Changes in emissions of ozone (O3) precursors affect both air quality and climate. We first examine the sensitivity of surface O3 concentrations (O3srf) and net radiative forcing of climate (RFnet) to reductions in emissions of four precursors – nitrogen oxides (NOx), non-methane volatile organic compounds, carbon monoxide, and methane (CH4). We show that long-term CH4-induced changes in O3, known to be important for climate, are also relevant for air quality; for example, NOx reductions increase CH4, causing a long-term O3 increase that partially counteracts the direct O3 decrease. Second, we assess the radiative forcing resulting from actions to improve O3 air quality by calculating the ratio of ΔRFnet to changes in metrics of O3srf. Decreases in CH4 emissions cause the greatest RFnet decrease per unit reduction in O3srf, while NOx reductions increase RFnet. Of the available means to improve O3 air quality, therefore, CH4 abatement best reduces climate forcing.
 Concentrations of ozone (O3) have increased historically at the surface in industrialized regions, and in the global background troposphere [Vingarzan, 2004]. Actions to decrease tropospheric O3 to improve air quality have emphasized reducing emissions of short-lived O3 precursors – nitrogen oxides (NOx), non-methane volatile organic compounds (NMVOCs), and carbon monoxide (CO) – on urban and regional scales. Methane (CH4), a long-lived O3 precursor (lifetime ∼9 yr), contributes to the global background concentration of tropospheric O3, and CH4 emission controls have received recent attention as a means of international O3 air quality management [Fiore et al., 2002; Dentener et al., 2005; West and Fiore, 2005; West et al., 2006].
 Reducing emissions of any precursor to improve O3 air quality also influences climate, as such changes affect the concentrations of and radiative forcing from both O3 and CH4. Reductions in surface NOx emissions, for example, decrease O3 and global concentrations of the hydroxyl radical (OH). Since reaction with OH is the major sink for CH4, NOx reductions increase the atmospheric lifetime and concentrations of CH4. The resulting positive radiative forcing from increased CH4 roughly cancels, or slightly exceeds, the negative forcing from decreased O3 globally [Fuglestvedt et al., 1999; Wild et al., 2001; Fiore et al., 2002; Berntsen et al., 2005; Naik et al., 2005; Shindell et al., 2005]. In contrast, reductions in emissions of NMVOCs, CO, and CH4 increase OH and decrease the forcing from both O3 and CH4 [Prather et al., 2001; Fiore et al., 2002].
 These changes in CH4 will also affect O3 on the longer time scale of the CH4 perturbation lifetime. In the case of NOx controls, the CH4 increase induces a long-term O3 increase, which partially counteracts the short-term O3 decrease. Conversely, controls on NMVOCs and CO cause a long-term O3 decrease in addition to the short-term O3 decrease. These long-term changes in O3 have been included in estimates of net radiative forcing [Wild et al., 2001; Berntsen et al., 2005; Naik et al., 2005]; here we also quantify their relevance for long-term surface O3 air quality.
 Local actions to improve O3 air quality therefore affect surface O3 and radiative forcing on global scales; while the global effects of emission changes in a single airshed are expected to be small, the cumulative effects of such actions in many polluted regions are analyzed here. The first objective of this study is to simulate the sensitivity of surface O3 concentrations, and of the net radiative forcing due to O3 and CH4, to global emission reductions of NOx, NMVOCs, CO, and CH4, using consistent methods. In doing so, we account for the long-term changes in O3 via CH4. These estimates can be used together with information on the costs and feasibility of reducing precursor emissions, in the integrated planning of actions to address O3 air quality and climate change. Changes in aerosol radiative forcing are not included in our estimate of net radiative forcing, but are estimated here to be small (section 3.1.). The second objective addresses the effects of actions to improve surface O3 air quality on radiative forcing. Because precursor emissions are being reduced in many nations to improve O3 air quality, and because such actions affect climate, we compare the radiative forcing consequences of decreasing surface O3 through reductions in emissions of each precursor. We normalize the change in net radiative forcing by changes in surface O3 concentration metrics, and present this ratio as an indicator of the climate forcing resulting per unit improvement in O3 air quality.
 We use the MOZART-2 global three-dimensional model of tropospheric chemistry and transport, driven with meteorology from the middle atmosphere version of the Community Climate Model (MACCM3), with a horizontal resolution of about 2.8° by 2.8° and 34 vertical levels. Emissions in the base simulation are representative of the early 1990s. The base simulation includes only minor model updates from previous model versions [Horowitz et al., 2003; Naik et al., 2005] that have been thoroughly evaluated with measurements [Horowitz et al., 2003].
 We simulate the effects of sustained uniform 20% reductions in global anthropogenic emissions of four precursors individually (NOx, NMVOCs, CO, and CH4), relative to the base simulation. The anthropogenic sources reduced in these experiments include the combustion of biofuels (among other sources), but exclude other forms of biomass burning. Changes in emissions of each gas-phase precursor are assumed to occur independently; we neglect possible changes in emissions of co-emitted species due to the control action. A fixed global CH4 mixing ratio of 1700 ppbv is used in the base and NOx, NMVOC, and CO control simulations. In the CH4 control case, we decrease the CH4 mixing ratio to 1460 ppbv, the steady-state mixing ratio that would result from a 20% anthropogenic emission reduction, based on the CH4-OH feedback factor of 1.33 derived from the model (see auxiliary material). All simulations are conducted for 25 months, with results presented for the final 12 months.
 The short-term O3 response to changes in NOx, NMVOC, and CO is simulated directly in the model, and is assumed constant into the future. This short-term O3 change adds to the long-term change via CH4 (the “primary mode” [Prather, 1996]) to give the net O3 change at steady state [Wild et al., 2001]. Previous studies have typically focused only on the short-term O3 effects of NOx, NMVOC and CO reductions [Fuglestvedt et al., 1999; Fiore et al., 2002], or used other model results to estimate the long-term response [Berntsen et al., 2005; Naik et al., 2005]. We estimate the long-term O3 response by scaling the hourly O3 change in each model grid cell from our CH4 emission reduction simulation to the implied CH4 changes derived from the other precursor reduction simulations (based on the simulated change in CH4 lifetime and the CH4-OH feedback factor; see equations in auxiliary material). Adding the estimated long-term change in O3 to the simulated short-term change, and assuming O3 scales linearly with CH4, is supported by previous model results [Fiore et al., 2002; Shindell et al., 2005].
 Long-term O3 changes will gradually approach their final response according to the modeled perturbation lifetime of CH4 of 12.1 yr, reaching 56% of the final response in 10 yr, 81% in 20 yr, and 92% in 30 yr. Projected changes in emissions of O3 precursors over 30 yr are expected to have only minor effects on the CH4-O3 sensitivity [West et al., 2006] and on the CH4-OH feedback factor [Prather et al., 2001].
3.1. Changes in Surface O3 Metrics and Net Radiative Forcing
Figure 1 shows the response of surface ozone (O3srf) to 20% decreases in anthropogenic precursor emissions. We show the daily maximum 8-hour O3 averaged over the “O3 season,” which we define for each grid cell individually as the consecutive three-month period with highest average 8-hr. O3 in the base simulation. The reduction in O3srf due to the decrease in CH4 emissions is widespread globally, with the largest effects (>1 ppbv) in North Africa, the Middle East, and Europe (due to greater down-welling from the free troposphere and availability of NOx) in agreement with previous results using different models or meteorology [Fiore et al., 2002; West and Fiore, 2005; West et al., 2006]. In contrast, the responses to changes in NOx, NMVOCs and CO are concentrated near source regions, with changes in background concentrations occurring mainly in the temperate Northern Hemisphere. In some locations (e.g., Northern Russia), NOx reductions increase the modeled O3srf due to the decreased local destruction of O3 by reaction with fresh NO emissions.
Figure 2c also includes changes in O3srf metrics relevant for air quality and human health, including the population-weighted daily maximum 8-hr. O3, averaged over the whole year and over the O3 season. The 20% NOx reduction yields the greatest decrease in the population-weighted O3srf metrics, as the ΔO3srf is concentrated in populous source regions.
 In the United States, O3srf responds strongly to the 20% NOx reduction, reflecting high regional NOx emissions and NOx-sensitive chemistry due to the high emissions of biogenic NMVOCs in the eastern US. In Europe, O3srf is particularly sensitive to changes in CH4. The NMVOC emission reduction causes rather small changes in O3srf, in part because of large biogenic NMVOC emissions (the 20% anthropogenic reduction is only 1.0% of total NMVOC emissions). Note that the global model likely underestimates the effects of changes in NMVOC emissions on O3 air quality in urban areas, particularly for the population-weighted metrics, due to its coarse resolution and lack of detail for highly reactive VOCs. For NOx decreases, global models may either under- or overestimate O3srf changes in urban regions [Liang and Jacobson, 2000; Karamchandani et al., 2002]. Local sensitivity to changes in emissions of NMVOCs and NOx is best determined using a local or regional model which has been tested for local conditions.
Figure 2c also shows a notable difference between the short-term and steady-state ΔO3srf metrics, due to long-term changes in CH4. The 20% NOx reduction causes long-term increases in the population-weighted O3srf metrics of about 0.2 ppbv, which counteract the short-term O3srf decreases (1.3 to 3.4 ppbv) by 6 to 14%. For the CO reduction, the short-term ΔO3srf (0.4 to 0.5 ppbv) is amplified by about 0.1 ppbv at steady state (a 16–21% increase). For NMVOCs, the additional long-term O3srf decrease is small (about 0.01 ppbv). These long-term changes in O3srf are globally widespread, following the spatial pattern of ΔO3srf from the CH4 perturbation (Figure 1).
 Changes in net radiative forcing (RFnet) are shown in Figure 2d. The largest negative RFnet results from the CH4 reduction, with 76% of the ΔRFnet from the decrease in CH4 itself, and the remainder from the decrease in O3. While reducing each precursor causes a negative forcing due to O3, the NOx reduction causes a larger positive CH4 forcing, producing a positive global average ΔRFnet. This agrees with previous results showing that the ΔRFnet is positive for surface NOx emission reductions in all world regions [Fuglestvedt et al., 1999; Berntsen et al., 2005; Naik et al., 2005] (see auxiliary material for a comparison with other RFnet estimates).
 Changes in ozone precursor emissions also affect the concentrations of aerosols, through changes in oxidant chemistry. We find that the CH4 reduction decreases the tropospheric burden of sulfate aerosol by 0.23%, in agreement with previous results [West et al., 2006]. This decrease results from a competition between decreased heterogeneous sulfate production by hydrogen peroxide (−0.9 Tg yr−1, or −1.1%) and increased gas-phase production by OH (+0.55 Tg yr−1, or +2.5%). Decreases in other precursor emissions are likewise estimated to decrease the global sulfate burden: −0.16% when decreasing CO, −0.14% for NMVOCs, and −0.06% for NOx. The global mean positive radiative forcing by the direct effect resulting from each of these decreases in sulfate are estimated to be less than ∼0.002 W m−2 (see auxiliary material), smaller than the forcings in Figure 2 but potentially important for local forcing [Unger et al., 2006].
3.2. Change in RFnet Per Unit Change in O3srf Metrics
 We consider the ratio ΔRFnet/ΔO3srf as an indicator of the effects on radiative forcing of actions to decrease metrics of O3 air quality by a given amount. High positive values indicate that actions to reduce O3srf by one unit cause a large negative radiative forcing. In Figure 2e, the CH4 emission reduction causes the largest decrease in RFnet per unit ΔO3srf, mainly due to the decreased forcing from CH4 itself. Reductions in NOx increase RFnet, causing a negative ΔRFnet/ΔO3srf. Abatement of CO emissions leads to a greater ΔRFnet/ΔO3srf than for NMVOCs, as the ratio of the reduction in CH4 to O3 is greater for CO. We find that the order of ΔRFnet/ΔO3srf (CH4, CO, NMVOCs, NOx) is the same for all ΔO3srf metrics considered, and we expect this order to be robust over model uncertainties as well. If ΔRFnet were evaluated after a short time period (<10 yr), however, the change in short-term O3 forcing would dominate, and the order of the precursors would likely differ.
 We first consider the sensitivity of surface O3 air quality and net radiative climate forcing to reductions in global O3 precursor emissions. For O3 air quality, the 20% reduction in anthropogenic NOx emissions has the greatest effect on the population-weighted O3srf metrics analyzed, followed by CH4, although the long lifetime of CH4 delays realization of the O3 decrease. The CH4 reduction causes the greatest decrease in RFnet, mainly because of the direct reduction in CH4 forcing, while NOx emission reductions increase RFnet. Second, we present an indicator of the climate forcing resulting from actions to improve O3 air quality (ΔRFnet/ΔO3srf). We find that of the means of decreasing O3srf metrics by one unit, abatement of CH4 emissions best reduces radiative forcing.
 This research also demonstrates that the long-term changes in surface O3 concentrations via changes in CH4 are substantial for NOx and CO reductions, and relevant for air quality management. While the 20% anthropogenic NOx emission reduction decreases short-term O3srf, it also causes a long-term global increase of about 0.2 ppbv in population-weighted O3srf metrics, which counteracts the short-term decrease by 6 to 14%. For CO, the long-term O3srf reduction of 0.1 ppbv adds 16 to 21% to the short-term decrease, while the long-term O3srf reduction of the 20% NMVOC reduction is small (0.01 ppbv).
 We also find that reducing emissions of each precursor decreases the global sulfate aerosol burden (<0.3% for the 20% anthropogenic reductions). Future research should further address the effects of these emission changes on all relevant aerosol species (including secondary organics), and evaluate impacts on both surface air quality and global and regional aerosol radiative forcing. Further, while the methods presented here are convenient for considering short-term and steady-state changes, future research should conduct transient simulations with CH4 modeled explicitly, to explore possible interactions between reductions in different species under future scenarios [Dentener et al., 2005].
 In addition to the traditional focus of O3 air quality management on local and regional scales, changes in precursor emissions also cause long-term changes in O3srf and radiative climate forcing on a global scale. Although the long-term effects of actions to manage O3srf in a single airshed are expected to be small, the cumulative effects of such actions globally are significant. Because of these long-term air quality and climate effects, therefore, it may be desirable to emphasize CH4 abatement, and to increase the emphasis of O3 air quality control on CO and NMVOC abatement. While the ΔRFnet/ΔO3srf indicators presented here are useful in considering the climate effects of actions to improve air quality, we would ideally like to know the least-cost combination of emission reduction measures that would jointly achieve air quality and climate objectives [West et al., 2004], and future work should combine our results with relevant control costs. In the case of CH4, a large potential for low-cost and cost-saving emission controls has been identified, mainly in industrial sectors [West and Fiore, 2005].
 We thank H. Levy, P. Ginoux, and anonymous reviewers. This work was supported by the National Oceanic and Atmospheric Administration, and a National Aeronautics and Space Administration New Investigator Program grant to D.L.M.