The Fifth-Generation NCAR/Penn State Mesoscale meteorological Model (MM5) [22] is used to downscale (i.e., increase the spatial and temporal resolution over the chosen modeling domain) NASA’s Goddard Institute for Space Studies (GISS) Global Climate Model (GCM) [21] outputs for years 2001 and 2050 [14,25]. The simulation followed the Intergovernmental Panel on Climate Change (IPCC) A1B emission scenario [17] for greenhouse gases. The GISS GCM was applied at a horizontal resolution of 4° latitude by 5° longitude to simulate current and future climate at global scale [14] while the MM5 is applied in a nested configuration with 108 km horizontal resolution for the outer domain and 36 km for the inner one [25]. The inner domain covers the continental United States, part of Canada, Mexico and nearby oceans.

The Community Multiscale Air Quality model (CMAQ) [23] with the SAPRC-99 [26] chemical mechanism is used to simulate pollutant concentrations (i.e., O₃ and PM_2.5) for both historic and future periods keeping constant boundary conditions for 2001 and 2050 simulations [24]. A uniform grid of 36 km × 36 km horizontal cells with 9 vertical layers is employed in the simulations. Although the emission inventory is kept the same (i.e., emission sources, population, activity levels and pollution controls) emissions are not since some pollutant emissions (e.g., biogenic and mobile sources) depend on meteorology. Higher ambient temperatures lead to higher biogenic VOC emissions, suggesting that climate induced emission changes in a warmer environment will affect pollutant formation. The Decoupled Direct Method 3D (DDM-3D) [27–30] is incorporated in the CMAQ to quantify sensitivities of air pollutants to precursor emissions [13]. These sensitivities represent how pollutant concentrations respond to precursor emission changes as if the systems were linear [13]. Although the system is not linear, extensive testing suggests the linear (first-order) response is accurate up to emission changes of the order of 30% for O₃ and 20–50% for PM_2.5 (depending on species) [31–33].

The U.S. EPA’s Environmental Benefits Mapping and Analysis Program (BenMAP) ver. 2.4.8 (http://www.epa.gov/air/benmap) was employed to estimate the potential health impact of ambient O₃ and PM_2.5 concentration changes due to climate change over the U.S. [20]. BenMAP includes a database of age-specific population and disease incidence rates, and a concentration-response functions library for use in analyzing the health effects driven by changes in air quality. The concentrationresponse functions used are consistent with those in recent regulatory analyses [34–37]. The O₃ mortality toxicity factor is 0.00052, (i.e., a 1 ppbv change in O₃ concentrations would lead to a 0.052% change in the expected number of premature deaths) [38] while the PM_2.5 mortality toxicity factor is 0.0058 (i.e., a 1 μgm⁻³ change in PM_2.5 concentrations would lead to a 0.58% change in the expected number of premature deaths) [39]. BenMAP does not account for the potential variability in the impacts of different components of PM_2.5, and the exposure-response estimates are viewed as uncertain and may vary between parts of the country. Here, the default BenMAP ozone-premature mortality relationship is used and is based on 24-hour averaged ozone levels [20]. Since population, mortality rates and disease incidence rates obtained from 2000 are used the anticipated changes in the population (increasing by 2050) and age-specific mortality rates (expected to continue to decrease) would affect future health estimates.

In order to estimate, here, the relative contribution of PM_2.5 and O₃ precursor emissions in premature mortality changes for each state in the continental U.S. the following formula is used: E X ( Y ) = Δ M X Δ C X S X ( Y )where:

➢ E_X(Y) is the mortality change induced by changes in pollutant X concentration due to a 1% reduction in precursor Y emissions over the domain

Linear responses of the pollutant concentrations, and the resulting changes in premature mortality, to precursor emissions (i.e., how premature mortality would change to a 1% reduction in SO₂, NOx, NH₃ or VOC emissions) can be used for emission reductions of up to 25–50%, depending on pollutant and environment, as mentioned above. In this way, the reduction needed in precursor emissions to offset air pollution-induced mortality due to climate change could be estimated for each state. This is the first time, to our knowledge, that an analysis of premature mortality sensitivity to air pollutant precursor emissions is performed.

Temperatures in 2050s are modeled to be higher over the U.S. with an average increase between 1 and 3 degrees [24]. During winter and spring warming is between 0 and 3 degrees. Throughout summer warming between 2 and 4 degrees is simulated over the southwestern U.S. [25]. Warming over the midwestern U.S. is found to be less, while in some regions a small cooling is related to changes in cloud cover. During fall, warming of up to 4 degrees occurs over much of the western U.S. Daily rainfall intensity increases in most regions across the continental U.S., but the change in daily rainfall frequency is more spatially variable. As modeled, changes in rainfall frequency are small during winter and spring [25]. Regional changes in precipitation up to ±5 cm yr⁻¹ are simulated for the majority of the states, while in a few states the changes will be higher than ±20 cm yr⁻¹ (more rain is simulated in the southeastern states). Extreme positive changes (higher than 50 cm yr⁻¹) are simulated over the Atlantic Ocean and Gulf of Mexico [40]. During winter and spring the changes in downward solar radiation is about 8 W/m² in the U.S. [25]. During summer, it reaches 30 W/m² over Texas. In the Midwest, cloud cover changes reduce solar radiation by up to 30 W/m². During fall, the change is positive everywhere, with a maximum over the western U.S. of 15 W/m². The changes in the number of stagnation days during winter and spring are much smaller compared to summer and fall, where the percentage change in stagnation occurrence is very significant [25].

Climate change modifies mean summer daily maximum 8-hour average O₃ concentration levels by ±3% and mean annual PM_2.5 concentrations by −3% to 6% [24]. The lengthening of stagnation events tends to increase summer O₃ concentrations particularly during intense episodes near cities while a spatially mixed impact on annual PM_2.5 levels is simulated. The latter effect is mainly due to a variable change in precipitation. Stagnation events are predicted to have the most impact in the west, northeast and plains and a small impact is anticipated in the southeast. Climate change alone leads to increasing O₃ concentrations in all the examined cities (i.e., Los Angeles, Houston, Chicago, New York, and Atlanta) and more days with daily maximum 8-hour average O₃ concentration over the air quality standard are predicted in Los Angeles, New York and Houston. First-order (linear) sensitivities suggest [13] that a 10% reduction in anthropogenic NO_X emissions causes 2–4% decreases in maximum ozone concentrations. Reductions in VOC emissions are also beneficial for decreasing O₃ levels. Overall, O₃ sensitivities to anthropogenic NO_X, biogenic VOC, and anthropogenic VOC emissions are predicted to increase only slightly in 2050 compared to 2001 due to climate change. SO₂, NH₃, anthropogenic NOx and biogenic VOCs were found to be important precursors for PM_2.5 formation, with climate change modeled to affect slightly these sensitivities.

Air pollution-related premature mortality will be higher in the future in more than 2/3 of the states due to climate changes. Model results find that New York, along with the states in the Great Lakes and the northeastern U.S. will be affected more. Conversely, Texas and the southeastern states will experience a smaller effect [20]. The PM_2.5-related health effects dominate the O₃-related health effects but the geographic pattern of changes in O₃ concentrations is significantly different than the patterns observed for PM_2.5. About 4,000 more PM_2.5-related premature deaths are projected nationally for 2050 compared to 2001 with more incidents in the Great Lakes area and the northeastern U.S. and less in the southern states. In addition, about 300 more O₃-related premature deaths are projected nationally for 2050 compared to 2001. Climate change-related increased O₃ health effects are less pronounced in the Great Lakes area and more pronounced for the southern states.

PM_2.5 concentrations are more sensitive to SO₂, NO_X and NH₃ emissions (Figure 1) than other species (e.g., VOCs). Atmospheric SO₂ is oxidized to sulfuric acid which reacts with ammonia to form ammonium sulfate while gas-phase NO_X, oxidizes to nitric acid which reacts with ammonia to form ammonium nitrate [41]. States which are simulated to be more sensitive to SO₂ emissions are those with elevated SO₂ emissions such as the eastern states [42] (decreases up to 0.035 μgm⁻³ in daily state average PM_2.5 concentration for a 1% reduction in SO₂ emissions), while the western states are less sensitive (decrease between 0.003 and 0.008 μgm⁻³ in daily state average PM_2.5 concentration for a 1% reduction in SO₂ emissions). Midwest states are simulated to be more sensitive to anthropogenic NO_X emissions since this sub-region experiences relatively large NO_X and NH₃ emissions [42] (decrease up to 0.026 μgm⁻³ in daily state average PM_2.5 concentration for a 1% reduction in anthropogenic NO_X emissions). The sensitivity of PM_2.5 to NH₃ emissions follows SO₂ and NO_X spatial distributions and contributes to a decrease of up to 0.04 μgm⁻³ in daily state average PM_2.5 concentration for a 1% reduction in NH₃ emissions. The impact of both biogenic and anthropogenic VOC emissions changes to PM_2.5 concentration is less important. A 1% reduction in VOC emissions decreases daily state average PM_2.5 concentration in few states (up to 0.01 μgm⁻³ and 0.004 μgm⁻³ for biogenic and anthropogenic emissions, respectively) and increases daily state average PM_2.5 concentration in other states (up to 0.003 μgm⁻³ and 0.001μgm⁻³ for biogenic and anthropogenic emissions, respectively).

When NO_X and VOCs mix in the presence of sunlight, ground level O₃ is formed [41]. The response of ambient O₃ formation to reductions in NOx and VOC emissions depends on the relative abundance of NO_X and VOCs, as well as meteorological factors. The majority of the states have a positive response to anthropogenic NO_X emissions, with a decrease of up to 0.067 ppb in daily state average O₃ concentration for a 1% emissions reduction. A few states located in the Midwest and Northeast sub-regions have a negative response, with an increase of up to 0.044 ppb in daily state average O₃ concentration for a 1% reduction in anthropogenic NO_X emissions. VOC emissions are also important to O₃ responses. A 1% reduction in anthropogenic and biogenic VOC emissions are simulated to reduce O₃ concentrations up to 0.021 ppb in the eastern states while a minor negative response to biogenic VOC emissions is noticed for the northwestern states (an increase of up to 0.005 ppb in daily state average O₃ concentration for a 1% reduction in biogenic VOC emissions). SO₂ and NH₃ emissions are simulated to have a minor impact (a decrease of up to 0.004 ppb and increase of up to 0.006 ppb in daily state average O₃ concentrations for a 1% reduction in SO₂ and NH₃ emissions, respectively).

States with high emission rates [42] and significant premature mortality increases induced by PM_2.5 concentrations modulated by climate change (i.e., midwestern and northeastern U.S. sub-regions) are estimated to substantially benefit from SO₂, anthropogenic NO_X or NH₃ emissions reduction (Table 1). Illinois is simulated to be the state where emissions reduction will most significantly decrease PM_2.5-induced premature mortality: a 1% reduction in SO₂, anthropogenic NO_X or NH₃ emissions results in 28, 27 and 40 less incidents, respectively. States with fewer related incidents in the future (e.g., Texas, Florida) will also benefit from emissions reduction. In general, reduction in both anthropogenic and biogenic VOC emissions plays a minor role compared to SO₂, anthropogenic NO_X and NH₃ emissions reduction.

States with premature mortality increases induced by O₃ concentrations modulated by climate change are estimated to benefit mainly from anthropogenic NO_X emissions reduction. In the majority of the states, anthropogenic NO_X emissions reduction will reduce premature death, however, in a few states where VOCs are the limiting precursor for O₃ formation, NO_X emissions reductions are found to result in an increase (e.g., NJ, IL, OH, PA, IN). This is partly an artifact of using the exposure-response relationship for O₃ based on a 24-hour average. 24-hour average O₃ levels can respond negatively to NO_X emissions when 8-hour maximum levels would respond positively. Texas and California are simulated to be the states that will benefit most: a 1% reduction in anthropogenic NO_X emissions results in about 4 less premature deaths. Reduction in both anthropogenic and biogenic VOC emissions are also simulated to be beneficial for the states with high premature mortality increase induced by O₃ concentrations modulated by climate change (between 0.6 and 1.7 fewer incidents for a 1% reduction of VOCs). Northwestern states are simulated to have a small increase in premature mortality due to biogenic VOC emissions reduction. As anticipated, SO₂ or NH₃ emissions reductions only slightly modify O₃-related premature mortality since these two pollutants do not have a large impact on ozone formation.

Generally, the effect of emissions reduction in cumulative (total) premature mortality induced by both PM_2.5 and O₃ changes follows the PM_2.5 trend since PM_2.5 related mortality has been found higher than that due to O₃ [20]. In a few states O₃ related premature mortality modulated by anthropogenic NO_X and VOC emissions reduction play an important role in the cumulative results (e.g., NJ and RI for NO_X, IL, AR, KY and TN for biogenic VOCs, AZ and AL for anthropogenic VOCs) (Table 1).

Reduction in one precursor emission class (i.e., SO₂, anthropogenic NO_X, NH₃, or VOCs) is estimated to be able to offset premature mortality induced by PM_2.5 and O₃ changes modulated by climate change in most of the states (Table 2). States with increases of more than 400 premature deaths will be able to offset those incidents by reducing SO₂ or NH₃ emissions. For the majority of the states with less than 400 deaths, the reduction in anthropogenic NO_X emissions is estimated to be another feasible option to offset the increased premature mortality from climate-related air pollution increases. Reduction in VOC emissions works best in a few states. Nine states (i.e., IL, LA, KY, MS, IA, NM, DE, KS and VT) will be able to offset premature mortality by reducing 17% or less of their SO₂, anthropogenic NO_X or NH₃ emissions while seven states (i.e., MA, CT, WA, OR, NH, ME, and RI) need reductions in more than one precursor emission class. Although in this study a domain wide emissions reduction has been applied, impacts of precursor emissions on air quality drop quickly with increasing distance between receptor and emission sources [43]. This suggests that emission controls in a specific state will have the major impact in air quality and the induced health effects within that state, except for some of the smaller, downwind states.