Decision-Making Tools for the Mitigation of Particulate Matter Health Impacts

EASIUR Organics: Reduced-Complexity Modeling of Organic PM_2.5 Health Damages 

Reduced-complexity models (RCMs) currently do not estimate health damages from organic aerosol or overlook important nuances in the modeling of organic aerosol. EASIUR is a parameterization of health damages built on simulations from chemical transport models. Therefore, EASIUR is well-positioned to incorporate state-of-thescience modeling frameworks for organic aerosol. In this work, we critically expand the EASIUR model to include seasonal and annual marginal health damages for anthropogenic VOCs, semi-volatile POA, and intermediate-VOCs by training regression models on source-tagged simulations from the PMCAMx model. Emissions for OA precursors are individually perturbed at a representative sample of locations in the U.S., and health damages for each species and location are derived for EASIUR model training and testing. When compared with PMCAMx estimates, EASIUR-predicted marginal health damages perform well, with most models having R² above 0.8 with respect to explaining the spatial variation in health damages, fractional biases ±0.15, and fractional errors below 0.35. Nationally averaged health damages across seasons range from 2,200 to 8,900 $/tonne for high-order alkanes, 24,000 to 110,000 $/tonne for aromatics, 8,500 to 44,000 $/tonne for olefines, 300,000 to 690,000 $/tonne for semivolatile POA, and 51,000 to 120,000 $/tonne for intermediate-VOCs. We estimate total damages from organic aerosol to be $1.8 trillion in 2016. Of these damages, we find that 30% are attributable to biomass burning (including wildfires), 24% are attributable to solvent emissions, and 20% are attributable to mobile sources. We also develop source-specific health damages to facilitate their use in policy analyses. 

Changes to PM_2.5 Health Damages: Robustness of Reduced-Complexity Models for Future Policy Scenarios

Reduced-complexity models (RCMs) generally produce PM_2.5 health damages by marginally perturbing precursors emissions, and then quantifying the downwind impacts on human mortality. However, marginal health damages may be used to evaluate policies that result in non-marginal emission changes, or policies in future scenarios where the emission baseline differs from those used to develop RCMs. In both cases, marginal health damages from RCMs may not capture potential changes to PM_2.5 sensitivity. In this work, we explore changes to PM_2.5 sensitivities resulting from future changes to the emission baseline. Simulations from the PMCAMx chemical transport model are used to develop health damages resulting from a -25% emission reduction from 2005 and 2055 emission baselines. Emissions are scaled to match emission changes from 2005 to 2055 as estimated by the RCP4.5 scenario. Month-long simulations for January and July are used to account for seasonal differences in emissions and chemistry. Simulation outputs are used to estimate seasonal and annual nationwide health damages for precursors of PM_2.5, including NH₃, NO_x, SO₂, anthropogenic VOCs, and low-, semi-, and intermediate-VOCs. Three classes of anthropogenic VOCs are evaluated, including high-order alkanes, aromatics, and olefins. Emissions of NH₃, NO_x, SO₂, anthropogenic VOCs, and POA change by +24%, -70%, -82%, -42%, and -10% respectively, between the 2005 and 2055 baselines. Changes to annually averaged marginal health damages are -67% for NH₃, +16% for SO₂, +17% for NO_x (not including impacts on OA), -106% for alkanes, -35% for aromatics, -36% for olefins, -19% for IVOCs, and negligible for SVOCs and LVOCs. Changes are modest and within a factor of 2 for most precursors, except for NH₃ and alkanes, with the latter contributing little to PM_2.5 formation. While health damages for SO₂ and NO_x change modestly year-round, there are key seasonal differences resulting from seasonal changes to oxidant chemistry. Lower wintertime damages for SO₂ and NO_x increase +350% and +260%, while higher summertime damages decrease -17% and -36%, respectively. While not included in RCM estimates, impacts of NO_x on organic components of PM_2.5 become increasingly important due to NO_x-limited conditions in the future. Changes in health damages resulting from population changes and increases to wealth can potentially exceed changes resulting from chemistry. Ultimately, we recommend that analysts practice caution when using RCMs to evaluate future scenarios where NH₃, or seasonal differences are of concern. Changes to annual health damages tend to be modest for most pollutants, however this is dependent on the cancellation of opposing seasonal effects. This study provides critical guidance for RCM users to better understand uncertainties associated with the use of RCM marginal health damages in future or non-marginal scenarios. 

Bias Corrections for Species- and Source-Resolved PM_2.5 Chemical Transport Model Simulations Using a Geographically Weighted Regression 

Chemical transport models provide information – albeit with biases – about PM_2.5 composition and source that is lacking in observations. Correcting biases in simulated PM_2.5 could facilitate their use in exposure assessment, environmental epidemiology, and other applications. Here, we develop a novel dataset of species- and sourceresolved PM_2.5 estimates across the continuous U.S. for 2001 and 2010. We use geographically weighted regressions to predict and then correct for the bias in PM_2.5 concentration predictions for five chemical species (elemental carbon, organic aerosol, ammonium, nitrate, and sulfate). The regression models are trained using speciated measurements, empirical PM_2.5 exposure estimates, CTM predictions, and other geographic information. A 10-fold cross-validation shows minimal bias across all simulated PM_2.5 species (-25 – 39% before; 0 – 3 % after) and improved correlations with ground-level monitors for elemental carbon, nitrate, and organic aerosol (R² : 0.30 – 0.53 before; 0.53 – 0.87 after). GWR also outperforms OLS for all PM_2.5 species except elemental carbon, where performance is comparable. Corrected fields also show improved performance in predicting fractional composition. The species- and sourceresolved exposure estimates developed in this study are suitable for use in health analyses of PM_2.5 toxicity.