Session Presentations

Jacob S. Becker, Marissa N. DeLang, Kai-Lan Chang, Marc L. Serre, Owen R. Cooper, Martin G. Schultz, Sabine Schroeder, Xiao Lu, Lin Zhang, Makoto Deushi, Beatrice Josse, Christoph A. Keller, Jean-Francois Lamarque, Meiyun Lin, Junhua Liu, Virginie Marecal, Sarah A. Strode, Kengo Sudo, Simone Tilmes, Stephanie Cleland, Elyssa Collins, J. Jason West

Station based observations provide high fidelity estimates of ground-level ozone, but provide low spatial coverage, while models are less accurate but have the advantage of global coverage. We use the Bayesian Maximum Entropy (BME) and Regionalized Air Quality Model Performance (RAMP) frameworks to integrate observations from 7,269 globally distributed surface monitoring sites from the Tropospheric Ozone Assessment Report and 1,565 sites from the Chinese National Environmental Monitoring Center Network, with nine global atmospheric chemistry models. Observations and models from 1990-2017 were first aggregated into daily maximum 8-hour average (mda8) values for each year. Then, a multi-model composite is created with the M³ Fusion method which weights models in a given continent and year by their predictive ability. We then further refine this composite with RAMP, a non-homogenous, non-linear, non-homoscedastic bias correction. RAMP corrects on a smaller scale than M³ fusion by correcting each model point based on the model trends at the nearest observations. Posterior to this, BME corrects the RAMP estimate by matching observations, with the observational influence decreasing across space and time until the output matches the multi-model or RAMP composite. The method successfully incorporates observations, as quantified by an R² increase from 0.28 when using only the M³ Model to 0.49 when using the RAMP corrected model when compared to monitor values. BME greatly increases R² to 0.81 when using the M³ Model and 0.83 when using the RAMP corrected model as the global background. Our final product estimates annual global surface ozone for each year between 1990 and 2017 at 0.1km resolution.

Martin Gauthier, Jeff Lundgren, Greg Conley, Julia Veerman, Jyotsna Kashyap, Golnoosh Bizhani, Louise Aubin, Nancy Lotecki, Kathie Brown

Over the past six years the Regional Municipality of Peel has undertaken a project to develop a flexible and comprehensive air quality modelling system, the purpose of which is to study the impacts of potential emission scenarios, and to guide policy decisions relating to public health, urban growth, and sustainability programs. The modeling system is based on WRF/SMOKE/CMAQ and has been used with nested 36-km, 12-km, 4-km and 1-km resolution grids to perform year-long model simulations for year 2015. The Regional Municipality of Peel has simulated the impact on air quality of an Off-peak delivery (OPD) emission scenario, where delivery of goods occurs during evening and overnight hours. Emissions processing was performed using SMOKE to temporally and spatially allocate the truck delivery emissions in the off-peak period. This presentation will lay out assumptions and methodology and will present preliminary results of the impact of OPD on air quality and the potential for using the modelled results for Health impact assessment.

Michael Breen, Catherine Seppanen, Vlad Isakov, Sarav Arunachalam, Miyuki Breen, James Samet, Haiyan Tong, Wayne Cascio

Epidemiological studies of ambient fine particulate matter (PM2.5) and ozone (O3) often use outdoor concentrations from central-site monitors as exposure surrogates. Failure to account for variability of indoor infiltration of ambient PM2.5 and O3, time spent indoors, and personal inhalation rates can introduce exposure errors. Personal air pollution measurements are often not feasible, and currently available exposure models require substantial technical expertise and near real-time exposure estimates are not possible since they require collection, organization and processing of large and diverse input data. To address these limitations, we developed a smartphone-based exposure model called TracMyAir, which automatically determines individual-level exposure metrics in real-time for ambient PM2.5 and O3. The input data for TracMyAir includes: (1) outdoor concentrations from the nearest air pollution monitors, (2) outdoor temperatures and wind speeds from the nearest weather stations, (3) the user's geolocation and physical activity level (PAL) from smartphone sensors, and (4) the user's home building characteristics and operating conditions. We previously linked a building infiltration model, geolocation-based microenvironment model, and PAL-based inhalation rate model to determine personal exposures and inhaled doses. In this study, we integrated the CMAQ air quality model into the TracMyAir smartphone app to account for the spatio-temporal variability of outdoor PM2.5 and O3 concentrations, and enable future forecasting capabilities. The extended TracMyAir app is being evaluated in a pilot study in central North Carolina. Using CMAQ to account for the variability of outdoor air pollutants can improve TracMyAir's exposure assessments for epidemiological investigations, in support of improving health risk assessments. Also, CMAQ's forecasting capability would enable TracMyAir to be used for public health strategies to help at risk individuals reduce their exposures to ambient air pollutants, such as wildfire smoke.

E.L. D'Ambro1, H.O.T. Pye2, C. Allen3, K. Talgo3, L. Reynolds3, K. Brehme3, R. Gilliam2, J.O. Bash2, B.N. Murphy2

1Oak Ridge Institute for Science and Education, Office of Research and Development, US Environmental Protection Agency, Research Triangle Park, NC 27711, USA

2Office of Research and Development, US Environmental Protection Agency, Research Triangle Park, NC 27711, USA

3General Dynamics Information Technology, Research Triangle Park, NC 27709, USA

Per- and polyfluoroalkyl substances (PFAS) are a class of man-made compounds whose emissions to air may contribute, via transport and deposition, to concentrations in surface water, ground water, and private well water in the vicinity of large point sources. Air quality modeling techniques can be used to quantify air concentrations and deposition fluxes to help parse the role of exposure pathways such as direct inhalation and ingestion via contaminated water. We apply the Community Multiscale Air Quality model (CMAQ) version 5.3.2 to a case study in Eastern North Carolina to model the PFAS emissions and transport at fine scale (1 km) from the Chemours Inc. Fayetteville-Works. The 26 PFAS with the largest emissions by mass are identified and added explicitly to CMAQ, along with an aggregate "other PFAS" species to represent the balance of the emissions. An updated deposition parameterization (Surface Tiled Aerosol and Gaseous Exchange, STAGE) is utilized along with estimated chemical properties for each species to simulate the deposition flux to specific, sub-grid land surface types. Thus, the updated model (CMAQ-PFAS) captures the dynamic transformations and removal processes that affect the extent of atmospheric transport of PFAS. The Fayetteville-Works site is unique as it is the only fluoropolymer manufacturer to our knowledge to provide a detailed accounting of quantified speciated emissions. We evaluate the impacts on air concentration and deposition of assuming consistent emissions throughout the year versus daily varying emissions. Next, we evaluate the role of explicit speciated emissions relative to different lumping mechanisms within CMAQ-PFAS. Finally, we examine the role of chemistry on fate, namely phase transformations, by assuming a hydrolysis of acyl fluorides to carboxylic acids with a simplified approach assuming instantaneous conversion. Our findings will help inform the level of emissions detail needed from other facilities for accurate future modeling efforts.

Disclaimer: The views expressed in this presentation are those of the authors and do not necessarily reflect the views or policies of the U.S. EPA.

James T. Kelly¹, Carey Jang¹, Brian Timin¹, Qian Di², Joel Schwartz³, Yang Liu⁴, Aaron van Donkelaar^5,6, Randall V. Martin^5-7, Veronica Berrocal⁸, and Michelle L. Bell⁹

¹Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC, USA

²Vanke School of Public Health, Tsinghua University, Beijing, China

³Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA

⁴Gangarosa Department of Environmental Health, Rollins School of Public Health, Emory University, Atlanta, GA, USA

⁵Department of Physics and Atmospheric Science, Dalhousie University, Halifax, Nova Scotia, Canada

⁶Department of Energy, Environmental & Chemical Engineering, Washington University, St. Louis, MO, USA

⁷Harvard-Smithsonian Centre for Astrophysics, Cambridge, Massachusetts 02138, United States

⁸Donald Bren School of Information and Computer Sciences, University of California, Irvine, CA, USA

⁹School of the Environment, Yale University, New Haven, CT, USA

Epidemiologic studies have found associations between fine particulate matter (PM_2.5) exposure and adverse health effects using exposure models that incorporate monitoring data and other relevant information. Here, we use nine PM_2.5 concentration models (i.e., exposure models) that span a wide range of methods to investigate i) PM_2.5 concentrations in 2011, ii) potential changes in PM_2.5 concentrations between 2011 and 2028 due to on-the-books regulations, and iii) PM_2.5 exposure (population-weighted concentrations) for the U.S. population and four racial/ethnic groups. The exposure models include two geophysical chemical transport models (CTMs), two interpolation methods, a satellite-derived aerosol optical depth-based method, a Bayesian statistical regression model, and three machine learning methods. We focus on annual predictions that were re-gridded to 12-km resolution over the conterminous U.S., but also considered 1-km predictions in sensitivity analyses. The exposure models predicted broadly consistent PM_2.5 concentrations; however, differences in national concentration distributions (median standard deviation: 1.00 ug m^-3) and spatial distributions over urban areas were evident. PM_2.5 concentrations were estimated to decrease by about 1 ug m^-3 on average due to modeled emission changes between 2011 and 2028, with decreases of more than 3 ug m^-3 in areas with relatively high 2011 concentrations. About 50% of the population was estimated to experience PM_2.5 concentrations less than 10 ug m^-3 in 2011 and PM_2.5 improvements of about 2 ug m^-3 due to modeled emission changes between 2011 and 2028. Two inequality metrics generally yielded consistent information and suggest that the modeled emission reductions between 2011 and 2028 would reduce exposure inequality on average for the four racial/ethnic groups.

Cheng-Pin Kuo ^a, Joshua S. Fu ^a

^a Department of Civil and Environmental Engineering, University of Tennessee Knoxville, Knoxville, TN, USA

Global Burden of Disease (GBD) of PM_2.5 exposure has been estimated by several previous studies and reports with different methodologies. However, although most of the researches considered the heterogeneity of spatial PM_2.5 exposure, they still used national-wide risk to estimate GBD and also overlooked the distribution of population and heterogeneous risks for people living at diverse levels of urbanizations. Thus, the following estimation of GBD must be largely biased. To approximate the facts, the diverse PM_2.5 exposure risk in diverse urbanizations and ambient PM_2.5 concentration should be applied simultaneously to estimate GBD. Also, the uncertainty between different methods should be quantified. Our objective is to identify the potential concerns and quantify the uncertainty of present GBD methodology with different risk and PM_2.5 exposure data. In this study, we developed a systematic methodology to calculate localized PM_2.5 exposure risk estimates and their heterogeneity in a rural and urban area. With collecting hospital emergency visit data, address data, air quality data, meteorological data, 1 km* 1km PM_2.5 emission data and 1 km * 1 km land-use patterns data, we used case-crossover study design and conditional logistic regressions to model the relationship between short-term PM_2.5 exposure and the consequent risks. Furthermore, to quantify the urbanization levels, we applied the heterogeneity index of the land-use living pattern (HLUL) and neighboring PM_2.5 emission densities of patients, estimating the risk for patients living in diverse levels of urbanizations. Also, to evaluate the uncertainty of using different PM_2.5 exposure data, we compared the measurement-applied estimates with CMAQ-fused results for areas with and without monitoring sites. We found that people residing in more urbanized areas would have a higher vulnerability of short-term PM_2.5 exposure for cardiovascular disease and these areas have higher hospital admissions due to higher population density. Meanwhile, we identified the overall uncertainty using different methods ranging from 14%-138% in our study area. In conclusion, we suggest the city or country without the local burden of disease estimations could utilize our approach to build their own estimations. Furthermore, the proposed burden of disease map would also facilitate re-assessment of the potential risk of present urban planning strategies, and provide a quantified reference for air quality implementation plans and emergency episode-response plans.

Brandon Lewis*, Viney P. Aneja, William H. Battye

Concentrated animal feeding operations (CAFOs) produces tons of animal waste, which can inherently pollute air, soil and water when not properly processed and filtered. The concentration of hog production in North Carolina have raised concerns of the disproportionate exposure of air pollution on vulnerable communities. Pollutants such as ammonia, hydrogen sulfide, acetaldehyde, and methanol are emitted by CAFOs and at high enough concentrations could affect human health. This research investigates the exposure of ammonia and hydrogen sulfide and possible health impacts of nearby community members looking at the disparities that may exist between different subpopulations. Characteristic data from 483 hog facilities within Duplin County including locations and hog inventory were gathered and processed for point source dispersion modeling. Emission factors from the U.S. Environmental Protection Agency in cooperation with Carnegie-Mellon University were used in the calculation of ammonia and hydrogen sulfide emission rates. We used HEM-3, Human Exposure Model, to estimate ambient concentrations of ammonia and hydrogen sulfide and the subsequent exposures on communities within Duplin County. We combine this with Census demographic data (2010) using spatial analysis to investigate whether exposures to these pollutants differ by race/ethnicity, age, income, education, and language proficiency. Based on these estimations, we assess associated health risks extenuating estimated concentrations. In this work, we limit our analysis to Duplin County, North Carolina. Results show that the average annual estimated concentration of ammonia in Duplin county is 6.05 g/m3, and the average annual estimated concentration of hydrogen sulfide is 0.06 g/m3. The max average annual ambient concentration estimated is 51.67 g/m3 and 0.52 g/m3 for ammonia and hydrogen sulfide, respectively. Among vulnerable populations within Duplin County results show that people of low income, minorities, people with low educational attainment, and the linguistically isolated are disproportionately exposure to higher levels of ammonia and hydrogen sulfide on average. The linguistically isolated are estimated to experience 101% higher levels of exposure and people with low educational attainment more that 45% higher levels. Block group characterizations indicate a high prevalence of hogs and associated higher pollutant exposures in the two highest quintile compared with the lowest quintiles for each of these subpopulations. The partial distribution of ammonia and hydrogen sulfide exposures and hog facilities among communities may have adverse health effects and environmental impacts.

Havala Pye,¹ Cavin Ward-Caviness,¹ Ben Murphy,¹ Wyat Appel,¹ Karl Seltzer²

¹Office of Research and Development, US Environmental Protection Agency, Research Triangle Park, NC 27711

²Oak Ridge Institute for Science and Education Postdoctoral Fellow in the Office of Research and Development, US Environmental Protection Agency, Research Triangle Park, NC 27711

Fine particulate matter, PM_2.5, is associated with negative health outcomes including cardiovascular and respiratory disease deaths. Over the past few decades, the composition of PM_2.5 in the United States has undergone significant changes including a decrease in sulfate and increase in relative abundance of secondary organic aerosol (SOA). Combined with advances in modeling of SOA formation pathways (e.g oxidation of monoterpenes, isoprene, and anthropogenic volatile organic compounds) the role of SOA and its components in cardiovascular and respiratory disease death rates can be examined in a way not previously possible. In this work, we use PM_2.5 constituent concentrations from the Community Multiscale Air Quality (CMAQ) modeling system v5.3.1 (ww.epa.gov/cmaq) to examine the relationship between PM_2.5 organic aerosol (OA) components and combined cardiovascular and respiratory disease deaths. Using county-level data from the Centers for Disease Control and Prevention, we associated county-level cardiorespiratory mortality rates with SOA while adjusting for a broad array of relevant confounders. We find SOA is strongly associated with mortality independent of total PM2.5 mass. Spatial variability in SOA across the U.S. is associated with a larger increase (per unit mass) in cardiorespiratory mortality rates than total PM_2.5; with the largest associations of SOA with mortality coming from counties in the southeastern U.S. Results indicate biogenic and anthropogenic carbon sources both play a role in the overall SOA association.

Saravanan Arunachalam, Komal Shukla*, Catherine Seppanen, Brian Naess, Charles Chang, David Cooley, Frank Divita, Andreas Maier,Sarah Johnson

Increasing PM_2.5 concentration levels are linked with deteriorating human health in megacities of the world. Current tools for estimating air quality, exposure, and health impacts provide county-wide, annual estimates but tools are needed for analysis at the neighborhood scales for the evaluation of energy and air planning policies. The goal of this study is to integrate two reduced form models - COBRA (Co-Benefits Risk Assessment Health Impacts screening tool) and C-TOOLS (Community Air Quality Tools) - that predict air pollution exposures and health impacts. Along with integrating these tools, we added custom higher resolution input datasets for emissions, health incidences and population for New York City (NYC) to allow for a finer-grained, more localized analysis (by zip code tabulation area (ZCTA) rather than county level) of the health benefits from air quality management decisions and perform rapid evaluation of various policy options. In this study, we used emissions inventories from the EPA's 2016 modeling platform, and augmented with NYC-specific high resolution activity-based inventories for the energy sector, buildings, commercial cooking and transportation. These were spatially allocated at each of 192 ZCTAs in all 5 boroughs in NYC, while retaining the ability to model air quality and health benefits due to implementing various policy scenarios for each of 12 individual emissions source categories. C-TOOLS uses these emissions to calculate PM_2.5 concentrations from all these sources at a network of receptors at high spatial resolution, and are then averaged for each ZCTA. The total PM_2.5 concentration is composed of primary PM2.5 from C-TOOLS, and secondary PM_2.5 estimates from COBRA. Secondary PM_2.5 estimates in COBRA are derived using a source-receptor matrix that uses estimates of PM_2.5 precursor emissions at the zip-code level. The total PM_2.5 concentrations are then used with NYC-specific population and health incidence data (focused on both mortality and morbidity) in an updated version of COBRA. The novel integrated combined tool (C-TOOLS + COBRA for NYC) with a single web-based user interface will enable high-resolution assessments for each ZCTA or NYC-wide, and perform rapid evaluation of various policy options, specifically those that are already on the books such as the NYC Roadmap to 80x50 . The tool would allow users to analyze air quality and health benefits due to PM_2.5 from emissions reductions (or increase) scenarios. The key outputs of the novel integration would include changes in PM_2.5 concentrations, changes in incidence of mortality and morbidity due to PM_2.5, and monetary value of the public health impacts. In this presentation, we will present initial results from the air quality estimates at the zip-code level for PM_2.5 and its components, and comparisons against observations in NYC.

Alejandro Valencia, Saravanan Arunachalam, Vlad Isakov, Brian Naess, and Marc Serre

Isolating air pollution sources in a complex transportation environment to quantify their contribution is challenging, particularly with sparse stationary measurements. Mobile measurements can add finer spatial resolution to support source apportionment, but they exhibit limitations when characterizing long term concentrations. Dispersion models can help overcome these limitations. However, they are only as reliable as their input emissions inventories. Herein, we developed a method to revise emissions and improve dispersion modeling predictions using stationary/mobile measurements. One specific revision estimated an adjustment factor of ~306 for warehouse emissions, indicating a significant underestimation of our initial estimates. This revised emission rate scaled up nationally would correspond to ~3.5% of the total Black Carbon emissions in the U.S. Nevertheless, domain revisions only contribute to a 4% increase of area source emissions while improving R² values by up to 56%. After revising emissions for the dispersion model, we combine this model with stationary/mobile measurements through Bayesian Maximum Entropy (I-DISP BME) data fusion and compare to BME using only measurements (Flat BME). A 10-fold conventional cross-validation (representative of months with mobile measurements) shows that all BME methods have R² values that range from 0.787 to 0.798. A 2-fold cross-validation (representative of months with no mobile measurements) shows that the R² for I-DISP BME increases by a factor 90 when compared to Flat BME.

Robson Will

Leonardo Hoinaski

This study analyzes the performance of the microphysics and boundary layer parameterizations of the WRF model in three cities in the south of Brazil, with different geographical conditions: one with high altitude (R1), one in the coastal region (R2), and one in the middle of the continental region (R3). The use of mesoscale meteorological models such as WRF is essential to enable exposure-response studies between atmospheric conditions and air quality-related diseases. The absence of local meteorological data makes it impossible to carry out assessments on the relations between air pollution, climate, meteorology, and hospital admissions. The errors and the association measures are performed using Bias, RMSE, correlation index, and Dpielke for five meteorological elements: temperature, atmospheric pressure, relative humidity, wind speed, and planetary boundary layer. Results for temperature show that different parameterizations are best suited for each of the analyzed regions, R1 (mp_physics-6, bl_pbl_physics-1, = 0.46), R2 (mp_physics - 10, bl_pbl_physics - 7, = 0.54) and R3 ( mp_physics-6, bl_pbl_physics-7, = 0.38). The differences between the configurations are repeated for the other climatic elements, demonstrating that the use of a single parameterization is not the best option to describe the meteorological conditions on a regional scale, requiring the division of each region in smaller areas considering geographic factors. We found combinations of parameters that can be a reference for new studies involving atmospheric pollution and meteorology in the entire high latitude region of the southern hemisphere, which has been little studied so far.

Matthew Woody, Ted Palma, Mark Morris, Darcie Smith, Steve Fudge, Chris Stolte, David Lindsey, and Jill Mozier

The U.S. Environmental Protection Agency (EPA) has released an updated version of its Human Exposure Model (HEM). EPA uses HEM to perform risk assessments for sources emitting air toxics to ambient air by combining (1) the atmospheric dispersion model, AERMOD, with included meteorological data, and (2) U.S. Census Bureau population data at the Census block level. Based on the inputs for source parameters and the meteorological data, AEMROD estimates the magnitude and distribution of ambient air concentrations in the vicinity of each source. Exposure estimates generated in HEM are based on the ambient air concentrations predicted by AERMOD. These exposure estimates are combined with pollutant health reference values (a library of which is also included) to estimate cancer risks and noncancer hazards, cancer incidence, and other risk measures. New features in HEM4 will include the ability to model outside the US (with user-supplied receptors), additional mapping and data viewing functionality, and the latest version of AERMOD with additional AERMOD options beyond those available in the previous version. This presentation will showcase the features of the new HEM4 and demonstrate the usefulness of the model and the summary programs.

Shunliu Zhao, Burak Oztaner, Amir Hakami, Petros Vasilakos, Armistead G. Russell, Amanda Pappin, and the Adjoint Development Team

Benefit-per-ton (BPT) estimates are used by the U.S. EPA to quantify the human health impacts of reducing air pollutant emissions (Fann et al., 2009). Forward modelling approaches have been normally used with regional or sector-based emission reductions for estimating BPTs. Here we use the backward CMAQ adjoint model (CMAQ-ADJ; Zhao et al., 2020) to calculate the monetized benefits of reducing PM2.5 primary and precursor emissions including those of PM2.5, NOX, NH3, and SO2.

Source attribution of health impacts is based on various concentration response functions (CRFs) which link average PM2.5 concentrations with mortality counts, and subsequent monetization using the value of statistical life. The variations in time (e.g., seasons) and space in BPT values shed additional light on the cost effectiveness of emission reduction measures. We evaluate BPTs for the U.S. at 36 and 12 km resolutions. We find that our BPT estimates at the two resolutions generally conform with each other and with existing estimates using other approaches; however, the adjoint model allows for greater granularity in delineating source impacts.

In addition to presenting our BPT estimates, our presentation will also include results from sensitivity analyses as to how BPT estimates are affected by the choice of model resolution and the CRF. We also discuss how using episodic simulations to represent annual health benefit estimation would impact BPT estimates. Finally, we compare our BPT estimates with those from reduced form models, and discuss limitations of adjoint BPTs.