Human-health impacts of controlling secondary air pollution precursors

Abstract

Exposure to ozone and fine particle (PM_2.5) air pollution results in premature death. These pollutants are predominantly secondary in nature and can form from nitrogen oxides (NO_X), sulfur oxides (SO_X), and volatile organic compounds (VOCs). Predicted health benefits for emission reduction scenarios often incompletely account for VOCs as precursors as well as the secondary organic aerosol (SOA) component of PM_2.5. Here, we show that anthropogenic VOC emission reductions are more than twice as effective as equivalent fractional reductions of SO_X or NO_X at reducing air pollution-associated cardiorespiratory mortality in the United States. A 25% reduction in anthropogenic VOC emissions from 2016 levels is predicted to avoid 13,000 premature deaths per year, and most (85%) of the VOC-reduction benefits result from reduced SOA with the remainder from ozone. While NO_X (−5.7 ± 0.2 % yr⁻¹) and SO_X (−12 ± 1 % yr⁻¹) emissions have declined precipitously across the U.S. since 2002, anthropogenic VOC emissions (−1.8 ± 0.3 % yr⁻¹) and concentrations of non-methane organic carbon (−2.4 ± 1.0 % yr⁻¹) have changed less. This work indicates preferentially controlling VOCs could yield significant benefits to human health.

Graphical Abstract

Introduction

More than 100 million Americans live in counties with unhealthy levels of ozone or fine particles (PM_2.5) and numbers could increase as a result of climate change.¹ Criteria pollutants, including ozone and PM_2.5, can be emitted or formed via secondary processes such as reaction in the atmosphere. Depending on the meteorology, chemical regime, and fine particle composition, different precursor or source controls may be more effective than others in reducing adverse health outcomes. For example, ozone concentrations could be reduced via controls on nitrogen oxides (NO_x) or volatile organic compound (VOC) precursors. Fine particles can be mitigated by controlling direct emissions, sulfur dioxide (SO₂), NO_x, ammonia, or VOCs. Historically, ozone-related benefits and those due to VOC reductions have not been quantified in benefit-per-ton analyses attributing monetized benefits to specific mass of emissions due to the complexity and non-linearity of ozone chemistry and uncertainty in predicting secondary organic aerosol (SOA).^{2, 3}

The health impacts of ambient fine particles generally exceed those of ozone,⁴ and SOA is a major contributor to PM_2.5 in the United States.⁵ SOA formation from the oxidation of VOCs from vegetation and anthropogenic sources^6–10 generally dominates over direct emission of organic PM_2.5.¹¹ Significant advances in the understanding of SOA chemistry have resulted in chemical transport model algorithms that can realistically predict ambient fine particle concentrations including organic carbon.¹² In addition, SOA may be an important driver of the adverse cardiorespiratory impacts of fine particle pollution in the United States.¹³

In this work, we examine how changes in secondary pollutant precursors impact fine particle and ozone concentrations. Specifically, we ask whether equivalent fractional reductions in NO_X, SO_X, or VOC emissions in the U.S. are most effective in reducing mortality and which pollutant component changes drive the reduced mortality. In addition, results are placed in the context of historical changes in emissions and concentrations to highlight opportunities to improve public health.

Materials and Methods

The Community Multiscale Air Quality (CMAQ) model was used to create air pollution concentration fields for baseline and emission reduction scenarios. Relationships between cardiorespiratory mortality and pollutant concentrations were used with CMAQ-predicted changes in concentration to quantify changes in mortality for each emission scenario. The emission reductions in the hypothetical scenarios were compared to 2002 through 2019 trends.

Pollutant concentrations

Output from CMAQv5.3.1¹⁴ for year 2016 conditions over the contiguous U.S. was obtained from the work of Appel et al.¹² CMAQv5.3.1 predictions are generally within 10% of observations^{12, 13} and the responsiveness of CMAQ predictions to emission changes has been characterized in previous work.^{15, 16} Emissions were based on 2014 National Emissions Inventory methods¹⁷ with year-specific information when available. Three simulations were conducted with changes in anthropogenic emissions of SO_X (SO₂ and sulfuric acid), NO_X (NO, NO₂, and HONO), and VOCs (including all gas-phase non-methane organic carbon, NMOC, precursors to ozone and SOA) in CMAQ using the Detailed Emissions Scaling, Isolation, and Diagnostic module.¹⁸ Anthropogenic sectors include electric generating utilities (EGUs), non-EGU point sources, residential wood combustion, oil and gas sources, mobile sources, and other point and area sources. Emission sectors held constant in all simulations include fires, biogenic activity (e.g. vegetation), lightning, and sea spray (Table S4).

Health impact quantification

Health impacts of simulated pollutant reductions were quantified based on changes in concentrations and the association between individual pollutants and cardiorespiratory mortality. The association between individual pollutant concentrations and county-level age-adjusted cardiorespiratory mortality rates, R’, were calculated in a cross-sectional framework using single pollutant models for 2016 and data from the Centers for Disease Control and Prevention (CDC)¹⁹ as well as relevant confounders (Table S1), C_j, following the single pollutant multiple linear regression methods of Pye et al.¹³:

$$\mathit{R}\prime ={\beta }_0+{\beta }_{\text{i}}{\mathit{P}}_{\mathit{i}}+\sum _{j=1}^N{\beta }_j^{\text{′}}{\mathit{C}}_{\mathit{j}}$$ (1)

Where all β are regressed coefficients and β_i represent the pollutant (P_i) coefficients connecting a change in pollutant concentration with a change in mortality. Pollutants were SOA, sulfate (SO₄), nitrate (NO₃), and ammonium (NH₄) in PM_2.5 (defined by Pye et al.¹³) as well as O₃ due to their regional nature and the fact that they can be impacted by changes to NO_X, SO_X, and VOC emissions. Other PM_2.5 components (sea spray, soot, dust, and primary organic aerosol) were included in all CMAQ simulations but were not considered in the health impact calculations as they are emitted, can show large variability within counties, and were not directly modified compared to the baseline in any simulation. O₃ was converted to mass per volume at standard conditions using 101.325 kPa and 298K (1 ppb ~ 2 μg sm⁻³). Single pollutant models were the same as specified in Pye et al.¹³ with one exception: race was reduced to one confounder as it reduced the standard error on the regressed pollutant coefficients. Changes in county-level cardiorespiratory mortality were estimated using the regression coefficients (1) and the PM_2.5 component changes from the CMAQ emission scenarios (Supplementary Note 1).

Trends in emissions and concentrations

Nationwide anthropogenic emissions and observed pollutant concentrations were obtained from U.S. Environmental Protection Agency (EPA) datasets. Anthropogenic emissions of criteria pollutants from 2002 to 2019 were obtained from the National Emissions Inventory (NEI).²⁰ Trends in mean concentrations of NMOC (24 HOUR), sulfur dioxide (24-HR BLK AVG), and oxides of nitrogen (1 HOUR) were created from Air Quality System (AQS) measurements (Supplementary Note 2).²¹

Results and Discussion

Reductions in NO_X, SO_X, or VOC emissions generally resulted in reduced annual-mean concentrations of fine particles and ozone (Figure 1a). The 25% reduction in NO_X emissions led to a 0.11 μg m⁻³ reduction in predicted ammonium and nitrate and a 0.22 ppb (0.44 μg sm⁻³) reduction in predicted ozone (nationwide, population-weighted). Simulated ozone increased up to 2 ppb in urban areas because of reduced NO_X titration but decreased throughout the Southeast and rural parts of the country by 2–5%. Ammonium and nitrate together decreased up to ~0.4 μg m⁻³ in locations with significant agricultural activity such as the Midwest, eastern North Carolina, and California’s Central Valley in addition to the Northeast U.S. SOA decreased up to 5% (~0.2 μg m⁻³) in the Southeast but increased up to 6% in urban locations where ozone also increased. SOA from monoterpene nitrates²² was predicted to decrease ~15% in the southeastern U.S., but other SOA systems were less sensitive to NO_X due to a lack of direct NO_X dependence or competing effects.^{22, 23} Predicted sulfate generally increased by 1–2% across the Midwest through Northeast as a result of greater oxidation (Figures S2–S6).

Figure 1:

Predicted (a) population-weighted change in annual-mean concentration, (b) change in mortality, and (c) avoided mortality per Tg of emission reduction resulting from a 25% reduction in anthropogenic NO_X (−2.2 Tg), SO_X (−0.6 Tg), or VOC (−2.5 Tg) emissions compared to 2016 levels. In (a), PM_2.5 component concentrations (SOA, SO₄, NH₄, and NO₃) are μg m⁻³ while ozone is μg sm⁻³.

The overall, predicted impact of SO_X emission reductions on fine particles was comparable to that from the reduction in NO_X, but SO_X emission reductions did not meaningfully affect O₃ (<0.07% change in predicted concentration, Figure 1a). As expected, SO_X emission reductions led to less sulfate over the entire eastern U.S. by up ~14% or 0.1 to 0.2 μg m⁻³ with a population-weighted decrease in concentration of 0.10 μg m⁻³. SOA showed modest decreases of 1.4% or 0.03 μg m⁻³ in the East as a result of decreased acid-catalyzed reactions of isoprene expoxydiols.²⁴ While the epoxydiol pathways responded almost linearly to the reduction in SO_X, other SOA pathways were generally not responsive to SO_X emissions.²⁵ As SO_X emissions continue to decrease, changes in diffusivity²⁶ or acidity^{27, 28} could slow uptake of isoprene epoxydiols, further modifying the responsiveness of SOA to SO_X emissions.

A 25% reduction in anthropogenic VOC emissions was predicted to lead to greater PM_2.5 and ozone reductions than the equivalent fractional reductions in NO_X and SO_X due to substantial reductions in predicted SOA and ozone. Anthropogenic VOC-induced reductions in SOA exceeded 15% (up to 22%) over a wide portion of the US including the Midwest, Northeast, and parts of California. The largest magnitude absolute concentrations changes (as much as −0.8 μg m⁻³, generally −0.3 μg m⁻³) occurred throughout the East, southern California, and California’s Central Valley, and resulted primarily from decreased anthropogenic SOA (reductions up to 0.6 μg m⁻³) coupled with more modest reductions in SOA from monoterpene oxidation (up to 0.1 μg m⁻³) due to absorptive partitioning and oxidant feedbacks. Almost no net change in the SOA from isoprene was predicted. Predicted ozone decreased almost universally throughout the eastern U.S. and California by as much as 2%. In much of the remaining western U.S., almost no change in ozone concentrations were found. Concentration decreases in ozone were highest in urban areas and on the order of 0.5 to 1 ppb. Predicted inorganic aerosol concentrations were not strongly affected by reductions in anthropogenic VOCs.

The fine particle and ozone concentrations for 2016 conditions were associated with 110,000 ± 58,000 and 70,000 ± 23,000 cardiorespiratory deaths, respectively, in the contiguous U.S. (Supplementary Note 1), consistent with previous work.^29–31 The reductions in fine particle mass and ozone summed across the three individual emission reduction scenarios (effective change of −8.0% in PM_2.5 and −1.6% in O₃) were associated with 24,000 fewer cardiorespiratory deaths. Individual fine particle components showed some variation in their health impacts per unit mass (β ranged from 8.0 to 13 deaths per 100,000 in population for 1 μg m⁻³, Table S2), but the cardiorespiratory mortality impacts of those components generally exceeded that of O₃ per unit mass by around a factor of 10. As a result, the change in mortality compared to change in concentration (Figure 1 a,b) is mainly explained by the difference in particulate component- vs. ozone-associated mortality. Individually, the 25% reductions in NO_X, SO_X, and VOC emissions were associated with 5,500; 5,100; and 13,000 fewer deaths, respectively.

The largest predicted mortality benefits resulted from the anthropogenic VOC reduction and were driven by the predicted reduction in SOA, which accounted for 85% of the avoided deaths. VOC reductions led to less predicted mortality throughout the East and California with the largest impacts on mortality rates in urban areas (Figure 2). The spatial pattern of mortality illustrates that as NO_x decreases from urban to rural locations, ozone and pollutants that rely on oxidant levels switch from being VOC-limited to NOx-limited³² and changes in mortality due to one emission change will depend on the other precursor levels. When viewed in terms of nationwide, mean county-level changes in predicted mortality rates, NO_X and VOC emission reductions showed associations of 2.4 and 2.7 fewer deaths per 100,000 in population (age-adjusted). SO_X emission reductions were associated with 1.5 fewer age-adjusted deaths per 100,000 in population.

Figure 2:

Change in cardiorespiratory mortality rate (age-adjusted deaths per 100,000 in population) predicted to result from secondary pollutant concentration changes due to a 25% reduction in anthropogenic (a) NO_X, (b) SO_X, or (d) VOC emissions.

Fractional reductions from 2016 conditions corresponded to different absolute emission reductions (Table S4). Normalizing the avoided mortality by the mass of avoided emissions, indicated 2,500; 9,100; and 5,200 fewer deaths per Tg of NO_X, SO_X, or VOC reduced (Figure 1c). This suggests that on both a relative and per mass basis, VOC reduction benefits could exceed those of NO_X by at least a factor of 2. While SO_X reductions can have large benefits per unit mass³³ due to a more direct relationship between precursor and particulate component, marginal abatement costs are also higher as emissions are reduced.³⁴

U.S. NO_X and SO_X emission levels have been decreasing at a rate of −0.79 ± 0.05 Tg yr⁻¹ (−5.7 ± 0.2 % yr⁻¹) and −0.81± 0.08 Tg yr⁻¹ (−12 ± 1 % yr⁻¹) while VOCs have decreased more slowly at a rate of −0.24 ± 0.04 Tg yr⁻¹ (−1.8 ± 0.3 % yr⁻¹) since 2002 (Figure 3). Ambient concentrations show good correspondence with the trend in emissions³⁵ although an exact correspondence is not expected due to limited monitor coverage (particularly for VOCs, n=14 monitors), uncertainty in emission reporting, and changes in atmospheric chemical lifetime.³⁶ The observed VOC proxy could also include biogenic VOCs and is operationally defined. The trend in NEI emissions suggests that the 25% SO_X and NO_X emissions reductions examined here have likely already occurred.

Figure 3:

Trend in U.S. National Emissions Inventory anthropogenic emissions and gas-phase concentrations of NO_X, SO₂, and VOCs. Concentrations are from the EPA AQS and use total non-methane organic compounds as a proxy for VOC concentrations. Values on right indicate the 2019 emissions level relative to 2002.

However, given the current trend in anthropogenic VOC emissions, the 25% reduction (2.5 Tg) examined here would take about a decade to achieve. Observed NMOC concentrations decreased consistent with VOC emissions at a rate of −4.2 ± 1.6 ppbC yr⁻¹ (−2.4 ± 1.0 % yr⁻¹, Table S6). While Azusa, California indicated a downward trend in NMOC (−7.2 % yr⁻¹, Table S7) consistent with historical trends in that area from 1960 to 2010,^{37, 38} other locations showed less change and Beltsville, Maryland even indicated increasing NMOC since 2002 (7.4 ppbC yr⁻¹, 4.3 % yr⁻¹). Trends were not uniform across demographics and counties with high fractions of Black populations were predicted to experience higher SOA concentrations and less NMOC mitigation (Figure S8) indicating the potential for an exposure disparity that could grow in time. In addition to more steady levels of NMOC over the 18-year time period, ambient concentrations of non-methane organic compounds (160 ± 48 ppb C) exceeded those of NO_X (9.7 ± 6.2 ppb N) and SO_X (0.8 ± 0.8 ppb S) in 2019 (Table S6).

Fine particle pollution is a driver of adverse health outcomes, and reducing anthropogenic emissions of VOCs could reduce cardiorespiratory mortality associated with air pollution by reducing SOA. VOCs differ from NO_X and SO_X in terms of their emission sources with SO_X and NO_X having large combustion sources that have been targeted for control³⁹ as evident in their trends since 2002. VOC sources are varied and concentrations are monitored at 10x fewer sites than are NO_X and SO_X. Sparse observations limit our ability to understand exposure disparities for vulnerable populations. Further efforts to understand the composition and magnitude of emissions of organic compounds, for example, using co-located VOC and PM_2.5 speciation measurements, as well as how they are transformed in the atmosphere could inform public health strategies. This information may become more critical in the future as VOC concentrations in urban areas respond to temperature⁴⁰ and thus could increase with climate change. Air quality model development will also benefit from this information as model studies^{33, 41, 42} often lack anthropogenic SOA sources comparable to those used here, and many models for future air quality predictions only include SOA from biogenic precursors.⁴³ In addition, time series or long-term exposure cohort analysis could refine the epidemiologic evidence used to quantify pollutant-specific risks.

While fine particles and ozone have known linkages to negative health outcomes, emerging pollutants are not considered in calculations of the impacts of environmental pollution on mortality.⁴⁴ A significant fraction of VOCs evaporate from products,⁹ and result in near-field exposure before reaching air.¹⁰ VOCs, and more generally total reactive organic carbon,⁴⁵ include hundreds of thousands of species including known hazardous air pollutants like benzene and species with unquantified adverse health impacts like per- and polyfluoroalkyl substances.^{46, 47} Addressing VOCs released to air could reduce chemical pollutant exposure, bringing additional health benefits beyond reduced criteria pollutant formation.