Co-benefits of global, domestic, and sectoral greenhouse gas mitigation for US air quality and human health in 2050

Exposure to fine particulate matter (PM_2.5) and ozone (O₃) is associated with both morbidity (e.g. hospitalizations, emergency department visits, school absences, and asthma-related health effects) and premature human mortality (e.g. deaths from cardiovascular and respiratory disease and lung cancer), as revealed in epidemiological studies (US EPA 2009, 2013). Several cohort studies have shown evidence for chronic effects of PM_2.5 on mortality (Laden et al 2006, Krewski et al 2009, Lepeule et al 2012), whereas fewer have demonstrated the chronic effects of O₃ on mortality (Jerrett et al 2009).

Previous research has quantified future air quality changes and their effects on human health under projected emission scenarios, at both the global (West et al 2007, Selin et al 2009, Silva et al 2016a) and regional scales (Fann et al 2013, Kim et al 2015, Jiang et al 2015, Sun et al 2015). Climate change can also affect air quality through several mechanisms, including photochemical reactions, natural emissions, deposition rates, and air stagnation events (Weaver et al 2009, Jacob and Winner 2009, Fiore et al 2012, 2015). Related studies have quantified the effect of global and regional climate change on air quality and human health (Bell et al 2007, Tagaris et al 2009, Post et al 2012, Fang et al 2013, Fann et al 2015). Post et al (2012) used an ensemble of atmospheric models to study the effect of climate change in 2050 on air quality and human health in the US, and found significant variability when using different models.

Many studies have also investigated the co-benefits of greenhouse gas (GHG) mitigation for air quality and avoided premature mortality, as actions to reduce GHG emissions also tend to reduce co-emitted air pollutants (Bell et al 2008, Cifuentes et al 2001, Nemet et al 2010). When monetized, the health co-benefits of GHG mitigation were found to range across the literature from $2–$196/tCO₂ (Nemet et al 2010), comparable to the costs of GHG reductions. Other recent studies have also analyzed the effects of GHG mitigation on future air quality and human health co-benefits in the US (Driscoll et al 2015, Markandya et al 2009, Thompson et al 2014, Trail et al 2015, Plachinski et al 2014). Thompson et al (2014) studied the co-benefits of different climate policies in the US on domestic air quality in 2030, finding that human health benefits due to improved air quality can offset 26%–1050% of the cost of carbon polices. Other studies also investigate the co-benefits of climate policy on food security, energy savings, and other health co-benefits of active transportation (walking, biking) and changes in diet (Capps et al 2016, Chuwah et al 2015, Friel et al 2009, Jakob 2006, McCollum et al 2013, Wilkinson et al 2009, Woodcock et al 2009), but they are not the focus of our study.

Previous co-benefits studies have been limited by only considering the co-benefits of regional or local climate policies on regional air quality and human health, neglecting (i) the co-benefits of those actions for other nations or regions, and (ii) the co-benefits gained domestically from global actions where one country's actions are coordinated with reductions internationally. Both PM_2.5 and O₃ have long enough lifetimes in the atmosphere to transport intercontinentally, suggesting that emissions from one source region can affect air quality and human health on multiple receptor regions (Anenberg et al 2009, 2014, Liu et al 2009). For O₃, the health benefits of O₃ precursor reductions may even be greater outside of the source region than within due to the greater population over several receptor regions (Duncan et al 2008, Anenberg et al 2009, West et al 2009). PM_2.5 has a much shorter lifetime than O₃, but the mortality impacts of intercontinental transport of PM_2.5 are comparable to that of ozone due to the stronger effects of PM_2.5 on mortality (Anenberg et al 2014). To address these limitations, West et al (2013), (referred to as WEST2013 hereafter) were the first to use a global chemical transport model (CTM) to address the co-benefits of global GHG mitigation on air quality and human health. WEST2013 were also the first to estimate co-benefits via two mechanisms: reduced co-emitted air pollutants, and slowing climate change and its effects on air quality. They found that global GHG mitigation could avoid 2.2 ± 0.8 million premature deaths in 2100 due to the improved air quality, accounting for both PM_2.5 and O₃ mortality. The co-benefits from the first mechanism of reduced co-emitted air pollutants are much greater than those from the second mechanism of slowing climate change and its effect on air quality. The monetized co-benefits for health were estimated at $50–$380/tCO₂, globally averaged, higher than previous estimates (Nemet et al 2010).

WEST2013 applied a global CTM (horizontally 2˚ × 2.5˚) to study the co-benefits. We increased the horizontal resolution using a limited area model framework to further investigate the co-benefits for US air quality in 2050 at much finer resolution (Zhang et al 2016). Here we use the simulations performed by Zhang et al (2016) and focus on quantifying the co-benefits of global GHG reductions for avoided air pollution-related mortality in the continental US in 2050. We study the total co-benefits through the two mechanisms, following WEST2013 and Zhang et al (2016), and separate the co-benefits of GHG mitigation in the US versus the contributions from foreign countries. By embedding this study within the previous global study of WEST2013, we are the first to investigate the co-benefits of foreign GHG mitigation for US air quality and human health. Previous studies have also investigated the effects of air pollution from specific emission sectors on premature mortality, both globally (Lelieveld et al 2015, Morita et al 2014, Yim et al 2015, Silva et al 2016b) and regionally (Caiazzo et al 2013, Fann et al 2012, 2013). Here we conduct three new sensitivity simulations to quantify the air quality and health co-benefits of GHG reductions in three US emission sectors: industry, residential and energy.

2.1. Air quality changes in the US in 2050 at fine scale

Air quality changes in the US under different GHG scenarios centered on 2050 were downscaled from WEST2013 by Zhang et al (2016). Meteorological fields from the NOAA Geophysical Fluid Dynamics Laboratory (GFDL) atmospheric model AM3 (Donner et al 2011, Naik et al 2013), used by WEST2013, was first downscaled to the regional scale over the continental US domain to a 36 km horizontal resolution using the Weather Research Forecast model (WRF, v3.4.1, Skamarock and Klemp 2008). The WRF configuration applies spectral nudging to maintain the large-scale atmospheric circulation resolved by global model (Otte et al 2012, Bowden et al 2012, 2013). Further information on the WRF configuration can be found in Zhang et al (2016). GFDL-AM3 was downscaled using WRF for two periods, a historical period (2000–2003; here considered 2000), and a future period (2049–2052; here considered 2050) for the Representative Concentration Pathway 4.5 (RCP4.5) and RCP8.5 scenarios with one year of spin-up. Global anthropogenic emissions from RCP4.5 and its reference scenario (REF) were directly processed to the regional scale using the Sparse Matrix Operator Kernel Emissions (SMOKE, v3.5, Houyoux et al 2000) program. Dynamical chemical boundary conditions were acquired from the global CTM outputs of WEST2013. The Community Multiscale Air Quality model (CMAQ, v5.0.1, Byun and Schere 2006), with the CB05 chemical mechanism with updated toluene reactions and the latest aerosol module (AE6), was used to simulate air pollutant concentrations (i.e. PM_2.5 and O₃) in 2000 and 2050. Most of the CMAQ simulations used in this study (table 1) were completed by Zhang et al (2016), but three new sensitivity simulations are performed here to quantify the co-benefits of GHG mitigation from domestic emission sectors in the US. The CMAQ simulations from Zhang et al (2016) and the three additional sensitivity simulations are run for 40 consecutive months, with the first four months as spin-up, and the results are presented as three-year averages.

The total co-benefits from global GHG mitigation are obtained by comparing scenarios S_RCP45 and S_REF (table 1). As discussed by WEST2013 and Zhang et al (2016), RCP4.5 was developed based upon REF, which is a self-consistent representation of future energy and land use development, with regionally specific air pollutants emissions, developed consistently with the assumed future development to 2100 but without considering climate policy (Smith et al 2011). Relative to REF, RCP4.5 is created by applying a global carbon policy spanning all world regions and emission sectors (Thomson et al 2011); the only difference between these two scenarios is therefore the carbon policy. These self-consistent scenarios therefore uniquely isolate the effects of GHG mitigation (RCP8.5 is used as a proxy for REF meteorology, since no climate model simulated REF). The total co-benefits from global GHG mitigation are obtained by comparing scenarios S_RCP45 and S_REF (table 1). As discussed by Zhang et al (2016), the sensitivity run S_Emis applies emissions from RCP4.5 and meteorology from RCP8.5. To separate the total co-benefits from the two mechanisms, we use S_Emis minus S_REF to give the co-benefits from co-emitted air pollutant reductions, and S_RCP45 minus S_Emis for the co-benefits from slowing climate change. The sensitivity simulation S_Dom applies GHG mitigation from the RCP4.5 scenario in the US only, so the co-benefits of domestic GHG mitigation are estimated as S_Dom minus S_REF, and foreign co-benefits as S_RCP45 minus S_Dom.

In addition, we simulate three more scenarios to identify the co-benefits from actions to reduce GHG emissions in individual sectors domestically. We choose to simulate reductions in the industry (S_indUS, manufacturing industries, industrial process emissions other than solvents, construction, mining, and agricultural machinery), residential and commercial buildings (S_resUS, primarly from cooking, heating and hot water), and energy sectors (S_eneUS, from electric power generation and energy extraction and transformation), because air pollutant emission reductions in RCP4.5 in 2050 are greatest from these sectors in the US. Although ground transportation is the largest contributor for most air pollutants in the US in 2000 and 2050, we did not select transportation as little air pollutants reductions are seen from this sector in 2050. The air pollutant emission reductions from the three sectors selected here account for more than 98% of the total SO₂ and NO_x reductions in RCP4.5 relative to REF in the US in 2050, 80% of the CO reductions, and more than 50% of the EC and OC reductions. However, these three sectors only account for 11% of the total non-methane volatile organic compound (NMVOC) decreases (supplementary table S1 available at stacks.iop.org/ERL/12/114033/mmedia).

Zoom In Zoom Out Reset image size

2.2. Human health analysis

We use the environmental Benefits Mapping and Analysis Program–Community Edition (BenMAP-CE, v1.08) (US EPA 2014) to calculate the avoided human mortality associated with future surface air quality changes for both PM_2.5 and O₃. BenMAP-CE calculates the relationship between air pollution and certain health effects, using a health impact function (HIF) from epidemiological studies. The HIFs for PM_2.5 and O₃ used in this study are based on a log-linear relationship between relative risk (RR) and air pollutant concentrations defined by epidemiology studies (Jerrett et al 2009, Krewski et al 2009), which are also used by WEST2013. RR is used to calculate attributable fraction (AF), the fraction of the disease burden attributable to the risk factor, which is defined as:

$$\begin{equation} {\rm{AF}} = \frac{{{\rm{RR}} - 1}}{{{\rm{RR}}}} = 1 - {\exp ^{ - \beta \Delta x}} \end{equation} \tag{ 1 }$$

where β is the concentration–response factor (CRF; i.e. the estimated slope of the log-linear relation between concentration and mortality) and Δx is the change in air pollutant concentration between two scenarios. AF is multiplied by the baseline mortality rate (y₀), and the exposed population (Pop) to yield an estimate of excess deaths attributable to changes in air pollution (ΔMort):

$$\begin{equation} {\rm{\Delta Mort}} = {y_0} \times (1 - {\exp ^{ - \beta \Delta x}}) \times {\rm{Pop}}. \end{equation} \tag{ 2 }$$

We present results for all-cause mortality from the PM_2.5 changes, rather than cardiopulmonary disease (CPD) and lung cancer (LC), as all-cause mortality is the most comprehensive estimate of PM-related mortality appropriate for the US. However, we also estimate the PM-related mortality from CPD and LC to compare with the results of WEST2013. We also quantify the premature mortality from respiratory disease (RESP) associated with O₃ changes. The 90% confidence intervals (CI) presented in this study are calculated using a full Monte Carlo analysis inside BenMAP-CE considering only uncertainty in the HIF.

BenMAP-CE uses county-level baseline mortality rates for the present day and projected to 2050 at five-year intervals, including RESP for O₃, and all-cause, CPD, and LC for PM_2.5 (RTI International 2015). Overall, the projected baseline mortality rates within BenMAP-CE decrease from 2005–2050. However, the baseline mortality rates used by WEST2013 are projected to increase in 2050 in the US, derived from the International Futures (IFs, version 6.54, Hughes et al 2011) under the UNEPGEO Base Case scenario. For population, BenMAP-CE includes the future population projection at county level in the US until 2040 only (totalling 403 million, Woods and Poole 2012), but our study is focused on 2050 (the RCP4.5 projected total population is 384 million in 2050, Clarke et al 2007). To be consistent with WEST2013, we run BenMAP-CE with baseline mortality rates in 2005 and the population projection in 2040 (aged 30 and above), and then post-process the BenMAP-CE outputs by multiplying adjustment ratios to match the US population and US average baseline rates of WEST2013 (supplementary table S2). By doing so, we assume that future baseline mortality rates increase at a uniform national ratio in each county without age, gender or ethnic variations, and that the spatial distribution of population in 2050 of RCP4.5 is the same as that in 2040 projected by Woods and Poole (2012).

The total US PM_2.5 concentration co-benefits in 2050 from global GHG mitigation (−0.47 µg m⁻³ for three-year US annual average) are greatest in the East and California (CA), and less in the West (figure 1(a)). For O₃, we calculate the three-year average of the 6 month ozone-season average of 1 hr daily maximum O₃, to be consistent with Jerrett et al (2009), and the total US O₃ co-benefits in 2050 from global GHG mitigation (−2.96 ppbv for three-year US ozone-season average) are fairly uniform over the US domain (figure 1(b)), slightly higher over the Western US than the East. The population-weighted average (for the 2050 exposed population age 30 and older) for the PM_2.5 co-benefit (−0.84 µg m⁻³ for US average) is almost twice the simple average (table 2), as PM_2.5 has a short lifetime and is therefore distributed locally to regionally (Punger and West 2013). Population weighting has less of an impact on the O₃ estimates as the longer lifetime of O₃ produces a more uniform spatial distribution.

Zoom In Zoom Out Reset image size

For the human health benefits from the global GHG mitigation, our results show that 16 000 (90% CI: 11 700–20 300) premature deaths will be avoided annually in the US in 2050 due to PM_2.5 decreases (table 3). The states with the most avoided deaths are CA (2500 deaths, CI: 1800–3200), New York (NY, 1300 deaths, CI: 1000–1700) and Texas (TX, 1200 deaths, CI: 800–1500) (supplementary figure S1 and table S4), with each state having large population and large PM_2.5 decreases (figure 1, supplementary table S4). For O₃, the total avoided deaths in the US are 8000 (CI: 3600–12 400), 50% fewer than PM_2.5, and also highest in CA (1400, CI: 600–2200), NY (500, CI: 200–800) and TX (500, CI: 200–700). The spatial patterns of both PM_2.5 and O₃ related avoided premature mortality are shown in figure 2. We further quantify the human health co-benefits from global GHG mitigation by calculating the avoided mortality per capita (MPC, the avoided deaths per million people age 30 and older) in 2050, for both PM_2.5 and O₃ (supplementary figure S2, table S4). The MPC for PM_2.5 is much higher in the East than in the West (except for CA), with much greater variation than for O₃, consistent with the finding that the total concentration co-benefits vary locally to regionally for PM_2.5, and are more spatially uniform for O₃ (figure 1). Relative to the present, air quality improves and premature mortality decreases in the future under REF, due to the large projected emission reductions of conventional air pollutants (Silva et al 2016a, Smith et al 2011, West et al 2013, Zhang et al 2016).

Zoom In Zoom Out Reset image size

We then compare the health results in this study with WEST2013 for the avoided deaths from 2000 to 2050 under the REF (S_REF-S_2000) and RCP4.5 (S_RCP45-S_2000) scenarios, and the total co-benefits in 2050 (S_RCP45-S_REF). Zhang et al (2016) concluded that future PM_2.5 changes are greater using the regional CMAQ model simulations than those in WEST2013 for both REF (S_REF-S_2000) and RCP4.5 (S_RCP45-S_2000) scenarios, while the future O₃ changes in 2050 were comparable between CMAQ and WEST2013. When quantifying human health impacts, figure 3 shows that the avoided premature mortality for PM_2.5 for both REF and RCP4.5 relative to S_2000 are higher in this study than WEST2013, especially for CPD, which is consistent with the greater reductions in PM_2.5 predicted here. The avoided premature mortality for O₃ for both REF and RCP4.5 relative to S_2000 are comparable between this study and WEST2013. The total co-benefits for the population-weighted air quality changes are higher for WEST2013 (4.56 ppbv for O₃ and 1.30 µg m⁻³ for PM_2.5, figure S26 and S29 in WEST2013) than our estimations using the regional model (3.02 ppbv for O₃ and 0.84 µg m⁻³ for PM_2.5, table 3), but the estimated total co-benefits for avoided mortality are similar (figure 3 in this paper). The fact that the total co-benefits for avoided deaths are comparable between this study and WEST2013, even though air quality changes are different, may be in part due to the use of county-level baseline mortality rates here vs the national average of WEST2013. Note that the total avoided deaths from the sum of CPD (24 300 deaths yr⁻¹) and LC (3200 deaths yr⁻¹) is larger than the co-benefits calculated for all-cause mortality, as the RRs for CPD (1.13, 95%CI:1.1–1.16) and LC (1.14, 95%CI:1.06–1.23) are greater than that for all-cause mortality (1.06, 95%CI:1.04–1.08) (Krewski et al 2009).

We then separate the total co-benefits into the two mechanisms. The co-benefit of reductions in co-emitted air pollutants (the 'emission co-benefit') accounts for 98% of the total co-benefits (three-year population-weighted average of −0.84 µg m⁻³, table 2) for PM_2.5, and 96% of the total (three-year population-weighted average of −3.02 ppb) for O₃, consistent with WEST2013 and Zhang et al (2016). When calculating the co-benefits for human health, the emission co-benefit also dominates the total co-benefits, with 15 800 (CI: 11 500–20 000) avoided deaths for PM_2.5 (98% of the total), and 7600 (CI: 3400–11 700) for O₃ (94% of the total) (table 3, figure 4). The difference between the total co-benefit and the emission co-benefit is accounted for by the effect of slowing climate change and its effects on air quality (the 'climate co-benefit'). Notice that the climate co-benefit is negative in some locations, e.g. the Northern states for PM_2.5, and Southeast for O₃, where slowing climate change can cause concentrations and air pollution-related deaths to decrease as a result of more precipitation and lower temperature (see figure 1 in Zhang et al 2016). For the climate co-benefits, we only simulate three years, which may reflect climate variability in addition to climate change (Deser et al 2012). However, since we estimate that the emission co-benefits are much greater than the climate co-benefits, we conclude that more years of simulations would not affect this conclusion.

Zoom In Zoom Out Reset image size

GHG reductions from foreign countries account for 2400 avoided deaths (CI: 1800–3100) for PM_2.5-related all-cause mortality, and 5000 (CI: 2200–7800) deaths for O₃-related RESP, which are 15% and 62% of the total deaths for PM_2.5 and O₃ (table 3). Foreign GHG mitigation likewise contributes 15% (−0.13 µg m⁻³ for the three-year US population-weighted average) of the total air quality co-benefits for PM_2.5, and 65% (−1.95 ppbv) of the total co-benefits for O₃, emphasizing that PM_2.5 is more influenced by emission reductions in US, while O₃ is more influenced by the global methane reductions and intercontinental air pollutant transport (Zhang et al 2016). Foreign co-benefits for both PM_2.5- and O₃-related mortality are centred in urban areas (figure 5), where population density is high, even though foreign GHG mitigation reduces surface O₃ pretty uniformly in the US (see supplementary figure S3). The contributions from domestic GHG mitigation on population-weighted average PM_2.5 (85% of the total) and O₃ (35%) are higher than those for the simple average (74% for PM_2.5 and 27% for O₃ in table 2), as air quality improvements from domestic GHG mitigation occur in densely-populated areas. CA has the largest human health benefits from foreign GHG mitigation, with 400 deaths (CI: 300–500) avoided from PM_2.5-related all-cause mortality, and 800 deaths (CI: 400–1300) avoided from O₃. We have calculated total, domestic, and foreign mortality co-benefits for each state (see supplementary tables S4–S6). In quantifying the domestic co-benefits, we neglect the effect of US GHG mitigation on global climate change, and assume that global and regional climate will be controlled by foreign GHG emissions, which introduces a small error into our results. We also attribute the global methane concentration change to the effect of foreign GHG reductions, as US emissions are relatively small (6%–10% of global emissions).

Zoom In Zoom Out Reset image size

Among emission sectors, the residential sector has the largest co-benefits for PM_2.5-related human health, avoiding 2800 deaths (CI: 2000–3600), accounting for 21% of the total domestic co-benefits for PM_2.5, followed by industry (2100, CI: 1500–2700) and energy (1700, CI: 1300–2200). Residential also has the largest change in the population-weighted annual average PM_2.5 (−0.15 µg m⁻³), even though its simple annual average change is comparable to that from the industry sector, demonstrating that residential emissions have a greater influence near where people live. GHG mitigation from industry has the largest effect on O₃-related human health, avoiding 500 deaths (200–800) or 17% of the total domestic co-benefits for O₃, followed by energy (300, CI:100–500), and residential (200, CI:100–300). The total air quality co-benefits for O₃ are also highest in industry (population-weighted average of −0.20 ppb and simple average of −0.22 ppbv). These three sectors together account for 50% of the total avoided PM_2.5-related deaths from domestic GHG reductions and 33% of the total avoided O₃-related deaths, even though the sectors account for a larger fraction of emissions of most pollutants, possibly reflecting the smaller NMVOC emissions decreases from these sectors in RCP4.5. These findings of greater avoided deaths for residential GHG reductions suggest that residential sources might be targeted in policy efforts. Future research should attempt to evaluate air quality and health co-benefits for more specific GHG mitigation measures, including for other sources such as transportation, so that these co-benefits can be evaluated alongside the cost of GHG mitigation.

The total co-benefits of avoided premature mortality are monetized using high ($9.81 million) and low ($3.25 million) values of a statistical life (VSLs) for the US in 2050, as estimated by WEST2013 (in 2005 US$) based on projected income growth. Adding avoided mortality from O₃ and PM_2.5, and dividing monetized benefits by US CO₂ reductions in 2050, we estimate monetized co-benefits in 2050 of $137 ($87–$187) per ton CO₂ reduced at a high VSL, and $45 ($29–$62) per ton CO₂ reduced at a low VSL, very similar to the 2050 estimates of WEST2013 for the US. As for WEST2013, these monetized estimates do not account for avoided deaths outside of the US. These benefits at high VSL exceed the full range of GHG marginal abatement cost estimates from 13 energy-economic models (West et al 2013), and at low VSL are greater than the median cost. Of these total co-benefits, foreign GHG reductions are responsible for monetized benefits of $42 ($23–$62) per ton CO₂ at high VSL, and $14 ($8–$21) at low VSL, which is 31% of the total monetized benefits.

We quantify the co-benefits of global GHG mitigation under the RCP4.5 scenario on US air quality and human health in 2050 using dynamical downscaling. We find that 16 000 (11 700–20 300) deaths yr⁻¹ will be avoided for PM_2.5-related all-cause mortality, and 8000 (3600–12 400) deaths yr⁻¹ will be avoided for O₃-related respiratory mortality. When separating the total co-benefits into two mechanisms, the emission co-benefits have a larger impact than the climate co-benefits for both PM_2.5 and O₃, accounting for 98% and 94% of the total avoided deaths. Foreign GHG mitigation contributes 15% of the total PM_2.5-related and 62% of the total O₃-related deaths. Among the three domestic emission sectors with the greatest reductions in air pollutants under RCP4.5, residential has the highest co-benefits for PM_2.5-related mortality, leading to a reduction of 2800 deaths, and industry has the highest co-benefits for O₃, avoiding 500 deaths in the US. Monetized co-benefits of the GHG mitigation, accounting for avoided deaths from reductions in both PM_2.5 and O₃, are $137 ($87–$187) per ton CO₂ at a high VSL and $45 ($29–$62) at a low VSL. Of these co-benefits, 31% come from the influence of foreign GHG reductions. These benefits likely exceed the marginal costs of GHG reductions in 2050.

Significant uncertainties exist in our results. For PM_2.5, we compare the uncertainty for the future concentration change under RCP4.5 of −2.92% ± 2.3% µg m⁻³ (−2.79% ± 22.0% µg m⁻³ for the PM₂₅ estimated as a sum of species) based on the spread of ACCMIP models (Silva et al 2013, Zhang et al 2016), and the uncertainty for the CRF is 0.0058% ± 32.8%. For O₃, the uncertainty for the future concentration change under RCP4.5 is −5.87% ± 48.8% ppbv, and the uncertainty for the CRF is 0.0039% ± 69.2%. Therefore, the uncertainty in the CRF likely contributes more to the overall uncertainty than the uncertainty in modeled concentration changes, although, for ozone, concentration uncertainty is of similar magnitude to the CRF uncertainty. When quantifying the avoided deaths from improved air quality, we only account for adults above 30. Additional uncertainty arises from downscaling from the global to the regional scale chemistry model, including the conversion of chemical mechanisms in the models, particularly from the addition of new inorganic species for primary PM_2.5 (Zhang et al 2016). Different components of PM_2.5 may have different effects on human health, like black carbon particles (Li et al 2016, Zanobetti and Schwartz 2009). However, we consider all of the components of PM_2.5 to have equal toxicity. Only a single modelling system (AM3-WRF-SMOKE-CMAQ) is used in this study, and as pointed out previously (Post et al 2012, Silva et al 2013), results may differ among different models and ensembles of models can better characterize the range of results. Similarly, increasing the number of years simulated by the models used here can reduce uncertainty related to inter-annual variability (Deser et al 2012). Our conclusions are specific to the REF and GHG mitigation (RCP4.5) scenarios we choose, including their simulation of future emission pathways, which depend on economic drivers and air pollution control policies, and would differ for other scenarios. For example, the new shared socio-economic pathways 4 (SSP4) have different climate policy assumptions considering economic, institutional and technological limitations (Rao et al 2017), and different emission reductions for co-emitted air pollutants in 2050 (supplementary table S3). We only account for the co-benefits from air quality changes due to the GHG mitigation, neglecting other impacts of climate change on health, like heat-waves, elevated temperatures, and infectious disease (Smith et al 2014). Despite these uncertainties, both those quantified and unquantified, our major conclusion that global GHG mitigation can have significant co-benefits for air quality and avoided mortality in the US is unlikely to be altered.

Future studies should estimate co-benefits at both the global and regional scales with finer-resolution air quality model simulations. Uncertainties could be reduced by improving emission estimates for multiple species, the chemical and aerosol mechanisms (CB05 and AE6), and using multi-year simulations and ensemble model experiments (Rao et al 2016). Future air pollutant reference-case emission trajectories are also uncertain (e.g. Rao et al 2017), and use of multiple future scenarios would also be valuable. Future studies should also evaluate benefits beyond health, such as for agriculture and energy. Previous studies have shown that using coarse resolution models tends to underestimate mortality near urban areas for PM_2.5 (Punger and West 2013, Li et al 2015). Improving horizontal resolution in future studies can produce more robust estimates of health benefits, and may cause estimates to increase.

Previous studies have estimated co-benefits of GHG mitigation mainly on local, national, or continental scales (Bell et al 2008, Cifuentes et al 2001, Nemet et al 2010). These studies have presumed that most co-benefits are realized on those scales, and that the contributions of foreign GHG mitigation to total co-benefits would be small. Here we show that the US can gain significantly greater co-benefits for air quality and human health, especially for ozone, when coordinating its GHG emission reductions with concurrent reductions in other nations to combat global climate change. Similar results would also be expected for foreign countries, which will likely also benefit from GHG mitigation in other countries. Previous studies, which only estimate co-benefits from regional or local GHG mitigation may significantly underestimate the full co-benefits of coordinated global actions to mitigate climate.

This work was funded by the US Environmental Protection Agency STAR grant #834285, National Institute of Environmental Health Sciences grant #1 R21 ES022600-01, and NASA Health Air Quality Applied Sciences Team #NNX16AQ80G. The contents are solely the responsibility of the grantee and do not necessarily represent the official views of the US EPA or other funding sources. We also thank the three anonymous reviewers for their thoughtful comments and suggestions which make the paper more complete.