔEeff-2030 for the baseline CLE scenario (+63 Tg CH4 a−1), to estimate the steady state increase of +26 Tg in the O3 burden. Relative to the steady state CLE value, the CH4-700 simulation yields a decrease of 49 Tg O3 (included in Figure 1) or a sensitivity of 0.11 Tg tropospheric O3 (11 ppt surface O3) per Tg CH4 a−1 effective emission change, consistent with that diagnosed above from RGLOB-BASE.
 The similarity in the steady state O3 response to CH4 in our simulations and in the published literature implies that the CH4 and O3 changes resulting from perturbations to CH4 emissions can be accurately approximated, even when other OH precursors are evolving, by scaling to the effective CH4 emission change relative to the baseline CH4 emission scenario. Multiple computationally expensive transient simulations should only be needed in situations where OH changes substantially, as would be expected if CH4 emissions doubled [Prather, 1996]. In most cases, one steady state simulation relative to the baseline scenario should be sufficient to determine the model sensitivity of O3 (and its spatial distribution; see section 4) to CH4. For estimating changes in global annual mean O3, results can be approximated to within ∼30% without using a model, given the range of 0.11–0.16 Tg O3 burden or 11–15 ppt surface O3 per Tg a−1 change in CH4 emission diagnosed from our simulations and the prior modeling studies in Figure 1.
4. Spatial Distribution of the O3 Response to Changes in CH4 Emissions
 When anthropogenic CH4 emissions are reduced (in the RGLOB versus BASE simulations), gross O3 production and O3 concentrations decrease everywhere (Table 3 and Figure 6a), with the largest percentage changes occurring in the southern hemisphere (Figure 6b) where CH4 contributes to a larger fraction of the O3 production (since NMVOC abundances are lower than in the northern hemisphere). Annual mean O3 concentrations decrease by 0.5–2.0 ppb throughout the troposphere (Figure 6a), with the largest decreases centered around 30°N where OH and NOx are enhanced in the lower troposphere compared to other latitudes (Figure 4). West and Fiore  previously noted an enhanced response of surface O3 to CH4 decreases centered around 30°N and suggested that it reflects O3 production from CH4 in the free troposphere at these latitudes where NOx is abundant, followed by downwelling of that air to the surface, a hypothesis that we examine further in section 5.
 We plot in Figure 7 the tropospheric O3 columns, relevant for climate forcing, as well as MDA8 O3 concentrations in surface air, an indicator of air quality, in the BASE simulation. The highest annual mean tropospheric O3 columns and MDA8 surface O3 levels occur in the northern hemisphere over and downwind of continental regions where anthropogenic emissions of O3 precursors are largest (Figures 7a and 7e). Figure 7 also shows the O3 response to CH4 emission changes, which exhibits considerable spatial structure.
 Similar patterns of decreases in tropospheric O3 columns in response to CH4 controls occur in the RGLOB (Figure 7b) and scenario B (Figure 7c) simulations (r = 0.98). While maximum O3 decreases are centered at 30°N (as in Figure 6a), large decreases in tropospheric O3 columns also extend southward into the tropical Atlantic ocean and Africa. Figures 7f and 7g show broadly similar patterns in the decreases of annual mean MDA8 surface O3 from CH4 controls, but the correlation is weaker than for the tropospheric O3 columns (r = 0.88). Figure 7f indicates that the largest benefits to air quality should occur over land, whereas the maximum decreases in surface O3 in Figures 7g and 7h are over the Gulf of Mexico, North Atlantic, and extend eastward across northern Africa and the Middle East into the Arabian Sea and Bay of Bengal. These discrepancies stem from differences in the NOx emissions in the BASE and CLE inventories. For example, the model predicts a strong enhancement of O3 production (and a strong sensitivity to CH4) in NOx-saturated ship track plumes in the CLE scenario, which does not occur in the sustained CH4 reduction scenarios where ship NOx emissions were excluded (section 2.1). The dispersion of ship NOx emissions into a coarse model grid box likely results in an artificially high O3 production efficiency [Kasibhatla et al., 2000; Liang and Jacobson, 2000] and we thus expect that the sensitivity of O3 to CH4 oxidation is overstated in the ship tracks in Figures 7g and 7h. Nevertheless, this result highlights that the O3 response to CH4 depends strongly on NOx; accurately representing the global NOx distribution is thus important for simulating this sensitivity.
 The geographical patterns of the O3 response to CH4 controls are identical among the future CH4 control scenarios (A, B, C), implying that the O3 response is determined by the distribution of the shorter-lived species OH and NOx which govern the O3 production from CH4 oxidation. As such, we focus in Figure 7 and thereafter on results from scenario B, noting that the magnitude of the O3 change scales linearly to the “effective CH4 emission controls” defined in section 3.
 Comparing the O3 response in scenario B with that in the CH4-700 simulation (third and fourth columns in Figure 7) illustrates that the O3 response exhibits an identical spatial pattern in a full transient CH4 simulation and in a simulation with CH4 set to a globally uniform, fixed value (spatial correlation of r = 1.0 between Figures 7c and 7d, and between Figures 7g and 7h). This result indicates that the O3 response is insensitive to biases in the simulated CH4 distribution. Furthermore, the linearity of the O3 response to CH4 (Figure 1 and section 3) implies that the spatial pattern can be scaled to different magnitudes of CH4 controls.
 Finally, we examine the extent to which the spatial pattern of the surface O3 response varies with changes in meteorology in the final 15 years of our sustained CH4 reduction simulations (RGLOB-BASE). The interannual variability is <5% in most world regions, except for in the tropics, on the west coast of central South America, and off the southwestern coast of United States where it is 10-20% (not shown). We conclude that the spatial patterns in Figure 7 are fairly robust to fluctuations in present-day meteorology.
5. Surface O3 Contribution from CH4 Oxidation in the Free Troposphere Versus Boundary Layer
 Surface O3 is influenced by CH4 oxidation both within the boundary layer and in the free troposphere (followed by subsidence of this air to the surface). With our steady state simulations (section 2.3), we separate the contributions to surface O3 from CH4 oxidation in these two regions. We focus our analysis on the PBL1460 and FT1460 simulations relative to GLOB1760 since we find our results to be linear; the O3 response to reductions of CH4 throughout the atmosphere (GLOB1460-GLOB1760) are virtually identical to the sum of the O3 changes from reducing CH4 in the boundary layer (PBL1460-GLOB1760) and from reducing CH4 in the free troposphere (FT1460-GLOB1760).
 At nearly all surface locations, oxidation of CH4 in the boundary layer contributes more than 50% of the change in annual mean MDA8 O3 resulting from CH4 controls (Figure 8), with a global mean contribution of 64%. The 300 ppb decrease in boundary layer CH4 concentrations decreases MDA8 O3 in surface air by up to 1 ppb O3 in the NOx-rich regions of the northern midlatitudes and in southern Africa (Figure 8a). Surface O3 is also decreased by the reduction in CH4 oxidation in the free troposphere near the downwelling branches of the Hadley cell, at roughly 30°N and 30°S, particularly over the Sahara region of North Africa, the Middle East, and the Caribbean Sea where the 300 ppb decrease in free tropospheric CH4 concentrations lowers surface O3 by ∼0.5 ppb (Figure 8b).
 As a case study, we plot the composite maximum MDA8 O3 decrease over the United States resulting from a 300 ppb decrease in CH4 in the planetary boundary layer (below 724 hPa; PBL1460-GLOB1760) in Figure 9. Southern California and the New York-New Jersey region in the northeastern United States exhibit a larger peak O3 decrease (>2–3 ppb) from the CH4 reduction than the rest of the country. In these NOx-saturated regions, CH4 oxidation within the planetary boundary layer accounts for 70–80% of the total O3 decrease from CH4 reductions (Figure 8c). We conclude that the surface O3 response to CH4 is strongly enhanced in locations with NOx-saturated chemistry, and weakly enhanced in regions of downwelling air.
 The question arises as to whether our coarse resolution global model accurately represents the response of O3 to CH4 at the urban scale. Higher-resolution model calculations of O3 production in urban airsheds support the conclusion that O3 production in polluted urban areas is sensitive to CH4. For example, it has been estimated that despite its low reactivity, CH4 oxidation contributes approximately 10–20% of the O3 formed downwind of London in the urban plume [Hough and Derwent, 1987; Derwent et al., 1991]. Martien and Harley  used adjoint methods to identify a strong VOC-sensitivity in and downwind of core urban areas. The consistency of our global model results with the conclusions from these studies implies that the global model is useful in assessing the qualitative O3 response to CH4 in urban areas, although the accuracy of the results for a particular urban core will depend upon the model representation of the VOC-NOx sensitivity.
6. Sensitivity of Tropospheric O3 to the Location of CH4 Emissions
 Given the long lifetime of CH4, the O3 produced from CH4 oxidation is expected to be independent of the location of the CH4 emissions. Instead, the spatial pattern of the O3 from CH4 oxidation should be controlled by the distributions of OH and NOx, which have much shorter lifetimes and affect both the location of the CH4 oxidation and the amount of O3 produced per CH4 molecule oxidized. We directly test this assumption by comparing the results from the RASIA simulation, in which Asian anthropogenic CH4 emissions are set to zero, with those from the RGLOB simulation, in which an equivalent CH4 emission reduction was distributed globally.
Figure 10 compares the MDA8 surface O3 in RASIA and RGLOB simulations in year 11 (note that The RASIA simulation was stopped after 11 years since there was little difference in the O3 response from that in RGLOB). Except for the immediate source region over Asia, there is <10% difference in other northern hemispheric source regions, <5% difference over the rest of the northern hemisphere, and <1% difference in the southern hemisphere. Annual mean tropospheric O3 columns in year 11 are even more similar between the simulations (<0.5% differences; not shown). These results imply that control options could be targeted starting with the least cost reductions available anywhere in the world since the global air quality and climate benefits of CH4 controls do not depend strongly upon the location of the emission reductions.
7. Regional Air Quality
 We first investigate the seasonality of the regional surface O3 response to CH4 using the simulations with sustained CH4 reductions (RGLOB-BASE; Figure 11). Both Europe and South Asia show some seasonal variation (0.8 and 0.5 ppb amplitude, respectively) whereas there is little seasonality over the United States, East Asia and Africa. These results are consistent with the findings from a multimodel intercomparison of the surface O3 response to a 20% decrease in global CH4 mixing ratios, although the regional definitions in that study are not identical [Task Force on Hemispheric Transport of Air Pollution (TF HTAP), 2007]. The enhanced response over Europe in summertime and over South Asia in April may reflect a more NOx-saturated O3 formation regime than during the peak O3 seasons in the United States and Africa, where biogenic VOC are more abundant.
 Next, we examine the percentage of model grid box-days where MDA8 O3 ≥ 70 ppb in surface air (a metric for polluted conditions) by season for the selected world regions in the future scenarios. Figure 12 shows that air quality degrades substantially under the baseline CLE scenario from 2005 to 2030 in South Asia, where large growth in emissions (+58% for NOx) from the transport and power generation sectors is projected [Dentener et al., 2005]. From 2005 to 2030, the percentage of model grid cell-days with MDA8 O3 ≥ 70 ppb increases by 4%, 1%, and 2% in the United States, Europe, and Africa, respectively, during summer, the season in which high-O3 events peak for those regions (Figure 12). Peak incidences occur in spring in the Asian regions, with increases of 4% (for East Asia) and 14% (for South Asia) in the percentage of model grid cell-days with MDA8 O3 ≥ 70 ppb from 2005 to 2030. These findings are consistent with prior modeling studies in which the CLE scenario produced surface O3 increases of ∼10 ppb over the Indian subcontinent from 2000 to 2030, and of ∼2 ppb globally, with little change over some regions of North America, Europe and Asia [Dentener et al., 2006b; Stevenson et al., 2006].
 We evaluate here the ability of the global model to represent the threshold statistic presented in Figure 12, and determine whether those results can be extended to other thresholds. A similar statistic was previously used by Fiore et al. [2002a] who found that a global model was able to simulate adequately the percentage of total grid box-days above 70 ppb in the United States during the summer of 1995 (observed values of 5%, 15% and 36% for thresholds of 80, 70, and 60 ppb and simulated values of 1%, 10%, and 37%). In our CLE simulation for 2005, we find that 5%, 13% and 26% of grid box-days over the United States in summer exceed 80, 70, and 60 ppb thresholds, respectively. We further find that the changes due to CH4 emission controls are relatively insensitive to the chosen threshold, such that in 2030, scenario A yields a ∼5% decrease in the incidence of grid box-days above 60, 70, and 80 ppb relative to the CLE scenario. For scenarios B, C, and CH4-700, the corresponding percentage changes are ∼10%, ∼15%, and ∼35%. Below, we focus on results for the 70 ppb threshold.
 All of our CH4 reduction scenarios decrease the percentage of model grid cell-days with MDA8 O3 ≥ 70 ppb in 2030, by up to 3% (Europe) to 12% (South Asia) when all anthropogenic CH4 is removed (Figure 12). The cost-effective CH4 reductions in scenario B yield percentage decreases of 1%, 1%, and 2% in Africa, Europe, and the United States in summer, and of 3% in both Asian regions in spring compared to CLE. For all regions, the largest decreases in the percentage of high-O3 events occurs in the same season as the peak number of model grid cell-days with MDA8 O3 ≥ 70 ppb, except for the United States where the decrease is slightly larger in spring.
 During summer in Europe and fall in East Asia, aggressive CH4 controls (scenarios B and C) would decrease high-O3 events to below 2005 levels. For the United States and Africa, only the CH4-700 simulation reduces high-O3 events below the 2005 levels. The dramatic growth of O3 precursor emissions over South Asia leads to additional model grid-cell days where MDA8 O3 ≥ 70 ppb, even with the drastic reduction of CH4 to pre-industrial levels.
 In Figure 13, we assess the potential for CH4 controls to reduce MDA8 O3 in surface air under background, average, and highly polluted conditions. We focus on the high-O3 season for each region, when the incidence of grid-square days ≥70 ppb is maximum in the year 2030, and plot the distribution of the change in MDA8 surface O3 in response to the CH4 controls in scenario B (B-CLE), as a function of the MDA8 O3 values in the baseline CLE simulation for the year 2030. Within the United States, results are shown for the four quadrants in order to illustrate differences that can occur within large continental-scale regions.
 The median O3 decrease tends to grow linearly to a maximum of ∼2 ppb at CLE 2030 MDA8 O3 concentrations of 60–90 ppb (Figure 13). In contrast, the median response in the western U.S. quadrants increases linearly over the entire O3 distribution, with days when total O3 > 90 ppb nearly always associated with decreases of >1 ppb and >2.5 ppb for the northwestern and southwestern quadrants, respectively; these conditions are typically associated with boundary layer stagnation and the strong O3 response reflects the relatively high sensitivity to CH4 oxidation in the polluted boundary layer in southern California (Figures 8 and 9). In Figure 13, O3 also exhibits a strong sensitivity to CH4 on the most polluted grid-cell days (>90 ppb) in Europe, East Asia, and Africa, often decreasing by >1 ppb under scenario B.
 The median response in the southeastern U.S. quadrant peaks at ∼40 ppb total O3 and then weakens as total O3 increases further. This feature reflects the meteorology in the southeastern U.S. in summer. The cleanest conditions are associated with inflow of marine air from the Gulf of Mexico; the stronger median sensitivity at 10-40 ppb total O3 over the southeastern U.S. compared to the other U.S. regions in Figure 13, stems from the enhanced contribution of CH4 oxidation in the free troposphere (followed by mixing into the boundary layer) over the Gulf of Mexico (Figure 8). The most polluted conditions are associated with stagnation events that suppress mixing between the free troposphere and boundary layer [Logan, 1989; Eder et al., 1993; Jacob et al., 1993]. Given that O3 chemistry over the southeastern United States is NOx-sensitive due to abundant biogenic VOC emissions [Chameides et al., 1988], CH4 contributes less to O3 on the most polluted days than in the NOx-saturated regions of the western and northeastern United States.
 Peak decreases of ≥4 ppb (indicated by the lower extent of the vertical lines in Figure 13) occur near the center of the overall distribution, within the range of 40–60 ppb for Europe; 20–70 ppb for Africa; 40–70 ppb for East Asia, 30–60 ppb for South Asia, and 30–100 ppb in the United States. Emissions of NOx, CO, and NMVOC emissions on foreign continents were previously shown to exert a maximum influence on U.S. surface O3 near the middle of the total O3 distribution, under conditions of strong mixing with the free troposphere [e.g., Fiore et al., 2002b; Li et al., 2002]. Similarly, the highest U.S. background O3 concentrations (estimated in simulations where North American anthropogenic emissions of NOx, CO, and NMVOC were turned off but O3 generated from CH4 oxidation was included) were found at the center of the overall surface O3 distribution and attributed to O3 mixing down from the free troposphere [Fiore et al., 2003]. The qualitative similarity of the results in Figure 13 to prior studies provides some confidence that our findings are robust to the model bias in surface O3 (section 2.4) although we cannot rule out some influence of the bias on the quantitative results, particularly at the high tail of the O3 distribution.
 We conclude that the scatter in the MDA8 O3 changes in Figure 13 likely reflects differences in meteorology (mixing with the free troposphere versus local stagnation) as well as in chemistry (the local sensitivity of ozone production to CH4). The smaller sensitivity to CH4 at the low end of the total O3 distribution reflects cleaner air and does not necessarily imply a lack of mixing between the boundary layer and the free troposphere. The larger than average response in the upper tail of the total O3 distribution over most regions typically occurs under stagnant conditions and is driven by the local NOx-saturated chemistry.
 The global mean O3 radiative forcing under the CLE baseline scenario (2030–2005) is 0.065 W m−2, near the multimodel mean in the CLE scenario from 2000 to 2030 (0.063 ± 0.015 W m−2) reported by Stevenson et al. . The strongest forcing occurs in the tropics, particularly over South Asia and the Middle East regions (0.15–0.21 W m−2) where the increases in tropospheric O3 columns are largest (not shown). All of the CH4 control scenarios yield a similar spatial pattern of decreases in O3 forcing, with the largest decreases occurring broadly in the tropics and over the Middle East and northern Africa (Figure 14). The larger forcings in the tropics as compared to the poles reflect the spatial pattern of the change in tropospheric O3 columns (Figure 7c) and the higher sensitivity of the forcing to the O3 column at these latitudes (forcing efficiencies of 0.04–0.06 versus a global mean of 0.036 W m−2 DU−1). Table 4 shows the global mean forcing from both CH4 and O3 for each 2030 sensitivity simulation. Aggressive CH4 controls (scenarios B and C) would offset the positive net forcing from CH4 and O3 predicted to occur otherwise from 2005 to 2030 under the baseline CLE scenario. While the global mean forcing from CH4 and O3 roughly cancel in scenario B, we expect regional variations; for example, the positive forcing from O3 may exceed the negative forcing from CH4 in the tropics, with the opposite impact near the poles. Eliminating anthropogenic CH4 emissions would reduce global mean radiative forcing from CH4 and O3 by 0.6 W m−2 relative to 2005 (by 0.7 W m−2 relative to the CLE 2030 baseline).
Table 4. Adjusted Radiative Forcing (W m−2) in 2030 Versus 2005 Due to Changes in Tropospheric CH4 and O3
Calculated in the GFDL AM2 radiative transfer model, including stratospheric adjustment following the approach of Naik et al. . The tropospheric O3 forcing is dominated by longwave radiation, with a 21–25% contribution from shortwave.
 The potential to improve both climate and air quality by regulating CH4 emissions has sparked discussion of CH4 controls as a component of future air pollution policy [EMEP, 2005; TF HTAP, 2007]. Our analysis expands upon prior modeling studies [Fiore et al., 2002a; Dentener et al., 2005; West et al., 2006, 2008] to provide a basis for more fully assessing the monetary costs and benefits associated with managing global O3 pollution by controlling CH4 emissions [West et al., manuscript in preparation, 2007]. We employed two sets of full-chemistry multidecadal transient simulations in the MOZART-2 global CTM to characterize the response of CH4 and O3 to changes in CH4 emissions, and to estimate the O3 air quality and climate benefits that would result from CH4 controls. We further diagnosed the relative impact of CH4 oxidation in the free troposphere versus in the boundary layer on surface O3, as well as the anthropogenic CH4 contribution to tropospheric O3, with a set of steady state simulations.
 Cost-effective future CH4 controls through 2030 (our scenario B; <$315 per ton CH4) would offset the projected positive climate forcing from CH4 and O3, and reduce the incidence of high surface O3 events in all regions relative to the CLE baseline scenario; air quality would improve even relative to 2005 in Europe (summer and fall) and East Asia (fall). The mean O3 decrease from CH4 controls is often largest near the middle of the total O3 distribution (e.g., ∼40–70 ppb total O3). Even during high-O3 events, O3 often decreases by >1 ppb in scenario B by 2030. Controlling CH4 emissions could thus help to achieve compliance with air quality standards, particularly in situations where high-O3 events are frequently within a few ppb of a threshold concentration, as is often the case in the United States.
 Combining our results with estimates from the published literature, we find that global tropospheric O3 decreases approximately linearly with reductions in CH4 emissions: 0.11–0.16 Tg tropospheric O3 or 11–15 ppt surface O3 per Tg CH4 a−1. This sensitivity implies a total contribution from present-day anthropogenic CH4 emissions of ∼50 Tg to the tropospheric O3 burden and ∼5 ppb to surface O3. The similarity of the global mean sensitivity of O3 to CH4 in our simulations and those from other models shown in Figure 1, and of the spatial pattern in our transient and steady state simulations (Figures 7c and 7g versus Figures 7d and 7h) indicates that the O3 response to CH4 is insensitive to biases in the simulated CH4 distribution and thus to errors in the spatial distribution of emissions. We further expect that the sensitivity of O3 to CH4 in MOZART-2 is fairly robust to the positive surface O3 bias versus observations over the eastern United States [Fiore et al., 2005; Murazaki and Hess, 2006], based on the consistency of our results with the other models in Figure 1, and of the results in Figures 12 and 13 with prior simulations with the GEOS-Chem model which has a much smaller bias compared to the observations [Fiore et al., 2002a, 2002b, 2005].
 We defined an “effective CH4 emission change” to facilitate comparisons between transient and steady state results. In the case of our future simulations, the effective CH4 emission change in 2030 (ΔEeff-2030) is the sum of the emission controls applied in each year from 2005 to 2030, weighted by the fraction of the steady state response that should be realized by 2030. We showed that once the relationship between ΔEeff-2030 and the resulting CH4 concentration is established for a baseline scenario (in which emissions of NOx and other OH precursors are evolving), the CH4 and O3 changes that would result from perturbing CH4 emissions by a different magnitude relative to that scenario can be accurately approximated, eliminating the need for multiple computationally expensive transient simulations (as long as the OH is relatively constant). In many cases, one steady state simulation relative to the baseline scenario should be sufficient to determine the model sensitivity of O3 to CH4, including the spatial distribution of the O3 response. For estimating changes in annual global mean O3, results can be approximated to within ∼30% using the sensitivity estimated here, without needing additional model simulations.
 The decreases in surface O3 and tropospheric O3 columns (relevant for air quality and climate, respectively) arising from CH4 emission control are largely independent of source location, with the exception of ∼10–20% enhancements in source regions. Although the surface O3 response to CH4 emission reductions is relatively homogenous across the globe compared with the response to controls on NOx emissions [West et al., 2008], the decreases in surface O3 are not uniform, reflecting a combination of local meteorological and chemical conditions.
 We find in the model that global annual mean maximum daily 8-h (MDA8) surface O3 is nearly twice as sensitive to CH4 in the planetary boundary layer (below ∼2.5 km) than to CH4 in the free troposphere. The surface O3 response to CH4 is strongly enhanced in locations with NOx-saturated chemistry, including at the high tail of the O3 distribution (e.g., in southern California). Weaker enhancements occur in regions where surface air mixes frequently with the free troposphere, either due to subsiding air masses (such as over northern Africa and the Middle East) or due to active convection (such as over the Caribbean Sea and Gulf Coast of the United States).
 Since the O3 response to CH4 depends strongly on NOx, we underscore the need for a better understanding of the global NOx distribution. A key policy implication from our study is that the efficacy of CH4 emission reductions for addressing global air quality and climate goals is nearly independent of the location of the emissions. This result is particularly important given the rising cost of implementing additional controls on the traditional O3 precursors in many nations where these emissions have already been regulated for decades. Accurate determination of CH4 emissions by region and sector will nevertheless be critical for estimating the costs and technical feasibility of various options for CH4 control, as well as for understanding the relative contributions from anthropogenic and natural CH4 sources.
 We are grateful to F.J. Dentener for providing the baseline CLE and MFR emissions and J.S. Wang for providing the biogenic emissions, and to E. Baum, J. Chaisson, R.G. Derwent, H. Levy II, R. A. Harley, D. D. Parrish, D. Shindell, S. Sillman, and A. Zuber for insightful discussions. We acknowledge funding from Luce Foundation via Clean Air Task Force.