Co-benefits of global and regional greenhouse gas mitigation on U.S. air quality in 2050

Policies to mitigate greenhouse gas (GHG) emissions will not only slow climate change, but can also have ancillary benefits of improved air quality. Here we examine the co-benefits of both global and regional GHG mitigation on U.S. air quality in 2050 at fine resolution, using dynamical downscaling methods, building on a previous global co-benefits study (West et al., 2013). The co-benefits for U.S. air quality are quantified via two mechanisms: through reductions in coemitted air pollutants from the same sources, and by slowing climate change and its influence on air quality, following West 15 et al. (2013). Additionally, we separate the total co-benefits into contributions from domestic GHG mitigation versus mitigation in foreign countries. We use the WRF model to dynamically downscale future global climate to the regional scale, the SMOKE program to directly process global anthropogenic emissions into the regional domain, and we provide dynamical boundary conditions from global simulations to the regional CMAQ model. The total co-benefits of global GHG mitigation from the RCP4.5 scenario compared with its reference are estimated to be higher in the eastern U.S. (ranging from 20 0.6-1.0 μg m-3) than the west (0-0.4 μg m-3) for PM2.5, with an average of 0.47 μg m-3 over U.S.; for O3, the total co-benefits are more uniform at 2-5 ppb with U.S. average of 3.55 ppb. Comparing the two mechanisms of co-benefits, we find that reductions of co-emitted air pollutants have a much greater influence on both PM2.5 (96% of the total co-benefits) and O3 (89% of the total) than the second co-benefits mechanism via slowing climate change, consistent with West et al. (2013). GHG mitigation from foreign countries contributes more to the U.S. O3 reduction (76% of the total) than that from domestic 25 GHG mitigation only (24%), highlighting the importance of global methane reductions and the intercontinental transport of air pollutants. For PM2.5, the benefits of domestic GHG control are greater (74% of total). Since foreign contributions to the co-benefits are comparable to that from the domestic reductions, especially for O3, previous studies that focus on local or regional co-benefits may greatly underestimate the total co-benefits of global GHG reductions. We conclude that the U.S. can gain significantly greater domestic air quality co-benefits by engaging with other nations to control GHGs. 30 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-1054, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 21 January 2016 c © Author(s) 2016. CC-BY 3.0 License.


Introduction
Climate change and air quality are interrelated problems. First, climate change can affect the formation, destruction and transport of major air pollutants, through changes in meteorological variables of temperature, precipitation, air stagnation events, etc. (Weaver et al., 2009;Jacob and Winner, 2009;Fiore et al., 2012Fiore et al., , 2015. It can also affect natural emissions (biogenic, dust, fire and lighting) that influence air quality. Second, air pollutants such as particulate matter (PM) and ozone 5 (O3) can change the climate by altering the solar and terrestrial radiation balance through direct and indirect effects (Myhre et al., 2013). Third, the sources of emissions of greenhouse gases (GHGs) and air pollutants are usually shared, particularly through the combustion of fossil fuels, so actions to control one can also influence emissions of the other. Policies to control GHG emissions will therefore not only slow climate change in the future, but will also provide co-benefits of improvements to air quality and consequently to human health (Bell et al., 2008;Nemet et al., 2010). 10 the two RCP scenarios (RCP8.5 and RCP4.5) in 2050 used by WEST2013 is downscaled using a one-way nesting configuration for five consecutive years. WRF is initialized at 0000 Coordinated Universal Time (UTC) 1 January 2048 and run for a 12-month spin-up, then run continuously through 0000 UTC 1 January 2053. A historical period from GFDL AM3 is also downscaled with WRF initialized at 0000UTC 1 January 1999 and run for a 12-month spin-up, then run continuously through 0000 UTC 1 January 2004. The WRF physics options include the Rapid Radiative Transfer Model for global climate 5 models (Iacono et al., 2008) for longwave and shortwave radiation, WRF single-moment 6-class microphysics scheme (Hong and Lim, 2006), the Grell ensemble convective parameterization scheme (Grell and Devenyi, 2002), the Yonsei University planetary boundary layer scheme , and the Noah land surface model (Chen and Dudhia, 2001).
The WRF configuration also applies spectral nudging. Otte et al. (2012) and Bowden et al. (2012Bowden et al. ( , 2013 demonstrated that using nudging in WRF improves the overall accuracy of the simulated climate over the CONUS at 36-km and does not 10 squelch extremes in temperature and precipitation. In particular, spectral nudging affects the model solution through a nonphysical term in the prognostic equations based on the difference between the spectral decomposition of the model solution and the reference analysis. Spectral nudging is used to constrain WRF toward synoptic-scale wavelengths resolved by GFDL AM3 exceeding 1200 km. Nudging is applied equally to potential temperature, wind, and geopotential with a nudging coefficient of 1.0×10 -4 , which is equivalent to a time scale of 2.8 hours. The downscaled meteorology from WRF is 15 used to provide meteorological inputs to CMAQ. Hourly WRF outputs are processed using Meteorology-Chemistry Interface Processor (MCIP v4.1; Otte and Pleim, 2010) to provide meteorological inputs for CMAQ.
Comparing the downscaling results between WRF with the GFDL AM3 simulation for three-year averages of the 2-m temperature (we present three-year averages instead of four to be consistent with CMAQ outputs below), we see that the large-scale spatial patterns for temperature are similar (Fig. S1). However, the downscaling clearly improves the resolved 20 features related to topography and provides a different realization of average regional climate throughout the CONUS.
Comparing WRF future projected change centered on 2050 with 2000, we see that the three-year average of 2-m temperature generally increases over the entire U.S. for both RCP8.5 and RCP4.5 ( Fig. S2-S3). Temperature increases are largest for extreme northeastern latitudes, the Southeast and Southwest U.S. in both scenarios, with U.S. average warming of 3.05°C and 2.59°C for RCP8.5 and RCP4.5, respectively. Additionally, precipitation is projected to increase over most of the U.S. in 25 both scenarios with U.S. average increases of 8.16 and 7.63 mm day -1 in RCP8.5 and RCP4.5. Comparing the changes between scenarios (RCP8.5 minus RCP4.5), Fig. 1 illustrates that temperature increases are smaller in RCP4.5 throughout the CONUS, except in the Northwest. The precipitation difference between scenarios has a larger spatial variability than the 2-m temperature. However, the only region where the regional climate is warmer and drier in RCP4.5 is in the Northwest U.S. Ignoring other influences of climate change, increases in precipitation would be expected to increase PM wet 30 scavenging, and decrease PM concentration.

Regional emissions
Similar studies in the past have typically chosen to run SMOKE with the present-day U.S. National Emission Inventory (NEI), and then scale the SMOKE outputs into future years, using the mass ratio of projected future to present-day emissions from global inventories (e.g., Hogrefe et al., 2004;Nolte et al., 2008;Avise et al., 2009;Chen et al., 2009;Gao et al., 2013).
Instead, we use SMOKE to directly process the global emissions in 2000 and in 2050 from REF and RCP4.5 to provide 5 temporally-and spatially-resolved CMAQ emission input files. We first regrid the global emissions datasets at 0.5º×0.5º into finer resolution (36km×36km), and then take advantage of the temporal and speciation profiles inside SMOKE to assign temporal variations and re-speciate the PM and VOCs species. By doing this, we account better for the spatial distribution changes of future emissions projected in the RCPs (Figs. S4-S10), whereas the traditional method only considers changes in the magnitude of air pollutants in the future, assuming a constant spatial and sectoral distribution. 10 In addition, the RCP datasets report only elemental carbon (EC) and organic carbon (OC), but ignore emissions of other primary PM species. Here we back-calculate the total PM2.5 and PM coarse (PMC) primary emissions for all sectors from the reported EC and OC. We first derive the emission fractions of EC and OC in each sector by cross-comparing the definitions of the sectors in IPCC, the Source Clarification Codes (SCC) in the speciation cross-reference file  (Table S1). If multiple sources are included in one IPCC sector (e.g., energy and industries in Table S1), we use the mass ratio from the source that contributes the largest fraction by referring to previous studies (Reff et al., 2009;Xing et al., 2013).
Then we calculate the total PM2.5 and PMC in each grid cell by dividing the reported EC and OC by their emission fractions 20 individually, and average these two. By doing this, we increase the total PM2.5 emissions of the RCPs by incorporating the inorganic components of primary PM, such as sulfate and nitrate. We check these results by comparing the total 2000 PM2.5 emissions of 4.14 Tg yr -1 in this study (Table 2) with other studies, finding that it is comparable to the total of 4.69 Tg yr -1 in 2001 from the U.S. NEI (http://www.epa.gov/ttnchie1/trends/, accessed 5 October 2013). Our calculated PM2.5 emission is also lower than the estimated 5.53 Tg yr -1 in 2000 by Xing et al. (2013), which used an activity data based approach to 25 develop consistent temporally-resolved emissions from 1999 to 2010.
In and O3 precursors also decrease, including EC (7.59%) and OC (6.17%), with NOx and NMVOC decreasing by more than 30 10%. SO2 has the largest relative decreases between RCP4.5 and REF in 2050 (28.78%). Large spatial variations in emissions reductions are also seen over the U.S., with the largest reductions seen on the east and west urban areas of U.S. for most air pollutants and smaller reductions in the Great Plains (Figs. S4-S10).
Biogenic emissions are estimated using the Biogenic Emission Inventory System (BEIS v3.14), which responds to the changing climate for different scenarios. It is configured to run on-line in CMAQ, and calculates the emissions of 35 chemical species including 14 monoterpenes and 1 sesquiterpene. We assume that land use and land cover will stay constant in the future for the purpose of estimating biogenic emissions. The on-line option of lightning is also turned on to calculate the NOx emissions by estimating the number of lightning flashes based on the modeled convective precipitation, which also 5 changes with climate. We prepare the ocean/land mask for the domain to calculate sea salt emissions which can be significant in coastal environments (Kelly et al., 2010). We also use the BEIS on-line calculation for natural soil NOx emissions.

Regional air quality model and dynamical chemical BCs
The latest CMAQ model (https://www.cmascenter.org/cmaq/index.cfm, accessed 15 June 2012) is used to perform the 10 regional air quality simulations with the CB05 chemical mechanism and updated toluene reactions. The model incorporates the newest aerosol module (AE6), including features of new PM speciation (Reff et al., 2009), oxidative aging of primary organic carbon (Simon and Bhave, 2012), and an updated treatment and tracking of crustal species (e.g., Ca 2 +, K + , Mg 2+ ) and trace metals (e.g., Fe, Mn) (Fountoukis and Nenes, 2007). Several other enhancements in v5.0 of CMAQ were discussed by Appel et al. (2013) and Nolte et al. (2015), and there are no significant changes for the aerosol module between v5. The dynamical BCs for this study are provided by the global MZ4 simulations of WEST2013. The hourly boundary values from MZ4 are horizontally interpolated from coarser resolution to the regional finer resolution, and also vertically interpolated as MZ4 and CMAQ have different vertical layers. Chemical species are mapped between MZ4 and CMAQ 25 v5.0.1, due to the different chemical mechanisms used by these two models, following the descriptions of Emmons et al.
(2010) and ENVIRON (http://www.camx.com/download/support-software.aspx, accessed 19 September 2013). For the chemical species in CMAQ that do not exist in MZ4, values are set to defaults as suggested by the CMAQ website.

Scenarios
We simulate scenarios in CMAQ comparable to WEST2013, except that we carry out one extra scenario to quantify the co-30 benefits from domestic versus foreign GHG mitigation (Table 1). S_2000 is conducted to evaluate CMAQ model performance and to compare with future scenarios. For this study, we run four scenarios in 2050. The differences between Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-1054 Manuscript under review for journal Atmos. Chem. Phys. Published: 21 January 2016 c Author(s) 2016. CC-BY 3.0 License. S_RCP45 and S_REF are the total co-benefits on U.S. air quality from global GHG mitigation. The emission benefit from the first mechanism is calculated as the difference between S_Emis and S_REF, and the meteorology benefit is calculated as S_RCP45 minus S_Emis. By comparing S_Dom (applying GHG mitigation from RCP4.5 scenario in the U.S. only) with S_REF, and S_RCP45 with S_Dom, we quantify the co-benefits from domestic and foreign GHG mitigation. In estimating the co-benefits of domestic reductions, we account for the influences of global climate change as a foreign influence (as most 5 GHG emissions are global), assuming that U.S. air pollutant emissions have small effects on global or regional climate, such as through aerosol forcing. In each scenario, we fix global methane at concentrations given by the RCPs (Table 1)

CMAQ model evaluation
The CMAQ model has been broadly used to study regional future air quality (Hogrefe et al., 2004;Tagaris et al., 2007;Nolte et al., 2008;Lam et al., 2011;Gao et al., 2013) and has been evaluated in many applications (Appel et al., 2010(Appel et al., , 2011(Appel et al., , 201315 Nolte et al., 2015). Here we evaluate the CMAQ v5. Model performance is not expected to be perfect as meteorology does not correspond with actual year 2000 meteorology, and emissions are derived from global datasets rather than specific emissions for the U.S. 30 For total PM2.5, overall model performance is good and the NMdnE for IMPROVE and CSN are less than 50%, with slight differences in performance (Table 3). CMAQ underestimates PM2.5 in these two networks and also its components in all Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-1054 Manuscript under review for journal Atmos. Chem. Phys. Published: 21 January 2016 c Author(s) 2016. CC-BY 3.0 License. three networks (Table S2), except that it overestimates SO4 2compared with IMPROVE, and NH3 + with CSN. Compared with other components, OC and EC are not well predicted, with higher NMdnB, -63.55% and -37.00% in IMPROVE (OC and EC are not measured in the other two networks). In simulating PM2.5 and its species, model performance is better in winter than in summer (not shown here). The model overestimates surface O3 as indicated by the positive MdnB (ppb) and NMdnB (%). The NMdnE for the 1hr-O3 (MDA8-O3) declines from 27.60% (33.35%) to 17.36% (16.95%) after we apply 5 the cutoff value of 40 ppb. The overprediction is slightly higher for 1hr-O3 than for MDA8-O3, however this difference becomes smaller when we consider the cutoff values.

Air quality changes in 2050
Here we show the seasonal and spatial patterns of future air quality changes centered in 2050 relative to 2000 from REF and  (Table 2), as also reported in other studies (Gao et al., 2013;Fiore et al., 2015). The magnitude of the decreases between S_REF and S_2000 is lower than that between S_RCP45 and S_2000, as the REF scenario did not apply a GHG mitigation policy, and thus has less emission reductions.
We then compare these air quality changes in 2050 with the MZ4 simulations of WEST2013 for both S_REF ( (3.23 µg m -3 ) is modestly higher than both the results from MZ4 and the ACCMIP ensemble mean, but within the range of ACCMIP models when PM2.5 is calculated as a sum of species. The future O3 changes in our study (5.20 ppb) are clearly in the range of ACCMIP results, and nearly identical to MZ4 (5.13 ppb). Comparisons of the air quality changes in 2050 for 25 S_REF relative to 2000 between CMAQ and MZ4 are similar, except that the magnitudes of the changes are smaller than those for S_RCP45 (Fig. S15).

Total co-benefits for U.S. air quality from global GHG mitigation
Projected three-year average PM2.5 concentrations in 2050 in both scenarios (S_REF and S_RCP45) are higher in the Eastern U.S. and the west coast of CA, and lower in the Western U.S. (Fig.3). The total co-benefits for U.S. air quality (S_RCP45 30 minus S_REF) show notable decreases of major air pollutants in 2050. The total co-benefits for PM2.5 over the U.S. show a significant spatial gradient over the U.S. domain, greatest in the eastern U.S., especially urban areas, as well as CA, ranging from 0.4 to 1.0 µg m -3 , and least in the Rocky Mountains and Northwest with values below 0.4 µg m -3 . The total co-benefits for PM2.5 averaged over the U.S. is 0.47 µg m -3 , with the largest contribution from organic matter (OM, including primary OC, SOA and NCOM), accounting for the 45% of the total (0.21 µg m -3 ), followed by sulfate (0.11 µg m -3 ) and ammonia (0.05 µg m -3 ) (Fig. S16). The total co-benefits are highest in fall, with U.S. domain average of 0.55 µg m -3 , and lowest in spring (0.41 µg m -3 ) (Fig. 4). Notice that the region with greatest co-benefits shifts from Central areas in winter and spring to 5 the East in summer and fall, with the largest component of OM also shifting from primary OC to SOA (Fig. S17).
Future O3 is presented here as the ozone-season average (from May to October) of MDA8. In general, 2050 O3 concentrations in S_REF and S_RCP45 are projected to be high in the Southern U.S., especially over the coastal areas, and higher in the West than the East (Fig. 5). The total co-benefits for O3 are fairly uniformly significant over the entire U.S. domain, slightly higher in the Northeast and Northwest, and range from 2-5 ppb with a domain average of 3.55 ppb, unlike 10 PM2.5 which is higher over urban regions. The uniformity of the total O3 co-benefits suggests that they are strongly influenced by global O3 reductions.
The total co-benefit for PM2.5 from this study (0.47 µg m -3 over U.S.) is lower than WEST2013 (area-weighted three-year averages of 0.72 µg m -3 over U.S.), especially over the Northwest and Central of U.S. (Fig. S18). Analyzing the components of PM2.5, we find that this difference is mainly caused by OM, with a U.S. annual average of 0.40 µg m -3 in WEST2013 and 15 0.21 µg m -3 in this study (Fig. S19). For other components (EC, SO4 2-, NO3as reported in MZ4 of WEST2013), the CMAQ results are slightly lower than WEST2013 but share a similar spatial pattern (Figs. S20-S22). We expect that the total cobenefits of PM2.5 in this study might be higher than WEST2013, as we account for inorganic primary PM emissions in SMOKE. A possible explanation may be that different chemical mechanisms and deposition processes are adopted for organic aerosols in MZ4 and CMAQ, which may make a shorter atmospheric lifetime for PM in CMAQ than that in MZ4. 20 The differences of the meteorology (e.g., the precipitation and temperature) between the downscaled WRF and the GFDL could also contribute to this difference. Total co-benefit of O3 from this study (3.55 ppb over U.S.) is comparable to WEST2013 (3.71 ppb) in both the magnitude and spatial distribution (Fig. S23).

Co-benefits from the two mechanisms
We quantify the co-benefits of global GHG mitigation on PM2.5 and O3 through the two mechanisms: reduced co-emitted air 25 pollutants (S_Emis-S_REF) and slowing climate change and its effect on air quality (S_RCP45-S_Emis). The reduction of co-emitted air pollutants has a much greater effect than slowing climate change for PM2.5, accounting for 96% of the U.S. average PM2.5 decrease. The emission benefit for PM2.5 over the U.S. domain is 0.45 µg m -3 , greatest near urban areas where emissions are reduced (Fig. 6), with the largest contribution from OM (0.172 µg m -3 over the U.S.), followed by sulfate For O3, the emission benefit is also larger than the climate benefit, accounting for 89% of the total O3 decreases averaged over the U.S. The emission benefit for O3 over the U.S. domain is 3.16 ppb, and much more uniform over the U.S., slightly higher over Northeast and Northwest. Slowing climate change accounts for 0.39 ppb O3 decreases, 11% of the total and mainly in the Great Plains and the East, where temperatures are cooler under RCP4.5 compared with RCP8.5 (Fig. 1). The dominance of the emission co-benefit over the climate co-benefit for both PM2.5 and O3 is consistent with WEST2013. 5

Co-benefits from domestic and foreign GHG mitigation
We also investigate the co-benefits from domestic GHG mitigation by comparing S_Dom with S_REF, versus foreign GHG reductions by comparing S_RCP45 with S_Dom (Fig. 7). For PM2.5, domestic GHG mitigation accounts for 74% (0.35 µg m -3 ) of the total PM2.5 decrease over the whole U.S., with the greatest effect over the East and CA, where emissions of PM2.5 and its precursors are greatly reduced (Figs. S3-S9). The benefits from foreign GHG reductions on the U.S. PM2.5 10 change are only obvious in the Southern U.S., influenced by emission reductions in Mexico and global climate change. We conclude that domestic GHG mitigation has a greater influence on U.S. PM2.5 than reductions in foreign countries, but that foreign reductions also make a noticeable contribution, accounting for 26% of total PM2.5 decreases over the U.S., and a greater fraction in the Southern U.S. For O3, foreign countries' GHG mitigation has a much larger influence on the U.S., accounting for 76% (2.69 ppb) of the 15 total O3 decrease, compared with 24% from domestic GHG mitigation (Fig. 7). The U.S. experiences greater O3 decreases in the North than the South, which is likely influenced in part by the air quality improvement in Western Canada as a result of slowing deforestation due to the climate policy in RCP4.5 (West et al., 2013). This large influence of foreign reductions for O3 highlights the importance of global methane reductions in RCP4.5 and global emission reductions, particularly in Asia and intercontinental transport. 20

Regional co-benefits and variability
We then quantify the co-benefits over nine U.S. climate regions defined by the National Oceanic and Atmospheric Administration (Fig. S24), and their domestic and foreign components. The Central, Southeast, Northeast and South regions have the largest total co-benefits for PM2.5 (regional annual means of 0.78, 0.75, 0.62 and 0.62 µg m -3 ), and the Northwest has the lowest total co-benefits (0.16 µg m -3 ) (Fig. 8). Domestic GHG mitigation has the largest effect over these same 25 regions and lowest effects over Northwest and West North Central, with means of 0.13 µg m -3 . Foreign co-benefits are greatest over the South, Southwest, Central and Southeast, and lowest over Northwest (Table S3). As a fraction of the total co-benefits, the domestic co-benefit is highest in the Northeast, East North Central and Central accounting for more than 80% of the total, while foreign co-benefits are highest over Southwest, South and West North Central, accounting for about 40% of the total. 30 For O3, the Northeast, East North Central, and Northwest have the highest total co-benefits, (regional means of 4.61, 4.25, 4.15 ppb; Fig. 9 and Table S3), although the total co-benefits for O3 are fairly uniform over the U.S (Fig. 5). The Southeast Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-1054 Manuscript under review for journal Atmos. Chem. Phys. Published: 21 January 2016 c Author(s) 2016. CC-BY 3.0 License.
has the lowest total co-benefits, with 2.67 ppb for the regional mean. Domestic co-benefits are higher over the Central, Northeast and Southeast, with regional means of 1.25, 1.16 and 1.14 ppb, and lowest over Northwest (0.4 ppb). In general, foreign mitigation contributes more in the west than the east, most likely influenced by intercontinental transport from Asia.
It is highest in the Northwest, West North Central and Northeast, with regional means of 3.75, 3.45 and 3.45 ppb. The fraction of co-benefits from foreign mitigation is larger than 60% in most regions, highest over the Northwest (90%), and 5 lowest over the Southeast (57%).
We also evaluate the variability in co-benefits for the three years simulated (Table S3). Over the U.S., the coefficient of variation (CV) for the total co-benefits for PM2.5 (7%) is much lower than that of the total co-benefits for O3 (37%), which is controlled by the intercontinental transport and global CH4. The Southeast has the highest CV (29%) for the total co-benefits of PM2.5, while other regions are lower than 15%, lowest in the East North Central and Northeast (3%). Southwest and South 10 have the highest CV (70%, 69%) for the total co-benefits of O3, and lowest in Northwest (21%). For regions with higher variability, longer simulations would be desirable to better quantify the annual average co-benefits.

Discussion
The co-benefits we present here are specific to the reference (REF) and mitigation (RCP4.5) scenarios we choose, and results would differ for other baseline and mitigation scenarios. The estimated co-benefits also depend on participation of many 15 nations in the mitigation policies, and delaying participation will likely change the co-benefits.
The total co-benefits for O3 when downscaled are comparable to the global study in both magnitude and spatial pattern, but the downscaled simulations capture some local features better than the global model, such the effects of topography and urban areas. For PM2.5, significant differences are seen from the downscaling due to the fine resolution and different chemical mechanisms between the global and the regional model. The resolution we are using for this study (36km by 36 20 km) is fine enough for us to analyze the co-benefits at a state level, but insufficient to fully resolve urban areas. Finer resolution simulations (such as 12 km by 12 km) with CMAQ or other CTMs can be carried out to better quantify the cobenefits over urban areas.
For this study, uncertainties and errors may exist under the assumptions and choices we make for each model. For example, the co-benefits of PM2.5 have large contributions from OC and SOA over the Central and East U.S. (Fig. 4, Fig. S16). 25 However, our model evaluations show that CMAQ greatly underestimates the OC concentration compared with surface observations. New gas-phase and aqueous-phase oxidation pathways for SOA formation are found to play significant roles in producing organic aerosols (Lin et al., 2014;Pye and Pouliot, 2012;, which are missing in the CMAQ version used in this study. We use BEIS model to estimate the biogenic VOC (BVOC) emissions, but studies have shown that the BVOCs from the Model of Emissions of Gases and Aerosols from Nature (MEGAN) are higher than those from 30 BEIS by a factor 2 (Pouliot, 2008;Pouliot and Pierce, 2009), which highlights the uncertainty in representing these emissions and simulating both PM2.5 and O3 (Hogrefe et al., 2011). Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-1054 Manuscript under review for journal Atmos. Chem. Phys. Published: 21 January 2016 c Author(s) 2016. CC-BY 3.0 License.
We assume constant land use in the GCM, WRF and CMAQ when simulating the global and regional climate and estimating the biogenic emissions, which could introduce errors in our results (Unger, 2014;Heald and Spracklen, 2015). When we process the global anthropogenic emissions with SMOKE, we back-calculate the total PM2.5 and PMC from OC and BC, which introduces inorganic PM emissions and may make our results for co-benefits of PM2.5 higher. By doing this, we account for missing emissions but also increase the total uncertainties in the emission inventory. Spectral nudging is adopted 5 in this study to restrain WRF from drifting from the GCM, which has been shown to be better for some meteorological variables, but spectral nudging better for others (Bowden et al., , 2013Liu et al., 2012;Otte et al., 2012). Moreover, only one model is used at each step during downscaling, and ensemble model means can be used to reduce the single model's variability. Simulations are based on three-year averages, due to computational limitations, but these three years may reflect meteorological variability and not only climate change. This uncertainty may be greater for the total co-benefits 10 of O3, for which we see greater year-to-year variations than for PM2.5. CMAQ simulations could be performed over more years to reduce the influence of the climate variability. In separating domestic and foreign co-benefits, we assume that global and regional climate will be controlled by foreign GHGs emissions, and not influenced by GHG mitigation in the U.S., which may also introduce errors into our results. We similarly attribute the global methane change as a foreign influence, as U.S. methane emissions are a small fraction of the global. 15

Conclusions
Climate polices to control GHG emissions will not only have the benefit of slowing climate change, but can also have cobenefits of improved air quality. Previous co-benefits studies focus mostly on local or regional GHG reductions. As a result, these studies omit air quality benefits outside of the domain considered, and neglect benefits from global GHG mitigation. In this study we adopt a systematic approach to quantify the co-benefits from both the global and regional GHG mitigation on 20 regional air quality over U.S. at fine resolution in 2050, building on the global co-benefits study from West et al. (2013). The co-benefits of global GHG mitigation on U.S. air quality are discussed through two mechanisms: reduced co-emitted air pollutants and slowing climate change and its influence on air quality. We also quantify the co-benefits from domestic GHG mitigation versus foreign countries' reduction.
We find that there are significant benefits for both PM2.5 and O3 over U.S. by 2050 from the global GHG mitigation in 25 RCP4.5. The total co-benefits for PM2.5 are higher in the east than the west, with an average of 0.47 µg m -3 over U.S. For O3, the total co-benefits are fairly uniform across the U.S. at 2-5 ppb, with U.S. average of 3.55 ppb. The co-benefits from reductions of co-emitted air pollutants have a greater influence on both PM2.5 (accounting for 96% of total decreases) and O3 (89% of the total decreases) than the second mechanism via slowing climate change, consistent with West et al. (2013).
Foreign countries' GHG reductions have a much greater influence on the U.S. O3 reduction (76% of the total), compared 30 with that from domestic GHG mitigation only (24%), highlighting the importance of global methane reductions and the intercontinental transport of air pollutants. For PM2.5, the benefits of foreign GHG control are less than domestic, but still a Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-1054 Manuscript under review for journal Atmos. Chem. Phys. Published: 21 January 2016 c Author(s) 2016. CC-BY 3.0 License.
considerable portion of the total (26%). We conclude that the U.S. can gain significantly greater domestic air quality cobenefits by engaging with other nations for GHG control to combat climate change, especially for O3. This also applies to other nations which can be expected to have ancillary air quality benefits from foreign countries' GHG mitigation. We also conclude that previous studies that estimate co-benefits for one nation or region (e.g., Thomson et al., 2014), may significantly underestimate the full co-benefits when many countries reduce GHGs together, particularly for O3. 5 funding sources do not endorse the purchase of any commercial products or services mentioned in the publication.