* _ \
KWJ
*1 PRO1^
Policy Assessment for the Review of the
National Ambient Air Quality Standards for
Particulate Matter

-------

-------
EPA-452/R-20-002
January 2020
Policy Assessment for the Review of the National Ambient Air Quality Standards for
Particulate Matter
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Health and Environmental Impacts Division
Research Triangle Park, NC

-------
DISCLAIMER
This Policy Assessment has been prepared by staff in the U.S. Environmental Protection
Agency's (EPA) Office of Air Quality Planning and Standards. Any findings and conclusions are
those of the authors and do not necessarily reflect the views of the EPA. Questions or comments
related to this document should be addressed to Dr. Scott Jenkins, U.S. Environmental Protection
Agency, Office of Air Quality Planning and Standards, C539-06, Research Triangle Park, North
Carolina 27711 (email: jenkins.scott@epa.gov).

-------
TABLE OF CONTENTS
LIST OF APPENDICES	iii
LIST OF TABLES	iv
LIST OF FIGURES	v
LIST 01 ACRONYMS AND ABBREVIATIONS	ix
1	INTRODUCTION	1-1
1.1	Purpose	1-1
1.2	Legislative Requirements	1-3
1.3	Hi story of Revi ews of the PM NAAQ S	1-5
1.3.1	Reviews Completed in 1971 and 1987	1-7
1.3.2	Review Completed in 1997	1-7
1.3.3	Review Completed in 2006	1-9
1.3.4	Review Completed in 2012	1-11
1.4	Current Review of the PM NAAQS	1-11
References 	1-16
2	PM AIR QUALITY	2-1
2.1	Distribution of Particle Size in Ambient Air	2-1
2.1.1 Sources of PM Emissions	2-3
2.2	Ambient PM Monitoring Methods and Networks	2-14
2.2.1	Total Suspended Particulates (TSP) Sampling	2-15
2.2.2	PMio Monitoring	2-16
2.2.3	PM2.5 Monitoring	2-17
2.2.4	PM10-2.5 Monitoring	2-22
2.2.5	Additional PM Measurements and Metrics	2-23
2.3	Ambient Air Concentrations	2-25
2.3.1	Trends in Emissions of PM and Precursor Gases	2-25
2.3.2	Trends in Monitored Ambient Concentrations	2-26
2.3.3	Predicted Ambient PM2.5 Based on Hybrid Modeling Approaches	2-43
2.4	Background PM	2-53
2.4.1	Natural Sources	2-55
2.4.2	International Transport	2-57
2.4.3	Estimating Background PM with Recent Data	2-59
References 	2-62
3	REVIEW OF THE PRIMARY STANDARDS FOR PM2.5	3-1
3.1	Approach	3-1
3.1.1	Approach UsedintheLast Revi ew	3-1
3.1.2	General Approach in the Current Review	3-9
3.2	Evidence-Based Considerations	3-15
3.2.1 Nature of Effects	3-16
1

-------
3.2.2	Potential At-Risk Populations	3-44
3.2.3	PM2.5 Concentrations in Key Studies Reporting Health Effects	3-45
3.3	Risk-Based Considerations	3-80
3.3.1	Overview of Approach to Estimating Risks	3-81
3.3.2	Results of the Risk Assessment	3-86
3.3.3	Conclusions from the Risk Assessment	3-97
3.4	CASAC Advice and Public Comments	3-98
3.5	Conclusions on the Primary PM2.5 Standards	3-100
3.5.1	Current Standards	3-101
3.5.2	Potential Alternative Standards	3-107
3.6	Areas for Future Research and Data Collection	3-121
References 	3-123
4	REVIEW OF THE PRIMARY STANDARD FOR PM10	4-1
4.1	Approach	4-1
4.1.1	Approach Used in the Last review	4-1
4.1.2	Approach in the Current Review	4-4
4.2	Evidence-Based Considerations	4-5
4.2.1 Nature of Effects	4-5
4.3	CASAC Advice and Public Comments	4-13
4.4	Conclusions on The Adequacy of the Current Standard	4-14
4.5	Areas for Future Research and Data Collection	4-16
References 	4-18
5	REVIEW OF THE SECONDARY STANDARDS	5-1
5.1	Approach	5-1
5.1.1	Approach UsedintheLast Revi ew	5-2
5.1.2	General Approach Used in the Current Review	5-8
5.2	Adequacy of the Current Secondary PM Standards	5-10
5.2.1	Visibility Effects	5-10
5.2.2	Non-Visibility Effects	5-23
5.3	CASAC Advice	5-35
5.4	Conclusions on the Secondary PM Standards	5-36
5.5	Areas for Future Research and Data Collection	5-41
References 	5-43
ii

-------
LIST OF APPENDICES
APPENDIX A. Supplemental Information on PM Air Quality Analyses
APPENDIX B. Data Inclusion Criteria and Sensitivity Analyses
APPENDIX C. Supplemental Information Related to the Human Health Risk Assessment
APPENDIX D. Quantitative Analyses for Visibility Impairment
111

-------
LIST OF TABLES
Table 1-1. Summary of NAAQS promulgated for particulate matter 1971-2012	 1-6
Table 2-1. Percent Changes in PM and PM precursor emissions in the NEI for the time
periods 1990-2014 and 2002-2014	2-26
Table 2-2. Daily and annual PM2.5 design values for the near-road sites in major CBSAs
(2015-2017)	2-33
Table 2-3. Mean 2011 PM2.5 concentration by region for predictions in Figure 2-24 .... 2-48
Table 3-1. Key causality determinations for PM2.5 and UFP exposures	3-18
Table 3-2. Summary of information from PM2.5 controlled human exposure studies	3-47
Table 3-3. Epidemiologic studies examining the health impacts of long-term reductions in
ambient PM2.5 concentrations	3-62
Table 3-4. Epidemiologic studies used to estimate PIVh.s-associated risk	3-84
Table 3-5. Estimates of PIVh.s-associated mortality for air quality adjusted to just meet the
current or alternative standards (47 urban study areas)	3-87
Table 3-6. Estimated reduction in PIVh.s-associated mortality for alternative annual and 24-
hour standards (47 urban study areas)	3-88
Table 3-7. Estimates of PIVh.s-associated mortality for the current and potential alternative
annual standards in the 30 study areas where the annual standard is controlling..
	3-90
Table 3-8. Estimated delta and percent reduction in PIVh.s-associated mortality for the
current and potential alternative annual standards in the 30 study areas where
the annual standard is controlling	3-91
Table 3-9. Estimates of PIVh.s-associated mortality for the current 24-hour standard, and an
alternative, in the 11 study areas where the 24-hour standard is controlling. 3-94
Table 4-1. Key Causality Determinations for PM10-2.5 Exposures	4-6
Table 5-1. Key causality determinations for PM-related welfare effects	5-10
iv

-------
LIST OF FIGURES
Figure 2-1. Comparisons of PM2.5 and PM10 diameters to human hair and beach sand.
(Adapted from: https://www. epa.gov/pm-pollution/particulate-matter-pm-
basics)	2-2
Figure 2-2. Percent contribution of PM2.5 emissions by national source sectors. (Source:
2014 Mil)	2-5
Figure 2-3. 2014 NEI PM2.5 Emissions Density Map, tons per square mile	2-6
Figure 2-4. Percent contribution of PM10 emissions by national source sectors. (Source:
2014 NEI)	2-7
Figure 2-5. PM10 Emissions Density Map, tons per square mile	2-8
Figure 2-6. Percent contribution to organic carbon (top panel) and elemental carbon
(bottom panel) national emissions by source sectors. (Source: 2014 NEI)	2-9
Figure 2-7. Elemental Carbon Emissions Density Map, tons per square mile	2-10
Figure 2-8. Percent contribution to sulfur dioxide (panel A), oxides of nitrogen (panel B),
ammonia (panel C), and anthropogenic volatile organic compounds (panel D)
national emissions by source sectors. (Source: 2014 NEI)	2-11
Figure 2-9. SO2 Emissions Density Map, tons per square mile	2-12
Figure 2-10. NOx Emissions Density Map, tons per square mile	2-12
Figure 2-11. NH3 Emissions Density Map, tons per square mile	2-13
Figure 2-12. Anthropogenic (including wildfires) VOC Emissions Density Map, tons per
square mile	2-13
Figure 2-13. PM Monitoring stations reporting to EPA's AQS database by PM size fraction,
1970-2018	2-15
Figure 2-14. National emission trends of PM2.5, PM10, and precursor gases from 1990 to
2014	2-26
Figure 2-15. Annual average and 98th percentile PM2.5 concentrations (in |J,g/m3) from 2015-
2017 (top) and linear trends and their associated significance (based on p-
values) in PM2.5 concentrations from 2000-2017 (bottom)	2-28
Figure 2-16. Seasonally-weighted annual average PM2.5 concentrations in the U.S. from
2000 to 2017 (429 sites). (Note: The white line indicates the mean
concentration while the gray shading denotes the 10th and 90th percentile
concentrations.)	2-29
Figure 2-17. Pearson's correlation coefficient between annual average and 98th percentile of
24-hour PM2.5 concentrations from 2000-2017	2-30
Figure 2-18. Scatterplot of CBSA maximum annual versus daily design values (2015-2017)
with the solid black line representing the ratio of daily and annual NAAQS
values	2-31
Figure 2-19. Network-wide average of the hourly near-road PM2.5 increment through 2017....
	2-32
v

-------
Figure 2-20. Annual average near-road increment for PM2.5 at the Elizabeth, NJ site	2-34
Figure 2-21. Frequency distribution of 2015-2017 2-hour averages for sites meeting both or
violating either PM2.5 NAAQS for October to March (blue) and April to
September (red)	2-35
Figure 2-22. Annual average PM2.5 sulfate, nitrate, organic carbon, and elemental carbon
concentrations (in |ig/m3) from 2015-2017	2-36
Figure 2-23. Annual average and 2nd highest PM10 concentrations (in |ag/m3) from 2015-
2017 (top) and linear trends and their associated significance in PM10
concentrations from 2000-2017 (bottom)	2-38
Figure 2-24. National trends in Annual 2nd Highest 24-Hour PM10 concentrations from 2000
to 2017 (131 sites). (Note: The white line indicates the mean concentration
while the gray shading denotes the 10th and 90th percentile concentrations.) 2-39
Figure 2-25. Annual average PM2.5/PM10 ratio for 2015-2017	2-40
Figure 2-26. PM2.5/PM10 ratio for the second highest PM10 concentrations for 2015-2017	
	2-40
Figure 2-27. Annual average and 98th percentile PM10-2.5 concentrations (|ag/m3) from 2015-
2017 (top) and linear trends and their associated significance in PM10-2.5
concentrations from 2000-2017 (bottom)	2-41
Figure 2-28. Average hourly particle number concentrations from three locations in the State
of New York for 2014 to 2015 (green is Steuben County, orange is Buffalo, red
is New York City). (Source: Figure 2-18 in U.S. EPA, 2019a)	2-42
Figure 2-29. Time series of annual average mass and number concentrations (left) and
scatterplot of mass vs. number concentration (right) between 2000-2017 in
Bondvilie. IL	2-43
Figure 2-30. R2 for ten-fold cross-validation of daily PM2.5 predictions in 2015 from three
methods for individual sites as a function of observed concentration. Text
indicates the number of monitors in the PM2.5 concentration range. Downscaler:
Bayesian downscaler of CMAQ predictions; VNA: Voronoi Neighbor
Averaging; eVNA: enhanced-VNA. From Kelly et al. (2019)	2-47
Figure 2-31. Comparison of 2011 annual average PM2.5 concentrations from four methods.
(Note: These four methods include: downscaler (Berrocal et al., 2012), DI2016
(Di et al., 2016), HU2017 (Hu et al., 2017), and VD2019 (van Donkelaar et al.,
2019). Predictions have been averaged to a common 12-km grid for this
comparison	2-48
Figure 2-32. Comparison of 2011 annual average PM2.5 concentrations from four methods
for regions centered on the (a) California (b) New Jersey, and (c) Arizona.
Predictions are shown at their native resolution (i.e., about 1-km for DI2016
and VD2019 and 12-km for downscaler and HU2017)	2-50
Figure 2-33. (a) Spatial distribution of the CV (i.e., standard deviation divided by mean) in
percentage units for the four models in Figure 2-31. (b) Boxplot distributions
of CV for grid cells binned by the average PM2.5 concentration for the four
models. (Note: The box brackets the interquartile range (IQR), the horizontal
vi

-------
line within the box represents the median, the whiskers represent 1.5 times the
IQR from either end of the box, and circles represent individual values less than
and greater than the range of the whiskers.)	2-51
Figure 2-34. Distance from the center of the 12-km grid cells to the nearest PM2.5 monitoring
site for PM2.5 measurements from the AQS database and IMPROVE network. ...
	2-51
Figure 2-35. Location of PM2.5 predictions by range in annual average concentration for the
four prediction methods at their native resolution. (Note: Concentration ranges:
< 5 ng/m3, 5-7 ng/m3, 7-9 ng/m3, 9-11 ng/m3, and >11 ng/m3.)	2-52
Figure 2-36. Annual mean PM2.5 from the VD2019 method (van Donkelaar et al., 2019) for
2001, 2006, 2011, and 2016	2-53
Figure 2-37. Smoke and fire detections observed by the MODIS instrument onboard the
Aqua satellite on August 4th, 2017 accessed through NASA Worldview	2-56
Figure 2-38. Fine PM mass time series during 2017 from the North Cascades IMPROVE site
in north central Washington state	2-57
Figure 2-39. Speciated annual average IMPROVE PM2.5 in |ig/m3 at select remote monitors
during 2004 and 2016. (Note: Monitor locations are shown in Figure 2-33.)2-61
Figure 2-40. Site locations for the IMPROVE monitors in Figure 2-39. (Note: Monitors also
assessed in the 2009 ISA are shown in blue. Monitors only examined in this
assessment are shown in red.)	2-61
Figure 3-1. Overview of general approach for review of primary PM2.5 standards	3-14
Figure 3-2. Estimated concentration-response function and 95% confidence intervals
between PM2.5 and cardiovascular mortality in the Six Cities Study (1974-2009)
(from Lepeule et al., 2012, supplemental material, figure 1; Figure 6-26 in U.S.
EPA, 2019)	3-53
Figure 3-3. Epidemiologic studies examining associations between long-term PM2.5
exposures and mortality	3-57
Figure 3-4. Epidemiologic studies examining associations between long-term PM2.5
exposures and morbidity	3-58
Figure 3-5. Epidemiologic studies examining associations between short-term PM2.5
exposures and mortality	3-59
Figure 3-6. Epidemiologic studies examining associations between short-term PM2.5
exposures and morbidity	3-61
Figure 3-7. Monitored PM2.5 concentrations in key epidemiologic studies	3-65
Figure 3-8. Hybrid model-predicted PM2.5 concentrations in key epidemiologic studies. 3-68
Figure 3-9. PM2.5 annual pseudo-design values (in |ig/m3) corresponding to various
percentiles of study area populations or health events for studies of long-term
and short-term PM2.5 exposures	3-75
Figure 3-10. Map of 47 urban study areas included in risk modeling	3-83
vii

-------
Figure 3-11. Illustration of approach to adjusting air quality to simulate just meeting annual
standards with levels of 11.0 and 9.0 |ig/m3	3-85
Figure 3-12. Distribution of absolute risk estimates (PIVh.s-associated mortality) for the
current and alternative annual standards for the subset of 30 urban study areas
where the annual standard is controlling (blue and green lines represent the Pri-
PM2.5 and Sec-PM2.5 estimates, respectively)	3-92
Figure 3-13. Distribution of the difference in risk estimates between the current annual
standard (level of 12.0 ng/m3) and alternative annual standards with levels of
11.0, 10.0, and 9.0 ng/in3 for the subset of 30 urban study areas where the
annual standard is controlling	3-93
Figure 5-1. Overview of general approach for review of secondary PM standards	5-9
Figure 5-2. Relationship of viewer acceptability ratings to light extinction	5-16
Figure 5-3. Comparison of 90th percentile of daily light extinction, averaged over three
years, and 98th percentile of daily PM2.5 concentrations, averaged over three
years, for 2015-2017 using the original IMPROVE equation	5-20
Figure 5-4.
Comparison of 90th percentile of daily light extinction, averaged over three
years, and 98th percentile of daily PM2.5 concentrations, averaged over three
years, for 2015-2017 using the Lowenthal and Kumar equation	5-21

-------
LIST OF ACRONYMS AND ABBREVIATIONS
AAMS
Ambient Air Monitoring Subcommittee
ACS
American Cancer Society
AMTIC
Ambient Monitoring Technology Information Center
APEX
Air Pollutants Exposure model
AQCD
Air Quality Criteria Document
AQI
Air Quality Index
AQS
Air Quality System
ATUS
American Time Use Survey
BC
Black carbon
BenMAP-CE
Environmental Benefits Mapping and Analysis Program - Community Edition
CAA
Clean Air Act
CASAC
Clean Air Scientific Advisory Committee
CBS A
Core-based statistical area
CHAD
Consolidated Human Activity Database
CPL
Candidate protection level
C-R
Concentration-response
CSN
Chemical Speciation Network
dv
Deciview
EC
Elemental carbon
U.S. EPA
United States Environmental Protection Agency
FEM
Federal Equivalent Method
FR
Federal Register
FRM
Federal Reference Method
HERO
Health and Environmental Research Online
HREA
Health Risk and Exposure Assessment
IARC
International Agency for Research on Cancer
IHD
Ischemic heart disease
IMPROVE
Interagency Monitoring of Protected Visual Environments
IPCC
Intergovernmental Panel on Climate Change
IRP
Integrated Review Plan
ISA
Integrated Science Assessment
LML
Lowest measured level
Mm-i
Megameters
N
Nitrogen
NAAQS
National Ambient Air Quality Standards
IX

-------
NATTS	National Air Toxics 1 Trends Stations
NCEA	National Center for Environmental Assessment
NCore	National Core
NO2	Nitrogen dioxide
NOx	Oxides of nitrogen
O3	Ozone
OAR	Office of Air and Radiation
OAQPS	Office of Air Quality Planning and Standards
OC	Organic carbon
OMB	Office of Management and Budget
ORD	Office of Research and Development
PA	Policy Assessment
PM	Particulate matter
PM2.5	In general terms, particulate matter with an aerodynamic diameter less than or
equal to a nominal 2.5 [j,m; a measurement of fine particles
In regulatory terms, particles with an upper 50% cut-point of 2.5 [j,m aerodynamic
diameter (the 50% cut point diameter is the diameter at which the sampler collects
50% of the particles and rejects 50% of the particles) and a penetration curve as
measured by a reference method based on Appendix L of 40 CFR Part 50 and
designated in accordance with 40 CFR Part 53, by an equivalent method
designated in accordance with 40 CFR Part 53, or by an approved regional
method designated in accordance with Appendix C of 40 CFR Part 58
PM10	In general terms, particulate matter with an aerodynamic diameter less than or
equal to a nominal 10 [j.m; a measurement of thoracic particles (i.e, that subset of
inhalable particles thought small enough to penetrate beyond the larynx into the
thoracic region of the respiratory tract.
In regulatory terms, particles with an upper 50% cut-point of 10 ± 0.5 [j,m
aerodynamic diameter (the 50% cut point diameter is the diameter at which the
sampler collects 50% of the particles and rejects 50% of the particles) and a
penetration curve as measured by a reference method based on Appendix J of 40
CFR Part 50 and designated in accordance with 40 CFR Part 53, or by an
equivalent method designated in accordance with 40 CFR Part 53
PMio-2.5 In general terms, particulate matter with an aerodynamic diameter less than or
equal to a nominal 10 pm and greater than a nominal 2.5 pm; a measurement of
thoracic particles or the coarse fraction of PM10
In regulatory terms, particles with an upper 50% cut-point of 10 pm aerodynamic
diameter and a lower 50% cut-point of 2.5 pm aerodynamic diameter (the 50%
cut point diameter is the diameter at which the sampler collects 50% of the
particles and rejects 50% of the particles) as measured by a reference method
based on Appendix O of 40 CFR Part 50 and designated in accordance with 40
CFR Part 53, or by an equivalent method designated in accordance with 40 CFR
Part 53
x

-------
PRB	Policy relevant background
QA	Quality assurance
QMP	Quality Management Plan
REA	Risk and Exposure Assessment
RIA	Regulatory impact analysis
S	Sulfur
SES	Socioeconomic status
SIP	State Implementation Plan
SLAMS	State and Local Air Monitoring Stations
SO2	Sulfur dioxide
SOx	Sulfur oxides
SOPM	Secondary Organic Particulate Matter
STN	Speciation Trends Network
TAD	Technical Assistance Document
TRIM	Total Risk Integrated Methodology
TSP	Total Suspended Particles
UFP	Ultrafine Particles: Generally considered as particulates with a diameter less than
or equal to 0.1 |im, typically based on physical size, thermal diffusivity or
electrical mobility
UFVA	Urban-Focused Visibility Assessment
VAQ	Visual air quality
VOC	Volatile organic compound
WHO	World Health Organization
WREA	Welfare Risk and Exposure Assessment
XI

-------
1 INTRODUCTION
This document, Policy Assessment for the Review of the National Ambient Air Quality
Standards for Particulate Matter (hereafter referred to as the PA), presents the policy assessment
for the U.S. Environmental Protection Agency's (EPA's) current review of the national ambient
air quality standards (NAAQS) for particulate matter (PM). The overall plan for this review was
presented in the Integrated Review Plan for the National Ambient Air Quality Standards for
Particulate Matter (IRP; U.S. EPA, 2016). The IRP also identified key policy-relevant issues to
be addressed in this review and discussed the key documents that generally inform NAAQS
reviews, including an Integrated Science Assessment (ISA) and a PA.
This document is organized into five chapters. Chapter 1 presents introductory
information on the purpose of the PA, legislative requirements for reviews of the NAAQS, an
overview of the history of the PM NAAQS, including background information on prior reviews,
and a summary of the progress to date for the current review. Chapter 2 provides an overview of
the available information on PM-related emissions, atmospheric chemistry, monitoring and air
quality. Chapters 3 and 4 focus on policy-relevant aspects of the currently available health
effects evidence and exposure/risk information, identifying and summarizing key considerations
related to this review of the primary standards for PM2.5 and PM10, respectively. Chapter 5
focuses on policy-relevant aspects of the currently available welfare evidence and associated
quantitative analyses, identifying and summarizing key considerations related to this review of
the PM secondary standards.1
1.1 PURPOSE
The PA evaluates the potential policy implications of the available scientific evidence, as
assessed in the ISA, and the potential implications of the available air quality, exposure or risk
analyses. The role of the PA is to help "bridge the gap" between the Agency's scientific
assessments and quantitative technical analyses, and the judgments required of the Administrator
in determining whether it is appropriate to retain or revise the NAAQS.
1 The welfare effects considered in this review include visibility impairment, climate effects, and materials effects
(i.e., damage and soiling). Ecological effects associated with PM, and the adequacy of protection provided by the
secondary PM standards for them, are being addressed in the separate review of the secondary NAAQS for oxides
of nitrogen, oxides of sulfur and PM in recognition of the linkages between oxides of nitrogen, oxides of sulfur,
and PM with respect to atmospheric chemistry and deposition, and with respect to ecological effects. Information
on the current review of the secondary NAAQS for oxides of nitrogen, oxides of sulfur and PM can be found at
https://www.epa.gov/naaas/nitrogen-dioxide-no2-and-sulfur-dioxide-so2-secondarv-air-qualitv-standards.
1-1

-------
In evaluating the question of adequacy of the current standards, and whether it may be
appropriate to consider alternative standards, the PA focuses on information that is most
pertinent to evaluating the standards and their basic elements: indicator, averaging time, form,
and level.2 These elements, which together serve to define each standard, must be considered
collectively in evaluating the health and welfare protection the standards afford.
The PA is also intended to facilitate advice to the Agency and recommendations to the
Administrator from an independent scientific review committee, the Clean Air Scientific
Advisory Committee (CASAC), as provided for in the Clean Air Act (CAA). As discussed below
in section 1.2, the CASAC is to advise on subjects including the Agency's assessment of the
relevant scientific information and on the adequacy of the current standards, and to make
recommendations as to any revisions of the standards that may be appropriate. The EPA
generally makes available to the CASAC and the public one or more drafts of the PA for
CASAC review and public comment.
In this PA, we3 take into account the available scientific evidence, as assessed in the
Integrated Science Assessment for Particulate Matter (Final Report) (ISA [U.S. EPA, 2019]),
and additional policy-relevant analyses of air quality and risks. Our approach to considering the
available evidence and analyses in this PA has been informed by the advice received from the
CASAC, based on its review of the draft IRP and the draft ISA, and also by public comment
received thus far in the review. This final PA is also informed by the advice and
recommendations received from the CASAC during its review of the draft PA, and also by
public comments received. The final PA is intended to help the Administrator in considering the
currently available scientific and technical information, and in formulating judgments regarding
the adequacy of the current standards and regarding alternative standards, as appropriate.
Beyond informing the Administrator and facilitating the advice and recommendations of
the CASAC, the PA is also intended to be a useful reference to all parties interested in the review
of the PM NAAQS. In these roles, it is intended to serve as a source of policy-relevant
information that informs the Agency's review of the NAAQS for PM, and it is written to be
understandable to a broad audience.
2	The indicator defines the chemical species or mixture to be measured in the ambient air for the purpose of
determining whether an area attains the standard. The averaging time defines the period over which air quality
measurements are to be averaged or otherwise analyzed. The form of a standard defines the air quality statistic
that is to be compared to the level of the standard in determining whether an area attains the standard. For
example, the form of the annual NAAQS for fine particulate matter is the average of annual mean concentrations
for three consecutive years, while the form of the 8-hour NAAQS for carbon monoxide is the second-highest 8-
hour average in a year. The level of the standard defines the air quality concentration used for that purpose.
3	The terms "we," "our," and "staff' throughout this document refer to the staff in the EPA's Office of Air Quality
Planning and Standards (OAQPS).
1-2

-------
1.2 LEGISLATIVE REQUIREMENTS
Two sections of the Clean Air Act (CAA) govern the establishment and revision of the
NAAQS. Section 108 (42 U.S.C. 7408) directs the Administrator to identify and list certain air
pollutants and then to issue air quality criteria for those pollutants. The Administrator is to list
those pollutants "emissions of which, in his judgment, cause or contribute to air pollution which
may reasonably be anticipated to endanger public health or welfare"; "the presence of which in
the ambient air results from numerous or diverse mobile or stationary sources"; and for which he
"plans to issue air quality criteria...." (42 U.S.C. § 7408(a)(1)). Air quality criteria are intended
to "accurately reflect the latest scientific knowledge useful in indicating the kind and extent of all
identifiable effects on public health or welfare which may be expected from the presence of [a]
pollutant in the ambient air...." 42 U.S.C. § 7408(a)(2).
Section 109 [42 U.S.C. 7409] directs the Administrator to propose and promulgate
"primary" and "secondary" NAAQS for pollutants for which air quality criteria are issued [42
U.S.C. § 7409(a)], Section 109(b)(1) defines primary standards as ones "the attainment and
maintenance of which in the judgment of the Administrator, based on such criteria and allowing
an adequate margin of safety, are requisite to protect the public health."4 Under section
109(b)(2), a secondary standard must "specify a level of air quality the attainment and
maintenance of which, in the judgment of the Administrator, based on such criteria, is requisite
to protect the public welfare from any known or anticipated adverse effects associated with the
presence of [the] pollutant in the ambient air."5
In setting primary and secondary standards that are "requisite" to protect public health
and welfare, respectively, as provided in section 109(b), the EPA's task is to establish standards
that are neither more nor less stringent than necessary. In so doing, the EPA may not consider the
costs of implementing the standards. See generally, Whitman v. American Trucking Associations,
531 U.S. 457, 465-472, 475-76 (2001). Likewise, "[attainability and technological feasibility are
not relevant considerations in the promulgation of national ambient air quality standards."
American Petroleum Institute v. Costle, 665 F.2d 1176, 1185 (D.C. Cir. 1981). At the same time,
courts have clarified the EPA may consider "relative proximity to peak background ...
concentrations" as a factor in deciding how to revise the NAAQS in the context of considering
4	The legislative history of section 109 indicates that a primary standard is to be set at "the maximum permissible
ambient air level. . . which will protect the health of any [sensitive] group of the population," and that for this
purpose "reference should be made to a representative sample of persons comprising the sensitive group rather
than to a single person in such a group." S. Rep. No. 91-1196, 91st Cong., 2d Sess. 10 (1970).
5	Under CAA section 302(h) (42 U.S.C. § 7602(h)), effects on welfare include, but are not limited to, "effects on
soils, water, crops, vegetation, manmade materials, animals, wildlife, weather, visibility, and climate, damage to
and deterioration of property, and hazards to transportation, as well as effects on economic values and on personal
comfort and well-being."
1-3

-------
standard levels within the range of reasonable values supported by the air quality criteria and
judgments of the Administrator. American Trucking Associations, Inc. v. EPA, 283 F.3d 355, 379
(D C. Cir. 2002).
The requirement that primary standards provide an adequate margin of safety was
intended to address uncertainties associated with inconclusive scientific and technical
information available at the time of standard setting. It was also intended to provide a reasonable
degree of protection against hazards that research has not yet identified. See Lead Industries
Association v. EPA, 647 F.2d 1130, 1154 (D.C. Cir 1980), cert, denied, 449 U.S. 1042 (1980);
American Petroleum Institute v. Costle, 665 F.2d at 1186 (D.C. Cir. 1981), cert, denied, 455 U.S.
1034 (1982); Coalition of Battery Recyclers Ass'n v. EPA, 604 F.3d 613, 617-18 (D.C. Cir.
2010); Mississippi v. EPA, 744 F.3d 1334, 1353 (D.C. Cir. 2013). Both kinds of uncertainties are
components of the risk associated with pollution at levels below those at which human health
effects can be said to occur with reasonable scientific certainty. Thus, in selecting primary
standards that include an adequate margin of safety, the Administrator is seeking not only to
prevent pollution levels that have been demonstrated to be harmful but also to prevent lower
pollutant levels that may pose an unacceptable risk of harm, even if the risk is not precisely
identified as to nature or degree. The CAA does not require the Administrator to establish a
primary NAAQS at a zero-risk level or at background concentration levels, see Lead Industries
v. EPA, 647 F.2d at 1156 n.51, Mississippi v. EPA, 744 F.3d at 1351, but rather at a level that
reduces risk sufficiently so as to protect public health with an adequate margin of safety.
In addressing the requirement for an adequate margin of safety, the EPA considers such
factors as the nature and severity of the health effects involved, the size of the sensitive
population(s), and the kind and degree of uncertainties. The selection of any particular approach
to providing an adequate margin of safety is a policy choice left specifically to the
Administrator's judgment. See Lead Industries Association v. EPA, 647 F.2d at 1161-62;
Mississippi v. EPA, 744 F.3d at 1353.
Section 109(d)(1) of the Act requires a review be completed every five years and, if
appropriate, revision of existing air quality criteria to reflect advances in scientific knowledge on
the effects of the pollutant on public health and welfare. Under the same provision, the EPA is
also to review every five years and, if appropriate, revise the NAAQS, based on the revised air
quality criteria.6
Section 109(d)(2) addresses the appointment and advisory functions of an independent
scientific review committee. Section 109(d)(2)(A) requires the Administrator to appoint this
6 This section of the Act requires the Administrator to complete these reviews and make any revisions that may be
appropriate "at five-year intervals."
1-4

-------
committee, which is to be composed of "seven members including at least one member of the
National Academy of Sciences, one physician, and one person representing State air pollution
control agencies." Section 109(d)(2)(B) provides that the independent scientific review
committee "shall complete a review of the criteria.. .and the national primary and secondary
ambient air quality standards...and shall recommend to the Administrator any new... standards
and revisions of existing criteria and standards as may be appropriate...." Since the early 1980s,
this independent review function has been performed by the Clean Air Scientific Advisory
Committee (CAS AC) of the EPA's Science Advisory Board. A number of other advisory
functions are also identified for the committee by section 109(d)(2)(C), which reads:
Such committee shall also (i) advise the Administrator of areas in which
additional knowledge is required to appraise the adequacy and basis of existing,
new, or revised national ambient air quality standards, (ii) describe the research
efforts necessary to provide the required information, (iii) advise the
Administrator on the relative contribution to air pollution concentrations of
natural as well as anthropogenic activity, and (iv) advise the Administrator of any
adverse public health, welfare, social, economic, or energy effects which may
result from various strategies for attainment and maintenance of such national
ambient air quality standards.
As previously noted, the Supreme Court has held that section 109(b) "unambiguously bars cost
considerations from the NAAQS-setting process" (Whitman v. Am. Trucking Associations, 531
U.S. 457, 471 [2001]). Accordingly, while some of these issues regarding which Congress has
directed the CASAC to advise the Administrator are ones that are relevant to the standard setting
process, others are not. Issues that are not relevant to standard setting may be relevant to
implementation of the NAAQS once they are established.7
1.3 HISTORY OF REVIEWS OF THE PM NAAQS
This section summarizes the PM NAAQS that have been promulgated in past reviews
(Table 1-1). Each of these reviews is discussed briefly below.
7 Some aspects of CASAC advice may not be relevant to EPA's process of setting primary and secondary standards
that are requisite to protect public health and welfare. Indeed, were EPA to consider costs of implementation
when reviewing and revising the standards "it would be grounds for vacating the NAAQS." Whitman, 531 U.S. at
471 n.4. At the same time, the Clean Air Act directs CASAC to provide advice on "any adverse public health,
welfare, social, economic, or energy effects which may result from various strategies for attainment and
maintenance" of the NAAQS to the Administrator under section 109(d)(2)(C)(iv). In Whitman, the Court
clarified that most of that advice would be relevant to implementation but not standard setting, as it "enable[s] the
Administrator to assist the States in carrying out their statutory role as primary implementers of the NAAQS." Id.
at 470 (emphasis in original). However, the Court also noted that CASAC's "advice concerning certain aspects of
'adverse public health ... effects' from various attainment strategies is unquestionably pertinent" to the NAAQS
rulemaking record and relevant to the standard setting process. Id. at 470 n.2.
1-5

-------
Table 1-1. Summary of NAAQS promulgated for particulate matter 1971-2012.
Review
Completed
Indicator
Averaging
Time
Level
Form
1971
Total
Suspended
Particles
(TSP)
24-hour
260 |jg/m3
(primary)
150 pg/m3
(secondary)
Not to be exceeded more than once per year
Annual
75 |jg/m3
(primary)
60 |jg/m3
(secondary)
Annual geometric mean
1987
PM10
24-hour
150 |jg/m3
Not to be exceeded more than once per year on
average over a 3-year period
Annual
50 |jg/m3
Annual arithmetic mean, averaged over 3 years
1997
PM2.5
24-hour
65 |jg/m3
98th percentile, averaged over 3 years
Annual
15.0 |jg/m3
Annual arithmetic mean, averaged over 3 years3
PM10
24-hour
150 |jg/m3
99th percentile, averaged over 3 yearsb
Annual
50 |jg/m3
Annual arithmetic mean, averaged over 3 years
2006
PM2.5
24-hour
35 |jg/m3
98th percentile, averaged over 3 years
Annual
15.0 |jg/m3
Annual arithmetic mean, averaged over 3 years0
PM10
24-hourd
150 |jg/m3
Not to be exceed more than once per year on average
over a 3-year period
2012
PM2.5
24-hour
35 |jg/m3
98th percentile, averaged over 3 years
Annual
12.0 pg/m3
(primary)
15.0 pg/m3
(secondary)
Annual mean, averaged over 3 yearse
PM10
24-hour
150 pg/m3
Not to be exceeded more than once per year on
average over 3 years
Note: When not specified, primary and secondary standards are identical.
a The level of the 1997 annual PM2.5 standard was to be compared to measurements made at the community-
oriented monitoring site recording the highest concentration or, if specific constraints were met, measurements
from multiple community-oriented monitoring sites could be averaged (i.e., "spatial averaging") (62 FR 38652,
July 18, 1997).
b When the 1997 standards were vacated (see below), the form of the 1987 standards remained in place (i.e., not
to be exceeded more than once per year on average over a 3-year period).
c The EPA tightened the constraints on the spatial averaging criteria by further limiting the conditions under which
some areas may average measurements from multiple community-oriented monitors to determine compliance (71
FR 61144, October 17, 2006).
d The EPA revoked the annual PM10 NAAQS in 2006 (71 FR 61144, October 17, 2006).
e In the 2012 decision, the EPA eliminated the option for spatial averaging (78 FR 3086, January 15, 2013).
1-6

-------
1.3.1	Reviews Completed in 1971 and 1987
The EPA first established NAAQS for PM in 1971 (36 FR 8186, April 30, 1971), based
on the original Air Quality Criteria Document (AQCD) (DHEW, 1969).8 The federal reference
method (FRM) specified for determining attainment of the original standards was the high-
volume sampler, which collects PM up to a nominal size of 25 to 45 micrometers (|im) (referred
to as total suspended particulates or TSP). The primary standards were set at 260 |ig/m3, 24-hour
average, not to be exceeded more than once per year, and 75 |ig/m3, annual geometric mean. The
secondary standards were set at 150 |ig/m3, 24-hour average, not to be exceeded more than once
per year, and 60 |ig/m3, annual geometric mean.
In October 1979 (44 FR 56730, October 2, 1979), the EPA announced the first periodic
review of the air quality criteria and NAAQS for PM. Revised primary and secondary standards
were promulgated in 1987 (52 FR 24634, July 1, 1987). In the 1987 decision, the EPA changed
the indicator for particles from TSP to PMio, in order to focus on the subset of inhalable particles
small enough to penetrate to the thoracic region of the respiratory tract (including the
tracheobronchial and alveolar regions), referred to as thoracic particles.9 The level of the 24-hour
standards (primary and secondary) was set at 150 |ig/m3, and the form was one expected
exceedance per year, on average over three years. The level of the annual standards (primary and
secondary) was set at 50 |ig/m3, and the form was annual arithmetic mean, averaged over three
years.
1.3.2	Review Completed in 1997
In April 1994, the EPA announced its plans for the second periodic review of the air
quality criteria and NAAQS for PM, and in 1997 the EPA promulgated revisions to the NAAQS
(62 FR 38652, July 18, 1997). In the 1997 decision, the EPA determined that the fine and coarse
fractions of PMio should be considered separately. This determination was based on evidence
that serious health effects were associated with short- and long-term exposures to fine particles in
areas that met the existing PMio standards. The EPA added new standards, using PM2.5 as the
indicator for fine particles (with PM2.5 referring to particles with a nominal mean aerodynamic
diameter less than or equal to 2.5 |im). The new primary standards were as follows: (1) an annual
standard with a level of 15.0 |ig/m3, based on the 3-year average of annual arithmetic mean
8	Prior to the review initiated in 2007 (see below), the AQCD provided the scientific foundation (i.e., the air quality
criteria) for the NAAQS. Beginning in that review, the Integrated Science Assessment (ISA) has replaced the
AQCD.
9	PMio refers to particles with a nominal mean aerodynamic diameter less than or equal to 10 |im. More specifically,
10 |im is the aerodynamic diameter for which the efficiency of particle collection is 50 percent.
1-7

-------
PM2.5 concentrations from single or multiple community-oriented monitors;10 and (2) a 24-hour
standard with a level of 65 |ig/m3, based on the 3-year average of the 98th percentile of 24-hour
PM2.5 concentrations at each monitor within an area. Also, the EPA established a new reference
method for the measurement of PM2.5 in the ambient air and adopted rules for determining
attainment of the new standards. To continue to address the health effects of the coarse fraction
of PM10 (referred to as thoracic coarse particles or PM10-2.5; generally including particles with a
nominal mean aerodynamic diameter greater than 2.5 |im and less than or equal to 10 |im), the
EPA retained the annual primary PM10 standard and revised the form of the 24-hour primary
PM10 standard to be based on the 99th percentile of 24-hour PM10 concentrations at each monitor
in an area. The EPA revised the secondary standards by setting them equal in all respects to the
newly established primary standards.
Following promulgation of the 1997 PM NAAQS, petitions for review were filed by
several parties, addressing a broad range of issues. In May 1999, the U.S. Court of Appeals for
the District of Columbia Circuit (D.C. Circuit) upheld the EPA's decision to establish fine
particle standards, holding that "the growing empirical evidence demonstrating a relationship
between fine particle pollution and adverse health effects amply justifies establishment of new
fine particle standards." American Trucking Associations v. EPA, 175 F. 3d at 1027, 1055-56
(D.C. Cir. 1999). The D.C. Circuit also found "ample support" for the EPA's decision to regulate
coarse particle pollution, but vacated the 1997 PM10 standards, concluding that the EPA had not
provided a reasonable explanation justifying use of PM10 as an indicator for coarse particles.
American Trucking Associations v. EPA, 175 F. 3d at 1054-55. Pursuant to the D.C. Circuit's
decision, the EPA removed the vacated 1997 PM10 standards, and the pre-existing 1987 PM10
standards remained in place (65 FR 80776, December 22, 2000). The D.C. Circuit also upheld
the EPA's determination not to establish more stringent secondary standards for fine particles to
address effects on visibility. American Trucking Associations v. EPA, 175 F. 3d at 1027.
The D.C. Circuit also addressed more general issues related to the NAAQS, including
issues related to the consideration of costs in setting NAAQS and the EPA's approach to
establishing the levels of NAAQS. Regarding the cost issue, the court reaffirmed prior rulings
holding that in setting NAAQS the EPA is "not permitted to consider the cost of implementing
those standards." American Trucking Associations v. EPA, 175 F. 3d at 1040-41. Regarding the
10 The 1997 annual PM2 5 standard was to be compared with measurements made at the community-oriented
monitoring site recording the highest concentration or, if specific constraints were met, measurements from
multiple community-oriented monitoring sites could be averaged (i.e., "spatial averaging"). In the last review
(completed in 2012) the EPA replaced the term "community-oriented" monitor with the term "area-wide"
monitor. Area-wide monitors are those sited at the neighborhood scale or larger, as well as those monitors sited at
micro- or middle-scales that are representative of many such locations in the same CBSA (78 FR 3236, January
15,2013).
1-8

-------
levels of NAAQS, the court held that the EPA's approach to establishing the level of the
standards in 1997 (i.e., both for PM and for the ozone NAAQS promulgated on the same day)
effected "an unconstitutional delegation of legislative authority." American Trucking
Associations v. EPA, 175 F. 3d at 1034-40. Although the court stated that "the factors EPA uses
in determining the degree of public health concern associated with different levels of ozone and
PM are reasonable," it remanded the rule to the EPA, stating that when the EPA considers these
factors for potential non-threshold pollutants "what EPA lacks is any determinate criterion for
drawing lines" to determine where the standards should be set.
The D.C. Circuit's holding on the cost and constitutional issues were appealed to the
United States Supreme Court. In February 2001, the Supreme Court issued a unanimous decision
upholding the EPA's position on both the cost and constitutional issues. Whitman v. American
Trucking Associations, 531 U.S. 457, 464, 475-76. On the constitutional issue, the Court held
that the statutory requirement that NAAQS be "requisite" to protect public health with an
adequate margin of safety sufficiently guided the EPA's discretion, affirming the EPA's
approach of setting standards that are neither more nor less stringent than necessary.
The Supreme Court remanded the case to the Court of Appeals for resolution of any
remaining issues that had not been addressed in that court's earlier rulings. Id. at 475-76. In a
March 2002 decision, the Court of Appeals rejected all remaining challenges to the standards,
holding that the EPA's PM2.5 standards were reasonably supported by the administrative record
and were not "arbitrary and capricious" American Trucking Associations v. EPA, 283 F. 3d 355,
369-72 (D.C. Cir. 2002).
1.3.3 Review Completed in 2006
In October 1997, the EPA published its plans for the third periodic review of the air
quality criteria and NAAQS for PM (62 FR 55201, October 23, 1997). After the CASAC and
public review of several drafts, the EPA's NCEA finalized the AQCD in October 2004 (U.S.
EPA, 2004a, U.S. EPA, 2004b). The EPA's OAQPS finalized a Risk Assessment and Staff Paper
in December 2005 (Abt Associates, 2005, U.S. EPA, 2005).11 On December 20, 2005, the EPA
announced its proposed decision to revise the NAAQS for PM and solicited public comment on a
broad range of options (71 FR 2620, January 17, 2006). On September 21, 2006, the EPA
announced its final decisions to revise the primary and secondary NAAQS for PM to provide
increased protection of public health and welfare, respectively (71 FR 61144, October 17, 2006).
11 Prior to the review initiated in 2007, the Staff Paper presented the EPA staff's considerations and conclusions
regarding the adequacy of existing NAAQS and, when appropriate, the potential alternative standards that could
be supported by the evidence and information. More recent reviews present this information in the Policy
Assessment.
1-9

-------
With regard to the primary and secondary standards for fine particles, the EPA revised the level
of the 24-hour PM2.5 standards to 35 |ig/m3, retained the level of the annual PM2.5 standards at
15.0 |ig/m3, and revised the form of the annual PM2.5 standards by narrowing the constraints on
the optional use of spatial averaging. With regard to the primary and secondary standards for
PM10, the EPA retained the 24-hour standards, with levels at 150 |ig/m3, and revoked the annual
standards.12 The Administrator judged that the available evidence generally did not suggest a link
between long-term exposure to existing ambient levels of coarse particles and health or welfare
effects. In addition, a new reference method was added for the measurement of PM10-2.5 in the
ambient air in order to provide a basis for approving federal equivalent methods (FEMs) and to
promote the gathering of scientific data to support future reviews of the PM NAAQS.
Several parties filed petitions for review following promulgation of the revised PM
NAAQS in 2006. These petitions addressed the following issues: (1) selecting the level of the
primary annual PM2.5 standard; (2) retaining PM10 as the indicator of a standard for thoracic
coarse particles, retaining the level and form of the 24-hour PM10 standard, and revoking the
PM10 annual standard; and (3) setting the secondary PM2.5 standards identical to the primary
standards. On February 24, 2009, the U.S. Court of Appeals for the District of Columbia Circuit
issued its opinion in the case American Farm Bureau Federation v. EPA, 559F. 3d512 (D.C.
Cir. 2009). The court remanded the primary annual PM2.5 NAAQS to the EPA because the
Agency failed to adequately explain why the standards provided the requisite protection from
both short- and long-term exposures to fine particles, including protection for at-risk populations.
American Farm Bureau Federation v. EPA, 559 F. 3d 512, 520-27 (D.C. Cir. 2009). With regard
to the standards for PM10, the court upheld the EPA's decisions to retain the 24-hour PM10
standard to provide protection from thoracic coarse particle exposures and to revoke the annual
PM10 standard. American Farm Bureau Federation, 559 F. 2d at 533-38. With regard to the
secondary PM2.5 standards, the court remanded the standards to the EPA because the Agency
failed to adequately explain why setting the secondary PM standards identical to the primary
standards provided the required protection for public welfare, including protection from visibility
impairment. American Farm Bureau Federation, 559 F. 2d at 528-32. The EPA responded to the
12 In the 2006 proposal, the EPA proposed to revise the 24-hour PM10 standard in part by establishing a new PM10-2.5
indicator for thoracic coarse particles (i.e., particles generally between 2.5 and 10 |im in diameter). The EPA
proposed to include any ambient mix of PM10-2 5 that was dominated by resuspended dust from high density
traffic on paved roads and by PM from industrial sources and construction sources. The EPA proposed to exclude
any ambient mix of PMi0-2.5 that was dominated by rural windblown dust and soils and by PM generated from
agricultural and mining sources. In the final decision, the existing PM10 standard was retained, in part due to an
"inability.. .to effectively and precisely identify which ambient mixes are included in the [PM10-2.5] indicator and
which are not" (71 FR 61197, October 17, 2006).
1-10

-------
court's remands as part of the next review of the PM NAAQS, which was initiated in 2007
(discussed below).
1.3.4 Review Completed in 2012
In June 2007, the EPA initiated the fourth periodic review of the air quality criteria and
the PM NAAQS by issuing a call for information in the Federal Register (72 FR 35462, June 28,
2007). Based on the NAAQS review process, as revised in 2008 and again in 2009,13 the EPA
held science/policy issue workshops on the primary and secondary PM NAAQS (72 FR 34003,
June 20, 2007; 72 FR 34005, June 20, 2007), and prepared and released the planning and
assessment documents that comprise the review process (i.e., IRP (U.S. EPA, 2008), ISA (U.S.
EPA, 2009a), REA planning documents for health and welfare (U.S. EPA, 2009b, U.S. EPA,
2009c), a quantitative health risk assessment (U.S. EPA, 2010a) and an urban-focused visibility
assessment (U.S. EPA, 2010b), and PA (U.S. EPA, 2011)). In June 2012, the EPA announced its
proposed decision to revise the NAAQS for PM (77 FR 38890, June 29, 2012).
In December 2012, the EPA announced its final decisions to revise the primary NAAQS
for PM to provide increased protection of public health (78 FR 3086, January 15, 2013). With
regard to primary standards for PM2.5, the EPA revised the level of the annual PM2.5 standard14 to
12.0 |ig/m3 and retained the 24-hour PM2.5 standard, with its level of 35 |ig/m3. For the primary
PM10 standard, the EPA retained the 24-hour standard to continue to provide protection against
effects associated with short-term exposure to thoracic coarse particles (i.e., PM10-2.5). With
regard to the secondary PM standards, the EPA generally retained the 24-hour and annual PM2.5
standards15 and the 24-hour PM10 standard to address visibility and non-visibility welfare effects.
As with previous reviews, petitioners challenged the EPA's final rule. Petitioners argued
that the EPA acted unreasonably in revising the level and form of the annual standard and in
amending the monitoring network provisions. On judicial review, the revised standards and
monitoring requirements were upheld in all respects. NAMv EPA, 750 F.3d 921 (D.C. Cir.
2014).
1.4 CURRENT REVIEW OF THE PM NAAQS
In December 2014, the EPA announced the initiation of the current periodic review of the
air quality criteria for PM and of the PM2.5 and PM10 NAAQS and issued a call for information
13	The history of the NAAQS review process, including revisions to the process, is discussed at
http://www3 .epa. gov/ttn/naaas/review2.html.
14	The EPA also eliminated the option for spatial averaging.
15	Consistent with the primary standard, the EPA eliminated the option for spatial averaging with the annual
standard.
1-11

-------
in the Federal Register (79 FR 71764, December 3, 2014). On February 9 to 11, 2015, the EPA's
NCEA and OAQPS held a public workshop to inform the planning for the current review of the
PM NAAQS (announced in 79 FR 71764, December 3, 2014). Workshop participants, including
a wide range of external experts as well as EPA staff representing a variety of areas of expertise
(e.g., epidemiology, human and animal toxicology, risk/exposure analysis, atmospheric science,
visibility impairment, climate effects), were asked to highlight significant new and emerging PM
research, and to make recommendations to the Agency regarding the design and scope of this
review. This workshop provided for a public discussion of the key science and policy-relevant
issues around which the EPA has structured the current review of the PM NAAQS and of the
most meaningful new scientific information that would be available in this review to inform our
understanding of these issues.
The input received at the workshop guided the EPA staff in developing a draft IRP,
which was reviewed by the CASAC Particulate Matter Panel and discussed on public
teleconferences held in May 2016 (81 FR 13362, March 14, 2016) and August 2016 (81 FR
39043, June 15, 2016). Advice from the CASAC, supplemented by the Particulate Matter Panel,
and input from the public were considered in developing the final IRP for this review (U.S. EPA,
2016). The final IRP discusses the approaches to be taken in developing key scientific, technical,
and policy documents in this review and the key policy-relevant issues that will frame the EPA's
consideration of whether the current primary and/or secondary NAAQS for PM should be
retained or revised.
In May 2018, the Administrator issued a memorandum describing a "back-to-basics"
process for reviewing the NAAQS (Pruitt, 2018). This memo announced the Agency's intention
to conduct the current review of the PM NAAQS in such a manner as to ensure that any
necessary revisions are finalized by December 2020. Following this memo, on October 10, 2018
the Administrator additionally announced that the role of reviewing the key science assessments
developed as part of the ongoing review of the PM NAAQS (i.e., drafts of the ISA and PA)
would be performed by the seven-member chartered CASAC (i.e., rather than the CASAC
Particulate Matter Panel that reviewed the draft IRP).16
The EPA released the draft ISA in October 2018 (83 FR 53471, October 23, 2018). The
draft ISA was reviewed by the chartered CASAC at a public meeting held in Arlington, VA in
December 2018 (83 FR 55529, November 6, 2018) and was discussed on a public teleconference
in March 2019 (84 FR 8523, March 8, 2019). The CASAC provided its advice on the draft ISA
in a letter to the EPA Administrator dated April 11, 2019 (Cox, 2019a). In that letter, the
16 Announcement available at: https://www.epa.gov/newsreleases/acting-administrator-wheeler-announces-science-
advisors-kev-clean-air-act-committee
1-12

-------
CASAC's recommendations address both the draft ISA's assessment of the science for PM-
related effects and the process under which this review of the PM NAAQS is being conducted.
Regarding the assessment of the evidence, the CASAC letter states that "the Draft ISA
does not provide a sufficiently comprehensive, systematic assessment of the available science
relevant to understanding the health impacts of exposure to particulate matter (PM)" (Cox, 2019,
p. 1 of letter). The CASAC recommends that this and other limitations (i.e., "[inadequate
evidence for altered causal determinations" and the need for a "[cjlearer discussion of causality
and causal biological mechanisms and pathways") be remedied in a revised ISA (Cox, 2019, p. 1
of letter). The EPA has taken steps to address these comments in the Final PM ISA (U.S. EPA,
2019). In particular, the final ISA includes additional text and a new appendix to clarify the
comprehensive and systematic process employed by the EPA to develop the PM ISA. In
addition, several causality determinations were re-examined and the final ISA reflects a revised
causality determination for long-term ultrafine particle exposures and nervous system effects
(i.e., from "likely to be causal" to "suggestive of, but not sufficient to infer, a causal
relationship"). The final ISA also contains additional text to clarify the evidence for biological
pathways of particular PM-related effects and the role of that evidence in causality
determinations.
Among its comments on the process, the chartered CASAC recommended "that the EPA
reappoint the previous CASAC PM panel (or appoint a panel with similar expertise)" (Cox, 2019a).
The Agency's response to this advice was provided in a letter from the Administrator to the
CASAC chair dated July 25, 2019.17 As indicated in that letter, on September 13, 2019 the
Administrator announced the selection of a pool of non-member subject matter experts. These
experts were intended to "provide technical expertise to help CASAC ensure a rigorous and
timely review of the National Ambient Air Quality Standards for particulate matter and ozone."18
Input from members of this pool of experts informed the CASAC's review of the draft PA.
The EPA released the draft PA in September 2019 (84 FR 47944, September 11, 2019).
The draft PA was reviewed by the chartered CASAC and discussed in October 2019 at a public
meeting held in Cary, NC. Public comments were received via a separate public teleconference
(84 FR 51555, September 30, 2019). A public meeting to discuss the chartered CASAC letter
and response to charge questions on the draft PA was held in Cary, NC in December 2019 (84
FR 58713, November 1, 2019), and the CASAC provided its advice on the draft PA, including its
17	Available at:
httPs://vosemite.epa.gov/sab/sabproduct.nsf/0/6CBCBBC3025E13B4852583D90047B352/$File/EPA-CASAC-
19-002 Response.pdf
18	Available at: https://www.epa.gov/newsreleases/administrator-wheeler-announces-new-casac-member-pool-
naaas-subiect-matter-experts
1-13

-------
advice on the current primary and secondary PM standards, in a letter to the EPA Administrator
dated December 16, 2019 (Cox, 2019b).
With regard to the primary standards, the CASAC recommends retaining the current 24-
hour PM2.5 and PM10 standards, but does not reach consensus on the adequacy of the current
annual PM2.5 standard. With regard to the secondary standards, the CASAC recommends
retaining the current standards. The CASAC's advice on the primary and secondary PM
standards is discussed in detail in chapters 3 (primary PM2.5 standards), 4 (primary PM10
standards), and 5 (secondary standards) of this final PA.
The CASAC additionally makes a number of recommendations regarding the information
and analyses presented in the draft PA. Specifically, the CASAC recommends that a revised PA
include (1) additional discussion of the current CASAC and NAAQS review process; (2)
additional characterization of PM-related emissions, monitoring and air quality information,
including uncertainties in that information; (3) additional discussion and examination of
uncertainties in the PM2.5 health evidence and the risk assessment; (4) updates to reflect changes
in the ISA's causality determinations; and (5) additional discussion of the evidence for PM-
related welfare effects, including uncertainties (Cox, 2019b, pp. 2-3 in letter). In response to the
CASAC's comments, we have incorporated a number of changes into this final PA, including the
following:
(1)	We have added text to Chapter 1 (see above) to clarify the process followed for this
review of the PM NAAQS, including how the process has evolved since the initiation of
the review.
(2)	We have added text and figures to Chapter 2 on emissions of PM and PM precursors, and
we have added a section discussing uncertainty in emissions estimates. We have also
added new discussion of measurement uncertainty for FRM, FEM, CSN, and IMPROVE
monitors.
(3)	In Chapter 3 and Appendices B and C, we have made a number of changes:
a.	We have reduced the emphasis on evidence for long-term ultrafine particle exposures
and nervous system effects to reflect the change in the final ISA's causality
determination from "likely to be causal" to "suggestive of, but not sufficient to infer, a
causal relationship."
b.	We have expanded the characterization and discussion of the evidence related to
exposure measurement error, the potential confounders examined by key studies, the
shapes of concentration-response functions, and the results of causal inference and
quasi-experimental studies.
1-14

-------
c.	We have expanded and clarified the discussion of uncertainties in the risk
assessment,19 and we have added additional air quality model performance evaluation
for each of the urban study areas included in the risk assessment.
d.	We have provided additional detail on the procedure used to derive concentration-
response functions used in the risk assessment.
(4) Throughout the document (Chapters 3, 4 and 5), we have added summaries of the CAS AC
advice on the PM standards, and we have expanded the discussion of data gaps and areas
for future research in the health and welfare effects evidence.
19 The CASAC's comments on the risk assessment include recommending additional analyses to quantify
uncertainty in estimates of how PM2 5-related risks may change with changing ambient PM2 5 concentrations
(Cox, 2019b, p. 7 of consensus responses). While this final PA includes additional discussion of sources of
uncertainty in the risk assessment, and additional qualitative consideration of the potential impacts of those
uncertainties on risk estimates, we have not conducted additional analyses to further quantify uncertainty. This
approach to addressing the CASAC's comments on the risk assessment reflects our consideration of the timeline
for this review as well as the likely impact of such additional analyses on decision making.
1-15

-------
REFERENCES
Abt Associates, Inc. (2005). Particulate matter health risk assessment for selected urban areas:
Draft report. Research Triangle Park, NC, U.S. Environmental Protection Agency: 164.
Cox, LA. (2019a). Letter from Louis Anthony Cox, Jr., Chair, Clean Air Scientific Advisory
Committee, to Administrator Andrew R. Wheeler. Re: CASAC Review of the EPA's
Integrated Science Assessment for Particulate Matter (External Review Draft - October
2018). April 11, 2019. EPA-CASAC-19-002. U.S. EPA HQ, Washington DC. Office of
the Administrator, Science Advisory Board. Available at:
https://vosemite.epa.gov/sab/sabproduct.nsf/LookupWebReportsLastMonthCASAC/932
D1DF8C2A9043F852581000048170D?QpenDocument&TableRow=2.3#2.
Cox, LA. (2019b). Letter from Louis Anthony Cox, Jr., Chair, Clean Air Scientific Advisory
Committee, to Administrator Andrew R. Wheeler. Re: CASAC Review of the EPA's
Policy Assessment for the Review of the National Ambient Air Quality Standards for
Particulate Matter (External Review Draft - September 2019). December 16, 2019. EPA-
CASAC-20-001. U.S. EPA HQ, Washington DC. Office of the Administrator, Science
Advisory Board. Available at:
https://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c852574020Q7446a4/E2F
6C7173 72016128525 84D20069DFB l/$File/EP A-C AS AC-20-001 .pdf.
U.S. Department of Health. (1969). Air Quality Criteria for Particulate Matter. Washington, D.C.
National Air Pollution Control Administration. DHEW. January 1969.
Pruitt, E. (2018). Memorandum from E. Scott Pruitt, Administrator, U.S. EPA to Assistant
Administrators. Back-to-Basics Process for Reviewing National Ambient Air Quality
Standards. May 9, 2018. U.S. EPA HQ, Washington DC. Office of the Administrator.
Available at: https://www.epa.gov/criteria-air-pollutants/back-basics-process-reviewing-
national-ambient-air-qualitv-standards.
U.S. EPA. (2004a). Air Quality Criteria for Particulate Matter. (Vol I of II). Research Triangle
Park, NC. Office of Research and Development. U.S. EPA. EPA-600/P-99-002aF.
October 2004. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100LFIQ.txt.
U.S. EPA. (2004b). Air Quality Criteria for Particulate Matter. (Vol II of II). Research Triangle
Park, NC. Office of Research and Development. U.S. EPA. EPA-600/P-99-002bF.
October 2004. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100LG7Q.txt.
U.S. EPA. (2005). Review of the National Ambient Air Quality Standards for Particulate Matter:
Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper.
Research Triangle Park, NC. Office of Air Quality Planning and Standards. U.S. EPA.
EPA-452/R-05-005a. December 2005. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 1009MZM.txt.
1-16

-------
U.S. EPA. (2008). Integrated Review Plan for the National Ambient Air Quality Standards for
Particulate Matter Research Triangle Park, NC. Office of Research and Development,
National Center for Environmental Assessment; Office of Air Quality Planning and
Standards, Health and Environmental Impacts Division. U.S. EPA. EPA 452/R-08-004.
March 2008. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 1001FB9.txt.
U.S. EPA. (2009a). Integrated Science Assessment for Particulate Matter (Final Report).
Research Triangle Park, NC. Office of Research and Development, National Center for
Environmental Assessment. U.S. EPA. EPA-600/R-08-139F. December 2009. Available
at: https://cfpub.epa.gov/ncea/risk/recordisplav.cfm?deid=216546.
U.S. EPA. (2009b). Particulate Matter National Ambient Air Quality Standards: Scope and
Methods Plan for Health Risk and Exposure Assessment Research Triangle Park, NC.
Office of Air Quality Planning and Standards, Health and Environmental Impacts
Division. U.S. EPA. EPA-452/P-09-002. February 2009. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100FLWP.txt.
U.S. EPA. (2009c). Particulate Matter National Ambient Air Quality Standards: Scope and
Methods Plan for Urban Visibility Impact Assessment Research Triangle Park, NC.
Office of Air Quality Planning and Standards, Health and Environmental Impacts
Division. U.S. EPA. EPA-452/P-09-001. February 2009. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100FLUX.txt.
U.S. EPA. (2010a). Quantitative Health Risk Assessment for Particulate Matter (Final Report).
Research Triangle Park, NC. Office of Air Quality Planning and Standards, Health and
Environmental Impacts Division. U.S. EPA. EPA-452/R-10-005. June 2010. Available
at: https://nepis.epa.gov/Exe/ZyPURL.cgi?Dockev=P1007RFC.txt.
U.S. EPA. (2010b). Particulate Matter Urban-Focused Visibility Assessment (Final Document).
Research Triangle Park, NC. Office of Air Quality Planning and Standards, Health and
Environmental Impacts Division. U.S. EPA. EPA-452/R-10-004 July 2010. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100FQ5D.txt.
U.S. EPA. (2011). Policy Assessment for the Review of the Particulate Matter National Ambient
Air Quality Standards Research Triangle Park, NC. Office of Air Quality Planning and
Standards, Health and Environmental Impacts Division. U.S. EPA. EPA-452/R-11-003
April 2011. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100AUMY.txt.
U.S. EPA. (2016). Integrated review plan for the national ambient air quality standards for
particulate matter. Research Triangle Park, NC. Office of Air Quality Planning and
Standards. U.S. EPA. EPA-452/R-16-005. December 2016. Available at:
https://www3.epa.gov/ttn/naaqs/standards/pm/data/201612-final-integrated-review-
plan.pdf.
U.S. EPA. (2019). Integrated Science Assessment (ISA) for Particulate Matter (Final Report).
Washington, DC. U.S. Environmental Protection Agency, Office of Research and
1-17

-------
Development, National Center for Environmental Assessment. U.S. EPA. EPA/600/R-
19/188. December 2019. Available at: https://www.epa.gov/naaqs/particulate-matter-pm-
standards-integrated-science-assessments-current-review.
1-18

-------
2 PM AIR QUALITY
This chapter provides an overview of recent ambient air quality with respect to PM. It
summarizes information on the distribution of particle size in ambient air, including discussions
about size fractions and components (section 2.1), ambient monitoring of PM in the U.S. (section
2.2), ambient concentrations of PM in the U.S. (section 2.3), and background PM (section 2.4).
2.1 DISTRIBUTION OF PARTICLE SIZE IN AMBIENT AIR
In ambient air, PM is a mixture of substances suspended as small liquid and/or solid
particles. Particle size is an important consideration for PM, as distinct health and welfare effects
have been linked with exposures to particles of different sizes. Particles in the atmosphere range
in size from less than 0.01 to more than 10 micrometers (|J,m) in diameter (U.S. EPA, 2019a,
section 2.2). When describing PM, subscripts are used to denote the aerodynamic diameter1 of
the particle size range in micrometers (|im) of 50% cut points of sampling devices. The EPA
defines PM2.5, also referred to as fine particles, as particles with aerodynamic diameters
generally less than or equal to 2.5 [j.m. The size range for PM10-2.5, also called coarse or thoracic
coarse particles, includes those particles with aerodynamic diameters generally greater than 2.5
[j,m and less than or equal to 10 [j,m. PM10, which is comprised of both fine and coarse fractions,
includes those particles with aerodynamic diameters generally less than or equal to 10 [j,m.
Figure 2-1 provides perspective on these particle size fractions. In addition, ultrafine particles
(UFP) are often defined as particles with a diameter of less than 0.1 [j,m based on physical size,
thermal diffusivity or electrical mobility (U.S. EPA, 2019a, section 2.2).
1 Aerodynamic diameter is the size of a sphere of unit density (i.e., 1 g/cm3) that has the same terminal settling
velocity as the particle of interest (U.S. EPA, 2018, U.S. EPA, 2019a, section 4.1.1).
2-1

-------
HUMAN HA
50-70 (im
(microns) in diam
90 (im (microns) in diameter
FINE BEACH SAND
C-PMio
Dust, pollen, mold, etc.
<10|xm (microns) in diameter
• PM2.5
Combustion particles, organic
compounds, metals, etc.
<2.5jim (microns) in diameter

Figure 2-1. Comparisons of P.M2.5 and PM10 diameters to human hair and beach sand.
(Adapted from: https://www.epa.gov/pm-pollution/particulate-niatter-pm-basics)
Atmospheric distributions of particle size generally exhibit distinct modes that roughly
align with the PM size fractions defined above. The nucleation mode is made up of freshly
generated particles, formed either during combustion or by atmospheric reactions of precursor
gases. The nucleation mode is especially prominent near sources like heavy traffic, industrial
emissions, biomass burning, or cooking (Vu et al.. 2015). While nucleation mode particles are
only a minor contributor to overall ambient PM mass and surface area, they are the main
contributors to ambient particle number (U.S. EPA, 2019a, section 2.2). By number, most
nucleation mode parti cles fall into the UFP size range, though some fraction of the nucleation
mode number distribution can extend above 0.1 um in diameter. Nucleation mode particles can
grow rapidly through coagulation or uptake of gases by particle surfaces, giving rise to the
accumulation mode. The accumulation mode is typically the predominant contributor to PM2.5
mass and surface area, though only a minor contributor to particle number (U.S. EPA, 2019a,
section 2.2). PM2.5 sampling methods measure most of the accumulation mode mass, although a
small fraction of particles that make up the accumulation mode are greater than 2.5 (im in
diameter. Coarse mode particles are formed by mechanical generation, and through processes
like dust resuspension and sea spray formation (Whitby et al., 1972). Most coarse mode mass is
captured by PM10-2.5 sampling, but small fractions of coarse mode mass can be smaller than 2.5
pni or greater than 10 }im in diameter (U.S. EPA, 2019a, section 2.2).
Most particles are found in the lower troposphere, where they can have residence times
ranging from a few hours to weeks. Particles are removed from the atmosphere by wet
2-2

-------
deposition, such as when they are carried by rain or snow, or by dry deposition, when particles
settle out of suspension due to gravity. Atmospheric lifetimes are generally longest for PM2.5,
which often remains in the atmosphere for days to weeks (U.S. EPA, 2019a, Table 2-1) before
being removed by wet or dry deposition. In contrast, atmospheric lifetimes for UFP and PM10-2.5
are shorter. Within hours, UFP can undergo coagulation and condensation that lead to formation
of larger particles in the accumulation mode, or can be removed from the atmosphere by
evaporation, deposition, or reactions with other atmospheric components. PM10-2.5 are also
generally removed from the atmosphere within hours, through wet or dry deposition (U.S. EPA,
2019a, Table 2-1).
2.1.1 Sources of PM Emissions
PM is composed of both primary (directly emitted particles) and secondary chemical
components. Primary PM is derived from direct particle emissions from specific PM sources
while secondary PM originates from gas-phase chemical compounds present in the atmosphere
that have participated in new particle formation or condensed onto existing particles (U.S. EPA,
2019a, section 2.3). Primary particles, and gas-phase compounds contributing to secondary
formation PM, are emitted from both anthropogenic and natural sources.
Anthropogenic sources of PM include both stationary and mobile sources. Stationary
sources include fuel combustion for electricity production and other purposes, industrial
processes, agricultural activities, and road and building construction and demolition. Mobile
sources of PM include diesel- and gasoline-powered highway vehicles and other engine-driven
sources (e.g., ships, aircraft, and construction and agricultural equipment). Both stationary and
mobile sources directly emit primary PM to ambient air, along with secondary PM precursors
(e.g., SO2) that contribute to the secondary formation of PM in the atmosphere (U.S. EPA,
2019a, section 2.3, Table 2-2).
Natural sources of PM include dust from the wind erosion of natural surfaces, sea salt,
wildland fires, primary biological aerosol particles (PBAP) such as bacteria and pollen, oxidation
of biogenic hydrocarbons such as isoprene and terpenes to produce secondary organic aerosol
(SOA), and geogenic sources such as sulfate formed from volcanic production of SO2 (U.S.
EPA, 2009, section 3.3, Table 3-2). While most of the above sources release or contribute
predominantly to fine aerosol, some sources including windblown dust, and sea salt also produce
particles in the coarse size range (U.S. EPA, 2019a, section 2.3.3).
Generally, the sources of PM for different size fractions vary. While PM2.5 in ambient air
is largely emitted directly by sources such as those described above or through secondary PM
formation in the atmosphere, PM10-2.5 is almost entirely from primary sources (i.e., directly
emitted) and is produced by surface abrasion or by suspension of sea spray or biological
2-3

-------
materials such as microorganisms, pollen, and plant and insect debris (U.S. EPA, 2019a, section
2.3.2.1).
In sections 2.1.1.1 and 2.1.1.2 below, we describe the most recently available information
on sources contributing to PM2.5 and PM10-2.5 emissions into ambient air, respectively, based on
the U.S. EPA 2014 National Emissions Inventory (NEI).2 In section 2.1.1.3, we describe
information on sources contributing to emissions of PM components and precursor gases.
2.1.1.1 Sources Contributing to Primary PM2.5 Emissions
The National Emissions Inventory (NEI) is a comprehensive and detailed estimate of air
emissions of criteria pollutants, criteria precursors, and hazardous air pollutants from a
comprehensive set of air emissions sources, including point sources (electric generating units,
boilers, etc.), nonpoint (or area) sources (oil & gas, residential wood combustion, and many other
dispersed sources), mobiles sources, and events (large fires). There are over 3,000 sources for
which the NEI is developed. The NEI is released every three years based primarily upon data
provided by State, Local, and Tribal air agencies for sources in their jurisdictions and
supplemented by data developed by the U.S. EPA. The NEI is built using the Emissions
Inventory System (EIS) first to collect the data from State, Local, and Tribal air agencies and
then to blend that data with other data sources.
Based on the 2014 NEI, approximately 5.4 million tons/year of PM2.5 were estimated to
be directly emitted to the atmosphere from a number of source sectors in the U.S. This total
excludes sources that are not a part of the NEI (e.g., windblown dust, geogenic sources). As
shown in Figure 2-2, nearly half of the total primary PM2.5 emissions nationally are contributed
by the dust and fire sectors together. Dust includes agricultural, construction, and road dust. Of
these, agricultural dust and road dust in sum make the greatest contributions to PM2.5 emissions
nationally. Fires include wildfires, prescribed fires, and agricultural fires, with wildfires and
prescribed fires accounting for most of the fire-related primary PM2.5 emissions nationally (U.S.
EPA, 2019a, section 2.3.1.1). Other lesser-contributing anthropogenic sources of PM2.5
emissions nationally include stationary fuel combustion and agriculture sources (e.g., agricultural
tilling).
2 These sections do not provide a comprehensive list of all sources, nor does it provide estimates of emission rates or
emission factors for all source categories. Individual subsectors of source types were aggregated up to a sector
level as used in Figure 2-2 and Figure 2-4. More information about the sectors and subsectors can be found as a
part of the 2014 NEI available from https://www.epa.gov/sites/production/files/2018-
07/documents/nei2014v2 tsd 05iul2018.pdf.
2-4

-------
Mobile Sources
Miscellaneous
6%
Agriculture
18%
Industrial
Processes
Stationary Fuel ,
Combustion '
14%
Fires
32%
Figure 2-2. Percent contribution of PM2.5 emissions by national source sectors. (Source:
2014 NEI)
The relative contributions of specific sources to annual emissions of primary PM2.5 can
vary from location to location, with a notable difference in contributions of sources of PM2.5
emissions in urban areas compared to national emissions. For example, the ISA illustrates this
variation of primary PM2.5 emissions with data from five urban counties in the U.S. (U.S. EPA,
2019a, Figure 2-3).3 Across the majority of these urban areas, the largest PM2.5-emitting sectors
are mobile sources and fuel combustion. This is in contrast to fires, which account for the largest
fraction of primary emissions nationally but make much smaller contributions in many urban
counties (U.S. EPA, 2019a, section 2.3.1.2, Figure 2-3). While primary PM2.5 from mobile
sources are a dominant contributor in some urban areas, accounting for an estimated 13 to 30%
of the total primary PM2.5 emissions, mobile sources contribute only about 7% to total primary
PM2.5 emissions nationally as shown in Figure 2-2.
Another way to look at the emissions data shown in Figure 2-2 is by county. Figure 2-4
presents county-based total PM2.5 emissions divided by the area of the county to normalize for
3 The five counties included in the ISA analysis include Queens County, NY, Philadelphia County, PA, Los Angeles
County, CA, Sacramento County, CA, and Maricopa County (Phoenix), AZ (U.S. EPA, 2019a, section 2.3.1.2).
2-5

-------
differences in county size. This "emissions density" map highlights regions of the country with
the strongest emitting sectors for PM2.5.
.sTj:
?s
v
'¦ 9


Tons Per Sq Mi
3.3711-156.7583
1.4907-2.1136
0.9038-1.4669
0-0.8744
Figure 2-3. 2014 NEI PM2.5 Emissions Density Map, tons per square mile
2.1.1.2 Sources Contributing to Primary PM10 Emissions
Although the NEI does not estimate emissions of PM10-2.5 specifically, estimates of PM10
emissions can provide insight into sources of coarse particles. Thus, the discussion below
focuses on PM10 emissions. The relative contributions of key sources to national PM10 emissions,
based on the 2014 NEI, are shown in Figure 2-4. Total PM10 emissions are estimated to be about
13 million tons. National emissions of PM10 are dominated by dust and agriculture, contributing
a combined 75% of the total emissions. Current NEI estimates of dust emissions across the U.S.
are based on limited emissions profile and activity information. For a number of reasons,
quantification of dust emissions is highly uncertain. Much like wildfires, dust emissions are
common but intermittent emissions sources. Additionally, the suspension and resuspension of
dust is difficult to quantify. Moreover, some dust particles in the PM10-2.5 size range are also
transported internationally and considered as a part of the background component of PM as
opposed to a primary emission of coarse PM (U.S. EPA, 2019a, section 2.3.3).
As with PM2.5, the relative contributions of particular sources to total PM10 emissions
varies from location to location (e.g., depending on local climate, geography, degree of
urbanization, etc.). However, unlike with PM2.5, the sectors included in Figure 2-4 and found to
be the largest contributors to coarse PM emissions are expected to be among the most important
contributors at both the national and more regional levels, particularly given the sources of the
particles in these source categories (e.g., mineral dust, primary biological aerosols (including
pollen), sea spray). As noted previously, the NEI does not include sources such as pollen, sea
spray, windblown dust, or geogenic sources, though those sources also likely contribute to PM10
2-6

-------
emissions. Figure 2-4 shows the national contributions to PMio emissions from particular source
sectors and Figure 2-5 shows the emissions density map for PMio.
Industrial Processe
4%
Mobile Sources
3%
Miscellaneous
2%
Stationary Fuel
Combustion
5%
Agriculture
28%
Figure 2-4. Percent contribution of PMio emissions by national source sectors. (Source:
2014 NEI)
2-7

-------
Tons Per Sq Mi
12.4883-532.4948
7.9322-12.1370
5.6179-7.8035
3.2825-5.4779
0-3.1472
Figure 2-5. PMio Emissions Density Map, tons per square mile
2.1.1.3 Sources Contributing to Emissions of PM Components and Precursor Gases
Understanding the components of PM is particularly important for providing insight into
which sources contribute to PM mass, as well as for better understanding the health and welfare
effects of particles. Major components of PM2.5 mass include sulfate (SO42"), nitrate (NO3"),
elemental or black carbon (EC or BC), organic carbon (OC), and crustal materials. Some of these
PM components are emitted directly to the air (e.g., EC, BC) while others are formed secondarily
through reactions by gaseous precursors (e.g., sulfate, nitrate). The following sections
specifically discuss the sources that contribute to the specific PM2.5 components, including
particulate carbon (section 2.1.1.3.1) and precursor gases (section 2.1.1.3.2).
2.1.1.3.1 Sources Contributing to Emissions of Particulate Carbon
Of the directly emitted components of PM2.5, emissions of elemental (or black) carbon
and organic carbon often make up the largest percentage of directly emitted PM2.5 mass. Figure
2-6 illustrates the sources that contribute to national emissions of elemental and organic carbon
based on the 2014 NET The top panel of Figure 2-6 shows that fires account for most (i.e., 53%)
of the 1.5 milli on tons of particulate OC emi ssions estimated in the 2014 NEI, while the bottom
panel of Figure 2-6 shows that fires and mobile sources (mostly diesel sources) contribute 80%
of the estimated 431,000 tons of particulate EC in the 2014 NEI.
•r *
&
¦-, -
•- - • r»
{

% - !'Sir!"
_	T- THI *	.w »1%
2-8

-------
Organic Carbon
Miscellaneous
10%
Agriculture
Mobile Sources
Industrial Processes
Stationary Fuel
Combustion
Elemental Carbon
Mobile Sources
41%
Miscellaneous
6%
Agriculture
3%
	
Fires
39%
Industrial Processes
1%
Stationary Fuel
Combustion
9%
Figure 2-6. Percent contribution to organic carbon (top panel) and elemental carbon
(bottom panel) national emissions by source sectors. (Source: 2014 NEI)
2-9

-------
Figure 2-7 shows the emissions density map for elemental carbon. This map illustrates
that the elemental carbon emissions signals are strong in the Southeast U.S. and parts of the West
and Northwest U.S., where fires make substantial contributions to PM2.5. In addition, areas
where diesel off-road and on-road sources are a large part of the emissions mix also stand out
(urban and highway corridors). The OC density map (not shown) shows the highest emissions
density in locations with substantial biomass burning activity, consistent with most of the OC
emissions coming from fires (Figure 2-6).
Figure 2-7. Elemental Carbon Emissions Density Map, tons per square mile
2.1.1.3.2 Sources Contributing to Emissions of Precursor Gases
As discussed further in the ISA (U.S. EPA, 2019a, section 2.3.2.1), secondary PM is
formed in the atmosphere by photochemical oxidation reactions of both inorganic and organic
gas-phase precursors. Precursor gases include SO2, NOx, and volatile organic compound (VOC)
gases of anthropogenic or natural origin (U.S. EPA, 2019a, section 2.3.2.1). Anthropogenic SO2
and NOx are the predominant precursor gases in the formation of secondary PM2.5, and ammonia
also plays an important role in the formation of nitrate PM by neutralizing sulfuric acid and nitric
acid. In addition, atmospheric oxidation of VOCs, both anthropogenic and biogenic, is an
important source of organic aerosols, particularly in summer. The semi-volatile and non-volatile
products of VOC oxidation reactions can condense onto existing particles or can form new
particles (U.S. EPA, 2009, section 3.3.2; U.S. EPA, 2019a, section 2.3.2).
Emissions of each of the precursor gases noted above are estimated in the NEI and have
unique source signatures at the national level. Figure 2-8 illustrates the source contributions at
the national level for these PM2.5 precursor gases. As shown in Panel A in Figure 2-8, stationary
fuel combustion sources contribute nearly 80% of the estimated total of 4.8 million tons of
national SO2 national emissions. Within this source category, nearly all of the SO2 emitted to the
0.1428-0.2452
0.0585-0.0947
Tons Per Sq Mi
0-0.0562
2-10

-------
atmosphere comes from electricity generating units, or EGUs. NOx emissions, shown in panel B,
are emitted by a range of combustion sources, including mobile sources (58%) and stationary
fuel combustion sources (24%). In the 2014 NEI, there is an estimated total of 14.4 million tons
of NOx emitted. Of the total estimate of 3.6 million tons of ammonia (NH3) emissions shown in
panel C of Figure 2-8, NH3 emissions are dominated by the agriculture source categories. In
these categories, NH3 is predominantly emitted by livestock waste from animal husbandry
operations (55%) and fertilizer application (25%). In urban areas, on-road mobile sources may
also contribute significantly to NH3 emissions (U.S. EPA, 2019a, Figure 2-3; Sun et al., 2014).
Of the estimated 17 million tons of VOC emissions from anthropogenic sources, fires (26%) and
mobile sources (24%) are the largest contributors to national VOC emissions, along with
industrial processes (23%), as shown in panel D.
(A)S02
Industrial Processes
12%
. scellaneous
Mobile Sources
Stationary Fuel
Combustion
(B)NOx
Solvents
Agriculture r -1
| 0%
0% Biogenics
/ 0% |
I
Stationary Fuel
Combustion
Mobile Sources
58%
Industrial Processes
9%
(C) nh3
(D) Anthropogenic VOCs
Industrial Processes
Stationary Fuel	-jo/0
Combustion 	'— -7
3% , —
M seelaneas
Biogenics
Agriculture
Agriculture
Viscellaneous

Stationary Fuel
'•OHlbiJsllOf
Mobile Sources
Industrial Processes
23%
Figure 2-8. Percent contribution to sulfur dioxide (panel A), oxides of nitrogen (panel B),
ammonia (panel C), and anthropogenic volatile organic compounds (panel D) national
emissions by source sectors. (Source: 2014 NEI)
2-11

-------
Figure 2-9 to Figure 2-12 below show the emissions density maps corresponding to each
of the PM2.5 precursors included in Figure 2-8.
W . jjg
Tons Per Sq Mi
0.9877-1,017.0548
0.2074-0.9050
0.0776-0.1962
0.0354-0.0746
0-0.0337
Figure 2-9. SO2 Emissions Density Map, tons per square mile
f
&

r ' jT
i
.J.;,--
Figure 2-10. NOx Emissions Density Map, tons per square mile
Tons Per Sq Mi
9.1051-1,045.0268
4.5608-8.6456
2.7746-4.4274
1.6223-2.7050
0-1.5758
2-12

-------

r*
& »

l£j
,. 4
«*ii;
?*• ¦ w
js
.r\_ rsr59P5Tfc
' *£< *><•
' TV
Tons Per Sq Mi
2.0750-138.1133
1.2897-2.0271
0.7747-1.2547
0.4020-0.7569
0-0.3784
Figure 2-11. NHs Emissions Density Map, tons per square mile
Tons Per Sq Mi
35.7388-1,020.5877
24.0415-35.1864
15.0113-23.5361
8.9294-14.6469
0-8.6329
Figure 2-12. Anthropogenic (including wildfires) VOC Emissions Density Map, tons per
square mile
2.1.1.3.3 Uncertainty in Emission Estimates
Accuracy in an emissions inventory reflects the extent to which the inventory represents
the actual emissions that occurred. Anthropogenic emissions of air pollutants result from a
variety of sources such as power plants, industrial sources, motor vehicles and agriculture. The
emissions from any individual source typically vary in both time and space. It is not practically
possible to monitor each of the emission sources individually and, therefore, emission
inventories necessarily contain assumptions, interpolation and extrapolation from a limited set of
sample data.
The NEI process is based on a "bottom up" approach to developing emission estimates.
This means that a combination of activity and an appropriate emissions factor is used to estimate
emissions for all processes. For the thousands of sources that make up the NEI, there is
uncertainty in one or both of these factors. For some sources, such as EGUs, direct emission
measurements enable the emission factors to be more certain than for sources without such direct
measurements. For example, emission factors for residential wood combustion are taken from
information available in the literature, regardless of its pedigree and direct applicability to the
¦an
V*
%
£
'¦ \ ..
¦t . . y
Tv>
•lit	'
v.
2-13

-------
source in question. Many of these issues related to the analysis of uncertainty in the NEI are
discussed by Day et al., 2019).
It is not clear how uncertainties in emission estimates affect air quality modeling, as there
are no numerical empirical uncertainty estimates available for the NEI. However, by comparing
modeled concentrations to ambient measurements, overall uncertainty in model outputs can be
characterized. Some of this uncertainty in model outputs is likely due to uncertainty in emission
estimates.
2.2 AMBIENT PM MONITORING METHODS AND NETWORKS
To promote uniform enforcement of the air quality standards set forth under the CAA and
to achieve the degree of public health and welfare protection intended for the NAAQS, the EPA
established PM Federal Reference Methods (FRMs)4 for both PMio and PM2.5 (40 CFR
Appendix J and L to Part 50) and performance requirements for approval of Federal Equivalent
Methods (FEMs) (40 CFR Part 53). Amended following the 2006 and 2012 PM NAAQS
reviews, the current PM monitoring network relies on FRMs and automated continuous FEMs, in
part to support changes necessary for implementation of the revised PM standards. The
requirements for measuring ambient air quality and reporting ambient air quality data and related
information are the basis for 40 CFR Appendices A through E to Part 58.
The EPA and its partners at state, local, and tribal monitoring agencies manage and
operate the nation's ambient air monitoring networks. The EPA provides minimum monitoring
requirements for criteria pollutants and related monitoring (e.g., the Chemical Speciation
Network (CSN)), including identification of an FRM for criteria pollutants and guidance
documents to support implementation and operation of the networks. Monitoring agencies carry
out and perform ambient air monitoring in accordance with the EPA's requirements and
guidance as well as often meeting their own state monitoring needs that may go beyond the
minimum federal requirements. Data from the ambient air monitoring networks are available
from two national databases: 1) the Air Quality System (AQS) database, which is the EPA's
long-term repository of ambient air monitoring data and 2) the AirNow database, which provides
near real-time data used in public reporting and forecasting of the Air Quality Index (AQI).5
4	FRMs provide the methodological basis for comparison to the NAAQS and also serve as the "gold-standard" for
the comparison of other methods being reviewed for potential approval as equivalent methods. The EPA keeps a
complete list of designated reference and equivalent methods available on its Ambient Monitoring Technology
Information Center (AMTIC) website (https://www.epa.gov/amtic/air-monitoring-methods-criteria-pollutants').
5	The AQI translates air quality data into numbers and colors to help people understand when to take action to
protect their health against ambient air concentrations of criteria pollutants.
2-14

-------
The EPA and monitoring agencies manage and operate robust national networks for both
PMio and PM2.5, as these are the two measurement programs directly supporting the PM
NAAQS. PM10 measurements are based on gravimetric mass, while PM2.5 measurements include
gravimetric mass and chemical speciation. A smaller network of stations is operating and
reporting data for PM10-2.5 gravimetric mass and a few monitors are operated to support special
projects, including pilot studies, for continuous speciation and particle count data. Monitoring
networks and additional monitoring efforts for each of the various PM size fractions and for PM
composition are discussed below.6 Section 2.2.1 provides information on monitoring for total
suspended particulates (TSP), section 2.2.2 provides information on monitoring for PM10, section
2.2.3 provides information on monitoring PM2.5, section 2.2.4 provides information on
monitoring for PM10-2.5, and section 2.2.5 provides information on additional PM metrics. All
sampler and monitor counts provided in these sections are based on data submitted to the EPA
for calendar year 2018, unless otherwise noted. Figure 2-13 below illustrates the changes in PM
monitoring stations reporting to the EPA's AQS database by size fraction since 1970.
PM Monitoring Stations Reporting to
EPA's AQS database by Size Fraction 1970 - 2018
4DDD
3 ODD
20DD
1Q0D
0
r-.rKFlSiocciooocococncncncricnooooo-H^H
TSP Pm 10 Total O-lOum Stp	PmZ.5-Local Conditions PfvllP-2.5
Figure 2-13. PM Monitoring stations reporting to EPA's AQS database by PM size
fraction, 1970-2018.
2.2.1 Total Suspended Particulates (TSP) Sampling
The EPA first established NAAQS for PM in 1971, based on the original air quality
criteria document (DHEW, 1969). The reference method specified for determining attainment of
6 More information on ambient monitoring networks can be found at https://www.epa. gov/amtic
2-15

-------
the original standards was the high-volume sampler, which collects PM up to a nominal size of
25 to 45 [j,m (referred to as total suspended particles or TSP). TSP was replaced by PMio as the
indicator for the PM NAAQS in the 1987 final rule (52 FR 24854, July 1, 1987). TSP sampling
remains in operation at a limited number of locations primarily to provide aerosol collection for
TSP lead (Pb) analysis as well as for instances where a state may continue to have state standards
for TSP. The size of the TSP network peaked in the mid-1970s when over 4,300 TSP samplers
were in operation. As of 2018, there were 164 TSP samplers still in operation as part of the Pb
monitoring program; of these, 41 also report TSP mass.
2.2.2 PMio Monitoring
To support the 1987 PMio NAAQS, the EPA and its state and local partners implemented
the first size-selective PM monitoring network in 1990 with the establishment of a PMio network
consisting of mainly high-volume samplers. The network design criteria emphasize monitoring at
middle7 and neighborhood8 scales to effectively characterize the emissions from both mobile and
7	For PMio, middle-scale is defined as follows: Much of the short-term public exposure to PMio is on this scale and
on the neighborhood scale. People moving through downtown areas or living near major roadways or stationary
sources, may encounter particulate pollution that would be adequately characterized by measurements of this
spatial scale. Middle scale PMio measurements can be appropriate for the evaluation of possible short-term
exposure public health effects. In many situations, monitoring sites that are representative of micro-scale or
middle-scale impacts are not unique and are representative of many similar situations. This can occur along traffic
corridors or other locations in a residential district. In this case, one location is representative of a neighborhood
of small scale sites and is appropriate for evaluation of long-term or chronic effects. This scale also includes the
characteristic concentrations for other areas with dimensions of a few hundred meters such as the parking lot and
feeder streets associated with shopping centers, stadia, and office buildings. In the case of PMio, unpaved or
seldomly swept parking lots associated with these sources could be an important source in addition to the
vehicular emissions themselves.
8	For PMio, neighborhood scale is defined as follows: Measurements in this category represent conditions
throughout some reasonably homogeneous urban sub-region with dimensions of a few kilometers and of
generally more regular shape than the middle scale. Homogeneity refers to the particulate matter concentrations,
as well as the land use and land surface characteristics. In some cases, a location carefully chosen to provide
neighborhood scale data would represent not only the immediate neighborhood but also neighborhoods of the
same type in other parts of the city. Neighborhood scale PMio sites provide information about trends and
compliance with standards because they often represent conditions in areas where people commonly live and
work for extended periods. Neighborhood scale data could provide valuable information for developing, testing,
and revising models that describe the larger-scale concentration patterns, especially those models relying on
spatially smoothed emission fields for inputs. The neighborhood scale measurements could also be used for
neighborhood comparisons within or between cities.
2-16

-------
stationary sources, although not ruling out microscale9 monitoring in some instances (40 CFR
Part 58 Appendix D, 4.6 (b)). The PMio monitoring network peaked in size in 1995 with 1,665
stations reporting data.
In 2018, there were 714 PMio stations in operation to support comparison of the PMio
data to the NAAQS, trends, and reporting and forecasting of the AQI. Though the PMio network
is relatively stable, monitoring agencies may continue divesting of some of the PMio monitoring
stations where concentration levels are low relative to the NAAQS.
While the PMio network is national in scope, there are areas of the west, such as
California and Arizona, with substantially higher PMio station density than the rest of the
country. In the PMio mass network, 365 of the stations operate automated continuous mass
monitors approved as FEMs and 391 operate FRMs. About 40 of the PMio stations have
collocation with both continuous FEMs and FRMs. About two thirds of the PMio stations with
FRMs operate on a sample frequency of one in every sixth day, with about 70 operating every
third day and 60 operating every day.
2.2.3 PM2.5 Monitoring
To support the 1997 PM2.5 NAAQS, the first PM standard with PM2.5 as an indicator, the
EPA and states implemented a PM2.5 network consisting of ambient air monitoring sites with
mass and/or chemical speciation measurements. Network operation began in 1999 with nearly
1,000 monitoring stations operating FRMs to measure fine particle mass. The PM2.5 monitoring
program remains one of the major ambient air monitoring programs operated across the country.
For most urban locations PM2.5 monitors are sited at the neighborhood scale,10 where
PM2.5 concentrations are reasonably homogeneous throughout an entire urban sub-region. In each
9 For PMio, microscale is defined as follows: This scale would typify areas such as downtown street canyons, traffic
corridors, and fence line stationary source monitoring locations where the general public could be exposed to
maximum PMio concentrations. Microscale particulate matter sites should be located near inhabited buildings or
locations where the general public can be expected to be exposed to the concentration measured. Emissions from
stationary sources such as primary and secondary smelters, power plants, and other large industrial processes
may, under certain plume conditions, likewise result in high ground level concentrations at the microscale. In the
latter case, the microscale would represent an area impacted by the plume with dimensions extending up to
approximately 100 meters. Data collected at microscale sites provide information for evaluating and developing
hot spot control measures.
i° for PM2 5, neighborhood scale is defined as follows: Measurements in this category would represent conditions
throughout some reasonably homogeneous urban sub-region with dimensions of a few kilometers and of
generally more regular shape than the middle scale. Homogeneity refers to the particulate matter concentrations,
as well as the land use and land surface characteristics. Much of the PM2 5 exposures are expected to be associated
with this scale of measurement. In some cases, a location carefully chosen to provide neighborhood scale data
would represent the immediate neighborhood as well as neighborhoods of the same type in other parts of the city.
PM2.5 sites of this kind provide good information about trends and compliance with standards because they often
represent conditions in areas where people commonly live and work for periods comparable to those specified in
the NAAQS. In general, most PM2 5 monitoring in urban areas should have this scale.
2-17

-------
CBS A with a monitoring requirement, at least one PM2.5 monitoring station representing area-
wide air quality is to be sited in an area of expected maximum concentration. Sites that represent
relatively unique microscale, localized hot-spot, or unique middle scale impact sites are only
eligible for comparison to the 24-hour PM2.5 NAAQS.
There are three main components of the current PM2.5 monitoring program: FRMs, PM2.5
continuous mass monitors, and CSN samplers. The FRMs are primarily used for comparison to
the NAAQS, but also serve other important purposes such as developing trends and evaluating
the performance of PM2.5 continuous mass monitors. PM2.5 continuous mass monitors are
automated methods primarily used to support forecasting and reporting of the AQI, but are also
used for comparison to the NAAQS where approved as FEMs. The CSN and related Interagency
Monitoring of Protected Visual Environments (IMPROVE) network are used to provide
chemical composition of the aerosol which serve a variety of objectives. This section provides an
overview of each of these components of the PM2.5 monitoring program and of recent changes to
PM2.5 monitoring requirements.
2.2.3.1 Federal Reference Method and Continuous Monitors
As noted above, the PM2.5 monitoring network began operation in 1999 with nearly 1,000
monitoring stations operating FRMs. The PM2.5 FRM network peaked in operation in 2001 with
over 1,150 monitoring stations. In the PM2.5 network, in 2018 there were 624 FRM filter-based
samplers that provide 24-hour PM2.5 mass concentration data. Of these operating FRMs, 70 are
providing daily PM2.5 data, 422 every third day, and 132 every sixth day.
As of 2018, there are 940 continuous PM2.5 mass monitors that provide hourly data on a
near real-time basis reporting across the country. A total of 579 of the PM2.5 continuous monitors
are FEMs and therefore used both for comparison with the NAAQS and to report the AQI.
Another 361 monitors not approved as FEMs are operated primarily to report the AQI. These
legacy PM2.5 continuous monitors were largely purchased prior to the availability of PM2.5
continuous FEMs.
The first method approved as a continuous PM2.5 FEM was the Met One BAM 1020. This
method, approved in 2008, accounts for just over 50% of the operating PM2.5 continuous FEMs
in the country. The EPA has approved a total of 11 PM2.5 continuous methods as FEMs. Other
methods approved as continuous PM2.5 FEMs include beta attenuation from multiple instrument
manufacturers; optical methods such as the GRIMM and Teledyne T640; and methods
employing the Tapered Element Oscillating Microbalance (TEOM) with a Filter Dynamic
Measurement System (FDMS) manufactured by Thermo Fisher Scientific.
The quality of the data generated by PM2.5 FRMs and automated FEMs were analyzed for
years 2016-2018. Data quality terms for measurement uncertainty regularly assessed in the
2-18

-------
PM2.5 monitoring program include precision and bias. Precision is calculated by comparing data
from collocated methods of the same make and model operated by the same monitoring
organization. Bias is calculated by comparing data from routinely operated FRMs or automated
FEMs by the monitoring organization and comparing that to data from reference method audit
samplers temporarily collocated and operated independently from the staff in the monitoring
organization. Goals for measurement uncertainty are defined in Appendix A to 40 CFR Part 58.
They state "Measurement Uncertainty for Automated and Manual PM2.5 Methods. The goal for
acceptable measurement uncertainty is defined for precision as an upper 90 percent confidence
limit for the coefficient of variation (CV) of 10 percent and ±10 percent for total bias." The most
recent three-year average estimate of national aggregate PM2.5 FRM precision is 8.2% and bias is
-4.7%.
Automated PM2.5 FEMs include a wide variety of approved methods which can have
different measurement principles. Data aggregated across all automated FEMs result in a
collocated precision of 18.6% and a bias as compared to the reference method audit program of
+7.6%). When evaluating automated FEMs as individual methods, only two of the seven
methods with available collocated precision data meet the measurement uncertainty goal;
however, as explained in the Notice of Proposed Rulemaking, January 17, 200611 when
considering a requirement for approval of candidate FEMs: "Statistical analyses based on the
DQO model show that the precision of a candidate method is not, statistically, very important to
annual concentration averages used for NAAQS attainment decisions, but would be important
for a daily standard." When evaluating automated FEMs as individual methods for bias, eight of
ten methods with data available to calculate a performance evaluation bias meet this goal. In
summary, PM2.5 automated FEMs tend to have higher collocated precision than FRMs and tend
to have a positive bias relative to both State and local operated FRMs as well as performance
evaluation audit FRMs.
2.2.3.2 Chemical Speciation and IMPROVE Networks
Due to the complex nature of fine particles, the EPA and states implemented the CSN to
better understand the components of fine particle mass at selected locations across the country.
The CSN was first piloted at 13 sites in 2000, and after the pilot phase, the program continued
with deployment of the Speciation Trends Network (STN) later that year. The CSN ultimately
grew to 54 trends sites and peaked in operation in 2005 with 252 stations: the 54 trends stations
and nearly 200 supplemental stations. The original CSN program had multiple sampler
configurations including the Thermo Andersen RAAS, Met One SASS/SuperSASS, and URG
11 https://www.govinfo.gov/content/pkg/FR-2006-01-17/pdf/06-177.pdf
2-19

-------
MASS. During the 2000s, the EPA and states worked to align the network to one common
sampler for elements and ions, which was the Met One SASS/SuperSASS. In 2005, the CASAC
provided recommendations to the EPA for making changes to the CSN. These changes were
intended to improve data comparability with the rural IMPROVE carbon concentration data. To
accomplish this, the EPA replaced the existing carbon channel sampling and analysis methods
with a new modified IMPROVE version III module C sampler, the URG 3000N. Implementation
of the new carbon sampler and analysis was broken into three phases starting in May 2007
through October 2009.
In the 2018 PM2.5 CSN, long-term measurements are made at about 76 largely urban
locations comprised of either the STN or the National Core (NCore) network.12 NCore is a
multipollutant network measuring particles, gases, and basic meteorology that has been in formal
operation since January 1, 2011. Particle measurements made at NCore include PM2.5 filter-based
mass, which is largely the FRM, except in some rural locations that utilize the IMPROVE
program PM2.5 mass filter-based measurement; PM2.5 speciation using either the CSN program or
IMPROVE program; and PM10-2.5 mass utilizing an FRM, FEM or IMPROVE for some of the
rural locations. As of 2018, the NCore network includes a total of 78 stations of which 63 are in
urban or suburban stations designed to provide representative population exposure and another
15 rural stations designed to provide background and transport information. The NCore network
is deployed in all 50 States, DC, and Puerto Rico with at least one station in each state and two or
more stations in larger population states (California, Florida, Illinois, Michigan, New York,
North Carolina, Ohio, Pennsylvania, and Texas).
Both the STN and NCore networks are intended to remain in operation indefinitely. The
CSN measurements at NCore and STN stations operate every third day. Another approximately
72 CSN stations, known as supplemental sites, are intended to be potentially less permanent
locations used to support State Implementation Plan (SIP) development and other monitoring
objectives.13 Supplemental CSN stations typically operate every sixth day. In January 2015, 38
supplemental CSN stations that are largely located in the eastern half of the country stopped
operations to ensure a sustainable CSN network moving forward.14
12	In most cases where a city has an STN station, it is located at the same site as the NCore station. In a few cases, a
city may have an STN station located at a different location than the NCore station.
13	See https://www3.epa.gov/ttn/amtic/speciepg.html for more information on the PM2 5 speciation monitoring
program.
14	Based on assessments of the CSN network and IMPROVE protocol sites, monitoring resources were redistributed
to focus on new or high priorities. More information on the CSN and IMPROVE protocol assessments is
available at https://www.sdas.battelle.org/CSNAssessment/html/Default.html.
2-20

-------
Specific components of fine particles are also measured through the IMPROVE
monitoring program15 which supports regional haze characterization and tracks changes in
visibility in Class I areas as well as many other rural and some urban areas. As of 2018, the
IMPROVE network includes 110 monitoring locations that are part of the base network
supporting regional haze and another 46 locations operated as IMPROVE protocol sites where a
monitoring agency has requested participation in the program. These IMPROVE protocol sites
operate the same way as the IMPROVE program, but they may serve several monitoring
objectives (i.e., the same objectives as the CSN) and are not explicitly tied to the Regional Haze
Program. Samplers at IMPROVE stations operate every third day. In January 2016, eight
IMPROVE protocol stations stopped operating to ensure a sustainable IMPROVE program
moving forward. Details on the process and outcomes of the CSN supplemental and IMPROVE
protocol assessments used to identify sites that would no longer be funded are available on an
interactive website.16 Together, the CSN and IMPROVE data provide chemical species
information for fine particles that are critical for use in health and epidemiologic studies to help
inform reviews of the primary PM NAAQS and can be used to better understand visibility
through calculation of light extinction using the IMPROVE algorithm17 to support reviews of the
secondary PM NAAQS.
The quality of the data generated by the PM2.5 speciation networks (CSN and IMPROVE)
is assessed regularly, using a variety of metrics. Overall network precision, including
uncertainties associated with both field operations and laboratory analyses, is assessed using the
subset of sites with collocated samplers. Fractional uncertainty is one metric that both speciation
networks regularly calculates using collocated data pairs above the MDL and reflects the overall
percent uncertainty for the measurements. For CSN data collected between November 2015 and
December 2016, the fractional uncertainties range from 6.6% for sulfate to 31.4% for chlorine.18
15	Recognizing the importance of visual air quality, Congress included legislation in the 1977 Clean Air Act to
prevent future and remedy existing visibility impairment in Class I areas. To aid the implementation of this
legislation, the IMPROVE program was initiated in 1985 and substantially expanded in 2000-2003. This program
implemented an extensive long-term monitoring program to establish the current visibility conditions, track
changes in visibility and determine causal mechanism for the visibility impairment in the National Parks and
Wilderness Areas. For more information, see httos://www3.epa. gov/ttn/amtic/visdata.html.
16	See the Chemical Speciation Network Assessment Interactive Website at:
https://www.sdas.battelle.org/CSNAssessment/html/Default.html.
17	The IMPROVE algorithm is an equation to estimate light extinction based on the measured concentration of
several PM components and is used to track visibility progress in the Regional Haze Rule. More information
about the IMPROVE algorithm is at available at: http://vista.cira.colostate.edu/Improve/the-improve-algorithm.
18	https://airqualitv.ucdavis.edu/sites/g/files/dgvnskl671/files/inline-
files/CSN AnnualReport 2016Data 03.06.2019 FINAL APPROVED.pdf
2-21

-------
For IMPROVE data collected in 2016 and 2017, the fractional uncertainties range from 2% for
sulfur and sulfate to 27% for phosphorous.19 In general, uncertainties are higher for species with
concentrations near the detection limit. Bias for the speciation networks can be assessed using
reports from interlaboratory comparisons.20
2.2.3.3 Recent Changes to PM2.5 Monitoring Requirements
Key changes made to the EPA's monitoring requirements as a result of the 2012 PM
NAAQS review included the addition of PM2.5 monitoring at near-road locations in core-based
statistical areas (CBSAs) over 1 million in population; the clarification of terms used in siting of
PM2.5 monitors and their applicability to the NAAQS; and the provision of flexibility on data
uses to monitoring agencies where their PM2.5 continuous monitors are not providing data that
meets the performance criteria used to approve the continuous method as an FEM. The addition
of PM2.5 monitoring at near-road locations was phased in from 2015 to 2017. On January 1,
2015, 22 CBSAs with a population of 2.5 million or more were required to have a PM2.5 FRM or
FEM operating at a near-road monitoring station. On January 1, 2017, 30 CBSAs with a
population between 1 million and 2.5 million were required to have a PM2.5 FRM or FEM
operating are a near-road monitoring station.
The terms clarified as a part of the 2012 rulemaking ensure consistency with all other
NAAQS and long-standing definitions used by the EPA (78 FR 3234, January 15, 2013). The
flexibility provided to monitoring agencies ensures that the incentives of utilizing PM2.5
continuous monitors (e.g., efficiencies in operation and availability of hourly data in near-real
time) are realized without having potentially poor performing data be used in situations where
the data is not applicable to the NAAQS (78 FR 3241, January 15, 2013).
2.2.4 PM10-2.5 Monitoring
In the 2006 PM NAAQS review, the EPA promulgated a new FRM for the measurement
of PM10-2.5 mass in ambient air. Although the standard for coarse particles uses a PM10 indicator,
a new FRM for PM10-2.5 mass was developed to provide a basis for approving FEMs and to
promote the gathering of scientific data to support future reviews of the PM NAAQS. The
PM10-2.5 FRM (or approved FEMs, where available) was implemented at required NCore stations
by January 1, 2011. In addition to NCore, there are other collocated PM10 and PM2.5 low-volume
FRMs operating across the country that are essentially providing the PM10-2.5 FRM measurement
by the difference method.
19	http://vista.cira.colostate.edU/improve/wp-content/uploads/2019/l 1/IMPROVE OAReport 11.15.2019.pdf
20	https://www3.epa.gov/ttn/amtic/pmspec.html
2-22

-------
PMio—2.5 measurements are currently performed across the country at NCore stations,
IMPROVE monitoring stations, and at a few additional locations where state or local agencies
choose to operate a PM10-2.5 method. For urban NCore stations and other State and Local Air
Monitoring Stations (SLAMS) the method employed is either a PM10-2.5 FRM, which is
performed using a low-volume PM10 FRM collocated with a low volume PM2.5 FRM of the same
make and model, or FEMs for PM10-2.5, including filter-based dichotomous methods and
continuous methods of which several makes and models are approved. Filter-based PM10-2.5
measurements at NCore (i.e., the FRM or dichotomous filter-based FEM) operate every third
day, while continuous methods have data available every hour of every day. PM10-2.5 filter-based
methods at other SLAMS typically operate every third or sixth day. For IMPROVE, which is
largely a rural network, PM10-2.5 measurements are made with two sample channels; one each for
PM10 and PM2.5. All IMPROVE program samplers operate every third day. All together there
were 279 stations in 2018 where PM10-2.5 data were being reported to the AQS database.
There is no operating chemical speciation network for characterizing the specific
components of coarse particles. In 2015, Washington University at St. Louis, under contract to
the U.S. EPA, reported on a coarse particle speciation pilot study with several objectives aimed
at addressing this issue, such as evaluating a coarse particle species analyte list and evaluating
sampling and analytical methods (U.S. EPA, 2015). The coarse particle speciation pilot study
provides useful information for any organization wishing to pursue coarse particle speciation.
2.2.5 Additional PM Measurements and Metrics
There are additional PM measurements and metrics made at a much smaller number of
stations. These measurements may be associated with special projects or are complementary
measurements to other networks where the monitoring agency has prioritized having the
measurements. None of these measurements are required by regulation. They include PM
measurements such as particle counts, continuous carbon, and continuous sulfate.
The EPA and state and local agencies have also been working together to pilot additional
PM methods at near-road monitoring stations that may be of interest to data users. These
methods include such techniques as particle counters, particle size distribution, and black carbon
by aethalometer. These methods and their rationale for use at near-road monitoring stations are
described in a Technical Assistance Document (TAD) on NO2 near-road monitoring (U.S. EPA,
2012, section 16).
Aethalometer measurements of the concentration of optically absorbing particles have
been submitted to AQS for many years. Data uses include characterizing black carbon and wood
smoke. Ambient air monitoring stations that may have aethalometers include some of the near-
road monitoring stations and National Air Toxics Trends Stations (NATTS). Data from about 72
2-23

-------
monitoring sites across the county are being reported from aethalometers. While aethalometer
data is available at high time resolutions (e.g., 5-minute data), it is typically reported to the AQS
database in 1-hour periods.
Continuous elemental and organic carbon data were monitored at select locations
participating in a pilot of the Sunset EC/OC analyzer as well as a few additional sites that were
already operating before the EPA initiated the pilot study.21 The Sunset EC/OC analyzer
provides high time resolution carbon data, typically every hour, but in some remote locations the
instrument is programmed to run every two hours to ensure collection of enough aerosol. The
data from the Sunset EC/OC analyzer was compared to filter-based carbon methods from the
carbon channel of the CSN program. The Sunset EC/OC analyzer was operated at each of the
study sites for at least three years. Results from this pilot study are available in an EPA report
(U.S. EPA, 2019b). A key finding from the study suggests that when the Sunset instrument was
working well, OC and optical EC were comparable to CSN OC and EC; however, the time and
resources needed to keep a Sunset analyzer operational did not merit replacement of CSN OC
and EC measurements.
As of 2018, continuous sulfate is measured at four remaining monitoring sites, one in
Maine and three in New York State. Several other stations have historical data but are no longer
monitoring continuous sulfate. Discontinuing monitoring efforts for continuous sulfate is likely
an outcome of the significantly lower sulfate concentrations throughout the east where these
methods were operated. The continuous sulfate analyzer provides hourly data and these data can
be readily compared to 24-hour sulfate data which are collected from the ion channel in both the
CSN and IMPROVE programs.
In addition, over the last few years, the EPA has investigated the use of several PM
sensor technologies as one of several areas of research intended to address the next generation of
air measurements. The investigation into air sensors is envisioned to work towards near real-time
or continuous measurement options that are smaller, cheaper, and more portable than traditional
FRM or FEM methods. These sensor devices have the potential to be used in several applications
such as identifying hotspots, informing network design, providing personal exposure monitoring,
supporting risk assessments, and providing background concentration data for permitting. The
EPA has hosted workshops and published several documents and peer-reviewed articles on this
work.22
21	The six sites that participated in the study were Washington, DC; Chicago, IL; St. Louis, MO; Houston, TX; Las
Vegas, NV; and Los Angeles, CA.
22	For more information, see https://www.epa.gov/sciencematters/epas-next-generation-air-measuring-research and
https://www.epa.gOv/air-sensor-toolbox/air-sensor-toolbox-what-epa-doing#pane-l.
2-24

-------
2.3 AMBIENT AIR CONCENTRATIONS
This section summarizes available information on recent ambient PM concentrations.
Section 2.3.1 presents trends in emissions of PM and precursor gases, while section 2.3.2
presents trends in monitored ambient concentrations of PM in the U.S. Section 2.3.3 discusses
approaches for predicting ambient PM2.5 by hybrid modeling approaches.
2.3.1 Trends in Emissions of PM and Precursor Gases
Direct emissions of PM have remained relatively unchanged in recent years, while
emissions of some precursor gases have declined substantially.23 As illustrated in Figure 2-14,24
from 1990 to 2014, SO2 emissions have undergone the largest declines while NH3 emissions
have undergone the smallest change. Declining SO2 emissions during this time period are
primarily a result of reductions at stationary sources such as EGUs, with substantial reductions
also from mobile sources (U.S. EPA, 2019a, section 2.3.2.1). In more recent years (i.e., 2002 to
2014), emissions of SO2 and NOx have undergone the largest declines, while direct PM2.5 and
NH3 emissions have undergone the smallest changes, as shown in Table 2-1. Regional trends in
emissions can differ from the national trends illustrated in Figure 2-14 and Table 2-1.25 For
example, Hand et al. (2012) studied reductions in EGU-related annual SO2 emissions during the
2001-2010 period and found that while SO2 emissions decreased throughout the U.S. by an
average of 6.2% per year, the amount of change varied across the U.S. with the largest percent
reductions in the western U.S. at 20.1% per year.
23	More information on these trends, including details on methods and explanations on the noted changes over time
is available at https://gispub.epa.gov/neireport/2014/.
24	Emission trends in Figure 2-14 do not include wildfire emissions.
25	State-specific emission trends data for 1990 to 2014 can be found at: https://www.epa.gov/air-emissions-
inventories/air-pollutant-emissions-trends-data.
2-25

-------
35,000
30,000
o
? 25,000
§ 20,000
i—
£ 15,000
.£2 10,000
5,000
0
# K#	K# o# <#0# o#
Year
^^NH3 ^^NOx	PM2.5 —•— PM10 -•—S02 —^VOCs
Figure 2-14. National emission trends of PM2.5, PM10, and precursor gases from 1990 to
2014.
Table 2-1. Percent Changes in PM and PM precursor emissions in the NEI for the time
periods 1990-2014 and 2002-2014.
Pollutant
Percent Change
in Emissions:
1990 to 2014
Percent Change
in Emissions:
2002 to 2014
Major Sources
nh3
-21%
-10%
Agricultural Sources (Fertilizer and
Livestock Waste), Fires
NOx
-50%
-48%
EGUs, Mobile Sources
S02
-80%
-69%
EGUs, other Stationary Sources
VOCs
-38%
-15%
Solvents, Fires, Mobile Sources
PM2.5
-40%
-4%
Dust, Fires
PM10
-38%
-15%
Dust, Fires
2.3.2 Trends in Monitored Ambient Concentrations
2.3.2.1 National Characterization of PM2.5 Mass
At long-term monitoring sites in the U.S., annual PM2.5 concentrations from 2015 to 2017
averaged 8.0 [j,g/m3 (ranging from 3.0 to 18.2 (J,g/m3) and the 98th percentiles of 24-hour
concentrations averaged 20.9 [j,g/m3 (ranging from 9.2 to 111 (J,g/m3). Figure 2-15 (top panels)
shows that the highest ambient PM2.5 concentrations occur in the west, particularly in California
and the Pacific northwest. Much of the eastern U.S. has lower ambient concentrations, with
annual average concentrations generally at or below 12.0 [j,g/m3 and 98th percentiles of 24-hour
concentrations generally at or below 30 [j,g/m3.
2-26

-------
These concentrations are distinct from design values in part because they include days
with episodic events like wildfires and dust storms which can have very high PM2.5 and/or PM10
concentrations. The EPA's Exceptional Events Rule,26 most recently updated in 2016, describes
the process by which these events can be excluded from the design values used for comparison to
the NAAQS. For the remainder of Chapter 2, episodic events are included in the calculations of
PM concentrations. When design values are discussed in Chapter 2, regionally-concurred
exceptional events (as of July 2019) have been excluded from the analysis.27
26	The final version of the 2016 Exceptional Events Rule can be accessed at
https://www.epa.gov/sites/production/files/2018-10/documents/exceptional events rule revisions 2060-
as02 final.pdf.
27	Regionally-concurred exceptional events are unusual or naturally-occurring events such as wildfires or high wind
dust events that have 1) resulted in PM2 5 concentrations above the level of the NAAQS, 2) been submitted by
tribal, state or local air agencies under the EPA's Exceptional Events Rule to their respective EPA Region, and 3)
received concurrence.
2-27

-------
2015-2017 Annual Average
•/•	/^jyy^hwi a
yj			fat> ¦.¦;(ij	: jv '.7 \ •
>V . /••* •; | , ." >• '* >; r^^fr	%\ . •" • * I V ." * /%3
'•¦ \ u_ • i * *y t. .£mT*	* \ /-•_ i • I • * *, \
} •,' nM -rv-—_i- _ * ' *~ «©	I •,• "JA —j~r——rJ » * •» •« . * '• *» i'h i
V", / ••	V-, i •• r^-3tw
• . .. r-v—? • V^j• I • j. • J	J,--*—4 » p)' * t. • 7
¦t? . I Vn .• ' - h	«? . .• J3r^ ~n-\
2015-2017 98th Percentile
A
P


<8
•	8-10 V3
•	10-12 >5
•	12-14
•	>14
^ ¦"¦ #'
. (—!• L—<«-*-
25-30
• 30-35
35-40
2000-2017 Annual Average Trend
W
2000-2017 98th Percentile Trend
|v t y* -- ^
—-r-J	~ ! Yj
T if _ * ' .**CtV
*¦ T ^ _
T T —f tt;VV*
f *|V
/
^tVTt T t-
¦	' v,
\	(-, No Trend
Significant	I—J(p>0,15) T^Of
\ ? ~ Reduction	Possible '-, f
^ (pso.05)	A increase ™
Possible	(0.050.15l
~nificant
Reduction
. Possible
(p<0.05) A increase
V Possible	(0.05
-------
Analysis of monthly data indicate distinct peaks in national ambient PM2.5 concentrations
during the summer and the winter (U.S. EPA, 2019a, Figure 2-22). Through 2008, the summer
peaks reflected the highest national average PM2.5 concentrations. These summer peaks in
ambient PM2.5 concentrations were largely a consequence of summertime peaks in SO2
emissions from power plants in the eastern U.S., and subsequent sulfate formation. However,
substantial reductions in SO2 emissions (see above and U.S. EPA, 2019a, sections 2.5.1.1.1 and
2.5.2.2.1) have changed this pattern. Starting in 2009, winter peaks in national average PM2.5
concentrations have been higher than those in the summer (U.S. EPA, 2019a, section 2.5.2.2.1).
This pattern is illustrated by data from 2013 to 2015, when average winter PM2.5 concentrations
were about 11 (J,g/m3, average summer concentrations were about 9 (j,g/m3, and average spring
and fall concentrations were about 7 [j,g/m3 (Chan et al., 2018).
The ambient PM2.5 concentrations in Figure 2-15 reflect the substantial reductions that
have occurred across much of the U.S. over recent years (Figure 2-15, bottom panels and Figure
2-16). From 2000 to 2017, national annual average PM2.5 concentrations have declined from 13.5
[j,g/m3 to 8.0 (j,g/m3, a 41% decrease (Figure 2-16).28 These declines have occurred at both urban
and rural monitoring sites, although urban PM2.5 concentrations remain consistently higher than
those in rural areas (Chan et al., 2018) due to the so-called "urban increment" of PM2.5 from
local sources in an urban area that is additive to the regional and natural background PM2.5
concentrations.
301	
25-
05
^ 20-
National Standard
°~t—1—1	1—1—1	1—1—1	1—1—1	1—1—1—1	1—r
222222222222222222
000000000000000000
000000000011 1111 11
012345678901 234567
Figure 2-16. Seasonally-weighted annual average PM2.5 concentrations in the U.S. from
2000 to 2017 (429 sites). (Note: The white line indicates the mean concentration while the
gray shading denotes the 10th and 90th percentile concentrations.)
28 See https://www.epa.gov/air-trends/particulate-matter-pm25 -trends and https://www.epa.gov/air-
trends/particulate-matter-pm25-trends#pmnat for more information.
2-29

-------
Analyses at individual monitoring sites indicate that declines in ambient PM2.5
concentrations have been most consistent across the eastern U.S. and in parts of coastal
California, where both annual average and 98th percentiles of 24-hour concentrations have
declined significantly (Figure 2-15, bottom panels). In contrast, trends in ambient PM2.5
concentrations have been less consistent over much of the western U.S., with no significant
changes since 2000 observed at some sites in the Pacific northwest, the northern Rockies and
plains, and the southwest, particularly for 98th percentiles of 24-hour concentrations (Figure 2-
15, bottom panels). Trends in annual average PM2.5 concentrations have been highly correlated
with trends in 98th percentiles of 24-hour concentrations at individual sites (Figure 2-17). Such
correlations are highest across the eastern U.S. and in coastal California, and are somewhat
lower, though still generally positive, at sites in the Central and Western U.S. (i.e., outside of
coastal California).
I •
> <*•
••
O <0.0 «
° 0.0-0.5
° 0.5-0.75
• >0.75
Figure 2-17. Pearson's correlation coefficient between annual average and 98th percentile
of 24-hour PM2.5 concentrations from 2000-2017.
2.3.2.2 Characterization of PM2.5 Mass at Finer Spatial and Temporal Scales
2.3.2.2.1 CBSA Maximum Annual Versus Daily Design Values
Analysis of recent air quality indicates that maximum annual and daily PM2.5 design
values within a CBSA are positively correlated with some noticeable regional variability (Figure
2-18). In the Southeast, Northeast, and Industrial Midwest regions, the annual design values are
high relative to the daily design values due in part to the infrequent impacts of episodic events
2-30

-------
like wildfire or dust storms. On the other hand, the Northwest region has very high daily design
values relative to the annual design values. This is due to episodically high PM2.5 concentrations
that affect the region, both from wintertime stagnation events and summer/fall wildfire smoke
events.29 The relatively small population and low emissions in the region result in much lower
PM2.5 concentrations during the other parts of the year not affected by these episodes.
Other (AK, HI)
•	IndustMidwest
•	Northeast
•	Northwest
•	SoCal
•	Southeast
•	Southwest
•	UpperMidwest
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
CBSA Maximum 2015-2017 Annual Design Value (^g rrr3)
Figure 2-18. Scatterplot of CBSA maximum annual versus daily design values (2015-2017)
with the solid black line representing the ratio of daily and annual NAAQS values.
2.3.2.2.2 PM2.5 Near Major Roadways
Because of its longer atmospheric lifetime (U.S. EPA, 2019a, section 2.2), PM2.5 is
expected to exhibit less spatial variability on an urban scale than UFP or PM 10-2.5 (U.S. EPA,
2019a, section 2.5.1.2.1). Analyses in the 2009 ISA for PM indicated that correlations between
PM2.5 monitoring sites up to a distance of 100 km from each other were greater than 0.75 in most
29 Due to the recent time period shown in Figure 2-18, it is likely that some of the annual and daily design values are
affected by potential exceptional events associated with wildfire smoke that have yet to be regionally-concurred
and removed from the design value calculations. The EPA defines exceptional events as unusual or natural-
occurring events that that affect air quality but are not reasonably controllable using techniques that tribal, state,
or local air agencies may implement. This is especially likely for the daily design values in the Northwest region
which experienced frequent wildfire smoke events during the 2015-2017 period.
2-31

-------
urban areas. However, more substantial spatial variation has been reported for some urban areas,
due in part to proximity between monitors and emissions sources (U.S. EPA, 2019a, section
2.5.1.2.1). The recent deployment of PM2.5 monitors near major roads in large urban areas
provides some insight into this spatial variation.
As discussed above, in the last review of the PM NAAQS the EPA required monitoring
of PM2.5, along with NO2 and CO, near major roads in CBSAs with populations greater than 1
million. PM2.5 monitoring was required to start for the largest CBSAs at the beginning of 2015,
and several years of data are now available for analysis at these sites. DeWinter et al. (2018)
analyzed these data and found that the average near-road increment (difference between near-
road PM2.5 concentrations and the concentrations at other sites in the same CBSA) was 1.2 [j,g/m3
for 2014 to 2015. The near-road increment has a diurnal cycle, with a peak during the morning
rush hour (Figure 2-19). This near-road increment likely is additive to the urban increment of
PM2.5 from local sources in the CBSA including mobile sources on the numerous non-highway
roads that are not monitored by the near-road network.
1.5
¦*->
c
<1>
£

-------
road site while 24% measured the highest 24-hour design value at the near-road site (Table 2-2).
Of the CBS As with highest annual design values at near-road sites, those design values were, on
average, 0.7 |j,g/m3 higher than at the highest measuring non-near-road sites (range is 0.1 to 2.0
|j,g/m3 higher at near-road sites).
Table 2-2. Daily and annual PM2.5 design values for the near-road sites in major CBSAs
(2015-2017).

Maximum
Maximum
Maximum
Maximum
Non-Near-
Road
Annual
Design
Value

Near-Road
Non-Near-
Near-Road
CBSA Name
Daily
Design
Value
Road Daily
Design
Value
Annual
Design
Value
New York-Newark-Jersey City, NY-NJ-PA
22
23
NA
9.7
Los Angeles-Long Beach-Anaheim, CA
33
39
12.6
12.1
Dallas-Fort Worth-Arlington, TX
18
18
8.7
8.9
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
24
25
9.5
10.6
Atlanta-Sandy Springs-Roswell, GA
23
20
10.5
9.9
Boston-Cambridge-Newton, MA-NH
16
16
7
7.2
San Francisco-Oakland-Hayward, CA
27
30
10.1
10.6
Phoenix-Mesa-Scottsdale, AZ
18
27
7.9
9.6
Riverside-San Bernardino-Ontario, CA
37
39
14.7
13.6
Detroit-Warren-Dearborn, Ml
22
28
8.5
11.2
Seattle-Tacoma-Bellevue, WA
24
34
8.4
8.7
Minneapolis-St. Paul-Bloomington, MN-WI
18
19
8
7.5
St. Louis, MO-IL
19
21
8.7
9.8
Baltimore-Columbia-Towson, MD
20
23
9.1
8.9
Denver-Aurora-Lakewood, CO
23
20
8.5
7.1
Portland-Vancouver-Hillsboro, OR-WA
25
28
7.4
7.4
Kansas City, MO-KS
16
21
7.1
9.0
Indianapolis-Carmel-Anderson, IN
22
22
10.5
10.2
San Jose-Sunnyvale-Santa Clara, CA
28
27
9.4
9.3
Providence-Warwick, RI-MA
20
18
9.1
7.1
Louisville/Jefferson County, KY-IN
21
22
9.4
9.7
New Orleans-Metairie, LA
18
19
8.2
8.5
Hartford-West Hartford-East Hartford, CT
20
18
8.2
6.7
Birmingham-Hoover, AL
22
22
11
10.4
Buffalo-Cheektowaga-Niagara Falls, NY
17
18
7.8
7.6
Rochester, NY
17
16
7
6.5
Although most near-road monitoring sites do not have sufficient data to evaluate long-
term trends in near-road PM2.5 concentrations, analyses of the data at one near-road-like site in
2-33

-------
Elizabeth, NJ,31 show that the annual average increment has generally decreased between 1999
and 2017 from about 2.0 [j,g/m3 to about 1.3 [j,g/m3 (Figure 2-20). The trend in the near-road
increment of elemental carbon at the Elizabeth, NJ site has shown a similar reduction, with
values of-1.0 [j,g/m3 in 2000 decreasing to -0.5 [j,g/m3 in 2017. These data are consistent with
the timing of EPA emission standards for motor vehicles.32 Although long-term data are not
available at other near-road sites, the national scope of the diesel vehicle controls suggests the
near-road environment across the U.S. likely experienced similar decreasing trends in near-road
PM2.5 increments.
3 1
DO
3.
C
CD
OJ
i—
u
c
D.
a>
2.5 -
2 -
1.5 -
1 -
T3
(T3
o
TO
0)
0.5 -
0
• •
	P.. •
• •
y = -0.0401X + 82.207
R2 = 0.294
1999 2002 2005 2008 2011 2014 2017
Figure 2-20. Annual average near-road increment for PM2.5 at the Elizabeth, NJ site.
2.3.2.2.3 Sub-Daily Concentrations of PM2.5
Ambient PM2.5 concentrations can exhibit a diurnal cycle that varies due to impacts from
intermittent emission sources, meteorology, and atmospheric chemistry. The PM2.5 monitoring
network in the U.S. has an increasing number of continuous FEM monitors reporting hourly
PM2.5 mass concentrations that reflect this diurnal variation. The ISA describes a two-peaked
diurnal pattern in urban areas, with morning peaks attributed to rush-hour traffic and afternoon
peaks attributed to a combination of rush hour traffic, decreasing atmospheric dilution, and
31	The Elizabeth Lab site in Elizabeth, NJ is situated approximately 30 meters from travel lanes of the Interchange
13 toll plaza of the New Jersey Turnpike and within 200 meters of travel lanes for Interstate 278 and the New
Jersey Turnpike.
32	See https://www.epa.g0v/diesel-fuel-standards/diesel-fuel-standards-and-rulemakings#n0nr0ad-diesel.
2-34

-------
nucleation (U.S. EPA, 2019a, section 2.5.2.3, Figure 2-32). Because a focus on annual average
and 24-hour average PM2.5 concentrations could mask sub-daily patterns, and because some
health studies examine PM exposure durations shorter than 24-hours, it is useful to understand
the broader distribution of sub-daily PM2.5 concentrations across the U.S. Figure 2-21 below
presents the frequency distribution of 2-hour average PM2.5 mass concentrations from all FEM
PM2.5 monitors in the U.S. for 2015-2017.33 At sites meeting the current primary PM2.5
standards, these 2-hour concentrations generally remain below 11 ng/m3, and virtually never
exceed 32 |ag/m3. Two-hour concentrations are higher at sites violating the current standards,
generally remaining below 19 |ag/m3 and virtually never exceeding 69 |ag/m3.
Sites meeting both NAAQS
>.
o
c
a>
Z3
CT
CD
1(T
10'
10°
10 1
Sites violating either NAAQS
Percentiles (|jg m 3)
10"
10"1
Percentiles (|jg nr3)
143.5
Percentiles (|jg nr3)
Percentiles (|jg nr3)
50th:
8.0
75th:
12.5
95th:
26.8
99th:
65.5
99.9th:
172.8
Concentration (jag m )
Concentration (fig m )
Figure 2-21. Frequency distribution of 2015-2017 2-hour averages for sites meeting both or
violating either PM2.5 NAAQS for October to March (blue) and April to September
(red).
The extreme upper end of the distribution of 2-hour PM2.5 concentrations is shifted higher
during the warmer months (red in Figure 2-21), generally corresponding to the period of peak
wildfire frequency (April to September) in the U.S. At sites meeting the current primary
standards, the highest 2-hour concentrations measured virtually never occur outside of the period
of peak wildfire frequency. Most of the sites measuring these very high concentrations are in the
northwestern U.S. and California, where wildfires have been relatively common in recent years
33 As discussed further in section 3.2, PM2.5 controlled human exposure studies often examine 2-hour exposures.
Thus, when evaluating those studies in the context of the current primary PM2.5 standards, it is useful to consider
the distribution of 2-hour PM2.5 concentrations. Similar analyses of 5-hour PM2.5 concentrations are presented in
Appendix A, Figure A-2.
2-35

-------
(see Appendix A, Figure A-l). When the period of peak wildfire frequency is excluded from the
analysis (blue in Figure 2-21), the extreme upper end of the di stribution is reduced.
2.3.2.3 Chemical Composition of PM2.5
Based on recent air quality data, the major chemical components of PM2.5 have distinct
spatial distributions. Sulfate concentrations tend to be highest in the eastern U.S., while in the
Ohio Valley, Salt Lake Valley, and California nitrate concentrations are highest and relatively
high concentrations of organic carbon are widespread across most of the Continental U.S., as
shown in Figure 2-22. Elemental carbon, crustal material, and sea-salt are found to have the
highest concentrations in the northeast U.S., southwest U.S., and coastal areas, respectively.
2015-2017 Sulfate	2015-2017 Nitrate
•	<0.5
•	0.5-1.0
•/
• >2.0
2015-2017 Organic Carbon
2015-2017 Elemental Carbon (2x)
•/
Figure 2-22. Annual average PM2.5 sulfate, nitrate, organic carbon, and elemental carbon
concentrations (in jig/m3) from 2015-2017.
An examination of PM2.5 composition trends can provide insight into the factors
contributing to overall reductions in ambient PM2.5 concentrations. The biggest change in PM2.5
composition that has occurred in recent years is the reduction in sulfate concentrations due to
reductions in SO2 emissions. Between 2000 and 2015, the nationwide annual average sulfate
concentration decreased by 17% at urban sites and 20% at rural sites. This change in sulfate
2-36

-------
concentrations is most evident in the eastern U.S. and has resulted in organic matter or nitrate
now being the greatest contributor to PM2.5 mass in many locations (U.S. EPA, 2019a, Figure 2-
19). The overall reduction in sulfate concentrations has contributed substantially to the decrease
in national average PM2.5 concentrations as well as the decline in the fraction of PM10 mass
accounted for by PM2.5 (U.S. EPA, 2019a, section 2.5.1.1.6; section 2.3.1 above).
2.3.2.4 National Characterization of PM10 Mass
At long-term monitoring sites in the U.S., the 2015-2017 average of 2nd highest 24-hour
PM10 concentration was 56 [j,g/m3 (ranging from 18 to 173 (J,g/m3) (Figure 2-23, top panels).34
The highest PM10 concentrations tend to occur in the western U.S. Seasonal analyses indicate
that ambient PM10 concentrations are generally higher in the summer months than at other times
of year, though the most extreme high concentration events are more likely in the spring (U.S.
EPA, 2019a, Table 2-5). This is due to fact that the major PM10 emission sources, dust and
agriculture, are more active during the warmer and drier periods of the year.
34 The form of the current 24-hour PM10 standard is one-expected-exceedance, averaged over three years.
2-37

-------
2015-2017 Annual Average
15-20
20-25

• 25-30
• >30
2015-2017 2nd Highest
I *
T-d • • " '
- i c	'
•rv.f t
!_• I • *# » I
y vt >.*
:( r * r ; . ^
^ 2000-2017 Annual Average Trend
~/ _
7	M—J ? S"1'" /"T V
,<75 v\
y	• 75-100 \ I
2000-2017 2nd Highest Trend
—a.—n
•V
•	<75
•	75-100
100-125 y
•	125-150
•	>150
- -^tw;
T *	V '
w \ \y
A^C'	
~ No Trend
(p>0.15)
T Reduction Possible
(pso.05) A Increase
~ Possible	(G.050.i5)
Reduction
05)
~ Possible
Reduction
(0.05
-------
Recent ambient PMio concentrations reflect reductions that have occurred across much of the
U.S. (Figure 2-23, bottom panels). From 2000 to 2017, 2nd highest 24-hour PMio concentrations
have declined by about 30% (Figure 2-24).35 Analyses at individual monitoring sites indicate that
annual average PMio concentrations have declined at most sites across the U.S., with much of the
decrease in the eastern U.S. associated with reductions in PM2.5 concentrations. Annual second
highest 24-hour PMio concentrations have generally declined in the eastern U.S., while
concentrations in the much of the midwest and western U.S. have remained unchanged or
increased since 2000 (Figure 2-23, bottom panels).
200
National Standard
150
01
=5
d
o
100-
50-
222222222222222222
000000000000000000
000000000011111111
0 1 2345678901 234567
Figure 2-24. National trends in Annual 2nd Highest 24-Hour PMio concentrations from
2000 to 2017 (131 sites). (Note: The white line indicates the mean concentration while the
gray shading denotes the 10th and 90th percentile concentrations.)
Compared to previous reviews, data available from the NCore monitoring network in the
current review allows a more comprehensive analysis of the relative contributions of PM2.5 and
PM10-2.5 to PMio mass. PM2.5 generally contributes more to annual average PMio mass in the
eastern U.S. than the western U.S. (Figure 2-25). At most sites in the eastern U.S., the majority
of PMio mass is comprised of PM2.5. Similar east-west patterns are observed for both
urban/suburban and rural sites. As ambient PM2.5 concentrations have declined in the eastern
U.S. (section 2.3.2.2, above), the ratios of PM2.5 to PMio have also declined.
35 For more information, see https://www.epa.gOv/air-trends/particulate-matter-pml0-trends#pmnat.
2-39

-------
2015-2017 Urban/Suburban Sites
Figure 2-25. Annual average PM2.5/PM10 ratio for 2015-2017.
For days with very high PM10 concentrations (Figure 2-26), the PM2.5/PM10 ratios are
typically higher than the annual average ratios. This is particularly true in the northwestern U.S.
where the high PM10 concentrations can occur during wildfires with high PM2.5.
2015-2017 Urban/Suburban Sites
Figure 2-26. PM2.5/PM10 ratio for the second highest PM10 concentrations for 2015-2017.
2.3.2.5 National Characterization of PM10-2.5 Mass
Since the last review, the availability of PM10-2.5 ambient concentration data has greatly
increased. As illustrated in Figure 2-2736 (top panels), annual average and 98th percentile PM10-2.5
concentrations exhibit less distinct differences between the eastern and western U.S. than for
either PM2.5 or PM10. Additionally, compared to PM2.5 and PM10, changes in PM10-2.5
concentrations have been small in magnitude and inconsistent in direction (Figure 2-27, lower
panels).
36 The sites shown in Figure 2-27 have a data completeness of either 75% or >182 valid days in each year.
2-40

-------
2015-2017 Annual Average
2000-2017 Annual Average Trend
~ No Trend
(p>0.15)
Significant
y Reduction
(p<0.05)
Possible
A Increase
(0.05
Significant 1—' (p>0.15)
T Reduction	Possible
(p<0.05) A jncrease
~ Possible	(0.05
-------
2.3.2.6 Characterization of the Ultrafine Fraction of PM2.5 Mass
Compared to PM2.5 mass, there is relatively little data on U.S. particle number
concentrations, which are dominated by UFP. In the published literature, annual average particle
number concentrations reaching about 20,000 to 30,000 cm 3 have been reported in U.S. cities
(U.S. EPA, 2019a). In addition, based on UFP measurements in two urban areas (New York
City, Buffalo) and at a background site (Steuben County) in New York, there is a pronounced
difference in particle number concentration between different types of locations (Figure 2-28;
U.S. EPA, 2019a, Figure 2-18). Urban particle number counts were several times higher than at
the background site, and the highest particle number counts in an urban area with multiple sites
(Buffalo) were observed at a near-road location. Hourly data indicate that particle numbers
remain fairly constant throughout the day at the background site, that they peak around 8:00 a.m.
in Buffalo and New York City (NYC), and that they remain high into the evening hours with
distinct rush hour and early afternoon peaks.
OK
0123456789 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hour of the Day
Figure 2-28. Average hourly particle number concentrations from three locations in the
State of New York for 2014 to 2015 (green is Steuben County, orange is Buffalo, red is
New York City). (Source: Figure 2-18 in U.S. EPA, 2019a).
Long-term trends in UFP are generally not available at U.S. monitoring sites. However,
data on number size distribution have been reported for an 8-year period from 2002 to 2009 in
Rochester, NY. Number concentrations averaged 4,730 cnT3 for 0.01 to 0.05 [j,m particles and
1,838 cnT3 for 0.05 to 0.1 [j,m particles (Wang et al., 2011). On average over the 8 years that
UFP data were collected in Rochester, total particle number concentrations declined from the
earlier period evaluated (i.e., 2001 to 2005) to the later period (2006 to 2009). This decline was
5K
2-42

-------
most evident for particles between 0.01 and 0.1 [j,m and was attributed to changes in local
sources resulting from the 2007 Heavy Duty Highway Rule, a reduction in local industrial
activity, and the closure of a nearby coal-fired power plant (Wang et al., 2011; U.S. EPA, 2019a,
section 2.5.2.1.4).
In addition, at a site in Illinois the annual average particle number concentration declined
between 2000 and 2017, closely matching the reductions in annual PM2.5 mass over that same
period (Figure 2-29, below). Particle number concentrations at this site are closer to those of the
background site in Figure 2-28 than the urban sites. A recent study found that particle number
concentrations in an urban area (Pittsburgh, PA) decreased between 2001-2002 and 2016-2017
along with decreases in PM2.5 associated with SO2 emission reductions (Saha et al., 2018).
However, the relationship between changes in ambient PM2.5 and UFPs cannot be
comprehensively characterized due to the high variability and limited monitoring of UFPs.

16


ro
1
14
E

cn
ZL
12
v—'

C
O
10
j-j

ro

-M
8
c

CD

U
c
6
0

u

->
c
CD
U
c
o
u
1_
CD
-O
E
=3
Year
2010
2015
Figure 2-29. Time series of annual average mass and number concentrations (left) and
scatterplot of mass vs. number concentration (right) between 2000-2017 in Bondville, IL.
2.3.3 Predicted Ambient PM2.5 Based on Hybrid Modeling Approaches
Ambient concentrations of PM2.5 are often characterized using measurements from
national monitoring networks due to the accuracy and precision of the measurements and the
public availability of data. For applications requiring PM2.5 characterizations across urban areas,
data averaging techniques such as area-wide and population-weighted averaging of monitors are
sometimes used to provide complete coverage from the site measurements (U.S. EPA, 2019a,
chapter 3). Yet data averaging methods may not adequately represent the spatial heterogeneity of
2-43

-------
PM2.5 within an area and are not practical for large unmonitored areas or time periods. As a
result, additional methods have been developed to improve PM2.5 characterizations in areas
where monitoring is relatively sparse or unavailable. Methods include interpolation of monitored
data, land-use regression models, chemical-transport models (CTMs), models based on satellite-
derived aerosol optical depth (AOD), and hybrid spatiotemporal models that combine
information from the individual approaches (U.S. EPA, 2019a, chapter 3). A number of recent
studies have employed such methods to estimate PM2.5 air quality concentrations across the U.S.
and Canada, and to estimate population exposures for use in epidemiologic analyses (U.S. EPA,
2019a, sections 3.3 and 3.4). Given the increasing availability and application of these methods,
in this section we provide an overview of recently developed hybrid modeling methods, their
predictions and performance, and how predictions from various methods compare to each other.
2.3.3.1.1 Overview of Hybrid Methods
Hybrid methods are broadly classified into four categories: (1) methods based primarily
on interpolation of monitor data, (2) Bayesian statistical downscalers, (3) methods based
primarily on satellite-derived AOD, and (4) methods based on machine-learning algorithms.
Each method is discussed briefly below.
Interpolation-based methods are the simplest approach for developing spatial fields of
PM2.5 concentrations and rely on the moderate degree of spatial autocorrelation in PM2.5 in many
areas of the U.S. Interpolation methods often use inverse-distance or inverse-distance-squared
weighted averaging of monitoring data to predict PM2.5 concentrations at unmonitored receptor
points. Examples include the Voronoi neighbor averaging (VNA) approach and the enhanced
VNA approach (eVNA). The VNA approach applies weighted averaging to the concentrations
monitored in the Voronoi cells neighboring the cell containing the prediction point (Abt
Associates, 2014). In the eVNA approach, monitored data are further weighted by the ratio of
CTM predictions in the grid-cell containing the prediction point to the grid-cell containing the
monitor (Abt Associates, 2014).
Bayesian statistical modeling has been used to calibrate CTM PM2.5 predictions or
satellite-derived AOD estimates to surface measurements (Berrocal et al., 2012; Wang et al.,
2018b). This approach, commonly referred to as a Bayesian downscaler because it "downscales"
grid-cell average values to points, first regresses the PM2.5 predictions or AOD estimates on
monitoring data. The resulting relationships are then used to develop a gridded PM2.5 field from
the CTM or AOD input field. Bayesian downscalers have been applied to develop gridded daily
PM2.5 fields at 12-km resolution for the conterminous U.S. (Wang et al., 2018b; U.S. EPA,
2017). An ensemble technique that optimally combines predictions of CTM and AOD
2-44

-------
downscalers has also been developed to predict PM2.5 at high resolution over Colorado during
the fire season (Geng et al., 2018).
Surface PM2.5 concentrations can also be predicted based on satellite retrievals of AOD
and the relationship between surface PM2.5 and AOD from CTM simulations (van Donkelaar et
al., 2010). For example, in van Donkelaar et al. (2015a), satellite-based approaches (van
Donkelaar et al., 2010; van Donkelaar et al., 2013) were used to estimate a gridded field of
global mean PM2.5 concentration for the 2001-2010 period that was combined with information
from radiometrically stable satellite instruments (Boys et al., 2014) to develop global PM2.5
fields over the 1998-2012 period (van Donkelaar et al., 2015a). Motivated by the limited use of
surface measurements in this approach, van Donkelaar et al. (2015b) developed an updated
method that incorporates additional information from PM2.5 monitoring networks to improve
performance. Specifically, geographically weighted regression (GWR) of residual PM2.5 (i.e., the
difference between monitored PM2.5 and predictions based on satellite-derived AOD) with land-
use and other variables is performed to improve PM2.5 concentration estimates in areas such as
North America where monitoring is relatively dense (van Donkelaar et al., 2019; van Donkelaar
et al., 2015b). This approach has been used to create long-term PM2.5 fields globally and for
North America at about 1-km resolution. However, the developers caution that PM2.5 gradients
may not be fully resolved at 1-km resolution due to the influence of coarser-scale data used in
the model37 and report that mean error variance decreases when averaging the 1-km fields to
coarser resolution (van Donkelaar et al., 2019).
Daily PM2.5 fields based on non-parametric (i.e., machine learning) methods have also
been developed to characterize PM2.5 over the U.S. Non-parametric methods facilitate the use of
large numbers of predictor variables that may have complex nonlinear relationships with PM2.5
concentrations that would be challenging to specify with a parametric method. For example, a
neural network algorithm was used to predict daily PM2.5 fields at 1-km resolution over the
conterminous U.S. during 2000-2012 using more than 50 predictor variables including satellite-
derived AOD, CTM predictions, satellite-derived absorbing aerosol index, meteorological data,
and land-use variables (Di et al., 2016). A random forest algorithm was also applied to develop
daily PM2.5 fields at 12-km resolution over the conterminous U.S. in 2011 and provide variable
importance information for about 40 predictor variables including CTM results and satellite-
derived AOD (Hu et al., 2017). Satellite-derived AOD and the convolution layer for nearby
PM2.5 measurements are ranked among the top five most important predictor variables for the
importance metrics considered. A wide range of parametric and non-parametric hybrid PM2.5
models have recently been reviewed in Chapter 3 of the ISA (U.S. EPA, 2019a).
37 See http://fizz.phvs.dal.ca/~atmos/martin/7page id= 140
2-45

-------
2.3.3.1.2 Performance of the Methods
The performance of hybrid modeling methods is often evaluated against surface
measurements using n-fold cross validation (i.e., 1/n of the data are reserved for validation with
the rest used for model training, and the process is repeated n times). Although model evaluation
methods are not consistent across studies, ten-fold cross-validation statistics are often reported
and support use of the hybrid methods just described. For example, the neural network achieved
total R2 of 0.84 and root-mean-square error (RMSE) of 2.94 |Lxg m"3 for daily PM2.5 predictions at
sites in the conterminous U.S. during 2000-2012 (Di et al., 2016). The random forest achieved
total R2 of 0.80 and RMSE of 2.83 |Lxg m"3 for daily PM2.5 predictions at U.S. sites in 2011 (Hu et
al., 2017). The satellite-derived AOD approach with GWR yielded an R2 of 0.79 and RMSE of
1.7 |Lxg m"3 in cross validation for longer-term PM2.5 predictions at sites in North America (van
Donkelaar et al., 2015b). The Bayesian downscalers had weaker performance in cross validation
(e.g., national R2: 0.66-0.70; Wang et al., 2018b; Kelly et al., 2019) than the other methods,
possibly due to the relatively small number of predictor variables. However, the downscalers
have advantages of simplicity, computational efficiency, and lower potential for overfitting
compared with the machine learning methods.
Although model validation analyses often report favorable performance in terms of
aggregate cross-validation statistics, studies have reported heterogeneity in performance by
season, region, and concentration range. For example, several methods had relatively high cross-
validation R2 in summer compared with other seasons (Kelly et al., 2019; Hu et al., 2017; Di et
al., 2016; van Donkelaar et al., 2015b). Also, studies have noted relatively weak performance in
parts of the western U.S., possibly due to the complex terrain, low concentrations (and therefore
signal-to-noise ratio), less dense monitoring, prevalence of wildfire, and challenges in satellite
retrievals and CTM modeling (Di et al., 2016; Wang et al., 2018b; Hu et al., 2017; Kelly et al.,
2019). Predictive capability in terms of cross-validation R2 has also been reported to weaken
with decreasing PM2.5 concentration in several studies (e.g., Kelly et al., 2019; Di et al., 2016;
van Donkelaar et al., 2019). Trends in model performance associated with PM2.5 concentration
(e.g., Figure 2-30) could be due in part to the relatively sparse monitoring in remote areas, where
PM2.5 concentrations tend to be low. Consistent with this hypothesis, studies have reported
degradation of model performance metrics with increasing distance to the nearest in-sample
monitor, suggesting that predictions are most reliable in densely monitored urban areas (Jin et
al., 2019; Huang et al., 2018; Kelly et al., 2019).
2-46

-------
1.00 -
0.75-
0.50-
0.25-
0.00-

lL
46
0-3
e
0 1
171	487
3-6 6-9
Observed PM2.5

F-|-| Downscaler
^ VNA
^ eVNA
37
>12
(H9 m
Figure 2-30. R2 for ten-fold cross-validation of daily PM2.5 predictions in 2015 from three
methods for individual sites as a function of observed concentration. Text indicates the
number of monitors in the PM2.5 concentration range. Downscaler: Bayesian downscaler of
CMAQ predictions; VNA: Voronoi Neighbor Averaging; eVNA: enhanced-VNA. From
Kelly etal. (2019).
A limited number of studies have intercompared concentration predictions based on
different PM2.5 characterization methods. Huang et al. (2018) compared PM2.5 concentrations
from the method of Di et al., 2016 with concentrations from the CTM-based data fusion method
of Friberg et al. (2016) and the satellite-derived AOD approach of Hu et al. (2014) for North
Carolina. They reported general agreement in concentrations among methods, with some
differences along the coast and in forested regions where monitoring is less dense. Yu et al.
(2018) compared PM2.5 concentrations from fourteen approaches of varying complexity for
developing PM2.5 spatial fields over the Atlanta, Georgia region. They reported that predictions
of the methods can differ considerably, and the hybrid approaches that incorporate CTM
predictions generally outperformed the simpler techniques (e.g., monitor interpolation). Also,
model predictions appeared to be more reliable in the urban center based on relatively low cross
validation R2 for sites away from the urban core. Jin et al. (2019) reported increasing uncertainty
in hybrid model predictions with distance to the nearest AQS monitor. Keller and Peng (2019)
reported that a prediction model incorporating CTM output outperformed a monitor averaging
approach and error reduction could be achieved by restricting the study to areas near monitors.
2.3.3.1.3 Comparison of PM2.5 Fields Across Approaches
To illustrate features of the spatial fields reported in the literature, the annual mean PM2.5
concentrations for 2011 from four methods is shown in Figure 2-31, where predictions from the
methods were averaged to a common 12-km grid. The fields were developed using a Bayesian
downscaler (downscaler, Berrocal et al., 2012), neural network (D12016, Di et al., 2016), random
forest (HU2017, Hu et al., 2017), and GWR of residuals from satellite-based PM2.5 estimates
(VD2019; van Donkelaar et al., 2019). Annual mean concentrations were developed from daily
PM2.5 predictions in the downscaler, DI2016, and HU2017 cases and from monthly PM2.5
predictions in the VD2019 case. General features of the 2011 fields are in reasonable agreement
2-47

-------
across methods, with elevated concentrations across broad areas of the eastern U.S. and in the
San Joaquin Valley and South Coast Air Basin of California. The nati onal mean PM2.5
concentration for the VD2019 case (7.06 |.ig rrf !) is slightly lower than those of the other cases
(7.36-7.44 jag m~3), possibly because the VD2019 fields were developed using monthly (rather
than daily) PM2.5 measurements. Use of monthly averages provides greater influence on the
annual mean of sites with less frequent monitoring that tend to be in rural areas with relatively
low concentrations. Mean PM2.5 concentrations predicted by the four methods in nine U.S.
climate regions (Karl and Koss, 1984) are provided in Table 2-3.
downscaler
DI2016
Gulf of
HU2017
VD2019
Gulf of
Longitude
Figure 2-31. Comparison of 2011 annual average PM2.5 concentrations from four methods.
(Note: These four methods include: downscaler (Berrocal et al.. 2012), DI2016 (Di et al.,
2016), HU2017 (Hu et al., 2017), and VD2019 (van Donkelaar et al., 2019). Predictions have
been averaged to a common 12-km grid for this comparison.
Table 2-3. Mean 2011 PM2.5 concentration by region for predictions in Figure 2-24
Region1
downscaler
HU2017
DI2016
VD2019
Northeast
8.5
8.0
8.2
7.5
Southeast
9.9
10.0
9.4
9.8
Ohio Valley
10.7
9,6
9.8
10.0
Upper Midwest
8.8
7.9
7.9
7.1
2-48

-------
South
00
CO
8.9
9.0
8.7
Southwest
5.0
5.3
5.2
5.1
N. Rockies & Plains
5.6
5.9
5.6
4.5
Northwest
5.0
5.3
6.1
4.9
West
5.5
5.7
6.0
6.5
1 U.S. climate reaion: httos://www.ncdc.noaa.aov/monitorina-references/maps/us-climate-reaions.php.
In Figure 2-32, PM2.5 concentrations predicted by the four methods are shown at their
native resolution for regions centered on California, New Jersey, and Arizona. Predictions span a
wider range of concentrations for the western regions centered on California and Arizona (Figure
2-32, panels a and c) than the eastern region centered on New Jersey (Figure 2-32, panel b).
Despite general agreement among predictions for the California and the eastern U.S. areas, the
spatial texture of the concentration fields differs among methods. For instance, the 12-km
Bayesian downscaler produces the smoothest PM2.5 concentration field, and the 1-km neural
network (DI2016) produces the field with the greatest variance. Some of the largest differences
in PM2.5 concentration among methods occurred over southwest Arizona. The DI2016 and
VD2019 methods predict higher concentrations in this area than the downscaler and HU2017
methods, and the DI2016 approach predicts distinct spatial features associated with Interstate 40,
10, and 8 that are not apparent in the other fields (Figure 2-32, panel c).
2-49

-------
(a)
downscaler
DI2016
(b)
downscaler
DI2016
0Modesto
VfejEresno
\S \
^¦uBakersfield

42-
41 -
40-
39-
oHarrisburg
0 ffnliadelphia
HF
¥
Av^ 6.62 ug/m3
£vg:7.06ug/m3 ^
38-
Avg: 9,27 Ug/m3
Avg: 9lig/m3
HU2017
VD2019

HU2017
VD2019


42-




41-



V
40-
39-
0
0

Avg: 6.56 UQ'mS
Avg: 7.68 ug.'m3
38-
Avg: 8 72 ijj$/m3
Avg: 8.49 ug,'ni3
-122 -120
-118 -116 -122
Longitude
-120 -118 -116
(C)
38-
downscaler
DI2016
-72 -78
Longitude
ug/m3
I
36
35-
34-
33-1
> &
l
i
\ 38-
37-|f
36 -
35-^
o Flagstaff
0 Phoenix
3: 4.91 ug/m3
HU2017
VD2019
ug/m3
> 16
I
Avg: 5.44 ug/m3
-114 -112
Avg 5 49 ug.TTiS
-114 -112 -110
Longitude
Figure 2-32. Comparison of 2011 annual average PM2.5 concentrations from four methods
for regions centered on the (a) California (b) New Jersey, and (c) Arizona. Predictions
are shown at their native resolution (i.e., about 1-km for DI2016 and VD2019 and 12-km for
downscaler and HU2017).
In Figure 2-33, the coefficient of variation (CV; i.e., the standard deviation divided by the
mean) among methods is shown in percentage units based on predictions that were averaged to a
common 12-km grid. The largest values occur in the western U.S. (Figure 2-33, panel a), where
terrain is complex, wildfire is prevalent, monitoring is relatively sparse, and PM2.5 concentrations
tend to be low. The distance from the grid-cell center to the nearest monitor is greater than 100
km for broad areas of the west (Figure 2-34).
2-50

-------
11442 13687 11302 14501 1451
(a)
45
"O 40
13
15
-1 35
30
50-
Figure 2-33. (a) Spatial distribution of the CV (i.e., standard deviation divided by mean) in
percentage units for the four models in Figure 2-31. (b) Boxplot distributions of CV for
grid cells binned by the average PM2.5 concentration for the four models. (Note: The box
brackets the interquartile range (IQR), the horizontal line within the box represents the
median, the whiskers represent 1.5 times the IQR from either end of the box, and circles
represent individual values less than and greater than the range of the whiskers.)
-120 -110 -100	-90	-80	-70
Longitude
3-5 5-7 7-9 9-11 11-16
Average PM2.5 (ng rrf3)
« 40
13
03
35
30
Figure 2-34. Distance from the center of the 12-km grid cells to the nearest P1VI2.5
monitoring site for PM2.5 measurements from the AQS database and IMPROVE
network.
25
-120 -110
-100	-90
Longitude
Concentrations less than 5 |ig m"3 occur exclusively in the western U.S. for the
downscaler and HU2017 methods, and the western U.S. plus a few areas along the northern U,S
border in the eastern U.S. for the DI2016 and VD2019 methods (Figure 2-35, top row).
Concentrations between 5 and 7 |j.g m"( are predicted in the western U.S. and parts of New
England for all methods and over Florida by the downscaler and DI2016 approaches (Figure 2-
35, second row). The CV among methods increases with decreasing concentration (Figure 2-33
above, panel b), and the median CV is about 15% for grid cells with mean concentrations less
than 7 |jg m"\ As illustrated by Figure 2-33 and Figure 2-35, the low-concentration areas with
2-51

-------
relatively large CVs are in the western U.S. and along the northern and southern border of the
eastern U.S.
downscaler	DI2016	HU2017	VD2019
45- ¦ — • ¦ -- 4 V i— ym ,-rg, -"rY-TH 'v"„ nji-	nfr-
40-	v
»:'•	'fm-W& '¦ ;v v , f*> ¦*
-120-110-100 -90 -80 -70-120-110-100 -90 -80 -70-120-110-100 -90 -80 -70-120-110-100 -90 -80 -70
Longitude
Figure 2-35. Location of PM2.5 predictions by range in annual average concentration for
the four prediction methods at their native resolution. (Note: Concentration ranges: < 5
jj,g/m3, 5-7 |Jg/m3, 7-9 fig/in3, 9-11 |Jg/m3, and >11 j.ig/m3.)
The comparison of PM2.5 concentrations across approaches was based on the 2011 period
due to the availability of predictions from multiple methods for that year. As discussed earlier in
this chapter, PM2.5 concentrations have declined over the U.S. in the last several decades. Annual
mean PM2.5 concentrations predicted by the VD2019 method for 2011 are compared with
predictions for 2001, 2006, and 2016 in Figure 2-36. The VD2019 fields capture the trend of
decreasing PM2.5 over the U.S. during this period, and the areas with annual mean PM2.5
concentration greater than 11 |jg rrf ' in 2016 are limited to California and southwest Arizona.
2-52

-------
2006
2016
CD 45
P 40
ug/m3

j > 15

10

5
0
-120-110-100 -90 -80 -70 -120-110-100 -90 -80 -70 -120-110-100 -90 -80 -70-120-110-100 -90 -80 -70
Longitude
Figure 2-36. Annual mean PM2.5 from the VD2019 method (van Donkelaar et al., 2019) for
2001,2006, 2011, and 2016.
2.3.3.1.4 Summary
Hybrid PM2.5 modeling methods have improved the ability to estimate PM2.5 exposure for
populations throughout the conterminous U.S. compared with the earlier approaches based on
monitoring data alone. Excellent performance in cross-validation tests suggests that hybrid
methods are reliable for estimating PM2.5 exposure in many applications. As discussed in
Chapter 3 of this PA, good agreement in health study results between monitor- and model-based
methods for urban areas (McGuinn et aL 2017) and general consistency in results for the
conterminous U.S. (Jerrett et al., 2017; Di et al., 2016) also suggests that the fields are reliable
for use in health studies. However, there are also important limitations associated with the
modeled fields. First, performance evaluations for the methods are weighted toward densely
monitored urban areas at the scales of representation of the monitoring networks. Predictions at
different scales or in sparsely monitored areas are relatively untested. Second, studies have
reported heterogeneity in performance with relatively weak performance in parts of the western
U.S., at low concentrations, at greater distance to monitors, and under conditions where the
reliability and availability of key input datasets (e.g., satellite retrievals and air quality modeling)
are limited. Differences in predictions among different hybrid methods have also been reported
and tend to be most important under conditions with the performance issues just noted.
Differences in predictions could also be related to the different approaches used to create long-
term PM2.5 fields (e.g., averaging daily PM2.5 fields vs. developing long-term average fields),
which is important due to variable monitoring schedules. More work on comprehensively
characterizing the performance of modeled fields is warranted and will further inform our
understanding of the implications of using these fields to estimate PM2.5 exposures in health
studies.
2.4 BACKGROUND PM
For the purposes of this assessment, we define background PM as all particles that are
formed by sources or processes that cannot be influenced by actions within the jurisdiction of
concern. For this document, U.S. background PM is defined as any PM formed from emissions
2-53

-------
other than U.S. anthropogenic (i.e. manmade) emissions. Potential sources of U.S. background
PM include both natural sources (i.e. PM that would exist in the absence of any anthropogenic
emissions of PM or PM precursors) and transboundary sources originating outside U.S. borders.
Ambient monitoring networks provide long-term records of speciated PM concentrations
across the U.S., which can inform estimates of individual source contributions to background PM
levels in different parts of the country. However, even the most remote monitors within the U.S.
can be periodically affected by U.S. anthropogenic emissions. Monitor data are also limited in
more remote areas due to a sparser monitoring network where PM concentrations are more likely
influenced by background sources. Chemical transport models (CTMs) offer complementary
information to ambient monitor networks by providing more spatially and temporally
comprehensive estimates of atmospheric composition. CTMs can also be applied to isolate
contributions from specific emission sources to PM concentrations in different areas via source
apportionment or "zero-out" modeling (i.e., estimating what the residual concentrations would be
were emissions from the emission source of interest to be entirely removed).
At annual and national scales, estimated background PM concentrations in the U.S. are
small compared to contributions from domestic anthropogenic emissions. For example, based on
zero-out modeling in the last review of the PM NAAQS, annual background PM2.5
concentrations were estimated to range from 0.5-3 |ig/m3 across the sites examined. The
magnitude and sources of background PM can vary widely by region and time of year. Coastal
sites may experience a consistent contribution of PM from sea spray aerosol, while other areas
covered with dense vegetation may be impacted by biogenic aerosol production during the
summertime. Sources of background PM also operate across a range of time scales. While some
sources like biogenic aerosol vary at monthly to seasonal scales, many sources of background
PM are episodic in nature. These episodic sources (e.g. large wildfires) can be characterized by
infrequent contributions to high-concentration events occurring over shorter periods of time (e.g.,
hours to several days). Such episodic events are sporadic and do not necessarily occur in all
years. While these exceptional episodes can lead to violations of the daily PM2.5 standard (35 |ig
m"3) in some cases (Schweizer et al., 2017), such events are routinely screened for and usually
identifiable in the monitoring data. As described further below, contributions to background PM
in the U.S. result mainly from sources within North America. Contributions from
intercontinental events have also been documented (e.g., transport from dust storms occurring in
deserts in North Africa and Asia), but these events are less common and represent a relatively
small fraction of background PM in most places.
While the potential sources of background PM discussed above include sources of both
fine (PM2.5) and coarse (PM10) particles, background contributions to ambient UFP are less well
characterized and are not discussed here due to lack of information. Section 2.4.1 below further
2-54

-------
discusses background PM from natural sources inside the U.S. Section 2.4.2 characterizes the
role of international transport of PM from sources outside U.S. borders.
2.4.1 Natural Sources
As noted in section 2.1.1, sources that contribute to natural background PM include dust
from the wind erosion of natural surfaces, sea salt, wildland fires, primary biological aerosol
particles (PBAP) such as bacteria and pollen, oxidation of biogenic hydrocarbons such as
isoprene and terpenes to produce SOA, and geogenic sources such as sulfate formed from
volcanic production of SO2 and oceanic production of dimethyl-sulfide (DMS). While most of
the above sources release or contribute predominantly to fine aerosol, some sources including
windblown dust, and sea salt also produce particles in the coarse size range (U.S. EPA, 2019a,
section 2.3.3).
Biogenic emissions from plants are perhaps the most ubiquitous sources of background
PM in the U.S. Certain species of plants and trees can release large amounts of VOCs such as
isoprene and monoterpenes that are oxidized in the atmosphere to form organic aerosol. SOA
production from biogenic emissions is largest in the southeastern U.S., where conditions are
warm, humid, and sunny for much of the year. Many of the processes involved with biogenic
SOA formation are complex and remain highly uncertain. Results from radiocarbon techniques
applied to distinguish modern (biogenic or fires) from fossil (anthropogenic) carbon fractions in
organic aerosol have suggested comparable contributions from both carbon types in the
Southeast where SOA concentrations are high (Schichtel et al., 2008). However, SOA formation
from biogenic emission sources can also be facilitated by the presence of anthropogenic
precursors (Xu et al., 2015). More work characterizing the interactions of anthropogenic and
biogenic emissions is needed to determine the implications of such processes for background PM
concentrations.
Soil dust and sea salt have been estimated to account for less than 10% of urban PM2.5 on
average in the U.S. (Karagulian et al., 2015), although episodic contributions from these sources
can be much higher in some locations. For example, during a dust storm affecting Phoenix in
July of 2011, peak hourly average PM10 concentrations were greater than 5,000 |ig/m3, with area-
wide average hourly concentrations ranging from a few hundred to a few thousand |ig/m3
(Vukovic et al., 2014). Dust can also account for much of the PM that originates from outside the
U.S., which we discuss further below (U.S. EPA, 2019a, section 2.5.4.2). In addition to sea salt
aerosol, biological production of the sulfate precursor DMS can also occur in some marine
environments, although the impact of DMS emissions on annual mean sulfate concentrations is
likely very small in the U.S. (<0.2 |ig/m3) and confined to coastal areas (Sarwar et al., 2018).
2-55

-------
Wildfires release large amounts of particles and gaseous PM precursors. Invasive species,
historical fire management practices, frequency of drought, and extreme heat have resulted in
longer fire seasons (Jolly et al., 2015) and more large fires (Dennison et al., 2014) over time. In
addition to emissions from fires in the U.S., emissions from fires in other countries can be
transported to the U.S. Transport of smoke from fires in Canada, Mexico, Central America, and
Siberia have been documented in multiple studies (U.S. EPA, 2009). According to the NEI,
wildfire smoke contributes between 10 and 20% of primary PM emissions in the U.S. per year
(U.S. EPA, 2019a, section 2.3.1), with much higher localized contributions near fire-affected
areas.
To illustrate how episodic impacts from a large natural source can affect PM
concentrations in the U.S., Figure 2-37 and Figure 2-38 show an example from a recent wildfire
event. In summer 2017, smoke from wildfires in British Columbia, Canada led to severe air
quality degradation in parts of the Pacific Northwest. A NASA Worldview38 image from August
4, 2017 (Figure 2-37) shows smoke from multiple fire detections across southern British
Columbia crossing into northern Washington state. Smoke from these fires was also captured at
the North Cascades IMPROVE monitor (Figure 2-38), where daily fine PM concentrations were
increased from a typical baseline of less than 10 jag/m3 to -100 jig/m3 during this time.
iBOgSGlL
Figure 2-37. Smoke and fire detections observed by the MODIS instrument onboard the
Aqua satellite on August 4th, 2017 accessed through NASA Worldview.
38 Available from https://worldview.earthdata.nasa.gov.
2-56

-------
CO
CO

-------
2.4.2.1	Transboundary Transport in North America
As discussed above, some of the largest potential international sources of U.S.
background PM originate elsewhere in North America. PM produced from fires in both Canada
and Mexico can affect air quality in the U.S., particularly in border states (Park et al., 2007;
Miller et al., 2011; Wang et al., 2018a). Anthropogenic emissions from Canada and Mexico can
also influence U.S. PM air quality. An inverse modeling study by Henze et al. (2009) estimated
that in 2001 anthropogenic SOx emissions from Canada and Mexico accounted for 6% and 4%
respectively of total daily inorganic PM2.5 in the U.S. These authors also estimated that SOx
emissions related to international shipping accounted for approximately 2% of total inorganic
PM in the U.S.
2.4.2.2	Long Range Transport from Anthropogenic Sources
Due to the relatively short atmospheric lifetime of particles (-days to weeks), long range
transport of aerosols does not contribute significant PM mass to the U.S. Heald et al. (2006)
estimated that transport from Asia accounted for less than 0.2 |ig/m3 of sulfate PM2.5 in the
Northwestern U.S. in spring, and Leibensperger et al. (2011) estimated intercontinental
contributions from Asian anthropogenic SO2 and NOx emissions of 0.1 - 0.25 |ig/m3 annually in
the western U.S. Leibensperger et al. (2011) also concluded that much of the intercontinental
influence captured by the GEOS-Chem model was in fact local PM production attributable to
domestic emissions in receptor countries arising from changes in global oxidant budgets, rather
than impacts from PM directly transported across geopolitical boundaries. The studies above are
also consistent with findings from other analyses. A report from the United Nations on global air
quality synthesizing results across many studies estimated an annual average contribution of
approximately 0.1 |ig/m3 sulfate PM in North America due to transport from East Asia
(TFHTAP, 2006).
2.4.2.3	Long Range Transport from Natural Sources
Long range transport of dust from both Asia (Vancuren and Cahill, 2002; Yu et al., 2008)
and North Africa (Prospero, 1999a; Prospero, 1999b; Chiapello et al., 2005; McKendry et al.,
2007) has been shown to occasionally contribute to surface PM concentrations in some regions
of the U.S. The likelihood of such long-range dust transport events depends on large-scale
meteorological patterns, which can vary significantly across seasons and between years. Yu et al.
(2015) found that the transport of North African dust across the Atlantic Ocean is strongly
negatively correlated with precipitation in the Sahel during the preceding year. Dust from Africa
has also shown a decreasing trend of approximately 10% per decade from 1982 to 2008 based on
measurements of aerosol optical depth and surface concentrations in Barbados. This trend was
attributed to a corresponding decrease in surface winds over source regions (Ridley et al., 2014).
2-58

-------
Variability in springtime Asian dust transport to the U.S. has been linked to north-south shifts in
trans-Pacific flow modulated by the El Nino-Southern Oscillation (Achakulwisut et al., 2017), as
well as to variations in regional precipitation affecting both dust emissions in Asia and
atmospheric residence times during transport (Fischer et al., 2009).
On average, intercontinental dust transport is estimated to contribute about 1-2 |ig/m3 to
annual PM2.5 at some U.S. sites (Jaffe et al., 2005; TFHTAP, 2006; Creamean et al., 2014).
However, daily concentrations can be substantially larger for individual events, especially for
coarser particles. For example, Jaffe et al. (2003) found evidence of Asian dust events in 1998
and 2001 contributing 30-40 |ig/m3 to daily PM10 at sites throughout the U.S., although the
authors also note that large events of this scale are rare and only occurred twice during their 15-
year study period. Similar magnitudes have also been reported for individual North African
events; analysis of a multidecadal record of African dust reaching Miami indicated
concentrations of PM ranging from -10 to 120 |ig/m3 (Prospero, 1999b; Prospero, 1999a).40
2.4.3 Estimating Background PM with Recent Data
As discussed above, the 2009 PM ISA estimated background PM concentrations at
several remote IMPROVE sites in different regions of the U.S. for 2004 using a combination of
monitor data and zero-out air quality modeling. Revisiting the speciated IMPROVE PM data that
the monitors included in the last assessment provides some insights into how contributions from
different PM sources may have changed, and what those changes (or lack thereof) mean for our
current understanding of background PM in the U.S.
Figure 2-39 shows observed annual average PM2.5 in 2004 and 2016 at the same remote
monitors examined in the last ISA. The comparisons show decreases in both total PM2.5 and
ammonium sulfate across all sites examined, consistent with decreases in anthropogenic SO2 and
other PM precursors observed over this time period. It is likely that most of the remaining
ammonium sulfate observed at these sites is also a result of domestic anthropogenic emissions
and therefore not relevant for assessments of background PM.
Sea salt and dust aerosol are likely natural in origin at these remote sites. With the
exception of REDW1, a coastal site in California, soil and sea salt aerosol together account for
less than about 0.5 |ig/m3 of the annual average PM2.5 at all monitors examined here, which is
below the values cited from the literature for long range dust contributions discussed above.
Contributions from ammonium nitrate and elemental carbon could be from either anthropogenic
or natural sources, but together represent less than about 0.5 |ig/m3 at most of the sites in 2016.
40 Sample collection began in 1974, before network PM10 and PM2 5 samplers were developed, and no size cut was
specified (Prospero, 1999b).
2-59

-------
The largest contribution from nitrate occurs at the BRIG1 monitor in New Jersey and is likely
anthropogenic given the high density of NOx from vehicle emissions in that region.
After ammonium sulfate, the next largest contributing species for most of the sites is
organic matter, which for many of the monitors in Figure 2-39 represents 50% or more of total
PM in both 2004 and 2016. In addition to the IMPROVE sites from the last ISA, Figure 2-31
also shows comparisons for three sites in the Southeast U.S. As a region, the Southeast has the
highest levels of biogenic aerosol production in the country, so the organic matter contribution at
these three sites likely represents an upper bound for the country of what natural biogenic
organic aerosol production could be under present atmospheric conditions. The organic aerosol
components shown in Figure 2-39 will also include the influence of fires for some monitors. The
highest organic matter contribution for any of the sites shown in Figure 2-39, including the three
Southeast monitors, is approximately 2 |ig/m3. While contributions from ammonium sulfate have
decreased substantially at some of the monitors, particularly the eastern sites, contributions from
organic aerosol are roughly consistent between 2004 and 2016, as are the contributions from the
other species assumed to be mostly natural in origin (soil and sea salt). Therefore, while no new
zero-out modeling was done for the current review, revisiting these monitors with more recent
data suggests that estimates of background concentrations at these monitors are still around 1-3
|ig/m3 and have not changed significantly since the last PM NAAQS Review.
While estimates of total annual background concentrations have generally not changed
significantly since the last review, our scientific understanding of organic aerosol formation has
evolved. Organic aerosol can be produced from a variety of natural and anthropogenic processes,
which presents a challenge for source attribution techniques. Additionally, new research over the
past decade has identified a host of new sources and chemical pathways for SOA formation that
have only recently begun to be implemented into CTMs. Further research implementing these
new sources and pathways into CTMs is needed to understand 1) the behavior of these different
algorithms under a range of possible atmospheric conditions, and 2) what the implications are for
understanding SOA formation in the U.S.
2-60

-------
ACAD1
BRIG1
D0SO1
V0YA2
2034 2010
BRID1
15"
oi
2004 2016
REDW1
2034 2016
2004 2016
CANY 1
2504 2016
GICL1
15-
5-
o-
15"
10"
oi
2004 2016
0KEF1
2004 2316
SHR01


¦




2034
2316
¦
Amirs, nit.

Amm. sulf.
2034 2016
2034 2016
GLAC1
15-
10"
oi
2034 2016
SIPS1
2034 2016
Figure 2-39. Speciated annual average IMPROVE PM2.5 in jug/m3 at select remote monitors
during 2004 and 2016. (Note: Monitor locations are shown in Figure 2-40.)
Figure 2-40. Site locations for the IMPROVE monitors in Figure 2-39. (Note: Monitors also
assessed in the 2009 ISA are shown in blue. Monitors only examined in this assessment are
shown in red.)
2-61

-------
REFERENCES
Abt Associates, Inc. (2014). User's guide: modeled attainment test software. Research Triangle
Park, NC, U.S. Environmental Protection Agency.
Achakulwisut, P, Shen, L and Mickley, LJ (2017). What controls springtime fine dust variability
in the western United States? Investigating the 2002-2015 increase in fine dust in the US
southwest. Journal of Geophysical Research: Atmospheres 122(22): 12449-12467.
Berrocal, VJ, Gelfand, AE and Holland, DM (2012). Space-time data fusion under error in
computer model output: an application to modeling air quality. Biometrics 68(3): 837-
848.
Boys, BL, Martin, RV, van Donkelaar, A, MacDonell, RJ, Hsu, NC, Cooper, MJ, Yantosca, RM,
Lu, Z, Streets, DG, Zhang, Q and Wang, SW (2014). Fifteen-year global time series of
satellite-derived fine particulate matter. Environ Sci Technol 48(19): 11109-11118.
Chan, EAW, Gantt, B and McDow, S (2018). The reduction of summer sulfate and switch from
summertime to wintertime PM2.5 concentration maxima in the United States. Atmos
Environ 175: 25-32.
Chiapello, I, Moulin, C and Prospero, JM (2005). Understanding the long-term variability of
African dust transport across the Atlantic as recorded in both Barbados surface
concentrations and large-scale Total Ozone Mapping Spectrometer (TOMS) optical
thickness. J Geophys Res 110(D18): D18S10.
Creamean, JM, Spackman, JR, Davis, SM and White, AB (2014). Climatology of long-range
transported Asian dust along the West Coast of the United States. Journal of Geophysical
Research: Atmospheres 119(21): 12171-12185.
Day, M, Pouliot, G, Hunt, S, Baker, KR, Beardsley, M, Frost, G, Mobley, D, Simon, H,
Henderson, B and Yelverton, T (2019). Reflecting on progress since the 2005 NARSTO
emissions inventory report. Journal of the Air Waste Management Association(just-
accepted).
Dennison, PE, Brewer, SC, Arnold, JD and Moritz, MA (2014). Large wildfire trends in the
western United States, 1984-2011. Geophys Res Lett 41(8): 2928-2933.
DeWinter, J, Brown, S, Seagram, A, Landsberg, K and Eisinger, D (2018). A national-scale
review of air pollutant concentrations measured in the U.S. near-road monitoring network
during 2014 and 2015. Atmos Environ 183.
U.S. Department of Health. (1969). Air Quality Criteria for Particulate Matter. Washington, D.C.
National Air Pollution Control Administration. DHEW. January 1969.
Di, Q, Kloog, I, Koutrakis, P, Lyapustin, A, Wang, Y and Schwartz, J (2016). Assessing PM2.5
exposures with high spatiotemporal resolution across the Continental United States.
Environ Sci Technol 50(9): 4712-4721.
2-62

-------
Fischer, EV, Hsu, NC, Jaffe, DA, Jeong, MJ and Gong, SL (2009). A decade of dust: Asian dust
and springtime aerosol load in the US Pacific Northwest. Geophys Res Lett 36.
Friberg, MD, Zhai, X, Holmes, HA, Chang, HH, Strickland, MJ, Sarnat, SE, Tolbert, PE,
Russell, AG and Mulholland, JA (2016). Method for fusing observational data and
chemical transport model simulations to estimate spatiotemporally resolved ambient air
pollution. Environ Sci Technol 50(7): 3695-3705.
Geng, G, Murray, NL, Tong, D, Fu, JS, Hu, X, Lee, P, Meng, X, Chang, HH and Liu, Y (2018).
Satellite-based daily PM2.5 estimates during fire seasons in Colorado. Journal of
Geophysical Research: Atmospheres 123(15): 8159-8171.
Hand, JL, Schichtel, BA, Malm, WC and Pitchford, ML (2012). Particulate sulfate ion
concentration and S02 emission trends in the United States from the early 1990s through
2010. Atmos ChemPhys 12(21): 10353-10365.
Heald, CL, Jacob, DJ, Park, RJ, Alexander, B, Fairlie, TD, Yantosca, RM and Chu, DA (2006).
Transpacific transport of Asian anthropogenic aerosols and its impact on surface air
quality in the United States. Journal of Geophysical Research: Atmospheres 111(D14).
Henze, DK, Seinfeld, JH and Shindell, DT (2009). Inverse modeling and maing US air quality
influences of inorganic PM(2.5) precursor emissions using the adjoint of GEOS-Chem.
Atmos Chem Phys 9(16): 5877-5903.
Hu, X, Waller, LA, Lyapustin, A, Wang, Y and Liu, Y (2014). 10-year spatial and temporal
trends of PM2.5 concentrations in the southeastern US estimated using high-resolution
satellite data. Atmos Chem Phys 14(12): 6301-6314.
Hu, X, Belle, JH, Meng, X, Wildani, A, Waller, LA, Strickland, MJ and Liu, Y (2017).
Estimating PM2.5 Concentrations in the Conterminous United States Using the Random
Forest Approach. Environ Sci Technol 51(12): 6936-6944.
Huang, Ra, Zhai, X, Ivey, CE, Friberg, MD, Hu, X, Liu, Y, Di, Q, Schwartz, J, Mulholland, JA
and Russell, AG (2018). Air pollutant exposure field modeling using air quality model-
data fusion methods and comparison with satellite AOD-derived fields: application over
North Carolina, USA. Air Quality, Atmosphere and Health 11(1): 11-22.
Jaffe, D, Snow, J and Cooper, O (2003). The 2001 Asian Dust Events: Transport and impact on
surface aerosol concentrations in the U.S. Eos 84(46): 501-506.
Jaffe, D, Tamura, S and Harris, J (2005). Seasonal cycle and composition of background fine
particles along the west coast of the US. Atmos Environ 39(2): 297-306.
Jerrett, M, Turner, MC, Beckerman, BS, Pope, CA, van Donkelaar, A, Martin, RV, Serre, M,
Crouse, D, Gapstur, SM, Krewski, D, Diver, WR, Coogan, PF, Thurston, GD and
Burnett, RT (2017). Comparing the health effects of ambient particulate matter estimated
using ground-based versus remote sensing exposure estimates. Environ Health Perspect
125(4): 552-559.
2-63

-------
Jin, X, Fiore, AM, Civerolo, K, Bi, J, Liu, Y, van Donkelaar, A, Martin, RV, Al-Hamdan, M,
Zhang, Y, Insaf, TZ, Kioumourtzoglou, M-A, He, MZ and Kinney, PL (2019).
Comparison of multiple PM2.5 exposure products for estimating health benefits of
emission controls over New York State, USA. Environmental Research Letters 14(8):
084023.
Jolly, WM, Cochrane, MA, Freeborn, PH, Holden, ZA, Brown, TJ, Williamson, GJ and
Bowman, DMJS (2015). Climate-induced variations in global wildfire danger from 1979
to 2013. Nature Communications 6.
Karagulian, F, Belis, CA, Dora, CFC, Pruess-Ustuen, AM, Bonjour, S, Adair-Rohani, H and
Amann, M (2015). Contributions to cities' ambient particulate matter (PM): A systematic
review of local source contributions at global level. Atmos Environ 120: 475-483.
Karl, T and Koss, WJ (1984). Regional and national monthly, seasonal, and annual temperature
weighted by area, 1895-1983. Asheville, NC, National Climatic Data Center.
Kaulfus, AS, Nair, U, Jaffe, D, Christopher, SA and Goodrick, S (2017). Biomass burning smoke
climatology of the United States: implications for particulate matter air quality. Environ
Sci Technol 51(20): 11731-11741.
Keller, JP and Peng, RD (2019). Error in estimating area-level air pollution exposures for
epidemiology. Environmetrics 0(0): e2573.
Kelly, JT, Koplitz, SN, Baker, KR, Holder, AL, Pye, HOT, Murphy, BN, Bash, JO, Henderson,
BH, Possiel, NC, Simon, H, Eyth, AM, Jang, CJ, Phillips, S and Timin, B (2019).
Assessing PM2.5 model performance for the conterminous U.S. with comparison to model
performance statistics from 2007-2015. Atmos Environ 214: 116872.
Leibensperger, EM, Mickley, LJ, Jacob, DJ and Barrett, SRH (2011). Intercontinental influence
of NO(x) and CO emissions on particulate matter air quality. Atmos Environ 45(19):
3318-3324.
McGuinn, LA, Ward-Caviness, C, Neas, LM, Schneider, A, Di, Q, Chudnovsky, A, Schwartz, J,
Koutrakis, P, Russell, AG, Garcia, V, Kraus, WE, Hauser, ER, Cascio, W, Diaz-Sanchez,
D and Devlin, RB (2017). Fine particulate matter and cardiovascular disease:
Comparison of assessment methods for long-term exposure. Environ Res 159: 16-23.
McKendry, IG, Strawbridge, KB, O'Neill, NT, MacDonald, AM, Liu, PSK, Leaitch, WR,
Anlauf, KG, Jaegle, L, Fairlie, TD and Westphal, DL (2007). Trans-Pacific transport of
Saharan dust to western North America: A case study. J Geophys Res 112(D1): DO 1103.
Miller, DJ, Sun, K, Zondlo, MA, Kanter, D, Dubovik, O, Welton, EJ, Winker, DM and Ginoux,
P (2011). Assessing boreal forest fire smoke aerosol impacts on U.S. air quality: A case
study using multiple data sets. J Geophys Res 116(D22): D22209.
Park, RJ, Jacob, DJ and Logan, JA (2007). Fire and biofuel contributions to annual mean aerosol
mass concentrations in the United States. Atmos Environ 41(35): 7389-7400.
2-64

-------
Prospero, JM (1999a). Long-range transport of mineral dust in the global atmosphere: Impact of
African dust on the environment of the southeastern United States. Proc Natl Acad Sci
USA 96(7): 3396-3403.
Prospero, JM (1999b). Long-term measurements of the transport of African mineral dust to the
southeastern United States: implications for regional air quality. J Geophys Res
104(D13): 15917-15927.
Ridley, DA, Heald, CL and Prospero, JM (2014). What controls the recent changes in African
mineral dust aerosol across the Atlantic? Atmos Chem Phys 14(11): 5735-5747.
Saha, PK, Robinson, ES, Shah, RU, Zimmerman, N, Apte, JS, Robinson, AL and Presto, AA
(2018). Reduced ultrafine particle concentration in urban air: changes in nucleation and
anthropogenic emissions. Environ Sci Technol 52(12): 6798-6806.
Sarwar, G, Xing, Ji, Fahey, K, Foley, K, Wong, D, Mathur, R, Gan, CM, Gantt, B and Simon, H
(2018). Dimethylsulfide Chemistry: Annual, Seasonal, and Spatial Impacts on Sulfate.
International Technical Meeting on Air Pollution Modelling and its Application: 347-
352.
Schichtel, BA, Malm, WC, Bench, G, Fallon, S, McDade, CE, Chow, JC and Watson, JG (2008).
Fossil and contemporary fine particulate carbon fractions at 12 rural and urban sites in the
United States. Journal of Geophysical Research: Atmospheres 113(D2).
Schweizer, D, Cisneros, R, Traina, S, Ghezzehei, TA and Shaw, G (2017). Using National
Ambient Air Quality Standards for fine particulate matter to assess regional wildland fire
smoke and air quality management. J Environ Manage 201: 345-356.
TFHTAP (2006). Hemispheric transport of air pollution 2010. Part A: Ozone and particulate
matter. HTAP 2010. F. Dentener, T. Keating and H. Akimoto. Geneva, Switzerland,
United Nations Publications.
U.S. EPA. (2009). Integrated Science Assessment for Particulate Matter (Final Report). Research
Triangle Park, NC. Office of Research and Development, National Center for
Environmental Assessment. U.S. EPA. EPA-600/R-08-139F. December 2009. Available
at: https://cfpub.epa.gov/ncea/risk/recordisplav.cfm?deid=216546.
U.S. EPA. (2012). Near-road N02 monitoring technical assistance document. Research Triangle
Park, NC. U.S. Environmental Protection Agency, Office of Air Quality Planning and
Standards. U.S. EPA. EPA-454/B-12-002. June 2012. Available at:
http://nepis.epa.gov/exe/ZyPURL.cgi?Dockev=P 100EPRZ.txt.
U.S. EPA (2015). Pilot study on coarse PM monitoring. Research Triangle Park, NC, U.S. EPA
Office of Air Quality Planning and Standards.
U.S. EPA. (2017). Bayesian space-time downscaling fusion model (downscaler) -Derived
Estimates of Air Quality for 2013. Research Triangle Park, NC. Office of Air Quality
2-65

-------
Planning and Standards, Office of Air and Radiation. U.S. EPA. EPA-450/R-17-001.
Available:
https://www.epa.gov/hesc/rsig-related-downloadable-data-files.
U.S. EPA. (2018). Integrated Science Assessment (ISA) for Particulate Matter (External Review
Draft). Washington, DC. U.S. Environmental Protection Agency, Office of Research and
Development, National Center for Environmental Assessment. U.S. EPA. EPA/600/R-
18/179. October 2018. Available at:
https://cfpub.epa.gov/ncea/isa/recordisplav.cfm?deid=341593
U.S. EPA. (2019a). Integrated Science Assessment (ISA) for Particulate Matter (Final Report).
Washington, DC. U.S. Environmental Protection Agency, Office of Research and
Development, National Center for Environmental Assessment. U.S. EPA. EPA/600/R-
19/188. December 2019. Available at: https://www.epa.gov/naaqs/particulate-matter-pm-
standards-integrated-science-assessments-current-review.
U.S. EPA. (2019b). Review of Sunset Organic and Elemental Carbon (OC and EC)
Measurements During EPA's Sunset Carbon Evaluation Project. Research Triangle Park,
NC. U.S. Environmental Protection Agency, Office of Air Quality Planning and
Standards. U.S. EPA. EPA-454/R-19-005. May 2019. Available at:
https://www.epa.gov/sites/production/files/2019-05/documents/sunset epa-454r-19-
005 l.pdf.
van Donkelaar, A, Martin, RV, Brauer, M, Kahn, R, Levy, R, Verduzco, C and Villeneuve, PJ
(2010). Global estimates of ambient fine particulate matter concentrations from satellite-
based aerosol optical depth: development and application. Environ Health Perspect
118(6): 847-855.
van Donkelaar, A, Martin, RV, Spurr, RJD, Drury, E, Remer, LA, Levy, RC and Wang, J (2013).
Optimal estimation for global ground-level fine particulate matter concentrations. Journal
of Geophysical Research: Atmospheres 118(11): 5621-5636.
van Donkelaar, A, Martin, RV, Brauer, M and Boys, BL (2015a). Use of satellite observations
for long-term exposure assessment of global concentrations of fine particulate matter.
Environ Health Perspect 123(2): 135-143.
van Donkelaar, A, Martin, RV, Spurr, RJ and Burnett, RT (2015b). High-resolution satellite-
derived PM2.5 from optimal estimation and geographically weighted regression over
North America. Environ Sci Technol 49(17): 10482-10491.
van Donkelaar, A, Martin, RV, Li, C and Burnett, RT (2019). Regional estimates of chemical
composition of fine particulate matter using a combined geoscience-statistical method
with information from satellites, models, and monitors. Environ Sci Technol 53(5).
Vancuren, RA and Cahill, TA (2002). Asian aerosols in North America: Frequency and
concentration of fine dust. J Geophys Res 107(D24): 4804.
2-66

-------
Vu, TV, Delgado-Saborit, JM and Harrison, RM (2015). Review: Particle number size
distributions from seven major sources and implications for source apportionment
studies. Atmos Environ 122: 114-132.
Vukovic, A, Vujadinovic, M, Pejanovic, G, Andric, J, Kumjian, MR, Djurdjevic, V, Dacic, M,
Prasad, AK, El-Askary, HM, Paris, BC, Petkovic, S, Nickovic, S and Sprigg, WA (2014).
Numerical simulation of "an American haboob". Atmos Chem Phys 14(7): 3211-3230.
Wang, S-C, Wang, Y, Estes, M, Lei, R, Talbot, R, Zhu, L and Hou, P (2018a). Transport of
Central American fire emissions to the U.S. Gulf Coast: climatological pathways and
impacts on ozone and PM2.5. Journal of Geophysical Research: Atmospheres 123(15):
8344-8361.
Wang, Y, Hopke, PK, Chalupa, DC and Utell, MJ (2011). Long-term study of urban ultrafine
particles and other pollutants. Atmos Environ 45(40): 7672-7680.
Wang, Y, Hu, X, Chang, H, Waller, L, Belle, J and Liu, Y (2018b). A Bayesian downscaler
model to estimate daily PM2.5 levels in the conterminous U.S. Int J Env Res Public
Health 15(9): 1999.
Whitby, KT, Husar, RB and Liu, BYH (1972). The aerosol size distribution of Los Angeles
smog. J Colloid Interface Sci 39: 177-204.
Xu, L, Guo, H, Boyd, CM, Klein, M, Bougiatioti, A, Cerully, KM, Hite, JR, Isaacman-Vanwertz,
G, Kreisberg, NM, Knote, C, Olson, K, Koss, A, Goldstein, AH, Hering, SV, de Gouw, J,
Baumann, K, Lee, SH, Nenes, A, Weber, RJ and Ng, NL (2015). Effects of
anthropogenic emissions on aerosol formation from isoprene and monoterpenes in the
southeastern United States. ProcNatl Acad Sci USA 112(1): 37-42.
Yu, H, Remer, LA, Chin, M, Bian, H, Kleidman, RG and Diehl, T (2008). A satellite-based
assessment of transpacific transport of pollution aerosol. J Geophys Res 113(D14): 1-15.
Yu, H, Chin, M, Bian, H, Yuan, T, Prospero, JM, Omar, AH, Remer, LA, Winker, DM, Yang, Y,
Zhang, Ya and Zhang, Z (2015). Quantification of trans-Atlantic dust transport from
seven-year (2007-2013) record of CALIPSO lidar measurements. Remote Sens Environ
159: 232-249.
Yu, H, Russell, A, Mulholland, J, Odman, T, Hu, Y, Chang, HH and Kumar, N (2018). Cross-
comparison and evaluation of air pollution field estimation methods. Atmos Environ 179:
49-60.
2-67

-------
3 REVIEW OF THE PRIMARY STANDARDS FOR PM25
This chapter presents our key policy-relevant considerations and conclusions regarding
the public health protection provided by the current suite of primary PM2.5 standards and the
protection that could be provided by potential alternative standards. These considerations and
conclusions are framed by a series of policy-relevant questions, including the following
overarching questions:
•	Does the currently available scientific evidence, air quality and quantitative risk
information support or call into question the adequacy of the public health
protection afforded by the current annual and 24-hour PM2.5 standards?
•	What range of potential alternative standards could be supported by the available
scientific evidence, air quality and risk information?
The answers to these questions are informed by our evaluation of a series of more specific
policy-relevant questions, which expand upon those presented at the outset of this review in the
IRP (U.S. EPA, 2016). Answers to these questions are intended to inform decisions by the
Administrator on whether, and if so how, to revise the current suite of primary fine particle
standards.
Section 3.1 presents our approach for reviewing the primary standards for PM2.5. Sections
3.2 and 3.3 present our consideration of the available scientific evidence and our consideration of
information from the PM2.5 risk assessment, respectively. Section 3.4 summarizes CAS AC
advice and public comments and section 3.5 summarizes our conclusions regarding the adequacy
of the public health protection provided by the current primary PM2.5 standards and the
protection that could be provided by potential alternative standards. Section 3.6 discusses areas
for future research and data collection to improve our understanding of fine particle-related
health effects in future reviews.
3.1 APPROACH
3.1.1 Approach Used in the Last Review
The last review of the primary PM NAAQS was completed in 2012 (78 FR 3086, January
15, 2013). As noted above (section 1.3), in the last review the EPA lowered the level of the
primary annual PM2.5 standard from 15.0 to 12.0 |ag/m3,' and retained the existing 24-hour PM2.5
standard with its level of 35 |j,g/m3. The 2012 decision to strengthen the suite of primary PM2.5
1 The Agency also eliminated spatial averaging provisions as part of the form of the annual standard.
3-1

-------
standards was based on the Administrator's consideration of the extensive body of scientific
evidence assessed in the 2009 ISA (U.S. EPA, 2009); the quantitative risk analyses presented in
the 2010 HREA (U.S. EPA, 2010);2 the advice and recommendations of the CASAC (e.g.,
Samet, 2009; Samet, 2010c; Samet, 2010b); and public comments on the proposed rule (78 FR
3086, January 15, 2013; U.S. EPA, 2012). The Administrator particularly noted the "strong and
generally robust body of evidence of serious health effects associated with both long- and short-
term exposures to PM2.5" (78 FR 3120, January 15, 2013). This included epidemiologic studies
reporting health effect associations based on long-term average PM2.5 concentrations ranging
from about 15.0 |j,g/m3 or above (i.e., at or above the level of the then-existing annual standard)
to concentrations "significantly below the level of the annual standard" (78 FR 3120, January 15,
2013). The Administrator further observed that such studies were part of an overall pattern
across a broad range of studies reporting positive associations, which were frequently
statistically significant. Based on her "confidence in the association between exposure to PM2.5
and serious public health effects, combined with evidence of such an association in areas that
would meet the current standards" (78 FR 3120, January 15, 2013), the Administrator concluded
that revision of the suite of primary PM2.5 standards was necessary in order to provide increased
public health protection. Specifically, she concluded that the then-existing suite of primary PM2.5
standards was not sufficient, and thus not requisite, to protect public health with an adequate
margin of safety. This decision was consistent with advice received from the CASAC (Samet,
2010c).
The Administrator next considered what specific revisions to the existing primary PM2.5
standards were appropriate, given the available evidence and quantitative risk information. She
considered both the annual and 24-hour PM2.5 standards, focusing on the basic elements of those
standards (i.e., indicator, averaging time, form, and level). These considerations, and the
Administrator's conclusions, are summarized in sections 3.1.1.1 to 3.1.1.4 below.
3.1.1.1 Indicator
In initially setting standards for fine particles in 1997, the EPA concluded it was
appropriate to control fine particles as a group, based on PM2.5 mass, rather than singling out any
particular component or class of fine particles (62 FR 38667, July 18, 1997). In the review
completed in 2006, based on similar considerations, the EPA concluded that the available
information supported retaining the PM2.5 indicator and remained too limited to support a distinct
2 In the last review, the EPA generated a quantitative health risk assessment for PM, and did not conduct a
microenvironmental exposure assessment (U.S. EPA, 2010). To be consistent with our general process for
reviewing the NAAQS (section 1.2, above), and with our discussion of potential quantitative analyses in the
current review, we refer to the 2010 health risk assessment as the 2010 HREA.
3-2

-------
standard for any specific PM2.5 component or group of components associated with particular
source categories of fine particles (71 FR 61162 to 61164, October 17, 2006).
In the last review, the EPA again considered issues related to the appropriate indicator for
fine particles, with a focus on evaluating support for the existing PM2.5 mass-based indicator and
for potential alternative indicators based on the ultrafine particle fraction or on fine particle
composition (78 FR 3121, January 15, 2013).3 With regard to PM2.5 mass, as in the 1997 and
2006 reviews, the health studies available during the last review continued to link adverse health
outcomes (e.g., premature mortality, hospital admissions, emergency department visits) with
long- and short-term exposures to fine particles indexed largely by PM2.5 mass (78 FR 3121,
January 15, 2013). With regard to the ultrafine fraction of ambient PM, the PA noted the limited
body of health evidence assessed in the ISA (summarized in U.S. EPA, 2009, section 2.3.5 and
Table 2-6) and the limited monitoring information available to characterize ambient
concentrations of ultrafine particles (U.S. EPA, 2011, section 1.3.2). With regard to PM
composition, the ISA concluded that "the evidence is not yet sufficient to allow differentiation of
those constituents or sources that are more closely related to specific health outcomes" (U.S.
EPA, 2009, pp. 2-26 and 6-212; 78 FR 3123, January 15, 2013). The PA further noted that
"many different constituents of the fine particle mixture as well as groups of components
associated with specific source categories of fine particles are linked to adverse health effects"
(U.S. EPA, 2011, p. 2-55; 78 FR3123, January 15, 2013). Consistent with the considerations
and conclusions in the PA, the CASAC advised that it was appropriate to consider retaining
PM2.5 as the indicator for fine particles. The CASAC specifically stated that "[t]here [is]
insufficient peer-reviewed literature to support any other indicator at this time" (Samet, 2010a, p.
12). In light of the evidence and the CASAC's advice, the Administrator concluded that it was
"appropriate to retain PM2.5 as the indicator for fine particles" (78 FR 3123, January 15, 2013).
3.1.1.2 Averaging Time
In 1997, the EPA set an annual PM2.5 standard to provide protection from health effects
associated with long- and short-term exposures to PM2.5, and a 24-hour standard to supplement
the protection afforded by the annual standard (62 FR 38667 to 38668, July, 18, 1997). In the
2006 review, the EPA retained both annual and 24-hour averaging times (71 FR 61164, October
17, 2006).
In the last review, the EPA again considered issues related to the appropriate averaging
times for PM2.5 standards, with a focus on evaluating support for the existing annual and 24-hour
3 In the last review, the ISA defined ultrafine particles as generally including particles with a mobility diameter less
than or equal to 0.1 |im. Mobility diameter is defined as the diameter of a particle having the same diffusivily or
electrical mobility in air as the particle of interest, and is often used to characterize particles of 0.5 |im or smaller
(U.S. EPA, 2009, pp. 3-2 to 3-3).
3-3

-------
averaging times and for potential alternative averaging times based on sub-daily or seasonal
metrics. Based on the evidence assessed in the ISA, the PA noted that the overwhelming
majority of studies that had been conducted since the 2006 review continued to utilize annual (or
multi-year) or 24-hour PM averaging periods (U.S. EPA, 2011, section 2.3.2). With regard to
potential support for an averaging time shorter than 24-hours, the PA noted that studies of
cardiovascular effects associated with sub-daily PM concentrations had evaluated a variety of
PM metrics (e.g., PM2.5, PM10, PM10-2.5, ultrafine particles), averaging periods (e.g., 1, 2, and 4
hours), and health outcomes (U.S. EPA, 2011, section 2.3.2). The PA concluded that this
evidence, when viewed as a whole, was too uncertain to serve as a basis for establishing a
primary PM2.5 standard with an averaging time shorter than 24-hours (U.S. EPA, 2011, p. 2-57).4
With regard to potential support for a seasonal averaging time, few studies were available to
deduce a general pattern in PM2.5-related risk across seasons, and these studies did not provide
information on health effects associated with season-long exposures to PM2.5 (U.S. EPA, 2011,
p. 2-58; 78 FR3124, January 15,2013).
The PA reached the overall conclusions that the available information provided strong
support for considering retaining the current annual and 24-hour averaging times and did not
provide support for considering alternative averaging times (U.S. EPA, 2011, p. 2-58). The
CASAC agreed that these conclusions were reasonable (Samet, 2010a, p. 13). The Administrator
concurred with the PA conclusions and with the CASAC's advice. Specifically, she judged that it
was "appropriate to retain the current annual and 24-hour averaging times for the primary PM2.5
standards to protect against health effects associated with long- and short-term exposure periods"
(78 FR 3124, January 15, 2013).
3.1.1.3 Form
In 1997, the EPA established the form of the annual PM2.5 standard as an annual
arithmetic mean, averaged over 3 years, from single or multiple community-oriented monitors.5
That is, the level of the annual standard was to be compared to measurements made at each
community-oriented monitoring site or, if specific criteria were met, measurements from
multiple community-oriented monitoring sites could be averaged together (i.e., spatial
4	For respiratory effects specifically, the Administrator further noted the ISA conclusion that the strongest
associations were observed with 24-hour average or longer exposures, not with exposures less than 24-hours
(U.S. EPA, 2009, section 6.3).
5	As noted above (section 1.3), in the last review the EPA replaced the term "community-oriented" monitor with the
term "area-wide" monitor. Area-wide monitors are those sited at the neighborhood scale or larger, as well as those
monitors sited at micro- or middle scales that are representative of many such locations in the same core-based
statistical area (CBSA; 78 FR 3236, January 15, 2013). CBSAs are required to have at least one area-wide
monitor sited in the area of expected maximum PM2 5 concentration.
3-4

-------
averaging)6 (62 FR 38671 to 38672, July 18, 1997). In the 1997 review, the EPA also established
the form of the 24-hour PM2.5 standard as the 98th percentile of 24-hour concentrations at each
monitor within an area (i.e., no spatial averaging), averaged over three years (62 FR at 38671 to
38674, July 18, 1997). In the 2006 review, the EPA retained these standard forms but tightened
the criteria for using spatial averaging with the annual standard (71 FR 61167, October 17,
2006).7
In the last review, the EPA's consideration of the form of the annual PM2.5 standard again
included a focus on the issue of spatial averaging. An analysis of air quality and population
demographic information indicated that the highest PM2.5 concentrations in a given area tended
to be measured at monitors in locations where the surrounding populations were more likely to
live below the poverty line and to include larger percentages of racial and ethnic minorities (U.S.
EPA, 2011, p. 2-60). Based on this analysis, the PA concluded that spatial averaging could result
in disproportionate impacts in at-risk populations, including minority populations and
populations with lower socioeconomic status (SES). Therefore, the PA concluded that it was
appropriate to consider revising the form of the annual PM2.5 standard such that it did not allow
for the use of spatial averaging across monitors (U.S. EPA, 2011, p. 2-60). The CASAC agreed
with the PA conclusions that it was "reasonable" for the EPA to eliminate the spatial averaging
provisions (Samet, 2010c, p. 2), stating the following: "Given mounting evidence showing that
persons with lower SES levels are a susceptible group for PM-related health risks, [the] CASAC
recommends that the provisions that allow for spatial averaging across monitors be eliminated"
(Samet, 2010a, p. 13).
The Administrator concluded that public health would not be protected with an adequate
margin of safety in all locations, as required by law, if disproportionately higher PM2.5
concentrations in low income and minority communities were averaged together with lower
concentrations measured at other sites in a large urban area. Therefore, she concluded that the
form of the annual PM2.5 standard should be revised to eliminate spatial averaging provisions (78
FR 3124, January 15, 2013). Thus, the level of the annual PM2.5 standard established in the last
review is to be compared with measurements from each appropriate monitor in an area, with no
allowance for spatial averaging.
6	The original criteria for spatial averaging included: (1) the annual mean concentration at each site shall be within
20% of the spatially averaged annual mean, and (2) the daily values for each monitoring site pair shall yield a
correlation coefficient of at least 0.6 for each calendar quarter (62 FR 38671 to 38672, July 18, 1997).
7	Specifically, the Administrator revised spatial averaging criteria such that "(1) [t]he annual mean concentration at
each site shall be within 10 percent of the spatially averaged annual mean, and (2) the daily values for each
monitoring site pair shall yield a correlation coefficient of at least 0.9 for each calendar quarter (71 FR 61167,
October 17, 2006).
3-5

-------
In the last review, the EPA also considered the form of the 24-hour PM2.5 standard. The
Agency recognized that the existing 98111 percentile form for the 24-hour standard was originally
selected to provide a balance between limiting the occurrence of peak 24-hour PM2.5
concentrations and identifying a stable target for risk management programs.8 Updated air
quality analyses in the last review provided additional support for the increased stability of the
98th percentile PM2.5 concentration, compared to the 99th percentile (U.S. EPA, 2011, Figure 2-2,
p. 2-62). Consistent with the PA conclusions based on this analysis, the Administrator concluded
that it was appropriate to retain the 98th percentile form for the 24-hour PM2.5 standard (78 FR
3127, January 15,2013).
3.1.1.4 Level
The EPA's approach to considering alternative levels of the PM2.5 standards in the last
review was based on evaluating the public health protection afforded by the annual and 24-hour
standards, taken together, against mortality and morbidity effects associated with long-term or
short-term PM2.5 exposures. This approach recognized that there is no bright line clearly
directing the choice of level. Rather, the choice of what is appropriate is a public health policy
judgment entrusted to the Administrator. In the last review, this judgment included consideration
of the strengths and limitations of the evidence and the appropriate inferences to be drawn from
the evidence and the risk assessments.
In evaluating alternative standards, the Agency considered the extent to which potential
alternative annual and 24-hour standard levels would be expected to reduce the mortality and
morbidity risks associated with both long-term and short-term PM2.5 exposures. Results of the
2010 HREA indicated that, compared to revising the 24-hour standard level, lowering the level
of the annual standard would result in more consistent risk reductions across urban study areas,
thereby potentially providing a more consistent degree of public health protection across the U.S.
(U.S. EPA, 2010, pp. 5-15 to 5-17; 78 FR 3128, January 15, 2013). Based on risk results,
together with the available evidence, the Administrator concluded that it was appropriate to
lower the level of the annual standard in order to increase protection against both long- and
short-term PM2.5 exposures. She further concluded that it was appropriate to retain the 24-hour
standard in order to provide supplemental protection, particularly for areas with high peak-to-
mean ratios of 24-hour PM2.5 concentrations (e.g., areas with important local or seasonal sources)
and for PIVh.s-related effects that may be associated with shorter-than daily exposure periods.
The Administrator judged that this approach was the "most effective and efficient way to reduce
8 See ATA III, 283 F.3d at 374-376 which concludes that it is legitimate for the EPA to consider overall stability of
the standard and its resulting promotion of overall effectiveness of NAAQS control programs in setting a standard
that is requisite to protect the public health.
3-6

-------
total PM2.5-related population risk and to protect public health with an adequate margin of
safety" (78 FR 3158, January 15, 2013).
In selecting the level of the annual PM2.5 standard, the Administrator recognized the
substantial increase in the number and diversity of studies available in the last review, including
extended analyses of seminal studies of long-term PM2.5 exposures (i.e., American Cancer
Society (ACS) and Harvard Six Cities studies), important new long-term exposure studies, and
new U.S. multi-city epidemiologic studies that greatly expanded and reinforced our
understanding of mortality and morbidity effects associated with short-term PM2.5 exposures.
She placed the greatest emphasis on health endpoints for which the evidence was strongest,
based on the assessment of the evidence in the ISA and on the ISA's causality determinations
(U.S. EPA, 2009, section 2.3.1). She particularly noted that the evidence was sufficient to
conclude a causal relationship exists between PM2.5 exposures and mortality and cardiovascular
effects (i.e., for both long- and short-term exposures) and that the evidence was sufficient to
conclude a causal relationship is "likely" to exist between PM2.5 exposures and respiratory
effects (i.e., for both long- and short-term exposures). The Administrator also noted additional,
but more limited, evidence for a broader range of health endpoints, including evidence
"suggestive of a causal relationship" between long-term exposures and developmental and
reproductive effects as well as carcinogenic effects (78 FR 3158, January 15, 2013).
Based on information discussed and presented in the ISA, the Administrator recognized
that health effects may occur over the full range of ambient PM2.5 concentrations observed in
epidemiologic studies, since no discernible population-level threshold could be identified based
on the evidence available in the last review (78 FR 3158, January 15, 2013; U.S. EPA, 2009,
section 2.4.3). To inform her decisions on an appropriate level for the annual standard in the
absence of a discernible population-level threshold, the Administrator considered the degree to
which epidemiologic studies indicate confidence in the reported health effect associations over
distributions of ambient PM2.5 concentrations. In doing so, she recognized that epidemiologic
studies provide greater confidence in the observed associations for the part of the air quality
distribution corresponding to the bulk of the health events evaluated, generally at and around the
long-term mean PM2.5 concentrations. Accordingly, the Administrator weighed most heavily the
long-term mean concentrations reported in key multi-city epidemiologic studies. She also took
into account additional population-level information from a subset of studies, beyond the long-
term mean concentrations, to identify a broader range of PM2.5 concentrations to consider in
3-7

-------
judging the need for public health protection.9 In doing so, the Administrator recognized that
studies indicate diminished confidence in the magnitude and significance of observed
associations in the lower part of the air quality distribution, corresponding to where a relatively
small proportion of the health events are observed.
In revising the level of the annual standard to 12.0 |ig/m3, the Administrator noted that
such a level was below the long-term mean PM2.5 concentrations reported in key epidemiologic
studies that provided evidence of an array of serious health effects, including premature mortality
and increased hospitalizations for cardiovascular and respiratory effects (78 FR 3161, January
15, 2013). The Administrator further noted that 12.0 |ig/m3 generally corresponded to the lower
portions (i.e., about the 25th percentile) of distributions of health events in the limited number of
epidemiologic studies for which population-level information was available. The Administrator
viewed this population information as helpful in guiding her determination as to where her
confidence in the magnitude and significance of the PM2.5 associations were reduced to such a
degree that a standard set at a lower level was not warranted. The Administrator also recognized
that a level of 12.0 |ig/m3 reflected placing some weight on studies of reproductive and
developmental effects, for which the evidence was more uncertain (78 FR 3161-3162, January
15, 2013).10
In conjunction with a revised annual standard with a level of 12.0 |ig/m3, the
Administrator concluded that the evidence supported retaining the 35 |ig/m3 level of the 24-hour
PM2.5 standard. Specifically, she judged that by lowering the level of the annual standard, the
distribution of 24-hour PM2.5 concentrations would be lowered as well, affording additional
protection against effects associated with short-term PM2.5 exposures.11 She noted that the
existing 24-hour standard, with its 35 |ig/m3 level and 98th percentile form, would provide
supplemental protection, particularly for areas with high peak-to-mean ratios possibly associated
with strong local or seasonal sources and for areas with PIVh.s-related effects that may be
associated with shorter than daily exposure periods (78 FR 3163, January 15, 2013).
9	This information characterized the distribution of health events in the studies, and the corresponding long-term
mean PM25 concentrations (78 FR 3130 to 3134, January 15, 2013). The additional population-level data helped
inform the Administrator's judgment of how far below the long-term mean concentrations to set the level of the
annual standard (78 FR 3160).
10	With respect to cancer, mutagenic, and genotoxic effects, the Administrator observed that the PM2 5
concentrations reported in studies evaluating these effects generally included ambient concentrations that are
equal to or greater than ambient concentrations observed in studies that reported mortality and cardiovascular and
respiratory effects (U.S. EPA, 2009, section 7.5). Therefore, the Administrator concluded that, in selecting a
standard level that provides protection from mortality and cardiovascular and respiratory effects, it is reasonable
to anticipate that protection will also be provided for carcinogenic effects (78 FR 3161-3162, January 15, 2013).
11	This judgment is supported by risk results presented in the 2010 HREA. For example, see section 4.2.2, and
Figures 4-4 and 4-6 (U.S. EPA, 2010).
3-8

-------
The Administrator recognized that uncertainties remained in the scientific information.
She specifically noted uncertainties related to understanding the relative toxicity of the different
components in the fine particle mixture, the role of PM2.5 in the complex ambient mixture,
exposure measurement errors in epidemiologic studies, and the nature and magnitude of
estimated risks related to relatively low ambient PM2.5 concentrations. Furthermore, the
Administrator noted that epidemiologic studies had reported heterogeneity in responses both
within and between cities and in geographic regions across the U.S. She recognized that this
heterogeneity may be attributed, in part, to differences in fine particle composition in different
regions and cities. With regard to evidence for reproductive and developmental effects, the
Administrator recognized that there were a number of limitations associated with this body of
evidence, including the following: the limited number of studies evaluating such effects;
uncertainties related to identifying the relevant exposure time periods of concern; and limited
toxicological evidence providing little information on the mode of action(s) or biological
plausibility for an association between long-term PM2.5 exposures and adverse birth outcomes.
On balance, the Administrator found that the available evidence, interpreted in light of
the remaining uncertainties (noted above), did not justify an annual standard level set below 12.0
|ig/m3 as being "requisite" (i.e., neither more nor less stringent than necessary) to protect public
health with an adequate margin of safety. Thus, the Administrator concluded that the available
evidence and information supported an annual standard with a level of 12.0 |ig/m3, combined
with a 24-hour standard with a level of 35 |ig/m3. She noted that this combination of standard
levels was consistent with the CASAC's advice to consider an annual standard level within the
range of 13 to 11 |_ig/m3 and a 24-hour standard level from 35 to 30 |j,g/m3 (Samet, 2010c). Taken
together, the Administrator concluded that the revised annual PM2.5 standard, with its level of
12.0 |ig/m3 and a form that does not allow for spatial averaging, combined with the existing 24-
hour standard, would be requisite to protect the public health with an adequate margin of safety
from effects associated with long- and short-term PM2.5 exposures.
3.1.2 General Approach in the Current Review
The approach for this review builds on the substantial body of work completed during the
last review, taking into account the more recent scientific information and air quality data now
available to inform our understanding of the key policy-relevant issues. The approach
summarized below is most fundamentally based on using the EPA's assessment of the current
scientific evidence for health effects attributable to fine particle exposures (i.e., in the ISA, U.S.
EPA, 2019), along with quantitative assessments of PM2.5-associated health risks and analyses of
PM2.5 air quality, and CASAC advice, to inform the Administrator's judgments regarding the
primary standards for fine particles that are requisite to protect the public health with an adequate
3-9

-------
margin of safety. In this PA, we seek to provide as broad an array of policy options as is
supportable by the available scientific and technical information, recognizing that the selection
of a specific approach to reaching final decisions on the primary PM2.5 standards will reflect the
judgments of the Administrator as to what weight to place on the various types of information
and associated uncertainties.
In considering the public health protection provided by the current primary PM2.5
standards, and the protection that could be provided by alternatives, we emphasize health
outcomes for which the ISA determines that the evidence supports either a "causal" or a "likely
to be causal" relationship with PM2.5 exposures (U.S. EPA, 2019). We consider the PIVh.s-related
health effects documented in studies that support these causality determinations and, together
with other analyses (i.e., air quality analyses, risk assessment), what they may indicate regarding
the primary PM2.5 standards. In doing so, we specifically focus on information from key
epidemiologic and controlled human exposure studies.
Epidemiologic studies represent a large part of the evidence base supporting several of
the ISA's "causal" and "likely to be causal" determinations. As discussed below in section
3.2.3.2, the use of information from epidemiologic studies to inform conclusions on the primary
PM2.5 standards is complicated by the fact that such studies evaluate associations between
distributions of ambient PM2.5 and health outcomes and do not identify the specific exposures
that cause reported effects. Rather, health effects can occur over the entire distributions of
ambient PM2.5 concentrations evaluated, and epidemiologic studies do not identify a population-
level threshold below which it can be concluded with confidence that PM-associated health
effects do not occur (U.S. EPA, 2019, section 1.5.3). In the absence of a discernible threshold,
we use two approaches to consider information from epidemiologic studies (section 3.2.3.2).
In one approach, we evaluate the PM2.5 air quality distributions over which epidemiologic
studies support health effect associations and the degree to which such distributions are likely to
occur in areas meeting the current (or alternative) standards. As discussed further in section
3.2.3.2.1, epidemiologic studies generally provide the strongest support for reported health effect
associations over the part of the air quality distribution corresponding to the bulk of the
underlying data (i.e., estimated exposures and/or health events), often falling in the middle part
of the distribution (i.e., rather than at the extreme upper or lower ends). In support of this, a
number of epidemiologic studies report that confidence intervals around concentration-response
functions are relatively narrow around the overall means of the PM2.5 concentrations examined
and wider at the extreme upper and lower ends of the distributions. The observed narrowing of
confidence intervals over the middle portions of these distributions likely reflects the relatively
large amount of data available (i.e., the numerous "typical" daily or annual PM2.5 exposures
estimated). As described in greater detail in section 3.2.3.2.1, in using PM2.5 air quality data from
3-10

-------
epidemiologic studies to inform conclusions on standards we evaluate study-reported means (or
medians) of daily and annual average PM2.5 concentrations as proxies for the middle portions of
the air quality distributions that support reported associations. When data are available, we also
consider the broader PM2.5 air quality distributions around the overall mean concentrations, with
a focus on the lower quartiles of data, to provide insight into the concentrations below which
data supporting reported associations become relatively sparse.
A key uncertainty in using study-reported PM2.5 concentrations to inform conclusions on
the primary PM2.5 standards is that they reflect the averages of daily or annual PM2.5 air quality
concentrations or exposure estimates in the study population over the years examined by the
study, and are not the same as the PM2.5 design values used by the EPA to determine whether
areas meet the NAAQS (section 3.2.3.2.1).12 Therefore, as described in section 3.2.3.2.2, in this
review we also consider a second approach to evaluating information from epidemiologic
studies. In this approach, we calculate study area air quality metrics similar to PM2.5 design
values (i.e., referred to in this PA as "pseudo-design values") and consider the degree to which
such metrics indicate that study area air quality would likely have met or violated the current or
alternative standards during study periods. When pseudo-design values in individual study
locations are linked with the populations living in those locations, or with the number of study-
specific health events recorded in those locations, these values can provide insight into the
degree to which reported health effect associations are based on air quality likely to have met or
violated the current (or alternative) primary PM2.5 standards.
To the extent the application of these two approaches indicates that health effect
associations are based on PM2.5 air quality likely to have met the current or alternative standards,
those standards are likely to allow the daily or annual average PM2.5 exposures that provide the
foundation for reported associations. Alternatively, to the extent reported health effect
associations reflect air quality violating the current or alternative standards, there is greater
uncertainty in the degree to which those standards would allow the PM2.5 exposures that provide
the foundation for reported associations. Sections 3.2.3.2.1 and 3.2.3.2.2 discuss each of these
approaches in detail, and present our key observations based on their application.
Beyond epidemiologic studies, we additionally consider what controlled human exposure
studies may indicate regarding the current and alternative primary PM2.5 standards. Controlled
human exposure studies examine short-term PM2.5 exposures (i.e., up to several hours) under
12 The design value is a statistic that describes the air quality status of a given area relative to the NAAQS. As
discussed further in section 3.2.3.2.1, to determine whether areas meet or violate the NAAQS, the EPA measures
air pollution concentrations at individual monitors (i.e., concentrations are not averaged across monitors) and
calculates design values at monitors meeting appropriate data quality and completeness criteria. For an area to
meet the NAAQS, all valid design values in that area, including the highest annual and 24-hour monitored values,
must be at or below the levels of the standards.
3-11

-------
carefully controlled laboratory conditions. Drawing from the ISA, such studies report PM2.5-
induced changes in markers of cardiovascular function and provide strong support for the
biological plausibility of the more serious cardiovascular-related outcomes observed in
epidemiologic studies (sections 3.2.1 and 3.2.3.1). Unlike most epidemiologic studies, available
controlled human exposure studies provide support for effects following single, short-term PM2.5
exposures to concentrations that typically correspond to the upper end of the PM2.5 air quality
distribution in the U.S. (i.e., "peak" concentrations). In evaluating what such controlled human
exposure studies may indicate regarding the primary standards, we consider the effects reported
following PM2.5 exposures, the exposure concentrations/durations reported to cause those effects,
and the degree to which air quality analyses indicate that such exposures are likely to occur in
areas meeting the current or alternative PM2.5 standards.13
Consideration of the evidence and related air quality analyses, as summarized above,
informs our evaluation of the public health protection provided by the combination of the current
annual and 24-hour primary PM2.5 standards, as well as the protection that could be provided by
alternative annual and 24-hour standards with revised levels (section 3.4). There are various
ways to combine an annual standard (based on arithmetic mean concentrations) and a 24-hour
standard (based on 98th percentile concentrations), to achieve an appropriate degree of public
health protection. The extent to which the standards are interrelated in any given area depends in
large part on the relative levels of the standards, the peak-to-mean ratios that characterize air
quality patterns in the area, and whether changes in air quality designed to meet a given suite of
standards are likely to be of a more regional or more localized nature. In considering the
combined effects of the standards, we recognize that changes in PM2.5 air quality designed to
meet an annual standard would likely result not only in lower short- and long-term PM2.5
concentrations near the middle of the air quality distribution (i.e., around the mean of the
distribution), but also in fewer and lower short-term peak PM2.5 concentrations. Additionally,
changes designed to meet a 24-hour standard, with a 98th percentile form, would result not only
in fewer and lower peak 24-hour PM2.5 concentrations, but also in lower annual average PM2.5
concentrations.
However, while either standard could be viewed as providing some measure of protection
against both average exposures and peak exposures, the 24-hour and annual standards are not
expected to be equally effective at limiting both types of exposures. Specifically, the 24-hour
standard (with its 98th percentile form) is more directly tied to short-term peak PM2.5
concentrations than to the more typical concentrations that make up the middle portion of the air
13 As discussed further in section 3.2.3.1, animal toxicology studies can be similarly evaluated, though there is
greater uncertainty in extrapolating the effects seen in animals, and the PM2 5 exposures and doses that cause
those effects, to human populations.
3-12

-------
quality distribution, and thus more likely to appropriately limit exposures to peak concentrations.
Compared to a standard that is directly tied to the middle of the air quality distribution, the 24-
hour standard is less likely to appropriately limit the typical exposures that are most strongly
associated with the health effects observed in epidemiologic studies. In contrast, the annual
standard, with its form based on the arithmetic mean concentration, is more likely to effectively
limit the PM2.5 concentrations that comprise the middle portion of the air quality distribution,
affording protection against the daily and annual PM2.5 exposures that strongly support
associations with the most serious PIVh.s-related effects in epidemiologic studies (e.g., mortality,
hospitalizations).
For these reasons, as in the last review (78 FR 3161-3162, January 15, 2013), we focus
on the annual PM2.5 standard as the principle means of providing public health protection against
the bulk of the distribution of short- and long-term PM2.5 exposures, and thus protecting against
the exposures that provide strong support for associations with mortality and morbidity in key
epidemiologic studies. We additionally consider the 24-hour standard, with its 98th percentile
form, as a means of providing supplemental protection against the short-term exposures to peak
PM2.5 concentrations that can occur in areas with strong contributions from local or seasonal
sources, even when overall mean PM2.5 concentrations remain relatively low (section 3.4).
Figure 3-1 summarizes our general approach to informing conclusions on the current
primary standards and on potential alternatives. Subsequent sections of this chapter provide
additional detail on this general approach.
3-13

-------
Adequacy of Existing Primary PM2 5 Standards
1
'

'
Evidence-Based Considerations
'-Degree to which available evidence strengthens supportfor, or calls
into question, health effects attributable to fine particle exposures
> Evidence for at-risk populations
'-Degree to which uncertainties have been addressed or new
uncertainties identified
'-Supportfor adverse effects at ambient concentrations meeting current
PM2 5 standards
*
Risk-Based Considerations
Nature, magnitude, and importance of
estimated risks associated with current
primary PM2 5 standards
'-Uncertainties in the risk estimates


Does information call
into question adequacy
of current Primary PM25
standards?
Appropnateto
consider retaining
current standards
Further evaluate the scientific evidence and risk assessmentto
inform Identification of potential alternatives
Identify range of potential alternative standards for consideration
^-Support for PM-attributable adverse effects at PM2 5 concentrations meeting potential alternative standards with
various levels?
^-Support from risk assessment for public health improvements by meeting potential alternative standards with
various levels?
^Uncertainties and limitations in the extent to which revised standards with various levels could result in public
health improvements, compared to existing standards
Indicator
Support for PM25?
Support for alternatives?
Averaging Time
Support for current 24-hour and/or annual?
Supportfor alternatives?
Level
^-Supportfor existing forms?
> Support for alternatives?
Form
Figure 3-1. Overview of general approach for review of primary P1VI2.5 standards.
3-14

-------
In adopting the approach outlined above, we recognize that decisions on the primary
PM2.5 standards are largely public health policy judgments to be made by the Administrator. The
Administrator's final decisions will draw upon the scientific evidence for PM-related health
effects, information from the quantitative assessment of population health risks, information
from analyses of air quality, and judgments about how to consider the uncertainties and
limitations that are inherent in the evidence and information. To inform the Administrator's
public health policy judgments and decisions, the PA considers support for, and the potential
implications of, placing more or less weight on various aspects of this evidence, air quality and
risk information, and associated uncertainties and limitations.
This approach is consistent with the requirements of the NAAQS provisions of the CAA
and with how the EPA and the courts have historically interpreted these CAA provisions. The
CAA requires primary standards that, in the judgment of the Administrator, are requisite to
protect public health with an adequate margin of safety. In setting primary standards that are
"requisite" to protect public health, the EPA's task is to establish standards that are neither more
nor less stringent than necessary for this purpose. The requirement that primary standards
provide an "adequate margin of safety" is meant to address uncertainties associated with
inconclusive scientific and technical information. Thus, as discussed in section 1.1 of this PA,
the CAA does not require that primary standards be set at a zero-risk level, but rather at a level
that, in the judgment of the Administrator, limits risk sufficiently so as to protect public health
with an adequate margin of safety.
3.2 EVIDENCE-BASED CONSIDERATIONS
In this section, we draw from the EPA's synthesis and assessment of the scientific
evidence presented in the ISA (U.S. EPA, 2019) to consider the following policy-relevant
question:
• To what extent does the currently available scientific evidence, as assessed in the
ISA, support or call into question the public health protection afforded by the
current suite of PM2.5 standards?
The ISA uses a weight-of-evidence framework for characterizing the strength of the available
scientific evidence for health effects attributable to PM exposures (U.S. EPA, 2015, Preamble,
Section 5). This framework provides the basis for robust, consistent, and transparent evaluation
of the scientific evidence, including its uncertainties, and for drawing conclusions on PM-related
health effects. As in the last review (U.S. EPA, 2009), the ISA for this review has adopted a five-
level hierarchy to classify the overall weight of evidence into one of the following categories:
causal relationship; likely to be a causal relationship; suggestive of, but not sufficient to infer, a
causal relationship; inadequate to infer a causal relationship; and not likely to be a causal
3-15

-------
relationship (U.S. EPA, 2015, Preamble Table II). In using the weight-of-evidence approach to
inform judgments about the likelihood that various health effects are caused by PM exposures,
evidence is evaluated for major outcome categories or groups of related outcomes (e.g.,
respiratory effects), integrating evidence from across disciplines, including epidemiologic,
controlled human exposure, and animal toxicological studies and evaluating the coherence of
evidence across a spectrum of related endpoints (U.S. EPA, 2015, Preamble, Section 5.c.). In this
PA, we consider the full body of health evidence, placing the greatest emphasis on the health
effects for which the evidence has been judged in the ISA to demonstrate a "causal" or a "likely
to be causal" relationship with PM exposures. The ISA defines these causality determinations as
follows (U.S. EPA, 2019, p. p-20):
•	Causal relationship: the pollutant has been shown to result in health effects at relevant
exposures based on studies encompassing multiple lines of evidence and chance,
confounding, and other biases can be ruled out with reasonable confidence.
•	Likely to be a causal relationship: there are studies in which results are not explained by
chance, confounding, or other biases, but uncertainties remain in the health effects evidence
overall. For example, the influence of co-occurring pollutants is difficult to address, or
evidence across scientific disciplines may be limited or inconsistent.
In the sections below, we consider the nature of the health effects attributable to long-
and short-term fine particle exposures (Section 3.2.1), the populations potentially at increased
risk for PM-related effects (Section 3.2.2), and the PM2.5 concentrations at which effects have
been shown to occur (Section 3.2.3).
3.2.1 Nature of Effects
In considering the available evidence for health effects attributable to PM2.5 exposures
presented in the ISA, this section poses the following policy-relevant questions:
•	To what extent does the currently available scientific evidence strengthen, or otherwise
alter, our conclusions from the last review regarding health effects attributable to long-
or short-term fine particle exposures? Have previously identified uncertainties been
reduced? What important uncertainties remain and have new uncertainties been
identified?
In answering these questions, as noted above, we consider the full body of evidence assessed in
the ISA, placing particular emphasis on health outcomes for which the evidence supports either a
"causal" or a "likely to be causal" relationship. While the strongest evidence focuses on PM2.5,
the ISA also assesses the evidence for the ultrafine fraction of PM2.5 (ultrafine particles or UFP),
3-16

-------
generally considered as particulates with a diameter less than or equal to 0.1 [j,m14 (typically
based on physical size, thermal diffusivity or electrical mobility) (U.S. EPA, 2019, Preface, p.
11). Table 3-1 lists the health outcomes for which the ISA concludes the evidence supports
either a causal, a likely to be causal, or a suggestive relationship (adapted from U.S. EPA, 2019,
Table 1-4).
14 Though definitions of UFP vary across the scientific literature and, as discussed in sections 3.2.1.5 and 3.2.1.6,
UFP exposures in animal toxicological and controlled human exposure studies typically use a particle
concentrator, which can result in exposures to particles > 0.1 |im in diameter in some studies of UFP-related
health effects.
3-17

-------
Table 3-1. Key causality determinations for PM2.5 and UFP exposures.
Health Outcome
Size
Fraction
Exposure
Duration
2009 PM ISA
2019 PM ISA
Mortality
PM2.5
Long-term
Short-term
Causal
Causal
Cardiovascular
effects
PM2.5
Long-term
Short-term
Causal
Causal
UFP
Short-term
Suggestive of, but not
sufficient to infer
Suggestive of, but not
sufficient to infer
Respiratory
effects
PM2.5
Long-term
Short-term
Likely to be causal
Likely to be causal
UFP
Short-term
Suggestive of, but not
sufficient to infer
Suggestive of, but not
sufficient to infer
Cancer
PM2.5
Long-term
Suggestive of, but not
sufficient to infer
Likely to be causal
Nervous
System effects
PM2.5
Long-term
...
Likely to be causal
Short-term
Inadequate
Suggestive of, but not
sufficient to infer
UFP
Long-term
...
Suggestive of, but not
sufficient to infer
Short-term
Inadequate
Suggestive of, but not
sufficient to infer
Metabolic effects
PM2.5
Long-term
...
Suggestive of, but not
sufficient to infer
Short-term
...
Suggestive of, but not
sufficient to infer
Reproduction
and Fertility
Pregnancy and
Birth Outcomes
PM2.5
Long-,
Short-term
Suggestive of, but not
sufficient to infer
Suggestive of, but not
sufficient to infer
Table 3-1 lists the health outcomes for which the ISA concludes the evidence supports either a causal, a likely
to be causal, or a suggestive relationship. For other health outcomes, the ISA concludes the evidence is
inadequate to infer a causal relationship (U.S. EPA, 2019, Table 1-4).
The 2009 ISA (U.S. EPA, 2009) made causality determinations for the broad category of "Reproductive and
Developmental Effects." Causality determinations for 2009 represent this broad category and not specifically
for "Male and Female Reproduction and Fertility" and "Pregnancy and Birth Outcomes".
For reproductive and developmental effects, the ISA's causality determinations reflect the combined evidence
for both short- and long-term exposures (U.S. EPA, 2019, Chapter 9).
3-18

-------
Sections 3.2.1.1 to 3.2.1.5 summarize the evidence supporting the ISA's "causal" and "likely to
be causal" determinations for PM2.5 (bold, italics in Table 3-1). Section 3.2.1.6 briefly
summarizes the evidence supporting the ISA's "suggestive" determinations. Each of these
sections focuses on addressing the policy-relevant questions posed above. Section 3.2.1.7
summarizes the evidence in preceding sections and revisits the policy-relevant questions posed
above.
3.2.1.1 Mortality
Long-term PM2.5 exposures
In the last review, the 2009 PM ISA reported that the evidence was "sufficient to
conclude that the relationship between long-term PM2.5 exposures and mortality is causal" (U.S.
EPA, 2009, p. 7-96). The strongest evidence supporting this conclusion was provided by
epidemiologic studies, particularly those examining two seminal cohort, the American Cancer
Society (ACS) and the Harvard Six Cities cohorts. Analyses of the Harvard Six Cities cohort
included demonstrations that reductions in ambient PM2.5 concentrations are associated with
reduced mortality risk (Laden et al., 2006) and with increases in life expectancy (Pope et al.,
2009). Further support was provided by other cohort studies conducted in North America and
Europe that also reported positive associations between long-term PM2.5 exposures and risk of
mortality (U.S. EPA, 2009).
Recent cohort studies, which have become available since the 2009 ISA, continue to
provide consistent evidence of positive associations between long-term PM2.5 exposures and
mortality. These studies add support for associations with total and non-accidental mortality,15 as
well as with specific causes of death, including cardiovascular disease and respiratory disease
(U.S. EPA, 2019, section 11.2.2). Many of these recent studies have extended the follow-up
periods originally evaluated in the ACS and Harvard Six Cities cohorts and continue to observe
positive associations between long-term PM2.5 exposures and mortality (U.S. EPA, 2019, section
11.2.2.1; Figures 11-18 and 11-19). Adding to recent evaluations of the ACS and Six Cities
cohorts, studies conducted in other cohorts also demonstrate consistent, positive associations
between long-term PM2.5 exposure and mortality across various demographic groups (e.g., age,
sex, occupation), spatial and temporal extents, exposure assessment metrics, and statistical
techniques (U.S. EPA, 2019, sections 11.2.2.1, 11.2.5). This includes some of the largest cohort
studies conducted to date, with analyses of the U.S. Medicare cohort that include nearly
61 million enrollees (Di et al., 2017b) and studies that control for a range of individual and
15 The majority of these studies examined non-accidental mortality outcomes, though some Medicare studies lack
cause-specific death information and, therefore, examine total mortality.
3-19

-------
ecological covariates, such as race, age, socioeconomic status, smoking status, body mass index,
and annual weather variables (e.g., temperature, humidity).
A recent series of retrospective studies has additionally tested the hypothesis that past
reductions in ambient PM2.5 concentrations have been associated with increased life expectancy
or a decreased mortality rate (U.S. EPA, 2019, section 11.2.2.5). In their original study, Pope et
al. (2009) used air quality data in a cross-sectional analysis from 51 metropolitan areas across the
U.S., beginning in the 1970s through the early 2000s, to demonstrate that a 10 |ig/m3 decrease in
long-term PM2.5 concentration was associated with a 0.61-year increase in life expectancy. In a
subsequent analysis, these authors extended the period of analysis to include 2000 to 2007
(Correia et al., 2013), a time period with lower ambient PM2.5 concentrations. In this follow-up
study, a decrease in long-term PM2.5 concentration continued to be associated with an increase in
life expectancy, though the magnitude of the increase was smaller than during the earlier time
period (i.e., a 10 |ig/m3 decrease in long-term PM2.5 concentration was associated with a
0.35-year increase in life expectancy). Additional studies conducted in the U.S. or Europe
similarly report that reductions in ambient PM2.5 are associated with improvements in longevity
(U.S. EPA, 2019, section 11.2.2.5).
The ISA specifically evaluates the degree to which recent studies that examine the
relationship between long-term PM2.5 exposure and mortality have addressed key policy-relevant
issues and/or previously identified data gaps in the scientific evidence. For example, based on its
assessment of the evidence, the ISA concludes that positive associations between long-term
PM2.5 exposures and mortality are robust across recent analyses using various approaches to
estimate PM2.5 exposures (e.g., based on monitors, modeling, satellites, or hybrid methods that
combine information from multiple sources) (U.S. EPA, 2019, section 11.2.5.1). This includes a
recent study (Hart et al. (2015) reporting that correction for bias due to exposure measurement
error increases the magnitude of the hazard ratios (confidence intervals widen but the association
remains statistically significant), suggesting that failure to correct for exposure measurement
error could result in attenuation or underestimation of risk estimates. The ISA additionally
concludes that positive associations between long-term PM2.5 exposures and mortality are robust
across statistical models that use different approaches to control for confounders or different sets
of confounders (U.S. EPA, 2019, sections 11.2.3 and 11.2.5), across diverse geographic regions
and populations, and across a range of temporal periods including the periods of declining PM
concentrations (U.S. EPA, 2019, sections 11.2.2.5 and 11.2.5.3). Recent evidence further
demonstrates that associations with mortality remain robust in copollutants analyses (U.S. EPA,
2019, section 11.2.3), and that associations persist in analyses restricted to long-term exposures
below 12 |j,g/m3 (Di et al., 2017b) or 10 |j,g/m3 (Shi et al., 2016) (i.e., indicating that risks are not
disproportionately driven by the upper portions of the air quality distribution).
3-20

-------
An emerging group of studies explores the use of causal inference methods to further
evaluate the causal nature of relationships between long-term PM2.5 exposure and mortality (U.S.
EPA, 2019, section 11.2.2.4). The goal of these methods is to "estimate the difference (or ratio)
in the expected value of [an] outcome in the population under the exposure they received versus
what it would have been had they received an alternative exposure" (Schwartz et al., 2015). For
example, Wang et al. (2016)) observe a positive and statistically significant relationship between
long-term exposure to PM2.5 and total (nonaccidental) mortality in New Jersey using a
difference-in-difference approach to control for geographical differences, long-term temporal
trends, and temperature. Additionally, a few recent studies use statistical techniques to reduce
uncertainties related to potential confounding in order to further inform conclusions on causality
for long-term PM2.5 exposure and mortality. For example, studies by Greven et al. (2011) and
Pun et al. (2017) decompose ambient PM2.5 into "spatial" and "spatiotemporal" components in
order to evaluate the potential for bias due to unmeasured spatial confounding. The results of
these analyses suggest the presence of unmeasured confounding for several health outcomes,
though they do not indicate the direction or magnitude of the bias that could result.1617
An additional important consideration in characterizing the public health impacts
associated with PM2.5 exposure is whether concentration-response relationships are linear across
the range of concentrations or if nonlinear relationships exist along any part of this range. Several
recent studies examine this issue, and continue to provide evidence of linear, no-threshold
relationships between long-term PM2.5 exposures and all-cause and cause-specific mortality (U.S.
EPA, 2019, section 11.2.4). Though available studies have not systematically evaluated
alternatives to a linear fitted model of concentration-response relationships, potential deviations
from linearity have been assessed in individual studies using a variety of approaches (U.S. EPA,
2019, Table 11-7). However, interpreting the shapes of these relationships, particularly at PM2.5
concentrations near the lower end of the air quality distribution, can be complicated by relatively
low data density in the lower concentration range, the possible influence of exposure measurement
error, and variability among individuals with respect to air pollution health effects. These sources
16	In public comments on the draft PA, the authors of the Pun et al. study further note that "the presence of
unmeasured confounding.. .was expected given that we did not control for several potential confounders that may
impact PM2 5-mortality associations, such as smoking, socio-economic status (SES), gaseous pollutants, PM2 5
components, and long-term time trends in PM2 5" and that "spatial confounding may bias mortality risks both
towards and away from the null" (Docket ID EPA-HQ-OAR-2015-0072-0065; accessible in
https://www.re gulations. gov/)
17	In its letter on the draft PA, the CASAC cites the study by Eum et al. (2018), which evaluates approaches similar
to those in Greven et al. (2011) and Pun et al. (2017). Eum et al. (2018) concludes that associations between 1-
year PM2 5 exposures and mortality "were likely confounded by long-term temporal trends in PM2 5" but that
controlling for this confounding still resulted in a statistically significant "11.7% increase in all-cause mortality
among Medicare beneficiaries for a 10 (ig/m3 increase inPM2 5."
3-21

-------
of variability and uncertainty tend to smooth and "linearize" population-level concentration-
response functions, and thus could obscure the existence of a threshold or nonlinear relationship
(U.S. EPA, 2015, section 6.c).
The biological plausibility of PIVh.s-attributable mortality is supported by the coherence
of effects across scientific disciplines (i.e., animal toxicological, controlled human exposure
studies, and epidemiologic), including in recent studies evaluating the morbidity effects that are
the largest contributors to total (nonaccidental) mortality. The ISA outlines the available
evidence for plausible pathways by which inhalation exposure to PM2.5 could progress from
initial events (e.g., pulmonary inflammation, autonomic nervous system activation) to endpoints
relevant to population outcomes, particularly those related to cardiovascular diseases such as
ischemic heart disease, stroke and atherosclerosis (U.S. EPA, 2019, section 6.2.1), and to
metabolic disease and diabetes (U.S. EPA, 2019, section 7.3.1). The ISA notes "more limited
evidence from respiratory morbidity" (U.S. EPA, 2019, p. 11-101) such as exacerbation of
COPD (U.S. EPA, 2019, section 5.2.1) to support the biological plausibility of mortality due to
long-term PM2.5 exposures (U.S. EPA, 2019, section 11.2.1).
Taken together, recent studies reaffirm and further strengthen the body of evidence from
the 2009 ISA for the relationship between long-term PM2.5 exposure and mortality. Recent
epidemiologic studies consistently report positive associations with mortality across different
geographic locations, populations, and analytic approaches. Such studies reduce key
uncertainties identified in the last review, including those related to potential copollutant
confounding, and provide additional information on the shape of the concentration-response
curve. Recent experimental and epidemiologic evidence for cardiovascular effects, and
respiratory effects to a more limited degree, supports the plausibility of mortality due to long-
term PM2.5 exposures. The ISA concludes that, "collectively, this body of evidence is sufficient
to conclude that a causal relationship exists between long-term PM2.5 exposure and total
mortality" (U.S. EPA, 2019, section 11.2.7; p. 11-102).
Short-term PM2.5 exposures
The 2009 PM ISA concluded that "a causal relationship exists between short-term
exposure to PM2.5 and mortality" (U.S. EPA, 2009). This conclusion was based on the evaluation
of both multi- and single-city epidemiologic studies that consistently reported positive
associations between short-term PM2.5 exposure and non-accidental mortality. These associations
were strongest, in terms of magnitude and precision, primarily at lags of 0 to 1 days.
Examination of the potential confounding effects of gaseous copollutants was limited, though
evidence from single-city studies indicated that gaseous copollutants have minimal effect on the
PM2.5-mortality relationship (i.e., associations remain robust to inclusion of other pollutants in
copollutant models). The evaluation of cause-specific mortality found that effect estimates were
3-22

-------
larger in magnitude, but also had larger confidence intervals, for respiratory mortality compared
to cardiovascular mortality. Although the largest mortality risk estimates were for respiratory
mortality, the interpretation of the results was complicated by the limited coherence from studies
of respiratory morbidity. However, the evidence from studies of cardiovascular morbidity
provided both coherence and biological plausibility for the relationship between short-term PM2.5
exposure and cardiovascular mortality.
Recent multicity studies evaluated since the 2009 ISA continue to provide evidence of
primarily positive associations between daily PM2.5 exposures and mortality, with percent
increases in total mortality ranging from 0.19% (Lippmann et al., 2013) to 2.80% (Kloog et al.,
2013)18 at lags of 0 to 1 days in single-pollutant models. Whereas most studies rely on assigning
exposures using data from ambient monitors, associations are also reported in recent studies that
employ hybrid modeling approaches using additional PM2.5 data (i.e., from satellites, land use
information, and modeling, in addition to monitors), allowing for the inclusion of more rural
locations in analyses (Kloog et al., 2013, Shi et al., 2016, Lee et al., 2015).
Some recent studies have expanded the examination of potential confounders, including
long-term temporal trends, weather, and co-occurring pollutants. Mortality associations were
found to remain positive, although in some cases were attenuated, when using different
approaches to account for temporal trends or weather covariates (U.S. EPA, 2019, section
11.1.5.1). For example, Sacks et al. (2012) examined the influence of model specification using
the approaches for confounder adjustment from models employed in several recent multicity
studies within the context of a common data set (U.S. EPA, 2019, section 11.1.5.1). These
models use different approaches to control for long-term temporal trends and the potential
confounding effects of weather. The authors report that associations between daily PM2.5 and
cardiovascular mortality were similar across models, with the percent increase in mortality
ranging from 1.5-2.0%) (U.S. EPA, 2019, Figure 11-4). Thus, alternative approaches to
controlling for long-term temporal trends and for the potential confounding effects of weather
may influence the magnitude of the association between PM2.5 exposures and mortality but have
not been found to influence the direction of the observed association (U.S. EPA, 2019, section
11.1.5.1). Taken together, the ISA concludes that recent multicity studies conducted in the U.S.,
Canada, Europe, and Asia continue to provide consistent evidence of positive associations
between short-term PM2.5 exposures and total mortality across studies that use different
approaches to control for the potential confounding effects of weather (e.g., temperature) (U.S.
EPA, 2019, section 1.4.1.5.1).
18 As detailed in the Preface to the ISA, risk estimates are for a 10 ng/m3 increase in 24-hour avg PM2 5
concentrations, unless otherwise noted (U.S. EPA, 2019).
3-23

-------
With regard to copollutants, recent studies provide additional evidence that associations
between short-term PM2.5 exposures and mortality remain positive and relatively unchanged in
copollutant models with both gaseous pollutants and PM10-2.5 (U.S. EPA, 2019, Section 11.1.4).
Additionally, the low (r < 0.4) to moderate correlations (r = 0.4-0.7) between PM2.5 and gaseous
pollutants and PM10-2.5 increase the confidence in PM2.5 having an independent effect on
mortality (U.S. EPA, 2019, section 11.1.4).
The generally positive associations reported with mortality are supported by a small group
of studies employing causal inference or quasi-experimental statistical approaches (U.S. EPA,
2019, section 11.1.2.1). For example, two studies by Schwartz et al. (Schwartz et al., 2015;
Schwartz et al., 2017) report associations between PM2.5 instrumental variables and mortality
(U.S. EPA, 2019, Table 11-2), including in an analysis limited to days with 24-hour average
PM2.5 concentrations <30 [j,g/m3 (Schwartz et al., 2017). In addition to the main analyses, these
studies conducted Granger-like causality tests as sensitivity analyses to examine whether there
was evidence of an association between mortality and PM2.5 after the day of death, which would
support the possibility that unmeasured confounders were not accounted for in the statistical
model. Neither study reports evidence of an association with PM2.5 after death (i.e., they do not
indicate unmeasured confounding). A recent quasi-experimental study examines whether a
specific regulatory action in Tokyo, Japan (i.e., a diesel emission control ordinance) resulted in a
subsequent reduction in daily mortality (Yorifuji et al., 2016). The authors report a reduction in
mortality in Tokyo due to the ordinance, compared to Osaka, which did not have a similar diesel
emission control ordinance in place.
The positive associations for total mortality reported across the majority of studies
evaluated are further supported by analyses reporting generally consistent, positive associations
with both cardiovascular and respiratory mortality (U.S. EPA, 2019, section 11.1.3). For both
cardiovascular and respiratory mortality, there has been only limited assessment of potential
copollutant confounding, though initial evidence indicates that associations remain positive and
relatively unchanged in models with gaseous pollutants and PM10-2.5. This evidence further
supports the copollutant analyses conducted for total mortality. The strong evidence for ischemic
events and heart failure, as detailed in the assessment of cardiovascular morbidity (U.S. EPA,
2019, Chapter 6), provides biological plausibility for PIVh.s-related cardiovascular mortality,
which comprises the largest percentage of total mortality (i.e., -33%) (NHLBI, 2017). Although
there is evidence for exacerbations of COPD and asthma, the collective body of respiratory
morbidity evidence provides only limited biological plausibility for PIVh.s-related respiratory
mortality (U.S. EPA, 2019, Chapter 5).
In the 2009 ISA, one of the main uncertainties identified was the regional and city-to-city
heterogeneity in PM2.5-mortality associations. Recent studies examine both city-specific as well
3-24

-------
as regional characteristics to identify the underlying contextual factors that could contribute to
this heterogeneity (U.S. EPA, 2019, section 11.1.6.3). Analyses focusing on effect modification
of the PM2.5-mortality relationship by PM2.5 components, regional patterns in PM2.5 components
and city-specific differences in composition and sources indicate some differences in the PM2.5
composition and sources across cities and regions, but these differences do not fully explain the
observed heterogeneity. Additional studies find that factors related to potential exposure
differences, such housing stock and commuting, as well as city-specific factors (e.g., land-use,
port volume, and traffic information), may explain some of the observed heterogeneity (U.S.
EPA, 2019, section 11.1.6.3). Collectively, recent studies indicate that the heterogeneity in
PM2.5-mortality risk estimates cannot be attributed to one factor, but instead a combination of
factors including, but not limited to, PM composition and sources as well as community
characteristics that could influence exposures (U.S. EPA, 2019, section 11.1.12).
A number of recent studies conducted systematic evaluations of the lag structure of
associations for the PM2.5-mortality relationship by examining either a series of single-day or
multiday lags and these studies continue to support an immediate effect (i.e., lag 0 to 1 days) of
short-term PM2.5 exposures on mortality (U.S. EPA, 2019, section 11.1.8.1). Recent studies also
conducted analyses comparing the traditional 24-hour average exposure metric with a sub-daily
metric (i.e., 1-hour max). These initial studies provide evidence of a similar pattern of
associations for both the 24-hour average and 1-hour max metric, with the association larger in
magnitude for the 24-hour average metric.
Recent multicity studies indicate that positive and statistically significant associations
with mortality persist in analyses restricted to short-term PM2.5 exposures below 35 |j,g/m3 (Lee
et al., 2015),19 below 30 |j,g/m3 (Shi et al., 2016), and below 25 |j,g/m3 (Di et al., 2017a),
indicating that risks associated with short-term PM2.5 exposures are not disproportionately driven
by the peaks of the air quality distribution. Additional studies examine the shape of the
concentration-response relationship and whether a threshold exists specifically for PM2.5 (U.S.
EPA, 2019, section 11.1.10). These studies have used various statistical approaches and
consistently demonstrate a linear relationship with no evidence of a threshold. Recent analyses
provide initial evidence indicating that PM2.5-mortality associations persist and may be stronger
(i.e., a steeper slope) at lower concentrations (e.g., Di et al., 2017a; Figure 11-12 in U.S. EPA,
2019). However, given the limited data available at the lower end of the distribution of ambient
PM2.5 concentrations, the shape of the concentration-response curve remains uncertain at these
low concentrations and, to date, studies have not conducted extensive analyses exploring
19 Lee et al. (2015) also report that positive and statistically significant associations between short-term PM2 5
exposures and mortality persist in analyses restricted to areas with long-term concentrations below 12 |.ig/m\
3-25

-------
alternatives to linearity when examining the shape of the PIVh.s-mortality concentration-response
relationship.
Overall, recent epidemiologic studies build upon and extend the conclusions of the 2009
ISA for the relationship between short-term PM2.5 exposures and total mortality. Supporting
evidence for PM2.5-related cardiovascular morbidity, and more limited evidence from respiratory
morbidity, provides biological plausibility for mortality due to short-term PM2.5 exposures. The
primarily positive associations observed across studies conducted in diverse geographic locations
is further supported by the results from co-pollutant analyses indicating robust associations,
along with evidence from analyses of the concentration-response relationship. The ISA states
that, collectively, "this body of evidence is sufficient to conclude that a causal relationship exists
between short-term PM2.5 exposure and total mortality" (U.S. EPA, 2019, pp. 11-58).
3.2.1.2 Cardiovascular Effects
Long-term PM2.5 exposures
The scientific evidence reviewed in the 2009 PM ISA was "sufficient to infer a causal
relationship between long-term PM2.5 exposure and cardiovascular effects" (U.S. EPA, 2009).
The strongest line of evidence comprised findings from several large epidemiologic studies of
U.S. cohorts that consistently showed positive associations between long-term PM2.5 exposure
and cardiovascular mortality (Pope et al., 2004, Krewski et al., 2009, Miller et al., 2007, Laden et
al., 2006). Studies of long-term PM2.5 exposure and cardiovascular morbidity were limited in
number. Biological plausibility and coherence with the epidemiologic findings were provided by
studies using genetic mouse models of atherosclerosis demonstrating enhanced atherosclerotic
plaque development and inflammation, as well as changes in measures of impaired heart
function, following 4- to 6-month exposures to PM2.5 concentrated ambient particles (CAPs), and
by a limited number of studies reporting CAPs-induced effects on coagulation factors, vascular
reactivity, and worsening of experimentally induced hypertension in mice (U.S. EPA, 2009).
Consistent with the evidence assessed in the 2009 PM ISA, the 2019 ISA concludes that
recent studies, together with the evidence available in previous reviews, support a causal
relationship between long-term exposure to PM2.5 and cardiovascular effects. As discussed above
(section 3.2.1.1), results from recent U.S. and Canadian cohort studies consistently report
positive associations between long-term PM2.5 exposure and cardiovascular mortality (U.S. EPA,
2019, Figure 6-19) in evaluations conducted at varying spatial scales and employing a variety of
exposure assessment and statistical methods (U.S. EPA, 2019, section 6.2.10). Positive
associations between long-term PM2.5 exposures and cardiovascular mortality are generally
robust in copollutant models adjusted for ozone, NO2, PM10-2.5, or SO2. In addition, most of the
results from analyses examining the shape of the concentration-response relationship for
3-26

-------
cardiovascular mortality support a linear relationship with long-term PM2.5 exposures and do not
identify a threshold below which effects do not occur (U.S. EPA, 2019, section 6.2.16; Table 6-
52).20
The body of literature examining the relationship between long-term PM2.5 exposure and
cardiovascular morbidity has greatly expanded since the 2009 PM ISA, with positive
associations reported in several cohorts (U.S. EPA, 2019, section 6.2). Though results for
cardiovascular morbidity are less consistent than those for cardiovascular mortality (U.S. EPA,
2019, section 6.2), recent studies provide some evidence for associations between long-term
PM2.5 exposures and the progression of cardiovascular disease. Positive associations with
cardiovascular morbidity (e.g., coronary heart disease, stroke) and atherosclerosis progression
are observed in several epidemiologic studies (U.S. EPA, 2019, sections 6.2.2. to 6.2.9).
Associations in such studies are supported by toxicological evidence for increased plaque
progression in mice following long-term exposure to PM2.5 collected from multiple locations
across the U.S. (U.S. EPA, 2019, section 6.2.4.2). A small number of epidemiologic studies also
report positive associations between long-term PM2.5 exposure and heart failure, changes in
blood pressure, and hypertension (U.S. EPA, 2019, sections 6.2.5 and 6.2.7). Associations with
heart failure are supported by animal toxicological studies demonstrating decreased cardiac
contractility and function, and increased coronary artery wall thickness following long-term
PM2.5 exposure (U.S. EPA, 2019, section 6.2.5.2). Similarly, a limited number of animal
toxicological studies demonstrating a relationship between long-term exposure to PM2.5 and
consistent increases in blood pressure in rats and mice are coherent with epidemiologic studies
reporting positive associations between long-term exposure to PM2.5 and hypertension.
Longitudinal epidemiologic analyses also report positive associations with markers of
systemic inflammation (U.S. EPA, 2019, section 6.2.11), coagulation (U.S. EPA, 2019, section
6.2.12), and endothelial dysfunction (U.S. EPA, 2019, section 6.2.13). These results are coherent
with animal toxicological studies generally reporting increased markers of systemic
inflammation, oxidative stress, and endothelial dysfunction (U.S. EPA, 2019, section 6.2.12.2
and 6.2.14).
In summary, the ISA concludes that there is consistent evidence from multiple
epidemiologic studies illustrating that long-term exposure to PM2.5 is associated with mortality
from cardiovascular causes. Associations with CHD, stroke and atherosclerosis progression were
observed in several additional epidemiologic studies providing coherence with the mortality
findings. Results from copollutant models generally support the independence of the PM2.5
associations. Additional evidence of the independent effect of PM2.5 on the cardiovascular
20 As noted above for mortality, uncertainty in the shape of the concentration-response relationship increases near
the upper and lower ends of the distribution due to limited data.
3-27

-------
system is provided by experimental studies in animals, which demonstrate biologically plausible
pathways by which long-term inhalation exposure to PM2.5 could potentially result in outcomes
such as CHD, stroke, CHF and cardiovascular mortality. The combination of epidemiologic and
experimental evidence results in the ISA conclusion that "a causal relationship exists between
long-term exposure to PM2.5 and cardiovascular effects" (U.S. EPA, 2019, section 6.2.18).
Short-term PM2.5 exposures
The 2009 PM ISA concluded that "a causal relationship exists between short-term
exposure to PM2.5 and cardiovascular effects" (U.S. EPA, 2009). The strongest evidence in the
2009 PM ISA was from epidemiologic studies of ED visits and hospital admissions for IHD and
HF, with supporting evidence from epidemiologic studies of cardiovascular mortality (U.S. EPA,
2009). Animal toxicological studies provided coherence and biological plausibility for the
positive associations reported with myocardial ischemia ED visit and hospital admissions. These
included studies reporting reduced myocardial blood flow during ischemia and studies indicating
altered vascular reactivity. In addition, effects of PM2.5 exposure on a potential indicator of
ischemia (i.e., ST segment depression on an electrocardiogram) were reported in both animal
toxicological and epidemiologic panel studies.21 Key uncertainties from the last review resulted
from inconsistent results across disciplines with respect to the relationship between short-term
exposure to PM2.5 and changes in blood pressure, blood coagulation markers, and markers of
systemic inflammation. In addition, while the 2009 PM ISA identified a growing body of
evidence from controlled human exposure and animal toxicological studies, uncertainties
remained with respect to biological plausibility.
A large body of recent evidence confirms and extends the evidence from the 2009 ISA
indicating that there is a causal relationship between short-term PM2.5 exposure and
cardiovascular effects. This includes generally positive associations observed in multicity
epidemiologic studies of emergency department visits and hospital admissions for ischemic heart
disease (IHD), heart failure (HF), and combined cardiovascular-related endpoints. In particular,
nationwide studies of older adults (65 years and older) using Medicare records report positive
associations between PM2.5 exposures and hospital admissions for HF (U.S. EPA, 2019,
section 6.1.3.1). Additional multicity studies conducted in the northeast U.S. report positive
associations between short-term PM2.5 exposures and emergency department visits or hospital
admissions for IHD (U.S. EPA, 2019, section 6.1.2.1) while studies conducted in the U.S. and
Canada reported positive associations between short-term PM2.5 exposures and emergency
department visits for HF. Epidemiologic studies conducted in single cities contribute some
21 Some animal studies included in the 2009 PM ISA examined exposures to mixtures, such as motor vehicle
exhaust or woodsmoke. In these studies, it was unclear if the resulting cardiovascular effects could be attributed
specifically to the particulate components of the mixture.
3-28

-------
support, though associations reported in single-city studies are less consistently positive than in
multicity studies, and include a number of studies reporting null associations (U.S. EPA, 2019,
sections 6.1.2 and 6.1.3). When considered as a whole; however, the recent body of IHD and HF
epidemiologic evidence supports the evidence from previous ISAs reporting mainly positive
associations between short-term PM2.5 concentrations and emergency department visits and
hospital admissions.
In addition, a number of more recent controlled human exposure, animal toxicological,
and epidemiologic panel studies provide evidence that PM2.5 exposure could plausibly result in
IHD or HF through pathways that include endothelial dysfunction, arterial thrombosis, and
arrhythmia (U.S. EPA, 2019, section 6.1.1). The most consistent evidence from recent controlled
human exposure studies is for endothelial dysfunction, as measured by changes in brachial artery
diameter or flow mediated dilation. All but one of the available controlled human exposure
studies examining the potential for endothelial dysfunction report an effect of PM2.5 exposure on
measures of blood flow (U.S. EPA, 2019, section 6.1.13.2). These studies report variable results
regarding the timing of the effect and the mechanism by which reduced blood flow occurs
(i.e., availability vs sensitivity to nitric oxide). Some controlled human exposure studies using
CAPs report evidence for small increases in blood pressure (U.S. EPA, 2019, section 6.1.6.3). In
addition, although not entirely consistent, there is also some evidence across controlled human
exposure studies for conduction abnormalities/arrhythmia (U.S. EPA, 2019, section 6.1.4.3),
changes in heart rate variability (HRV) (U.S. EPA, 2019, section 6.1.10.2), changes in
hemostasis that could promote clot formation (U.S. EPA, 2019, section 6.1.12.2), and increases
in inflammatory cells and markers (U.S. EPA, 2019, section 6.1.11.2). Thus, when taken as a
whole, controlled human exposure studies are coherent with epidemiologic studies in that they
demonstrate short-term exposures to PM2.5 may result in the types of cardiovascular endpoints
that could lead to emergency department visits and hospital admissions in some people.
Animal toxicological studies published since the 2009 ISA also support a relationship
between short-term PM2.5 exposure and cardiovascular effects. A recent study demonstrating
decreased cardiac contractility and left ventricular pressure in mice is coherent with the results of
epidemiologic studies reporting associations between short-term PM2.5 exposure and heart failure
(U.S. EPA, 2019, section 6.1.3.3). In addition, and as with controlled human exposure studies,
there is generally consistent evidence in animal toxicological studies for indicators of endothelial
dysfunction (U.S. EPA, 2019, section 6.1.13.3). Studies in animals also provide evidence for
changes in a number of other cardiovascular endpoints following short-term PM2.5 exposure.
Although not entirely consistent, these studies provide some evidence of conduction
abnormalities and arrhythmia (U.S. EPA, 2019, section 6.1.4.4), changes in HRV (U.S. EPA,
3-29

-------
2019, section 6.1.10.3), changes in blood pressure (U.S. EPA, 2019, section 6.1.6.4), and
evidence for systemic inflammation and oxidative stress (U.S. EPA, 2019, section 6.1.11.3).
In summary, recent evidence further supports and extends the conclusions of the evidence
base reported in the 2009 ISA. In support of epidemiologic studies reporting robust associations
in copollutant models, direct evidence for an independent effect of PM2.5 on cardiovascular
effects can be found in a number of controlled human exposure and animal toxicological studies.
Coherent with these results are epidemiologic panel studies reporting that PM2.5 exposure is
associated with some of the same cardiovascular endpoints reported in experimental studies. For
these effects, there are inconsistencies in results across some animal toxicological, controlled
human exposure, and epidemiologic panel studies, though this may be due to substantial
differences in study design and/or study populations. Overall, the results from epidemiologic
panel, controlled human exposure, and animal toxicological studies, in particular those related to
endothelial dysfunction, impaired cardiac function, ST segment depression, thrombosis,
conduction abnormalities, and changes in blood pressure provide coherence and biological
plausibility for the consistent results from epidemiologic studies observing positive associations
between short-term PM2.5 concentrations and IHD and HF, and ultimately cardiovascular
mortality. The ISA concludes that, overall, "there continues to be sufficient evidence to conclude
that a causal relationship exists between short-term PM2.5 exposure and cardiovascular effects"
(U.S. EPA, 2019, p. 6-138).
3.2.1.3 Respiratory Effects
Long-term PM2 5 exposures
The 2009 PM ISA concluded that "a causal relationship is likely to exist between
long-term PM2.5 exposure and respiratory effects" (U.S. EPA, 2009). This conclusion was based
mainly on epidemiologic evidence demonstrating associations between long-term PM2.5
exposure and changes in lung function or lung function growth in children. Biological
plausibility was provided by a single animal toxicological study examining pre- and post-natal
exposure to PM2.5 CAPs, which found impaired lung development. Epidemiologic evidence for
associations between long-term PM2.5 exposure and other respiratory outcomes, such as the
development of asthma, allergic disease, and COPD; respiratory infection; and the severity of
disease was limited, both in the number of studies available and the consistency of the results.
Experimental evidence for other outcomes was also limited, with one animal toxicological study
reporting that long-term exposure to PM2.5 CAPs results in morphological changes in nasal
airways of healthy animals. Other animal studies examined exposure to mixtures, such as motor
vehicle exhaust and woodsmoke, and effects were not attributed specifically to the particulate
components of the mixture.
3-30

-------
Recent cohort studies provide additional support for the relationship between long-term
PM2.5 exposure and decrements in lung function growth (as a measure of lung development),
indicating a robust and consistent association across study locations, exposure assessment
methods, and time periods (U.S. EPA, 2019, section 5.2.13). This relationship is further
supported by a recent retrospective study that reports an association between declining PM2.5
concentrations and improvements in lung function growth in children (U.S. EPA, 2019,
section 5.2.11). Epidemiologic studies also examine asthma development in children (U.S. EPA,
2019, section 5.2.3), with recent prospective cohort studies reporting generally positive
associations, though several are imprecise (i.e., they report wide confidence intervals).
Supporting evidence is provided by studies reporting associations with asthma prevalence in
children, with childhood wheeze, and with exhaled nitric oxide, a marker of pulmonary
inflammation (U.S. EPA, 2019, section 5.2.13). A recent animal toxicological study showing the
development of an allergic phenotype and an increase in a marker of airway responsiveness
provides biological plausibility for allergic asthma (U.S. EPA, 2019, section 5.2.13). Other
epidemiologic studies report a PIVh.s-related acceleration of lung function decline in adults, while
improvement in lung function was observed with declining PM2.5 concentrations (U.S. EPA,
2019, section 5.2.11). A recent longitudinal study found declining PM2.5 concentrations are also
associated with an improvement in chronic bronchitis symptoms in children, strengthening
evidence reported in the 2009 ISA for a relationship between increased chronic bronchitis
symptoms and long-term PM2.5 exposure (U.S. EPA, 2019, section 5.2.11). A common
uncertainty across the epidemiologic evidence is the lack of examination of copollutants to
assess the potential for confounding. While there is some evidence that associations remain
robust in models with gaseous pollutants, a number of these studies examining copollutant
confounding were conducted in Asia, and thus have limited generalizability due to high annual
pollutant concentrations.
When taken together, the ISA concludes that the "epidemiologic evidence strongly
supports a relationship with decrements in lung function growth in children" and "with asthma
development in children, with increased bronchitic symptoms in children with asthma, with an
acceleration of lung function decline in adults, and with respiratory mortality and cause-specific
respiratory mortality for COPD and respiratory infection" (U.S. EPA, 2019, p. 1-34). In support
of the biological plausibility of such associations reported in epidemiologic studies of respiratory
health effects, animal toxicological studies continue to provide direct evidence that long-term
exposure to PM2.5 results in a variety of respiratory effects. Recent animal studies show
pulmonary oxidative stress, inflammation, and morphologic changes in the upper (nasal) and
lower airways. Other results show that changes are consistent with the development of allergy
and asthma, and with impaired lung development. Overall, the ISA concludes that "the collective
3-31

-------
evidence is sufficient to conclude that a causal relationship is likely to exist between long-term
PM2.5 exposure and respiratory effects" (U.S. EPA, 2019, section 5.2.13).
Short-term PM2.5 exposures
The 2009 PM ISA (U.S. EPA, 2009) concluded that a "causal relationship is likely to
exist" between short-term PM2.5 exposure and respiratory effects. This conclusion was based
mainly on the epidemiologic evidence demonstrating positive associations with various
respiratory effects. Specifically, the 2009 ISA described epidemiologic evidence as consistently
showing PM2.5-associated increases in hospital admissions and emergency department visits for
chronic obstructive pulmonary disease (COPD) and respiratory infection among adults or people
of all ages, as well as increases in respiratory mortality. These results were supported by studies
reporting associations with increased respiratory symptoms and decreases in lung function in
children with asthma, though the epidemiologic evidence was inconsistent for hospital
admissions or emergency department visits for asthma. Studies examining copollutant models
showed that PM2.5 associations with respiratory effects were robust to inclusion of CO or SO2 in
the model, but often were attenuated (though still positive) with inclusion of O3 or NO2. In
addition to the copollutant models, evidence supporting an independent effect of PM2.5 exposure
on the respiratory system was provided by animal toxicological studies of PM2.5 CAPs
demonstrating changes in some pulmonary function parameters, as well as inflammation,
oxidative stress, injury, enhanced allergic responses, and reduced host defenses. Many of these
effects have been implicated in the pathophysiology for asthma exacerbation, COPD
exacerbation, or respiratory infection. In the few controlled human exposure studies conducted in
individuals with asthma or COPD, PM2.5 exposure mostly had no effect on respiratory
symptoms, lung function, or pulmonary inflammation. Available studies in healthy people also
did not clearly demonstrate respiratory effects following short-term PM2.5 exposures.
Recent epidemiologic studies provide evidence for a relationship between short-term
PM2.5 exposure and several respiratory-related endpoints, including asthma exacerbation (U.S.
EPA, 2019, section 5.1.2.1), COPD exacerbation (U.S. EPA, 2019, section 5.1.4.1), and
combined respiratory-related diseases (U.S. EPA, 2019, section 5.1.6), particularly from studies
examining emergency department visits and hospital admissions. The generally positive
associations between short-term PM2.5 exposure and asthma and COPD emergency department
visits and hospital admissions are supported by epidemiologic studies demonstrating associations
with other respiratory-related effects such as symptoms and medication use that are indicative of
asthma and COPD exacerbations (U.S. EPA, 2019, sections 5.1.2.2 and 5.4.1.2). The collective
body of epidemiologic evidence for asthma exacerbation is more consistent in children than in
adults. Additionally, epidemiologic studies examining the relationship between short-term PM2.5
3-32

-------
exposure and respiratory mortality provide evidence of consistent positive associations,
demonstrating a continuum of effects (U.S. EPA, 2019, section 5.1.9).
Building off the studies evaluated in the 2009 ISA, recent epidemiologic studies expand
the assessment of potential copollutant confounding. There is some evidence that PM2.5
associations with asthma exacerbation, combined respiratory-related diseases, and respiratory
mortality remain relatively unchanged in copollutant models with gaseous pollutants (i.e., O3,
NO2, SO2, with more limited evidence for CO) and other particle sizes (i.e., PM10-2.5) (U.S. EPA,
2019, section 5.1.10.1).
The uncertainty related to whether there is an independent effect of PM2.5 on respiratory
health is also partially addressed by findings from animal toxicological studies. Specifically,
short-term exposure to PM2.5 enhanced asthma-related responses in an animal model of allergic
airways disease and enhanced lung injury and inflammation in an animal model of COPD (U.S.
EPA, 2019, sections 5.1.2.4.4 and 5.1.4.4.3). The experimental evidence provides biological
plausibility for some respiratory-related endpoints, including limited evidence of altered host
defense and greater susceptibility to bacterial infection as well as consistent evidence of
respiratory irritant effects. Animal toxicological evidence for other respiratory effects is
inconsistent.
The ISA concludes that "[t]he strongest evidence of an effect of short-term PM2.5
exposure on respiratory effects is provided by epidemiologic studies of asthma and COPD
exacerbation. While animal toxicological studies provide biological plausibility for these
findings, some uncertainty remains with respect to the independence of PM2.5 effects" (U.S.
EPA, 2019, p. 5-155). When taken together, the ISA concludes that this evidence "is sufficient to
conclude that a causal relationship is likely to exist between short-term PM2.5 exposure and
respiratory effects" (U.S. EPA, 2019, p. 5-155).
3.2.1.4 Cancer - Long-term PM2.5 Exposures
The 2009 ISA concluded that the overall body of evidence was "suggestive of a causal
relationship between relevant PM2.5 exposures and cancer" (U.S. EPA, 2009). This conclusion
was based primarily on positive associations observed in a limited number of epidemiologic
studies of lung cancer mortality. The few epidemiologic studies that had evaluated PM2.5
exposure and lung cancer incidence or cancers of other organs and systems generally did not
show evidence of an association. Toxicological studies did not focus on exposures to specific
PM size fractions, but rather investigated the effects of exposures to total ambient PM, or other
source-based PM such as wood smoke. Collectively, results of in vitro studies were consistent
with the larger body of evidence demonstrating that ambient PM and PM from specific
combustion sources are mutagenic and genotoxic. However, animal inhalation studies found
3-33

-------
little evidence of tumor formation in response to chronic exposures. A small number of studies
provided preliminary evidence that PM exposure can lead to changes in methylation of DNA,
which may contribute to biological events related to cancer.
Since the 2009 ISA, additional cohort studies provide evidence that long-term PM2.5
exposure is positively associated with lung cancer mortality and with lung cancer incidence, and
provide initial evidence for an association with reduced cancer survival (U.S. EPA, 2019, section
10.2.5). Reanalyses of the ACS cohort using different years of PM2.5 data and follow-up, along
with various exposure assignment approaches, provide consistent evidence of positive
associations between long-term PM2.5 exposure and lung cancer mortality (U.S. EPA, 2019,
Figure 10-3). Additional support for positive associations with lung cancer mortality is provided
by recent epidemiologic studies using individual-level data to control for smoking status, by
studies of people who have never smoked (though such studies generally report wide confidence
intervals due to the small number of lung cancer mortality cases within this population), and in
analyses of cohorts that relied upon proxy measures to account for smoking status (U.S. EPA,
2019, section 10.2.5.1.1). Although studies that have evaluated lung cancer incidence, including
studies of people who have never smoked, are limited in number, recent studies generally report
positive associations with long-term PM2.5 exposures (U.S. EPA, 2019, section 10.2.5.1.2). A
subset of the studies focusing on lung cancer incidence also examined histological subtype,
providing some evidence of positive associations for adenocarcinomas, the predominate subtype
of lung cancer observed in people who have never smoked (U.S. EPA, 2019, section 10.2.5.1.2).
Associations between long-term PM2.5 exposure and lung cancer incidence were found to remain
relatively unchanged, though in some cases confidence intervals widened, in analyses that
attempted to reduce exposure measurement error by accounting for length of time at residential
address or by examining different exposure assignment approaches (U.S. EPA, 2019, section
10.2.5.1.2).
The ISA evaluates the degree to which recent epidemiologic studies have addressed the
potential for confounding by copollutants and the shape of the concentration-response
relationship. To date, relatively few studies have evaluated the potential for copollutant
confounding of the relationship between long-term PM2.5 exposure and lung cancer mortality or
incidence. The small number of such studies have generally focused on O3 and report that PM2.5
associations remain relatively unchanged in copollutant models (U.S. EPA, 2019, section
10.2.5.1.3).	However, available studies have not systematically evaluated the potential for
copollutant confounding by other gaseous pollutants or by other particle size fractions (U.S.
EPA, 2019, section 10.2.5.1.3). Compared to total (non-accidental) mortality (see section
3.2.1.1), fewer studies have examined the shape of the concentration-response curve for
cause-specific mortality outcomes, including lung cancer. Several studies have reported no
3-34

-------
evidence of deviations from linearity in the shape of the concentration-response relationship
(Lepeule et al., 2012; Raaschou-Nielsen et al., 2013; Puett et al., 2014), though authors provided
only limited discussions of results (U.S. EPA, 2019, section 10.2.5.1.4).
In support of the biological plausibility of an independent effect of PM2.5 on cancer, the
ISA notes evidence from recent experimental and epidemiologic studies demonstrating that
PM2.5 exposure can lead to a range of effects indicative of mutagenicity, genotoxicity, and
carcinogenicity, as well as epigenetic effects (U.S. EPA, 2019, section 10.2.7). For example,
both in vitro and in vivo toxicological studies have shown that PM2.5 exposure can result in DNA
damage (U.S. EPA, 2019, section 10.2.2). Although such effects do not necessarily equate to
carcinogenicity, the evidence that PM exposure can damage DNA, and elicit mutations, provides
support for the plausibility of epidemiologic associations with lung cancer mortality and
incidence. Additional supporting studies indicate the occurrence of micronuclei formation and
chromosomal abnormalities (U.S. EPA, 2019, section 10.2.2.3), and differential expression of
genes that may be relevant to cancer pathogenesis, following PM exposures. Experimental and
epidemiologic studies that examine epigenetic effects indicate changes in DNA methylation,
providing some support for PM2.5 exposure contributing to genomic instability (U.S. EPA, 2019,
section 10.2.3).
Epidemiologic evidence for associations between PM2.5 and lung cancer mortality and
incidence, together with evidence supporting the biological plausibility of such associations,
contributes to the ISA's conclusion that the evidence "is sufficient to conclude that a causal
relationship is likely to exist between long-term PM2.5 exposure and cancer" (U.S. EPA, 2019,
section 10.2.7).
3.2.1.5 Nervous System Effects
Long-term PM2.5 exposures
Reflecting the very limited evidence available in the last review, the 2009 ISA did not
make a causality determination for long-term PM2.5 exposures and nervous system effects (U.S.
EPA, 2009). Since the last review, this body of evidence has grown substantially (U.S. EPA,
2019, section 8.2). Recent animal toxicology studies report that long-term PM2.5 exposures can
lead to morphologic changes in the hippocampus and to impaired learning and memory. This
evidence is consistent with epidemiologic studies reporting that long-term PM2.5 exposure is
associated with reduced cognitive function (U.S. EPA, 2019, section 8.2.5). Further, while the
evidence is limited, the presence of early markers of Alzheimer's disease pathology has been
demonstrated in rodents following long-term exposure to PM2.5 CAPs. These findings support
reported associations with neurodegenerative changes in the brain (i.e., decreased brain volume),
all-cause dementia, or hospitalization for Alzheimer's disease in a small number of
3-35

-------
epidemiologic studies (U.S. EPA, 2019, section 8.2.6). Additionally, loss of dopaminergic
neurons in the substantia nigra, a hallmark of Parkinson disease, has been reported in mice (U.S.
EPA, 2019, section 8.2.4), though epidemiologic studies provide only limited support for
associations with Parkinson's disease (U.S. EPA, 2019, section 8.2.6). Overall, the lack of
consideration of copollutant confounding introduces some uncertainty in the interpretation of
epidemiologic studies of nervous system effects, but this uncertainty is partly addressed by the
evidence for an independent effect of PM2.5 exposures provided by experimental animal studies.
In addition to the findings described above, which are most relevant to older adults,
several recent studies of neurodevelopmental effects in children have also been conducted.
Positive associations between long-term exposure to PM2.5 during the prenatal period and autism
spectrum disorder (ASD) are observed in multiple epidemiologic studies (U.S. EPA, 2019,
section 8.2.7.2), while studies of cognitive function provide little support for an association (U.S.
EPA, 2019, section 8.2.5.2). Interpretation of these epidemiologic studies is limited due to the
small number of studies, their lack of control for potential confounding by copollutants, and
uncertainty regarding the critical exposure windows. Biological plausibility is provided for the
ASD findings by a study in mice that found inflammatory and morphologic changes in the
corpus collosum and hippocampus, as well as ventriculomegaly (i.e., enlarged lateral ventricles)
in young mice following prenatal exposure to PM2.5 CAPs.
Taken together, the ISA concludes that recent studies indicate long-term PM2.5 exposures
can lead to effects on the brain associated with neurodegeneration (i.e., neuroinflammation and
reductions in brain volume), as well as cognitive effects in older adults (U.S. EPA, 2019, Table
1-2). Animal toxicology studies provide evidence for a range of nervous system effects in adult
animals, including neuroinflammation and oxidative stress, neurodegeneration, and cognitive
effects, and effects on neurodevelopment in young animals. The epidemiologic evidence is more
limited but studies generally support associations between long-term PM2.5 exposure and
changes in brain morphology, cognitive decrements and dementia. There is also initial, and
limited, evidence for neurodevelopmental effects, particularly ASD. The consistency and
coherence of the evidence supports the ISA's conclusion that "the collective evidence is
sufficient to conclude that a causal relationship is likely to exist between long-term PM2.5
exposure and nervous system effects" (U.S. EPA, 2019, section 8.2.9).
3.2.1.6 Other Effects
Compared to the health outcomes discussed above, the ISA concludes that there is greater
uncertainty in the evidence linking PM2.5, or UFP, exposures with other health outcomes,
reflected in conclusions that the evidence is "suggestive of, but not sufficient to infer, a causal
relationship." The sections below summarize the daft ISA conclusions for these "suggestive"
3-36

-------
outcomes for long-term (Section 3.2.1.6.1) and short-term (Section 3.2.1.6.2) PM2.5 and UFP
exposures.
3.2.1.6.1 Long-term Exposures
As indicated in Table 3-1 above, the ISA concludes that the evidence is "suggestive of,
but not sufficient to infer, a causal relationship" between long-term PM2.5 exposures and
metabolic effects and reproductive and developmental effects (reproduction and fertility;
pregnancy and birth outcomes). These conclusions reflect evidence that is "generally supportive
but not entirely consistent or is limited overall" where "[cjhance, confounding, and other biases
cannot be ruled out" (U.S. EPA, 2019, Preface, p. P-20). The basis for these causality
determinations is summarized briefly below.
PM2.5 - Metabolic effects
There were no causality determinations for long-term PM2.5 exposure and metabolic
effects in the 2009 ISA (U.S. EPA, 2009). However, the literature pertaining to the effect of
long-term exposure to PM2.5 and metabolic effects has expanded substantially since the 2009
ISA, and consists of both epidemiologic and experimental evidence (U.S. EPA, 2019, section
7.2). Epidemiologic studies report positive associations between long-term PM2.5 exposure and
diabetes-related mortality. In addition, although results were not consistent across cohorts, there
is some evidence from epidemiologic studies for positive associations with incident diabetes,
metabolic syndrome, and alterations in glucose and insulin homeostasis. Consideration of
copollutant confounding was limited. In animal toxicologic studies, there is some support for a
relationship between long-term PM2.5 exposure and metabolic effects from experimental studies
demonstrating increased blood glucose, insulin resistance, and inflammation and visceral
adiposity but the experimental evidence was not entirely consistent. Based on this evidence, the
ISA concludes that, "[ojverall, the collective evidence is suggestive of, but is not sufficient to
infer, a causal relationship between long-term PM2.5 exposure and metabolic effects" (U.S. EPA,
2019, p. 7-52).
PM2.5 - Reproductive and developmental effects
The 2009 ISA determined that the evidence was "suggestive of a causal relationship" for
the association between long-term PM2.5 exposure and reproductive and developmental
outcomes. The body of literature characterizing these relationships has grown since the 2009
ISA, with much of the evidence focusing on reproduction and fertility or pregnancy and birth
outcomes, though important uncertainties persist (U.S. EPA, 2019, sections 9.1.1, 9.1.2, 9.1.5).
Effects of PM2.5 exposure on sperm have been studied in both epidemiology and
toxicology studies and shows the strongest evidence in epidemiologic studies for impaired sperm
motility and in animal toxicological studies for impaired spermiation. Epidemiologic evidence on
3-37

-------
sperm morphology have reported inconsistent results. Evidence for effects of PM2.5 exposure on
female reproduction also comes from both epidemiology and toxicology studies. In the
epidemiologic literature, results on human fertility and fecundity is limited, but the evidence on
in vitro fertilization indicates a modest association of PM2.5 exposures with decreased odds of
becoming pregnant. Studies in rodents have shown ovulation and estrus are affected by PM2.5
exposure. Biological plausibility for outcomes related to male and female fertility and
reproduction comes from laboratory animal studies demonstrating genetic and epigenetic
changes in germ cells with PM2.5 exposure. The ISA concludes that, "[cjollectively, the evidence
is suggestive of, but not sufficient to infer, a causal relationship between PM2.5 exposure and
male and female reproduction and fertility" (U.S. EPA, 2019, p. 9-43).
With regard to pregnancy and birth outcomes, while the collective evidence for many of
the outcomes examined is not consistent, there are some animal toxicology and epidemiologic
studies that indicate an association between PM2.5 exposures and reduced fetal growth, low birth
weight and preterm birth. Most of the epidemiologic studies do not control for co-pollutant
confounding and do not identify a specific sensitive window of exposure, but results from animal
toxicologic studies provide biological plausibility for these outcomes, as well as support for
multiple sensitive windows for PM2.5 exposure-associated outcomes. There is also epidemiologic
evidence for congenital heart defects of different types, as well as biological plausibility to
support this outcome from the animal toxicology literature. However, evidence for a relationship
between PM2.5 exposure and various pregnancy-related pathologies, including gestational
hypertension, pre-eclampsia and gestational diabetes is inconsistent. Biological plausibility for
effects of PM2.5 exposure and various pregnancy and birth outcomes is provided by studies
showing that PM2.5 exposure in laboratory rodents resulted in impaired implantation and vascular
endothelial dysfunction. Coherence with toxicological studies is provided by epidemiologic
studies in humans reporting associations with epigenetic changes to the placenta and impaired
fetal thyroid function. When taken together, the ISA concludes that the available evidence,
including uncertainties that evidence, is "suggestive of, but not sufficient to infer, a causal
relationship between exposure to PM2.5 and pregnancy and birth outcomes" (U.S. EPA, 2019, p.
9-44).
UFP - Nervous System Effects
The 2009 ISA reported limited animal toxicological evidence of a relationship between
long-term exposure to UFP and nervous system effects, with no supporting epidemiologic
studies. Recent animal toxicological studies substantially add to this evidence base. Multiple
toxicological studies of long-term UFP exposure conducted in adult mice provide consistent
evidence of brain inflammation and oxidative stress in the whole brain, hippocampus, and
cerebral cortex (U.S. EPA, 2019, section 8.6.3). Studies also found morphologic changes,
3-38

-------
specifically neurodegeneration in specific regions of the hippocampus and pathologic changes
characteristic of Alzheimer's disease, and initial evidence of behavioral effects in adult mice
(U.S. EPA, 2019, sections 8.6.4 and 8.6.5). Toxicological studies examining pre- and post-natal
UFP exposures provide extensive evidence for behavioral effects, altered neurotransmitters,
neuroinflammation, and morphologic changes (U.S. EPA, 2019, section 8.6.6.2). Persistent
ventriculomegaly was observed in male, but not female, mice exposed postnatally to UFP (U.S.
EPA, 2019, section 8.6.6). Epidemiologic evidence is limited to a single study of school children
that provides support for the experimental results. This study, which did not consider copollutant
confounding, reports an association between long-term exposure to UFP, which was measured at
the school, and decrements on tests of attention and memory. Uncertainty results from the lack of
information on the spatial and temporal variability of UFP exposures on long-term UFP
exposures at the population level. Based primarily on the animal toxicological evidence of
neurotoxicity and altered neurodevelopment, the ISA concludes that the evidence is "suggestive
of, but not sufficient to infer, a causal relationship" between long-term UFP exposure and
nervous system effects (U.S. EPA, 2019, section 8.6.7).
3.2.1.6.2 Short-term Exposures
As indicated in Table 3-1 above, the ISA concludes that the evidence is "suggestive of,
but not sufficient to infer, a causal relationship" between short-term PM2.5 exposures and
metabolic effects and nervous system effects. Additionally, the ISA concludes that the evidence
is "suggestive" for short-term UFP exposures and cardiovascular effects, respiratory effects, and
nervous system effects. As for the outcomes related to long-term exposures, discussed above,
these conclusions reflect evidence that is "generally supportive but not entirely consistent or is
limited overall" where "[cjhance, confounding, and other biases cannot be ruled out" (U.S. EPA,
2019, Preface, p.P-20). The basis for these causality determinations is summarized briefly below.
I'M2.5 Metabolic effects
There were no studies of the effect of short-term PM2.5 exposure and metabolic effects
reviewed in the 2009 ISA (U.S. EPA, 2009). New evidence for a relationship between short-term
PM2.5 exposure and metabolic effects is based on a small number of epidemiologic and animal
toxicological studies reporting effects on glucose and insulin homeostasis and other indicators of
metabolic function such as inflammation in the visceral adipose tissue and liver (U.S. EPA,
2019, section 7.1). The ISA concludes that, overall, the collective evidence "is suggestive of, but
not sufficient to infer, a causal relationship between short-term PM2.5 exposure and metabolic
effects" (U.S. EPA, 2019, p. 7-11).
PM2.5 - Nervous system effects
3-39

-------
The evidence reviewed in the 2009 ISA was characterized as "inadequate to infer" a
causal relationship between short-term PM2.5 exposure and nervous system effects (U.S. EPA,
2009), based on a small number of experimental animal studies. Recent studies strengthen the
evidence that short-term exposure to PM2.5 can affect the nervous system (U.S. EPA, 2019,
section 8.1). The strongest evidence is provided by experimental studies in mice that show
effects on the brain. These toxicological studies demonstrate changes in neurotransmitters in the
hypothalamus that are linked to sympathetic nervous system and hypothalamic-pituitary-adrenal
(HP A) stress axis activation, as well as upregulation of inflammation-related genes, changes in
cytokine levels, and other changes that are indicative of brain inflammation. In addition, an
association of short-term PM2.5 exposure with hospital admissions for Parkinson's disease was
observed indicating the potential for exacerbation of neurological diseases. The ISA concludes
that, overall, the collective evidence "is suggestive of, but not sufficient to infer, a causal
relationship between short-term exposure to PM2.5 and nervous system effects" (U.S. EPA, 2019,
p. 8-15).
UFP - Cardiovascular effects
In the 2009 ISA, the evidence from toxicological studies, many of which examined
exposures to whole diesel exhaust or wood smoke rather than UFP alone, was suggestive of a
causal relationship between short-term UFP exposure and cardiovascular effects. Since the 2009
ISA, there have been only a limited number of studies published describing the relationship
between short-term UFP exposure and cardiovascular effects. This includes a small number of
epidemiologic panel studies that have observed positive associations between short-term
exposure to UFPs and measures of HRV (U.S. EPA, 2019, section 6.5.9.1) and markers of
coagulation (U.S. EPA, 2019, section 6.5.11.1) although there are also studies that did not report
such UFP-related effects. In addition, there is evidence from a single controlled human exposure
study indicating decreases in the anticoagulant proteins plasminogen and thrombomodulin in
individuals with metabolic syndrome (U.S. EPA, 2019, section 6.5.11.2). There is inconsistent
evidence from controlled human exposure and epidemiologic panel studies for endothelial
dysfunction, changes in blood pressure, and systemic inflammation following short-term
exposure to UFPs. Notably, there is little evidence of an effect when considering short-term UFP
exposure on other cardiovascular endpoints as well as cardiovascular-disease emergency
department visits or hospital admissions. The assessment of study results across experimental
and epidemiologic studies is complicated by differences in the size distributions examined
between disciplines and by the nonuniformity in the exposure metrics examined (e.g., particle
number concentration, surface area concentration, and mass concentration) (U.S. EPA, 2019,
section 1.4.3). When considered as a whole, the ISA concludes that the evidence is "suggestive
3-40

-------
of, but not sufficient to infer, a causal relationship between short-term exposure UFP exposure
and cardiovascular effects" (U.S. EPA, 2019, p. 6-304).
UFP - Respiratory effects
A limited number of studies examining short-term exposure to UFPs and respiratory
effects were reported in the 2009 ISA, which concluded that the relationship between short-term
exposure to UFP and respiratory effects is "suggestive of a causal relationship." This conclusion
was based on epidemiologic evidence indicating associations with combined respiratory-related
diseases, respiratory infection, and asthma exacerbation. In addition, personal exposures to
ambient UFP were associated with lung function decrements in adults with asthma. The few
available experimental studies provided limited coherence with epidemiologic findings for
asthma exacerbation. Recent studies add to this evidence base and support epidemiologic
evidence for asthma exacerbation and combined respiratory-related diseases but do not rule out
chance, confounding, and other biases (U.S. EPA, 2019, section 5.5). For example, associations
persist in one epidemiologic study with adjustment for NO2, but not in another. Additional
supporting evidence, showing decrements in lung function and enhancement of allergic
inflammation and other allergic responses, is provided by a controlled human exposure study in
adults with asthma and by animal toxicological studies in an animal model of allergic airway
disease. For combined respiratory-related diseases, recent findings add consistency for hospital
admissions and emergency department visits and indicate lung function changes among adults
with asthma or COPD. Uncertainty remains regarding the characterization of UFP exposures and
the potential for copollutant confounding in epidemiologic studies, which limits inference about
an independent effect of UFP exposures (U.S. EPA, 2019, section 5.5). The ISA concludes that,
overall, the evidence is "suggestive of, but not sufficient to infer, a causal relationship between
short-term UFP exposure and respiratory effects" (U.S. EPA, 2019, p. 5-303).
UFP- Nervous system effects
The 2009 ISA reported limited animal toxicological evidence of a relationship between
short-term exposure to UFP and nervous system effects, without supporting epidemiologic
studies. Several recent experimental studies add to this evidence base. In the current review, the
strongest evidence for a relationship between short-term UFP exposure and nervous system
effects is provided by animal toxicological studies that show inflammation and oxidative stress
in multiple brain regions following exposure to UFP. There is a lack of evidence from
epidemiologic studies (U.S. EPA, 2019, section 8.5). The ISA concludes that, overall, the
collective evidence is "suggestive of, but not sufficient to infer, a causal relationship between
short-term UFP exposure and nervous system effects" (U.S. EPA, 2019, p. 8-86).
3-41

-------
3.2.1.7 Summary
Based on the evidence assessed in the ISA (U.S. EPA, 2019), and summarized in sections
3.2.1.1 to 3.2.1.6 above, we revisit the policy-relevant questions posed at the beginning of this
section:
• To what extent does the currently available scientific evidence strengthen, or otherwise
alter, our conclusions from the last review regarding health effects attributable to long-
or short-term fine particle exposures? Have previously identified uncertainties been
reduced? What important uncertainties remain and have new uncertainties been
identified?
We consider these questions in the context of the evidence for effects of long- and short-term
PM2.5 exposures.
Studies conducted since the 2009 ISA have broadened our understanding of the health
effects that can result from long-term PM2.5 exposures and have reduced key uncertainties
identified in the last review. Recent epidemiologic studies consistently report positive
associations between long-term PM2.5 exposures and a wide range of health outcomes, including
total and cause-specific mortality, cardiovascular and respiratory morbidity, lung cancer, and
nervous system effects. Such associations have been reported in analyses examining a variety of
study designs, approaches to estimating PM2.5 exposures, statistical models, and long-term
exposure windows (i.e., the exposure period that is associated with the health outcome). Recent
evidence also includes retrospective studies that demonstrate improvements in health outcomes,
including increasing life expectancy, decreasing mortality, or decreasing respiratory effects, as a
result of past declines in ambient PM2.5 concentrations. Recent epidemiologic studies report that
associations with mortality (total, cardiovascular, and respiratory) remain relatively unchanged in
copollutant models, supporting the independence of these associations from co-occurring gases
or coarse PM. Recent studies additionally report that associations (i.e., primarily with mortality)
persist in analyses restricted to long-term PM2.5 exposures in the lower portions of the air quality
distribution, and such studies do not identify a threshold below which associations no longer
occur. The biological plausibility of health effect associations reported in epidemiologic studies
is supported by coherent results from experimental studies. Recent evidence from animal
toxicology and/or controlled human exposure studies provides stronger support, compared to
previous reviews, for potential biologic pathways by which long-term PM2.5 exposures could
lead to effects on the cardiovascular and respiratory systems, effects on the nervous system, and
3-42

-------
to lung cancer.22 23 In addition to providing insight into potential mechanisms, experimental
studies also demonstrate direct effects of PM2.5 exposures, providing further support for
independent effects of particle exposures on health (i.e., not confounded by co-occurring
pollutants). When taken together, the evidence available in this review (i.e., U.S. EPA, 2019)
reaffirms, and in some cases strengthens, the conclusions from the 2009 ISA regarding the health
effects of long-term PM2.5 exposures.
As with the evidence for effects of long-term exposures, since the 2009 ISA, much
progress has been made in assessing key uncertainties in our understanding of health effects
associated with short-term PM2.5 exposures. Recent epidemiologic studies build upon and further
reaffirm those studies evaluated in the 2009 PM ISA, providing evidence of positive associations
across a range of effects. The independence of the PM2.5 effects reported in such studies is
further supported by the results of copollutant analyses indicating that associations with short-
term PM2.5 remain robust. Some recent studies report that associations persist in analyses that
exclude short-term PM2.5 exposures near the upper end of the air quality distribution and that a
threshold below which associations no longer occur is not identifiable from the available data.
The plausibility of PIVh.s-associated mortality is supported by associations with cardiovascular
and respiratory morbidity. Direct evidence for PM2.5 exposure-related cardiovascular effects can
also be found in recent controlled human exposure and animal toxicological studies, supported
by results of epidemiologic panel studies, reporting that PM2.5 exposure can result in various
cardiovascular effects, including endothelial dysfunction, impaired cardiac function, ST segment
depression, thrombosis, conduction abnormalities, and increased blood pressure. Overall, the
results from these studies provide coherence and biological plausibility for the consistent results
from epidemiologic studies observing positive associations between short-term PM2.5
concentrations and ischemic heart disease and heart failure, and ultimately cardiovascular
mortality. While there are inconsistencies in results across some of the animal toxicological,
controlled human exposure, and epidemiologic panel studies, this may be due to substantial
differences in study design, study populations, or differences in PM composition across study
locations. While recent epidemiologic studies also demonstrate associations between short-term
PM2.5 exposures and respiratory effects, particularly asthma and COPD exacerbations, and while
animal toxicological studies provide biological plausibility for these findings, some uncertainty
22	For respiratory effects, nervous system effects, and cancer-related effects animal studies provide support for
potential biologic pathways while controlled human exposure studies are more limited.
23	Animal studies also provide stronger support in this review for effects following exposures to UFP (section 3.2.1),
though important uncertainties remain (e.g., inconsistent UFP definitions across studies, various methods of
administering UFP exposures in health studies, limited understanding of ambient UFP concentrations and
distributions in epidemiologic studies), limiting the potential for these studies to inform policy-relevant
conclusions.
3-43

-------
remains with respect to the independence of PM2.5 effects. Thus, when taken together, the
evidence available in this review (U.S. EPA, 2019) reaffirms, and in some cases strengthens, the
conclusions from the 2009 ISA regarding the health effects of short-term PM2.5 exposures.
3.2.2 Potential At-Risk Populations
The NAAQS are meant to protect the population as a whole, including groups that may
be at increased risk for pollutant-related health effects. In the last review, based on the evidence
assessed in the 2009 ISA (U.S. EPA, 2009), the 2011 PA focused on children, older adults,
people with pre-existing heart and lung diseases, and those of lower socioeconomic status as
populations that are "likely to be at increased risk of PM-related effects" (U.S. EPA, 2011, p. 2-
31). In the current review, the ISA cites extensive evidence indicating that "both the general
population as well as specific populations and lifestages are at risk for PM2.5-related health
effects" (U.S. EPA, 2019, p. 12-1). For example, in support of its "causal" and "likely to be
causal" determinations, the ISA cites substantial evidence for:
•	PM-related mortality and cardiovascular effects in older adults (U.S. EPA, 2019, sections
11.1, 11.2, 6.1, and 6.2);
•	PM-related cardiovascular effects in people with pre-existing cardiovascular disease (U.S.
EPA, 2019, section 6.1);
•	PM-related respiratory effects in people with pre-existing respiratory disease, particularly
asthma (U.S. EPA, 2019, section 5.1); and
•	PM-related impairments in lung function growth and asthma development in children (U.S.
EPA, 2019, sections 5.1 and 5.2; 12.5.1.1).
The ISA additionally notes that stratified analyses (i.e., analyses that directly compare
PM-related health effects across groups) provide strong evidence for racial and ethnic differences
in PM2.5 exposures and in PM2.5-related health risk. Such analyses indicate that minority
populations such as Hispanic and non-Hispanic black populations have higher PM2.5 exposures
than non-Hispanic white populations, thus contributing to adverse health risk in non-white
populations (U.S. EPA, 2019, section 12.5.4). Stratified analyses focusing on other groups also
suggest that populations with pre-existing cardiovascular or respiratory disease, populations that
are overweight or obese, populations that have particular genetic variants, and populations that
are of low socioeconomic status could be at increased risk for PM2.5-related adverse health
effects (U.S. EPA, 2019, Chapter 12).
Thus, the groups at risk of PM2.5-related health effects represent a substantial portion of
the total U.S. population. In evaluating the primary PM2.5 standards, an important consideration
is the potential PM2.5-related public health impacts in these populations.
3-44

-------
3.2.3 PM2.5 Concentrations in Key Studies Reporting Health Effects
To inform conclusions on the adequacy of the public health protection provided by the
current primary PM2.5 standards, this section evaluates the PM2.5 exposures and ambient
concentrations (i.e., used as surrogates for exposures in epidemiologic studies) in studies
reporting PIVh.s-related health effects. We specifically consider the following overarching
questions:
• What are the short- or long-term PM2.5 exposures that have been associated with health
effects and to what extent does the evidence support the occurrence of such effects for
air quality meeting the current primary PM2.5 standards?
In addressing these questions, we emphasize health outcomes for which the ISA has concluded
the evidence supports a "causal" or a "likely to be causal" relationship with PM exposures. As
discussed above, this includes mortality, cardiovascular effects, and respiratory effects associated
with short- or long-term PM2.5 exposures and cancer and nervous system effects associated with
long-term PM2.5 exposures. While the causality determinations in the ISA are informed by
studies evaluating a wide range of PM2.5 concentrations, this section considers the degree to
which the evidence supports the occurrence of PM-related effects at concentrations relevant to
informing conclusions on the primary PM2.5 standards. Section 3.2.3.1 considers the exposure
concentrations that have been evaluated in experimental studies and section 3.2.3.2 considers the
ambient concentrations in locations evaluated by epidemiologic studies.
3.2.3.1 PM Exposure Concentrations Evaluated In Experimental Studies
In the ISA, the evidence for a particular PIVh.s-related health outcome is strengthened
when results from experimental studies demonstrate biologically plausible mechanisms through
which adverse human health outcomes could occur (U.S. EPA, 2015, Preamble p. 20). Two types
of experimental studies are of particular importance in understanding the effects of PM
exposures: controlled human exposure and animal toxicology studies. In such studies,
investigators expose human volunteers or laboratory animals, respectively, to known
concentrations of air pollutants under carefully regulated environmental conditions and activity
levels. Thus, controlled human exposure and animal toxicology studies can provide information
on the health effects of experimentally administered pollutant exposures under highly controlled
laboratory conditions (U.S. EPA, 2015, Preamble, p. 11).
In this section, we consider the PM2.5 exposure concentrations shown to cause effects in
controlled human exposure studies and in animal toxicology studies. We particularly consider
the consistency of specific PIVh.s-related effects across studies, the potential adversity of such
effects, and the degree to which exposures shown to cause effects are likely to occur in areas
meeting the current primary standards. To address these issues, we consider the following
question:
3-45

-------
• To what extent does the evidence from controlled human exposure or animal toxicology
studies support the potential for adverse cardiovascular, respiratory, or other effects
following PM2.5 exposures likely to occur in areas meeting the current primary
standards?
Controlled Human Exposure Studies
As discussed in detail in the ISA (U.S. EPA, 2019, section 6.1), controlled human
exposure studies have reported that PM2.5 exposures lasting from less than one hour up to five
hours can impact cardiovascular function.24 The most consistent evidence from these studies is
for impaired vascular function (U.S. EPA, 2019, section 6.1.13.2). In addition, although less
consistent, the ISA notes that studies examining PM2.5 exposures also provide evidence for
increased blood pressure (U.S. EPA, 2019, section 6.1.6.3), conduction abnormalities/arrhythmia
(U.S. EPA, 2019, section 6.1.4.3), changes in heart rate variability (U.S. EPA, 2019, section
6.1.10.2), changes in hemostasis that could promote clot formation (U.S. EPA, 2019, section
6.1.12.2), and increases in inflammatory cells and markers (U.S. EPA, 2019, section 6.1.11.2).
The ISA concludes that, when taken as a whole, controlled human exposure studies demonstrate
that short-term exposure to PM2.5 may impact cardiovascular function in ways that could lead to
more serious outcomes (U.S. EPA, 2019, section 6.1.16). Thus, such studies can provide insight
into the potential for specific PM2.5 exposures to cause physiological changes that could increase
the risk of more serious effects.
Table 3-2 below summarizes information from the ISA25 on available controlled human
exposure studies that evaluate effects on markers of cardiovascular function following exposures
to PM2.5, either as concentrated ambient particles (CAP) or in unfiltered versus filtered exhaust.26
24	In contrast, controlled human exposure studies provide little evidence for respiratory effects following short-term
PM2.5 exposures (U.S. EPA, 2019, section 5.1, Table 5-18). Therefore, this section focuses on cardiovascular
effects evaluated in controlled human exposure studies of PM2 5 exposure.
25	Table 3-2 includes the controlled human exposure studies, and the endpoints from each study, that are discussed
in the ISA.
26	Table 3-2 identifies controlled human exposure studies included in the ISA that examine the potential for PM2 5
exposures to alter markers of cardiovascular function. Studies that focus on specific components of PM2.5 (e.g.,
endotoxin), or studies that evaluated PM2.5 exposures only in the presence of an intervention (e.g., dietary
intervention) or other pollutant (e.g., ozone), are not included.
3-46

-------
Table 3-2. Summary of information from PM2.5 controlled human exposure studies.
Study
Population
Exposure Details
(average concentration;
duration)
Results
Brauner et al.,
2008
Healthy adults
10.5 |jg/m3 PM2.5
(unfiltered) vs below
detection (filtered); 24 h
No significant effect on markers of vascular
function
Hemmingsen et
al., 2015a,
Hemmingsen et
al., 2015b
Healthy,
overweight
older adults
24 |jg/m3 (unfiltered) vs
3.0 |jg/m3 (filtered)
Copenhagen PM; 5 h
Impaired vascular function and altered heart rate
variability; no significant changes in blood
pressure or markers of inflammation or oxidative
stress
Urch et al., 2010
Non-asthmatic
and mild
asthmatic
adults
64 pg/m3 CAP (lower
exposure); 2 h
No significant change in blood markers of
inflammation or oxidative stress
Huang et al., 2012
Healthy adults
90 pg/m3 CAP; 2 h
No significant changes in heart rate variability
Devlin et al., 2003
Healthy older
adults
99 pg/m3 CAP27; 2 h
Decreased heart rate variability
Hazuchaet al.,
2013
Adult current
and former
smokers
109 pg/m3 CAP; 2 h
No significant changes in markers of
inflammation or coagulation
Ghio et al., 2000
Healthy young
adults
120 pg/m3 CAP; 2 h
Increased fibrinogen (coagulation)
Ghio et al., 2003
Healthy young
adults
120 pg/m3 CAP; 2 h
Increased fibrinogen; no significant effect on
markers of inflammation
Urch et al., 2010
Non-asthmatic
and mild
asthmatic
adults
140 pg/m3 CAP (higher
exposure); 2 h
Increased blood inflammatory markers
Brook et al., 2009
Healthy adults
149 pg/m3 CAP;2h
Impaired vascular function, increased blood
pressure; no significant change in markers of
inflammation (compared to filtered air)
Ramanathan et
al., 2016
Healthy adults
149 pg/m3 CAP;2h
Decreased anti-oxidant/anti-inflammatory
capacity when baseline capacity was low
27 The published study reports an average CAP concentration of 41 |ig/m3. but communication with the study
authors revealed an error in that reported concentration (Jenkins, 2016).
3-47

-------
Sivagangabalan et
al., 2011
Healthy adults
150 |jg/m3 CAP; 2 h
Increase in indicator of possible arrhythmia; no
significant effect on heart rate
Kushaet al., 2012
Healthy adults
154 pg/m3 CAP;2h
No significant effect on indicator of possible
arrhythmia
Gong et al., 2003
Adults with and
without asthma
174 pg/m3 CAP;2h
Increased heart rate; No significant effect on
indicators of arrhythmia, inflammation,
coagulation; inconsistent effects on blood
pressure
Gong et al., 2004
Older adults
with and
without COPD
200 pg/m3 CAP; 2 h
Decreased heart rate variability, increase in
markers of inflammation (without COPD only);
inconsistent effect on arrhythmia; no significant
effect on markers of blood coagulation
Liu et al., 2015
Healthy adults
238|jg/m3 CAP; 130 min
Increase in urinary markers of oxidative stress
and vascular dysfunction; no significant effect on
blood markers of oxidative stress, vascular
function, or inflammation
Bellaviaet al.,
2013
Healthy adults
-242 pg/m3 CAP; 130 min
Increased blood pressure
Behbod et al.,
2013
Healthy adults
-250 pg/m3 CAP; 130 min
Increase in markers of inflammation
Tong et al., 2015
Healthy older
adults
253 pg/m3 CAP; 2 h
Impaired vascular function and increased blood
pressure; no significant change in markers of
inflammation or coagulation
Lucking et al.,
2011
Healthy young
men
320 pg/m3 (unfiltered) vs
7.2 pg/m3 (filtered); 1 h
Impaired vascular function and increased
potential for coagulation; no significant effect on
blood pressure, markers of inflammation, or
arterial stiffness
Vieira et al.,
2016a, Vieira et
al., 2016b
Healthy adults;
Heart failure
patients
325 pg/m3 (unfiltered) vs
25 pg/m3 (filtered) diesel
exhaust; 21-min
Increase in marker of potential impairment in
heart function, impaired vascular function (heart
failure patients); no significant effect on blood
pressure, heart rate or heart rate variability,
markers of inflammation, markers of coagulation,
or arterial stiffness
Most of the controlled human exposure studies in Table 3-2 have evaluated average
PM2.5 exposure concentrations at or above about 100 |ig/m3, with exposure durations typically up
to about two hours. Statistically significant effects on one or more indicators of cardiovascular
function are often, though not always, reported following 2-hour exposures to average PM2.5
concentrations at and above about 120 |ig/m3, with less consistent evidence for effects following
exposures to lower concentrations. Impaired vascular function, the effect identified in the ISA as
3-48

-------
the most consistent across studies (U.S. EPA, 2019, section 6.1.13.2), is shown following 2-hour
exposures to PM2.5 concentrations at and above 149 |ig/m3. Mixed results are reported in the
three studies that evaluate longer exposure durations (i.e., longer than 2 hours) and lower PM2.5
concentrations, with significant effects on some outcomes reported following 5-hour exposures
to 24 |ig/m3 in Hemmingsen et al. (2015b), but not for other outcomes following 5-hour
exposures in Hemmingsen et al. (2015a) and not following 24-hour exposures to 10.5 |ig/m3 in
Brauner et al. (2008).
To provide some insight into what these studies may indicate regarding the primary PM2.5
standards, we consider the degree to which 2-hour ambient PM2.5 concentrations in locations
meeting the current primary standards are likely to exceed the 2-hour exposure concentrations at
which statistically significant effects are reported in multiple studies for one or more indicators
of cardiovascular function. To this end, we refer to Figure 2-14 (Chapter 2, section 2.3.2.2.3),
which presents the frequency distribution of 2-hour average PM2.5 concentrations from all FEM
PM2.5 monitors in the U.S. for 2015-2017. At sites meeting the current primary PM2.5 standards,
most 2-hour concentrations are below 11 ng/m3, and almost never exceed 32 ng/in3. The extreme
upper end of the distribution of 2-hour PM2.5 concentrations is shifted higher during the warmer
months (April to September, denoted by red bars in Figure 2-14), generally corresponding to the
period of peak wildfire frequency in the U.S. At sites meeting the current primary standards, the
highest 2-hour concentrations measured almost never occur outside of the period of peak wildfire
frequency (i.e., 99.9th percentile of 2-hour concentrations is 68 ng/in3 during the warm season).
Most of the sites measuring these very high concentrations are in the northwestern U.S. and
California (see Appendix A, Figure A-l), where wildfires have been relatively common in recent
years. When the typical fire season is excluded from the analysis (blue in Figure 2-14), the
extreme upper end of the distribution is reduced (i.e., 99.9th percentile of 2-hour concentrations is
59 |ag/m3).28
Thus, while controlled human exposure studies support the plausibility of the serious
cardiovascular effects that have been linked with ambient PM2.5 exposures (U.S. EPA, 2019,
Chapter 6), the PM2.5 exposure concentrations evaluated in most of these studies are well-above
the ambient concentrations typically measured in locations meeting the current primary
standards. Therefore, controlled human exposure studies are of limited utility in informing
conclusions on the adequacy of the public health protection provided by the current standards.
Additional controlled human exposure studies that examine longer exposure periods (e.g., 24-
hour as in Brauner et al. (2008); 5-hour as in Hemmingsen et al. (2015b)), or repeated exposures,
28 Similar analyses of 5-hour PM2 5 concentrations are presented in Appendix A, Figure A-2.
3-49

-------
to concentrations typical in the ambient air across much of the U.S. may provide additional
insight into this issue in future reviews.
Animal Toxicology Studies
The ISA relies on animal toxicology studies to support the plausibility of a wide range of
PM2.5-related health effects. While animal toxicology studies often examine more severe health
outcomes and longer exposure durations than controlled human exposure studies, there is
uncertainty in extrapolating the effects seen in animals, and the PM2.5 exposures and doses that
cause those effects, to human populations. We consider these uncertainties when evaluating what
the available animal toxicology studies may indicate with regard to the current primary PM2.5
standards.
Most of the animal toxicology studies assessed in the ISA have examined effects
following exposures to PM2.5 concentrations well-above the concentrations likely to be allowed
by the current PM2.5 standards. Such studies have generally examined short-term exposures to
PM2.5 concentrations from 100 to >1,000 |j,g/m3 and long-term exposures to concentrations from
66 to >400 |j,g/m3 (e.g., see U.S. EPA, 2019, Table 1-2). Two exceptions are a study reporting
impaired lung development following long-term exposures (i.e., 24 hours per day for several
months prenatally and postnatally) to an average PM2.5 concentration of 16.8 |j,g/m3 (Mauad et
al., 2008) and a study reporting increased carcinogenic potential following long-term exposures
(i.e., 2 months) to an average PM2.5 concentration of 17.7 |j,g/m3 (Cangerana Pereira et al., 2011).
These two studies demonstrate serious effects following long-term exposures to PM2.5
concentrations similar to the ambient concentrations reported in some PM2.5 epidemiologic
studies (U.S. EPA, 2019, Table 1-2), though still above the ambient concentrations likely to
occur in areas meeting the current primary standards. Thus, as is the case with controlled human
exposure studies, animal toxicology studies support the plausibility of various adverse effects
that have been linked to ambient PM2.5 exposures (U.S. EPA, 2019), but have not evaluated
PM2.5 exposures likely to occur in areas meeting the current primary standards. Given this, and
the additional uncertainty of extrapolating from effects in animals to those in human populations,
animal toxicology studies are of limited utility in informing conclusions on the public health
protection provided by the current or alternative primary PM2.5 standards.
3.2.3.2 Ambient PM Concentrations in Locations of Epidemiologic Studies
As summarized in section 3.2.1 above, epidemiologic studies examining associations
between daily or annual average PM2.5 exposures and mortality or morbidity represent a large
part of the evidence base supporting several of the ISA's "causal" and "likely to be causal"
determinations. In this section, we consider the ambient PM2.5 concentrations present in areas
where epidemiologic studies have evaluated associations with mortality or morbidity, and what
3-50

-------
such concentrations may indicate regarding the primary PM2.5 standards. The approaches
discussed in this section are also summarized above in section 3.1.2.
As noted in section 3.1.2, the use of information from epidemiologic studies to inform
conclusions on the primary PM2.5 standards is complicated by the fact that such studies evaluate
associations between distributions of ambient PM2.5 and health outcomes, and do not identify the
specific exposures that cause reported effects. Rather, health effects can occur over the entire
distributions of ambient PM2.5 concentrations evaluated, and epidemiologic studies do not
identify a population-level threshold below which it can be concluded with confidence that PM-
associated health effects do not occur (U.S. EPA, 2019, section 1.5.3).
In the absence of discernible thresholds, we use two approaches to consider information
from epidemiologic studies. In one approach, we evaluate the PM2.5 air quality distributions
reported by key epidemiologic studies (i.e., and used to estimate exposures in these studies) and
the degree to which such distributions are likely to occur in areas meeting the current (or
alternative) standards (section 3.2.3.2.1). We recognize uncertainty in using this approach to
inform conclusions on the primary standards because study-reported PM2.5 concentrations are not
the same as the design values used by the EPA to determine whether areas meet the NAAQS
(discussed further below). Therefore, in an additional approach, we calculate study area air
quality metrics similar to PM2.5 design values and consider the degree to which such metrics
indicate that study area air quality would likely have met or violated the current or alternative
standards during study periods (section 3.2.3.2.2).
To the extent these approaches indicate that health effect associations are based on PM2.5
air quality likely to have met the current or alternative standards, such standards are likely to
allow the daily or annual average PM2.5 exposures that provide the foundation for reported
associations. Alternatively, to the extent reported health effect associations reflect air quality
violating the current or alternative standards, there is greater uncertainty in the degree to which
such standards would allow the PM2.5 exposures that provide the foundation for reported
associations. The sections below (i.e., 3.2.3.2.1, 3.2.3.2.2) discuss each of these approaches in
more detail, and present our key observations based on their application. The potential
implications of these observations for the current and alternative primary PM2.5 standards are
discussed below in section 3.4.
3.2.3.2.1 PM2.5 Air Quality Distributions Associated with Mortality or Morbidity in Key
Epidemiologic Studies
In this section, we consider the PM2.5 air quality distributions associated with mortality or
morbidity in key epidemiologic studies, with a focus on the parts of the distributions over which
those studies provide the strongest support for reported associations. As discussed further below,
while health effects may occur at PM2.5 concentrations across the air quality distribution,
3-51

-------
epidemiologic studies often provide the strongest support for reported health effect associations
over the part of the distribution corresponding to the bulk of the underlying data (i.e., estimated
exposures and/or health events). This is the case both for studies of daily PM2.5 exposures and for
studies of annual average PM2.5 exposures.
Studies of daily PM2.5 exposures examine associations between day-to-day variation in
PM2.5 concentrations and health outcomes, often over several years. While there can be
considerable variability in daily exposures over a multi-year study period, most of the estimated
exposures reflect days with ambient PM2.5 concentrations around the middle of the air quality
distributions examined (i.e., "typical" days rather than days with extremely high or extremely
low concentrations). Similarly, for studies of annual PM2.5 exposures, most of the estimated
exposures reflect annual average PM2.5 concentrations around the middle of the air quality
distributions examined. In both cases, epidemiologic studies provide the strongest support for
reported health effect associations for this middle portion of the PM2.5 air quality distribution,
which corresponds to the bulk of the underlying data, rather than the extreme upper or lower
ends of the distribution. Consistent with this, as noted above in section 3.2.1.1, several
epidemiologic studies report that associations persist in analyses that exclude the upper portions
of the distributions of estimated PM2.5 exposures, indicating that "peak" PM2.5 exposures are not
disproportionately responsible for reported health effect associations.
An example of the relationship between data density and reported health effect
associations is illustrated in Figure 3-2 below (from Lepeule et al., 2012, Figure 1 in
supplemental material; U.S. EPA, 2019, Figure 6-26). For the years 1974 to 2009, Lepeule et al.
(2012) report a positive and statistically significant association between estimated long-term
PM2.5 exposures and cardiovascular mortality in six U.S. cities. Based on a visual inspection of
the concentration-response function reported in this study (i.e., presented in Figure 3-2), 95%
confidence intervals are narrowest for long-term PM2.5 concentrations near the overall mean
concentration reported in the study (i.e., 15.9 |j,g/m3). Confidence intervals widen at lower and
higher long-term PM2.5 concentrations, particularly at concentrations < -10 |j,g/m3 and > -20
Hg/m3. This widening in the confidence intervals is likely due in part to the comparative lack of
data at concentrations approaching the lower and upper ends of the air quality distribution (i.e.,
exposure estimates are indicated by hash marks on the horizontal axis).
3-52

-------
Cardiovascular mortality
CD
O
O
o
03
N
10 15 20 2S 30 35 40
PM2.s (ng/m3)-moving average
1-3 years
Figure 3-2. Estimated concentration-response function and 95% confidence intervals
between PM2.5 and cardiovascular mortality in the Six Cities Study (1974-2009) (from
Lepeule et al., 2012, supplemental material, figure 1; Figure 6-26 in U.S. EPA, 2019).
Similar to the information presented in Figure 3-2, other recent studies have also reported
that confidence intervals around concentration-response functions are relatively narrow at PM2.5
concentrations around the overall mean concentrations reported by those studies, likely reflecting
high data density in the middle portions of the distributions (e.g., Crouse et al., 2015; Villeneuve
et al., 2015; Shi et al., 2016 as discussed in U.S. EPA, 2019, section 11.2.4). Thus, consistent
with the approach in the last review (78 FR 3161, January 15, 2013; U.S. EPA, 2011, sections
2.1.3 and 2.3.4.1), we use study-reported means (or medians) of daily and annual average PM2.5
concentrations as proxies for the middle portions of the air quality distributions, over which
studies generally provide strong support for reported associations. As described further below,
when considering the PM2.5 air quality distributions in epidemiologic studies in this section, we
focus on PM2.5 concentrations around these overall means (including concentrations somewhat
below means).
To evaluate the PM2.5 air quality distributions in key studies in this review, we first
identify the epidemiologic studies assessed in the ISA that have the potential to be most
informative in reaching conclusions on the primary PM2.5 standards. As for the experimental
studies discussed above, we focus on epidemiologic studies that provide strong support for
"causal" or "likely to be causal" relationships with PM2.5 exposures in the ISA. We focus on the
health effect associations that are determined in the ISA to be consistent across studies, coherent
3-53

-------
with the broader body of evidence (e.g., including animal and controlled human exposure
studies), and robust to potential confounding by co-occurring pollutants and other factors. We
emphasize multicity studies that examine health effect associations in the U.S. or Canada, as
such studies examine potential associations over large geographic areas with diverse atmospheric
conditions and population demographics (e.g., U.S. EPA, 2019, sections 11.1 and 11.2).
Additionally, studies examining associations outside the U.S. or Canada reflect air quality and
exposure patterns that may be less typical of the U.S., and thus less likely to be informative for
purposes of reviewing the NAAQS.29
Figure 3-3 to Figure 3-6 and Table 3-3 below summarize information from U.S. and
Canadian studies that are assessed in the ISA and that meet these criteria. For each study, Figure
3-3 to Figure 3-6 present the cohort and/or geographic area examined, the approach used to
estimate PM2.5 exposures (i.e., monitored versus predicted with hybrid modeling methods30), the
study years during which health events occurred, the years of PM2.5 air quality data used to
estimate exposures, and the effect estimate31 with 95% confidence intervals (per 5 |j,g/m3 for
long-term exposures; 10 |j,g/m3 for short-term exposures). When available, these figures also
include the overall means (or medians if means are not available) of the short- or long-term
PM2.5 exposure estimates reported by the study.
Figure 3-3 and Figure 3-4 summarize information from studies of long-term PM2.5
exposures. Figure 3-5 and Figure 3-6 summarize information from studies of short-term PM2.5
exposures. Table 3-3 summarizes information from the smaller group of retrospective studies
that have evaluated the potential for improvements in public health as ambient PM2.5
concentrations have declined over time. It is important to note that these retrospective studies
tend to focus on time periods during which ambient PM2.5 concentrations were substantially
higher than those measured more recently (e.g., see Chapter 2, Figure 2-8).
29	This emphasis on studies conducted in the U. S. or Canada is consistent with the approach in the last review of the
PM NAAQS (U.S. EPA, 2011, section 2.1.3).
30	As discussed further below, and in Chapter 2, hybrid methods incorporate data from several sources, often
including satellites and models, in addition to ground-based monitors.
31	The effect estimates presented in the forest plot figures (Figure 3-3 to Figure 3-6) show the associations of long or
short-term PM2 5 exposures with health endpoints presented either as hazard ratio or odds ratio or relative risk (for
which the bold dotted vertical line is at 1), or as per unit or percent change (for which the bold dotted vertical line
is at 0).
3-54

-------
All-cause mortality
Exposure Health Air Quality Reported PM Mean
Proxy Country Citation	Cohort Data	Data (Range)(ug/m3)
Modelled U.S. Dietal., 2017b Medicare 2000-2012 2000-2012 11.0 (Sth and 95th: 6.21-15.64)

•
Hart eta!., Nurses 2000.2006 1999.2006 12,0 (NR)
2015 Health


Popeetal., ACSCPS-II 1982-2004 1999-2004 12.6(1.0-28.0)
2015 1 '


Puetteta'., Nurses 1992.2002 1988.2002 13.9 (s8_27.6)
2009 Health


Puett etal.. Health Pro 1989.2QQ3 1988.2003 ( }
2011 fessionals


Shi etal., 2016 Medicare 2003-2008 2003-2008 8.12(0.8-20.22)

—
Thurston etal., N|H.AARp 2000.2009 2000-2008 12.2(2.9-28.0)
2016 1

*-
Turner etal., Acscps_n 1932-2004 1999-2004 12.6(1.4-27.9)
2016 1 '


20179 0t a' " Medicare 2000-2013 2000-2013 NR (Median: 10.7) (6.0-20.6)


Weichenthaj et „ , „„„„„„„„ „„„„ „„„ Iowa: 8.8; North Carolina: 11.1
, , Ag Health 1993-2009 2001-2006
al., 2014 3 (NR)


Canada Crouseeta'- CanCHEC 1991-2001 2001-2006 8.7(1.9-19.2)
2012 '


Grouse et ai1
2Q1 " CanCHEC 1991-2006 1984-2006 8.9 (0.9-17.6)

-~
Pinaultetal., CCHS 2000-2011 1998-2011 6.3(1.0-13.0)
2016 '

—
Goss et al.. U.S. Cvstic , ,
Monitor U.S. ' , 1999-2000 2000 13.7 (NR)
2004 Fibrosis x '


Hart etal.. Nurses 2000.2006 2000.2005 12.7 (NR)
2015 Health * '


Kiomourtzoglou dj 2000-2010 2000-2010 12.0 (Mean Range: 9.0-13.0) (NR)
etal., 2016 v y /v '


Lepeule et al.. Harvard „^,1974-2009:15.9; 2000 onwards
_ 2001-2009 1979-2009 '
2012 Six-City mean range: <15-<18(NR)


Lipfert et =1., Veterans i997.2ooi 1999-2001 14.3 (NR)
2006 1 '

•
Zegeret al., MCAPS 2000-2005 2000.2005 Central region: NR (Median: 10.7)
2008 (NR) , ,
Eastern region: NR (Median: 14.0)
(NR)
Western region: NR (Median: 13.1)
(NR)


Canada CrouseetEL- CanCHEC 1991-2001 1987-2001 11.2 (NR)
2012

-
T™alet CanCHEC 1991-2009 1998-2009 9.8(4.74-13.62)
al., 2016a

—

0.9 1.0 1.1 1.2 1.3 1.4
Hazard Ratio (95% CI)
3-55

-------
CVD mortality
Exposure Health Air Quality Reported PM Mean Health
Proxy Country Citation Cohort Data	Data (Range)(ug/m3) 	Outcome
Modelled U.S. ™etal-' £?, 1982-2004 2002-2004 12.0(1.5-26.6) IHD mortality
2016 CPS-II x 1 Age 30+


Popeetal., ACS 1982-2004 1999-2004 12.6(1.0-23.0) CVDmortality
Aqe 30+
2015 CPS-II y
IHD mortality
Age 30+
Other CVD-
CBVD Age 30+

~
Thurstonet NIH- CVDmortality
, 2000-2009 2000-2008 12.2(2.9-28.0) . '
al., 2016 AARP x ' Age 50-71

-
Turner et aL, ACS 1982-2004 1999-2004 12.6(1.4-27.9) CVDmortality
N ' Aqe 30+
2016 CPS-II y
IHD mortality
Age 30+
Other CVD-
CBVD Age 30+

~-
Weichenthal Ag 	, _	_ Iowa: 8.8; North Carolina: 11.1
1993-2009 2001-2006 ,.,-s CVDmortality
etal.,2014 Health (NR)


Chen et al., EFFECT „ „ „ _ . . CVDmortality
Can3da 2016 RCT 1999-2011 2001-2010 10.7 (NR) ^ ^


CrouseelaL. 1991-2001 2001-2006 8.7(1.9-19.2) CVDmortality
2012 x ' Age 25+

-~
Cro.se et al.. CanCHEC 1984.2006 8 g (Q g. CVD mortality
2015 N ' Age 25-90

~
Pinaultetal., 2000-2011 1998-2011 6.3(1,0-13.0) CVDmortality
2016 v ' Age 25-90


Vllleneuve et CNBSS 1980-2005 1998-2006 9.1(1.3-17.6) "l0^allty
, Age 40-59
al.. 2015 3
IHD mortality
Age 40-59


Monitor U.S. jou.813'" TrlPS 1985-2000 2000 14.1 (NR) CVDmortality


Lepeule et a!.. Harvard „„„„ „	^ „	 1974-2009:15.9; 2000 onwards CVDmortality
2001-2009 1979-2009 , ,
2012 Six-City mean range: <15-<18 (NR) Age 25-74


MiMer'etal WHI 1994-2002 2000 13.5(3.4-28.3) CVDmortality
2007 1 ' Age 50-79


Canada Wei,C™ CanCHEC 1991-2009 1998-2009 9.8 (4.74-13.62) 7™^
etal., 2016a Age 25-89

1.0 1.2 1.4 1.6
Hazard Ratio (95% CI)
3-56

-------
Respiratory mortality
Exposure	Health Air Quality Reported PM Mean
Proxy Country Citation	Cohort Data	Data	(Range)(ug/m3)
Modelled U.S. Popeetal.,2015 ACSCPS-I! 1982-2004 1999-2004 12.6(1.0-28.0)

—
Ih™Stal' NIH-AARP 2000-2009 2000-2008 12.2(2.9-28.0)
2016 '


Turner st al., 2016 ACSCPS-II 1982-2004 1999-2004 12.6(1.4-27.9)

—
Canada Crouse et al., 2015 CanCHEC 1991-2006 1984-2006 8.9(0.9-17.6)


Pinaultetal., 2016 CCHS 2000-2011 1998-2011 6.3(1.0-13.0)


Monitor U.S. Hart et al., 2011 TrIPS 1985-2000 2000 14.1 (NR)


Canada We'chenthal e-al- CanCHEC 1991-2009 1998-2009 9.8(4.74-13.62)
2016a x '



1.0 1.1 1.2 1.3
Hazard Ratio (95% CI)
Lung cancer mortality
Exposure	Health Air Quality Reported PM Mean
Proxy Country Citation	Cohort Data Data	(Range)(ug/m3)
Turner e^* al
Modelled U.S. 201g ' "" ACSCPS-II 1982-2004 1999-2004 12.6(1.4-27.9)


Canada grouse etal., CanCHEC 1991-2006 1984-2006 8.9(0.9-17.6)
2015 '

~
Pirmiltetal., CCHS 2000-2011 1998-2011 5.3(1.0-13.0)

Vi lieneuve et
' CNBSS 1980-2005 1998-2006 9.1(1.3-17.6)
al., 2015 * '


Monitor U.S. "f^eta'"' TrIPS 1985-2000 2000 14.1 (NR)
2011


Krewski et al., „„„„ 1979-1983; 1979-1983: 21.2; 1999- 2000;
ACSCPS-I 1982-2000 , ,
2009 1999-2000 14.0 (NR)


Laden et al.. Harvard 		 1979-1987; 16.4 (Mean Range: 10.2-29.0)
1974-1998
2006 Six-City 1985-1998 (NR)


Lepeule et al.. Harvard „„„„ 1974-2009:15.9; 2000 onwards
. 2001-2009 1979-2009 ' ,
2012 Six-City mean range: <15-<18 (NR)


Thurston et al.,
ACSCPS-II 1982-2004 2000-2005 14.2(NR)
2013 v '


Canada ®"tha'6t CanCHEC 1991-2009 1998-2009 9.8(4.74-13.62)
al.f 2016a x 1



1.0 1.5 2.0 2.5
Hazard Ratio (95% CI)
Figure 3-3. Epidemiologic studies examining associations between long-term PM2.5
exposures and mortality.
3-57

-------
Asthma incidence
Exposure
Proxy	Health Air Quality Reported PM Mean
(group) Country Citation	Cohort	Data	Data	(Range)(ug/m3)
Modelled Canada Tetreauitetal., Q|CDSS 1996-2011 2001-2006 9.86(NR)
2016 '

•
Monitor U.S. CHS 2003-2005 2003-2004 13.9(6.3-23.7)
2010 '


Nishimuraetal., 		 		 Mean Range: 8.1-17.0
GALA lii1 SAGE II 1986-2003 198o-2003 , ,
2013 ' (NR)


Lung cancer incidence
Exposure Health Air Quality Reported PM Mean
Proxy Country Citation Cohort Data Data (Range)(ug/m3)
1.0 1.5 2.0
Odds Ratio (95% CI)
Modelled Canada HYstade-al< necSS 1994-1997 1975-1994 11.9(NR)
2013 1


Tomczaketal., CNBSS 1980-2004 1998-2006 9.1(1.3-17.6)
2016 v '


Monitor U.S. Gharibvandet AH5MOG_2 2002-2011 2000-2001 12.9 (NR)
al., 2016 s '


Nurses
Puettetal.,2014 , 1994-2010 19S8-2007 13.1 (NR)
Health



1.0 1.1 1.2 1.3 1.4
Hazard Ratio (95% CI)
Lung development
Exposure	Health Air Quality Reported PM Mean
Proxy Country Citation Cohort Data	Data	(Range)(ug/m3) Health Outcome
Monitor U.S. Breton et al., CHS 1993-2000, 1994-2004 Mean Range: rwr^0^0^^
2011 1996-2004 6.0-28.0 (NR) FVC Age 10-18
Lung Development (Change) -
MMEF Age 10-18
Lung Development (Change)-
FEV1 Age 10-18



Gauderman CHS 1993-2000 1994-2000 Mean Range:
et al., 2004 6.0-28.0 (NR) FVC Age 10-18
Lung Development (Change) -
MMEF Age 10-18
Lung Development (Change)-
FEV1 Age 10-18








-80 -60 -40 -20 0
ml Change in Growth (95% CI)
Lung function
Exposure	Health Air Quality Reported PM Mean
Proxy Country Citation	Cohort Data	Data	(Range)(ug/m3) Health Outcome
Monitor U.S. UrmanetaL, CHS 2002-2007 2002-2007 6.0-28.0 (NR) Lung Function Decline (%)-
2014 98
Lung Function Decline (%)-
FVC Age 5-7




-1.5 -1.0 -0.5 0.0
Percent Difference (95% Cl)
Figure 3-4. Epidemiologic studies examining associations between long-term PM2.5
exposures and morbidity.
3-58

-------
All-cause mortality
Exposure
Proxy Country Citation
Cohort
Health Data
Reported PM Mean (Range)(ug/m3)

Modelled U.S. Dietal., 2017a
Medicare
2000-2012
11.6 (5th and 95th: 6.21-15.64)
i -
Lee et al., 2015b
State Dept
2007-2011
11.1(0.02-86.2)

Shi et al., 2016
Medicare
2003-2008
8.21(0.8-20.22)

Monitor U.S. Baxter et al., 2017
NCHS
2001-2005
Cluster Mean Range: 12.2-14.1 (NR)
: —
Dai et al., 2014
NCHS
2000-2006
13.3 (NR)
: —
Dominici et al., 2007
NMMAPS
1999-2000
NR (NR)

Franklin et al., 2007
NCHS/State Dept
1997-2002
15.6 (NR)

Franklin et al., 2008
NCHS/State Dept
2000-2005
14.8 (NR)

Klemm et al., 2003
Harvard Six-City
1979-1988
14.7 (Median: 9.0) (NR)

Krall et al.r 2013
NCHS
2000-2005
13.6 (NR)

Zanobetti and Schwartz, 2009
NCHS
1999-2005
13.2 (NR)
: —
Zanobetti et al., 2014
Medicare
1999-2010
Mean Range: 4.37-17.97 (NR)
j —
Canada Burnett et al., 2003
Statistics Canada
1986-1996
13.3 (NR)


Burnett et al., 2004
Statistics Canada
1981-1999
12.8 (NR)






0 12 3




Percent Increase (95% Cl)
CVD mortality




Exposure
Proxy Country Citation
Cohort
Health Data Reported PM Mean (Range)(ug/m3)

Modelled U.S. Lee et al., 2015b
State Dept
2007-2011
11.1(0.02-86.2)

Monitor U.S. Dai etal., 2014
NCHS
2000-2006
13.3 (NR)

Franklin et al., 2007
NCHS/State Dept 1997-2002
15.6 (NR)


Franklin et al., 2008
NCHS/State Dept 2000-2005
14.8 (NR)

Zanobetti and Schwartz, 200?
NCHS
1999-2005
13.2 (NR)





0 12 3




Percent Increase (95% CI)
Respiratory mortality




Exposure
Proxy Country Citation
Cohort
Health Data
Reported PM Mean (Range)(ug/m3)

Modelled U.S. Lee et al., 2015b
State Dept
2007-2011
11.1 (0.02- 86.2)
-r-
Monitor U.S. Dai et al., 2014
NCHS
2000-2006
13.3 (NR)

Franklin etal., 2007
NCHS/State Dept
1997-2002
15.6 (NR)

Franklin etal., 2008
NCHS/State Dept
2000-2005
14.8 (NR)

Zanobetti and Schwartz, 2009
NCHS
1999-2005
13.2 (NR)

0 12 3
Percent Increase (95% CI)
Figure 3-5. Epidemiologic studies examining associations between short-term PM2.5
exposures and mortality.32
32 As noted above, the overall mean PMSS concentrations reported in studies of short-term (24-hour) exposures
reflect averages across the study population and over the years of the study. Thus, mean concentrations reflect
long-term averages of 24-hour PMa5 exposure estimates.
3-59

-------
CVD morbidity
Exposure	Reported PM Mean
Proxy Country Citation	Cohort Health Data (Range)(ug/m3) Health Outcome
Modelled U.S. Bravo etal., 2017 Medicare 2002-2006 12.3 (NR) CVD HA Age 65+

•
Kloog et al., 2012 Medicare 2000-2006 9.6(0.01-72.59) CVD HA Age 65+

~-
Kloog et al., 2014 Medicare 2000-2006 11.9 (NR) CVD HA Age 65+
IHD HA Age 65+

~
Monitor U.S. Bell etal., 2008 Medicare 1999-2005 12.9 (NR) CVD HA Age 65+

•
Bell etal., 2014 Medicare 2000-2004 (Medlan- 11J> CVD HA Age 65+


Bell etaJ.r 2015 Medicare 1999-2010 12.3(6.4-20.2) Heart Failure HA
Age 65+
IHD HA Age 65+
~
~
Bravo etal., 2017 Medicare 2002-2006 12.5 (NR) CVD HA Age 65+

•
Dominici etal.. Medicare 1999-2002 13.4 (NR) Heart Failure HA
2006 3
IHD HA Age 65+

~-
Peng etal., 2009 Medicare 2000-2006 ^^Median-1-1-8) CVD HA Age 65+


Zanobettietal., Medicare 2000-2003 15.3 (NR) Heart Failure HA
Age 65+
2009 3
Ml HA Age 65+

,
Canada Stiebetal ,2009 Hospital 1992-2003 3.2(6.7-9.8) Angina/Ml ED
Database
Heart Failure ED


Szyszkowiczetal., Hospital 		,
2009 Database 1992"2003 S3W Angina ED


Welchentha'etaL, nacrs 2004-2011 6.9(NR) Ml ED
2016b v



0 5 10 15
Percent Increase (95% CI)
3-60

-------
Respiratory morbidity
Exposure	Reported PM Mean
Proxy Country Citation	Cohort	Health Data Health Outcome (Range)(ug/m3)
Modelled U.S. ^o°°4 Sta'"' Medicare 2000-2006 COPDHAAge65+ 11.9 (NR)

—
Monitor U.S. Alhanti at al.. Hospital 1993-2009 Asthma ED Age Mean Range.
2016 Database 5"18 11.1-14.1 (NR)


, , Mean Range:
Asthma ED Age 65+		
3 11.1-14.1 (NR)


Bell etal.,2015 Medicare 1999-2010 Asthma HA Age 65+ 12.3(6.4-20.2)


CORD HA Age 65+ 12.3(6.4-20.2)


Domimci etal., M d[ 1999-2002 COPDHAAge65+ 13.4(NR)
2006 3 * '

—
Maiigetal., Hospital 2005-2008 Asthma EDS HA
2013 Inpatient and 5.2-19.8 (NR)
Outpatient visits COPDED&HA Mean Range.
5.2-19.8 (NR)

—


Ostroetal., Hospital 2005-2009 Asthma ED a HA 16.5 (NR)
2016 Inpatient and
Outpatient visits COPD ED & HA 16.5 (NR)

—


v/ i -> ,, .4. . „„„„ __Asthma HA Age 1-9: Mean Range:
Yap etal, 2013 Hospital 2000-2005 a ,
.... Central Valley 12.8-20.8 (NR)

•
Asthma HA Age 1-9: Mean Range:
South Coast 14.0-24.6 (NR)

•
Canada StiebetaL, Hospital 1992-2003 Asthma ED 8.2(6.7-9.8)


COPD ED 8.2 (6.7-9.8)


Weichenthal et NACRS 2004-2011 Asthma ED 7.1 (NR)

—
COPD ED 7.1 (NR)

—

0.95 1.00 1.05
Relative Risk/ Odds Ratio (95% CI)
Figure 3-6. Epidemiologic studies examining associations between short-term PM2.5
exposures and morbidity.
3-61

-------
Table 3-3. Epidemiologic studies examining the health impacts of long-term reductions in
ambient PM2.5 concentrations.
Study
Reference
Study Area
Years of PM2.5
Air Quality
(monitored)
Starting PM2.5
Concentrations
(mean)
Ending PM2.5
concentrations
(mean)
Study Results
Pope et al.
(2009)
211 U.S.
counties
1979-1983
compared to
1999-2000
20.6 pg/m3
14.1 |a,g/m3
Statistically significant
association between
declining ambient
PlVtaand increasing
life expectancy
Correia et al.
(2013)
545 U.S.
counties
2000
compared to
2007
13.2 pg/m3
11.6 |a,g/m3
Statistically significant
association between
declining ambient
PlVtaand increasing
life expectancy
Berhane et al.
(2016)
4,602
children in 8
California
communities
1992-2000;
1995-2003;
2002-2011
20.5 pg/m3
14.4 pg/m3
Statistically significant
decrease in bronchitic
symptoms in 10-year
old children with and
without asthma
Gauderman et
al. (2015)
2,120
children in 5
California
communities
1994-1997;
1997-2000;
2007-2010
21.3-31.5 pg/m3
11.9-17.8 pg/m3
Statistically significant
improvements in 4-
year growth of lung
function
Based on the information in Figure 3-3 to Figure 3-6 and Table 3-3, key epidemiologic
studies conducted in the U.S. or Canada indicate generally positive and statistically significant
associations between estimated PM2.5 exposures (short- or long-term) and mortality or morbidity
across a wide range of ambient PM2.5 concentrations. As discussed above, considering the PM2.5
concentrations around (i.e., somewhat below to somewhat above) the overall means in these
studies can provide insight into the part of the air quality distribution over which studies provide
the strongest support for reported health effect associations. Evaluating whether such PM2.5 air
quality distributions would be likely to occur in areas meeting the current (or alternative) primary
standards can inform conclusions on the degree to which those standards would limit the
potential for the long-term and short-term PM2.5 exposures that provide strong support for
reported associations.
For a subset of key epidemiologic studies with available information, we characterize the
broader distributions of ambient concentrations, with a particular focus on the concentrations
below which data could become appreciably more limited (i.e., below which relatively few
estimated exposures, and/or few health events, occurred). As noted above, confidence in reported
health effect associations declines for portions of the air quality distribution accounting for
3-62

-------
comparatively little data (i.e., concentrations approaching the lower and upper ends of the
distribution). Thus, considering the concentrations below which data become relatively sparse
can provide insight into the ambient PM2.5 concentrations below which confidence in reported
health effect associations may decrease notably. While there is no single concentration below
which we lose confidence in reported associations, consistent with the approach in the last
review (U.S. EPA, 2011, section 2.3.4.1), we identify the PM2.5 concentrations corresponding to
the 25th and 10th percentiles of health data (when available) or exposure estimates to provide
insight into the concentrations that comprise the lower quartiles of the air quality distributions.33
To frame our evaluation of study-reported PM2.5 concentrations, we specifically consider
the following questions:
•	What are the overall mean PM2.5 concentrations reported by key epidemiologic studies?
•	For studies with available information on the broader distributions of exposure
estimates and/or health events, what are the PM2.5 concentrations corresponding to the
25th and/or 10th percentiles of those data?
Answers to these questions can provide insight into the range of PM2.5 concentrations, including
those below the overall means, over which key studies provide strong support for reported
associations. To this end, Figure 3-7 and Figure 3-8 below present information on the monitored
(Figure 3-7) and hybrid model-predicted (Figure 3-8) ambient PM2.5 concentrations used to
estimate PM2.5 exposures in key epidemiologic studies.
Drawing from the U.S. and Canadian multicity studies in Figure 3-3 to Figure 3-6
above,34 the studies included in Figure 3-7 and Figure 3-8 are those that report overall mean (or
median) PM2.5 concentrations and for which the years of PM2.5 air quality data used to estimate
exposures overlap entirely with the years during which health events are reported. Regarding this
latter issue, the PM2.5 concentrations reported by studies that estimate exposures from air quality
corresponding to only part of the study period, often including only the later years of the health
data (e.g., Miller et al., 2007; Hart et al., 2011; Thurston et al., 2013; Weichenthal et al., 2014;
Weichenthal et al., 2016a; Pope et al., 2015; Villeneuve et al., 2015; Turner et al., 2016), are not
33	In the last review of the PM NAAQS, the PA identified the long-term PM2 5 concentrations corresponding to the
25th and 10th percentiles of health events, or study populations. In doing so, the PA noted that a range of one
standard deviation around the mean represents approximately 68% of normally distributed data and, below the
mean, falls between the 25th and 10th percentiles.
34	Most of the studies included in Table 3-3 above (i.e., studies that examine relationships between declining
ambient PM2 5 concentrations and improving health) report mean ambient PM2 5 concentrations well-above those
in the studies highlighted in Figure 3-3 to Figure 3-6, and well-above the concentrations likely to be informative
for conclusions on the current primary PM2 5 standards. Therefore, our evaluation of mean concentrations focuses
on the key studies identified in Figure 3-3 to Figure 3-6.
3-63

-------
likely to reflect the full ranges of ambient PM2.5 concentrations that contributed to reported
associations.35
Figure 3-7 highlights the overall mean (or median) PM2.5 concentrations reported in key
studies that use ground-based monitors alone to estimate long- or short-term PM2.5 exposures.
For the subset of studies with available information on the broader distributions of underlying
data, Figure 3-7 also identifies the study-period mean PM2.5 concentrations corresponding to the
25th and 10th percentiles of health events36 (see Appendix B, Section B.2 for more information).
35	This is an issue only for some studies of long-term PM2 5 exposures. While this approach can be reasonable in the
context of an epidemiologic study evaluating health effect associations with long-term PM2 5 exposures, under the
assumption that spatial patterns in PM2 5 concentrations are not appreciably different during time periods for
which air quality information is not available (e.g., Chen et al., 2016), our interest is in understanding the
distribution of ambient PM2 5 concentrations that could have contributed to reported health outcomes.
36	That is, 25% of the total health events occurred in study locations with mean PM2 5 concentrations (i.e., averaged
over the study period) below the 25th percentiles identified in Figure 3-7 and 10% of the total health events
occurred in study locations with mean PM2 5 concentrations below the 10th percentiles identified.
3-64

-------
Franklin 2007 (US: 27 Cities)
Franklin 2008 {US: 25 Cities)
Klemm 2003 (US: Harvard 6 City)
Krall 2013 (US: 72 Cities)
Dai 2014 (US: 75 Cities)
Burnett 2003 (Canada: 8 Cities)
Zanobetti and Schwartz 2009 (US: 112 Cities)
Burnett 2004 (Canada: 12 Cities)
Ostro 2016 (US: 8 California Counties)
Zanobetti 2009 (US: 26 cities)
Bell 2014 (US: 4 Counties in MA & CT)
Dominici 2006 (US: 204 Urban Counties)
Bell 2008 (US: 202 Counties)
Bravo 2017 (US: 418 Counties)
Bell 2015 (US: 70 Urban Counties)
Peng 2009 (US: 119 Urban Counties)
Szyszkowicz 2009 (Canada: 6 Cities)
Stieb 2009 (Canada: 6 Cities)
Weichenthal 2016c (Canada: 15 Ontario Cities)
Weichenthal 2016b (Canada: 16 Ontario Cities)
Zeger 2008 (US: 421 Eastern Region Counties)
Zeger2008 (US: 62 Western Region Counties)
Hart 2015 (US: Nationwide)
Kioumourtzoglou 2016 (US: 207 Cities)
Crouse 2012 (Canada: 11 Cities)
Zeger 2008 (US: 185 Central Region Counties)
McConnell 2010 (US: 13 California Communities)
Gharibvand 2016 (US: Nationwide)
9
9
Study Types
I ST Exposure 8 Mortality
I ST Exposure S Morbidity
I LT Exposure & Mortality
LT Exposure S Morbidity
Summary Statistics
0	10th percentile
• 25th percentile
1	Mean or Median
ns: non-significant association
10 11
12
13
14 15 16 17
Overall PM2,S Concertation for the Study Period (|J.g/m3
Figure 3-7. Monitored P1M2.5 concentrations in key epidemiologic studies.
3-65

-------
We also consider the emerging body of studies that use predicted ambient PM2.5
concentrations from hybrid modeling methods to estimate long- or short-term PM2.5 exposures
(Figure 3-8, below). As discussed in Chapter 2 of this PA (section 2.3.3), hybrid methods
incorporate data from several sources, often including satellites and models in addition to
ground-based monitors. Compared to ground-based monitors alone, hybrid methods have the
potential to improve the characterization of PM2.5 exposures in areas with relatively sparse
monitoring networks (U.S. EPA, 2019, sections 3.3.2 to 3.3.5).
Figure 3-8 presents overall means of predicted PM2.5 concentrations for key studies, and
the concentrations corresponding to the 25th and 10th percentiles of estimated exposures or health
events37 when available (see appendix B, section B.3 for additional information).38 As for the
monitor-based studies highlighted above, Figure 3-8 focuses on multicity studies that examine
health outcomes supporting "causal" or "likely to be causal" determinations in the ISA and that
use air quality data to estimate PM2.5 exposures for the entire range of years during which health
events occurred.39 In addition to these criteria, we also consider the approach used to validate
hybrid model predictions. In particular, the studies included in Figure 3-8 are those for which
relatively robust model validation analyses are reported to have been conducted for the full range
of years during which PM2.5 exposures are estimated in the health study (e.g., regional or
37	For most studies in Figure 3-8, 25th percentiles of exposure estimates are presented. The exception is Di et al.
(2017a), for which Figure 3-8 presents the short-term PM2 5 exposure estimates corresponding to the 25th and 10th
percentiles of deaths in the study population (i.e., 25% and 10% of deaths occurred at concentrations below these
concentrations). In addition, the authors of Di et al. (2017b) provided population-weighted exposure values
(Chan, 2019). The 10th and 25th percentiles of these population-weighted exposure estimates are 7.9 and 9.5
|ig/m3. respectively.
38	In addition, 75th percentiles of exposure estimates are available for some studies. They are as follows: 14.4 g/nf'
(Di et al., 2017a), 12.9 ng/m3 (Di et al., 2017b), 11.7ng/m3 (Kloog et al., 2012), 10.7 ng/m3 (Shi et al. (2016),
short-term exposures), 10.0 |ig/m3 (Shietal. (2016), long-term exposures), 12.9 |ig/m3 (Wangetal., 2017).
39	All studies that meet the criteria for inclusion in Figure 3-8 were conducted in the U.S.
3-66

-------
national 10-fold cross validation performance statistics reported for the same years that
exposures are estimated).40
40 For example, due to lack of spatial field availability before 1998, Crouse et al. (2015) use median annual PM2 5
concentrations for the 1998-2006 time period (van Donkelaar et al., 2010; van Donkelaar et al., 2015a; van
Donkelaar et al., 2013) to predict exposures during the 1984-2006 period. Similarly, for Pinault et al. (2016),
model validation is for 2004 to 2008 (van Donkelaar et al., 2015b) while exposures are estimated for 1998 to
2012. Paciorek et al. (2009), which presents the model validation results for Puett et al. (2009) and Puett et al.
(2011), notes that PM2 5 monitoring was sparse prior to 1999, with many of the available PM2 5 monitors in rural
and protected areas. Therefore, Paciorek et al. (2009) conclude that coverage in the validation set for most of the
study period (1988-1998) is poor and that their model strongly underestimates uncertainty (Paciorek et al. (2009),
p. 392 in published manuscript). Hystad et al. (2013) used exposure fields developed by calibrating satellite-based
PM2 5 surfaces from a recent period (van Donkelaar et al., 2010) to estimate exposure for the 1975 to 1994
(Hystad et al., 2012). Hystad et al. (2012) noted that a random effect model was used to estimate PM25 based on
TSP measurements and metropolitan indicator variables because only small number of PM2 5 measurements were
available, and no measurements were made prior to 1984. Thus, these studies from Figures 3-3 to 3-6 are not
included in Figure 3-8.
3-67

-------
Di 2017a (US: Nationwide; ST Exposure)
Lee 2015b (US: 3 SE States)
Shi 2016 (US: 6 NE States; ST Exposure)
Bravo 2017 (US: 708 Counties)
Kloog 2014 (US: 7 Mid-Atlantic States and D.C.)
Kloog 2012 (US: 6 NE States)
Thurston 2016 (US: 6 States and 2 MSAs)
Hart 2015 (US: Nationwide)
Di 2017b (US: Nationwide; LT Exposure)
Wang 2017 (US: 7 SE States)
Di 2017b (US: Nationwide; LT Exposure < 12 ug/m3)
Shi 2016 (US: 6 NE States; LT Exposure)

5 6 7 8 9 10 11 12
Overall PM2.5 Concentration for the Study Period (|ig/m3)
Study Types
| ST Exposure & Mortality
I ST Exposure S Morbidity
I LT Exposure S Mortality
Summary Statistics
O 10th percentile
• 25th percentile
¦ Mean or Median
13
Figure 3-8. Hybrid model-predicted PM2.5 concentrations in key epidemiologic studies.
3-68

-------
Taking the information in Figure 3-7 and Figure 3-8 together, key epidemiologic studies
conducted in the U.S. or Canada report generally positive and statistically significant
associations between estimated PM2.5 exposures (short- or long-term) and mortality or morbidity
across a wide range of monitored or hybrid-model-predicted ambient PM2.5 concentrations. With
regard to these studies, we particularly note the following:
•	For the large majority of key studies, the PM2.5 air quality distributions that support reported
associations are characterized by overall mean (or median) PM2.5 concentrations ranging
from just above 8.0 |j,g/m3 to just above 16.0 |j,g/m3. There is substantial overlap between
mean concentrations based on monitoring alone and those based on hybrid modeling
approaches.
-	Most key studies that use monitors alone to estimate PM2.5 exposures, and all of
the U.S. studies in this group, report overall mean PM2.5 concentrations at or
above 10.7 |j,g/m3.
-	Four Canadian studies that use monitors alone report lower overall mean
concentrations. Two of these studies report overall means just above 8.0 |j,g/m3
(both report positive and statistically significant associations) and two studies
report overall means around 7.0 |j,g/m3 (positive and statistically significant
association in one of these studies).
-	Most key studies that use hybrid modeling approaches to estimate PM2.5
exposures report overall mean concentrations at or above 9.6 |j,g/m3. All of these
studies were conducted in the U.S. and report positive and statistically significant
health effect associations.
-	The hybrid modeling study with the lowest PM2.5 concentrations reports overall
means just above 8.0 |j,g/m3 (i.e., Shi et al., 2016). This study reports positive and
statistically significant health effect associations with both short- and long-term
PM2.5 exposures.41
•	Four U.S. studies examine health effect associations in analyses with the highest exposures
excluded. Only one of these restricted analyses is reflected in Figure 3-8 (i.e., Di et al.,
2017b; "LT exposure < 12 |j,g/m3"). In addition to this study, Lee et al. (2015), Di et al.
(2017a) and Shi et al. (2016) also report positive and statistically significant associations in
restricted analyses.
-	Lee et al. (2015) reports a positive and statistically significant association in an
analysis restricted to zip codes with annual average PM2.5 concentrations < 12
Hg/m3 and to days with 24-hour average PM2.5 concentrations < 35 |j,g/m3. This
study did not report an overall mean PM2.5 concentration for the restricted
analysis, though it was presumably somewhat below the mean reflected in Figure
3-8 (i.e., 11.1 |j,g/m3).
41 However, the authors report that, for associations with long-term PM2 5 exposures, most deaths occurred at or
above the 75th percentile of annual exposure estimates (i.e., 10 |ag/m3) (see Tables 1 and 2 in published
manuscript). Authors did not report this information for their analysis of short-term PM2 5 exposures.
3-69

-------
-	Di et al. (2017a) reports a positive and statistically significant association in an
analysis restricted to 24-hour PM2.5 exposure estimates < 25 |j,g/m3. This study did
not report an overall mean PM2.5 concentration for the restricted analysis, though
it was presumably somewhat below the mean reflected in Figure 3-8 (i.e., 11.6
Hg/m3).
-	Shi et al. (2016) report positive and statistically significant associations in
analyses restricted to annual PM2.5 exposure estimates <10 |j,g/m3 and in analyses
restricted to 24-hour exposure estimates < 30 |j,g/m3. This study does not report
the overall mean PM2.5 concentrations in restricted analyses, though such means
are presumably somewhat below those reflected in Figure 3-8 (i.e., 8.1 and 8.2
Hg/m3).
• For some key studies, information on the broader distributions of PM2.5 exposure estimates
and/or health events is available.
-	In U.S. studies that use monitors alone to estimate PM2.5 exposures, 25th
percentiles of health events correspond to mean PM2.5 concentrations (i.e.,
averaged over the study period for each study city) at or above 11.5 |j,g/m3 and
10th percentiles of health events correspond to mean PM2.5 concentrations at or
above 9.8 |j,g/m3 (i.e., 25% and 10% of health events, respectively, occur in study
locations with mean PM2.5 concentrations below these values).
-	In the Canadian studies that use monitors alone to estimate PM2.5 exposures, 25th
percentiles of health events correspond to mean PM2.5 concentrations at or above
6.5 |j,g/m3 and 10th percentiles of health events correspond to mean PM2.5
concentrations at or above 6.4 |j,g/m3.
-	Of the key studies that use hybrid modeling approaches to estimate long-term
PM2.5 exposures, the ambient PM2.5 concentrations corresponding to 25th
percentiles of estimated exposures are 6.2 and 9.1 |j,g/m3. In the one study with
data available on the 10th percentile of PM2.5 exposure estimates, the
concentration corresponding to that 10th percentile is 7.3 |j,g/m3.
-	In studies that use hybrid modeling approaches to estimate short-term PM2.5
exposures, the ambient concentrations corresponding to 25th percentiles of
estimated exposures, or health events, are generally at or above 6.4 |j,g/m3. In the
one study with lower concentrations, the ambient PM2.5 concentration
corresponding to the 25th percentile of estimated exposures is 4.7 |j,g/m3 42 In the
one study with information available on the 10th percentile of health events, the
ambient PM2.5 concentration corresponding to that 10th percentile is 4.7 |j,g/m3.
The information in Figure 3-7 and Figure 3-8 indicates consistent support for generally
positive and statistically significant health effect associations for PM2.5 air quality distributions
42 As noted above, in this study (Shi et al., 2016), the authors report that most deaths occurred at or above the 75th
percentile of annual exposure estimates (i.e., 10 |J.g/m3). The short-term exposure estimates accounting for most
deaths are not presented in the published study.
3-70

-------
characterized by overall mean (or median) concentrations above 8.0 |ag/m\ with most studies
(and all but one U.S. study) reporting overall mean (or median) concentrations at or above 9.6
Hg/m3. While the ambient PM2.5 concentrations around these overall means generally reflect the
part of the air quality distribution over which studies provide the strongest support for reported
PM2.5 effect estimates, there are uncertainties in using these concentrations to inform conclusions
on the primary PM2.5 standards. These uncertainties are summarized below and their potential
implications for conclusions on the current and alternative standards are discussed further in
section 3.5.
A key uncertainty in using study-reported mean PM2.5 concentrations to inform
conclusions on the primary PM2.5 standards is that such concentrations are not the same as the
ambient concentrations used by the EPA to determine whether areas meet or violate the PM
NAAQS. As discussed above, the overall mean PM2.5 concentrations reported by key
epidemiologic studies reflect averaging of short- or long-term PM2.5 exposure estimates across
locations (i.e., across multiple monitors or across modeled grid cells) and over time (i.e., over
several years). In contrast, to determine whether areas meet or violate the NAAQS, the EPA
measures air pollution concentrations at individual monitors (i.e., concentrations are not
averaged across monitors) and calculates "design values" at monitors meeting appropriate data
quality and completeness criteria. For the annual PM2.5 standard, design values are calculated as
the annual arithmetic mean PM2.5 concentration, averaged over 3 years. For the 24-hour standard,
design values are calculated as the 98th percentile of the annual distribution of 24-hour PM2.5
concentrations, averaged over three years (described in Appendix N of 40 CFR Part 50). For an
area to meet the NAAQS, all valid design values in that area, including the highest annual and
24-hour monitored values, must be at or below the levels of the standards.
Because of this approach to determining whether areas meet the NAAQS, and because
monitors are often required in locations with relatively high PM2.5 concentrations (section 2.2.3),
areas meeting a PM2.5 standard with a particular level would be expected to have average PM2.5
concentrations (i.e., averaged across space and over time in the area) somewhat below that
standard level. In support of this, analyses of recent air quality in U.S. CBS As indicate that
maximum annual PM2.5 design values for a given three-year period are often 10% to 20% higher
than average monitored concentrations (i.e., averaged across multiple monitors in the same
CBSA) (Appendix B, section B.7). The difference between the maximum annual design value
and average concentration in an area can be smaller or larger than this range, likely depending on
factors such as the number of monitors, monitor siting characteristics, and the distribution of
3-71

-------
ambient PM2.5 concentrations.43 When using this information to interpret key epidemiologic
studies in the context of the primary standards, it is also important to note that such ratios may
depend on how the average concentrations in a study are calculated (i.e., averaged across
monitors versus across modeled grid cells). Thus, as discussed further in section 3.5 below, when
evaluating what the mean PM2.5 concentrations reported by key epidemiologic studies may
indicate regarding the current or alternative PM2.5 standards, we consider the broader
relationships between mean PM2.5 concentrations, averaged across space and over time, and
PM2.5 design values.44
Additional uncertainties in using the PM2.5 concentrations reported by key epidemiologic
studies to inform conclusions on the primary PM2.5 standards include the following:
•	Effects can occur over the full distributions of ambient PM2.5 concentrations evaluated in
epidemiologic studies, and the evidence does not identify a threshold concentration below
which PM2.5-associated effects no longer occur. Thus, while conclusions on primary
standards can be informed by comparing the PM2.5 air quality distributions present in key
studies with the distributions likely to occur in areas meeting the current or alternative
standards, studies do not identify specific PM2.5 exposures that result in health effects or
exposures below which effects do not occur.
•	For studies that use hybrid model predictions to estimate PM2.5 exposures, the performance
of the recently developed modeling approaches depends on the availability of monitoring
data and varies by location. As noted in Chapter 2 (section 2.3.3), factors likely contributing
to poorer model performance often coincide with relatively low ambient PM2.5
concentrations, potentially accounting for the observations that model performance for
hybrid models weaken by some metrics with decreasing PM2.5 concentration and that the
normalized variability between predictions based on different hybrid modeling approaches
increases with decreasing concentrations. Thus, uncertainty in hybrid model predictions
becomes an increasingly important consideration as lower predicted concentrations are
considered.
The potential implications of these and other uncertainties for conclusions on the current and
alternative primary PM2.5 standards are discussed below in section 3.4.
3.2.3.2.2 PM2.5 Pseudo-Design Values in Locations of Key Epidemiologic Studies
In addition to considering the study-reported PM2.5 concentrations discussed above, we
also evaluate study area air quality using metrics more closely related to the design values
43	Given that higher PM2 5 concentrations have been reported at some near-road monitoring sites, relative to the
surrounding area (section 2.3.2.2.2), recent requirements for PM25 monitoring at near-road locations in large
urban areas (section 2.2.3) may increase the ratios of maximum annual design values to averaged concentrations
in some areas.
44	As discussed above in section 3.1.2, compared to the annual standard, the potential implications of overall mean
PM2.5 concentrations reported by key epidemiologic studies are less clear for the 24-hour PM2 5 standard with its
98th percentile form (section 3.4).
3-72

-------
employed by the EPA to determine whether areas meet or violate the primary PM2.5 standards.
To the extent these metrics suggest that reported health effect associations are based largely on
PM2.5 air quality that would have met the current or alternative standards during study periods,
we have greater confidence that those standards would allow the PM2.5 exposures that provide
the basis for reported associations. In contrast, to the extent these metrics suggest that reported
health effect associations are based largely on air quality that would have violated the current or
alternative standards, there is greater uncertainty in the degree to which those standards would
allow the PM2.5 exposures that provide the basis for reported associations.
To evaluate this issue, we calculate metrics similar to PM2.5 design values (referred to
here as "pseudo-design values") for the locations and time periods evaluated by key U.S. and
Canadian epidemiologic studies. Pseudo-design values are calculated as follows:
•	We first identify the study locations with one or more PM2.5 monitors operating during the
study period, and that have sufficient monitoring data available to calculate pseudo-design
values.45
o For key studies conducted in the U.S., study locations are defined as the
counties included in the study.
o For key studies conducted in Canada, study locations are defined as the cities
included in the study.
•	For each monitored study location, we then identify the highest annual and 24-hour PM2.5
pseudo-design values for each 3-year period of the study and calculate the study-period
average of these highest values.
•	We also identify the number of people living in each study location or, when available, the
number of health events that occurred in each location during the study period.46
•	To evaluate the percentages of study area populations living in locations likely to have met
the current standards over study periods (or the percentages of health events occurring in
45	Pseudo-design values are based on data from both FRM/FEM monitors and from high quality non-FRM/FEM
monitors. The non-regulatory data used to calculate pseudo-design values come from monitors typically used for
EPA applications like AirNow that are not FRM or FEM. Only monitors with 75% completeness for each of the
12 quarters in a 3-year design value period were included. Sensitivity analyses based only on data from
FRM/FEM regulatory monitors gave similar results (Appendix B, section B.5). For the pseudo-design values at
the Canadian sites, only sites with 75% completeness for each year of the 3-year design value period were
included. These criteria are slightly different than that of actual design values which have strict rounding
conventions and substitution tests for sites with less than 75% completeness for each quarter. Additional
information on the approach and data sources used to identify pseudo-design values in study locations is provided
in Appendix B (section B.4.3).
46	When available, we use the number of health events in each study location. However, for most key studies, health
event data was not available for each study location. For these studies, we evaluate the population living in each
study location. Comparison of these approaches in the subset of studies for which health events are available
demonstrate that distributions of annual pseudo-design values are comparable for the two approaches (Appendix
B, section B.6).
3-73

-------
such locations), we identify the percentages in locations with study-period average pseudo-
design values at or below the levels of the current annual (Figure 3-9; Appendix B, Tables B-
5 and B-6) and 24-hour (Appendix B, Figure B-9) PM2.5 standards.
In Figure 3-9, whiskers reflect annual PM2.5 pseudo-design values corresponding to 5th
and 95th percentiles of study area populations (or health events), boxes correspond to the 25th and
75th percentiles, and the vertical lines inside the boxes correspond to 50th percentiles. The vertical
dotted line in Figure 3-9 is drawn at 12.0 |ag/m3, the level of the current annual PM2.5 standard.
For studies with 25th percentiles < 12.0 ng/m3, at least 25% of the study area population (i.e., in
counties or cities with pseudo-design values) lived in locations likely to have met the current
annual standard over the study period (or at least 25% of health events occurred in such
locations).47 Similarly, for studies with 50th or 75fe percentiles < 12.0 |J,g/m3, at least 50% or 75%
of the study area population, respectively, lived in locations likely to have met the current annual
standard over the study period (or at least 50% or 75% of health events occurred in such
locations). The percentage of study area populations (or health events) in locations likely to have
met the current 24-hour standard over study periods was typically larger than the percentage in
locations likely to have met the current annual standard (i.e., Appendix B, Figure B-9).
47 As discussed below, among study locations with averaged PM2 5 pseudo-design values (i.e., averaged over the
study period) at or below 12.0 |.ig/ml almost all individual 3 -year pseudo-design values are also at or below 12.0
|ag/m3 (i.e., 89% for Di et al. (2017b); 98% for Shi et al. (2016)- see Appendix B, section B.9).
3-74

-------
Long-term exposure studies
Endpoint . Study „ —
Country „ Citation v/ Geographic Areas
Group Years
U.S. Mortality Lepeule et al., 2012* 2001-2009 6 U.S. Cities
Kiomourtzoglou et al., 2016* 2000-2010 207 U.S. Cities
Di et al2017b* 2000-2012 U.S. Nationwide
1 1 1—1
1 r—r—1 1
Wang et al., 2017* 2000-2013 7 SE U.S. States
I—1 i I—1
Shi et al., 2016* 2003-2008 6 NE U.S. States
Morbidity Urman et al., 2014* 2002-2007 8 CA Counties
1	r~n—;—1
1	;	1 1 1	1
Mcconnell etal., 2010 2003-2005 13 CA Communities
Canada Monality Pinault etal., 2016* 2000-2011 Multicity
1—1	1 1	1
1—1 1 1	1 :
5	10	15	20	25
Avg. Max PseudoDV
Short-term exposure studies
Coun..
Endpoint
Group
U.S. Mortality
Morbidity
Canada Morbidity
Citation
Franklin etal., 2008*
Dai etal.. 2014*
Study
Years
2000-2005
2000-2006
Geographic Areas
25 U.S. Cities
75 U.S. Cities
Baxteretal., 2017*	2001-2005	77U.S. Cities
Zanobatti et al., 2014*	1999-2010	121 U.S. Cities
Zancbetti and Schwartz, 2009*	1999-2005	112 U.S. Cities
Pi etal., 2017a*	2000-2012	U.S. Nationwide
Lee etal., 2015bs	2007-2011	3 SE U.S. States
Shi etal., 2016*
Yap etal, 2013*
Ostro et al., 2016*
Zanobetti et al., 2009*
Maligetal., 2013*
Peng et al., 2009*
Dominici et al., 2006*
Kloog et al., 2014*
Bell etal., 2008*
Bell etal., 2014*
Bravo etai., 2017*
Bell etal., 2015*
2003-2008	6 NE U.S. States	|
2000-2005	CA (Central a Southern Counties)
2005-2009	8 CA Counties
2000-2003	26 U.S. Cities
_2005-2008	35 CA Counties
2000-2006	119 U.S. Urban Counties
1999-2002	204 U.S. Urban Counties
2000-2006	7 U.S. Mid-Atlantic States S D.C.
1999-2005	202 U.S. Urban Counties
2000-2004	4 U.S. Counties. MA & CT
2002-2006	708 U.S. Counties
1999-2010	213 U.S. Urban Counties
Kloog et al., 2012*	2000-2006 6 NE U.S. States
Weichenthal etal., 2016b	2004-2011 16 Ontario Cities
Weicbenthal etal., 2016c*	2004-2011 15 Ontario Cities
5	10 15 20
Avg. Max PseudoDV
Figure 3-9. PM2.5 annual pseudo-design values (in jiig/1113) corresponding to various
percentiles48 of study area populations or health events for studies of long-term and
short-term PM2.5 exposures.49
18 Asterisks next to study citations denote statistically significant effect estimates.
49 For most of the studies included in Figure 3-9, pseudo-design values are available for >70% of study area
populations (or health events). Exceptions are Kloog et al. (2012), Lee et al. (2015), Pinault et al. (2016), and
Wang et al. (2017), with pseudo-design values available for 67%. 56%, 51%, and 65% of study area populations,
respectively.
3-75

-------
Drawing from the information in Figure 3-9 (and Figure B-9 in Appendix B), we
particularly note the following:
•	For most of the key studies (i.e., 18 of the 29 in Figure 3-950), about 25% or more of the
study area populations (i.e., of those in areas with pseudo-design values) lived in locations
with air quality likely to have met the current primary standards over study periods (or about
25% or more of health events occurred in locations with such air quality).
-	For the 15 U.S. studies included in this group, annual pseudo-design values from
8.7 to 11.9 |j,g/m3 correspond to 25th percentiles of study area populations (or
health events).
-	For the three Canadian studies included in this group, annual pseudo-design
values from 6.0 to 7.2 |j,g/m3 correspond to 25th percentiles of study area
populations (or health events).
•	For nine of the key studies, most of the study area population (i.e., > 50% of those living in
areas with pseudo-design values) lived in locations with air quality likely to have met the
current standards over study periods (or > 50% of health events occurred in locations with
such air quality).
-	For the six U.S. studies included in this group, annual pseudo-design values from
9.9 to 11.7 |j,g/m3 correspond to 50th percentiles of study area populations (or
health events).
-	For the three Canadian studies included in this group, annual pseudo-design
values from 7.3 to 7.4 |j,g/m3 correspond to 50th percentiles of study area
populations (or health events).
•	For four of the key studies, the large majority of the study area population (i.e., >75% of
those living in areas with pseudo-design values) lived in locations with air quality likely to
have met the current standards over study periods (or >75% of health events occurred in
locations with such air quality).
-	One of these studies (Shi et al., 2016) was conducted in the U.S. In this study, an
annual pseudo-design value of 11.0 |j,g/m3 corresponds to the 75th percentile of
the study area population.51
-	Three of these studies (Pinault et al., 2016; Weichenthal et al., 2016c; and
Weichenthal et al., 2016b) were conducted in Canada. In these studies, annual
pseudo-design values from 8.4 to 8.6 |j,g/m3 correspond to 75th percentiles of the
study area populations (or health events).
•	For the remaining 11 key studies, the large majority of the study area population (i.e., >75%
of those living in areas with pseudo-design values) lived in locations with air quality likely to
50	Shi et al. (2016) separately examined long- and short-term PM2 5 exposures and, therefore, is included twice in
Figure 3-9 and Figure B-9.
51	In Shi et al. (2016), 85% of all of the study areas with pseudo-design values would likely have met the current
annual standard over the entire study period (i.e., annual pseudo-design values for every three-year period
examined were < 12.0 |-ig/m3).
3-76

-------
have violated one or both of the current standards during study periods (or >75% of health
events occurred in locations with such air quality).
While the information in Figure 3-9 can inform conclusions regarding the degree to
which air quality present in study locations and during study periods would likely have met the
current primary PM2.5 standards, there are important uncertainties to consider when using such
information to inform conclusions on the primary PM2.5 standards. These include the following:
•	For most key multicity studies, some study locations would likely have met the current
primary standards over study periods while others would likely have violated one or both
standards. There is uncertainty in how to interpret such studies to inform conclusions on the
NAAQS. However, the importance of this uncertainty is lessened for studies that report
positive and statistically significant associations in populations that reside almost entirely in
areas likely to have met the current standards (e.g., Pinault et al., 2016; Shi et al., 2016;
Weichenthal et al., 2016c). This uncertainty is also lessened for key studies that report
positive and statistically significant associations in analyses restricted long-term average
PM2.5 concentrations below 12 |j,g/m3 (Di et al., 2017b) or 10 |j,g/m3 (Shi et al., 2016), which
account for about half of the total deaths in these studies (i.e., 54% in Di et al. (2017b), and
49% in Shi et al. (2016)). Effect estimates in these restricted analyses are slightly larger than
those based on the entire cohort.
•	For each study location, maximum 3-year pseudo-design values are averaged over study
periods. Depending on the years of air quality evaluated by the study, for some locations
those averages could reflect air quality that violated the current standards during part of the
study period and met the current standards during part of the study period. However, analysis
of this issue indicates that, among study locations with averaged PM2.5 pseudo-design values
(i.e., averaged over the study period) at or below 12.0 |ag/m3, almost all individual 3-year
pseudo-design values are also at or below 12.0 |j,g/m3 (i.e., 89% for Di et al. (2017b); 98%
for Shi et al. (2016)- see Appendix B, section B.9).
•	Analyses identifying pseudo-design values in study locations necessarily focus on locations
with at least one PM2.5 monitor. While this approach can account for the large majority of
study area populations for studies that use monitors alone to estimate PM2.5 exposures, some
recent key epidemiologic studies use hybrid modeling approaches to predict ambient PM2.5
concentrations in locations with and without nearby ground-based monitors (i.e., Figure 3-8,
above). For these studies, PM2.5 pseudo-design values are not available for unmonitored
study locations. For most of the key studies, pseudo-design values are available for locations
accounting for more than 70% of the study population. However, for some studies, the
percentages of study area populations living in locations with pseudo-design values are lower
(Kloog et al., 2012; Lee et al., 2015; Pinault et al., 2016; Wang et al., 2017). To the extent
unmonitored areas have generally lower ambient PM2.5 concentrations than monitored areas,
our analyses of pseudo-design values could be biased toward the higher values present in
monitored locations.
•	PM2.5 monitoring requirements have changed since the study periods covered by key studies.
In particular, PM2.5 pseudo-design values during study periods do not reflect the near-road
PM2.5 monitors that are now required in many large urban areas (discussed in section
2.3.2.2.2 above). Had current requirements for near-road monitors been in place during study
3-77

-------
periods, the maximum pseudo-design values in some counties could have been higher than
those identified. Early data from near road monitors indicates that about half of urban areas
with near-road monitors measured the highest annual design values at those monitors. Of the
CBS As with highest annual design values at near-road sites, those design values were, on
average, 0.7 |j,g/m3 higher than at the highest measuring non-near-road sites (range is 0.1 to
2.0 |j,g/m3 higher at near-road sites) (Table 2-2 above).
The potential implications of these and other uncertainties for the primary PM2.5 standards are
discussed in section 3.4 below.
3.2.3.3 Conclusions from the Evidence
In reaching conclusions based on the evidence considered in section 3.2.3, we revisit the
questions posed at the beginning of the section:
• What are the short- or long-term PM2.5 exposures that have been associated with health
effects and to what extent does the evidence support the occurrence of such effects for
air quality meeting the current primary PM2.5 standards?
To answer these questions, we draw on information from experimental studies, as discussed in
section 3.2.3.1, and information from epidemiologic studies, as discussed in section 3.2.3.2.
With regard to the experimental evidence, we note that available controlled human
exposure and animal toxicology studies provide general support for the plausibility of many of
the serious health outcomes associated with estimated PM2.5 exposures in epidemiologic studies
(U.S. EPA, 2019, Chapters 5 to 11). However, the PM2.5 exposure concentrations consistently
shown to elicit effects across these studies are considerably higher than the ambient
concentrations typically measured in the U.S. in recent years, and higher than the concentrations
likely to occur in areas meeting the current primary standards (section 3.2.3.1). A limited number
of experimental studies report effects following exposures to lower PM2.5 concentrations (Mauad
et al. (2008); Cangerana Pereira et al. (2011),52 though still above typical ambient concentrations
observed in locations meeting the current standards. Thus, while experimental studies support the
plausibility of serious PIVh.s-associated health effects, these studies provide limited insight into
the occurrence of effects following PM2.5 exposures likely to occur in the ambient air in areas
meeting the current primary PM2.5 standards.
With regard to the epidemiologic evidence, we first note that key studies conducted in the
U.S. or Canada indicate positive and often statistically significant associations between estimated
52 Mauad et al. (2008) and Cangerana Pereira et al. (2011) report respiratory and cancer-related effects, respectively,
in animals following long-term exposures to 16.8 and 17.7 )_Lg/m3 PM2 5. Hemmingsen et al. (2015b) reports
cardiovascular effects in human volunteers following 5-hour exposures to an average of 24 (.ig/m3 PM2 5.
Additionally, the controlled human exposure study by Brauner et al. (2008) reports no change in markers of
cardiovascular function following 24-hour PM exposures to an average PM2 5 concentration of 10.5 |.ig/m\
3-78

-------
PM2.5 exposures (short- or long-term) and mortality or morbidity across a broad range of ambient
concentrations. These include associations based on PM2.5 air quality distributions lower than
those in key studies from the last review.53 Based on the information in Figure 3-7 and Figure 3-
8, the large majority of key epidemiologic studies in the current review report health effect
associations for air quality distributions characterized by overall mean PM2.5 concentrations
ranging from 8.1 |j,g/m3 to 16.5 |ag/m3, with mean concentrations in most of these studies (and all
but one key U.S. study) at or above 9.6 |j,g/m3. These include studies that report associations in a
wide variety of populations, including studies examining substantial portions of the U.S.
population and studies examining groups that may be at comparatively high risk (e.g., older
adults, children). These studies employ various study designs and examine a wide variety of
health outcomes, geographic areas, approaches to estimating PM2.5 exposures, and approaches to
control for confounding. The evidence for associations at lower ambient concentrations (i.e.,
means < 8.0 |ag/m3) is more limited, with two studies conducted in Ontario reporting positive
associations (statistically significant in one study) for PM2.5 air quality distributions
characterized by overall mean concentrations around 7.0 |j,g/m3 (Weichenthal et al., 2016c;
Weichenthal et al., 2016b).
Considering the PM2.5 concentrations around these overall means can provide insight into
the part of the air quality distribution over which studies provide the strongest support for
reported health effect associations. Evaluating whether such PM2.5 air quality distributions would
be likely to occur in areas meeting the current (or alternative) primary standards can inform
conclusions on the degree to which those standards would limit the potential for the long- and
short-term PM2.5 exposures that support reported health effect associations. However, a
limitation of considering study-reported mean PM2.5 concentrations to inform conclusions on the
primary PM2.5 standards is that such concentrations, by themselves, do not indicate whether
study areas would likely have met or violated the current standards (or alternatives).
As discussed above (sections 3.2.3.2.1 and 3.2.3.2.2), the EPA uses design values at
individual monitors to determine whether areas meet the NAAQS. Based on analyses of recent
air quality in U.S. CBS As, maximum annual PM2.5 design values for a given three-year period
are often 10% to 20% higher than average concentrations over that period (i.e., averaged across
monitors in the same CBSA) (Appendix B, Figure B-7 and Table B-9). These relationships
suggest that areas with maximum annual PM2.5 design values of 12.0 |j,g/m3 (i.e., just meeting the
current annual standard) are likely to have long-term mean PM2.5 concentrations (i.e., averaged
53 In the last review key epidemiologic studies supporting "causal" or "likely to be causal" determinations examined
distributions of ambient PM2 5 with overall mean concentrations at or above 12.8 |-ig/m3 (U.S. EPA, 2011, Figure
2-8).
3-79

-------
across space and over time) that are somewhat below 12.0 |j,g/m3 but still higher than the overall
means reported by a number of key epidemiologic studies reporting PM2.5 health effect
associations. This indicates that the current standards are likely to allow the distributions of
short- and long-term PM2.5 exposures that are associated with health effects in some key studies.
Another approach to examine the potential implications of key epidemiologic studies for
the primary PM2.5 standards is to consider analyses of PM2.5 pseudo-design values in locations of
those studies, thereby focusing on a study-related air quality metric that is more directly
comparable to the levels of the primary PM2.5 standards. As illustrated in Figure 3-9, and in
Figure B-9 in Appendix B, for several key studies with available pseudo-design values (9 of the
studies evaluated), most of the study area populations lived in locations with air quality likely to
have met both the annual and 24-hour PM2.5 standards over study periods (or most of health
events occurred in such areas). For the U.S. studies in this group, annual pseudo-design values
from 9.9 to 11.7 |j,g/m3 correspond to 50111 percentiles of study area populations (or health
events). That is, 50% of the study area populations lived in locations with pseudo-design values
below these concentrations, or 50% of the health events occurred in such locations. For the U.S.
study reporting the lowest annual average concentrations (Shi et al., 2016), 75% of the study area
population lived in locations with annual pseudo-design values below 11.0 |j,g/m3. For the
Canadian studies with the lowest ambient PM2.5 concentrations, annual pseudo-design values of
about 7.3 to 7.4 |j,g/m3 correspond to 50th percentiles of study area populations (or health events),
and annual pseudo-design values from 8.4 to 8.6 |j,g/m3 correspond to 75th percentiles.
When the information summarized above is taken together, along with the uncertainties
discussed in section 3.2.3.2 above, we reach the conclusion that a number of key epidemiologic
studies report positive and statistically significant PM2.5 health effect associations based largely,
or entirely, on air quality that is likely to be allowed by the current primary PM2.5 standards. Our
consideration of the evidence and air quality information to inform conclusions on the primary
PM2.5 standards is discussed further in section 3.4 below.
3.3 RISK-BASED CONSIDERATIONS
To inform conclusions regarding the primary PM2.5 standards that are "requisite" to
protect the public health (i.e., neither more nor less stringent than necessary; section 1.2), it is
important to consider the health risks that would be allowed under those standards. For the
current standards, this means evaluating PIVh.s-related health risks in locations with three-year
annual PM2.5 design values of 12.0 |j,g/m3 and/or three-year 24-hour design values of 35 |j,g/m3
(i.e., neither above nor below the levels of the current standards). Therefore, in addition to our
evaluation of PM2.5 concentrations in locations of key epidemiologic studies (which are based on
existing air quality; section 3.2.3.2), we use information from those studies in a risk assessment
3-80

-------
that estimates population-level health risks associated with PM2.5 air quality that has been
adjusted to simulate "just meeting" the current standards (i.e., design values equal to 12.0 |j,g/m3
and/or 35 |j,g/m3). Given our conclusions based on the evidence (section 3.2.3.3), we also
estimate risks associated with PM2.5 air quality adjusted to simulate "just meeting" alternative
annual and 24-hour standards with lower levels. These risk estimates, when considered alongside
analyses of the evidence discussed above in section 3.2.3, are meant to inform conclusions on the
primary standards that would be requisite to protect the public health against long- and short-
term PM2.5 exposures. Our consideration of estimated risks focuses on addressing the following
policy-relevant questions:
•	What are the estimated PM2.5-associated health risks for air quality just meeting the
current primary PM2.5 standards?
•	To what extent are risks estimated to decline when air quality is adjusted to just meet
potential alternative standards with lower levels?
•	What are the uncertainties and limitations in these risk estimates?
The sections below summarize our approach to estimating risks (section 3.3.1) and the
results of the risk assessment (section 3.3.2). Additional detail on the risk assessment is provided
in Appendix C.
3.3.1 Overview of Approach to Estimating Risks
Our general approach to estimating PIVh.s-associated health risks combines
concentration-response functions from epidemiologic studies with ambient PM2.5 concentrations
corresponding to air quality scenarios of interest, baseline health incidence data, and population
demographics for locations included in the risk assessment. Below we summarize key aspects of
the risk modeling approach. Additional detail on the approach is provided in Appendix C
(section C.l).
•	Study area selection: In selecting U.S. study areas for inclusion in the risk assessment, we
focus on the following characteristics:
-	Available ambient monitors: We focus on areas with relatively dense ambient
monitoring networks, where we have greater confidence in adjustments to
modeled air quality concentrations in order to simulate "just meeting" the current
and alternative primary PM2.5 standards (air quality adjustments are described in
detail in Appendix C, section C.1.4).
-	Geographical Diversity. We focus on areas that represent a variety of regions
across the U.S. and that include a substantial portion of the U.S. population.
-	PM2.5 air quality concentrations: We balance the value of including a broad array
of study areas from across the U.S. against the larger uncertainty associated with
air quality adjustments in certain areas. For example, many areas have recent air
quality that meets the current primary PM2.5 standards. Inclusion of such areas in
3-81

-------
the risk assessment necessitates an upward adjustment to PM2.5 air quality
concentrations in order to simulate just meeting the current standards. Given
uncertainty in how such increases could potentially occur, we select areas (i.e.,
CBS As54) requiring either a downward adjustment to air quality or a relatively
modest upward adjustment (i.e., no more than 2.0 |j,g/m3 for the annual standard
and 5 |j,g/m3 for the 24-hour standard, based on the 2014-2016 design-value
period). In addition, as discussed further in Appendix C (section C.1.4), we
excluded several areas that appeared to be strongly influenced by exceptional
events. Forty-seven urban study areas met these criteria (Figure 3-10 and
Appendix C, section C.1.3), including 30 study areas where just meeting the
current standards is controlled by the annual standard,55 11 study areas where just
meeting the current standards is controlled by the daily standard,56 and 6 areas
where the controlling standard differed depending on the air quality adjustment
approach (Figure 3-10).57
54	CBS As (core-based statistical areas) can include one or more counties. Each CBS A selected included at least one
monitor with valid design values and several CBSAs had more than 10 monitors. See Table C-3 in Appendix C.
55	For these areas, the annual standard is the "controlling standard" because when air quality is adjusted to simulate
just meeting the current or potential alternative annual standards, that air quality also would meet the 24-hour
standard being evaluated.
56	For these areas, the 24-hour standard is the controlling standard because when air quality is adjusted to simulate
just meeting the current or potential alternative 24-hour standards, that air quality also would meet the annual
standard being evaluated. Some areas classified as being controlled by the 24-hour standard also violate the
annual standard.
57	In these 6 areas, the controlling standard depended on the air quality adjustment method used and/or the standard
scenarios evaluated.
3-82

-------
/ Ain P	J/^ I 1
m yffjA) H iljS		I f? Kfi flS» ftf^n^LA	, \	'	' * /	.7 A- V
Number of Urban Study
Areas (CBSAs)
Controlling
Standard
Population (>30
years old)
30
Annual (Blue)
~50M
11
Daily (Green)
-4M
6
Mixed (Grey)
~5M
Total: 47

~60M
Figure 3-10. Map of 47 urban study areas included in risk modeling.
•	Health outcomes: The health outcomes evaluated in the risk assessment are (a) total
mortality (all-cause and non-accidental), ischemic heart disease mortality, and lung cancer
mortality associated with long-term PM2.5 exposures and (b) total mortality associated with
short-term PM2.5 exposures (Table 3-4 below and Appendix C, section C. 1.1). Evidence for
these outcomes supports "causal" or "likely to be causal" determinations in the ISA (U.S.
EPA, 2019).
•	Concentration-response functions: Concentration-response functions used in this risk
assessment are from large, multicity U.S. epidemiologic studies that evaluate PM2.5 health
effect associations (drawn from those identified above in Figures 3-3 to 3-6). The selection of
specific epidemiologic studies and concentration-response functions for use in modeling risk
is based on criteria that take into account factors such as study design, geographic coverage,
demographic groups evaluated, and health endpoints examined. Information from these
studies is summarized in Table 3-4. Additional detail regarding the selection of
epidemiologic studies and specification of concentration-response functions can be found in
Appendix C (section C.l.l).
3-83

-------
Table 3-4. Epidemiologic studies used to estimate PM2.5-associated risk.
Epidemiology Study
Study Population3
Age Range
(years)
Mortality Categories
Covered
Long-term mortality studies
Jerrettet al., 2016
ACS
30+
IHD
Pope et al., 2015
ACS
30+
All-cause, IHD
Turner et al., 2016
ACS
30+
Lung cancer
Thurston et al., 2016
AARP
55-85
All-cause
Di et al., 2017b
Medicare
65+
All-cause
Short-term mortality
Baxter et al., 2017
77 cities
All ages
Non-accidental
Itoetal., 2013
NPACT
All ages
All cause
Zanobetti et al., 2014
121 communities
65+
All cause
aACS (American Cancer Survey), AARP (American Association of Retired Persons), NPACT (National Particle
Components Toxicity). See Appendix C Table C-1 for additional study details.
•	PM2.5 air quality scenarios evaluated: We first estimate health risks associated with air
quality adjusted to simulate "just meeting" the current primary PM2.5 standards (i.e., the
annual standard with its level of 12.0 |ig/m3 and the 24-hour standard with its level of 35
|ig/m3). We additionally evaluate the potential for alternative annual standards with levels of
9.0, 10.0 and 11.0 |ig/m3 to reduce estimated risk, relative to the current standards. As
discussed above (section 3.1.2), there is greater uncertainty regarding whether a revised 24-
hour standard (i.e., with a lower level) would appropriately limit PM2.5-associated health
risks by limiting the PM2.5 concentrations that make up the middle portion of the air quality
distribution (i.e., where epidemiologic studies provide the strongest support for reported
associations). However, we recognize the potential for considering a revised 24-hour
standard in this review (discussed below in section 3.5.2.4.2). Therefore, to provide insight
into the possible public health implications of a revised 24-hour standard, we also examine
an alternative 24-hour standard with a level of 30 |ig/m3.58
•	Model-based approach to adjusting air quality: Air quality modeling is used to simulate
just meeting the current standards and alternative standards with levels of 10.0 |ig/m3
(annual) and 30 |ig/m3 (24-hour). The air quality modeling employs a hybrid approach that
combines CMAQ-modeled surfaces59 and ambient monitoring data to generate ambient PM2.5
estimates for 2015 on a national grid with 12-km horizontal resolution (downscaler). The
modeled 2015 PM2.5 concentrations were then adjusted using one of two approaches60 for
each air quality scenario (discussed in detail in Appendix C, section C.1.4):
58	We also estimate population risks for recent (i.e., unadjusted) ambient PM2.5 concentrations (Appendix C).
59	https://www.epa.gov/cmaa
611 These two modeling approaches provided sensitivity analyses on key aspects of the HHRA and are not additive.
3-84

-------
-	Reductions in primarily-emitted PM2.5 (Pri-PM)\ This approach simulates air
quality scenarios of interest by preferentially adjusting modeled directly emitted
PM.61
-	Reductions in secondarily produced PM2.5 (Sec-PM): This approach simulates air
quality scenarios of interest by preferentially adjusting modeled SO2 and NOx
precursor emissions to simulate changes in secondarily formed PM2.5.62
• Linear interpolation/extrapolation to additional annual standard levels: In addition to
the hybrid modeling approach described above, we also employ linear interpolation and
extrapolation to simulate just meeting alternative annual standards with levels of 11.0 (i.e.,
interpolated between 12.0 and 10.0 |ag/m3) and 9.0 |j,g/m3 (i.e., extrapolated from 12.0 and
10.0 |jg/m3), respectively (illustrated in Figure 3-11). This interpolation/extrapolation was
only performed for the subset of 30 urban study areas where the annual standard was
controlling in all air quality scenarios evaluated.
11.23
Modeled
attainment of
current standard
(12)
10.55
Interpolated
attainment of
alternate standard
(11)
Modeled
attainment of
alternative
standard (10)
Extrapolated
attainment of
alternate standard
(9)
Linear slope
Figure 3-11. Illustration of approach to adjusting air quality to simulate just meeting
annual standards with levels of 11.0 and 9.0 jig/m3.
• Characterization of variability and uncertainty in the risk estimates: Both quantitative
and qualitative methods have been used to characterize variability and uncertainty in the risk
estimates (Appendix C, section C.3), including:
-	Inclusion of 95 percent confidence intervals for risk estimates: When modeling
risk, we generate confidence intervals for each risk estimate. The confidence
intervals reflect the standard error associated with the effect estimate reported in
the epidemiologic study that is used to estimate risk.
-	Sensitivity analyses: For several of the mortality endpoints, we include a range of
risk estimates reflecting epidemiology studies conducted in various populations
and using a variety of study designs (e.g., differing in the methods used to
estimate exposures and to control for potential confounders). We also estimate
risk using two approaches to adjust air quality to simulate just meeting the current
and alternative standards (i.e., Pri-PM and Sec-PM adjustment approaches).
61	In locations for which air quality scenarios cannot be simulated by adjusting modeled directly emitted PM alone,
modeled SO2 and NOx precursor emissions are additionally adjusted to simulate changes in secondarily formed
PM2.5 (Appendix C, section C.1.4).
62	In locations for which air quality scenarios cannot be simulated by adjusting modeled precursor emissions alone, a
proportional adjustment of air quality is subsequently applied (Appendix C, section C. 1.4).
3-85

-------
- Qualitative uncertainty assessment: We additionally perform qualitative
evaluations of the potential for key sources of uncertainty to impact the magnitude
and direction of risk estimates (Appendix C, section C.3.2).
3.3.2 Results of the Risk Assessment
This section presents estimates of PM2.5-associated mortality risks for urban study areas
(additional results are available in Appendix C, section C.2). These results are shown as point
estimates with 95 percent confidence intervals for air quality adjusted to simulate just meeting
the current, and potential alternative, standards. For alternative standards, we provide tables that
include the total or absolute risk, the change in or delta risk, and the percent risk reduction 63
We also quantify the percent of baseline incidence, which estimates the percent of total
incidence (i.e., the total public health burden associated with that health effect) that is associated
with ambient PM2.5 exposure.64 In addition to tables, we also provide figures to illustrate how
risks are distributed across annual average ambient PM25 concentrations. Figures present results
for IHD mortality associated with long-term PM2.5 exposures, based on the study by Jerrett et al.
(2016). Additional results are presented in Appendix C (section C.2).
The sections below present risk estimates for the full set of 47 modeled urban study areas
(section 3.3.2.1), the subset of 30 areas for which the annual PM2.5 standard is controlling
(section 3.3.2.2), and the subset of 11 areas for which the 24-hour PM2.5 standard is controlling
(section 3.3.2.3). Uncertainties in the risk assessment are summarized in section 3.3.2.4.
3.3.2.1 Summary of Risk Estimates for 47 Urban Study Areas
Risk estimates for the 47 urban study areas are presented in Table 3-5 and Table 3-6.
Table 3-5 presents absolute risk estimates for air quality just meeting the current primary PM2.5
standards and alternative standards. Table 3-6 presents differences in estimated risk between air
quality just meeting the current standards and air quality just meeting alternative standards. More
specifically, the risk estimates presented in the column labeled "Alternative Annual Standard (10
ug/m3)" reflect the reductions estimated (compared to the current standards) in the subset of
study areas for which the alternative annual standard, with a level of 10.0 ug/m3, is controlling.
Risk estimates presented in the column labeled "Alternative 24-hour Standard (30 ug/m3)" reflect
the reductions estimated in the subset of study areas for which the alternative 24-hour standard,
63	Absolute risk refers to risk associated with the full increment of exposure associated with either the current or
alternative standard. Both delta risk and percent risk reduction reflect the change in risk in going from the current
standard to a specific alternative standard, with delta risk referring to the change in incidence (i.e., premature
PM2 5-attributable mortality) and percent risk reduction referring to the percent change when comparing risk
under the current standard to risk under simulation of an alternative standard.
64	In other words, the percent of the effect associated with PM2 5 exposure. For example, risk results estimate that 13-
14% of all IHD mortality in 2015 was associated with PM2 5 exposure (Table 3-5).
3-86

-------
with a level of 30 ng/m3, is controlling. The smaller reductions estimated for the alternative 24-
hour standard reflect the smaller number of study areas controlled by the 24-hour standard and
the relatively small population in those areas. Key observations from these results are
summarized below.
Table 3-5. Estimates of PM2.5-associated mortality for air quality adjusted to just meet the
current or alternative standards (47 urban study areas).
Endpoint
Study
Air quality
simulation
approach*
Current Standad
Absolute Risk
(12/35 |jg/m3)
CS (12/35)
%of
baseline**
Alternative Standard Absolute Risk
Alternative Annual
(10 ng/m3)
Alternative 24-hr
(30 |jg/m3)
Long-term exposure related mortality
IHD Jerrett 2016 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
16,500 (12,600-20,300)
14.1
14,400 (11,000-17,700)
16,400 (12,500-20,000)
16,800 (12,800-20,500)
14.3
14,200 (10,900-17,500)
16,500 (12,600-20,200)
15,600 (11,600-19,400)
13.3
13,600 (10,100-17,000)
15,400 (11,500-19,200)
15,800 (11,800-19,600)
13.4
13,400 (9,970-16,700)
15,600 (11,600-19,400)
All-cause Di 2017 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
Thurston 2015 Pri-PM
Sec-PM
46,200 (45,000-47,500)
8.4
40,300 (39,200-41,400)
45,700 (44,500-47,000)
46,900 (45,600-48,200)
8.5
39,700 (38,600-40,800)
46,200 (44,900-47,500)
51,300 (41,000-61,400)
7.1
44,700 (35,700-53,500)
50,700 (40,500-60,700)
52,100 (41,600-62,300)
7.2
44,000 (35,100-52,700)
51,300 (41,000-61,400)
13,500 (2,360-24,200)
3.2
11,700 (2,050-21,100)
13,300 (2,330-24,000)
13,700 (2,400-24,600)
3.2
11,500 (2,010-20,700)
13,500 (2,360-24,200)
Lung cancer Turner 2016 Pri-PM
Sec-PM
3,890 (1,240-6,360)
8.9
3,390 (1,080-5,560)
3,850 (1,230-6,300)
3,950 (1,260-6,460)
9.1
3,330 (1,060-5,470)
3,890 (1,240-6,370)
Short-term exposure related mortality
All cause Baxter 2017 Pri-PM
Sec-PM
Ito 2013 Pri-PM
Sec-PM
Zanobetti 2014 Pri-PM
Sec-PM
2,490 (983-4,000)
0.4
2,160 (850-3,460)
2,460 (970-3,950)
2,530 (998-4,060)
0.4
2,120 (837-3,400)
2,490 (982-3,990)
1,180 (-16-2,370)
0.2
1,020 (-14-2,050)
1,160 (-16-2,340)
1,200 (-16-2,400)
0.2
1,000 (-14-2,020)
1,180 (-16-2,370)
3,810 (2,530-5,080)
0.7
3,300 (2,190-4,400)
3,760 (2,500-5,020)
3,870 (2,570-5,160)
0.7
3,250 (2,160-4,330)
3,810 (2,530-5,070)
*Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
** CS denotes the current standard.
3-87

-------
Table 3-6. Estimated reduction in PM2.5-associated mortality for alternative annual and 24-
hour standards (47 urban study areas).
Endpoint
Study
Air quality
simulation
approach*
Delta Risk
% Risk Reduction
CS-AS
Annual Standard
(10 ng/m3)**
CS-AS
24-hr Standard
(30 ng/m3)**
Annual
Standard
(12-10)
24-hr
Standard
(35-30)
Long-term exposure related mortality
IHD Jerrett 2016 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
2,390 (1,800-2,970)
200 (150-249)
12.6
1.1
2,870 (2,160-3,570)
266 (200-331)
15.0
1.4
2,240 (1,640-2,830)
187 (137-237)
12.7
1.1
2,690 (1,970-3,400)
250 (183-315)
15.1
1.4
All-cause Di 2017 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
Thurston 2015 Pri-PM
Sec-PM
6,440 (6,260-6,630)
573 (557-589)
12.9
1.2
7,800 (7,580-8,020)
772 (750-793)
15.4
1.5
7,100 (5,640-8,550)
644 (511-776)
13.0
1.2
8,630 (6,860-10,400)
828 (658-997)
15.6
1.5
1,830 (316-3,320)
168 (29-305)
13.2
1.2
2,230 (387-4,060)
209 (36-381)
15.9
1.5
Lung cancer Turner 2016 Pri-PM
Sec-PM
548 (170-921)
42 (13-70)
13.0
1.0
670 (208-1,120)
61 (19-102)
15.6
1.4
Short-term exposure related mortality
All cause Baxter 2017 Pri-PM
Sec-PM
Ito 2013 Pri-PM
Sec-PM
Zanobetti 2014 Pri-PM
Sec-PM
335 (132-537)
30 (12-48)
13.5
1.3
408 (160-654)
39 (15-62)
16.1
1.6
158 (-2-317)
14(0-29)
13.4
1.2
192 (-3-386)
18(0-37)
16.1
1.5
513 (341-684)
46 (30-61)
13.4
1.2
622 (413-830)
62 (41-82)
16.0
1.6
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
** CS denotes the current standard and AS denotes the alternative standard.
Drawing from the information in Table 3-5 and Table 3-6, we make the following key
observations:
•	Air quality adjusted to simulate just meeting the current PM2.5 standards
-	Long-term PM2.5 exposures are estimated to be associated with as many as 52,100
premature deaths (all-cause), including 16,800 IHD deaths and 3,950 lung cancer
deaths, annually across the 47 study areas (and approximately 54 million people
over the age of 30). These estimates account for approximately 3-9% of all-cause,
13-14% of IHD, and 9% of lung cancer mortality in these areas, respectively.65
-	Short-term PM2.5 exposures are estimated to be associated with up to 3,870 deaths
annually across the 47 study areas.
-	The approach used to adjust air quality (i.e., Pri-PM and Sec-PM) did not have a
substantial impact on overall risk estimates (also see Appendix C, section C.1.4)
•	Air quality adjusted to just meet potential alternative standards
65 Mortality risk estimates for specific endpoints (e.g., IHD and lung cancer) are distinct subsets of total mortality.
3-88

-------
-	Compared to the current standards, risks are estimated to decrease when air
quality is adjusted to just meet an alternative annual standard with a level of 10.0
|j,g/m3 or an alternative 24-hour standard with a level of 30 |j,g/m3 (Table 3-6).66
-	Substantially larger risk reductions are estimated in the urban study areas for
which the annual standard is controlling than in the study areas for which the 24-
hour standard is controlling, reflecting the larger population in the study areas
controlled by the annual standard.
-	The approach used to adjust air quality did not have a substantial impact on
estimated reductions in PM2.5-associated mortality.
3.3.2.2 Summary of Risk Estimates for a Broader Range of Alternative Annual Standards
This section explores the potential impacts of a range of alternative annual standard
levels using interpolation and extrapolation of the modeled PM2.5 concentrations. Table 3-7 and
Table 3-8 below present mortality risk estimates for potential alternative annual standards with
levels of 11.0, 10.0, and 9.0 |ig/m3, based on the subset of 30 urban study areas for which the
annual standard is controlling under all air quality scenarios evaluated. Figure 3-12 and Figure 3-
13 present distributions of absolute (total) risk associated with air quality adjusted to just meet
the current and alternative annual standards and the risk reductions estimated for each alternative
annual standard (relative to the current standard), respectively.67
66	In most study areas, the risk reductions presented for an annual standard with a level of 10.0 ng/m3 reflect the
difference between air quality with a maximum three-year design value of 12.0 |ig/m3 and air quality with a
maximum three-year design value of 10.0 |ig/m3. Similarly, in most study areas, the risk reduction presented for a
24-hour standard with a level of 30 ng/m3 reflects the difference between air quality with a maximum three-year
design value of 35 |ig/m3 and air quality with a maximum three-year design value of 30 |ig/m3. However, in a
small number of study areas, the "starting concentration" for the annual standard are below 12.0 |ig/m3 (four
study areas: Riverside-San Bernardino-Ontario, CA; Stockton-Lodi, CA; Bakersfield, CA; and Hanford-
Corcoran, CA) or the starting concentration for the 24-hr standard are below 35 |ig/m3 (two study areas
Pittsburgh, PA and South Bend-Mishawaka, IN-MI:). This is because, in these areas, the controlling standard for
air quality adjusted to just meet the current standards is different from the controlling standard for air quality
adjusted to simulate just meeting the alternatives evaluated.
67	As noted above, Figure 3-12 and Figure 3-13 present estimates of IHD mortality associated with long-term PM2 5
exposures, based on the study by Jerrett et al. (2016).
3-89

-------
Table 3-7. Estimates of PM2.5-associated mortality for the current and potential alternative
annual standards in the 30 study areas where the annual standard is controlling.
Endpoint
Study
Air quality
simulation
approach*
Current Standad
Absolute Risk
(12/35 Mg/m3)
CS (12/35
|jg/m3)
%of
baseline**
Alternative Annual Standard (absolute risk)
11 |jg/m3
10 |jg/m3
9 |jg/m3
Long-term exposure related mortality
IHD Jerrett 2016 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
14,300 (10,900-17,500)
14.1
13,300 (10,200-16,300)
12,300 (9,400-15,100)
11,300 (8,610-13,900)
14,600 (11,100-17,800)
14.3
13,300 (10,200-16,400)
12,100 (9,240-14,900)
10,900 (8,280-13,400)
13,500 (10,100-16,800)
13.3
12,500 (9,340-15,600)
11,600 (8,620-14,500)
10,600 (7,900-13,300)
13,700 (10,200-17,000)
13.4
12,600 (9,360-15,600)
11,400 (8,480-14,200)
10,200 (7,590-12,800)
All-cause Di 2017 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
Thurston 2015 Pri-PM
Sec-PM
39,800 (38,700-40,900)
8.4
36,900 (35,900-38,000)
34,100 (33,200-35,000)
31,200 (30,400-32,100)
40,500 (39,400-41,600)
8.5
37,000 (36,000-38,000)
33,500 (32,600-34,400)
29,900 (29,100-30,800)
44,200 (35,300-52,800)
7.1
41,000 (32,800-49,100)
37,800 (30,200-45,300)
34,600 (27,600-41,500)
45,000 (35,900-53,800)
7.2
41,000 (32,800-49,100)
37,100 (29,600-44,500)
33,200 (26,500-39,700)
11,600 (2,030-20,800)
3.2
10,700 (1,880-19,300)
9,900 (1,730-17,800)
9,050 (1,580-16,300)
11,800 (2,070-21,200)
3.2
10,800 (1,880-19,400)
9,710 (1,700-17,500)
8,650 (1,510-15,600)
Lung cancer Turner 2016 Pri-PM
Sec-PM
3,400(1,080-5,550)
8.9
3,160 (1,010-5,170)
2,920 (927-4,790)
2,670 (847-4,400)
3,460(1,110-5,650)
9.1
3,160 (1,010-5,180)
2,860 (908-4,700)
2,560 (809-4,210)
Short-term exposure related mortality
All cause Baxter 2017 Pri-PM
Sec-PM
llo 2013 Pri-PM
Sec-PM
Zanobetti 2014 Pri-PM
Sec-PM
2,150 (846-3,440)
0.4
1,990 (784-3,190)
1,830 (721-2,930)
1,670 (658-2,680)
2,190 (862-3,510)
0.4
1,990 (785-3,190)
1,790 (707-2,880)
1,600 (630-2,560)
1,010 (-14-2,040)
0.2
939 (-13-1,880)
864 (-12-1,730)
789 (-11-1,580)
1,030 (-14-2,070)
0.2
940 (-13-1,890)
847 (-11-1,700)
754 (-10-1,510)
3,280 (2,180-4,370)
0.7
3,040 (2,020-4,050)
2,790 (1,860-3,730)
2,550 (1,700-3,400)
3,340 (2,220-4,450)
0.7
3,040 (2,020-4,050)
2,740 (1,820-3,650)
2,440 (1,620-3,260)
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
** CS denotes the current standard.
3-90

-------
Table 3-8. Estimated delta and percent reduction in PM2.5-associated mortality for the
current and potential alternative annual standards in the 30 study areas where the
annual standard is controlling.
Endpoint
Study
Air quality
simulation
approach*
Delta Risk (CS-AS)**
% Risk Reduction
(CS-AS)**
12-11 (jg/m3
12-10 (jg/m3
12-9 (jg/m3
12-11
(jg'm3
12-10
(jg'm3
12-9
(jg'm3
Long-term exposure related mortality
IHD Jerrett 2016 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
1,140 (859-1,420)
2,270 (1,710-2,830)
3,390 (2,550-4,210)
7%
14%
21%
1,400(1,050-1,740)
2,770 (2,090-3,450)
4,130 (3,110-5,130)
8%
17%
25%
1,070 (785-1,360)
2,130 (1,560-2,690)
3,180 (2,340-4,010)
7%
14%
21%
1,310 (960-1,660)
2,600 (1,910-3,280)
3,880 (2,850-4,890)
8%
17%
25%
All-cause Di 2017 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
Thurston 2015 Pri-PM
Sec-PM
3,070 (2,980-3,160)
6,120 (5,950-6,300)
9,150 (8,890-9,410)
7%
14%
21%
3,800 (3,690-3,900)
7,560 (7,340-7,770)
11,300 (11,000-11,600)
9%
17%
26%
3,390 (2,690-4,080)
6,760 (5,370-8,140)
10,100 (8,030-12,200)
7%
14%
22%
4,190 (3,330-5,050)
8,350 (6,640-10,100)
12,500 (9,930-15,000)
9%
17%
26%
871 (151-1,590)
1,740 (301-3,170)
2,610 (452-4,740)
7%
15%
22%
1,080 (187-1,970)
2,160 (374-3,930)
3,230 (561-5,870)
9%
18%
27%
Lung cancer Turner 2016 Pri-PM
Sec-PM
262 (81-441)
522 (162-877)
780 (243-1,310)
7%
14%
21%
327 (101-550)
651 (202-1,090)
972 (303-1,630)
9%
17%
26%
Short-term exposure related mortality
All cause Baxter 2017 Pri-PM
Sec-PM
No 2013 Pri-PM
Sec-PM
Zanobetti 2014 Pri-PM
Sec-PM
160 (63-256)
319(126-512)
478 (188-767)
7%
15%
22%
197 (78-316)
394 (155-632)
592 (233-948)
9%
18%
27%
75 (-1-151)
150 (-2-302)
226 (-3-453)
7%
15%
22%
93 (-1-187)
186 (-2-374)
279 (-4-561)
9%
18%
27%
244 (162-325)
487 (324-650)
731 (486-975)
7%
15%
22%
301 (200-402)
603 (400-804)
904 (600-1,210)
9%
18%
27%
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
** CS denotes the current standard and AS denotes the alternative standard.
3-91

-------
Annual Modeling
Scenario
Just meeting
12 ng/m3
Just meeting
11 ng/m3
Just meeting
10 ng/m3
Just meeting
9 ng/m3
a>


to

6K-
E
tO
_c

TO
4K-
—
0)

l/>

2K-


0K-
a>


TO
,	s
6K-
E
l/>
JZ
10
TO
4K-
UJ
a;

to
Q
2K-
5

0K-
a>


TO
£
lo*
6K-
'¦M
(A
73
TO
4K-
_
a>

to

2K-
a:

0K-
ai


TO
,	„
6K-
E
to
JZ
CO
¦M
TO
4K-
UJ
a;


-------
Annual Standard
Change
Simulation
Method
Annual PM Concentration (1 |ig/m3 bins)
Total
2
3
4
5
6
7
8
9
10
11
12
13
12-11 ng/m3
Pri-PM
0
0
1
4
6
14
52
160
621
267
20
0
1,140
Sec-PM
0
0
1
3
9
14
54
258
731
295
30
0
1,400
12-10 ng/m3
Pri-PM
0
0
6
4
27
53
257
1,300
596
33
0
0
2,270
Sec-PM
0
0
8
9
30
121
639
1,350
583
28
0
0
2,770
12-9 ng/m
Pri-PM
0
1
9
27
37
281
1,860
1,110
60
0
0
0
3,390
Sec-PM
0
1
15
34
199
1,090
1,970
810
16
0
0
0
4,130
Figure 3-13. Distribution of the difference in risk estimates between the current annual
standard (level of 12.0 ng/m3) and alternative annual standards with levels of 11.0,10.0,
and 9.0 ng/m3 for the subset of 30 urban study areas where the annual standard is
controlling.69
Drawing from the information in Table 3-7, Table 3-8, Figure 3-12, and Figure 3-13, we
note the following key observations:
•	For air quality just meeting the current annual standard, in the subset of 30 study areas in
which the annual standard is controlling, long-term PM2.5 exposures are estimated to be
associated with as many as 45,000 total deaths and 14,600 IHD deaths annually, accounting
for approximately 3-9% and 13-14% of baseline mortality, respectively. The majority of this
estimated risk is associated with annual average PM2.5 concentrations from 10 to 12 |ig/m3
(Figure 3-12).
•	Compared to the current annual standards, air quality adjusted to meet alternative annual
standards with lower levels is associated with reductions in estimated IHD mortality risk
across the 30 study areas (i.e., 7 to 9% reduction for a level of 11.0 |ig/m3; 14 to 18%
reduction for a level of 10.0 |ig/m3; 21 to 27% reduction for a level of 9.0 |ig/m3) (Table 3-8
and Figure 3-12).
•	The magnitude of estimated risk reduction increases as alternative annual standards with
lower levels are simulated, and these estimated risk reductions are associated with lower
ambient PM2.5 concentrations. Specifically, for air quality adjusted to simulate just meeting
an annual standard with a level of 11.0 |ig/m3, the majority of risk reduction occurs in grid
cells with ambient PM2.5 concentrations between 9 and 11 |ig/m3; for air quality adjusted to
simulate just meeting an annual standard with a level of 10.0 |ig/m3, the majority of risk
reduction occurs in grid cells with ambient PM2.5 concentrations between 8 and 10 |ig/m3;
and for air quality adjusted to simulate just meeting an annual standard with a level of 9.0
|ig/m3, the majority of risk reduction occurs in grid cells with ambient PM2.5 concentrations
between 7 and 9 |ig/m3 70 (Figure 3-13).
69 Risks are presented as integers rounded to three significant digits and aggregated into 1 ng/m3 bins. Bins begin at
the whole number value indicated and include values up to, but not including, the next whole number (e.g., risk
occurring at PM concentrations of 6.00 to 6.99 are shown in the bin at 6). Risk is estimated in this figure using
Jerrett et al. (2016).
711 Compared to adjusting primary PM2 5 emissions, adjustment of PM precursor emissions resulted in substantially
larger estimated risk reductions at 7 ng/m3.
3-93

-------
3.3.2.3 Summary of Risk Estimates for a Potential Alternative 24-Hour Standard
Table 3-9 presents risk estimates and key observations for the subset of 11 urban study
areas in which the 24-hour standard controls the simulated attainment of all modeled standard
levels. For air quality just meeting the current 24-hour standard, long-term PM2.5 exposures are
estimated to be associated with as many as 2,970 total deaths and 870 IHD deaths annually,
accounting for approximately 3-8% and 12-13% of baseline mortality, respectively. Compared to
the current standard, air quality just meeting an alternative 24-hour standard with a level of 30
|ig/m3 is associated with reductions in estimated risk of 14 to 18%.
Table 3-9. Estimates of PM2.5-associated mortality for the current 24-hour standard, and
an alternative, in the 11 study areas where the 24-hour standard is controlling.
Endpoint
Study
Air quality
simulation
approach*
Current Standad
Absolute Risk
(12/35 pg/m3)
CS
(12/35 pg/m3)
% of baseline*
Alternative Standard
Absolute Risk
(30 pg/m3)
Delta Risk: CS-AS
(daily 30 |jg/m3)"
% Risk
Reduction
(CS-AS)"
Long-term exposure related mortality
IHD Jerrett 2016 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
870 (665-1,070)
13.3
769 (586-945)
115 (87-144)
14%
862 (658-1,060)
13.1
786 (599-965)
87 (65-108)
17%
820 (610-1,020)
12.5
724 (538-903)
108 (79-137)
14%
811 (604-1,010)
12.4
739 (550-922)
82 (60-103)
17%
All-cause Di 2017 Pri-PM
Sec-PM
Pope 2015 Pri-PM
Sec-PM
Thurston 2015 Pri-PM
Sec-PM
2,650 (2,570-2,720)
7.7
2,320 (2,260-2,390)
348 (338-358)
14%
2,630 (2,550-2,700)
7.6
2,390 (2,330-2,460)
249 (242-256)
17%
2,970 (2,370-3,560)
6.5
2,600 (2,080-3,120)
388 (308-467)
14%
2,950 (2,350-3,530)
6.4
2,680 (2,140-3,220)
279 (222-336)
17%
778 (136-1,400)
2.9
681 (119-1,230)
99(17-181)
15%
771 (135-1,390)
2.9
701 (123-1,260)
72(13-131)
18%
Lung cancer Turner 2016 Pri-PM
Sec-PM
183 (58-300)
8.4
161 (51-265)
24 (7-40)
14%
181 (58-297)
8.3
165 (52-270)
18(6-30)
17%
Short-term exposure related mortality
All cause Baxter 2017 Pri-PM
Sec-PM
ho 2013 Pri-PM
Sec-PM
Zanobetti 2014 Pri-PM
Sec-PM
142 (56-228)
0.3
124(49-199)
18(7-29)
15%
141 (56-226)
0.3
128 (51-206)
13(5-21)
18%
69 (-1-138)
0.1
60 (-1-120)
9(0-18)
15%
68 (-1-137)
0.1
62 (-1-124)
6(0-13)
18%
217 (145-290)
0.6
190 (126-253)
28 (18-37)
15%
216 (143-287)
0.6
196 (130-261)
20 (13-26)
18%
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
" CS denotes the current standard and AS denotes the alternative standard.
3.3.2.4 Variability and Uncertainty in Risk Estimates
We characterize variability and uncertainty associated with risk estimates using several
quantitative and qualitative approaches, as described in detail in Appendix C (section C.3).
Approaches to addressing key uncertainties include the following:
• Evaluating various effect estimates for the same health endpoint: In some instances, the
effect estimate used has only a small impact on risk estimates (i.e., IHD mortality using
effect estimates from Jerrett et al., 2016) versus Pope et al., 2015), see Table 3-5). By
contrast, for other mortality endpoints, such as all-cause mortality associated with long-term
exposures (e.g., Di et al., 2017b) and Pope et al. (2015) versus Thurston et al., 2016)), the use
of different effect estimates can have a larger impact (Table 3-5). The degree to which
3-94

-------
different concentration-response functions result in different risk estimates could reflect
differences in study design and/or study populations evaluated, as well as other factors.
•	Evaluating multiple methods for simulating air quality scenarios: The approach used to
adjust air quality (i.e., Pri-PM and Sec-PM adjustments) has little impact on overall estimates
of risk (e.g., see Table 3-5). However, the adjustment approach has a larger impact on the
distribution of risk reductions, particularly for the level of 9.0 |j,g/m3 (Figure 3-13).
•	Characterizing the 95 percent confidence intervals associated with risk estimates: There
is considerable variation in the range of confidence intervals associated with the point
estimates generated for this analysis (see Table 3-5), with some health endpoint/study
combinations displaying substantially greater variability than others (e.g., short-term PM2.5
exposure and all-cause mortality based on effect estimates from Ito et al. (2013) versus long-
term PM2.5 exposure IHD mortality estimates based on Jerrett et al. (2016)). There are a
number of factors potentially responsible for the varying degrees of statistical precision in
effect estimates, including sample size, exposure measurement error, degree of control for
confounders/effect modifiers, and variability in PM2.5 concentrations.
•	Qualitative assessment of additional sources of uncertainty: Based in part on WHO
(2008) guidance and on guidance documents developed by the EPA (U.S. EPA, 2001, U.S.
EPA, 2004), we have also completed a qualitative characterization of sources of uncertainty
including an assessment of both the magnitude and direction of impact of those uncertainties
on risk estimates. The classification of the magnitude of impact for sources of uncertainty
includes three levels: (a) low (unlikely to produce a sufficient impact on risk estimates to
affect their interpretation), (b) medium (potential to have a sufficient impact to affect
interpretation), and (c) high (likely to have an impact sufficient to affect interpretation). For
several of the sources, we provide a classification between these levels (e.g., low-medium,
medium-high).71 Sources of uncertainty given at least a medium classification include the
following (from Appendix C, Table C-32):72
o Use of air quality modeling to adjust PM2.5 concentrations: The baseline
and adjusted air quality concentration fields were developed using modeling
to fill spatial and temporal gaps in monitoring and explore "what if' scenarios.
State-of-the-science modeling methods were used, but modeling-related biases
and errors introduce uncertainty into the PM2.5 concentration estimates. In
addition, due to the national scale of the assessment, scenarios are based on
changing modeled emissions of primary PM2.5 or NOx and SO2 from all
anthropogenic sources throughout the U.S. by fixed percentages. Although
this approach tends to target the key sources in each area, it does not tailor
71	Additional information is available in Appendix C, section C.3.
72	We also identified several additional factors judged to have less than a medium classification of impact on the risk
estimates generate, including: (a) the temporal mismatch between ambient air quality data characterizing
exposure and mortality in long-term exposure-related epidemiology studies, (b) compositional and source
differences in PM, (c) exposure measurement error in epidemiology studies assessing the relationship between
mortality and exposure to ambient PM2 5, (d) lag structure in short-term exposure-related mortality epidemiology
studies, and (e) assumed causal association between PM and mortality that supports modeling changes in risk
associated with future changes in ambient PM2 5. See Table C-32 in Appendix C for additional discussion of these
sources of uncertainty.
3-95

-------
emission changes to specific periods or sources. The two adjustment cases
span a wide range of emission conditions, but these cases are necessarily a
subset of the full set of possible emission scenarios that could be used to
adjust PM2.5 concentrations to simulate "just meeting" standards.
Use of linear interpolation/extrapolation to adjust air quality: The use of
interpolation and extrapolation to simulate just meeting annual standards with
levels of 11.0 and 9.0 |ag/m3, respectively, does not fully capture potential
non-linearities associated with real-world changes in air quality.
Potential confounding of the PM2.5-mortality effect: Factors are considered
potential confounders if demonstrated in the scientific literature to be related
to health effects and correlated with PM2.5. Omitting potential confounders
from analyses could either increase or decrease the magnitude of PM2.5 effect
estimates (e.g., Di et al., 2017b, Figure S2 in Supplementary Materials). Thus,
not accounting for confounders can introduce uncertainty into effect estimates
and, consequently, into the risk estimates generated using those effect
estimates. Confounders vary according to study design, exposure duration,
and health effect. For studies of short-term exposures, confounders may
include meteorology (e.g., temperature, humidity), day of week, season,
medication use, allergen exposure, and long-term temporal trends. For studies
of long-term exposures, confounders may include socioeconomic status, race,
age, medication use, smoking status, stress, noise, and occupational
exposures. While various approaches to control for potential confounders have
been adopted across the studies used in the risk assessment, and across the
broader body of PM2.5 epidemiologic studies assessed in the ISA, no
individual study adjusts for all potential confounders.
Potential for exposure error: Epidemiologic studies have employed a
variety of approaches to estimate population-level PM2.5 exposures (e.g.,
stationary monitors, hybrid modeling approaches). These approaches are
based on using measured or predicted ambient PM2.5 concentrations as
surrogates for population exposures. As such, exposure estimates in
epidemiologic studies are subject to exposure error. The ISA notes that, while
bias in either direction can occur, exposure error tends to result in
underestimation of health effects in epidemiologic studies of PM exposure
(U.S. EPA, 2019, section 3.5). Consistent with this, a recent study by Hart et
al. (2015) reports that correction for PM2.5 exposure error using personal
exposure information results in a moderately larger effect estimate for long-
term PM2.5 exposure and mortality (though with wider confidence intervals).
This error in the underlying epidemiologic studies contributes to uncertainty
in the risk estimates that are based on concentration-response relationships in
those studies. Beyond the exposure error in epidemiologic studies themselves,
the use of a different approach to represent exposures in the risk assessment
(i.e., 12 x 12 km gridded surface based on modeling) could introduce
additional error into risk estimates.
Shape of the concentration-response relationship at low ambient PM
concentrations: Interpreting the shapes of concentration-response
3-96

-------
relationships, particularly at PM2.5 concentrations near the lower end of the air
quality distribution, can be complicated by relatively low data density in the
lower concentration range, the possible influence of exposure measurement
error, and variability among individuals with respect to air pollution health
effects. These sources of variability and uncertainty tend to smooth and
"linearize" population-level concentration-response functions, and thus could
obscure the existence of a threshold or nonlinear relationship (U.S. EPA,
2015, section 6.c).
3.3.3 Conclusions from the risk assessment
The risk assessment estimates that the current primary PM2.5 standards could allow a
substantial number of PIVh.s-associated deaths in the U.S. For example, when air quality in the 47
study areas is adjusted to simulate just meeting the current standards, the risk assessment
estimates from about 16,000 to 17,000 long-term PM2.5 exposure-related deaths from ischemic
heart disease in a single year (i.e., confidence intervals range from about 12,000 to 21,000
deaths). The absolute numbers of estimated PIVh.s-associated deaths vary widely across exposure
durations, endpoints, populations, and concentration-response functions. In addition, limitations
in the underlying data and approaches (summarized above) lead to uncertainty regarding absolute
estimates of PIVh.s-associated risk for any given air quality scenario. However, the general
magnitude of risk estimates supports the potential for significant public health impacts in
locations meeting the current primary PM2.5 standards. This is particularly the case given that the
large majority of PIVh.s-associated deaths for air quality just meeting the current standards are
estimated at annual average PM2.5 concentrations from about 10 to 12 |j,g/m3. These annual
average PM2.5 concentrations fall well-within the range of long-term average concentrations over
which key epidemiologic studies provide strong support for reported positive and statistically
significant PM2.5 health effect associations.
Compared to the current annual standard, meeting a revised annual standard with a lower
level is estimated to reduce PIVh.s-associated health risks by about 7 to 9% for a level of 11.0
|ig/m3, 14 to 18% for a level of 10.0 |ig/m3, and 21 to 27% for a level of 9.0 |ig/m3. As the
magnitude of estimated risk reductions increases at lower levels, these estimated risk reductions
are associated with lower ambient PM2.5 concentrations. Specifically, for air quality adjusted to
simulate just meeting an annual standard with a level of 11.0 |ig/m3, the majority of risk
reduction occurs at annual average PM2.5 concentrations between 9 and 11 |ig/m3; for air quality
adjusted to simulate just meeting an annual standard with a level of 10.0 |ig/m3, the majority of
risk reduction occurs at PM2.5 concentrations between 8 and 10 |ig/m3; and for air quality
adjusted to simulate just meeting an annual standard with a level of 9.0 |ig/m3, the majority of
risk reduction occurs at PM2.5 concentrations between 7 and 9 |ig/m3. Compared to a lower
annual standard level, revising the level of the 24-hour standard to 30 |j,g/m3 is estimated to
3-97

-------
lower PM2.5-associated risks across a more limited range of areas, largely confined to areas
located in the western U.S. (several of which are also likely to experience risk reductions upon
meeting a revised annual standard).
3.4 CASAC ADVICE AND PUBLIC COMMENTS
As part of its review of the draft PA, the CASAC has provided advice on the adequacy of
the public health protection afforded by the current primary PM2.5 standards. Its advice is
documented in a letter sent to the EPA Administrator (Cox, 2019). In this letter, the committee
does not reach consensus on whether the scientific and technical information support retaining or
revising the current annual PM2.5 standard.73 In particular, though the CASAC agrees that there is
a long-standing body of health evidence supporting relationships between PM2.5 exposures and
various health outcomes, including mortality and serious morbidity effects, individual CASAC
members "differ in their assessments of the causal and policy significance of these associations"
(Cox, 2019, p. 8 of consensus responses). Drawing from this evidence, "some CASAC
members" express support for retaining the current annual standard while "other members"
express support for revising that standard in order to increase public health protection (Cox,
2019, p.l of letter). These views are summarized below.
The CASAC members who support retaining the current annual standard express the
view that substantial uncertainty remains in the evidence for associations between PM2.5
exposures and mortality or serious morbidity effects. These committee members assert that "such
associations can reasonably be explained in light of uncontrolled confounding and other potential
sources of error and bias" (Cox, 2019, p. 8 of consensus responses). They note that associations do
not necessarily reflect causal effects, and they cite recent reviews (i.e.,Henneman et al., 2017;
Burns et al., 2019) to support their position that in intervention studies, "reductions of PM2.5
concentrations have not clearly reduced mortality risks" (Cox, 2019, p. 8 of consensus responses).
These members of the CASAC additionally contend that recent epidemiologic studies reporting
positive associations at lower estimated exposure concentrations mainly confirm what was
anticipated or already assumed in setting the 2012 NAAQS, and that such studies do not provide
new information calling into question the existing standard. Thus, they advise that, "while the data
on associations should certainly be carefully considered, this data should not be interpreted more
strongly than warranted based on its methodological limitations" (Cox, 2019, p. 8 of consensus
responses).
73 In contrast, the CASAC reaches the consensus conclusion that the recent scientific evidence does not call into
question the adequacy of the 24-hour PM2 5 standard (Cox 2019, p. 11 of consensus responses).
3-98

-------
These members of the CASAC further conclude that the PM2.5 risk assessment does not
provide a valid basis for revising the current standards. This conclusion is based on concerns that
1) "the risk assessment treats regression coefficients as causal coefficients with no justification or
validation provided for this decision;" 2) the estimated regression concentration-response
functions "have not been adequately adjusted to correct for confounding, errors in exposure
estimates and other covariates, model uncertainty, and heterogeneity in individual biological (causal)
[concentration-response] functions;" 3) the estimated concentration-response functions "do not
contain quantitative uncertainty bands that reflect model uncertainty or effects of exposure and
covariate estimation errors;" and 4) "no regression diagnostics are provided justifying the use of
proportional hazards... and other modeling assumptions" (Cox, 2019, p. 9 of consensus responses).
These committee members also contend that details regarding the derivation of concentration-
response functions, including specification of the beta values and functional forms, are not well-
documented, hampering the ability of readers to evaluate these design details. Thus, these
members "think that the risk characterization does not provide useful information about whether
the current standard is protective" (Cox, 2019, p. 11 of consensus responses).
Drawing from their evaluation of the evidence and the risk assessment, these committee
members conclude that "the Draft PM PA does not establish that new scientific evidence and
data reasonably call into question the public health protection afforded by the current 2012 PM2.5
annual standard" (Cox, 2019, p.l of letter).
In contrast, "[o]ther members of CASAC conclude that the weight of the evidence,
particularly reflecting recent epidemiology studies showing positive associations between PM2.5
and health effects at estimated annual average PM2.5 concentrations below the current standard,
does reasonably call into question the adequacy of the 2012 annual PM2.5 [standard] to protect
public health with an adequate margin of safety" (Cox, 2019, p.l of letter). The committee
members who support this conclusion note that the body of health evidence for PM2.5 includes
not only the repeated demonstration of associations in epidemiologic studies, but also includes
support for biological plausibility established by human clinical and animal toxicology studies.
They point to recent studies demonstrating that the associations between PM2.5 and health effects
occur in a diversity of locations, in different time periods, with different populations, and using
different exposure estimation and statistical methods. They conclude that "the entire body of
evidence for PM health effects justifies the causality determinations made in the Draft PM ISA"
(Cox, 2019, p. 8 of consensus responses).
The members of the CASAC who support revising the current annual standard
particularly emphasize recent findings of associations with PM2.5 in areas with average long-term
PM2.5 concentrations below the level of the annual standard and studies that show positive
associations even when estimated exposures above 12 [j,g/m3 are excluded from analyses. They
3-99

-------
find it "highly unlikely" that the extensive body of evidence indicating positive associations at
low estimated exposures could be fully explained by confounding or by other non-causal
explanations (Cox, 2019, p. 8 of consensus responses). They additionally conclude that "the risk
characterization does provide a useful attempt to understand the potential impacts of alternate
standards on public health risks" (Cox, 2019, p. 11 of consensus responses). These committee
members conclude that the evidence available in this review reasonably calls into question the
protection provided by the current primary PM2.5 standards and supports revising the annual
standard to increase that protection (Cox, 2019).
We also received a number of public comments on the adequacy of the current primary
PM2.5 standards. Some of these commenters, including several representing industry groups and
states, agree with the CASAC members who conclude that the evidence supports retaining the
current standards. These public commenters often cite the same types of uncertainties that are
highlighted by members of the CASAC who support retaining (e.g., potential for confounding,
exposure error, etc.). Other public commenters, including those representing environmental and
public health organizations and several members of the academic research community, conclude
that the current primary PM2.5 standards should be revised in order to increase public health
protection. These commenters generally cite the large body of evidence supporting relationships
between PM2.5 exposures and mortality or serious morbidity-related outcomes, including studies
reporting such outcomes for PM2.5 air quality likely to be allowed in locations meeting the
current standards. They conclude that the existing body of epidemiologic studies appropriately
considers potential confounders and sources of error, and that this evidence provides robust
support for revising the current standards.
3.5 CONCLUSIONS ON THE PRIMARY PM2.5 STANDARDS
This section describes our conclusions regarding the adequacy of the current primary
PM2.5 standards (section 3.5.1) and regarding potential alternatives for consideration (section
3.5.2). As described more fully in section 3.1.2, our approach to reaching conclusions is based on
considering the EPA's assessment of the current scientific evidence for health effects attributable
to PM2.5 exposures (discussed in detail in the ISA; U.S. EPA, 2019), quantitative assessments of
PM2.5-associated health risks, and analyses of PM2.5 air quality. We also consider the range of
advice received from the CASAC (Cox, 2019) and comments from the members of the public.
These considerations and conclusions are intended to inform the Administrator's judgments
regarding primary standards for fine particles that are requisite to protect public health with an
adequate margin of safety. We seek to provide as broad an array of policy options as is
supportable by the available science, recognizing that the selection of a specific approach to
3-100

-------
reaching final decisions on the primary PM2.5 standards will reflect the judgments of the
Administrator as to what weight to place on the various types of information.
3.5.1 Current Standards
We initially consider the adequacy of the current primary PM2.5 standards. As discussed
more fully in section 3.1.2, our approach recognizes that the current annual standard (based on
arithmetic mean concentrations) and 24-hour standard (based on 98th percentile concentrations),
together, are intended to protect the public health against the full distribution of short- and long-
term PM2.5 exposures. In considering the combined effects of these standards, we recognize that
changes in PM2.5 air quality designed to meet an annual standard would likely result not only in
lower short- and long-term PM2.5 concentrations near the middle of the air quality distribution,
but also in fewer and lower short-term peak PM2.5 concentrations. Additionally, changes
designed to meet a 24-hour standard, with a 98th percentile form, would result not only in fewer
and lower peak 24-hour PM2.5 concentrations, but also in lower annual average PM2.5
concentrations. Thus, our focus in evaluating the current primary standards is on the protection
provided by the combination of the annual and 24-hour standards against the distribution of both
short- and long-term PM2.5 exposures.
Our consideration of the adequacy of the current annual and 24-hour PM2.5 standards is
framed by the first overarching policy-relevant question posed at the beginning of this chapter:
• Does the currently available scientific evidence and risk-based information support
or call into question the adequacy of the public health protection afforded by the
current annual and 24-hour PM2.5 standards?
In answering this question, we consider the nature of the health effects reported to occur
following short- or long-term PM2.5 exposures, the strength of the evidence supporting those
effects, and the evidence that certain populations may be at increased risk (discussed in more
detail in sections 3.2.1 and 3.2.2); the PM2.5 exposures shown to cause effects and the ambient
concentrations in locations where PM2.5 health effect associations have been reported (section
3.2.3); estimates of PM2.5-associated health risks for air quality adjusted to simulate just meeting
the current annual and 24-hour primary PM2.5 standards (section 3.3); and advice from the
CASAC, based on its review of the draft PA (Cox, 2019). These considerations, and our
conclusions on the current primary PM2.5 standards, are summarized below.
As an initial matter, we note the longstanding body of health evidence supporting
relationships between PM2.5 exposures (short- and long-term) and mortality or serious morbidity
effects. The evidence available in this review (i.e., assessed in U.S. EPA, 2019 and summarized
above in section 3.2.1) reaffirms, and in some cases strengthens, the conclusions from the 2009
ISA regarding the health effects of PM2.5 exposures (U.S. EPA, 2009). Much of this evidence
3-101

-------
comes from epidemiologic studies conducted in North America, Europe, or Asia that
demonstrate generally positive, and often statistically significant, PM2.5 health effect
associations. Such studies report associations between estimated PM2.5 exposures and non-
accidental, cardiovascular, or respiratory mortality; cardiovascular or respiratory hospitalizations
or emergency room visits; and other mortality/morbidity outcomes (e.g., lung cancer mortality or
incidence, asthma development). Recent experimental evidence, as well as evidence from panel
studies, strengthens support for potential biological pathways through which PM2.5 exposures
could lead to the serious effects reported in many population-level epidemiologic studies. This
includes support for pathways that could lead to cardiovascular, respiratory, nervous system, and
cancer-related effects.
Epidemiologic studies report PM2.5 health effect associations with mortality and/or
morbidity across multiple U.S. cities and in diverse populations, including in studies examining
populations and lifestages that may be at comparatively higher risk of experiencing a PM2.5-
related health effect (e.g., older adults, children). Such studies employ various designs and
examine a variety of health outcomes, geographic areas, and approaches to controlling for
confounding variables. With regard to controlling for potential confounders in particular, key
studies use a wide array of approaches. Time-series studies control for potential confounders that
vary over short time intervals (e.g., including temperature, humidity, dew point temperature, and
day of the week) while cohort studies control for community- and/or individual-level
confounders that vary spatially (e.g., including income, race, age, socioeconomic status,
smoking, body mass index, and annual weather variables such as temperature and humidity)
(Appendix B, Table B-12). Sensitivity analyses indicate that adding covariates to control for
potential confounders can either increase or decrease the magnitude of PM2.5 effect estimates,
depending on the covariate, and that none of the covariates examined can fully explain the
association with mortality (e.g., Di et al., 2017b, Figure S2 in Supplementary Materials). Thus,
while no individual study adjusts for all potential confounders, a broad range of approaches have
been adopted across studies to examine confounding, supporting the robustness of reported
associations.
Available studies additionally indicate that PM2.5 health effect associations are robust
across various approaches to estimating PM2.5 exposures and across exposure windows. This
includes recent studies that estimate exposures using ground-based monitors alone and studies
that estimate exposures using data from multiple sources (e.g., satellites, land use information,
modeling), in addition to monitors. While none of these approaches eliminates the potential for
exposure error in epidemiologic studies, such error does not call into question the fundamental
findings of the broad body of PM2.5 epidemiologic evidence. In fact, the ISA notes that while
bias in either direction can occur, exposure error tends to lead to underestimation of health
3-102

-------
effects in epidemiologic studies of PM exposure (U.S. EPA, 2019, section 3.5). Consistent with
this, a recent study reports that correction for PM2.5 exposure error using personal exposure
information results in a moderately larger effect estimate for long-term PM2.5 exposure and
mortality (Hart et al., 2015). While most PM2.5 epidemiologic studies have not employed similar
corrections for exposure error, several studies report that restricting analyses to populations in
close proximity to a monitor (i.e., in order to reduce exposure error) result in larger PM2.5 effect
estimates (e.g., Willis et al., 2003; Kloog et al., 2013). The consistent reporting of PM2.5 health
effect associations across exposure estimation approaches, even in the face of exposure error,
together with the larger effect estimates reported in some studies that have attempted to reduce
exposure error, provides further support for the robustness of associations between PM2.5
exposures and mortality and morbidity.
Consistent findings from the broad body of epidemiologic studies are also supported by
an emerging body of studies employing "causal inference" or quasi-experimental statistical
approaches to further inform the causal nature of the relationship between long- or short-term
term PM2.5 exposure and mortality (U.S. EPA, 2019, sections 11.1.2.1, 11.2.2.4). These studies
are summarized above in section 3.2.1.1, including a recent accountability study that reports a
reduction in mortality following reductions in ambient PM2.5 due to the introduction of diesel
emission controls (Yorifuji et al., 2016).74 Other recent studies additionally report that declines in
ambient PM2.5 concentrations over a period of years have been associated with decreases in
mortality rates and increases in life expectancy, improvements in respiratory development, and
decreased incidence of respiratory disease in children, further supporting the robustness of PM2.5
health effect associations reported in the epidemiologic evidence (summarized in sections 3.2.1
to 3.2.3).
In addition to broadening our understanding of the health effects that can result from
exposures to PM2.5 and strengthening support for some key effects (e.g., nervous system effects,
cancer), recent epidemiologic studies strengthen support for health effect associations at
relatively low ambient PM2.5 concentrations. Studies that examine the shapes of concentration-
response functions over the full distribution of ambient PM2.5 concentrations have not identified
a threshold concentration, below which associations no longer exist (U.S. EPA, 2019, section
1.5.3). While such analyses are complicated by the relatively sparse data available at the lower
74 Air pollution accountability studies have reported mixed results overall (e.g., as reviewed in Burns et al., 2019 and
Henneman et al., 2017). However, many of the available studies have not focused on PM2 5, were not able to
attribute changes in ambient PM2 5 concentrations to the interventions under evaluation, and/or were not able to
disentangle health impacts of the intervention from background trends in health. The study by Yorifuji et al.
(2016), included in the review by Burns et al. (2019), is an example of a study that was able to link a particular
policy intervention to a decline in ambient PM2 5 concentrations, and that did include a control population to
correct for background trends in mortality.
3-103

-------
end of the air quality distribution (U.S. EPA, 2019, section 1.5.3), several studies report positive
and statistically significant associations in additional analyses restricted to annual average PM2.5
exposures below 12 |j,g/m3 (Lee et al., 2015; Di et al., 2017b) and 10 |j,g/m3 (Shi et al., 2016), or
to daily exposures below 25 |j,g/m3 (Di et al., 2017a), 30 |j,g/m3 (Shi et al., 2016), and 35 |j,g/m3
(Lee et al., 2015).
These and other recent studies provide support for health effect associations at lower
ambient PM2.5 concentrations than in previous reviews. For example, in the last review key
epidemiologic studies that were conducted in the U.S. or Canada, and that supported "causal" or
"likely to be causal" determinations in the ISA, reported generally positive and statistically
significant associations with mortality or morbidity for PM2.5 air quality distributions with
overall mean concentrations at or above 12.8 |j,g/m3 (U.S. EPA, 2011, Figure 2-8). In the current
review, a large number of key studies report positive and statistically significant associations for
air quality distributions with lower overall mean PM2.5 concentrations (i.e., Figure 3-7 and Figure
3-8). These key studies indicate such associations consistently for distributions with long-term
mean PM2.5 concentrations at or above 8.1 |j,g/m3 (8.2 |j,g/m3 based on studies that use monitors
alone to estimate PM2.5 exposures), with the large majority (and all but one key U.S. study)
reporting overall mean PM2.5 concentrations at or above 9.6 |j,g/m3 (10.7 |j,g/m3 based on studies
that use monitors alone). Air quality distributions with such low mean concentrations are likely
to be allowed by the current PM2.5 standards, based on analyses of the relationships between
maximum annual PM2.5 design values and annual average concentrations (i.e., averaged across
multiple monitors in the same area) (section 3.2.3.2.1; Appendix B, section B.7).75
In assessing the adequacy of the current standard, we also consider what key
epidemiologic studies may indicate for the current standards by calculating values similar to
PM2.5 design values, based on monitored air quality from the locations and time periods
evaluated by those studies (i.e., section 3.2.3.2.2). This approach identifies study-relevant PM2.5
air quality metrics similar to those used by the EPA to determine whether areas meet or violate
the PM NAAQS. Compared to study-reported mean PM2.5 concentrations, such "pseudo-design
values" also have the advantage of being consistently calculated across key studies, regardless of
how the studies themselves estimate PM2.5 exposures (e.g., averaging across monitors,
predictions from hybrid modeling approaches).
For some key studies that report positive and statistically significant PM2.5 health effect
associations, substantial portions of study area populations (e.g., > 50% or 75%) lived in
75 Given that the annual standard is the controlling standard across much of the U.S. (e.g., see section 3.3), the PM2 5
air quality distributions that occur in most locations meeting the current annual PM2 5 standard are also likely to
meet the current 24-hour standard (i.e., illustrated in Chapter 2, Figure 2-11).
3-104

-------
locations with air quality likely to have met both the current annual and 24-hour PM2.5 standards
over study periods (or substantial portions of health events occurred in such areas) (section
3.2.3.2.2). While there is uncertainty in interpreting analyses of PM2.5 pseudo-design values (e.g.,
some study locations and time periods would have met the current standards while others would
have violated those standards, unmonitored areas are excluded from analyses; section 3.2.3.2.2),
the importance of these uncertainties is lessened for studies with the large majority of the study
area population in locations with pseudo-design values well-below current standard levels (e.g.,
Pinault et al., 2016; Shi et al., 2016; Weichenthal et al., 2016c). This uncertainty is also lessened
for key studies reporting that positive and statistically significant associations persist in analyses
restricted to relatively low annual average PM2.5 exposure estimates (e.g., below 12 |j,g/m3 in Di
et al., 2017b; below 10 |j,g/m3 in Shi et al., 2016), particularly given that the excluded exposure
estimates account for about half of the deaths in the entire cohort.76 Thus, analyses of PM2.5
pseudo-design values support the occurrence of positive and statistically significant PM2.5 health
effect associations based largely on air quality likely to have met the current primary standards.
In addition to the evidence, we also consider what the risk assessment indicates with
regard to the adequacy of the current primary PM2.5 standards. The risk assessment estimates that
the current primary PM2.5 standards could allow a substantial number of deaths in the U.S., with
the large majority of those deaths associated with long-term PM2.5 exposures. For example, when
air quality in the 47 study areas is adjusted to simulate just meeting the current standards, the risk
assessment estimates from about 16,000 to 17,000 PIVh.s-related deaths from ischemic heart
disease in a single year (i.e., for long-term exposures; confidence intervals range from about
12,000 to 21,000 deaths). While the absolute numbers of estimated PIVh.s-associated deaths vary
widely across exposure durations, endpoints, populations, and concentration-response functions,
the general magnitude of risk estimates supports the potential for significant public health
impacts in locations meeting the current primary PM2.5 standards. This is particularly the case
given that the large majority of PM2.5-associated deaths for air quality just meeting the current
standards are estimated at annual average PM2.5 concentrations from about 10 to 12 |j,g/m3. These
annual average PM2.5 concentrations fall well-within the range of long-term average
concentrations over which key epidemiologic studies provide strong support for reported positive
and statistically significant PM2.5 health effect associations.
Based on the information summarized above, and discussed in more detail in sections 3.2
and 3.3 of this PA, we particularly note the following in reaching conclusions on the current
primary PM2.5 standards:
76 PM2 5 effect estimates in these restricted analyses are slightly larger than in those based on the entire cohort.
3-105

-------
•	There is a long-standing body of strong health evidence demonstrating relationships between
long- or short-term PM2.5 exposures and a variety of outcomes, including mortality and
serious morbidity effects. Studies published since the last review have reduced key
uncertainties and broadened our understanding of the health effects that can result from
exposures to PM2.5.
•	Recent U.S. and Canadian epidemiologic studies provide support for generally positive and
statistically significant health effect associations across a broad range of ambient PM2.5
concentrations, including for air quality distributions with overall mean concentrations lower
than in the last review and for distributions likely to be allowed by the current primary PM2.5
standards.
•	Analyses of PM2.5 pseudo-design values additionally support the occurrence of positive and
statistically significant health effect associations based largely on air quality likely to have
met the current annual and 24-hour primary standards.
•	The risk assessment estimates that the current primary PM2.5 standards could allow a
substantial number of PM2.5-associated deaths in the U.S. The large majority of these
estimated deaths are associated with the annual average PM2.5 concentrations near (and above
in some cases) the average concentrations in key epidemiologic studies reporting positive and
statistically significant health effect associations.
When taken together, we reach the conclusion that the available scientific evidence, air quality
analyses, and the risk assessment, as summarized above, can reasonably be viewed as calling
into question the adequacy of the public health protection afforded by the combination of the
current annual and 24-hour primary PM2.5 standards.
In contrast to this conclusion, a conclusion that the current primary PM2.5 standards do
provide adequate public health protection would place little weight on the broad body of
epidemiologic evidence reporting generally positive and statistically significant health effect
associations, particularly for PM2.5 air quality distributions likely to have been allowed by the
current primary standards, or on the PM2.5 risk assessment. Rather, such a conclusion would
place greater weight on uncertainties and limitations in the evidence and analyses (i.e., discussed
in sections 3.2.3 and 3.3.2 above), including the following:
•	Uncertainty in the biological pathways through which PM2.5 exposures could cause serious
health effects increases as the ambient concentrations being considered fall farther below the
PM2.5 exposure concentrations shown to cause effects in experimental studies. In the current
review, such studies generally examine the occurrence of PIVh.s-attributable effects following
exposures to PM2.5 concentrations well-above those likely to occur in the ambient air in areas
meeting the current primary PM2.5 standards (i.e., discussed in section 3.2.3.1).
•	Uncertainty in the potential public health impacts of air quality improvements increases as
the ambient concentrations being considered fall farther below those present in studies that
report improved health with reductions in PM2.5 concentrations. In the current review, such
3-106

-------
studies evaluate air quality improvements with "starting" mean PM2.5 concentrations (i.e.,
prior to the reductions being evaluated) from about 13 to > 20 |j,g/m3 (i.e., Table 3-3).77
• Uncertainty in the risk assessment results from uncertainties in the underlying epidemiologic
studies, in the air quality adjustments, and in the application of study and air quality
information to develop quantitative estimates of PIVh.s-associated mortality risks (section
3.3.2.4).
The considerations and conclusions discussed above are intended to inform the
Administrator's judgments regarding the current primary PM2.5 standards. In presenting these
considerations and conclusions, we seek to provide information on a range of policy options, and
on the potential approaches to viewing the scientific evidence and technical information that
could potentially support various options. We recognize that the selection of a particular
approach to reaching final decisions on the primary PM2.5 standards will reflect the judgments of
the Administrator as to what weight to place on the various types of evidence and information,
including associated uncertainties. Given that this PA seeks to provide information on the range
of policy options that could be supported by the scientific information, and given our conclusion
(noted above) that the evidence and information can reasonably be viewed as calling into
question the adequacy of the current primary PM2.5 standards, in the next section we additionally
consider support for potential alternative standards.
3.5.2 Potential Alternative Standards
In this section, we consider the potential alternative primary PM2.5 standards that could be
supported by the evidence and quantitative information available in this review. These
considerations are framed by the following overarching policy-relevant question, posed at the
beginning of this chapter:
• What is the range of potential alternative standards that could be supported by the
available scientific evidence and risk-based information to increase public health
protection against short- and long-term fine particle exposures?
In answering this question, we consider each of the elements of the annual and 24-hour PM2.5
standards: indicator, averaging time, form, and level. The sections below discuss our
consideration of these elements, and our conclusions that (1) it is appropriate to consider revising
the level of the current annual standard, in conjunction with retaining the current indicator,
averaging time, and form of that standard, to increase public health protection against fine
77 As noted above, these retrospective studies tend to include data from earlier time periods where ambient PM2 5
concentrations in the U.S. were considerably higher than they are at present.
3-107

-------
particle exposures and (2) depending on the decision made on the annual standard, consideration
could be given to either retaining or revising the level of the 24-hour PM2.5 standard.
3.5.2.1 Indicator
In initially setting standards for fine particles in 1997, the EPA concluded it was
appropriate to control fine particles as a group, rather than singling out any particular component
or class of fine particles. The Agency noted that community health studies had found significant
health effect associations using various indicators of fine particles, and that health effects in a
large number of areas had significant mass contributions from differing components or sources
of fine particles. In addition, a number of toxicological and controlled human exposure studies
had reported health effects following exposures to high concentrations of numerous fine particle
components (62 FR 38667, July 18, 1997). In establishing a size-based indicator in 1997 to
distinguish fine particles from particles in the coarse mode, the EPA noted that the available
epidemiologic studies of fine particles were based largely on PM2.5 mass. The selection of a 2.5
|j,m size cut additionally reflected the regulatory importance of defining an indicator that would
more completely capture fine particles under all conditions likely to be encountered across the
U.S. and the monitoring technology that was generally available (62 FR 38666 to 38668, July 18,
1997).
Since the 1997 review, studies that evaluate fine particle-related health effects continue to
provide strong support for such effects using PM2.5 mass as the metric for fine particle exposures.
Subsequent reviews have recognized the strength of this evidence, concluding that it has
continued to support a PM2.5 mass-based indicator for a standard meant to protect against fine
particle exposures. In the last review, some studies had additionally examined health effects of
exposures to particular sources or components of fine particles, or to the ultrafine fraction of fine
particles. Based on limitations in such studies, together with the continued strong support for
effects of PM2.5 exposures, the Agency retained PM2.5 mass as the indicator for fine particles and
did not supplement the PM2.5 standards with standards based on particle composition or on the
ultrafine fraction (78 FR 3123, January 15, 2013).
As in the last review, studies available in the current review continue to provide strong
support for health effects following long- and short-term PM2.5 exposures (U.S. EPA, 2019).
While some studies evaluate the health effects of particular sources of fine particles, or of
particular fine particle components, evidence from these studies does not identify any one source
or component that is a better predictor of health effects than PM2.5 mass (U.S. EPA, 2019,
section 1.5.4). The ISA specifically notes that the results of recent studies confirm and further
support the conclusion of the 2009 ISA that many PM2.5 components and sources are associated
with health effects, and the evidence does not indicate that any one source or component is
3-108

-------
consistently more strongly related with health effects than PM2.5 mass (U.S. EPA, 2019, section
1.5.4). In addition, the evidence for health effects following exposures specifically to the
ultrafine fraction of fine particles continues to be far more limited than the evidence for PM2.5
mass as a whole. As discussed in the ISA, the lack of a consistent UFP definition in health
studies and across disciplines, together with the variety of approaches to administering and
measuring UFP in those studies, contribute to such limitations (U.S. EPA, 2019, section 1.4.3).
Thus, for reasons similar to those discussed in the last review (78 FR 3121 to 3123, January 15,
2013), we conclude that the available information continues to support the PM2.5 mass-based
indicator and remains too limited to support a distinct standard for any specific PM2.5 component
or group of components, and too limited to support a distinct standard for the ultrafine fraction.
3.5.2.2 Averaging Time
In 1997, the EPA initially set an annual PM2.5 standard to protect against health effects
associated with both long- and short-term PM2.5 exposures, and a 24-hour standard to supplement
the protection afforded by the annual standard (62 FR 38667 to 38668, July 18, 1997). In
subsequent reviews, the EPA retained both annual and 24-hour averaging times, largely
reflecting the strong evidence for health effects associated with annual and daily PM2.5 exposure
estimates (71 FR 61164, October 17, 2006; 78 FR 3123 to 3124, January 15, 2013).
In the current review, epidemiologic and controlled human exposure studies have
examined a variety of PM2.5 exposure durations. Epidemiologic studies continue to provide
strong support for health effects associated with both long- and short-term PM2.5 exposures based
on annual (or multiyear) and 24-hour PM2.5 averaging periods, respectively.
With regard to short-term exposures in particular, a smaller number of epidemiologic
studies examine associations between sub-daily PM2.5 exposures and respiratory effects,
cardiovascular effects, or mortality. Compared to 24-hour PM2.5 exposure estimates, associations
with sub-daily estimates are less consistent and, in some cases, smaller in magnitude (U.S. EPA,
2019, section 1.5.2.1). In addition, studies of sub-daily exposures typically examine subclinical
effects, rather than the more serious population-level effects that have been reported to be
associated with 24-hour exposures (e.g., mortality, hospitalizations). Taken together, the ISA
concludes that epidemiologic studies do not indicate sub-daily averaging periods are more
closely associated with health effects than the 24-hour average exposure metric (U.S. EPA, 2019,
section 1.5.2.1).
Additionally, while recent controlled human exposure studies provide consistent evidence
for cardiovascular effects following PM2.5 exposures for less than 24 hours (i.e., < 30 minutes to
5 hours), exposure concentrations in these studies are well-above the ambient concentrations
typically measured in locations meeting the current standards (section 3.2.3.1). Thus, these
3-109

-------
studies also do not suggest the need for additional protection against sub-daily PM2.5 exposures,
beyond that provided by the current primary standards.
Drawing from the evidence assessed in the ISA, and the observations noted above, we
reach the conclusion that the available evidence continues to provide strong support for
consideration of retaining the current annual and 24-hour averaging times. The available
evidence suggests that PM2.5 standards with these averaging times, when coupled with
appropriate forms and levels, can protect against the range of long- and short-term PM2.5
exposures that have been associated with health effects. Thus, as in the last review, the currently
available evidence does not support considering alternatives to the annual and 24-hour averaging
times for standards meant to protect against long- and short-term PM2.5 exposures.
3.5.2.3 Form
The form of a standard defines the air quality statistic that is to be compared to the level
in determining whether an area attains that standard. As in other recent reviews, our foremost
consideration in reaching conclusions on form is the adequacy of the public health protection
provided by the combination of the form and the other elements of the standard.
As noted above, in 1997 the EPA initially set an annual PM2.5 standard to protect against
health effects associated with both long- and short-term PM2.5 exposures and a 24-hour standard
to provide supplemental protection, particularly against the short-term exposures to "peak" PM2.5
concentrations that can occur in some areas (62 FR 38667 to 38668, July 18, 1997). The EPA
established the form of the annual PM2.5 standard as an annual arithmetic mean, averaged over 3
years, from single or multiple community-oriented monitors. That is, the level of the annual
standard was to be compared to measurements made at each community-oriented monitoring site
or, if specific criteria were met, measurements from multiple community-oriented monitoring
sites could be averaged together (i.e., spatial averaging) (62 FR 38671 to 38672, July 18, 1997).
In the 1997 review, the EPA also established the form of the 24-hour PM2.5 standard as the 98th
percentile of 24-hour concentrations at each monitor within an area (i.e., no spatial averaging),
averaged over three years (62 FR at 38671 to 38674, July 18, 1997). In the 2006 review, the EPA
retained these standard forms but tightened the criteria for using spatial averaging with the
annual standard (78 FR 3124, January 15, 2013).78
In the last review, the EPA's consideration of the form of the annual PM2.5 standard again
included a focus on the issue of spatial averaging. An analysis of air quality and population
demographic information indicated that the highest PM2.5 concentrations in a given area tended
78 Specifically, the Administrator revised spatial averaging criteria such that "(1) [t]he annual mean concentration at
each site shall be within 10 percent of the spatially averaged annual mean, and (2) the daily values for each
monitoring site pair shall yield a correlation coefficient of at least 0.9 for each calendar quarter (71 FR 61167,
October 17, 2006).
3-110

-------
to be measured at monitors in locations where the surrounding populations were more likely to
live below the poverty line and to include larger percentages of racial and ethnic minorities (U.S.
EPA, 2011, p. 2-60). Based on this analysis, the PA concluded that spatial averaging could result
in disproportionate impacts in minority populations and populations with lower SES. The
Administrator concluded that public health would not be protected with an adequate margin of
safety in all locations, as required by law, if disproportionately higher PM2.5 concentrations in
low income and minority communities were averaged together with lower concentrations
measured at other sites in a large urban area. Therefore, she concluded that the form of the
annual PM2.5 standard should be revised to eliminate spatial averaging provisions (78 FR 3124,
January 15, 2013).
In the last review, the EPA also considered the form of the 24-hour PM2.5 standard. The
Agency recognized that the existing 98th percentile form for the 24-hour standard was originally
selected to provide a balance between limiting the occurrence of peak 24-hour PM2.5
concentrations and identifying a stable target for risk management programs. Updated air quality
analyses in the last review provided additional support for the increased stability of the 98th
percentile PM2.5 concentration, compared to the 99th percentile (U.S. EPA, 2011, Figure 2-2, p.
2-62). Thus, the Administrator concluded that it was appropriate to retain the 98th percentile form
for the 24-hour PM2.5 standard (78 FR 3127, January 15, 2013).
Nothing in the evidence that has become available since the last review calls into
question the current forms of the annual and 24-hour PM2.5 standards. As discussed above
(section 3.2.3.2), epidemiologic studies continue to provide strong support for health effect
associations with both long-term (e.g., annual or multi-year) and short-term (e.g., mostly 24-
hour) PM2.5 exposures. These studies provide the strongest support for such associations for the
part of the air quality distribution corresponding to the bulk of the underlying data, typically
around the overall mean concentrations reported (section 3.2.3.2.1). The form of the current
annual standard (i.e., arithmetic mean, averaged over three years) remains appropriate for
targeting protection against the annual and daily PM2.5 exposures around these means of the
PM2.5 air quality distribution. In addition, controlled human exposure studies provide evidence
for health effects following single short-term PM2.5 exposures near the peak concentrations
measured in the ambient air (section 3.2.3.1). Thus, the evidence also supports retaining a
standard focused on providing supplemental protection against short-term peak exposures.
Nothing in the evidence that has become available since the last review calls into question the
decision to use a 98th percentile form for a 24-hour standard that is meant to provide a balance
between limiting the occurrence of such peak 24-hour PM2.5 concentrations and identifying a
stable target for risk management programs. Thus, when the information summarized above is
taken together, we reach the conclusion that it is appropriate in the current review to consider
3-111

-------
retaining the forms of the current annual and 24-hour PM2.5 standards, in conjunction with a
revised level as discussed below.
3.5.2.4 Level
With regard to level, we specifically address the following policy-relevant question:
• For primary PM2.5 standards defined in terms of the current averaging times and
forms, what potential alternative levels are appropriate to consider in order to increase
public health protection against long- and short-term exposures to PM2.5 in ambient
air?
In answering this question, we consider key epidemiologic studies that evaluate associations
between PM2.5 air quality distributions and mortality or morbidity, controlled human exposure
studies examining effects following short-term PM2.5 exposures, air quality analyses that help to
place these studies into a policy-relevant context, and the risk assessment estimates of PM2.5-
associated mortality under various alternative standard scenarios.
As discussed above in section 3.1.2, consideration of the evidence and analyses, as
summarized in this chapter, informs our evaluation of the public health protection that could be
provided by alternative annual and 24-hour standards with revised levels. There are various ways
to combine an annual standard (based on arithmetic mean concentrations) and a 24-hour standard
(based on 98th percentile concentrations), to achieve an appropriate degree of public health
protection. In particular, as noted in section 3.1.2, we recognize that changes in PM2.5 air quality
designed to meet an annual standard would likely result not only in lower short- and long-term
PM2.5 concentrations near the middle of the air quality distribution (i.e., around the mean of the
distribution), but also in fewer and lower short-term peak PM2.5 concentrations. Additionally,
changes designed to meet a 24-hour standard, with a 98th percentile form, would result not only
in fewer and lower peak 24-hour PM2.5 concentrations, but also in lower average PM2.5
concentrations.
However, while either standard could be viewed as providing some measure of protection
against both average exposures and peak exposures, the 24-hour and annual standards are not
expected to be equally effective at limiting both types of exposures. Specifically, the 24-hour
standard (with its 98th percentile form) is more directly tied to short-term peak PM2.5
concentrations, and thus more likely to appropriately limit exposures to such concentrations, than
to the more typical concentrations that make up the middle portion of the air quality distribution.
Therefore, compared to a standard that is directly tied to the middle of the air quality distribution,
the 24-hour standard is less likely to appropriately limit the "typical" daily and annual exposures
that are most strongly associated with the health effects observed in epidemiologic studies. In
contrast, the annual standard, with its form based on the arithmetic mean concentration, is more
likely to effectively limit the PM2.5 concentrations that comprise the middle portion of the air
3-112

-------
quality distribution, affording protection against the daily and annual PM2.5 exposures that
strongly support associations with the most serious PIVh.s-related effects in epidemiologic studies
(e.g., mortality, hospitalizations).
For these reasons, as discussed in section 3.1.2, we focus on alternative levels of the
annual PM2.5 standard as the principle means of providing increased public health protection
against the bulk of the distribution of short- and long-term PM2.5 exposures, and thus protecting
against the exposures that provide strong support for associations with mortality and morbidity in
key epidemiologic studies. We additionally consider the 24-hour standard, with its 98th percentile
form, primarily as a means of providing supplemental protection against the short-term
exposures to peak PM2.5 concentrations that can occur in some areas (e.g., those with strong
contributions from local or seasonal sources), even when overall mean PM2.5 concentrations
remain relatively low.
To inform our consideration of potential alternative annual and 24-hour standard levels,
we specifically note the following key observations regarding (1) the overall mean PM2.5
concentrations reported in U.S. or Canadian epidemiologic studies, (2) the relationships between
long-term mean PM2.5 concentrations and annual design values in U.S. CBSAs, (3) the PM2.5
pseudo-design values in study locations, (4) the PM2.5 exposures shown to cause effects in
controlled human exposure studies, and (5) estimated PM2.5-associated risks.
(1) Long-Term Mean PM2.5 Concentrations in Key Epidemiologic Studies (section 3.2.3.2)
•	Key epidemiologic studies indicate consistently positive and statistically significant health
effect associations based on air quality distributions with overall long-term mean PM2.5
concentrations at and above 8.1 |j,g/m3 (8.2 |j,g/m3 based on studies that use monitors alone to
estimate PM2.5 exposures), with mean concentrations at or above 9.6 |j,g/m3 in most key
studies (10.7 |j,g/m3 based on studies that use monitors alone to estimate PM2.5 exposures).
The ranges of ambient PM2.5 concentrations accounting for the bulk of exposures and health
data in these studies are expected to extend at least somewhat below the overall long-term
mean concentrations reported.
•	Epidemiologic studies provide more limited support for health effect associations based on
air quality distributions with lower overall mean PM2.5 concentrations. Specifically, two key
studies report positive associations between short-term PM2.5 exposures and emergency room
visits based on cities in Ontario, Canada (Weichenthal et al., 2016b and Weichenthal et al.,
2016c and), with overall mean PM2.5 concentrations around 7.0 |j,g/m3 (one of these studies
reports an association that is statistically significant). Additionally, a U.S. study (Shi et al.
(2016) reports positive and statistically significant associations in analyses restricted to
relatively low annual or 24-hour PM2.5 exposure estimates. This study does not report the
overall mean PM2.5 concentrations in restricted analyses, though such means are presumably
somewhat below those based on the overall cohort (i.e., 8.1 and 8.2 |j,g/m3).
3-113

-------
Relationships between long-term mean PM2.5 concentrations and annual design values
(section 3.2.3.3; Appendix B, section B. 7)
Areas meeting a particular annual PM2.5 standard would be expected to have average PM2.5
concentrations (i.e., averaged across the area and over time) somewhat below the level of that
standard. This is supported by analyses of monitoring data in CBSAs across the U.S., which
show that maximum annual PM2.5 design values are often 10% to 20% higher than long-term
mean PM2.5 concentrations (Appendix B, Figure B-7; Table B-9).
PM2.5 Pseudo-Design Values in Study Locations (section 3.2.3.2.2 and Appendix B,
Figure B-9)
For most key epidemiologic studies with PM2.5 pseudo-design values available, about 25% or
more of study area populations lived in locations likely to have met the current primary PM2.5
standards over study periods (or about 25% or more of health events occurred in such
locations). For the U.S. studies in this group, annual pseudo-design values as low as 8.7
|j,g/m3 correspond to 25th percentiles of study area population (or health events). For the
smaller number of Canadian studies included in this group, annual pseudo-design values as
low as 6.0 |j,g/m3 correspond to the 25th percentiles of study area population (or health
events).
For several key epidemiologic studies, most of the study area populations (i.e., >50% of
those living in areas with pseudo-design values) lived in locations with air quality likely to
have met both standards over study periods (or >50% of health events occurred in locations
with such air quality). For the U.S. studies in this group, annual pseudo-design values from
9.9 to 11.7 |j,g/m3 correspond to 50111 percentiles of study area populations (or health events).
For the smaller number of Canadian studies included in this group, annual pseudo-design
values from 7.3 to 7.4 |j,g/m3 correspond to 50th percentiles of study area populations (or
health events).
For the U.S. study reporting the lowest annual average concentrations (Shi et al., 2016), an
annual pseudo-design value of 11.0 |j,g/m3 corresponds to the 75th percentile of the study area
population (i.e., 75% of the study area population lives in locations with pseudo-design
values < 11.0 |j,g/m3). For the Canadian studies with the lowest ambient PM2.5
concentrations, annual pseudo-design values from 8.4 to 8.6 |j,g/m3 correspond to 75th
percentiles of the study area populations (or health events).
PM2.5 exposures shown to cause effects in controlled human exposure studies (section
3.2.3.1)
While controlled human exposure studies support the plausibility of the serious
cardiovascular effects that have been linked with ambient PM2.5 exposures (U.S. EPA, 2019,
Chapter 6), the PM2.5 exposure concentrations evaluated in most of these studies are well-
above the ambient concentrations typically measured in locations meeting the current
primary standards (and thus well-above those likely to be measured in locations that would
meet revised standards with lower annual or 24-hour levels).
PM2.5-Associated Risk Estimates (section 3.3)
3-114

-------
•	The risk assessment estimates that, compared to the current standards, potential alternative
annual standards with levels from 11.0 down to 9.0 |j,g/m3 could reduce PM2.5-associated
mortality broadly across the U.S., including in most of the 47 urban study areas evaluated. In
such locations, estimated risk reductions range from about 7 to 9% for a level of 11.0 ng/m3,
14 to 18% for a level of 10.0 |ag/m3, and 21 to 27% for a level of 9.0 |j,g/m3. For each of these
standards, most of the risk remaining is estimated at annual average PM2.5 concentrations that
fall somewhat below the standard level.
•	Risk reductions estimated for an alternative 24-hour standard with a level of 30 |j,g/m3 are
concentrated in only a few study areas in the western U.S. (several of which could also
experience risk reductions in response to a revised annual standard with a level below 12.0
Hg/m3). In those few study areas for which risk reductions are estimated upon just meeting an
alternative 24-hour standard with a level of 30 ng/m3, reductions range from about 14 to
18%.
The information summarized in these key observations could support various decisions on
the levels of the annual and 24-hour PM2.5 standards, depending on the weight given to different
aspects of the evidence, air quality and risk information, including its uncertainties. As noted
above (section 3.1.2), in this PA we seek to provide as broad an array of policy options as is
supportable by the available evidence and quantitative information, recognizing that the selection
of a specific approach to reaching final decisions on the primary PM2.5 standards will reflect the
judgments of the Administrator as to what weight to place on the various types of evidence and
information, and on associated uncertainties. Potential approaches to considering support for
particular alternative annual and 24-hour standard levels are discussed below.
3.5.2.4.1 Alternative Annual Standard Levels
As discussed above, the degree to which particular alternative annual standard levels
below 12.0 |j,g/m3 are supported will depend on the weight placed on various aspects of the
scientific evidence, air quality and risk information, and its associated uncertainties. For
example, a level as low as about 10.0 |j,g/m3 could be supported to the extent weight is placed on
the following:
•	Setting a standard expected to maintain the PM2.5 air quality distribution below those present
in most key epidemiologic studies, recognizing that (1) the large majority of key studies
reporting positive and statistically significant health effect associations (and all but one key
U.S. study) examine distributions of ambient PM2.5 with overall mean concentrations at or
above 9.6 |ag/m3, while a few studies reporting such associations examine distributions with
overall mean concentrations just above 8.0 |j,g/m3 (section 3.2.3.2.1) and (2) analyses of
PM2.5 air quality in CBS As indicate that maximum annual PM2.5 design values are often 10%
to 20% higher than average PM2.5 concentrations (i.e., averaged across space and over
several years) suggesting that areas meeting a particular annual PM2.5 standard would be
expected to have average PM2.5 concentrations somewhat below the level of that standard
(section 3.2.3.2.2; Appendix B, section B.7);
3-115

-------
•	Setting the standard level at or below the pseudo-design values corresponding to about the
50th percentiles of study area populations (or health events) in most key studies (particularly
key U.S. studies), recognizing that a revised annual standard with a level as low as 10.0
Hg/m3 would be expected to maintain ambient PM2.5 concentrations below the concentrations
present during study periods for most of those populations (or below the concentrations in
locations accounting for most health events) (section 3.2.3.2.2);
•	Setting a standard estimated to reduce PIVh.s-associated health risks, such that a substantial
portion of the risk reduction is estimated at annual average PM2.5 concentrations > about 8
Hg/m3 and recognizing that these concentrations are within the range of overall means for
which key epidemiologic studies indicate consistently positive and statistically significant
health effect associations (section 3.3.2).
In selecting a particular level from 10.0 |j,g/m3 to < 12.0 |ag/m\ consideration of the
evidence could take into account individual study characteristics such as study design and
statistical approaches, precision of reported associations, study size and location, and
uncertainties in the study itself or in our analyses of study area air quality. For example, if less
weight is placed on the small number of studies reporting overall mean concentrations below 9.6
Hg/m3 and on the small number of studies with 50th percentile pseudo-design values below 10.0
Hg/m3, a standard higher than 10 |ig/m3 (but still below 12.0 |ag/m3) might be considered.
Similarly, consideration of the risk assessment could take into account the magnitude of
estimated risk reductions, compared to the current standards; the annual average PM2.5
concentrations at which those reductions are estimated to occur; and the uncertainties in the
underlying epidemiologic studies, in the air quality adjustments, or in other information that was
used to model risks. For example, concern about the uncertainty in the potential public health
importance of risk reductions estimated for a level as low as 10.0 ng/m3, much of which is
estimated at annual average PM2.5 concentrations around 8 |ag/m3, might focus consideration on
a standard level above 10 |ig/m3, where estimated risk reductions would occur at slightly higher
concentrations.
A decision to not consider annual standard levels below 10.0 |j,g/m3 might take into
account the increasing uncertainty in the degree to which lower levels would result in additional
public health improvements, due in part to the more limited amount of data available. Such a
decision could note the following regarding the increasing uncertainty at lower ambient
concentrations:
•	Few key epidemiologic studies (and only one key U.S. study) report positive and statistically
significant health effect associations for PM2.5 air quality distributions with overall mean
concentrations below 9.6 |ag/m\ and areas meeting a standard with a level of 10.0 |j,g/m3
would generally be expected to have lower long-term mean PM2.5 concentrations (and
potentially around 8.0 |j,g/m3 in some areas) (section 3.2.3.2.1; Appendix B, section B.7).
3-116

-------
•	There is increasing uncertainty in PM2.5 exposure estimates in some of the largest key studies
at lower ambient concentrations (i.e., those that use hybrid model predictions to estimate
exposures), given the more limited information available to develop and validate model
predictions (sections 2.3.3 and 3.2.3.2.1).
•	Pseudo-design values corresponding to the 50th percentiles of study area populations (or
health events) are > about 10.0 |j,g/m3 for almost all key studies, particularly those conducted
in the U.S. (section 3.2.3.2.2).
•	There is increasing uncertainty in quantitative estimates of PIVh.s-associated mortality risk for
standard levels below 10.0 |J,g/m3, given that a substantial proportion of the risk reductions
estimated for lower standard levels occur at annual average PM2.5 concentrations below 8
Hg/m3, and thus below the lower end of the range of overall mean PM2.5 concentrations in
key epidemiologic studies that consistently report positive and statistically significant
associations (section 3.3.2).
In contrast, an annual standard with a level below 10.0 ng/m3, and potentially as low as
8.0 ng/m3, could be supported to the extent greater weight is placed on the potential public health
improvements that could result from additional reductions in ambient PM2.5 concentrations (i.e.,
beyond those achieved by a standard with a level of 10.0 |ag/m3) and less weight is placed on the
limitations in the evidence that contribute to greater uncertainty at lower concentrations. For
example, a level below 10.0 |j,g/m3 could be supported to the extent greater weight is placed on
the following:
•	The two key studies in Canada with overall mean PM2.5 concentrations below 8.0 |j,g/m3 and
the potential for overall mean concentrations below 8.0 |j,g/m3 in restricted analyses in a key
U.S. study (section 3.2.3.2.1);
•	The ambient PM2.5 concentrations somewhat below overall means (e.g., corresponding the
lower quartile of underlying data), which contribute to the bulk of the data informing
reported associations (section 3.2.3.2.1);
•	Annual pseudo-design values corresponding to 25th percentiles of study area populations or
health events for most studies, recognizing that the revised standard would be expected to
maintain ambient PM2.5 concentrations below the concentrations present during study periods
for > -75% of those populations (or below the concentrations in locations accounting for >
75% of health events) (section 3.2.3.2.2);
•	Annual pseudo-design values for the smaller number of key studies conducted in Canada,
which tend to be somewhat lower than those in the U.S. (section 3.2.3.2.2);
•	The potential public health importance of the additional reductions in PM2.5-associated health
risks estimated for a level of 9.0 |j,g/m3 and the potential for continued reductions at lower
standard levels (i.e., below the lowest level examined in the risk assessment) (section 3.3).
As above, various levels from 8.0 |j,g/m3 to < 10.0 |j,g/m3 could be supported, depending
on the weight placed on specific aspects of the evidence and analyses. For example, compared to
a level of 8.0 |ag/m3, a higher level could be supported to the extent less weight is placed on the
3-117

-------
two key Canadian studies reporting overall mean concentrations below 8.0 ng/m3, on the
potential for overall mean concentrations below 8.0 |j,g/m3 in a U.S. study that reports
associations in restricted analyses, and on the three Canadian studies with the lowest pseudo-
design values. Such a judgment could also be informed by increasing uncertainty in the potential
public health importance of risks estimated for a level as low as 8.0 |ag/m\ given that such risks,
which were not quantified in the risk assessment, are likely to occur at annual average PM2.5
concentrations largely below 8 |j,g/m3 (i.e., below the mean concentrations in almost all key
epidemiologic studies).
3.5.2.4.2 Alternative 24-Hour Standard Levels
We additionally evaluate the degree to which the evidence supports considering potential
alternative levels for the 24-hour PM2.5 standard, in conjunction with the current 98th percentile
form of that standard. As discussed above (section 3.1.1), in the last review, the EPA recognized
that the annual standard would generally be the controlling standard across much of the U.S.,
except for certain areas in the western U.S. "where annual mean PM2.5 concentrations have
historically been low but where relatively high 24-hour concentrations occur, often related to
seasonal wood smoke emissions" (78 FR 3163, January 15, 2013). In such areas, the 24-hour
standard is the generally controlling standard. Thus, the EPA's approach in the last review was to
focus on the annual standard as the principle means of limiting both long- and short-term PM2.5
concentrations, recognizing that the 24-hour standard, with its 98th percentile form, would
provide supplemental protection against short-term peak exposures, particularly for areas with
high peak-to-mean ratios (e.g., areas with strong seasonal sources).
As discussed above (section 3.1.2), in the current review we again view the 24-hour
standard (with its 98fe percentile form) largely within the context of limiting short-term
exposures to peak PM2.5 concentrations. Compared to the annual standard, we recognize that the
24-hour standard is less likely to appropriately limit the more typical PM2.5 exposures (i.e.,
corresponding to the middle portion of the air quality distribution) that are most strongly
associated with the health effects observed in epidemiologic studies. Thus, as in the last review
(78 FR 3161-3162, January 15, 2013), we focus on the annual PM2.5 standard as the principle
means of providing public health protection against the bulk of the distribution of short- and
long-term PM2.5 exposures, and the 24-hour standard as a means of providing supplemental
protection against the short-term exposures to "peak" PM2.5 concentrations, such as can occur in
areas with strong contributions from local or seasonal sources.
Results of the risk assessment and of recent air quality analyses are consistent with our
reliance on the 24-hour standard to provide supplemental protection in areas with relatively low
long-term mean PM2.5 concentrations. In particular, the risk assessment indicates that the annual
3-118

-------
standard is the controlling standard across most of the urban study areas evaluated and revising
the level of the 24-hour standard to 30 |j,g/m3 would be estimated to lower PM2.5-associated risks,
compared to the current standards, largely in a few study areas located in the western U.S.
(several of which are also likely to experience risk reductions upon meeting a revised annual
standard). Additionally, recent air quality analyses indicate that almost all CBSAs with
maximum annual PM2.5 design values at or below 12.0 |j,g/m3 also have maximum 24-hour
design values below 35 |j,g/m3 (and below 30 |j,g/m3 in most areas) (Chapter 2, Figure 2-11). The
exceptions are a few CBSAs in the western U.S.
Thus, taking into account the approach described above, an important consideration is
whether additional protection is needed against short-term exposures to peak PM2.5
concentrations in areas meeting both the current 24-hour standard and the current, or a revised,
annual standard. To the extent the evidence indicates that such exposures can lead to adverse
health effects, it would be appropriate to consider alternative levels for the 24-hour standard. In
considering this issue, we evaluate the evidence from key health studies. With regard to these
studies, we particularly note the following:
•	To the extent a revised annual standard is determined to provide adequate protection against
the 24-hour and annual PM2.5 exposures associated with health effects in key epidemiologic
studies, those studies do not indicate the need for additional protection against short-term
exposures to peak PM2.5 concentrations. As discussed in detail above (section 3.2.3.2.1),
epidemiologic studies provide the strongest support for reported health effect associations for
the part of the air quality distribution corresponding to the bulk of the underlying data (i.e.,
estimated exposures and/or health events), often around the overall mean concentrations
evaluated rather than near the upper end of the distribution. Consistent with this, analyses
that exclude the upper end of the distribution of estimated exposures still find positive and
statistically significant associations with mortality. The magnitudes of the associations in
restricted analyses are similar to (Shi et al., 2016) or larger than (Di et al., 2017a) the
magnitudes of the associations based on the full cohorts, suggesting that, at a minimum,
short-term exposures to peak PM2.5 concentrations are not disproportionately responsible for
reported health effect associations.
•	Controlled human exposure studies do provide evidence for health effects following single,
short-term PM2.5 exposures to concentrations that typically correspond to upper end of the
PM2.5 air quality distribution in the U.S. (i.e., "peak" concentrations). However, most of these
studies examine exposure concentrations considerably higher than are typically measured in
areas meeting the current standards (section 3.2.3.1). In particular, while controlled human
exposure studies often report statistically significant effects on one or more indicators of
cardiovascular function following 2-hour exposures to PM2.5 concentrations at and above 120
|ig/m3 (at and above 149 |ig/m3 for vascular impairment, the effect shown to be most
consistent across studies), 2-hour ambient concentrations of PM2.5 at monitoring sites
meeting the current standards almost never exceed 32 ng/m3. In fact, even the extreme upper
end of the distribution of 2-hour PM2.5 concentrations at sites meeting the current standards
remains well-below the PM2.5 exposure concentrations consistently shown to elicit effects
3-119

-------
(i.e., 99.9th percentile of 2-hour concentrations at these sites is 68 ng/m3 during the warm
season). Thus, available PM2.5 controlled human exposure studies do not indicate the need
for additional protection against exposures to peak PM2.5 concentrations, beyond the
protection provided by the combination of the current 24-hour standard and the current or a
revised annual standard (section 3.2.3.1).
When the information summarized above is considered in the context of the 24-hour
standard, we reach the conclusion that, in conjunction with a lower annual standard level
intended to increase protection against short- and long-term PM2.5 exposures broadly across the
U.S., the evidence does not support the need for additional protection against short-term
exposures to peak PM2.5 concentrations. In particular, while epidemiologic studies do support the
need to consider increasing protection against the typical 24-hour and annual PM2.5 exposures
that provide strong support for reported health effect associations, these studies do not indicate
that such associations are strongly influenced by exposures to the peak concentrations in the air
quality distribution. Also, while controlled human exposure studies support the occurrence of
effects following single short-term exposures to PM2.5 concentrations that correspond to the peak
of the air quality distribution, these concentrations are well above those typically measured in
areas meeting the current standards. Thus, in the context of a 24-hour standard that is meant to
provide supplemental protection (i.e., beyond that provided by the annual standard alone) against
short-term exposures to peak PM2.5 concentrations, the available evidence supports consideration
of retaining the current 24-hour standard with its level of 35 |j,g/m3.
However, we also recognize that a different policy approach than that described above
could be applied to considering the level of the 24-hour standard. For example, consideration
could be given to lower 24-hour standard levels in order to increase protection across the U.S.
against the broader PM2 5 air quality distribution. If such an approach is evaluated in the current
review, consideration of 24-hour standard levels at least as low as 30 |j,g/m3 could be supported
(either alone or in conjunction with a lower annual standard level). The risk assessment estimates
that a level of 30 |j,g/m3 would increase protection compared to the current standards, though
only in a small number of study areas largely confined to the western U.S. (section 3.3.2).
Analyses of air quality in locations of some key epidemiologic studies indicate that substantial
portions of study area populations lived in locations with 24-hour PM2.5 pseudo-design values at
or below about 30 |j,g/m3 (or that substantial portions of study health events occurred in such
locations), providing additional support for considering lower levels.
If this alternative approach to revising the primary PM2.5 standards is adopted, the
uncertainty inherent in using the 24-hour standard to increase protection against the broad
distribution of PM2.5 air quality should be carefully considered. Specifically, the degree of
3-120

-------
protection provided by any particular 24-hour standard against the typical short- and long-term
PM2 5 exposures corresponding to the middle portion of the air quality distribution will vary
across locations and over time, depending on the relationship between those typical
concentrations and the short-term peak PM2 5 concentrations that are directly targeted by the 24-
hour standard (i.e., with its 98th percentile form). Thus, lowering the level of the 24-hour
standard is likely to have a more variable impact on public health than lowering the level of the
annual standard. Depending on the 24-hour standard level set, some areas could experience
reductions that are greater than warranted, based on the evidence, while others could experience
reductions that are less than warranted. Therefore, the rationale supporting this approach would
need to recognize and account for the uncertainty inherent in using 24-hour standard, with a 98th
percentile form, to increase protection against the broad distribution of PM2.5 air quality.
3.6 AREAS FOR FUTURE RESEARCH AND DATA COLLECTION
In this section, we identify key areas for additional research and data collection for fine
particles, based on the uncertainties and limitations that remain in the evidence and technical
information. Additional research in these areas could reduce uncertainties and limitations in
future reviews of the primary PM2.5 standards. Important areas for future research include the
following:
•	Further elucidating the physiological pathways through which exposures to the PM2.5
concentrations present in the ambient air across much of the U.S. could be causing mortality
and the morbidity effects shown in many epidemiologic studies. This could include the
following:
-	Controlled human exposure studies that examine longer exposure periods (e.g.,
24-hour as in Brauner et al. (2008); 5-hour as in Hemmingsen et al. (2015b)), or
repeated exposures, to concentrations typically measured in the ambient air across
the U.S.
-	Studies that evaluate the health impacts of decreasing PM2.5 exposures (e.g., due
to changes in policies or behavior, shifts in important emissions sources, or
targeted interventions).
-	Additional animal toxicological studies that evaluate exposures to low PM2.5
concentrations.
•	Additional research into "causal inference" methods in epidemiologic studies to evaluate the
causal nature of relationships between PM2.5 exposure and mortality or morbidity.
•	Improving our understanding of the PM2.5 concentration-response relationships near the
lower end of the PM2.5 air quality distribution, including the shapes of concentration-
response functions and the uncertainties around estimated functions for various health
outcomes and populations (e.g., older adults, people with pre-existing diseases, children).
3-121

-------
Understanding of the potential for particle characteristics, other than size-fractionated mass,
to influence PM toxicity (e.g., composition, oxidative potential, etc.) and the PM health
effect associations observed in epidemiologic studies.
Improving our understanding of the uncertainties inherent in the various approaches used to
estimate PM2.5 exposures in epidemiologic studies, including how those uncertainties may
vary across space and time, and over the PM2.5 air quality distribution. Approaches to
incorporating these uncertainties into quantitative estimates of PM2.5 concentration-response
relationships should also be explored.
Additional health research on ultrafine particles, with a focus on consistently defining UFPs
across studies and across disciplines (i.e., animal, controlled human exposure, and
epidemiologic studies), on using consistent exposure approaches in experimental studies, and
on improving exposure characterizations in epidemiologic studies. Also, further examine the
potential for translocation of ultrafine particles from the respiratory tract into other
compartments (i.e., blood) and organs (e.g., heart, brain), with particular emphasis on studies
conducted in humans.
Additional work to measure ultrafine particle emissions, using comparable methods to
measure emissions from various types of sources (e.g., mobile sources, fires, etc.).
Further evaluate the potential for some groups to be at higher risk of PM2.5-related effects
than the general population and the potential for PM2.5 exposures to contribute to the
development of underlying conditions that may then confer higher risk of PM2.5-related
effects. For example, research to address this latter need could include efforts to understand
the potential for long-term PM exposures to contribute to the development and progression of
atherosclerosis in adults and/or asthma in children. It could also include research to
understand the potential role of PM exposures in developmental outcomes (e.g.,
neurodevelopmental effects, reproductive and birth outcomes).
Research to further evaluate the combination of factors that contribute to differences in risk
estimates between cities, potentially including differences in exposures, demographics,
particle characteristics.
Research to improve our understanding of variability in PM2.5 exposures within and across
various populations (e.g., defined by life stage, pre-existing condition, etc.), the most health-
relevant exposure durations, as well as the temporal and spatial variability in ambient PM2.5
that is not captured by existing ambient monitors.
3-122

-------
REFERENCES
Alhanti, BA, Chang, HH, Winquist, A, Mulholland, JA, Darrow, LA and Sarnat, SE (2016).
Ambient air pollution and emergency department visits for asthma: A multi-city
assessment of effect modification by age. Journal of Exposure Science and
Environmental Epidemiology 26(2): 180-188.
Baxter, LK, Crooks, JL and Sacks, JD (2017). Influence of exposure differences on city-to-city
heterogeneity in PIVh.s-mortality associations in US cities. Environ Health 16(1): 1.
Behbod, B, Urch, B, Speck, M, Scott, JA, Liu, L, Poon, R, Coull, B, Schwartz, J, Koutrakis, P,
Silverman, F and Gold, DR (2013). Endotoxin in concentrated coarse and fine ambient
particles induces acute systemic inflammation in controlled human exposures. Occup
Environ Med 70(11): 761-767.
Bell, ML, Ebisu, K, Peng, RD, Walker, J, Samet, JM, Zeger, SL and Dominic, F (2008).
Seasonal and regional short-term effects of fine particles on hospital admissions in 202
U.S. counties, 1999-2005. Am J Epidemiol 168(11): 1301-1310.
Bell, ML, Ebisu, K, Leaderer, BP, Gent, JF, Lee, HJ, Koutrakis, P, Wang, Y, Dominici, F and
Peng, RD (2014). Associations of PM2.5 constituents and sources with hospital
admissions: analysis of four counties in Connecticut and Massachusetts (USA) for
persons > 65 years of age. Environ Health Perspect 122(2): 138-144.
Bell, ML, Son, JY, Peng, RD, Wang, Y and Dominici, F (2015). Brief report: Ambient PM2.5 and
risk of hospital admissions: do risks differ for men and women? Epidemiology 26(4):
575-579.
Bellavia, A, Urch, B, Speck, M, Brook, RD, Scott, JA, Albetti, B, Behbod, B, North, M, Valeri,
L, Bertazzi, PA, Silverman, F, Gold, D and Baccarelli, AA (2013). DNA
hypomethylation, ambient particulate matter, and increased blood pressure: Findings
from controlled human exposure experiments. Journal of the American Heart Association
2(3): e000212.
Berhane, K, Chang, CC, McConnell, R, Gauderman, WJ, Avol, E, Rapapport, E, Urman, R,
Lurmann, F and Gilliland, F (2016). Association of changes in air quality with bronchitic
symptoms in children in California, 1993-2012. J Am Med Assoc 315(14): 1491-1501.
Brauner, EV, M0ller, P, Barregard, L, Dragsted, LO, Glasius, M, Wahlin, P, Vinzents, P,
Raaschou-Nielsen, O and Loft, S (2008). Exposure to ambient concentrations of
particulate air pollution does not influence vascular function or inflammatory pathways in
young healthy individuals. Particle and Fibre Toxicology 5:13.
Bravo, MA, Ebisu, K, Dominici, F, Wang, Y, Peng, RD and Bell, ML (2017). Airborne fine
particles and risk of hospital admissions for understudied populations: Effects by
urbanicity and short-term cumulative exposures in 708 U.S. counties. Environ Health
Perspect 125(4): 594-601.
3-123

-------
Breton, CV, Salam, MT, Vora, H, Gauderman, WJ and Gilliland, FD (2011). Genetic variation in
the glutathione synthesis pathway, air pollution, and children's lung function growth. Am
J Respir Crit Care Med 183(2): 243-248.
Brook, RD, Urch, B, Dvonch, JT, Bard, RL, Speck, M, Keeler, G, Morishita, M, Marsik, FJ,
Kamal, AS, Kaciroti, N, Harkema, J, Corey, P, Silverman, F, Gold, DR, Wellenius, G,
Mittleman, MA, Rajagopalan, S and Brook, JR (2009). Insights into the mechanisms and
mediators of the effects of air pollution exposure on blood pressure and vascular function
in healthy humans. Hypertension 54(3): 659-667.
Burnett, RT and Goldberg, MS (2003). Size-fractionated particulate mass and daily mortality in
eight Canadian cities. Boston, MA, Health Effects Institute.
Burnett, RT, Stieb, D, Brook, JR, Cakmak, S, Dales, R, Raizenne, M, Vincent, R and Dann, T
(2004). Associations between short-term changes in nitrogen dioxide and mortality in
Canadian cities. Arch Environ Occup Health 59(5): 228-236.
Burns, J, Boogaard, H, Polus, S, Pfadenhauer, LM, Rohwer, AC, van Erp, AM, Turley, R and
Rehfuess, E (2019). Interventions to reduce ambient particulate matter air pollution and
their effect on health. Cochrane Database of Systematic Reviews(5).
Cangerana Pereira, FA, Lemos, M, Mauad, T, de Assuncao, JV and Nascimento Saldiva, PH
(2011). Urban, traffic-related particles and lung tumors in urethane treated mice. Clinics
66(6): 1051-1054.
Chan, E. (2019). Email and attachment from Elizabeth Chan, EPA, Question re: Request for
additional PM2.5 information from Di et al 2017. Available in PM NAAQS Review
Docket (EPA-HQ-OAR-2015-0072). December 3, 2019. Research Triangle Park, NC.
Office of Air Quality Planning and Standards.
Chen, H, Burnett, RT, Copes, R, Kwong, JC, Villeneuve, PJ, Goldberg, MS, Brook, RD, van
Donkelaar, A, Jerrett, M, Martin, RV, Brook, JR, Kopp, A and Tu, JV (2016). Ambient
fine particulate matter and mortality among survivors of myocardial infarction:
population-based cohort study. Environ Health Perspect 124(9): 1421-1428.
Correia, AW, Pope, CA, III, Dockery, DW, Wang, Y, Ezzati, M and Dominici, F (2013). Effect
of air pollution control on life expectancy in the United States: an analysis of 545 U.S.
counties for the period from 2000 to 2007. Epidemiology 24(1): 23-31.
Cox, LA. (2019). Letter from Louis Anthony Cox, Jr., Chair, Clean Air Scientific Advisory
Committee, to Administrator Andrew R. Wheeler. Re: CASAC Review of the EPA's
Policy Assessment for the Review of the National Ambient Air Quality Standards for
Particulate Matter (External Review Draft - September 2019). December 16, 2019. EPA-
CASAC-20-001. U.S. EPA HQ, Washington DC. Office of the Administrator, Science
Advisory Board. Available at:
https://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/E2F
6C7173 72016128525 84D20069DFB l/$File/EP A-C AS AC-20-001 .pdf.
3-124

-------
Crouse, DL, Peters, PA, van Donkelaar, A, Goldberg, MS, Villeneuve, PJ, Brion, O, Khan, S,
Atari, DO, Jerrett, M, Pope, CA, Brauer, M, Brook, JR, Martin, RV, Stieb, D and
Burnett, RT (2012). Risk of nonaccidental and cardiovascular mortality in relation to
long-term exposure to low concentrations of fine particulate matter: a Canadian national-
level cohort study. Environ Health Perspect 120(5): 708-714.
Crouse, DL, Peters, PA, Hystad, P, Brook, JR, van Donkelaar, A, Martin, RV, Villeneuve, PJ,
Jerrett, M, Goldberg, MS, Pope, CA, Brauer, M, Brook, RD, Robichaud, A, Menard, R
and Burnett, RT (2015). Ambient PM 2.5, O 3, and NO 2 exposures and associations with
mortality over 16 years of follow-up in the Canadian Census Health and Environment
Cohort (CanCHEC). Environ Health Perspect 123(11): 1180-1186.
Dai, L, Zanobetti, A, Koutrakis, P and Schwartz, JD (2014). Associations of fine particulate
matter species with mortality in the United States: a multicity time-series analysis.
Environ Health Perspect 122(8): 837-842.
Devlin, RB, Ghio, AJ, Kehrl, H, Sanders, G and Cascio, W (2003). Elderly humans exposed to
concentrated air pollution particles have decreased heart rate variability. Eur Respir J 40:
76S-80S.
Di, Q, Dai, L, Wang, Y, Zanobetti, A, Choirat, C, Schwartz, JD and Dominici, F (2017a).
Association of short-term exposure to air pollution with mortality in older adults. J Am
Med Assoc 318(24): 2446-2456.
Di, Q, Wang, Y, Zanobetti, A, Wang, Y, Koutrakis, P, Choirat, C, Dominici, F and Schwartz, JD
(2017b). Air pollution and mortality in the Medicare population. New Engl J Med
376(26): 2513-2522.
Dominici, F, Peng, RD, Bell, ML, Pham, L, McDermott, A, Zeger, SL and Samet, JL (2006).
Fine particulate air pollution and hospital admission for cardiovascular and respiratory
diseases. J Am Med Assoc 295(10): 1127-1134.
Dominici, F, Peng, RD, Zeger, SL, White, RH and Samet, JM (2007). Particulate air pollution
and mortality in the United States: did the risks change from 1987 to 2000? Am J
Epidemiol 166(8): 880-888.
Eum, KD, Suh, HH, Pun, VC and Manjourides, J (2018). Impact of long-term temporal trends in
fine particulate matter (PM2.5) on associations of annual PM2.5 exposure and mortality:
An analysis of over 20 million Medicare beneficiaries. Environmental Epidemiology
2(2): e009.
Franklin, M, Zeka, A and Schwartz, J (2007). Association between PM2.5 and all-cause and
specific-cause mortality in 27 US communities. Journal of Exposure Science and
Environmental Epidemiology 17(3): 279-287.
Franklin, M, Koutrakis, P and Schwartz, J (2008). The role of particle composition on the
association between PM2.5 and mortality. Epidemiology 19(5): 680-689.
3-125

-------
Gauderman, WJ, Avol, E, Gilliland, F, Vora, H, Thomas, D, Berhane, K, McConnell, R,
Kuenzli, N, Lurmann, F, Rappaport, E, Margolis, H, Bates, D and Peters, J (2004). The
effect of air pollution on lung development from 10 to 18 years of age. New Engl J Med
351(11): 1057-1067.
Gauderman, WJ, Urman, R, Avol, E, Berhane, K, McConnell, R, Rappaport, E, Chang, R,
Lurmann, F and Gilliland, F (2015). Association of improved air quality with lung
development in children. New Engl J Med 372(10): 905-913.
Gharibvand, L, Shavlik, D, Ghamsary, M, Beeson, WL, Soret, S, Knutsen, R and Knutsen, SF
(2016). The association between ambient fine particulate air pollution and lung cancer
incidence: Results from the AHSMOG-2 study. Environ Health Perspect 125(3): 378-
384.
Ghio, AJ, Kim, C and Devlin, RB (2000). Concentrated ambient air particles induce mild
pulmonary inflammation in healthy human volunteers. Am J Respir Crit Care Med
162(3): 981-988.
Ghio, AJ, Hall, A, Bassett, MA, Cascio, WE and Devlin, RB (2003). Exposure to concentrated
ambient air particles alters hematologic indices in humans. Inhal Toxicol 15(14): 1465-
1478.
Gong, H, Jr., Linn, WS, Sioutas, C, Terrell, SL, Clark, KW, Anderson, KR and Terrell, LL
(2003). Controlled exposures of healthy and asthmatic volunteers to concentrated
ambient fine particles in Los Angeles. Inhal Toxicol 15(4): 305-325.
Gong, H, Jr., Linn, WS, Terrell, SL, Anderson, KR, Clark, KW, Sioutas, C, Cascio, WE, Alexis,
N and Devlin, RB (2004). Exposures of elderly volunteers with and without chronic
obstructive pulmonary disease (COPD) to concentrated ambient fine particulate pollution.
Inhal Toxicol 16(11-12): 731-744.
Goss, CH, Newsom, SA, Schildcrout, JS, Sheppard, L and Kaufman, JD (2004). Effect of
ambient air pollution on pulmonary exacerbations and lung function in cystic fibrosis.
Am J Respir Crit Care Med 169(7): 816-821.
Greven, S, Dominici, F and Zeger, S (2011). An Approach to the Estimation of Chronic Air
Pollution Effects Using Spatio-Temporal Information. Journal of the American Statistical
Association 106(494): 396-406.
Hart, JE, Garshick, E, Dockery, DW, Smith, TJ, Ryan, L and Laden, F (2011). Long-term
ambient multi-pollutant exposures and mortality. Am J Respir Crit Care Med 183(1): 73-
78.
Hart, JE, Liao, X, Hong, B, Puett, RC, Yanosky, JD, Suh, H, Kioumourtzoglou, MA,
Spiegelman, D and Laden, F (2015). The association of long-term exposure to PM2.5 on
all-cause mortality in the Nurses' Health Study and the impact of measurement-error
correction. Environ Health 14: 38.
3-126

-------
Hazucha, MJ, Bromberg, PA, Lay, JC, Bennett, W, Zeman, K, Alexis, NE, Kehrl, H, Rappold,
A, Cascio, WE and Devlin, RB (2013). Pulmonary responses in current smokers and ex-
smokers following a two hour exposure at rest to clean air and fine ambient air particles.
Particle and Fibre Toxicology 10(1): 58.
Hemmingsen, JG, Jantzen, K, Miller, P and Loft, S (2015a). No oxidative stress or DNA
damage in peripheral blood mononuclear cells after exposure to particles from urban
street air in overweight elderly. Mutagenesis 30(5): 635-642.
Hemmingsen, JG, Rissler, J, Lykkesfeldt, J, Sallsten, G, Kristiansen, J, P, PM and Loft, S
(2015b). Controlled exposure to particulate matter from urban street air is associated with
decreased vasodilation and heart rate variability in overweight and older adults. Particle
and Fibre Toxicology 12(1): 6.
Henneman, LR, Liu, C, Mulholland, JA and Russell, AG (2017). Evaluating the effectiveness of
air quality regulations: A review of accountability studies and frameworks. Journal of the
Air Waste Management Association 67(2): 144-172.
Huang, YC, Rappold, AG, Graff, DW, Ghio, AJ and Devlin, RB (2012). Synergistic effects of
exposure to concentrated ambient fine pollution particles and nitrogen dioxide in humans.
Inhal Toxicol 24(12): 790-797.
Hystad, P, Demers, PA, Johnson, KC, Brook, J, van Donkelaar, A, Lamsal, L, Martin, R and
Brauer, M (2012). Spatiotemporal air pollution exposure assessment for a Canadian
population-based lung cancer case-control study. Environ Health 11: 22.
Hystad, P, Demers, PA, Johnson, KC, Carpiano, RM and Brauer, M (2013). Long-term
residential exposure to air pollution and lung cancer risk. Epidemiology 24(5): 762-772.
Ito, K, Ross, Z, Zhou, J, Nadas, A, Lippmann, M and Thurston, GD (2013). National Particle
Component Toxicity (NPACT) initiative: Study 3. Time-series analysis of mortality,
hospitalizations, and ambient PM2.5 and its components. Boston, MA, Health Effects
Institute: 95-125.
Jenkins, S. (2016). Email from Scott Jenkins to Robert Devlin. Re: Average PM2.5 concentration
in Devlin et al. (2003). Available in PM NAAQS Review Docket (EPA-HQ-OAR-2015-
0072-0021). March 2, 2016. Research Triangle Park, NC. Office of Air Quality Planning
and Standards.
Jerrett, M, Turner, MC, Beckerman, BS, Pope, CA, van Donkelaar, A, Martin, RV, Serre, M,
Crouse, D, Gapstur, SM, Krewski, D, Diver, WR, Coogan, PF, Thurston, GD and
Burnett, RT (2016). Comparing the health effects of ambient particulate matter estimated
using ground-based versus remote sensing exposure estimates. Environ Health Perspect
125(4): 552-559.
Kioumourtzoglou, MA, Schwartz, J, James, P, Dominici, F and Zanobetti, A (2016). PM2.5 and
mortality in 207 us cities: Modification by temperature and city characteristics.
Epidemiology 27(2): 221-227.
3-127

-------
Klemm, RJ and Mason, R (2003). HEI Special Report: Revised Analyses of Time-Series Studies
of Air Pollution and Health, Part II Replication of reanalysis of Harvard Six-City
mortality study. Boston, MA, Health Effects Institute.
Kloog, I, Coull, BA, Zanobetti, A, Koutrakis, P and Schwartz, JD (2012). Acute and chronic
effects of particles on hospital admissions in New-England. PLoS ONE 7(4): e34664.
Kloog, I, Ridgway, B, Koutrakis, P, Coull, BA and Schwartz, JD (2013). Long- and short-term
exposure to PM2.5 and mortality: Using novel exposure models. Epidemiology 24(4):
555-561.
Kloog, I, Nordio, F, Zanobetti, A, Coull, BA, Koutrakis, P and Schwartz, JD (2014). Short term
effects of particle exposure on hospital admissions in the Mid-Atlantic states: a
population estimate. PLoS ONE 9(2): e88578.
Krall, JR, Anderson, GB, Dominici, F, Bell, ML and Peng, RD (2013). Short-term exposure to
particulate matter constituents and mortality in a national study of U.S. urban
communities. Environ Health Perspect 121(10): 1148-1153.
Krewski, D, Jerrett, M, Burnett, RT, Ma, R, Hughes, E, Shi, Y, Turner, MC, Pope, CA, III,
Thurston, G, Calle, EE, Thun, MJ, Beckerman, B, Deluca, P, Finkelstein, N, Ito, K,
Moore, DK, Newbold, KB, Ramsay, T, Ross, Z, Shin, H and Tempalski, B (2009).
Extended follow-up and spatial analysis of the American Cancer Society study linking
particulate air pollution and mortality. Boston, MA, Health Effects Institute. 140: 5-114;
discussion 115-136.
Kusha, M, Masse, S, Farid, T, Urch, B, Silverman, FS, Brook, RD, Gold, DR, Mangat, I, Speck,
M, Nair, K, Poku, K, Meyer, C, Mittleman, MA, Wellenius, GA and Nanthakumar, K
(2012). Controlled exposure study of air pollution and t wave alternans in volunteers
without cardiovascular disease. Environ Health Perspect 120(8): 1157-1161.
Laden, F, Schwartz, J, Speizer, FE and Dockery, DW (2006). Reduction in fine particulate air
pollution and mortality: extended follow-up of the Harvard Six Cities study. Am J Respir
Crit Care Med 173(6): 667-672.
Lee, M, Koutrakis, P, Coull, B, Kloog, I and Schwartz, J (2015). Acute effect of fine particulate
matter on mortality in three Southeastern states from 2007-2011. Journal of Exposure
Science and Environmental Epidemiology 26(2): 173-179.
Lepeule, J, Laden, F, Dockery, D and Schwartz, J (2012). Chronic exposure to fine particles and
mortality: an extended follow-up of the Harvard Six Cities study from 1974 to 2009.
Environ Health Perspect 120(7): 965-970.
Lipfert, FW, Wyzga, RE, Baty, JD and Miller, JP (2006). Traffic density as a surrogate measure
of environmental exposures in studies of air pollution health effects: Long-term mortality
in a cohort of US veterans. Atmos Environ 40(1): 154-169.
3-128

-------
Lippmann, M, Chen, LC, Gordon, T, Ito, K and Thurston, GD (2013). National Particle
Component Toxicity (NPACT) Initiative: Integrated epidemiologic and toxicologic
studies of the health effects of particulate matter components: Investigators' Report.
Boston, MA, Health Effects Institute: 5-13.
Liu, L, Urch, B, Poon, R, Szyszkowicz, M, Speck, M, Gold, DR, Wheeler, AJ, Scott, JA, Brook,
JR, Thorne, PS and Silverman, FS (2015). Effects of ambient coarse, fine, and ultrafine
particles and their biological constituents on systemic biomarkers: A controlled human
exposure study. Environ Health Perspect 123(6): 534-540.
Lucking, AJ, Lundback, M, Barath, SL, Mills, NL, Sidhu, MK, Langrish, JP, Boon, NA,
Pourazar, J, Badimon, JJ, Gerlofs-Nijland, ME, Cassee, FR, Boman, C, Donaldson, K,
Sandstrom, T, Newby, DE and Blomberg, A (2011). Particle traps prevent adverse
vascular and prothrombotic effects of diesel engine exhaust inhalation in men.
Circulation 123(16): 1721-1728.
Malig, BJ, Green, S, Basu, R and Broadwin, R (2013). Coarse particles and respiratory
emergency department visits in California. Am J Epidemiol 178(1): 58-69.
Mauad, T, Rivero, DH, de Oliveira, RC, Lichtenfels, AJ, Guimaraes, ET, de Andre, PA,
Kasahara, DI, Bueno, HM and Saldiva, PH (2008). Chronic exposure to ambient levels of
urban particles affects mouse lung development. Am J Respir Crit Care Med 178(7): 721-
728.
McConnell, R, Islam, T, Shankardass, K, Jerrett, M, Lurmann, F, Gilliland, F, Gauderman, J,
Avol, E, Kiinzli, N, Yao, L, Peters, J and Berhane, K (2010). Childhood incident asthma
and traffic-related air pollution at home and school. Environ Health Perspect 118(7):
1021-1026.
Miller, KA, Siscovick, DS, Sheppard, L, Shepherd, K, Sullivan, JH, Anderson, GL and
Kaufman, JD (2007). Long-term exposure to air pollution and incidence of
cardiovascular events in women. New Engl J Med 356(5): 447-458.
NHLBI (2017). "NHLBI fact book, fiscal year 2012: Disease statistics." Retrieved August 23,
2017, from https://www.nhlbi.nih.gov/about/documents/factbook/2012/chapter4.
Nishimura, KK, Galanter, JM, Roth, LA, Oh, SS, Thakur, N, Nguyen, EA, Thyne, S, Farber, HJ,
Serebrisky, D, Kumar, R, Brigino-Buenaventura, E, Davis, A, Lenoir, MA, Meade, K,
Rodriguez-Cintron, W, Avila, PC, Borrell, LN, Bibbins-Domingo, K, Rodriguez-Santana,
JR, Sen, S, Lurmann, F, Balmes, JR and Burchard, EG (2013). Early-life air pollution and
asthma risk in minority children: The GALA II and SAGE II studies. Am J Respir Crit
Care Med 188(3): 309-318.
Ostro, B, Malig, B, Hasheminassab, S, Berger, K, Chang, E and Sioutas, C (2016). Associations
of source-specific fine particulate matter with emergency department visits in California.
Am J Epidemiol 184(6): 450-459.
3-129

-------
Paciorek, CJ, Yanosky, JD, Puett, RC, Laden, F and Suh, HH (2009). Practical large-scale
spatio-temporal modeling of particulate matter concentrations. Annals of Applied
Statistics 3(1): 370-397.
Peng, R, Bell, M, Geyh, A, McDermott, A, Zeger, S, Samet, J and Dominici, F (2009).
Emergency admissions for cardiovascular and respiratory diseases and the chemical
composition of fine particle air pollution. Environ Health Perspect 117(6): 957-963.
Pinault, L, Tjepkema, M, Crouse, DL, Weichenthal, S, van Donkelaar, A, Martin, RV, Brauer,
M, Chen, H and Burnett, RT (2016). Risk estimates of mortality attributed to low
concentrations of ambient fine particulate matter in the Canadian community health
survey cohort. Environ Health 15(1): 18.
Pope, CA, Turner, MC, Burnett, R, Jerrett, M, Gapstur, SM, Diver, WR, Krewski, D and Brook,
RD (2015). Relationships between fine particulate air pollution, cardiometabolic
disorders and cardiovascular mortality. Circul Res 116(1): 108-U258.
Pope, CA, III,, I, Burnett, RT, Thurston, GD, Thun, MJ, Calle, EE, Krewski, D and Godleski, JJ
(2004). Cardiovascular mortality and long-term exposure to particulate air pollution:
epidemiological evidence of general pathophysiological pathways of disease. Circulation
109(1): 71-77.
Pope, CA, III, Ezzati, M and Dockery, DW (2009). Fine-particulate air pollution and life
expectancy in the United States. New Engl J Med 360(4): 376-386.
Puett, RC, Hart, JE, Yanosky, JD, Paciorek, C, Schwartz, J, Suh, H, Speizer, FE and Laden, F
(2009). Chronic fine and coarse particulate exposure, mortality, and coronary heart
disease in the Nurses' Health Study. Environ Health Perspect 117(11): 1697-1701.
Puett, RC, Hart, JE, Suh, H, Mittleman, M and Laden, F (2011). Particulate matter exposures,
mortality and cardiovascular disease in the health professionals follow-up study. Environ
Health Perspect 119(8): 1130-1135.
Puett, RC, Hart, JE, Yanosky, JD, Spiegelman, D, Wang, M, Fisher, JA, Hong, B and Laden, F
(2014). Particulate matter air pollution exposure, distance to road, and incident lung
cancer in the Nurses' Health Study cohort. Environ Health Perspect 122(9): 926-932.
Pun, VC, Kazemiparkouhi, F, Manjourides, J and Suh, HH (2017). Long-term PM2.5 exposures
and respiratory, cancer and cardiovascular mortality in American older adults. Am J
Epidemiol 186(8): 961-969.
Raaschou-Nielsen, O, Andersen, ZJ, Beelen, R, Samoli, E, Stafoggia, M, Weinmayr, G,
Hoffmann, B, Fischer, P, Nieuwenhuijsen, MJ, Brunekreef, B, Xun, WW, Katsouyanni,
K, Dimakopoulou, K, Sommar, J, Forsberg, B, Modig, L, Oudin, A, Oftedal, B,
Schwarze, PE, Nafstad, P, De Faire, U, Pedersen, NL, Ostenson, CG, Fratiglioni, L,
Penell, J, Korek, M, Pershagen, G, Eriksen, KT, S0rensen, M, Tj0nneland, A, Ellermann,
T, Eeftens, M, Peeters, PH, Meliefste, K, Wang, M, Bueno-De-mesquita, B, Key, TJ, De
Hoogh, K, Concin, H, Nagel, G, Vilier, A, Grioni, S, Krogh, V, Tsai, MY, Ricceri, F,
3-130

-------
Sacerdote, C, Galassi, C, Migliore, E, Ranzi, A, Cesaroni, G, Badaloni, C, Forastiere, F,
Tamayo, I, Amiano, P, Dorronsoro, M, Trichopoulou, A, Bamia, C, Vineis, P and Hoek,
G (2013). Air pollution and lung cancer incidence in 17 European cohorts: Prospective
analyses from the European Study of Cohorts for Air Pollution Effects (ESCAPE). The
Lancet Oncology 14(9): 813-822.
Ramanathan, G, Yin, F, Speck, M, Tseng, CH, Brook, JR, Silverman, F, Urch, B, Brook, RD and
Araujo, JA (2016). Effects of urban fine particulate matter and ozone on HDL
functionality. Particle and Fibre Toxicology 13(1): 26.
Sacks, JD, Ito, K, Wilson, WE and Neas, LM (2012). Impact of covariate models on the
assessment of the air pollution-mortality association in a single- and multipollutant
context. Am J Epidemiol 176(7): 622-634.
Samet, J. (2009). Letter from Jonathan Samet, Chair, Clean Air Scientific Advisory Committee,
to Administrator Lisa Jackson. Re: CAS AC Particulate Matter Review of Integrated
Science Assessment for Particulate Matter (Second External Review Draft, July 2009).
November 24, 2009. EPA-CASAC-10-001. U.S. EPA HQ, Washington DC. Office of the
Administrator, Science Advisory Board. Available at:
http://nepis. epa.gov/Exe/ZyPURL. cgi?Dockev=P 1005PH9.txt.
Samet, J. (2010a). Letter from Jonathan Samet, Chair, Clean Air Scientific Advisory Committee,
to Administrator Lisa Jackson. Re: CAS AC Review of Policy Assessment for the Review
of the PM NAAQS - First External Review Draft (March 2010). May 17, 2010. EPA-
CASAC-10-011. U.S. EPA HQ, Washington DC. Office of the Administrator, Science
Advisory Board. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=9101XQXQ.txt.
Samet, J. (2010b). Letter from Jonathan Samet, Chair, Clean Air Scientific Advisory Committee,
to Administrator Lisa Jackson. Re: CASAC Review of Quantitative Health Risk
Assessment for Particulate Matter - Second External Review Draft (February 2010).
April 15, 2010. EPA-CASAC-10-008. U.S. EPA HQ, Washington DC. Office of the
Administrator, Science Advisory Board. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 1007CVB.txt.
Samet, J. (2010c). Letter from Jonathan Samet, Chair, Clean Air Scientific Advisory Committee,
to Administrator Lisa Jackson. Re: CASAC Review of Policy Assessment for the Review
of the PM NAAQS - Second External Review Draft (June 2010). September 10, 2010.
EPA-CASAC-10-015. U.S. EPA HQ, Washington DC. Office of the Administrator,
Science Advisory Board. Available at:
https://vosemite.epa. gov/sab/sabproduct.nsf/CCF9F4C0500C500F8525779D0073C593/$
File/EPA-CASAC-10-015-unsigned.pdf.
Schwartz, J, Austin, E, Bind, MA, Zanobetti, A and Koutrakis, P (2015). Estimating causal
associations of fine particles with daily deaths in Boston. Am J Epidemiol 182(7): 644-
650.
3-131

-------
Schwartz, J, Bind, MA and Koutrakis, P (2017). Estimating causal effects of local air pollution
on daily deaths: Effect of low levels. Environ Health Perspect 125(1): 23-29.
Shi, L, Zanobetti, A, Kloog, I, Coull, BA, Koutrakis, P, Melly, SJ and Schwartz, JD (2016).
Low-concentration PM2.5 and mortality: estimating acute and chronic effects in a
population-based study. Environ Health Perspect 124(1): 46-52.
Sivagangabalan, G, Spears, D, Masse, S, Urch, B, Brook, RD, Silverman, F, Gold, DR, Lukic,
KZ, Speck, M, Kusha, M, Farid, T, Poku, K, Shi, E, Floras, J and Nanthakumar, K
(2011). The effect of air pollution on spatial dispersion of myocardial repolarization in
healthy human volunteers. J Am Coll Cardiol 57(2): 198-206.
Stieb, DM, Szyszkowicz, M, Rowe, BH and Leech, JA (2009). Air pollution and emergency
department visits for cardiac and respiratory conditions: A multi-city time-series analysis.
Environ Health 8(25): 25.
Szyszkowicz, M (2009). Air pollution and ED visits for chest pain. American Journal of
Emergency Medicine 27(2): 165-168.
Tetreault, LF, Doucet, M, Gamache, P, Fournier, M, Brand, A, Kosatsky, T and Smargiassi, A
(2016). Childhood exposure to ambient air pollutants and the onset of asthma: an
administrative cohort study in Quebec. Environ Health Perspect 124(8): 1276-1282.
Thurston, GD, Ito, K, Lall, R, Burnett, RT, Turner, MC, Krewski, D, Shi, Y, Jerrett, M, Gapstur,
SM, Diver, WR and Pope, CA, III (2013). NPACT Study 4. Mortality and long-term
exposure to PM2.5 and its components in the American Cancer Society's Cancer
Prevention Study II Cohort. M. Lippmann, L. C. Chen, T. Gordon, K. Ito and G. D.
Thurston. Boston, MA, Health Effects Institute: 127-166.
Thurston, GD, Ahn, J, Cromar, KR, Shao, Y, Reynolds, HR, Jerrett, M, Lim, CC, Shanley, R,
Park, Y and Hayes, RB (2016). Ambient particulate matter air pollution exposure and
mortality in the NIH-AARP Diet and Health Cohort. Environ Health Perspect 124(4):
484-490.
Tomczak, A, Miller, AB, Weichenthal, SA, To, T, Wall, C, van Donkelaar, A, Martin, RV,
Crouse, DL and Villeneuve, PJ (2016). Long-term exposure to fine particulate matter air
pollution and the risk of lung cancer among participants of the Canadian National Breast
Screening Study. Int J Cancer 139(9): 1958-1966.
Tong, H, Rappold, AG, Caughey, M, Hinderliter, AL, Bassett, M, Montilla, T, Case, MW,
Berntsen, J, Bromberg, PA, Cascio, WE, Diaz-Sanchez, D, Devlin, RB and Samet, JM
(2015). Dietary supplementation with olive oil or fish oil and vascular effects of
concentrated ambient particulate matter exposure in human volunteers. Environ Health
Perspect 123(11): 1173-1179.
Turner, MC, Jerrett, M, Pope, A, III, Krewski, D, Gapstur, SM, Diver, WR, Beckerman, BS,
Marshall, JD, Su, J, Crouse, DL and Burnett, RT (2016). Long-term ozone exposure and
mortality in a large prospective study. Am J Respir Crit Care Med 193(10): 1134-1142.
3-132

-------
U.S. EPA. (2001). Risk Assessment Guidance for Superfund: Volume III - Part A, Process for
Conducting Probabilistic Risk Assessment. Washington, D.C. Office of Emergency and
Remedial Response. U.S. EPA. EPA 540-R-02-002. Available at:
https://www.epa.gov/sites/production/files/2015-09/documents/rags3adt complete.pdf.
U.S. EPA. (2004). An Examination of EPA Risk Assessment Principles and Practices.
Washington, D.C. Office of Research and Development. U.S. EPA. EPA/100/B-04/001.
Available at: https://semspub.epa.gov/work/10/500006305.pdf.
U.S. EPA. (2009). Integrated Science Assessment for Particulate Matter (Final Report). Research
Triangle Park, NC. Office of Research and Development, National Center for
Environmental Assessment. U.S. EPA. EPA-600/R-08-139F. December 2009. Available
at: https://cfpub.epa.gov/ncea/risk/recordisplav.cfm?deid=216546.
U.S. EPA. (2010). Quantitative Health Risk Assessment for Particulate Matter (Final Report).
Research Triangle Park, NC. Office of Air Quality Planning and Standards, Health and
Environmental Impacts Division. U.S. EPA. EPA-452/R-10-005. June 2010. Available
at: https://nepis.epa.gov/Exe/ZvPURL.cgi?Dockev=P1007RFC.txt.
U.S. EPA. (2011). Policy Assessment for the Review of the Particulate Matter National Ambient
Air Quality Standards Research Triangle Park, NC. Office of Air Quality Planning and
Standards, Health and Environmental Impacts Division. U.S. EPA. EPA-452/R-11-003
April 2011. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100AUMY.txt.
U.S. EPA. (2012). Responses to Significant Comments on the 2012 Proposed Rule on the
National Ambient Air Quality Standards for Particulate Matter (June 29, 2012; 77 FR
38890). Research Triangle Park, NC. U.S. EPA. Docket ID No. EPA-HQ-OAR-2007-
0492. Available at: https://www3.epa.gov/ttn/naaqs/standards/pm/data/20121214rtc.pdf.
U.S. EPA. (2015). Preamble to the integrated science assessments. Research Triangle Park, NC.
U.S. Environmental Protection Agency, Office of Research and Development, National
Center for Environmental Assessment, RTP Division. U.S. EPA. EPA/600/R-15/067.
November 2015. Available at:
https://cfpub.epa. gov/ncea/isa/recordisplav.cfm?deid=310244.
U.S. EPA. (2016). Integrated review plan for the national ambient air quality standards for
particulate matter. Research Triangle Park, NC. Office of Air Quality Planning and
Standards. U.S. EPA. EPA-452/R-16-005. December 2016. Available at:
https://www3.epa.gov/ttn/naaqs/standards/pm/data/201612-final-integrated-review-
plan.pdf.
U.S. EPA. (2019). Integrated Science Assessment (ISA) for Particulate Matter (Final Report).
Washington, DC. U.S. Environmental Protection Agency, Office of Research and
Development, National Center for Environmental Assessment. U.S. EPA. EPA/600/R-
19/188. December 2019. Available at: https://www.epa.gov/naaqs/particulate-matter-pm-
standards-integrated-science-assessments-current-review.
3-133

-------
Urch, B, Speck, M, Corey, P, Wasserstein, D, Manno, M, Lukic, KZ, Brook, JR, Liu, L, Coull,
B, Schwartz, J, Gold, DR and Silverman, F (2010). Concentrated ambient fine particles
and not ozone induce a systemic interleukin-6 response in humans. Inhal Toxicol 22(3):
210-218.
Urman, R, McConnell, R, Islam, T, Avol, EL, Lurmann, FW, Vora, H, Linn, WS, Rappaport,
EB, Gilliland, FD and Gauderman, WJ (2014). Associations of children's lung function
with ambient air pollution: Joint effects of regional and near-roadway pollutants. Thorax
69(6): 540-547.
van Donkelaar, A, Martin, RV, Brauer, M, Kahn, R, Levy, R, Verduzco, C and Villeneuve, PJ
(2010). Global estimates of ambient fine particulate matter concentrations from satellite-
based aerosol optical depth: development and application. Environ Health Perspect
118(6): 847-855.
van Donkelaar, A, Martin, RV, Spurr, RJD, Drury, E, Remer, LA, Levy, RC and Wang, J (2013).
Optimal estimation for global ground-level fine particulate matter concentrations. Journal
of Geophysical Research: Atmospheres 118(11): 5621-5636.
van Donkelaar, A, Martin, RV, Brauer, M and Boys, BL (2015a). Use of satellite observations
for long-term exposure assessment of global concentrations of fine particulate matter.
Environ Health Perspect 123(2): 135-143.
van Donkelaar, A, Martin, RV, Spurr, RJ and Burnett, RT (2015b). High-resolution satellite-
derived PM2.5 from optimal estimation and geographically weighted regression over
North America. Environ Sci Technol 49(17): 10482-10491.
Vieira, JL, Guimaraes, GV, de Andre, PA, Cruz, FD, Nascimento Saldiva, PH and Bocchi, EA
(2016a). Respiratory filter reduces the cardiovascular effects associated with diesel
exhaust exposure a randomized, prospective, double-blind, controlled study of heart
failure: the FILTER-HF trial. JACC: Heart Failure 4(1): 55-64.
Vieira, JL, Guimaraes, GV, de Andre, PA, Nascimento Saldiva, PH and Bocchi, EA (2016b).
Effects of reducing exposure to air pollution on submaximal cardiopulmonary test in
patients with heart failure: Analysis of the randomized, double-blind and controlled
FILTER-HF trial. International Journal of Cardiology 215: 92-97.
Villeneuve, PJ, Weichenthal, SA, Crouse, D, Miller, AB, To, T, Martin, RV, van Donkelaar, A,
Wall, C and Burnett, RT (2015). Long-term exposure to fine particulate matter air
pollution and mortality among Canadian women. Epidemiology 26(4): 536-545.
Wang, Y, Kloog, I, Coull, BA, Kosheleva, A, Zanobetti, A and Schwartz, JD (2016). Estimating
causal effects of long-term PM2.5 exposure on mortality in New Jersey. Environ Health
Perspect 124(8): 1182-1188.
Wang, Y, Shi, L, Lee, M, Liu, P, Di, Q, Zanobetti, A and Schwartz, JD (2017). Long-term
exposure to PM2.5 and mortality among older adults in the Southeastern US.
Epidemiology 28(2): 207-214.
3-134

-------
Weichenthal, S, Hatzopoulou, M and Goldberg, MS (2014). Exposure to traffic-related air
pollution during physical activity and acute changes in blood pressure, autonomic and
micro-vascular function in women: A cross-over study. Particle and Fibre Toxicology
11(1): 70.
Weichenthal, S, Crouse, DL, Pinault, L, Godri-Pollitt, K, Lavigne, E, Evans, G, van Donkelaar,
A, Martin, RV and Burnett, RT (2016a). Oxidative burden of fine particulate air pollution
and risk of cause-specific mortality in the Canadian Census Health and Environment
Cohort (CanCHEC). Environ Res 146: 92-99.
Weichenthal, S, Lavigne, E, Evans, G, Pollitt, K and Burnett, RT (2016b). Ambient PM2.5 and
risk of emergency room visits for myocardial infarction: Impact of regional PM2.5
oxidative potential: A case-crossover study. Environ Health 15: 46.
Weichenthal, S, Lavigne, E, Evans, GJ, Godri Pollitt, KJ and Burnett, RT (2016c). PM2.5 and
emergency room visits for respiratory illness: effect modification by oxidative potential.
Am J Respir Crit Care Med 194(5): 577-586.
WHO. (2008). Uncertainty and Data Quality in Exposure Assessment. Geneva. The
Internaltional Programme on Chemical Safety. WHO.
https://www.who.int/ipcs/publications/methods/harmonization/exposure assessment.pdf.
Willis, A, Jerrett, M, Burnett, RT and Krewski, D (2003). The association between sulfate air
pollution and mortality at the county scale: an exploration of the impact of scale on a
long-term exposure study. J Toxicol Environ Health, A: Curr Iss 66(16-19): 1605-1624.
Yap, P, Gilbreath, S, Garcia, C, Jareen, N and Goodrich, B (2013). The influence of
socioeconomic markers on the association between fine particulate matter and hospital
admissions for respiratory conditions among children. Am J Public Health 103(4): 695-
702.
Yorifuji, T, Kashima, S and Doi, H (2016). Fine-particulate air pollution from diesel emission
control and mortality rates in Tokyo: a quasi-experimental study. Epidemiology 27(6):
769-778.
Zanobetti, A, Franklin, M, Koutrakis, P and Schwartz, J (2009). Fine particulate air pollution and
its components in association with cause-specific emergency admissions. Environ Health
8: 58.
Zanobetti, A and Schwartz, J (2009). The effect of fine and coarse particulate air pollution on
mortality: A national analysis. Environ Health Perspect 117(6): 1-40.
Zanobetti, A, Dominici, F, Wang, Y and Schwartz, JD (2014). A national case-crossover analysis
of the short-term effect of PM2.5 on hospitalizations and mortality in subjects with
diabetes and neurological disorders. Environ Health 13(1): 38.
3-135

-------
Zeger, S, Dominici, F, McDermott, A and Samet, J (2008). Mortality in the Medicare population
and chronic exposure to fine particulate air pollution in urban centers (2000-2005).
Environ Health Perspect 116(12): 1614-1619.
3-136

-------
4 REVIEW OF THE PRIMARY STANDARD FOR PMio
This chapter presents key policy-relevant considerations and conclusions regarding the
public health protection provided by the current primary PMio standard. These considerations
and conclusions are framed by a series of policy-relevant questions, including the following
overarching policy-relevant question:
• Does the currently available scientific evidence support or call into question the
adequacy of the protection afforded by the current 24-hour primary PMio standard
against health effects associated with exposures to PM10-2.5?
The answer to this question is intended to inform decisions by the Administrator on whether, and
if so, how to revise the primary standard for PMio.
Section 4.1 summarizes the EPA's approach to reviewing the primary PMio standard in
the last review and our general approach to considering the updated scientific evidence in the
current review. Section 4.2 presents our consideration of the available evidence as assessed in the
ISA. Section 4.3 summarizes CASAC advice and public comments. Drawing from that
consideration of the evidence, section 4.4 summarizes our conclusions regarding the adequacy of
the current primary PMio standard. Section 4.5 discusses areas for future research and data
collection to improve our understanding of potential PMio-25-related health effects in future
reviews.
4.1 APPROACH
4.1.1 Approach Used in the Last review
The last review of the PMNAAQS was completed in 2012 (78 FR 3086, January 15,
2013). In that review the EPA retained the existing 24-hour primary PMio standard, with its level
of 150 |j,g/m3 and its one-expected-exceedance form on average over three years, to continue to
provide public health protection against exposures to PM10-2.5. In support of this decision, the
Administrator emphasized her consideration of three issues: the extent to which it was
appropriate to maintain a standard that provides some measure of protection against all PM10-2.5
(regardless of composition or source or origin), the extent to which a standard with a PMio
indicator can provide protection against exposures to PM10-2.5, and the degree of public health
protection provided by the existing PMio standard. Her consideration of each of these issues is
summarized below.
First, the Administrator judged that the evidence provided "ample support for a standard
that protects against exposures to all thoracic coarse particles, regardless of their location or
source of origin" (78 FR 3176, January 15, 2013). In support of this, she noted that
4-1

-------
epidemiologic studies had reported positive associations between PM10-2.5 and mortality or
morbidity in a large number of cities across North America, Europe, and Asia, encompassing a
variety of environments where PM10-2.5 sources and composition are expected to vary widely.
Though most of the available studies examined associations in urban areas, the Administrator
noted that some studies had also linked mortality and morbidity with relatively high ambient
concentrations of particles of non-urban crustal origin. In light of this body of available evidence,
and consistent with the CASAC's advice, the Administrator concluded that it was appropriate to
maintain a standard that provides some measure of protection against exposures to all thoracic
coarse particles, regardless of their location, source of origin, or composition (78 FR 3176,
January 15, 2013).
In next reaching the conclusion that it was appropriate to retain a PM10 indicator for a
standard meant to protect against exposures to ambient PM10-2.5, the Administrator noted that
PM10 mass includes both coarse PM (PM10-2.5) and fine PM (PM2.5). As a result, the
concentration of PM10-2.5 allowed by a PM10 standard set at a single level declines as the
concentration of PM2.5 increases. Because PM2.5 concentrations tend to be higher in urban areas
than rural areas (e.g., Chan et al., 2018), the Administrator observed that a PM10 standard would
generally allow lower PM10-2.5 concentrations in urban areas than in rural areas. She judged it
appropriate to maintain such a standard given that much of the evidence for PM10-2.5 toxicity,
particularly at relatively low particle concentrations, came from study locations where thoracic
coarse particles were of urban origin, and given the possibility that PM10-2.5 contaminants in
urban areas could increase particle toxicity. Thus, in the last review the Administrator concluded
that it remained appropriate to maintain a standard that allows lower ambient concentrations of
PM10-2.5 in urban areas, where the evidence was strongest that exposure to thoracic coarse
particles was associated with morbidity and mortality, and higher concentrations in non-urban
areas, where the public health concerns were less certain. The Administrator concluded that the
varying concentrations of coarse particles that would be permitted in urban versus non-urban
areas under the 24-hour PM10 standard, based on the varying levels of PM2.5 present,
appropriately reflected the differences in the strength of evidence regarding coarse particle health
effects.
Finally, in specifically evaluating the degree of public health protection provided by the
primary PM10 standard, with its level of 150 |j,g/m3 and its one-expected-exceedance form on
average over three years, the Administrator recognized that the available health evidence and air
quality information was much more limited for PM10-2.5 than for PM2.5. In particular, the
strongest evidence for health effects attributable to PM10-2.5 exposure was for cardiovascular
effects, respiratory effects, and/or premature mortality following short-term exposures. For each
of these categories of effects, the 2009 ISA concluded that the evidence was "suggestive of a
4-2

-------
causal relationship" (U.S. EPA, 2009, section 2.3.3). These determinations contrasted with those
for PM2.5, as described in Chapter 3 above, which were determined in the ISA to be either
"causal" or "likely to be causal" for mortality, cardiovascular effects, and respiratory effects
(U.S. EPA, 2009, Tables 2-1 and 2-2).
The Administrator judged that the important uncertainties and limitations associated with
the PM10-2.5 evidence and information raised questions as to whether additional public health
improvements would be achieved by revising the existing PM10 standard. She specifically noted
several uncertainties, including the following:
(1)	The number of epidemiologic studies that have employed copollutant models to address
the potential for confounding, particularly by PM2.5, was limited. Therefore, the extent to
which PM10-2.5 itself, rather than one or more copollutants, contributes to reported health
effects remained uncertain.
(2)	Only a limited number of experimental studies provided support for the associations
reported in epidemiologic studies, resulting in further uncertainty regarding the
plausibility of the associations between PM10-2.5 and mortality and morbidity reported in
epidemiologic studies.
(3)	Limitations in PM10-2.5 monitoring data (i.e., limited data available from FRM/FEM
sampling methods) and the different approaches used to estimate PM10-2.5 concentrations
across epidemiologic studies resulted in uncertainty in the ambient PM10-2.5
concentrations at which the reported effects occur, increasing uncertainty in estimates of
the extent to which changes in ambient PM10-2.5 concentrations would likely impact
public health.
(4)	While PM10-2.5 effect estimates reported for mortality and morbidity were generally
positive, most were not statistically significant, even in single-pollutant models. This
included effect estimates reported in some study locations with PM10 concentrations
above those allowed by the current 24-hour PM10 standard.
(5)	The composition of PM10-2.5, and the effects associated with various components, were
also key uncertainties in the available evidence. Without more information on the
chemical speciation of PM10-2.5, the apparent variability in associations across locations
was difficult to characterize.
In considering these uncertainties, the Administrator particularly emphasized the
considerable degree of uncertainty in the extent to which health effects reported in epidemiologic
studies are due to PM10-2.5 itself, as opposed to one or more co-occurring pollutants. This
uncertainty reflected the relatively small number of PM10-2.5 studies that had evaluated
copollutant models, particularly copollutant models that included PM2.5, and the very limited
body of controlled human exposure evidence supporting the plausibility of PMio-25-attributable
adverse effects at ambient concentrations.
4-3

-------
When considering the evidence as a whole, the Administrator concluded that the degree
of public health protection provided by the current PMio standard against exposures to PM10-2.5
should be maintained (i.e., neither increased nor decreased). The Administrator's judgment that
protection did not need to be increased was supported by her consideration of uncertainties in the
overall body of evidence. Her judgment that the degree of public health protection provided by
the current standard is not greater than warranted was supported by the observation that positive
and statistically significant associations with mortality were reported in some single-city U.S.
study locations likely to have violated the current PM10 standard. Thus, the Administrator
concluded that the existing 24-hour PM10 standard, with its one-expected exceedance form on
average over three years and a level of 150 ng/m3, was requisite to protect public health with an
adequate margin of safety against effects that have been associated with PM10-2.5. In light of this
conclusion, the EPA retained the existing PM10 standard.
4.1.2 Approach in the Current Review
As discussed above for PM2.5 (section 3.2.1), in this PA we place the greatest emphasis
on effects for which the evidence has been determined to demonstrate a "causal" or a "likely to
be causal" relationship with PM exposures (U.S. EPA, 2019). This approach focuses policy
considerations and conclusions on health outcomes for which the evidence is strongest. Unlike
for PM2.5, the ISA does not identify any PMio-25-related health outcomes for which the evidence
supports either a "causal" or a "likely to be causal" relationship. Thus, for PM10-2.5 this PA
considers the evidence determined to be "suggestive of, but not sufficient to infer, a causal
relationship," recognizing the greater uncertainty in such evidence.
The preamble to the ISA states that "suggestive" evidence is "limited, and chance,
confounding, and other biases cannot be ruled out" (U.S. EPA, 2015, Table II). In light of the
additional uncertainty in the evidence for PMio-2 5-related health outcomes, compared to the
evidence supporting "causal" or "likely to be causal" relationships for PM2.5, our approach to
evaluating the primary PM10 standard in this review is more limited than our approach to
evaluating the primary PM2.5 standards (discussed in Chapter 3). Specifically, our approach for
PM10 does not include evaluations of air quality distributions in locations of individual
epidemiologic studies, comparisons of experimental exposures with ambient air quality, or the
quantitative assessment of PM10-2.5 health risks. The substantial uncertainty in such analyses, if
they were to be conducted based on the currently available PM10-2.5 health studies, would limit
their utility for informing conclusions on the primary PM10 standard. Therefore, as discussed
further below, we focus our evaluation of the primary PM10 standard on the overall body of
evidence for PMio-25-related health effects. This includes consideration of the degree to which
uncertainties in the evidence from the last review have been reduced and the degree to which
4-4

-------
new uncertainties have been identified. In adopting this approach, we recognize that the
Administrator's decisions as to whether to retain or revise the primary PMio standard will largely
be public health policy judgments that will draw upon the scientific evidence for PMio-25-related
health effects and judgments about how to consider the uncertainties and limitations inherent in
that evidence.
4.2 EVIDENCE-BASED CONSIDERATIONS
This section draws from the EPA's synthesis and assessment of the scientific evidence
presented in the ISA (U.S. EPA, 2019) to consider the following policy-relevant questions:
• To what extent does the currently available scientific evidence strengthen, or otherwise
alter, our conclusions from the last review regarding health effects attributable to long-
or short-term PM10-2.5 exposures? Have previously identified uncertainties been
reduced? What important uncertainties remain and have new uncertainties been
identified?
Answers to these questions will inform our answer to the overarching question on the adequacy
of the current primary PM10 standard, posed at the beginning of this chapter. In section 4.2.1
below, we consider the nature of the effects attributable to long-term and short-term PM10-2.5
exposures.
4.2.1 Nature of Effects
As noted above, for the heath outcome categories and exposure duration combinations
evaluated, the ISA concludes that the evidence supports causality determinations for PM10-2.5 no
stronger than "suggestive of, but not sufficient to infer, a causal relationship." These outcomes,
along with their corresponding causality determinations from the 2009 ISA, are highlighted
below in Table 4-1 (adapted from U.S. EPA, 2019, Table 1-4).
4-5

-------
Table 4-1. Key Causality Determinations for PM10-2.5 Exposures
Health Outcome
Exposure
Duration
2009 PM ISA
2019 PM ISA
Mortality
Long-term
Inadequate

Short-term
Suggestive of, but not sufficient to infer

Cardiovascular
Long-term
Inadequate

effects
Short-term
Suggestive of, but not sufficient to infer

Respiratory effects
Short-term
Suggestive of, but not sufficient to infer
Suggestive of, but not
sufficient to infer
Cancer
Long-term
Inadequate

Nervous System
effects
Long-term
—

Metabolic effects
Long-term
—

While the evidence for some of the health outcomes listed in Table 4-1 has strengthened
since the last review, the ISA concludes that overall "the uncertainties in the evidence identified
in the 2009 PM ISA have, to date, still not been addressed" (U.S. EPA, 2019, section 1.4.2, p. 1-
41). For example, epidemiologic studies available in the last review relied on various methods to
estimate PM10-2.5 exposures, and these methods had not been systematically compared to
evaluate spatial and temporal correlations in exposure estimates. Methods included (1)
calculating the difference between PM10 and PM2.5 concentrations at co-located monitors, (2)
calculating the difference between county-wide averages of monitored PM10 and PM2.5 based on
monitors that are not necessarily co-located, and (3) direct measurement of PM10-2.5 using a
dichotomous sampler (U.S. EPA, 2019, section 1.4.2). In the current review, more recent
epidemiologic studies continue to use these approaches to estimate PM10-2.5 concentrations.
Additionally, some recent studies estimate long-term PM10-2.5 exposures as the difference
between PM10 and PM2.5 concentrations based on information from spatiotemporal or land use
regression (LUR) models, in addition to monitors. As in the last review, the various methods
used to estimate PM10-2.5 concentrations have not been systematically evaluated (U.S. EPA,
2019, section 3.3.1.1), contributing to uncertainty regarding the spatial and temporal correlations
in PM10-2.5 concentrations across methods and in the PM10-2.5 exposure estimates used in
epidemiologic studies (U.S. EPA, 2019, section 2.5.1.2.3). Given the greater spatial and temporal
variability of PM10-2.5 and fewer PM10-2.5 monitoring sites, compared to PM2.5, this uncertainty is
particularly important for the coarse size fraction.
4-6

-------
Beyond uncertainty associated with PM10-2.5 exposure estimates in epidemiologic studies,
the limited information on the potential for confounding by copollutants and the limited support
available for the biological plausibility of serious effects following PM10-2.5 exposures also
continue to contribute broadly to uncertainty in the PM10-2.5 health evidence. Uncertainty related
to potential confounding stems from the relatively small number of epidemiologic studies that
have evaluated PM10-2.5 health effect associations in copollutants models with both gaseous
pollutants and other PM size fractions. Uncertainty related to the biological plausibility of
serious effects caused by PM10-2.5 exposures results from the small number of controlled human
exposure and animal toxicology1 studies that have evaluated the health effects of experimental
PM10-2.5 inhalation exposures. The evidence supporting the ISA's "suggestive" causality
determinations for PM10-2.5, including uncertainties in this evidence, is summarized in sections
4.2.1.1 to 4.2.1.6 below.
4.2.1.1 Mortality
Long-term exposures
Due to the dearth of studies examining the association between long-term PM10-2.5
exposure and mortality, the 2009 PM ISA concluded that the evidence was "inadequate to
determine if a causal relationship exists" (U.S. EPA, 2009). Since the completion of the 2009
ISA, some recent cohort studies conducted in the U.S. and Europe report positive associations
between long-term PM10-2.5 exposure and total (nonaccidental) mortality, though results are
inconsistent across studies (U.S. EPA, 2019, Table 11-11). The examination of copollutant
models in these studies remains limited and, when included, PM10-2.5 effect estimates are often
attenuated after adjusting for PM2.5 (U.S. EPA, 2019, Table 11-11). Across studies, PM10-2.5
exposure concentrations are estimated using a variety of approaches, including direct
measurements from dichotomous samplers, calculating the difference between PM10 and PM2.5
concentrations measured at collocated monitors, and calculating difference of area-wide
concentrations of PM10 and PM2.5. As discussed above, temporal and spatial correlations between
these approaches have not been evaluated, contributing to uncertainty regarding the potential for
exposure measurement error (U.S. EPA, 2019, section 3.3.1.1 and Table 11-11). The ISA
concludes that this uncertainty "reduces the confidence in the associations observed across
studies" (U.S. EPA, 2019, p. 11-125). The ISA additionally concludes that the evidence for
long-term PM10-2.5 exposures and cardiovascular effects, respiratory morbidity, and metabolic
disease provide limited biological plausibility for PMio-25-related mortality (U.S. EPA, 2019,
sections 11.4.1 and 11.4). Taken together, the ISA concludes that, "this body of evidence is
1 Compared to humans, smaller fractions of inhaled PMi 0-2.5 penetrate into the thoracic regions of rats and mice
(U.S. EPA, 2018, section 4.1.6), contributing to the relatively limited evaluation of PM10-2.5 exposures in animal
studies.
4-7

-------
suggestive, but not sufficient to infer, that a causal relationship exists between long-term PM10-2.5
exposure and total mortality" (U.S. EPA, 2019, p. 11-125).
Short-term exposures
The 2009 ISA concluded that the evidence is "suggestive of a causal relationship between
short-term exposure to PM10-2.5 and mortality" (U.S. EPA, 2009). Since the completion of the
2009 ISA, multicity epidemiologic studies conducted primarily in Europe and Asia continue to
provide consistent evidence of positive associations between short-term PM10-2.5 exposure and
total (nonaccidental) mortality (U.S. EPA, 2019, Table 11-9). Although these studies contribute
to increasing confidence in the PMio-25-mortality relationship, the use of a variety of approaches
to estimate PM10-2.5 exposures continues to contribute uncertainty to the associations observed. In
addition, the ISA notes that an analysis by Adar et al. (2014) indicates "possible evidence of
publication bias, which was not observed forPlVb.s" (U.S. EPA, 2019, section 11.3.2, p. 11-106).
Recent studies expand the assessment of potential copollutant confounding of the
PMio-2.5-mortality relationship and provide evidence that PM10-2.5 associations generally remain
positive in copollutant models, though associations are attenuated in some instances (U.S. EPA,
2019, section 11.3.4.1, Figure 11-28, Table 11-10). The ISA concludes that, overall, the
assessment of potential copollutant confounding is limited due to the lack of information on the
correlation between PM10-2.5 and gaseous pollutants and the small number of locations in which
copollutant analyses have been conducted. Associations with cause-specific mortality provide
some support for associations with total (nonaccidental) mortality, though associations with
cause-specific mortality, particularly respiratory mortality, are more uncertain (i.e., wider
confidence intervals) and less consistent (U.S. EPA, 2019, section 11.3.7). The ISA concludes
that the evidence for PMio-25-related cardiovascular and respiratory effects provides only limited
support for the biological plausibility of a relationship between short-term PM10-2.5 exposure and
cardiovascular mortality (U.S. EPA, 2019, Section 11.3.7). Based on the overall evidence, the
ISA concludes that, "this body of evidence is suggestive, but not sufficient to infer, that a causal
relationship exists between short-term PM10-2.5 exposure and total mortality" (U.S. EPA, 2019, p.
11-120).
4.2.1.2 Cardiovascular Effects
Long-term exposures
In the 2009 PM ISA, the evidence describing the relationship between long-term
exposure to PM10-2.5 and cardiovascular effects was characterized as "inadequate to infer the
presence or absence of a causal relationship." The limited number of epidemiologic studies
reported contradictory results and experimental evidence demonstrating an effect of PM10-2.5 on
the cardiovascular system was lacking (U.S. EPA, 2019, section 6.4).
4-8

-------
The evidence relating long-term PM10-2.5 exposures to cardiovascular mortality remains
limited, with no consistent pattern of associations across studies and, as discussed above,
uncertainty stemming from the use of various approaches to estimate PM10-2.5 concentrations
(U.S. EPA, 2019, Table 6-70). The evidence for associations with cardiovascular morbidity has
grown and, while results across studies are not entirely consistent, some epidemiologic studies
report positive associations with ischemic heart disease (IHD) and myocardial infarction (MI)
(U.S. EPA, 2019, Figure 6-34); stroke (U.S. EPA, 2019, Figure 6-35); atherosclerosis; venous
thromboembolism (VTE); and blood pressure and hypertension (U.S. EPA, 2019, Section 6.4.6).
PM10-2.5 cardiovascular mortality effect estimates are often attenuated, but remain positive, in
copollutants models that adjust for PM2.5. For morbidity outcomes, associations are inconsistent
in copollutant models that adjust for PM2.5, NO2, and chronic noise pollution (U.S. EPA, 2019, p.
6-276). The lack of toxicological evidence for long-term PM10-2.5 exposures represents a
substantial data gap (U.S. EPA, 2019, section 6.4.10), resulting in the ISA conclusion that
"evidence from experimental animal studies is of insufficient quantity to establish biological
plausibility" (U.S. EPA, 2019, p. 6-277). Based largely on the observation of positive
associations in some high-quality epidemiologic studies, the ISA concludes that "evidence is
suggestive of, but not sufficient to infer, a causal relationship between long-term PM10-2.5
exposure and cardiovascular effects" (U.S. EPA, 2019, p. 6-277).
Short-term exposures
The 2009 ISA found that the available evidence for short-term PM10-2.5 exposure and
cardiovascular effects was "suggestive of a causal relationship." This conclusion was based on
several epidemiologic studies reporting associations between short-term PM10-2.5 exposure and
cardiovascular effects, including IHD hospitalizations, supraventricular ectopy, and changes in
heart rate variability (HRV). In addition, dust storm events resulting in high concentrations of
crustal material were linked to increases in total cardiovascular disease emergency department
visits and hospital admissions. However, the 2009 ISA noted the potential for exposure
measurement error and copollutant confounding in these epidemiologic studies. In addition, there
was only limited evidence of cardiovascular effects from a small number of experimental studies
(e.g. animal toxicological studies and controlled human exposure studies) that examined short-
term PM10-2.5 exposures (U.S. EPA, 2009, section 6.2.12.2). In the last review, key uncertainties
included the potential for exposure measurement error, copollutant confounding, and limited
evidence of biological plausibility for cardiovascular effects following inhalation exposure (U.S.
EPA, 2019, section 6.3.13).
The evidence for short-term PM10-2.5 exposure and cardiovascular outcomes has expanded
since the last review, though important uncertainties remain. The ISA notes that there are a small
number of epidemiologic studies reporting positive associations between short-term exposure to
4-9

-------
PMio-2.5 and cardiovascular-related morbidity outcomes. However, there is limited evidence to
suggest that these associations are biologically plausible, or independent of copollutant
confounding. The ISA also concludes that it remains unclear how the approaches used to
estimate PM10-2.5 concentrations in epidemiologic studies may impact exposure measurement
error. Taken together, the ISA concludes that "the evidence is suggestive of, but not sufficient to
infer, a causal relationship between short-term PM10-2.5 exposures and cardiovascular effects"
(U.S. EPA, 2019, p.6-254).
4.2.1.3	Respiratory Effects
Short-term exposures
Based on a small number of epidemiologic studies observing associations with some
respiratory effects and limited evidence from experimental studies to support biological
plausibility, the 2009 ISA (U.S. EPA, 2009) concluded that the relationship between short-term
exposure to PM10-2.5 and respiratory effects is "suggestive of a causal relationship."
Epidemiologic findings were consistent for respiratory infection and combined respiratory-
related diseases, but not for COPD. Studies were characterized by overall uncertainty in the
exposure assignment approach and limited information regarding potential copollutant
confounding. Controlled human exposure studies of short-term PM10-2.5 exposures found no lung
function decrements and inconsistent evidence for pulmonary inflammation. Animal
toxicological studies were limited to those using non-inhalation (e.g., intra-tracheal instillation)
routes of PM10-2.5 exposure.
Recent epidemiologic findings consistently link PM10-2.5 exposure to asthma
exacerbation and respiratory mortality, with some evidence that associations remain positive
(though attenuated in some studies of mortality) in copollutant models that include PM2.5 or
gaseous pollutants. Studies provide limited evidence for positive associations with other
respiratory outcomes, including COPD exacerbation, respiratory infection, and combined
respiratory-related diseases (U.S. EPA, 2019, Table 5-36). As noted above for other endpoints,
an uncertainty in these epidemiologic studies is the lack of a systematic evaluation of the various
methods used to estimate PM10-2.5 concentrations and the resulting uncertainty in the spatial and
temporal variability in PM10-2.5 concentrations compared to PM2.5 (U.S. EPA, 2019, sections
2.5.1.2.3 and 3.3.1.1). Taken together, the ISA concludes that "the collective evidence is
suggestive of, but not sufficient to infer, a causal relationship between short-term PM10-2.5
exposure and respiratory effects" (U.S. EPA, 2019, p. 5-270).
4.2.1.4	Cancer
Long-term exposures
4-10

-------
In the last review, little information was available from studies of cancer following
inhalation exposures to PM10-2.5. Thus, the 2009 ISA determined the evidence was "inadequate
to assess the relationship between long-term PM10-2.5 exposures and cancer" (U.S. EPA, 2009).
Since the 2009 ISA, the assessment of long-term PM10-2.5 exposure and cancer remains limited,
with a few recent epidemiologic studies reporting positive, but imprecise, associations with lung
cancer incidence. Uncertainty remains in these studies with respect to exposure measurement
error due to the use of PM10-2.5 predictions that have not been validated by monitored PM10-2.5
concentrations (U.S. EPA, 2019, sections 3.3.2.3 and 10.3.4). Relatively few experimental
studies of PM10-2.5 have been conducted, though available studies indicate that PM10-2.5 exhibits
two key characteristics of carcinogens: genotoxicity and oxidative stress. While limited, such
experimental studies provide some evidence of biological plausibility for the findings in a small
number of epidemiologic studies (U.S. EPA, 2019, section 10.3.4).
Taken together, the small number of epidemiologic and experimental studies, along with
uncertainty with respect to exposure measurement error, contribute to the determination in the
ISA that, "the evidence is suggestive of, but not sufficient to infer, a causal relationship between
long-term PM10-2.5 exposure and cancer" (U.S. EPA, 2019, p. 10-87).
4.2.1.5	Metabolic Effects
Long-term exposures
The 2009 ISA did not make a causality determination for PMio-25-related metabolic
effects. Since the last review, one epidemiologic study shows an association between long-term
PM10-2.5 exposure and incident diabetes, while additional cross-sectional studies report
associations with effects on glucose or insulin homeostasis (U.S. EPA, 2019, section 7.4). As
discussed above for other outcomes, uncertainties with the epidemiologic evidence include the
potential for copollutant confounding and exposure measurement error (U.S. EPA, 2019, Tables
7-14 and 7-15). The evidence base to support the biological plausibility of metabolic effects
following PM10-2.5 exposures is limited, but a cross-sectional study that investigated biomarkers
of insulin resistance and systemic and peripheral inflammation may support a pathway leading to
type 2 diabetes (U.S. EPA, 2019, sections 7.4.1 and 7.4.3). Based on the expanded, though still
limited evidence base, the ISA concludes that, "[ojverall, the evidence is suggestive of, but not
sufficient to infer, a causal relationship between [long]-term PM10-2.5 exposure and metabolic
effects" (U.S. EPA, 2019, p. 7-56).
4.2.1.6	Nervous system effects
Long-term exposures
The 2009 ISA did not make a causality determination for PMio-25-related nervous system
effects. In the current review, newly available epidemiologic studies report associations between
4-11

-------
PMio-2.5 and impaired cognition and anxiety in adults in longitudinal analyses (U.S. EPA, 2019,
Table 8-25, section 8.4.5). Associations of long-term exposure with neurodevelopmental effects
are not consistently reported in children (U.S. EPA, 2019, sections 8.4.4 and 8.4.5). Uncertainties
in these studies include the potential for copollutant confounding, as no studies examined
copollutants models (U.S. EPA, 2019, section 8.4.5), and for exposure measurement error, given
the use of various model-based subtraction methods to estimate PM10-2.5 concentrations (U.S.
EPA, 2019, Table 8-25). In addition, there is only limited animal toxicological evidence
supporting the biological plausibility of nervous system effects (U.S. EPA, 2019, sections 8.4.1
and 8.4.5). Overall, the ISA concludes that, "the evidence is suggestive of, but not sufficient to
infer, a causal relationship between long-term PM10-2.5 exposure and nervous system effects (U.S.
EPA, 2019, p. 8-75).
4.2.1.7 Conclusions Drawn from the Evidence
Based on the evidence available in the current review, as assessed in the ISA (U.S. EPA,
2019) and summarized in 4.2.1.1 to 4.2.1.6 above, we revisit the policy-relevant questions posed
at the beginning of this section:
• To what extent does the currently available scientific evidence strengthen, or otherwise
alter, our conclusions from the last review regarding health effects attributable to long-
or short-term PM10-2.5 exposures? Have previously identified uncertainties been
reduced? What important uncertainties remain and have new uncertainties been
identified?
In the last review, the strongest evidence for PMio-2 5-related health effects was for
cardiovascular effects, respiratory effects, and premature mortality following short-term
exposures. For each of these categories of effects, the ISA concluded that the evidence was
"suggestive of a causal relationship" (U.S. EPA, 2009, section 2.3.3). As summarized in the
sections above, key uncertainties in the evidence resulted from limitations in the approaches used
to estimate ambient PM10-2.5 concentrations in epidemiologic studies, limited examination of the
potential for confounding by co-occurring pollutants, and limited support for the biological
plausibility of the serious effects reported in many epidemiologic studies. Since 2009, the
evidence base for several PM10-2.5-related health effects has expanded, broadening our
understanding of the range of health effects linked to PM10-2.5 exposures. This includes expanded
evidence for the relationships between long-term exposures and cardiovascular effects, metabolic
effects, nervous system effects, cancer, and mortality. However, key limitations in the evidence
that were identified in the 2009 ISA persist in studies that have become available since the last
review. These limitations include the following:
• The use of a variety of methods to estimate PM10-2.5 exposures in epidemiologic studies
and the lack of systematic evaluation of these methods, together with the relatively high
4-12

-------
spatial and temporal variability in ambient PM10-2.5 concentrations and the small number
of monitoring sites, results in uncertainty in exposure estimates;
•	The limited number of studies that evaluate PM10-2.5 health effect associations in
copollutant models, together with evidence from some studies for attenuation of
associations in such models, results in uncertainty in the independence of PM10-2.5 health
effect associations from co-occurring pollutants;
•	The limited number of controlled human exposure and animal toxicology studies of
PM10-2.5 inhalation contributes to uncertainty in the biological plausibility of the PM10-2.5-
related effects reported in epidemiologic studies.
Thus, while new evidence is available for a broader range of health outcomes in the current
review, that evidence is subject to the same types of uncertainties that were identified in the last
review of the PM NAAQS. As in the last review, these uncertainties contribute to the
conclusions in the ISA that the evidence for the PMio-25-related health effects discussed in this
section is "suggestive of, but not sufficient to infer" causal relationships.
4.3 CASAC ADVICE AND PUBLIC COMMENTS
As part of its review of the draft PA, the CASAC has provided advice on the adequacy of
the public health protection afforded by the current primary PM10 standard. As for PM2.5 (section
3.4), the CASAC's advice is documented in a letter sent to the EPA Administrator (Cox, 2019).
In its comments on the draft PA, the CASAC concurs with the draft PA's overall
preliminary conclusions that it is appropriate to consider retaining the current primary PM10
standard without revision, stating that "[t]he CASAC agrees with the EPA conclusion that'.. .the
available evidence does not call into question the adequacy of the public health protection
afforded by the current primary PM10 standard and that evidence supports considering of
retaining the current standard in this review'" (Cox, 2019, p.3 of letter). The CASAC finds the
more limited approach taken for PM10, compared to PM2.5, to be "reasonable and appropriate"
given the less certain evidence and the conclusion that "key uncertainties identified in the last
review remain" (Cox, 2019, p. 13 of consensus responses). To reduce these uncertainties in
future reviews, the CASAC recommends improvements to PM10-2.5 exposure assessment,
including a more extensive network for direct monitoring of the PM10-2.5 fraction (Cox, 2019, p.
13 of consensus responses). The CASAC also recommends additional human clinical and animal
toxicology studies of the PM10-2.5 fraction to improve the understanding of biological causal
mechanisms and pathways (Cox, 2019, p. 13 of consensus responses).
We also received a limited number of public comments on the adequacy of the primary
PM10 standard. Of those who provided comments on the PM10 standard, most commenters
support the preliminary conclusion that it is appropriate to consider retaining the current PM10
standard, without revision. One group that includes members of the academic research
4-13

-------
community (i.e., the Independent PM Review Panel), however, supports lowering the level of the
primary PMio standard, consistent with their recommendation to also lower the level of the 24-
hour primary PM2.5 standard.
4.4 CONCLUSIONS ON THE ADEQUACY OF THE CURRENT
STANDARD
This section describes our conclusions regarding the adequacy of the current primary
PM10 standard. Our approach to reaching conclusions considers the EPA's assessment of the
current scientific evidence for PMio-2 5-related health effects in the ISA and takes into account
the advice received from the CAS AC (Cox, 2019) and comments from the members of the
public. We revisit the overarching question for this chapter:
• Does the currently available scientific evidence support or call into question the
adequacy of the protection afforded by the current primary PM10 standard against
health effects associated with exposures to PM10-2.5?
In answering this question, we consider the currently available evidence within the context of the
rationale supporting the decision in the last review to retain the primary PM10 standard. We
recognize that a final decision on the primary PM10 standard in the current review will be largely
a public health policy judgement in which the Administrator weighs the evidence, including its
associated uncertainties.
As discussed in section 4.1.1 above, the decision to retain the primary PM10 standard in
the last review recognized the importance of maintaining some degree of protection against
PM10-2.5 exposures, given the evidence for PMio-25-related health effects, but noted uncertainties
in the potential public health implications of revising the existing PM10 standard. Regarding
evidence for PMio-25-related health effects, the decision noted that epidemiologic studies had
reported positive associations between PM10-2.5 and mortality or morbidity in cities across North
America, Europe, and Asia, encompassing a variety of environments where PM10-2.5 sources and
composition are expected to vary widely. Although most of these studies examined PM10-2.5
health effect associations in urban areas, some studies had also linked mortality and morbidity
with relatively high ambient concentrations of particles of non-urban crustal origin. Drawing
from this evidence, it was judged appropriate to maintain a standard that provides some measure
of protection against exposures to PM10-2.5, regardless of location, source of origin, or particle
composition (78 FR 3176, January 15, 2013). As discussed above in section 4.1.1, it was further
judged appropriate to retain the PM10 indicator given that the varying concentrations of PM10-2.5
permitted in urban versus non-urban areas under a PM10 standard, based on the varying levels of
PM2.5 present (i.e., lower PM10-2.5 concentrations allowed in urban areas, where PM2.5
concentrations tend to be higher), appropriately reflected differences in the strength of PM10-2.5
4-14

-------
health effects evidence. With regard to uncertainties, limitations in the estimates of ambient
PMio-2.5 used in epidemiologic studies, the limited evaluation of copollutant models to address
the potential for confounding, and the limited number of experimental studies supporting
biologically plausible pathways for PMio-2 5-related effects were all highlighted. These and other
limitations in the PM10-2.5 evidence raised questions as to whether additional public health
improvements would be achieved by revising the existing PM10 standard.
Since the last review, the evidence for several PMio-2 5-related health effects has
expanded, particularly for long-term exposures, broadening our understanding of the range of
effects linked to PM10-2.5 exposures. As in the last review, epidemiologic studies continue to
report positive associations with mortality or morbidity in cities across North America, Europe,
and Asia, where PM10-2.5 sources and composition are expected to vary widely. Such studies
provide an important part of the body of evidence supporting the strengthened causality
determinations (and new determinations) for long-term PM10-2.5 exposures and mortality,
cardiovascular effects, metabolic effects, nervous system effects and cancer (U.S. EPA, 2019).
Thus, the scientific evidence that has become available since the last review does not call into
question the decision in that review to maintain a primary standard that provides some measure
of public health protection against PM10-2.5 exposures, regardless of location, source of origin, or
particle composition. In addition, recent epidemiologic studies do not call into question the
judgment in the last review that it is appropriate to retain the PM10 indicator, given that the
varying concentrations of coarse particles permitted in urban versus non-urban areas under a
PM10 standard (i.e., based on the varying concentrations of PM2.5 present) appropriately reflect
the differences in the strength of evidence regarding coarse particle health effects.
As in the last review, important uncertainties remain in the evidence base for PM10-2.5-
related health effects. As summarized in section 4.2.1 above, these include uncertainties in the
PM10-2.5 exposure estimates used in epidemiologic studies, in the independence of PM10-2.5 health
effect associations, and in the biological plausibility of the PMio-2 5-related effects. Thus, the
evidence available in the current review is subject to the same broad uncertainties as were
present in the last review. Consistent with the assessment of the evidence in the 2009 ISA (U.S.
EPA, 2009), these uncertainties contribute to the determinations in the current ISA that the
evidence for key PMio-2 5-related health effects is "suggestive of, but not sufficient to infer"
causal relationships (U.S. EPA, 2019). Drawing from this information, we reach the conclusion
that, as in the last review, such uncertainties raise questions regarding the degree to which
additional public health improvements would be achieved by revising the existing PM10
standard.
When the above information is taken together, we reach the conclusion that the available
evidence does not call into question the scientific judgments that informed the decision in the last
4-15

-------
review to retain the current primary PMio standard in order to protect against PM10-2.5 exposures.
Specifically, while the available evidence supports maintaining a PM10 standard to provide some
measure of protection against PM10-2.5 exposures, uncertainties in the evidence lead to questions
regarding the potential public health implications of revising the existing PM10 standard. Thus,
consistent with the approach taken in the last review and with the advice from the CASAC in this
review, we reach the conclusions that the available evidence does not call into question the
adequacy of the public health protection afforded by the current primary PM10 standard and that
evidence supports consideration of retaining the current standard in this review. As such, we
have not evaluated alternative standards in this PA.
4.5 AREAS FOR FUTURE RESEARCH AND DATA COLLECTION
As discussed above, a number of key uncertainties and limitations in the health evidence
have been considered in this review. In this section, we highlight areas for future health-related
research and data collection activities to address these uncertainties and limitations in the current
body of evidence. These efforts, if undertaken, could provide important evidence for informing
future reviews of the PM NAAQS. Key areas for future research efforts are summarized below.
•	The body of experimental inhalation studies of exposure to PM10-2.5 (e.g., controlled
human exposure and animal toxicology studies) is currently relatively sparse. While
coarse PM inhalation studies in rats and mice are complicated by substantial differences
in dosimetry (i.e., compared to humans), additional experimental studies of short- or
long-term PM10-2.5 exposures could play an important role in weight of evidence
judgments in future ISAs. Experimental evaluation of effects that are plausibly related to
the serious health outcomes documented in epidemiologic studies could be particularly
informative. Such effects could include changes in markers of cardiovascular or
respiratory function, similar to the effects that have been evaluated following PM2.5
exposures (e.g., vascular function, blood pressure, heart rate and heart rate variability,
markers of potential for coagulation, systemic and respiratory inflammation, respiratory
function, etc.).
•	The potential for exposure error is of particular concern for PM10-2.5, given its less
homogeneous atmospheric distribution compared to fine particles (U.S. EPA, 2019,
section 1.2.1.5) and the relatively sparse PM10-2.5 monitoring network. Therefore, efforts
to develop and validate new exposure estimation approaches, or to further validate
existing approaches, would be informative.
•	Existing epidemiologic studies have rarely examined associations with PM10-2.5 in
copollutant models, contributing to uncertainty in the degree to which reported health
effect associations are independent of potential confounding variables. Additional
epidemiologic studies that evaluate copollutants models would be informative.
•	Epidemiologic studies currently use a variety of approaches to measure/estimate PM10-2.5
concentrations, including: (1) difference method with co-located monitors, (2) difference
method with area-wide averages of monitored PM10 and PM2.5, (3) difference method
4-16

-------
with area-wide averages of modeled PMio and PM2.5 or (4) direct measurement of
PM10-2.5 using a dichotomous sampler. It is important that we better understand how these
methods compare to one another, both in terms of absolute estimated concentrations and
in terms of the spatial and temporal correlations in those estimated concentrations
between methods.
Measurement capabilities and the availability of PM10-2.5 ambient concentration data have
greatly increased since the 2009 ISA (U.S. EPA, 2019, section 2.5.1.1.3). Starting in
2011, PM10-2.5 has been monitored at NCore stations, IMPROVE stations, and several
sites run by State and local agencies. To date, epidemiologic studies have used a variety
of approaches to measure/estimate PM10-2.5 concentrations but have not used direct
measurements from NCore or IMPROVE stations to evaluate health effects associations
with PM10-2.5 exposure. A body of epidemiologic studies that evaluate health effect
associations using monitoring data from these stations could allow more direct
comparisons of results across studies.
Evaluate and expand the PM10-2.5 network, along with speciation of PM10-2.5 including
multi-elements, major ions, carbon (including carbonate carbon), and bioaerosols
Characterize PM10-2.5 in different health-relevant exposure environments (e.g., city center,
suburban, roadside, agricultural, and rural areas) for mass, elements (including potential
toxic species), carbonaceous materials (including selected organic compounds and
carbonate), water-soluble ions, and bioaerosols (including endotoxins, 1,3 beta glucan,
and total protein).
Additional areas of interest for future research include:
o Further evaluation of the potential for particular PM10-2.5 components, groups of
components, or other particle characteristics to contribute to exposure-related
health effects.
o Research to improve our understanding of concentration-response relationships
and the confidence bounds around these relationships, especially at lower ambient
PM10-2.5 concentrations.
o Identifying novel populations that could be at-risk of PMio-2 5-related health
effects.
o Modeling to estimate PM10-2.5 mass and composition in areas with sparse or less-
than-daily monitoring.
4-17

-------
REFERENCES
Adar, SD, Filigrana, PA, Clements, N and Peel, JL (2014). Ambient coarse particulate matter
and human health: A systematic review and meta-analysis. Current Environmental Health
Reports 1: 258-274.
Chan, E, Gantt, B and McDow, S (2018). The reduction of summer sulfate and switch from
summertime to wintertime PM2.5 concentration maxima in the United States. Atmos
Environ 175: 25-32.
Cox, LA. (2019). Letter from Louis Anthony Cox, Jr., Chair, Clean Air Scientific Advisory
Committee, to Administrator Andrew R. Wheeler. Re: CASAC Review of the EPA's
Policy Assessment for the Review of the National Ambient Air Quality Standards for
Particulate Matter (ExternalReview Draft - September 2019). December 16, 2019. EPA-
CASAC-20-001. U.S. EPA HQ, Washington DC. Office of the Administrator, Science
Advisory Board. Available at:
https://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/E2F
6C7173 72016128525 84D20069DFB l/$File/EP A-C AS AC-20-001 .pdf.
U.S. EPA. (2009). Integrated Science Assessment for Particulate Matter (Final Report). Research
Triangle Park, NC. Office of Research and Development, National Center for
Environmental Assessment. U.S. EPA. EPA-600/R-08-139F. December 2009. Available
at: https://cfpub.epa.gov/ncea/risk/recordisplav.cfm?deid=216546.
U.S. EPA. (2015). Preamble to the integrated science assessments. Research Triangle Park, NC.
U.S. Environmental Protection Agency, Office of Research and Development, National
Center for Environmental Assessment, RTP Division. U.S. EPA. EPA/600/R-15/067.
November 2015. Available at:
https://cfpub.epa. gov/ncea/isa/recordisplav.cfm?deid=310244.
U.S. EPA. (2019). Integrated Science Assessment (ISA) for Particulate Matter (Final Report).
Washington, DC. U.S. Environmental Protection Agency, Office of Research and
Development, National Center for Environmental Assessment. U.S. EPA. EPA/600/R-
19/188. December 2019. Available at: https://www.epa.gov/naaqs/particulate-matter-pm-
standards-integrated-science-assessments-current-review.
4-18

-------
5 REVIEW OF THE SECONDARY STANDARDS
This chapter presents key policy-relevant considerations and summary conclusions
regarding the public welfare protection provided by the current secondary PM standards to
protect against PM-related visibility impairment, climate effects, and materials effects. These
considerations and conclusions are framed by a series of policy-relevant questions, including the
following overarching question:
• Does the currently available scientific evidence and quantitative information support
or call into question the adequacy of the protection afforded by the current
secondary PM standards?
The answer to this question is informed by evaluation of a series of more specific policy-
relevant questions, which expand upon those presented at the outset of this review in the IRP
(U.S. EPA, 2016). Answers to these questions are intended to inform decisions by the
Administrator on whether, and if so how, to revise the secondary PM standards.
Section 5.1 presents our approach for reviewing the secondary standards for PM. Section
5.2.1 presents our consideration of the available scientific evidence and our consideration of
quantitative information for visibility effects, while section 5.2.2 considers the available
scientific evidence for each of the non-visibility welfare effects (climate effects and materials
effects) separately.1 Section 5.3 summarizes the advice and recommendations received from the
CASAC during its review of the draft PA, and by public comments received on the draft
document. Conclusions regarding the public welfare protection provided by the current
secondary PM standards are summarized in section 5.4. Section 5.5 discusses areas for future
research and data collection to improve our understanding of PM-related welfare effects in future
reviews.
5.1 APPROACH
In the last review of the PM NAAQS, completed in 2012, the EPA retained the secondary
24-hour PM2.5 standard, with its level of 35 |ig/m3, and the 24-hour PM10 standard, with its level
of 150 |ig/m3 (78 FR 3228, January 15, 2013). The EPA also retained the level, set at 15 |ig/m3,
and averaging time of the annual PM2.5 standard, while revising the form. With regard to the
1 Other welfare effects of PM, such as ecological effects, are being considered in the separate, on-going review of
the secondary NAAQS for oxides of nitrogen and oxides of sulfur. Accordingly, the public welfare protection
provided by the secondary PM standards against ecological effects such as those related to deposition of nitrogen-
and sulfur-containing compounds in vulnerable ecosystems is being considered in that separate review. Thus, the
Administrator's conclusion in this review will be focused only and specifically on the adequacy of public welfare
protection provided by the secondary PM standards from effects related to visibility, climate, and materials.
5-1

-------
form of the annual PM2.5 standard, the EPA removed the option for spatial averaging (78 FR
3228, January 15, 2013). Key aspects of the Administrator's decisions on the secondary PM
standards for non-visibility effects and visibility effects are described below in section 5.1.1.
5.1.1 Approach Used in the Last Review
The 2012 decision on the adequacy of the secondary PM standards was based on
consideration of the protection provided by those standards for visibility and for the non-
visibility effects of materials damage, climate effects and ecological effects. As noted earlier, the
current review of the public welfare protection provided by the secondary PM standards against
ecological effects is occurring in the separate, on-going review of the secondary NAAQS for
oxides of nitrogen and oxides of sulfur. Thus, the consideration of ecological effects in the 2012
review is not discussed here. Rather, the sections below focus on the Administrator's
consideration of climate and materials effects (section 5.1.1.1) and visibility effects (section
5.1.1.2).
5.1.1.1 Non- Visibility Effects
With regard to the role of PM in climate, the Administrator considered whether it was
appropriate to establish any distinct secondary PM standards to address welfare effects
associated with climate impacts. In considering the scientific evidence, she noted the 2009 ISA
conclusion "that a causal relationship exists between PM and effects on climate" and that
aerosols2 alter climate processes directly through radiative forcing and by indirect effects on
cloud brightness, changes in precipitation, and possible changes in cloud lifetimes (U.S. EPA,
2009, section 9.3.10). Additionally, the major aerosol components with the potential to affect
climate processes (i.e., black carbon (BC), organic carbon (OC), sulfates, nitrates and mineral
dusts) vary in their reflectivity, forcing efficiencies, and direction of climate forcing (U.S. EPA,
2009, section 9.3.10).
Noting the strong evidence indicating that aerosols affect climate, the Administrator
further considered what the available information indicated regarding the adequacy of protection
provided by the secondary PM standards. She noted that a number of uncertainties in the
scientific information affected our ability to quantitatively evaluate the standards in this regard.
2 In the climate sciences research community, PM is encompassed by what is typically referred to as aerosol. An
aerosol is defined as a solid or liquid suspended in a gas, but PM refers to the solid or liquid phase of an aerosol.
In this review of the secondary PM NAAQS the discussion on climate effects of PM uses the term PM throughout
for consistency with the ISA (U.S. EPA, 2019) as well as to emphasize that the climate processes altered by
aerosols are generally altered by the PM portion of the aerosol. Exceptions to this practice include the discussion
of climate effects in the last review, when aerosol was used when discussing suspending aerosol particles, and for
certain acronyms that are widely used by the climate community that include the term aerosol (e.g., aerosol
optical depth, or AOD).
5-2

-------
For example, the ISA and PA noted the spatial and temporal heterogeneity of PM components
that contribute to climate forcing, uncertainties in the measurement of aerosol components,
inadequate consideration of aerosol impacts in climate modeling, insufficient data on local and
regional microclimate variations and heterogeneity of cloud formations. In light of these
uncertainties and the lack of sufficient data, the 2011 PA concluded that it was not feasible in the
last review "to conduct a quantitative analysis for the purpose of informing revisions [to the
secondary PM NAAQS] based on climate" (U.S. EPA, 2011, pp. 5-11 to 5-12) and that there was
insufficient information available to base a national ambient air quality standard on climate
impacts associated with ambient air concentrations of PM or its constituents (U.S. EPA, 2011,
section 5.2.3). The Administrator agreed with this conclusion (78 FR 3225-3226, January 15,
2013).
With regard to materials effects, the Administrator also considered effects associated with
the deposition of PM (i.e., dry and wet deposition), including both physical damage (materials
effects) and aesthetic qualities (soiling effects). The deposition of PM can physically affect
materials, adding to the effects of natural weathering processes, by promoting or accelerating the
corrosion of metals; by degrading paints; and by deteriorating building materials such as stone,
concrete, and marble (U.S. EPA, 2009, section 9.5). Additionally, the deposition of PM from
ambient air can reduce the aesthetic appeal of buildings and objects through soiling. The ISA
concluded that evidence was "sufficient to conclude that a causal relationship exists between PM
and effects on materials" (U.S. EPA, 2009, sections 2.5.4 and 9.5.4). However, the 2011 PA
noted that quantitative relationships were lacking between particle size, concentrations, and
frequency of repainting and repair of surfaces and that considerable uncertainty exists in the
contributions of co-occurring pollutants to materials damage and soiling processes (U.S. EPA,
2011, p. 5-29). The 2011 PA concluded that none of the evidence available in the last review
called into question the adequacy of the existing secondary PM standards to protect against
material effects (U.S. EPA, 2011, p. 5-29). The Administrator agreed with this conclusion (78
FR 3225-3226, January 15, 2013).
In considering non-visibility welfare effects in the last review, as discussed above, the
Administrator concluded that, while it is important to maintain an appropriate degree of control
of fine and coarse particles to address non-visibility welfare effects, "[i]n the absence of
information that would support any different standards.. .it is appropriate to retain the existing
suite of secondary standards" (78 FR 3225-3226, January 15, 2013). Her decision was consistent
with the CASAC advice related to non-visibility effects. Specifically, the CASAC agreed with
the 2011 PA conclusions that, while these effects are important, "there is not currently a strong
technical basis to support revisions of the current standards to protect against these other welfare
effects" (Samet, 2010, p. 5). Thus, the Administrator concluded that it was appropriate to retain
5-3

-------
all aspects of the existing 24-hour PM2.5 and PM10 secondary standards. With regard to the
secondary annual PM2.5 standard, the Administrator concluded that it was appropriate to retain a
level of 15.0 |ig/m3 while revising only the form of the standard to remove the option for spatial
averaging (78 FR 3225-3226, January 15, 2013).
5.1.1.2 Visibility Effects
Having reached the conclusion to retain the existing secondary PM standards to protect
against non-visibility welfare effects, the Administrator next considered the level of protection
that would be requisite to protect public welfare against PM-related visibility impairment and
whether to adopt a distinct secondary standard to achieve this level of protection. In reaching her
final decision that the existing 24-hour PM2.5 standard provides sufficient protection against PM-
related visibility impairment (78 FR 3228, January 15, 2013), the Administrator considered the
evidence assessed in the 2009 ISA (U.S. EPA, 2009) and the analyses included in the Urban-
Focused Visibility Assessment (2010 UFVA; U.S. EPA, 2010) and the 2011 PA (U.S. EPA,
2011). She also considered the degree of protection for visibility that would be provided by the
existing secondary standard, focusing specifically on the secondary 24-hour PM2.5 standard with
its level of 35 |ig/m3. These considerations, and the Administrator's conclusions regarding
visibility are discussed in more detail below.
In the last review, the ISA concluded that, "collectively, the evidence is sufficient to
conclude that a causal relationship exists between PM and visibility impairment" (U.S. EPA,
2009, p. 2-28). Visibility impairment is caused by light scattering and absorption by suspended
particles and gases, including water content of aerosols.3 The available evidence in the last
review indicated that specific components of PM have been shown to contribute to visibility
impairment. For example, at sufficiently high relative humidity values, sulfate and nitrate are the
PM components that scatter more light and thus contribute most efficiently to visibility
impairment. Elemental carbon (EC) and OC are also important contributors, especially in the
northwestern U.S. where their contribution to PM2.5 mass is higher. Crustal materials can be
significant contributors to visibility impairment, particularly for remote areas in the arid
southwestern U.S. (U.S. EPA, 2009, section 2.5.1).
Visibility impairment can have implications for people's enjoyment of daily activities
and for their overall sense of well-being (U.S. EPA, 2009, section 9.2). In consideration of the
potential public welfare implication of various degrees of PM-related visibility impairment, the
3 All particles scatter light and, although a larger particle scatters more light than a similarly shaped smaller particle
of the same composition, the light scattered per unit of mass is greatest for particles with diameters from-0.3-1.0
|im (U.S. EPA, 2009, section 2.5.1). Particles with hygroscopic components (e.g., particulate sulfate and nitrate)
contribute more to light extinction at higher relative humidity than at lower relative humidity because they change
size in the atmosphere in response to relative humidity.
5-4

-------
Administrator considered the available visibility preference studies that were part of the overall
body of evidence in the 2009 ISA and reviewed as a part of the 2010 UFVA. These preference
studies provided information about the potential public welfare implications of visibility
impairment from surveys in which participants were asked questions about their preferences or
the values they placed on various visibility conditions, as displayed to them in scenic
photographs or in images with a range of known light extinction levels.4
In noting the relationship between PM concentrations and PM-related light extinction, the
Administrator focused on identifying an adequate level of protection against visibility-related
welfare effects. She first concluded that a standard in terms of a PM2.5 visibility index would
provide a measure of protection against PM-related light extinction that directly takes into
account the factors (i.e., species composition and relative humidity) that influence the
relationship between PM2.5 in ambient air and PM-related visibility impairment. A PM2.5
visibility index standard would afford a relatively high degree of uniformity of visual air quality
protection in areas across the country by directly incorporating the effects of differences of PM2.5
composition and relative humidity. In defining a target level of protection in terms of a PM2.5
visibility index, as discussed below, the Administrator considered specific elements of the index,
including the basis for its derivation, as well as an appropriate averaging time, level, and form.
With regard to the basis for derivation of a visibility index, the Administrator concluded
that it was appropriate to use an adjusted version of the original IMPROVE algorithm,5 in
conjunction with monthly average relative humidity data based on long-term climatological
means. In so concluding, the Administrator noted the CASAC conclusion on the reasonableness
of reliance on a PM2.5 light extinction indicator calculated from PM2.5 chemical composition and
relative humidity. In considering alternative approaches for a focus on visibility, the
Administrator recognized that the available mass monitoring methods did not include
measurement of the full water content of ambient PM2.5, nor did they provide information on the
composition of PM2.5, both of which contribute to visibility impacts (77 FR 38980, June 29,
2012). In addition, at the time of the proposal, the Administrator recognized that suitable
equipment and performance-based verification procedures did not then exist for direct
4	Preference studies were available in four urban areas in the last review. Three western preference studies were
available, including one in Denver, Colorado (Ely et al., 1991), one in the lower Fraser River valley near
Vancouver, British Columbia, Canada (Pryor, 1996), and one in Phoenix, Arizona (BBC Research & Consulting,
2003). A pilot focus group study was also conducted for Washington, DC (Abt Associates, 2001), and a replicate
study with 26 participants was also conducted for Washington, DC (Smith and Howell, 2009). More details about
these studies are available in Appendix D.
5	The revised IMPROVE algorithm (Pitchford et al., 2007) uses major PM chemical composition measurements and
relative humidity estimates to calculate light extinction. For more information about the derivation of and input
data required for the original and revised IMPROVE algorithms, see 78 FR 3168-3177, January 15, 2013.
5-5

-------
measurement of light extinction and could not be developed within the time frame of the review
(77 FR 38980-38981, June 29, 2012).
With regard to the averaging time of the index, the Administrator concluded that a 24-
hour averaging time would be appropriate for a visibility index (78 FR 3226, January 15, 2013).
Although she recognized that hourly or sub-daily (4- to 6-hour) averaging times, within daylight
hours and excluding hours with relatively high humidity, are more directly related to the short-
term nature of the perception of PM-related visibility impairment and relevant exposure periods
for segments of the viewing public than a 24-hour averaging time, she also noted that there were
data quality uncertainties associated with the instruments used to provide the hourly PM2.5 mass
measurements required for an averaging time shorter than 24 hours. The Administrator also
considered the results of analyses that compared 24-hour and 4-hour averaging times for
calculating the index. These analyses showed good correlation between 24-hour and 4-hour
average PM2.5 light extinction, as evidenced by reasonably high city-specific and pooled R-
squared values, generally in the range of over 0.6 to over 0.8. Based on these analyses and the
2011 PA conclusions regarding them, the Administrator concluded that a 24-hour averaging time
would be a reasonable and appropriate surrogate for a sub-daily averaging time.
With regard to the statistical form of the index, the Administrator settled on a 3-year
average of annual 90th percentile values. In so doing, she noted that a 3-year average form
provided stability from the occasional effect of inter-annual meteorological variability that can
result in unusually high pollution levels for a particular year (78 FR 3198, January 15, 2013; U.S.
EPA, 2011, p. 4-58). Regarding the annual statistic to be averaged, the 2010 UFVA evaluated
three different statistics: 90th, 95th, and 98th percentiles (U.S. EPA, 2010, chapter 4). In
considering these alternative percentiles, the 2011 PA noted that the Regional Haze Program
targets the 20 percent most impaired days for improvements in visual air quality in Federal Class
I areas and that the median of the distribution of these 20 percent worst days would be the 90th
percentile. The 2011 PA further noted that strategies that are implemented so that 90 percent of
days would have visual air quality that is at or below the level of the standard would reasonably
be expected to lead to improvements in visual air quality for the 20 percent most impaired days.
Lastly, the 2011 PA recognized that the available studies on people's preferences did not address
frequency of occurrence of different levels of visibility and did not identify a basis for a different
target for urban areas than that for Class I areas (U.S. EPA, 2011, p. 4-59). These considerations
led the Administrator to conclude that 90th percentile form was the most appropriate annual
statistic to be averaged across three years (78 FR 3226, January 15, 2013).
With regard to the level of the index, the Administrator considered the visibility
preferences studies conducted in four urban areas (U.S. EPA, 2011, p. 4-61). Based on these
5-6

-------
studies, the PA identified a range of levels from 20 to 30 deciviews (dv)6 as being a reasonable
range of "candidate protection levels" (CPLs).7 In considering this range of CPLs, the
Administrator noted the uncertainties and limitations in public preference studies, including the
small number of stated preference studies available; the relatively small number of study
participants and the extent to which the study participants may not be representative of the
broader study area population in some of the studies; and the variations in the specific materials
and methods used in each study. She concluded that the substantial degrees of variability and
uncertainty in the public preference studies should be reflected in a target protection level at the
upper end of the range of CPLs than if the information were more consistent and certain.
Therefore, the Administrator concluded that it was appropriate to set a target level of protection
in terms of a 24-hour PM2.5 visibility index at 30 dv (78 FR 3226-3227, January 15, 2013).
Based on her considerations and conclusions summarized above, the Administrator
concluded that the protection provided by a secondary standard based on a 3-year visibility
metric, defined in terms of a PM2.5 visibility index with a 24-hour averaging time, a 90th
percentile form averaged over 3 years, and a level of 30 dv, would be requisite to protect public
welfare with regard to visual air quality (78 FR 3227, January 15, 2013). Having reached this
conclusion, she next determined whether an additional distinct secondary standard in terms of a
visibility index was needed given the degree of protection from visibility impairment afforded by
the existing secondary standards. Specifically, she noted that the air quality analyses showed that
all areas meeting the existing 24-hour PM2.5 standard, with its level of 35 |ig/m3, had visual air
quality at least as good as 30 dv, based on the visibility index defined above (Kelly et al., 2012b,
Kelly et al., 2012a). Thus, the secondary 24-hour PM2.5 standard would likely be controlling
relative to a 24-hour visibility index set at a level of 30 dv. Additionally, areas would be unlikely
to exceed the target level of protection for visibility of 30 dv without also exceeding the existing
secondary 24-hour standard. Thus, the Administrator judged that the 24-hour PM2.5 standard
"provides sufficient protection in all areas against the effects of visibility impairment - i.e., that
the existing 24-hour PM2.5 standard would provide at least the target level of protection for
visual air quality of 30 dv which the Administrator judges appropriate" (78 FR 3227, January 15,
2013). She further judged that "[sjince sufficient protection from visibility impairment would be
provided for all areas of the country without adoption of a distinct secondary standard, and
adoption of a distinct secondary standard will not change the degree of over-protection for some
areas of the country.. .adoption of such a distinct secondary standard is not needed to provide
6	Deciview (dv) refers to a scale for characterizing visibility that is defined directly in terms of light extinction. The
deciview scale is frequently used in the scientific and regulatory literature on visibility.
7	For comparison, 20 dv, 25 dv, and 30 dv are equivalent to 64, 112, and 191 megameters (Mm1), respectively.
5-7

-------
requisite protection for both visibility and nonvisibility related welfare effects" (78 FR 3228,
January 15, 2013).
5.1.2 General Approach Used in the Current Review
To evaluate whether it is appropriate to consider retaining the current suite of secondary
PM standards, or whether consideration of revision is appropriate, we have adopted an approach
in this review that builds on the general approach used in the last review and reflects the body of
evidence and information now available. As summarized above, past approaches have been
based most fundamentally on using information from PM visibility studies and quantitative
analyses of PM-related visibility impairment to inform the selection of secondary PM standards
that, in the Administrator's judgment, protect the public welfare from any known or anticipated
effects. These fundamental considerations are again the basis for our approach in this review.
In conducting this assessment, we draw on the current evidence and quantitative
assessments of visibility impairment associated with PM in ambient air. In considering the
scientific and technical information, we consider both the information available at the time of the
last review and information newly available since the last review, including the evidence
assessed in the ISA and updated air quality-based analyses (Appendix D). Figure 5-1 below
illustrates our general approach in developing conclusions regarding the adequacy of the current
secondary standards and, as appropriate, potential alternative standards. In the boxes in Figure 5-
1, the range of questions that we consider in sections 5.2.1 and 5.2.2 below are represented by a
summary of policy-relevant questions that frame our consideration of the scientific evidence and
quantitative analyses.
5-8

-------
Adequacy of Pubiic Welfare Protection Provided by Existing Secondary PI Standards ~
>I3A weight-of-evidence determinations fcr welfare effects ie g .
visibility impairment climate, materials effects.!
>P:eie;eri£0suid',e$ tor Nature of relationship between PVI-
atributable visibility impairment and public perceptions'?
> Studies inkingPM re ckn&ie "roads Naturecf relationship between
ambient PM species and climate'^
>£f.ter-ce fo; PM-reSeietl ;r>6ier^s effects- Nature of relationships
between ambient PM arid materials effects'?
*Urce>la.nr.tss- >r e>defence How do uncertainties in the evidence affect
our understanding of the evidence of effects particularly for
concentrations bekwthe current standard levels"?
Quantitative Assessment-Based
Considerations
>	Nature, magnitude, and importance of
estimated welfare impacts associated
with current secondary PVt standards
>	Uncertainties trt assessment estimates,
including the ability to directly assess the
relationship between changes in PM
concentrations and chanpss in welfare
impacts
Does information call '
tntc question adequacy
oftheivelfare protection
provided ty the current
P-V) standards'?
NO i
Consider retaining
current secondary
\
PM standard{s)
?YES
Consider Potential Alternative Secondary Standards
Indicator
>Support for PM: = and'® PM..'*
>8upportfor indicators si based
on other size fraction PM
components light extinction etc'?
AveraalmTlme
y Support for current 24-hour
and'or annual
>Support for sub-daily,
seasonal or other amaging
time' sV5
Form
> Support for retaining existing
forms?
>Support for alternative form based
on daylight hours or other metric?

Level
>Supportfor PM-a!tribu!abte visibility, climate, or materials impacts at PM concentrations corresponding
to various potential standard levels'?
>Support from quantitative exposuie and-or risk assessments for public welfare improvements «tfi
various potential standard levels"7
^Uncertainties and limitations in the extent to which revised standard levels could result in public
welfare improvements compared to existing standards


Identify range of potential alternative secondary standards for consideration
Figure 5-1. Overview of general approach for review of secondary PM standards.
5-9

-------
5.2 ADEQUACY OF THE CURRENT SECONDARY PM STANDARDS
In considering the available evidence for welfare effects attributable to PM as presented
in the ISA, this section poses the following policy-relevant questions:
• Does the currently available scientific evidence and quantitative information support
or call into question the adequacy of the welfare protection afforded by the current
secondary PM standards?
In answering this question, we have posed a series of more specific questions to aid in
considering the currently available scientific evidence and quantitative information, as discussed
below. In considering the scientific and technical information, we reflect upon both the
information available in the last review and information that is newly available since the last
review as assessed and presented in the ISA (U.S. EPA, 2019), focusing on welfare effects for
which the evidence supports either a "causal" or a "likely to be causal" relationship as described
in the Preamble to the ISA (U.S. EPA, 2015). Table 5-1 lists such causality determinations from
the ISA for welfare effects. As in the last review, the evidence is sufficient to support a causal
relationship between PM and visibility effects (section 5.2.1), climate effects (section 5.2.2) and
materials effects (section 5.2.2).
Table 5-1. Key causality determinations for PM-related welfare effects.
Effect
2009 PM ISA
2019 PM ISA
Visibility effects
Causal
Causal
Climate effects
Causal
Causal
Materials effects
Causal
Causal
5.2.1 Visibility Effects
In the sections below, we consider the nature of visibility-related effects attributable to
PM (section 5.2.1.1) and the quantitative information currently available (section 5.2.1.2).
5.2.1.1 Evidence-Based Considerations
In considering the available evidence of visibility welfare effects attributable to PM as
presented in the ISA, this section addresses the following policy-relevant questions:
• Does the current evidence alter our conclusions from the last review regarding the
nature of visibility effects attributable to PM in ambient air?
Visibility refers to the visual quality of a human's view with respect to color rendition
and contrast definition. It is the ability to perceive landscape form, colors, and textures. Visibility
involves optical and psychophysical properties involving human perception, judgment, and
interpretation. Light between the observer and the object can be scattered into or out of the sight
5-10

-------
path and absorbed by PM or gases in the sight path. As recognized above, the conclusion of the
ISA that "the evidence is sufficient to conclude that a causal relationship exists between PM and
visibility impairment" is consistent with conclusions of causality in the last review (U.S. EPA,
2019, section 13.2.6). These conclusions are based on strong and consistent evidence that
ambient PM can impair visibility in both urban and remote areas (U.S. EPA, 2009, section 9.2.5).
These subsequent questions consider the characterization and quantification of light
extinction and preferences associated with varying degrees of visibility impairment.
• To what extent is new information available that changes or enhances our
understanding of the physics of light extinction and/or its quantification (e.g.,
through light extinction or other monitoring methods or through algorithms such as
IMPROVE)?
Our understanding of the relationship between light extinction and PM mass has changed
little since the 2009 ISA (U.S. EPA, 2009). The combined effect of light scattering and
absorption by particles and gases is characterized as light extinction, i.e., the fraction of light that
is scattered or absorbed per unit of distance in the atmosphere. Light extinction is measured in
units of 1/distance, which is often expressed in the technical literature as visibility per
megameter (abbreviated Mm"1). Higher values of light extinction (usually given in terms of Mm"1
or dv) correspond to lower visibility. When PM is present in the air, its contribution to light
extinction is typically much greater than that of gases (U.S. EPA, 2019, section 13.2.1). The
impact of PM on light scattering depends on particle size and composition, as well as relative
humidity. All particles scatter light, as described by the Mie theory, which relates light scattering
to particle size, shape and index of refraction (U.S. EPA, 2019, section 13.2.3; Van de Hulst,
1981; Mie, 1908). Fine particles scatter more light than coarse particles on a per unit mass basis
and include sulfates, nitrates, organics, light-absorbing carbon, and soil (Malm et al., 1994).
Hygroscopic particles like ammonium sulfate, ammonium nitrate, and sea salt increase in size as
relative humidity increases, leading to increased light scattering (U.S. EPA, 2019, section
13.2.3).
Direct measurements of PM light extinction, scattering, and absorption are considered
more accurate for quantifying visibility impairment than PM mass-based estimates because they
do not depend on assumptions about particle characteristics (e.g., size, shape, density, component
mixture, etc.). Measurements of light extinction can be made with high time resolution, allowing
for characterization of subdaily temporal patterns of visibility impairment. Measurement
methods include transmissometers for measurement of light extinction and the determination of
visual range and integrating nephelometers for measurement of light scattering, as well as
teleradiometers and telephotometers, and photography and photographic modeling (U.S. EPA,
2009; U.S. EPA, 2004). While some recent research confirms and adds to the body of knowledge
5-11

-------
available regarding direct measurements as is described in the ISA, no major new developments
have been made with these measurement methods since the last review (U.S. EPA, 2019, section
13.2.2.2).
A theoretical relationship between light extinction and PM characteristics has been
derived from Mie theory (U.S. EPA, 2019, Equation 13-5) and can be used to estimate light
extinction by combining mass scattering efficiencies of particles with particle concentrations
(U.S. EPA, 2019, section 13.2.3; U.S. EPA, 2009, sections 9.2.2.2 and 9.2.3.1). However,
routine ambient air monitoring rarely includes measurements of particle size and composition
information with sufficient detail for these calculations. Accordingly, a much simpler algorithm
has been developed to make estimating light extinction more practical.
This algorithm, known as the IMPROVE algorithm,8 provides for the estimation of light
extinction (bext), in units of Mm"1, using routinely monitored components of fine (PM2.5) and
coarse (PM10-2.5) PM. Relative humidity data are also needed to estimate the contribution by
liquid water that is in solution with the hygroscopic components of PM. To estimate each
component's contribution to light extinction, their concentrations are multiplied by extinction
coefficients and are additionally multiplied by a water growth factor that accounts for their
expansion with moisture. Both the extinction efficiency coefficients and water growth factors of
the IMPROVE algorithm have been developed by a combination of empirical assessment and
theoretical calculation using particle size distributions associated with each of the major aerosol
components (U.S. EPA, 2019, section 13.2.3.1, section 13.2.3.3).
The original IMPROVE algorithm (Equation D-l in Appendix D), so referenced here to
distinguish it from subsequent variations developed later, was found to underestimate the highest
light scattering values and overestimate the lowest values at IMPROVE monitors throughout the
U.S. (Malm and Hand, 2007; Ryan et al., 2005; Lowenthal and Kumar, 2004) and at sites in
China (U.S. EPA, 2019, section 13.2.3.3). To resolve these biases, a revised IMPROVE equation,
shown in Equation D-2 in Appendix D, was developed (Pitchford et al., 2007) that divides PM
components into smaller and larger sizes of particles in PM2.5, with separate mass scattering
efficiencies and hygroscopic growth functions for each size category. The revised IMPROVE
equation was described in detail in the 2009 ISA (U.S. EPA, 2009) and it both reduced bias at
the lowest and highest scattering values and improved the accuracy of the calculated light bext.
8 The algorithm is referred to as the IMPROVE algorithm as it was developed specifically to use monitoring data
generated at IMPROVE network sites and with equipment specifically designed ot support the IMPROVE
program and was evaluated using IMPROVE optical measurements at the subset of monitoring sites that make
those measurements (Malm et al., 1994).
5-12

-------
However, poorer precision was observed with the revised IMPROVE equation compared to the
original IMPROVE equation (U.S. EPA, 2009).9
Since the time of the last review, Lowenthal and Kumar (2016) have tested and evaluated
a number of modifications to the revised IMPROVE equation based on evaluations of
monitoring data from remote IMPROVE sites. In these locations, they observed that the
multiplier to estimate the concentration of organic matter, [OM], from the concentration of
organic carbon, [OC], was closer to 2.1 than the value of 1.8 used in the revised IMPROVE
equation.10 They also observed that water soluble organic matter absorbs water as a function of
relative humidity, which is not accounted for in either the original or revised IMPROVE
equations and was therefore underestimated in these equations. They further suggested that light
scattering by sulfate was overestimated because the assumption that all sulfate is fully
neutralized ammonium sulfate is not always true (U.S. EPA, 2019, section 13.2.3.3).
Modifications based on these points are reflected in Equation D-3 in Appendix D.
In summary, rather than altering our understanding from the previous review, we
continue to recognize that direct measurements are better at characterizing light extinction than
estimating light extinction with an algorithm. However, in the absence of advances in the
monitoring methods and/or network for directly measuring light extinction, the use of the
IMPROVE equation for estimating light extinction continues to be supported by the evidence,
with some new refinements to the inputs of the IMPROVE equation. Accordingly, as in the last
review, the current review focuses on calculated light extinction when quantifying visibility
impairment resulting from recent concentrations of PM in ambient air.
• What does the available information indicate with regard to factors that influence
light extinction and visibility, as well as variation in these factors and resulting light
extinction across the U.S.?
The ISA provides a comprehensive discussion of the spatial and temporal patterns of
PM2.5 composition and its contribution to light extinction from IMPROVE and CSN monitoring
9	In the most recent IMPROVE report, a combination of the original and revised IMPROVE equations (the modified
original IMPROVE equation) was used (Hand et al., 2011). This equation uses the sea salt term of the revised
equation but does not subdivide the components into two size classes. Further, it uses a factor of 1.8 to estimate
organic matter from organic carbon concentrations and also replaces the constant value of 10 Mm"1 used for
Rayleigh scattering in the original and revised equations with a site-specific term based on elevation and mean
temperature.
10	In areas near sources, PM is often less oxygenated, and therefore, in these locations, much of the organic PM mass
is present as OC (Jimenez et al., 2009). In areas further away from PM sources, organic PM mass is often more
oxygenated as a result of photochemical activity and interactions with other PM and gaseous components in the
atmosphere (Jimenez et al., 2009). Under these conditions, the multiplier to convert OC to OM may be higher
than in locations with less aged organic PM.
5-13

-------
sites, which are mostly rural and urban, respectively.11 The data from these sites for the periods
of 2005-2008 and 2011-2014 were used in the ISA to identify differences in species contributing
to light extinction in urban and rural areas by region and season. This is an expansion over the
analysis in the 2009 ISA, in that the measurements at that time were primarily based
measurements from monitors located in rural areas and at remote sites (U.S. EPA, 2019, section
13.2.4.1, Figures 13-1 through 13-14).
Focusing on the more recent time period of 2011-2014, some major differences in
estimated light extinction are apparent among regions of the U.S. Annual average calculated b ext
was considerably greater in the East and Midwest than in the Southwest. Based on IMPROVE
data, annual average bext was greater than 40 Mm"1 in the Southeast, East Coast, Mid-South,
Central Great Plains, and Appalachian regions, with the highest annual average bext (greater than
50 Mm"1) in the Ohio River Valley,12 while annual average bext was below 40 Mm"1 for all
Western IMPROVE regions. Annual average bext values were also generally higher in the East
than the West based on CSN data, although the highest annual average bext was in the
Sacramento/San Joaquin Valley and Los Angeles areas (U.S. EPA, 2019, section 13.2.4.1, Figure
13-1, Figure 13-3, Figure 13-5).
Components of PM2.5 contributing to light extinction vary regionally. For example, in the
Eastern regions, ammonium sulfate accounted for approximately 35 to 60% of the annual
average bext, with the greatest contributions typically occurring in the summer. The second
greatest contribution to light extinction came from particulate organic matter (POM), ranging
from about 20 to 30% of annual average bext with less seasonal variation than ammonium sulfate.
Ammonium nitrate also contributed approximately 10% to 35% of annual average b,,a, with
much higher concentrations in the winter than in the summer (U.S. EPA, 2019, section 13.2.4.1).
In the Northwest, POM was the largest contributor to annual average bext, up to 70%, in most
urban and rural regions with the greatest contributions in the fall. This seasonal contribution of
POM may be related to wildfires. A few exceptions included Boise and sites in North Dakota,
where ammonium nitrate was the greatest contributor, and sites in the Alaska IMPROVE region,
where ammonium sulfate was the greatest contributor (U.S. EPA, 2019, section 13.2.4.1). In the
Southwest, based on IMPROVE data, ammonium sulfate or POM were generally the greatest
contributors to annual average bext, with nearly equivalent contributions in several regions. Based
on CSN data, ammonium nitrate was often the greatest contributor, with especially high bext
contributions in the winter. While PM10-2.5 mass scattering was relatively small in the eastern and
11	Monitors were grouped into 28 IMPROVE regions and 31 CSN regions based on site location and PM
concentrations for major species. For comparison purposes, and where possible, CSN regions were defined
similarly to those forthe IMPROVE network (Hand et al., 2011; U.S. EPA, 2019, section 13.2.4.1).
12	A bext value of 40 Mm1 corresponds to a visual range of about 100 km.
5-14

-------
northwestern U.S., in the Southwest, PM10-2.5 mass scattering contributed to more than 20% of
light extinction (U.S. EPA, 2019, section 13.2.4.1).
Differences also exist between the urban CSN and the mainly rural IMPROVE data.
Light extinction is generally higher in CSN regions than the geographically corresponding
IMPROVE regions. Annual average be%t was greater than 50 Mm"1 in 11 CSN regions, compared
to only one IMPROVE region, and was greater than 20 Mm"1 in all CSN regions, compared to
just over half of the IMPROVE regions. Light absorbing carbon was the greatest contributor to
light extinction in several Western CSN regions but was not a large contributor in any of the
IMPROVE regions (U.S. EPA, 2019, Figure 13-11). Ammonium nitrate also accounted for more
light extinction in the CSN regions, while it was only a top contributor to be%t in one IMPROVE
region (U.S. EPA, 2019, section 13.2.4.1).
From the 2005-2008 time period to the 2011-2014 time period, the annual average be%t in
most CSN regions in the Eastern U.S. decreased by more than 20 Mm"1. This corresponds to an
improvement in average visual range in most Eastern U.S. regions of more than 6 Mm"1 (or 15
km) from 2005-2008 to 2011-2014. Additionally, the contribution of ammonium sulfate to light
extinction has also changed over this period. Due to decreased atmospheric sulfate
concentrations, the impact on visibility impairment is evident with a smaller fraction of the total
bext accounted for by ammonium sulfate in 2011-2014 compared to 2005-2008 (U.S. EPA, 2019,
section 13.2.4.1).
In summary, the spatial and temporal analysis of PM monitoring network data in the ISA
emphasizes that the extent of light extinction by PM2.5 depends on PM2.5 composition and
relative humidity. Regional differences in PM2.5 composition greatly influence light extinction.
Changes in PM2.5 composition over time can also affect light extinction based on concentrations
of specific PM components in ambient air.
• To what extent are new studies available that might inform judgments about the
potential adversity to public welfare of PM-attributable visibility impairment and
the nature of the relationship between PM-attributable visibility impairment and
public perceptions of such impairment?
In the last review, visibility preference studies were available from four areas in North
America,13 as described in section 5.1.1 above. Study participants were queried regarding
multiple images that, depending on the study, were either photographs of the same location and
scenery that had been taken on different days on which measured extinction data were available
13 As noted above, preference studies were available in four urban areas in the last review: Denver, Colorado (Ely et
al., 1991, Pryor, 1996), Vancouver, British Columbia, Canada (Pryor, 1996), Phoenix, Arizona (BBC Research &
Consulting, 2003), and Washington, DC (Abt Associates, 2001; Smith and Howell, 2009). More details about
these studies are available in Appendix D.
5-15

-------
or digitized photographs onto which a uniform "haze" had been superimposed. Results of these
studies indicated a wide range of judgments on what study participants considered to be
acceptable visibility across the different study areas, depending on the setting depicted in each
photograph. As a part of the 2010 UFVA, each study was evaluated separately, and figures were
developed to display the percentage of participants that rated the visual air quality depicted as
"acceptable" (U.S. EPA, 2010). Figure 5-2 represents a graphical summary of the results of the
studies in the four cities and identifies a range encompassing the PMzs visibility index values
from images that were judged to be acceptable by at least 50% of study participants across all
four of the urban preference studies (U.S. EPA, 2010, p. 4-24).14 As shown in Figure 5-2, much
lower visibility (considerably more haze resulting in higher values of light extinction) was
considered acceptable in Washington, D.C. than was in Denver. The median judgment for the
study groups in the two areas differed by 9.2 dv (which roughly corresponds to about 30 ug/nv'
of PM) (U.S. EPA, 2010).
100
_Q
CO
Q.
CD
a
0
<
o>
1	50
DC

"c
TO
Q.
'u
ro
a.
ss
0
5	10	15	20	25	30	35	40	45
PM2 5 Visibility Index (dv)
a Denver
~ Phoenix
~ BC
~ Washington
	Denver Logit
	Phoenix Logit
	BC Logit
	DC Logit
Figure 5-2. Relationship of viewer acceptability ratings to light extinction. (Source: U.S.
EPA, 2011, Figure 4-2; U.S. EPA, 2010, Figure 2-16)
14 Figure 5-2 shows the results of a logistical regression analysis using a logit model of the acceptable or
unacceptable ratings from participants of the studies. The logit model is a generalized linear model used for
binomial regression analysis which fits explanatory data about binary outcomes (in this case, a person rating an
image as acceptable or unacceptable) to a logistic function curve. A detailed description is available in Appendix
J of the 2010 UFVA (U.S. EPA 2010).
Light Extinction (Mm1)
20	50	100	200	400	800
~ ^
a \ \
A
—•	~	
~ :
t v ~
V \
\ \ ~
\ * \
\ \ \ * '
\ ~\ \
\ \ \ ! *
~
1 1
X \\ *
\ * \ * \ ~
\ \ *\ < \
\ ~ \ \ ~ ! \
V* W \
4. \ ~ \ \ ~ \
\ V ~\\ \
; ~ ~ ~ ~ t ~
r——	ft	
5-16

-------
Since the time of the last review, no new visibility preference studies have been
conducted in the U.S. Outside of the U.S., a visibility preference study was carried out in
Beijing, China (Fajardo et al., 2013). This study found a higher range of acceptable visibility
impairment among participants than was found in preference studies previously conducted in the
U.S. This finding may be related to the common occurrence of higher PM2.5 concentrations in
Beijing (with associated visibility impairment) than is typical in the U.S. (U.S. EPA, 2019,
section 13.2.5). Similarly, there is little newly available information regarding acceptable levels
of visibility impairment in the U.S.
•	To what extent have important uncertainties in the evidence from the last review
been addressed, and have new uncertainties emerged?
While some refinements have been made to the IMPROVE equation to better estimate
light extinction since the last review, there has been no expansion of monitoring efforts for direct
measurement of light extinction. At the time of the last review, it was noted that a PM2.5 light
extinction monitoring program could help with characterizing visibility conditions and the
relationships between PM component concentrations and light extinction.
Little to no new research is available that helps to expand our understanding of visibility
preferences or our characterization of visibility conditions. Uncertainties and limitations
consistent with those identified in the last review persist in this review.
•	Given the potential for people to have different preferences based on the visibility they are
used to based on conditions that they commonly encounter, and the potential for them to
also have different preferences for different types of scenes, the currently available
preference studies may not capture the range of preferences of people in the U.S.
•	The available preference studies were conducted 15 to 30 years ago and may not reflect
the visibility preferences of the U.S. population today. Given that air quality has
improved over the last several decades, the available studies may not reflect current
preferences of people in the U.S.
•	The available preference studies have used different methods to evaluate what level of
visibility impairment is acceptable. Variability in study methodology may influence an
individual's response as to what level of visibility impairment is deemed acceptable, and
thereby influence the results of the study.
•	Many factors that are not captured by the methods used in the currently available
preference studies may influence people's judgments on acceptable visibility. For
example, an individual's perception of an acceptable level of visibility impairment could
be influenced by the duration of visibility impairment experienced, the time of day during
which light extinction is greatest, and the frequency of episodes of visibility impairment,
as well as the intensity of the visibility impairment (i.e., the focus of the available
studies).
5-17

-------
Overall, the body of evidence regarding visibility effects remains largely unchanged since
the time of the last review. While one new study provides refinements to the methods for
estimating light extinction, uncertainties and limitations in the scientific evidence during the last
review remain.
5.2.1.2 Quantitative Assessment-Based Considerations
Beyond our consideration of the scientific evidence, discussed in section 5.2.1.1 above,
we have also considered quantitative analyses of PM air quality and visibility impairment with
regard to the extent they could inform conclusions on the adequacy of the public welfare
protection provided by the current secondary PM standards. In the last review, quantitative
analyses focused on daily visibility impairment, given the short-term nature of PM-related
visibility effects. Such quantitative analyses conducted as part of the last review informed the
decision on the secondary standards in that review (U.S. EPA, 2010, U.S. EPA, 2011; 78 FR
3189-3192, January 15, 2013). The information newly available in this review includes an
updated equation for estimating light extinction, summarized in section 5.2.1.1 above, as well as
more recent air monitoring data, that together allow for development of an updated assessment
with the potential to substantially add to our understanding of PM-related visibility impairment.
Thus, we have conducted updated analyses for this review based on the currently available
technical information, tools, and methods.
• How much visibility impairment is estimated to occur in areas that meet the current
secondary PM standards? What are the factors contributing to the estimates in areas
with higher values?
Consistent with the analyses conducted in the last review, we have conducted analyses
examining the relationship between PM mass concentrations and calculated light extinction
using the 3-year design values15 for the current secondary standards and a 3-year average
visibility metric based on light extinction estimated using IMPROVE equations.16 These analyses
are intended to inform our understanding of visibility impairment in the U.S. under recent air
quality conditions, particularly those conditions that meet the current standards, and the relative
influence of various factors on light extinction. Given the relationship of visibility with short-
term PM, we focus particularly on the short-term PM standards.
15	A design value is a statistic that summarizes the air quality data for a given area in terms of the indicator,
averaging time, and form of the standard. Design values can be compared to the level of the standard and are
typically used to designate areas as meeting or not meeting the standard and assess progress towards meeting the
NAAQS.
16	This is the 3-year visibility metric that was used to evaluate visibility impairment in the last review. Given that
there has been almost no new research since the time of the last review to better inform our understanding of
visibility preferences in the U.S., there is no new information available to inform selection of a visibility metric
for evaluating visibility impairment in the current review different from the one identified in the last review.
5-18

-------
Given that visibility-related effects are often associated with short-term PM
concentrations, and recognizing the relatively larger role of PM2.5 and its components in light
extinction and as inputs to the IMPROVE equation, we have given somewhat more attention to
consideration of the 24-hour PM2.5 standard. Analyses were conducted using three versions of
the IMPROVE equation (Equations D-l through D-3 in Appendix D) to estimate light extinction
to better understand the influence of variability in inputs across the three equations. This analysis
included 67 monitoring sites that are geographically distributed across the U.S. in both urban and
rural areas (see Figure D-l in Appendix D). These sites are those that have a valid 24-hour PM2.5
design value for the 2015-2017 period and met strict criteria for PM species for this analysis.17
We first present results for these 67 sites using the original IMPROVE equation, with
modifications to the equation consistent with those made in evaluating light extinction in the last
review (described in detail in section D.l of Appendix D). We then present results for these 67
sites with light extinction calculated using the Lowenthal and Kumar (2016) IMPROVE equation
described in section 5.2.1.1 above. For a subset of 20 of the 67 monitoring sites where PM10 data
were available and met completeness criteria for this analysis, we then present results of a second
analysis that included the coarse fraction as an input to the IMPROVE equations for calculating
light extinction to better characterize the influence of coarse PM on light extinction.
In considering the relationship between the 24-hour PM2.5 mass-based design value and
the 3-year visibility metric using recent air quality data, we first examine the relationship using
the original IMPROVE equation, consistent with the methods used in the last review (Kelly et
al., 2012b; 78 FR 3201, January 15, 2013; Appendix D). In those areas that meet the current 24-
hour PM2.5 standard, all sites have light extinction estimates at or below 27 dv (Figure 5-3; 78 FR
3218, January 15, 2013). This is also true for the one location that exceeds the current 24-hour
PM2.5 standard (Figure 5-3). These findings are consistent with the findings of the analysis in the
last review that used the same IMPROVE equation with data from 102 sites with data from
2008-2010. This indicates similar findings from this analysis as was the case with the similar
analysis in the last review, i.e., the updated quantitative analysis shows that the 3-year visibility
metric was no higher than 30 dv18 at sites meeting the current secondary PM standards, and at
17	For this analysis, completeness criteria for speciated PM data at these sites included having all 12 quarters in the
2015-2017 period with at least 11 days in each quarter with a valid PM2.5 mass, sulfate, nitrate, organic carbon,
elemental carbon, sea salt (chlorine or chloride), and fine soil (aluminum, silica, calcium, iron, and titanium)
measurement.
18	For comparison purposes in these air quality analyses, we use a 3 -year visibility metric with a level of 30 dv,
which is the highest level of visibility impairment judged to be acceptable by at least 50 percent of the
participants in the preference studies that were available at the time of the last review (78 FR 3191, January 15,
2013).
5-19

-------
most such sites the 3-year visibility index values are much lower (e.g., an average of 20 dv
across the 67 sites).
.2 ^
u
c
40
35
o>
'¦>->	>
x	~
^	-	to
=	•
T3	(V)
°	S;
CD	>
—	O
4->
C	T3
0)	
O	*
cn
30 -
25
20
15
10
0
•	Northeast (n = 19)
•	Southeast (n=9)
•	IndustMidwest (n = 13)
•	UpperMidwest (n = 10)
•	Southwest (n=4)
•	Northwest (n = 7)
•	SoCal (n=4)
•	Alaska (n = l)
0 5 10 15 20 25 30 35 40 45 50 55 60
98th percentile of daily PM2 5 concentration,
averaged over 3 years (pg m 3)
Figure 5-3. Comparison of 90th percentile of daily light extinction, averaged over three
years, and 98th percentile of daily PM2.5 concentrations, averaged over three years, for
2015-2017 using the original IMPROVE equation. (Note: Dashed lines indicate the level
of current 24-hour PM2.5 standard (35 |ig/m3) and the target level of protection identified for
the 3-year visibility metric (30 dv).)
When light extinction was calculated using the refined equation from Lowenthal and
Kumar (2016), the resulting 3-year visibility metrics are slightly higher at all sites compared to
light extinction estimates calculated using the original IMPROVE equation (Figure 5-4). As
noted in section 5.2.1.1, this version of the IMPROVE equation uses a multiplier of 2.1 to
convert the measured OC to OM for input into the equation and also accounts for water
absorption by water soluble organic matter as a function of relative humidity, likely contributing
to the slightly higher estimates of light extinction. As noted in section 5.2.1.1, the Lowenthal and
Kumar (2016) refinements to the IMPROVE equation are based on evaluations of monitoring
data from remote IMPROVE sites. More remote areas tend to have more aged organic particles
than urban areas, and these adjustments to the IMPROVE equation account for the higher
concentration of organic matter as a result of more aged organic particles at these sites. It is
5-20

-------
important to note that, since the Lowenthal and Kumar (2016) refinements to the IMPROVE
equation likely result in one of the higher estimates of light extinction, this equation may
overestimate light extinction in non-remote areas, including those urban areas in our analyses.
For those sites that meet the current 24-hour PM2.5 standard, the 3-year visibility metric is
at or below 30 dv when light extinction is calculated using the Lowenthal and Kumar (2016)
equation, with the exception of one site in Fairbanks, Alaska. This site just meets the current 24-
hour PM2.5 standard and has a 3-year visibility index value of 31 dv (compared to 27 dv when
light extinction is calculated with the original IMPROVE equation) (see Table D-3 in Appendix
D). The conditions at this site, however, may differ considerably from those under which the
Lowenthal and Kumar (2016) IMPROVE equation, with 2.1 as the multiplier to estimate OM
from OC, has been evaluated. Some of these differences, which include higher OC
concentrations, with OC as a much higher fraction of OM, much lower temperatures, and the
complete lack of sunlight for long periods, may affect the quantitative relationships of OC and
OM with visibility (e.g., Hand et al., 2012; Hand et al., 2013).
40
0 ^
u
c
0)
'-4-J >
X ~
Jc 3
(/>
—	1_
-
T3 (V)
O Si
CD >
—	o
4->
C	T3
OJ	CD
U	CD
to
O)
Q.
-C
o
cn
35
30 -
25
20
15
10
0
• •
• • •
• «M •
• • •
• • •
• • •
• • •
M •
• •
• • •
•	Northeast (n = 19)
•	Southeast (n=9)
•	IndustMidwest (n = 13)
•	UpperMidwest (n = 10)
•	Southwest (n=4)
•	Northwest (n = 7)
•	SoCal (n=4)
•	Alaska (n = l)
0 5 10 15 20 25 30 35 40 45 50 55 60
98th percentile of daily PM2 5 concentration,
averaged over 3 years (pg nrf3)
Figure 5-4. Comparison of 90th percentile of daily light extinction, averaged over three
years, and 98th percentile of daily PM2.5 concentrations, averaged over three years, for
2015-2017 using the Lowenthal and Kumar equation. (Note: Dashed lines indicate the
level of current 24-hour PM2.5 standard (35 |ig/m3) and the target level of protection
identified for the 3-year visibility metric (30 dv).)
5-21

-------
In considering visibility impairment under recent air quality conditions, we recognize that
the differences in the inputs to equations estimating light extinction can influence the resulting
values. For example, given the varying chemical composition of emissions from different
sources, the 2.1 multiplier in the Lowenthal and Kumar (2016) equation may not be appropriate
for all source types. At the time of the last review, the EPA judged that a 1.6 multiplier for
converting OC to OM was more appropriate, for the purposes of estimating visibility index at
sites across the U.S., than the 1.4 or 1.8 multipliers used in the original and revised IMPROVE
equations, respectively. A multiplier of 1.8 or 2.1 would account for the more aged and
oxygenated organic PM that tends to be found in more remote regions than in urban regions,
whereas a multiplier of 1.4 may underestimate the contribution of organic PM found in remote
regions when estimating light extinction (78 FR 3206, January 15, 2013; U.S. EPA, 2012b, p.
IV-5). The information and analyses available in the current review indicate that it may be
appropriate to select inputs to the IMPROVE equation (e.g., the multiplier for OC to OM) on a
regional basis rather than a national basis when calculating light extinction. This is especially
true when comparing sites with localized PM sources (such as sites in urban or industrial areas)
to sites with PM derived largely from biogenic precursor emissions (that contribute to
widespread secondary organic aerosol formation), such as those in the southeastern U.S. We
note, however, that conditions involving PM from such different sources have not been well
studied in the context of applying a multiplier to estimate light extinction, contributing
uncertainty to estimates of light extinction for such conditions.
At the time of the last review, the EPA noted that PM2.5 is the size fraction of PM
responsible for most of the visibility impairment in urban areas (77 FR 38980, June 29, 2012).
Data available at the time of the last review suggested that, generally, PM10-2.5 was a minor
contributor to visibility impairment most of the time (U.S. EPA, 2010) although the coarse
fraction may be a major contributor in some areas in the desert southwestern region of the U.S.
Moreover, at the time of the last review, there were few data available from PM10-2.5 monitors to
quantify the contribution of coarse PM to calculated light extinction. Since that time, an
expansion in PM10-2.5 monitoring efforts has increased the availability of data for use in
estimating light extinction with both PM2.5 and PM10-2.5 concentrations included as inputs in the
equations. Collocated PM10-2.5 monitoring data were available at 20 of the 67 PM2.5 sites (see
Appendix D) for 2015-2017. Thus, the analysis in this review addressed light extinction
estimated with coarse and fine PM at sites where feasible. All 20 of these sites met the 24-hour
PM2.5 standard and 24-hour PM10 standard, and they all had 3-year visibility metrics at or below
30 dv when light extinction was calculated with and without the coarse fraction for any of the
three versions of the IMPROVE equation. Generally, the contribution of the coarse fraction to
5-22

-------
light extinction at these sites is minimal, contributing less than 1 dv to the 3-year visibility
metric. However, we note that in our analysis, none of the locations included areas that would be
expected to have greater concentrations of coarse PM, such as the southwest. In such locations, if
PMio and PM10-2.5 data were available, the coarse fraction may be a more important contributor
to light extinction and visibility impairment than in those locations with lower concentrations of
coarse PM. These results are consistent with those in the analyses in the ISA, which found that
mass scattering from PM10-2.5 was relatively small (less than 10%) in the eastern and
northwestern U.S., whereas mass scattering was much larger in the Southwest (more than 20%)
particularly in southern Arizona and New Mexico (U.S. EPA, 2019, section 13.2.4.1, p. 13-36).
In summary, the findings of these updated quantitative analyses are generally consistent
with those in the last review. The 3-year visibility metric was generally below 25 dv in most
areas that meet the current 24-hour PM2.5 standard, with one location slightly above 30 dv,
rounding to 31 dv. Small differences in the 3-year visibility metric were observed between the
variations of the IMPROVE equation, which may suggest that it may be more appropriate to use
one version over another in different regions of the U.S. based on PM characteristics such as
particle size and composition to more accurately estimate light extinction. There was also very
little difference in estimates of light extinction when the coarse fraction was included in the
equation, although this may be more important in areas that have a higher concentration of
coarse PM than those included in this analysis.
5.2.2 Non- Visibility Effects
5.2.2.1 Evidence-Based Considerations
In considering the available evidence for non-visibility welfare effects attributable to PM
as presented in the ISA, this section poses the following policy-relevant questions:
•	To what extent has new scientific evidence improved our understanding of the
nature and magnitude of non-visibility welfare effects of PM in ambient air,
including the variability associated with such effects? To what extent have important
uncertainties in the evidence from the last review been addressed, and have new
uncertainties emerged?
We address these questions for PM and climate effects (section 5.2.2.1.1) and materials
effects (section 5.2.2.1.2) below.
5.2.2.1.1 Climate Effects
In considering the available evidence of climate effects attributable to PM, this section
poses the following policy-relevant question:
•	To what extent is new information available that changes or enhances our
understanding of the climate impacts of PM-related aerosols, particularly regarding
5-23

-------
a quantitative relationship between PM concentrations and effects on climate (e.g.,
through radiative forcing)?
In the last review, the 2009 PM ISA concluded that there was "sufficient evidence to
determine a causal relationship between PM and climate effects - specifically on the radiative
forcing of the climate system, including both direct effects of PM on radiative forcing and
indirect effects that involve cloud feedbacks that influence precipitation formation and cloud
lifetimes" (U.S. EPA, 2009, section 9.3.10).19 Since the last review, climate impacts have been
extensively studied and the ISA concludes that "overall the evidence is sufficient to conclude
that a causal relationship exists between PM and climate effects" (U.S. EPA, 2019, section
13.3.9). Recent research reinforces and strengthens the evidence evaluated in the 2009 ISA. New
evidence provides greater specificity about the details of these radiative forcing effects and
increased understanding of additional climate impacts driven by PM radiative effects. The
Intergovernmental Panel on Climate Change (IPCC) assesses the role of anthropogenic activity
in past and future climate change. In the last review, the 2009 ISA relied heavily on the Fourth
IPCC Assessment Report (AR4); since that time the IPCC has issued an updated report. The
Fifth IPCC Assessment Report (AR5; IPCC, 2013) reports on the key scientific advances in
understanding the climate effects of PM since AR4. The ISA draws substantially upon AR5 in
summarizing these effects.
Atmospheric PM has the potential to affect climate in multiple ways, including absorbing
and scattering of incoming solar radiation, alterations in terrestrial radiation, effects on the
hydrological cycle, and changes in cloud properties (U.S. EPA, 2019, section 13.3.1).
Atmospheric PM interacts with incoming solar radiation. Many species of PM (e.g., sulfate and
nitrate) efficiently scatter solar energy. By enhancing reflection of solar energy back to space,
scattering PM exerts a cooling effect on the surface below. Certain species of PM such as black
carbon (BC), brown carbon (BrC), or dust can also absorb incoming sunlight. A recent study
found that whether absorbing PM warms or cools the underlying surface depends on several
factors, including the altitude of the PM layer relative to cloud cover and the albedo of the
surface (Ban-Weiss et al., 2014). PM also perturbs incoming solar energy by influencing cloud
cover and cloud lifetime. For example, PM provides nuclei upon which water vapor condenses,
forming cloud droplets. Finally, absorbing PM deposited on snow and ice can diminish surface
albedo and lead to regional warming (U.S. EPA, 2019, section 13.3.2).
19 Radiative forcing (RF) for a given atmospheric constituent is defined as the perturbation in net radiative flux, at
the tropopause (or the top of the atmosphere) caused by that constituent, in watts per square meter (Wm~2), after
allowing for temperatures in the stratosphere to adjust to the perturbation but holding all other climate responses
constant, including surface and tropospheric temperatures (Fiore et al., 2015, Myhre et al., 2013). A positive
forcing indicates net energy trapped in the Earth system and suggests warming of the Earth's surface, whereas a
negative forcing indicates net loss of energy and suggests cooling (U.S. EPA, 2019, section 13.3.2.2).
5-24

-------
PM has direct and indirect effects on climate processes. PM interactions with solar
radiation through scattering and absorption, collectively referred to as aerosol-radiation
interactions (ARI), are also known as the direct effects of PM on climate, as opposed to the
indirect effects that involve aerosol-cloud interactions (ACI). The direct effects of PM on climate
result primarily from particles scattering light away from Earth and sending a fraction of solar
energy back into space, decreasing the transmission of visible radiation to the surface of the
Earth and resulting in a decrease in the heating rate of the surface and the lower atmosphere. The
IPCC AR5, taking into account both model simulations and satellite observations, reports a
radiative forcing from aerosol-radiation interactions (RFari) from anthropogenic PM of -0.35 ±
0.5 watts per square meter (Wm"2) (Boucher, 2013), which is slightly reduced compared to AR4.
Estimates of effective radiative forcing20 from aero sol-radiation interactions (ERFari), which
include the rapid feedback effects of temperature and cloud cover, rely mainly on model
simulations, as this forcing is complex and difficult to observe (U.S. EPA, 2019, section
13.3.4.1).	The IPCC AR5 best estimate for ERFari is -0.45 ± 0.5 Wm"2, which reflects this
uncertainty (Boucher, 2013).
By providing cloud condensation nuclei, PM increases cloud droplet number, thereby
increasing cloud droplet surface area and albedo (Twomey, 1977). The climate effects of these
perturbations are more difficult to quantify than the direct effects of aerosols with RF but likely
enhance the cooling influence of clouds by increasing cloud reflectivity (traditionally referred to
as the first indirect effect) and lengthening cloud lifetime (the second indirect effect). These
effects are reported as the radiative forcing from aerosol-cloud interactions (RFaci) and the
effective radiative forcing from aerosol-cloud interactions (ERFaci) (U.S. EPA, 2019, section
13.3.3.2).	IPCC AR5 estimates ERFaci at -0.45 Wm"2, with a 90% confidence interval of-1.2 to
0 Wm"2 (U.S. EPA, 2019, section 13.3.4.2).21 Studies have also calculated the combined
effective radiative forcing from aerosol-radiation and aerosol-cloud interactions (ERFari+aci)
(U.S. EPA, 2019, section 13.3.4.3). IPCC AR5 reports a best estimate of ERFari+aci of -0.90 (-
1.9 to -0.1) Wm"2, consistent with these estimates (Boucher, 2013).
PM can also strongly reflect incoming solar radiation in areas of high albedo, such as
snow- and ice-covered surfaces. The transport and subsequent deposition of absorbing PM such
as BC to snow- and ice-covered regions can decrease the local surface albedo, leading to surface
20	Effective radiative forcing (ERF), new in the IPCC AR5, takes into account not just the instantaneous forcing but
also a set of climate feedbacks, involving atmospheric temperature, cloud cover, and water vapor, that occur
naturally in response to the initial radiative perturbation (U.S. EPA, 2019, section 13.3.2.2).
21	While the ISA includes estimates of RFaci and ERFaci from a number of studies (U.S. EPA, 2019, sections
13.3.4.2, 13.3.4.3, 13.3.3.3), this PA focuses on the single best estimate with a range of uncertainty, as reported in
IPCC AR5 (Boucher, 2013).
5-25

-------
heating. The absorbed energy can then melt the snow and ice cover and further depress the
albedo, resulting in a positive feedback loop (U.S. EPA, 2019, section 13.3.3.3; Bond et al.,
2013; U.S. EPA, 2012a). Deposition of absorbing PM, such as BC, may also affect surface
temperatures over glacial regions (U.S. EPA, 2019, section 13.3.3.3). The IPCC AR5 best
estimate of RF from the albedo effect is +0.04 Wm'2, with an uncertainty range of +0.02 to +0.09
Wm"2 (Boucher, 2013).
While research on PM-related effects on climate has expanded since the last review, there
are still significant uncertainties associated with the accurate measurement of PM contributions
to the direct and indirect effects of PM on climate.
• To what extent does the currently available information provide evidence of a
quantitative relationship between specific PM constituents (i.e., BC, OC, sulfate) and
climate-related effects?
Since the last review, a number of new studies have examined the individual climate
effects associated with key PM components, including sulfate, nitrate, OC, BC, and dust, along
with updated quantitative estimates of the radiative forcing associated with the individual
species.
Sulfate particles form through oxidation of SO2 by OH in the gas phase and in the
aqueous phase by a number of pathways, including in particular those involving ozone and H2O2
(U.S. EPA, 2019, section 13.3.5.1). The main source of anthropogenic sulfate is from coal-fired
power plants, and global trends in the anthropogenic SO2 emissions are estimated to have
increased dramatically during the 20th and early 21st centuries, although the recent
implementation of more stringent air pollution controls on sources has led to a reversal in such
trends in many places (U.S. EPA, 2019, section 13.3.5.1). Sulfate particles are highly reflective.
Consistent with other recent estimates, on a global scale, the IPCC AR5 estimates that sulfate
contributes more than other PM types to RF, with RFari of -0.4 (-0.6 to -0.2) Wm"2, where the
5% and 95% uncertainty range is represented by the numbers in the parentheses (Myhre et al.,
2013). This uncertainty range indicates the challenges associated with estimating SO2 from
sources in developing regions and estimating the lifetime of sulfate against wet deposition.
Sulfate is also a major contributor to the influence of PM on clouds (Takemura, 2012). A total
effective radiative forcing (ERFari+aci) for anthropogenic sulfate has been estimated to be nearly
-1.0 Wm"2 (Zelinka et al., 2014, Adams et al., 2001).
Nitrate particles form through the oxidation of nitrogen oxides and occur mainly in the
form of ammonium nitrate. Ammonium preferentially associates with sulfate rather than nitrate,
leading to formation of ammonium sulfate at the expense of ammonium nitrate (Adams et al.,
2001). As anthropogenic emissions of SO2 decline, more ammonium will be available to react
with nitrate, potentially leading to future increases in ammonium nitrate particles in the
5-26

-------
atmosphere (U.S. EPA, 2019, section 13.3.5.2; Hauglustaine et al., 2014; Lee et al., 2013;
Shindell et al., 2013). Warmer global temperatures, however, may decrease nitrate abundance
given that it is highly volatile at higher temperatures (Tai et al., 2010). The IPCC AR5 estimates
RFari of nitrate of -0.11 (-0.3 to -0.03) Wm"2 (Boucher, 2013), which is one-fourth of the RFari
of sulfate.
Primary organic carbonaceous PM, including BrC, are emitted from wildfires,
agricultural fires, and fossil fuel and biofuel combustion. Secondary organic aerosols (SOA)
form when anthropogenic or biogenic nonmethane hydrocarbons are oxidized in the atmosphere,
leading to less volatile products that may partition into PM (U.S. EPA, 2019, section 13.3.5.3).
Organic particles are generally reflective, but in the case of BrC, a portion is significantly
absorbing at shorter wavelengths (<400 nm). The IPCC AR5 estimates an RFari for primary
organic PM from fossil fuel combustion and biofuel use of -0.09 (-0.16 to -0.03) Wm"2 and an
RFari estimate for SOA from these sources of -0.03 (-0.27 to +0.20) Wm"2 (Myhre et al., 2013).
The wide range in these estimates, including inconsistent signs for forcing, reflect uncertainties
in the optical properties of organic PM and its atmospheric budgets, including the production
pathways of anthropogenic SOA (Scott et al., 2014; Myhre et al., 2013; McNeill et al., 2012;
Heald et al., 2010). The IPCC AR5 also estimates an RFari of -0.2 Wm"2 for primary organic PM
arising from biomass burning (Boucher, 2013).
Black carbon (BC) particles occur as a result of inefficient combustion of carbon-
containing fuels. Like directly emitted organic PM, BC is emitted from biofuel and fossil fuel
combustion and by biomass burning. BC is absorbing at all wavelengths and likely has a large
impact on the Earth's energy budget (Bond et al., 2013). The IPCC AR5 estimates a RFari from
anthropogenic fossil fuel and biofuel use of +0.4 (+0.5 to +0.8) Wm"2 (Myhre et al., 2013).
Biomass burning contributes an additional +0.2 (+0.03 to +0.4) Wm"2 to BC RFari, while the
albedo effect of BC on snow and ice adds another +0.04 (+0.02 to +0.09) Wm"2 (Myhre et al.,
2013; U.S. EPA, 2019, section 13.3.5.4, section 13.3.4.4).
Dust, or mineral dust, is mobilized from dry or disturbed soils as a result of both
meteorological and anthropogenic activities. Dust has traditionally been classified as scattering,
but a recent study found that dust may be substantially coarser than currently represented in
climate models, and thus more light-absorbing (Kok et al., 2017). The IPCC AR5 estimates
RFari as -0.1 ± 0.2 Wm"2 (Boucher, 2013), although the results of the study by Kok et al. (2017)
would suggest that in some regions dust may have led to warming, not cooling (U.S. EPA, 2019,
section 13.3.5.5).
The new research available in this review expands upon the evidence available at the time
of the last review. Consistent with the evidence available in the last review, the key PM
5-27

-------
components, including sulfate, nitrate, OC, BC, and dust, that contribute to climate processes
vary in their reflectivity, forcing efficiencies, and direction of forcing.
•	To what extent does newly available evidence change or improve our understanding
of the spatial and temporal variation in climate responses to PM?
Radiative forcing due to PM elicits a number of responses in the climate system that can
lead to significant effects on weather and climate over a range of spatial and temporal scales,
mediated by a number of feedbacks that link PM and climate. Since the last review, the evidence
base has expanded with respect to the mechanisms of climate responses and feedbacks to PM
radiative forcing, described below, although considerable uncertainties continue to exist. We
focus our discussion primarily on the climate impacts in the U.S.
Unlike well-mixed, long-lived greenhouse gases in the atmosphere, PM has a very
heterogenous distribution across the Earth. As such, patterns of RFari and RFaci tend to correlate
with PM loading, with the greatest forcings centralized over continental regions. The climate
response is more complicated since the perturbation to one climate variable (e.g., temperature,
cloud cover, precipitation) can lead to a cascade of effects on other variables. While the initial
PM radiative forcing may be concentrated regionally, the eventual climate response can be much
broader spatially or be concentrated in remote regions (U.S. EPA, 2019, section 13.3.6). The
complex climate system interactions lead to variation among climate models, with some studies
showing relatively close correlation between forcing and surface response temperatures (e.g.,
Leibensperger et al., 2012), while other studies show much less correlation (e.g., Levy et al.,
2013). Many studies have examined observed trends in PM and temperature in the U.S. Climate
models have suggested a range of factors which can influence large-scale meteorological
processes and may affect temperature, including local feedback effects involving soil moisture
and cloud cover, changes in the hygroscopicity of the PM, and interactions with clouds alone
(U.S. EPA, 2019, section 13.3.7). While evidence in this review suggests that PM influenced
temperature trends across the southern and eastern U.S. in the 20th century, uncertainties
continue to exist and further research is needed to better characterize the effects of PM on
regional climate in the U.S.
•	To what extent have important uncertainties identified in the last review been
reduced and/or have new uncertainties emerged?
Since 2009, significant progress has been made in evaluating PM-related climate effects
and uncertainties. The IPCC AR5 states that "climate-relevant aerosol processes are better
understood, and climate-relevant aerosol properties are better observed, than at the time of the
AR4" (Boucher, 2013). However, significant uncertainties remain that make it difficult to
5-28

-------
quantify the climate effects of PM. Such uncertainties include those related to our understanding
of:
•	The magnitude of PM radiative forcing and the portion of that associated with
anthropogenic emissions;
•	The contribution of regional differences in PM concentrations, and of individual
components, to radiative forcing;
•	The mechanisms of climate responses and feedbacks resulting from PM-related radiative
forcing; and,
•	The process by which PM interacts with clouds and how to represent such interactions in
climate models.
While research has progressed significantly since the last review, substantial uncertainties
still remain with respect to key processes linking PM and climate, because of the small scale of
PM-relevant atmospheric processes compared to the resolution of state-of-the-art models, and
because of the complex cascade of indirect impacts and feedbacks in the climate system that
result from an initial PM-related radiative perturbation (U.S. EPA, 2019, section 13.3.9).
5.2.2.1.2 Materials Effects
In considering the available evidence on materials effects attributable to PM, this section
poses the following policy-relevant question:
•	To what extent is new information available to link PM to materials effects,
including degradation of surfaces, and deterioration of materials such as metal,
stone, concrete and marble?
In the last review, the 2009 ISA concluded that there was "a causal relationship between
PM and effects on materials" (U.S. EPA, 2009, sections 2.5.4 and 9.5.4). Rather than altering our
conclusions from the last review, the current evidence continues to support our prior conclusion
regarding materials effects associated with PM deposition. Effects of deposited PM, particularly
sulfates and nitrates,22 to materials include both physical damage and impaired aesthetic
qualities. Because of their electrolytic, hygroscopic, and acidic properties and their ability to sorb
corrosive gases, particles contribute to materials damage by adding to the effects of natural
weathering processes, by potentially promoting or accelerating the corrosion of metals,
degradation of painted surfaces, deterioration of building materials, and weakening of material
components. The majority of the newly available evidence on materials effects of PM are from
22 In the case of materials effects, it is difficult to isolate the effects of gaseous and particulate N and S wet
deposition so both will be considered along with other PM-related deposition effects on materials in this review
of the PM NAAQS.
5-29

-------
outside the U.S. on buildings and other items of cultural heritage; however, they provide limited
new data for consideration in this review (U.S. EPA, 2019, section 13.4).
Materials damage from PM generally involves one or both of two processes: soiling and
corrosion (U.S. EPA, 2019, section 13.4.2). Soiling and corrosion are complex, interdependent
processes, typically beginning with deposition of atmospheric PM or SO2 to exposed surfaces.
Constituents of deposited PM can interact directly with materials or undergo further chemical
and/or physical transformation to cause soiling, corrosion, and physical damage. Weathering,
including exposure to moisture, ultraviolet (UV) radiation and temperature fluctuations, affects
the rate and degree of damage (U.S. EPA, 2019, section 13.4.2).
Soiling is the result of PM accumulation on an object that alters its optical characteristics
or appearance. These soiling effects can impact the aesthetic value of a structure or result in
reversible or irreversible damage to the surface. The presence of air pollution can increase the
frequency and duration of cleaning and can enhance biodeterioration processes on the surface of
materials. For example, deposition of carbonaceous components of PM can lead to the formation
of black crusts on surfaces, and the buildup of microbial biofilms23 can discolor surfaces by
trapping PM more efficiently (U.S. EPA, 2009, p. 9-195; U.S. EPA, 2019, section 13.4.2). The
presence of PM may alter light transmission or change the reflectivity of a surface. Additionally,
the organic or nutrient content of deposited PM may enhance microbial growth on surfaces.
Since the last review, very little new evidence has become available related to deposition
of SO2 to materials such as limestone, granite, and metal. Deposition of SO2 onto limestone can
transform the limestone into gypsum, resulting in a rougher surface, which allows for increased
surface area for accumulation of deposited PM (Camuffo and Bernardi, 1993; U.S. EPA, 2019,
section 13.4.2). Oxidation of deposited SO2 that contributes to the transformation of limestone to
gypsum can be enhanced by the formation of surface coatings from deposited carbonaceous PM
(both elemental and organic carbon) (McAlister et al., 2008, Grossi et al., 2007). Ozga et al.
(2011) characterized damage to two concrete buildings in Poland and Italy. Gypsum was the
main damage product on surfaces of these buildings that were sheltered from rain runoff, while
PM embedded in the concrete, particularly carbonaceous particles, were responsible for
darkening of the building walls (Ozga et al., 2011).
Building on the evidence available in the 2009 ISA, research has progressed on the
theoretical understanding of soiling of cultural heritage in a number of studies. Barca et al.
(2010) developed and tested a new methodological approach for characterizing trace elements
and heavy metals in black crusts on stone monuments to identify the origin of the chemicals and
23 Microbial biofilms are communities of microorganisms, which may include bacteria, algae, fungi and lichens, that
colonize an inert surface. Microbial biofilms can contribute to biodeterioration of materials via modification of
the chemical environment.
5-30

-------
the relationship between the concentrations of elements in the black crusts and local
environmental conditions. Recent research has also used isotope tracers to distinguish between
contributions from local sources versus atmospheric pollution to black crusts on historical
monuments in France (Kloppmann et al., 2011). A study in Portugal found that biological
activity played a major role in soiling, specifically in the development of colored layers and in
the detachment process (de Oliveira et al., 2011). Another study found damage to cement
renders, often used for restoration, consolidation, and decorative purposes on buildings,
following exposure to sulfuric acid, resulting in the formation of gypsum (Lanzon and Garcia-
Ruiz, 2010).
Corrosion of stone and the decay of stone building materials by acid deposition and
sulfate salts were described in the 2009 ISA (U.S. EPA, 2009, section 9.5.3). Since that time,
advances have been made on the quantification of degradation rates and further characterization
of the factors that influence damage of stone materials (U.S. EPA, 2019, section 13.4.2). Decay
rates of marble grave stones were found to be greater in heavily polluted areas compared to a
relatively pristine area (Mooers et al., 2016). The time of wetness and the number of
dissolution/crystallization cycles were identified as hazard indicators for stone materials, with
greater hazard during the spring and fall when these indicators are relatively high (Casati et al.,
2015).
A study examining the corrosion of steel as a function of PM composition and particle
size found that changes in the composition of resulting rust gradually changed with particle size
(Lau et al., 2008). In a study of damage to metal materials under in Hong Kong, which generally
has much higher PM concentrations than those observed in the U.S., Liu et al. (2015) found that
iron and steel were corroded by both PM and gaseous pollutants (SO2 and NO2), while copper
and copper alloys were mainly corroded by gaseous pollutants (SO2 and O3) and aluminum and
aluminum alloy corrosion was mainly attributed to PM and NO2.
A number of studies have also found materials damage from PM components besides
sulfate and black carbon and atmospheric gases besides SO2. Studies have characterized impacts
of nitrates, NOx, and organic compounds on direct materials damage or on chemical reactions
that enhance materials damage (U.S. EPA, 2019, section 13.4.2). Other studies have found that
soiling of building materials can be attributed to enhanced biological processes and colonization,
including the development and thickening of biofilms, resulting from the deposition of PM
components and atmospheric gases (U.S. EPA, 2019, section 13.4.2).
Since the last review, other materials have been studied for damage attributable to PM,
including glass and photovoltaic panels. Soiling of glass can impact its optical and thermal
5-31

-------
properties, and can lead to increased cleaning costs and frequency. The development of haze24 on
modern glass has been measured and modeled, with a strong correlation between the size
distribution of particles and the evolution of the mass deposited on the surface of the glass.
Measurements showed that, under sheltered conditions, mass deposition accelerated regularly
with time in areas closest to sources of PM (i.e., near roadways) and coarse mineral particles
were more prevalent compared to other sites (Alfaro et al., 2012). Model predictions were found
to correctly simulate the development of haze at site locations when compared with
measurements (Alfaro et al., 2012).
Soiling of photovoltaic panels can lead to decreased energy efficiency. For example,
soiling by carbonaceous PM decreased solar efficiency by nearly 38%, while soil particles
reduced efficiency by almost 70% (Radonjic et al., 2017). The rate of photovoltaic power output
can also be degraded by soiling and has been found to be related to the rate of dust accumulation.
In five sites in the U.S. representing different meteorological and climatological conditions,25
photovoltaic module power transmission was reduced by approximately 3% for every g/m2 of
PM deposited on the cover plate of the photovoltaic panel, independent of geographical location
(Boyle et al., 2017). Another study found that photovoltaic module power output was reduced by
40% after 10 months of exposure without cleaning, although a number of anti-reflective coatings
can generally mitigate power reduction resulting from dust deposition (Walwil et al., 2017).
Energy efficiency can also be impacted by the soiling of building materials, such as light-colored
marble panels on building exteriors, that are used to reflect a large portion of solar radiation for
passive cooling and to counter the urban heat island effect. Exposure to acidic pollutants in urban
environments have been found to reduce the solar reflectance of marble, decreasing the cooling
effect (Rosso et al., 2016). Highly reflective roofs, or cool roofs, have been designed and
constructed to increase reflectance from buildings in urban areas, to both decrease air
conditioning needs and urban heat island effects, but these efforts can be impeded by soiling of
materials used for constructing cool roofs. Methods have been developed for accelerating the
aging process of roofing materials to better characterize the impact of soiling and natural weather
on materials used in constructing cool roofs (Sleiman et al., 2014).
24	In this discussion of non-visibility welfare effects (section 5.2.2), haze is used as it has been defined in the
scientific literature on soiling of glass, i.e., the ratio of diffuse transmitted light to direct transmitted light
(Lombardo et al., 2010). This differs from the definition of haze as used in the discussion of visibility welfare
effects in section 5.2.1, where it is used as a qualitative description of the blockage of sunlight by dust, smoke,
and pollution.
25	Of the five sites studied, three were in rural, suburban, and urban areas representing a semi-arid environment
(Front Range of Colorado), one site represented a hot and humid environment (Cocoa, Florida), and one
represented a hot and arid environment (Albuquerque, New Mexico) (U.S. EPA, 2019, section 13.4.2; Boyle et
al., 2017).
5-32

-------
• To what extent has new information emerged for quantifying material damage
attributable to PM through dose-response relationships or damage functions? Are
there studies linking perceptions of reduced aesthetic appeal of buildings and other
objects to PM or wet deposition of N and S species?
Some progress has been made since the last review in the development of dose-response
relationships for soiling of building materials, although some key relationships remain poorly
characterized. The first general dose-response relationships for soiling of materials were
generated by measuring contrast reflectance of a soiled surface to the reflectance of the unsoiled
substrate for different materials, including acrylic house paint, cedar siding, concrete, brick,
limestone, asphalt shingles, and window glass with varying total suspended particulate (TSP)
concentrations (Beloin and Haynie, 1975; U.S. EPA, 2019, section 13.4.3). Continued efforts to
develop dose-response curves for soiling have led to some advancements for modern materials,
but these relationships remain poorly characterized for limestone. A recent study quantified the
dose-response relationships between PMio and soiling for painted steel, white plastic, and
polycarbonate filter material, but there was too much scatter in the data to produce a dose-
response relationship for limestone (Watt et al., 2008). A dose-response relationship for silica-
soda-lime window glass soiling by PMio, NO2, and SO2 was quantified based on 31 different
locations (Lombardo et al., 2010; U.S. EPA, 2019, section 13.4.3, Figure 13-32, Equation 13-8).
The development of this dose-response relationship required several years of observation time
and had inconsistent data reporting across the locations.
Since the last review, there has also been progress in developing methods to more rapidly
evaluate soiling of different materials by PM mixtures. Modern buildings typically have simpler
lines, less detailed surfaces, and a greater use of glass, tile, and metal, which are easier to clean
than stone. There have also been major changes in the types of materials used for buildings,
including a variety of polymers available for use as coatings and sealants. New economic and
environmental considerations beyond aesthetic appeal and structural damage are emerging (U.S.
EPA, 2019, section 13.4.3). Changes in building materials and design, coupled with new
approaches in quantifying the dose-response relationship between PM and materials effects, may
reduce the amount of time needed for observations to support the development of material-
specific dose-response relationships.
In addition to dose-response functions, damage functions have also been used to quantify
material decay as a function of pollutant type and load. Damage can be determined from sample
surveys or inspection of actual damage and a damage function can be developed to link the rate
of material damage to time of replacement or maintenance. A cost function can then link the time
for replacement and maintenance to a monetary cost, and an economic function links cost to the
dose of pollution based on the dose-response relationship (U.S. EPA, 2019, section 13.4.3).
5-33

-------
Damage functions are difficult to assess because it depends on human perception of the level of
soiling deemed to be acceptable and evidence in this area remains limited in the current review.
Since the last review, damage functions for a wide range of building materials (i.e., stone,
aluminum, zinc, copper, plastic, paint, rubber, stone) have been developed and reviewed
(Brimblecombe and Grossi, 2010). One study estimated long-term deterioration of building
materials and found that damage to durable building material (such as limestone, iron, copper,
and discoloration of stone) is no longer controlled by pollution as was historically documented
but rather that natural weathering is a more important influence on these materials in modern
times (Brimblecombe and Grossi, 2009). Even as PM-attributable damage to stone and metals
has decreased over time, it has been predicted that there will be potentially higher degradation
rates for polymeric materials, plastic, paint, and rubber due to increased oxidant concentrations
and solar radiation (Brimblecombe and Grossi, 2009).
• To what extent have important uncertainties identified in the last review been
reduced and/or have new uncertainties emerged?
While there are a number of new studies in the ISA that investigate the effect of PM on
newly studied materials and further characterize the effects of PM on previously studied
materials, there remains insufficient evidence to relate soiling or damage to specific PM levels or
to establish a quantitative relationship between PM in ambient air and materials degradation.
Uncertainties that were identified in the last review still largely remain with respect to
quantitative relationships between particle size, concentration, chemical concentrations, and
frequency of repainting and repair. No new studies are available that link perceptions of reduced
aesthetic appeal of buildings and other objects to PM-related materials effects. Moreover,
uncertainties about the deposition rates of airborne PM to surfaces and the interaction of co-
pollutants still remain.
5.2.2.2 Quantitative Assessment-Based Considerations
Beyond our consideration of the scientific evidence, discussed above in section 5.2.2.1
above, we also consider the extent to which quantitative analyses of PM air quality and
quantitative assessments for climate and materials effects could inform conclusions on the
adequacy of the public welfare protection provided by the current secondary PM standards. We
have evaluated the potential support for conducting new analyses of PM air quality
concentrations and non-visibility welfare effects.
5.2.2.2.1 Climate Effects
While expanded since the last review, our current understanding of PM-related climate
effects is still limited by significant uncertainties. Large spatial and temporal heterogeneities in
direct and indirect PM climate forcing can occur for a number of reasons, including the
5-34

-------
frequency and distribution of emissions of key PM components contributing to climate forcing,
the chemical and microphysical processing that occurs in the atmosphere, and the atmospheric
lifetime of PM relative to other pollutants contributing to climate forcing (U.S. EPA, 2019,
section 13.3). These issues particularly introduce uncertainty at the local and regional scales in
the U.S. that would likely be most relevant to a quantitative assessment of the potential effects of
a national PM standard on climate in this review. Limitations and uncertainties in the evidence
make it difficult to quantify the impact of PM on climate and in particular how changes in the
level of PM mass in ambient air would result in changes to climate in the U.S. Thus, as in the last
review, the data remain insufficient to conduct quantitative analyses for PM effects on climate in
the current review.
5.2.2.2.2 Materials Effects
As at the time of the last review, sufficient evidence is not available to conduct a
quantitative assessment of PM-related soiling and corrosion effects. While soiling associated
with PM can lead to increased cleaning frequency and repainting of surfaces, no quantitative
relationships have been established between characteristics of PM or the frequency of cleaning
or repainting that would help inform our understanding of the public welfare implications of
soiling (U.S. EPA, 2019, section 13.4). Similarly, while some information is available with
regard to microbial deterioration of surfaces and the contribution of carbonaceous PM to the
formation of black crusts that contribute to soiling, the available evidence does not support
quantitative analyses (U.S. EPA, 2019, section 13.4). While some new evidence is available with
respect to PM-attributable materials effects, the data are insufficient to conduct quantitative
analyses for PM effects on materials in the current review.
5.3 CASAC ADVICE
As part of its review of the draft PA, the CASAC has provided advice on the adequacy of
the current PM secondary standards. In its comments on the draft PA, the CASAC concurs with
staffs overall preliminary conclusions that it is appropriate to consider retaining the current
secondary PM standards without revision (Cox, 2019). The CASAC "finds much of the
information.. .on visibility and materials effects of PM2.5 to be useful, while recognizing that
uncertainties and controversies remain about the best ways to evaluate these effects" (Cox, 2019,
p. 13 of consensus responses). Regarding climate, while the CASAC recommends that the EPA
consider recent research evaluating the impacts of reducing PM2.5 and suggests that the EPA
include quantitative analyses to more thoroughly address these effects,26 the committee also
26 While this final PA does consider research evaluating the impacts of PM on climate, we have not conducted
analyses to quantify the impacts of changes in U.S. ambient PM concentrations on regional and national climate
endpoints in the U.S. that would be of potential relevance for the NAAQS review. This approach to addressing
5-35

-------
agrees with the EPA that "the available evidence does not call into question the protection
afforded by the current secondary PM standards and concurs that they should be retained" (Cox,
2019, p. 3 of letter).
A number of public comments have been received in this review to date, including
comments focused on the draft PA. A limited number of public comment submissions on the
draft PA provide comments related to the adequacy of the secondary standards. Of those who
provide comments on the secondary standards, the majority of commenters support the
preliminary conclusion that it is appropriate to consider retaining the current secondary PM
standards, without revision. These commenters generally cite a lack of newly available evidence
and information that would inform quantitative assessments and consideration of alternate
secondary standards to protect against PM-related effects on visibility, climate, and materials.
One commenter (the Independent PM Review Panel), however, supports revision of the
secondary PM standards to provide additional protection against PM-related visibility effects,
citing inconsistencies between preliminary conclusions in the draft PA to consider retaining the
current secondary PM standards and the currently available scientific evidence regarding public
visibility preferences and indices for evaluating visibility impairment. This commenter also
recognizes the regional heterogeneity in PM2.5 mass and light extinction and that one single level
may not be appropriate in all regions of the country.
5.4 CONCLUSIONS ON THE SECONDARY PM STANDARDS
This section discusses staff conclusions for the Administrator's consideration in judging
the adequacy of the current secondary PM standards. These conclusions are based on
consideration of the assessment and integrative synthesis of evidence presented in the ISA, as
well as our analyses of recent air quality. Further, the staff conclusions have taken into account
advice from the CASAC and public comments on the draft PA and the associated preliminary
staff conclusions. Taking into consideration the responses to specific questions discussed above,
we revisit the overarching policy question for this chapter:
the CASAC's comments on climate reflects our consideration of the timeline for this review as well as the
uncertainties that would be inherent in such analyses and their likely impact on decision making. As discussed
above (section 5.2.2.2.1), limitations in the evidence would result in considerable uncertainty in analyses that
attempt to quantify the impact of changes in ambient PM in the US on climate in the U.S.
5-36

-------
• Does the currently available scientific evidence and quantitative information support
or call into question the adequacy of the protection afforded by the current
secondary PM standards?
As provided in section 109(b)(2) of the CAA, the secondary standard is to "specify a
level of air quality the attainment and maintenance of which in the judgment of the
Administrator.. .is requisite to protect public welfare from any known or anticipated adverse
effects associated with the presence of such air pollutant in the ambient air." Effects on welfare
include, but are not limited to, "effects on soils, water, crops, vegetation, man-made materials,
animals, wildlife, weather, visibility, and climate, damage to and deterioration of property, and
hazards to transportation, as well as effects on economic values and on personal comfort and
well-being" (CAA section 302(h)). The secondary standards are not meant to protect against all
known or anticipated PM-related effects, but rather those that are judged to be adverse to the
public welfare (78 FR 3212, January 15, 2013). Similarly, the extent to which secondary
standards are concluded to provide adequate protection from such effects also depends on
judgments by the Administrator.
Therefore, we recognize that, as is the case in NAAQS reviews in general, the extent to
which the current secondary PM standards are judged to be adequate will depend on a variety of
factors and judgments to be made by the Administrator. Such judgments include those
concerning the extent or severity of welfare effects that may be considered adverse to the public
welfare, and accordingly, what level of protection from such known or anticipated effects may be
judged requisite. In general, the public welfare significance of PM-related effects for different air
quality conditions and in different locations depend upon the type and severity of the effects, as
well as the strength of the underlying information and associated uncertainties. Thus, in the
discussion below, our intention is to focus on such aspects of the currently available evidence
and quantitative analyses.
With regard to visibility, climate, and materials effects of PM, our response to the
question above takes into consideration the discussions that address the specific policy-relevant
questions in prior sections of this chapter (see sections 5.2.1 and 5.2.2) and the approach
described in section 5.1 that builds on the approach from the last review. With respect to the
evidence-based considerations, we note that the currently available evidence, while somewhat
expanded since the last review, does not include evidence of effects at lower concentrations or
other welfare effects of PM than those identified at the time of the last review. There continue to
be significant uncertainties related to quantifying the relationships between PM mass
concentrations in ambient air and welfare effects, including visibility impairment, climate
effects, and materials effects.
5-37

-------
With respect to the visibility effects of PM, the currently available evidence continues to
support a causal relationship. With respect to evidence for visibility effects of PM, we note that
the currently available evidence, while somewhat expanded since the last review, does not
include evidence of effects at lower concentrations than those identified at the time of the last
review. Consistent with the evidence available at the time of the last review, significant
limitations remain in directly measuring light extinction. However, a number of small
refinements have been made to the algorithm commonly used to estimate light extinction (U.S.
EPA, 2019, section 13.2.3.3; section 5.2.1.1 above). Light extinction by PM2.5 is dependent on
PM2.5 composition and relative humidity, which varies regionally, with component contributions
to light extinction also changing over time with changes in emissions, as can be seen in analyses
of recent air quality. We also note that no new research is available on methods of characterizing
visibility or on how visibility is valued by the public, such as visibility preference studies. Thus,
while limited new research has further informed our understanding of the influence of
atmospheric components of PM2.5 on light extinction, the available evidence to inform
consideration of the public welfare implications of PM-related visibility impairment remains
relatively unchanged.
With respect to quantitative-based considerations, analyses using recent air quality and
considering updated and alternative methods for estimating visibility impairment provide results
generally similar to those given a focus in the decision for the last review. We recognize that
conclusions reached regarding visibility in the last review were based primarily on the
quantitative analyses that considered the relationship of estimated visibility impairment (light
extinction) with design values for the secondary 24-hour PM2.5 standard. These analyses
demonstrated that visibility index values were below 30 dv - the value identified as the target
level of protection for visibility-related welfare effects - at all locations that met the daily
standard. In our evaluation in this chapter, we have considered the currently available
information regarding the equations to estimate light extinction and the inputs to the equations
and regarding identification of the target level of protection. With regard to the equations, we
have utilized both the most recently published equations as well as alternatives considered in the
last review in recognition of the uncertainties inherent in the quantitative relationship between
PM and light extinction and the variability in applicability to different locations. Further, we
have considered key coefficients in estimating and adjusting concentrations of specific PM2.5
components, a key example of which is the multiplier used to estimate the concentration of
organic matter from the concentration of organic carbon. For consistency with the analysis on
which the decision was based in the last review, we have focused on a 3-year average of the 90th
percentile of daily light extinction (calculated using old and new algorithms) in considering
visibility impairment at the analyzed locations.
5-38

-------
In reaching a conclusion in the 2012 review with regard to the adequacy of visibility
protection provided by the secondary PM standards, the Administrator identified 30 dv as an
appropriate target level of protection. We have not identified new information in this review that
would challenge this public policy. Thus, in our consideration of the current information and
analyses in this document, we have compared the results of the updated analyses to the value of
30 dv, finding only one site that exceeds this target level of protection while meeting the current
daily standards, albeit just marginally at 31 dv. In so finding, we additionally note the
uncertainties recognized above regarding estimation of OM for use in the IMPROVE equations,
and also the variability across sites in characteristics that affect the relationship between PM in
ambient air and light extinction, and in characteristics that affect human visibility and
preferences in that regard. Based on the findings of this comparison, in light of all of these
considerations, we find it reasonable to conclude that the quantitative information available in
this review does not call into question the adequacy of visibility-related public welfare protection
provided by the current secondary PM standards. As a result, we have not conducted additional
analyses to evaluate the level of visibility protection that might be afforded by potential
alternative standards.
With respect to the non-visibility welfare effects of PM, the currently available evidence
continues to support causal relationships between climate effects and PM and materials effects
and PM. The currently available evidence related to climate effects and PM, while expanded
since the last review, has not appreciably improved our understanding of the spatial and temporal
heterogeneity of PM components that contribute to climate forcing. We note that, as at the time
of the last review, the evidence describes differences among individual PM components in their
reflective properties and direction of climate forcing. We also note that, while climate research
has continued, there are still significant limitations in our ability to quantify contributions of PM,
and of individual PM components, to the direct and indirect effects of PM on climate (e.g.
changes to the pattern of rainfall, changes to wind patterns, effects on vertical mixing in the
atmosphere). While climate models have been improved and refined since the last review,
climate models simulating aerosol-climate interactions on regional scales (e.g., -100 km) tend to
have more variability in estimates of the PM-related climate effects than simulations at the global
scale, and fewer studies are available that simulate specific regions (e.g., the U.S.) than that
provide global-scale simulations. While new research has added to the understanding of climate
forcing on a global scale, there remain significant limitations to quantifying potential adverse
effects from PM on climate in the U.S. and how they would vary in response to changes in PM
concentrations in the U.S. That is, the information currently available with regard to climate does
not provide a clear understanding of a quantitative relationship between concentrations of PM
mass in ambient air and associated climate-related effects, and consequently, precludes a
5-39

-------
quantitative evaluation of the level of protection provided by a PM concentration-based
secondary standard from adverse climate-related effects on the public welfare in the U.S. Thus,
on the whole, we do not find the currently available information to provide support for different
conclusions than were reached in the last review with regard to climate-related effects of PM in
ambient air.
In considering the currently available evidence related to materials effects and PM, we
note that there is newly available evidence that informs our understanding on the soiling process
and types of materials affected, and provides limited information on dose-response relationships
and damage functions, although most of the recent evidence comes from studies outside of the
U.S. In particular, there is a growing body of research on PM and energy efficiency-related
materials, such as solar panels and passive cooling building materials, affecting the optical and
thermal properties, thereby impacting the intended energy efficiency of these materials. While
new research has added to the understanding of PM-related materials effects, there remains a
lack of research related to quantifying materials effects and understanding the public welfare
implications of such effects.
In summary, with regard to the two main non-visibility effects - climate effects and
materials effects - the available evidence, as in the last review, documents a causal role for PM
in ambient air. This evidence, however, as in the last review, also includes substantial
uncertainties with regard to quantitative relationships with PM concentrations and concentration
patterns that limit our ability to quantitatively assess the public welfare protection provided by
the standards from these effects. Thus, as a whole, the current information, which is not
appreciably different from that available in the last review, does not call into question the
adequacy of protection provided by the current standards for these effects.
Based on all of the above considerations and consistent with CASAC advice, we find that
the available evidence does not call into question the protection afforded by the current
secondary PM standards against PM-related welfare effects. Thus, our conclusion for the
Administrator's consideration is that it is appropriate to consider retaining the current secondary
PM standards, without revision. In so concluding, we recognize, as noted above, that the final
decision on this review of the secondary PM standards to be made by the Administrator is largely
a public welfare judgment, based on his judgment as to the requisite protection of the public
welfare from any known or anticipated adverse effects. This final decision will draw upon the
available scientific evidence and quantitative analyses on PM-attributable welfare effects, along
with consideration of CASAC advice and public comments, and on judgments about the
appropriate weight to place on the range of uncertainties inherent in the evidence and analyses.
5-40

-------
5.5 AREAS FOR FUTURE RESEARCH AND DATA COLLECTION
In this section, we highlight key uncertainties in the available information related to the
effects of PM on public welfare. Such key uncertainties and areas for future research, model
development, and data gathering are outlined below. We note, however, that a full set of research
recommendations is beyond the scope of this discussion. Rather, listed below are key
uncertainties, research questions and data gaps that have been thus far highlighted in this review
of the secondary PM standards.
•	A critical aspect of our consideration of the evidence and quantitative information for
visibility impairment is our understanding of human perception of visibility impairment
in the preference studies. This is essential to the Administrator's consideration of the
public welfare implications of visibility effects and to decisions on the adequacy of
protection provided by the secondary PM standards from them. Additional information
related to several areas would reduce uncertainty in in our interpretation of the available
information for purposes of characterizing visibility impairment. These areas include the
following:
-	Expanding the number and geographic coverage of preference studies in urban,
rural and Class I areas to account for the potential for people to have different
preferences based on the conditions that they commonly encounter and potential
differences in preferences based on the scene types;
-	Evaluating visibility preferences of the U.S. population today, given that the
currently available preference studies were conducted more than 15 years ago,
during which time air quality in the U.S. has improved;
-	Accounting for the influence that varying study methods may have on an
individual's response as to what level of visibility impairment is acceptable; and
-	Providing insights regarding people's judgments on acceptable visibility based on
those factors that can influence an individual's perception of visibility
impairment, including the duration of visibility impairment experiences, the time
of day during which light extinction is greatest, and the frequency of episodes of
visibility impairment, as well as the intensity of the visibility impairment.
•	Direct monitoring of PM2.5 light extinction would help to characterize visibility and the
relationships between PM component concentrations and light extinction and to evaluate
and refine light extinction calculation algorithms for use in areas near anthropogenic
sources, and would provide measurements for future visibility effects assessments.
•	Substantial uncertainties still remain with respect to key processes linking PM and
climate, because of the small scale of PM-relevant atmospheric processes compared to
the resolution of state-of-the-art models, and because of the complex cascade of indirect
impacts and feedbacks in the climate system that result from an initial PM-related
radiative perturbation. Such uncertainties include those related to our understanding of:
-	The magnitude of PM radiative forcing and the portion of that associated with
anthropogenic emissions;
5-41

-------
-	The contribution of regional differences in PM concentrations, and of individual
components, to radiative forcing; and,
-	The process by which PM interacts with clouds and how to represent such
interactions in climate models.
•	Research on more accurate U.S. and global emission inventories would provide source-
specific data on PM and PM component contributions to climate effects, particularly
those effects resulting from climate forcing.
•	While CASAC highlighted a number of studies as providing quantitative information
regarding the impact of reductions in PM2.5 on direct and indirect climate effects, these
studies largely are conducted at a global scale and assume a zeroing out or near-zeroing
out of global PM emissions. Research is needed regarding the impacts of incremental
changes in PM mass on direct and indirect climate effects on a regional scale, thereby
limiting our ability to quantify the impact of these changes at this time.
•	Insufficient evidence is available to relate soiling or damage to specific PM
concentrations or to establish a quantitative relationship between PM concentrations in
ambient air and materials degradation. Additional information would reduce uncertainty
in in our interpretation of the available information, including in the following areas:
-	Identifying quantitative relationships between particle size, PM concentration,
chemical concentrations, and frequency of repainting and repair;
-	Understanding human perceptions of reduced aesthetic appeal of buildings, and
other objects to PM-related materials effects; and
-	Characterizing deposition rates of airborne PM to surfaces and the interaction of
co-pollutants.
5-42

-------
REFERENCES
Abt Associates, Inc. (2001). Assessing public opinions on visibility impairment due to air
pollution: Summary report. Research Triangle Park, NC, U.S. Environmental Protection
Agency.
Adams, PJ, Seinfeld, JH, Koch, D, Mickley, L and Jacob, D (2001). General circulation model
assessment of direct radiative forcing by the sulfate-nitrate-ammonium-water inorganic
aerosol system. J GeophysRes 106(D1): 1097-1111.
Alfaro, SC, Chabas, A, Lombardo, T, Verney-Carron, A and Ausset, P (2012). Predicting the
soiling of modern glass in urban environments: A new physically-based model. Atmos
Environ 60: 348-357.
Ban-Weiss, GA, Jin, L, Bauer, SE, Bennartz, R, Liu, X, Zhang, K, Ming, Yi, Guo, H and Jiang,
JH (2014). Evaluating clouds, aerosols, and their interactions in three global climate
models using satellite simulators and observations. Journal of Geophysical Research:
Atmospheres 119(18): 10876-10901.
Barca, D, Belfiore, CM, Crisci, GM, La Russa, MF, Pezzino, A and Ruffolo, SA (2010).
Application of laser ablation ICP-MS and traditional techniques to the study of black
crusts on building stones: a new methodological approach. Environmental Science and
Pollution Research 17(8): 1433-1447.
BBC Research & Consulting (2003). Phoenix area visibility survey. Denver, CO.
Beloin, NJ and Haynie, FH (1975). Soiling of building materials. J Air Waste Manage Assoc
25(4): 399-403.
Bond, TC, Doherty, SJ, Fahey, DW, Forster, PM, Berntsen, T, Deangelo, BJ, Flanner, MG,
Ghan, S, Karcher, B, Koch, D, Kinne, S, Kondo, Y, Quinn, PK, Sarofim, MC, Schultz,
MG, Schulz, M, Venkataraman, C, Zhang, H, Zhang, S, Bellouin, N, Guttikunda, SK,
Hopke, PK, Jacobson, MZ, Kaiser, JW, Klimont, Z, Lohmann, U, Schwarz, JP, Shindell,
D, Storelvmo, T, Warren, SG and Zender, CS (2013). Bounding the role of black carbon
in the climate system: A scientific assessment. Journal of Geophysical Research:
Atmospheres 118(11): 5380-5552.
Boucher, O (2013). Clouds and Aerosols. In Climate Change 2013: The Physical Science Basis.
Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge, United Kingdom and New York, NY, USA,
Cambridge University Press.
Boyle, L, Burton, PD, Danner, V, Hannigan, MP and King, B (2017). Regional and national
scale spatial variability of photovoltaic cover plate soiling and subsequent solar
transmission losses. 7(5): 1354-1361.
Brimblecombe, P and Grossi, CM (2009). Millennium-long damage to building materials in
London. Sci Total Environ 407(4): 1354-1361.
5-43

-------
Brimblecombe, P and Grossi, CM (2010). Potential damage to modern building materials from
21st century air pollution. ScientificWorldJournal 10: 116-125.
Camuffo, D and Bernardi, A (1993). Microclimatic factors affecting the trajan column. Sci Total
Environ 128(2-3): 227-255.
Casati, M, Rovelli, G, D'Angelo, L, Perrone, MG, Sangiorgi, G, Bolzacchini, E and Ferrero, L
(2015). Experimental measurements of particulate matter deliquescence and
crystallization relative humidity: Application in heritage climatology. Aerosol and Air
Quality Research 15(2): 399-409.
Cox, LA. (2019). Letter from Louis Anthony Cox, Jr., Chair, Clean Air Scientific Advisory
Committee, to Administrator Andrew R. Wheeler. Re: CASAC Review of the EPA's
Policy Assessment for the Review of the National Ambient Air Quality Standards for
Particulate Matter (External Review Draft - September 2019). December 16, 2019. EPA-
CASAC-20-001. U.S. EPA HQ, Washington DC. Office of the Administrator, Science
Advisory Board. Available at:
https://vosemite.epa.gov/sab/sabprodiict.nsf/264cbl227d55e02c852.57402007446a4/E2F
i \	: MD20069DFBl/$File/KP A-C ASAC-20-001 nlf.
de Oliveira, BP, de la Rosa, JM, Miller, AZ, Saiz-Jimenez, C, Gomez-Bolea, A, Braga, MAS and
Dionisio, A (2011). An integrated approach to assess the origins of black films on a
granite monument. Environmental Earth Sciences 63(6-7 SI): 1677-1690.
Ely, DW, Leary, JT, Stewart, TR and Ross, DM (1991). The establishment of the Denver
Visibility Standard. Denver, Colorado, Colorado Department of Health.
Fajardo, OA, Jiang, J and Hao, J (2013). Assessing young people's preferences in urban visibility
inbeijing. Aerosol and Air Quality Research 13(5): 1536-1543.
Fiore, AM, Naik, V and Leibensperger, EM (2015). Air quality and climate connections. J Air
Waste Manage Assoc 65(6): 645-685.
Grossi, CM, Brimblecombe, P, Esbert, RM and Javier Alonso, F (2007). Color changes in
architectural limestones from pollution and cleaning. Color Research and Application
32(4): 320-331.
Hand, JL, Copeland, SA, Day, DA, Dillner, AM, Indresand, H, Malm, WC, McDade, CE,
Moore, CT, Jr., Pitchford, ML, Schichtel, BA and Watson, JG (2011). Spatial and
seasonal patterns and temporal variability of haze and its constituents in the United
States, IMPROVE Report V. Fort Collins, CO, Colorado State University.
Hand, JL, Schichtel, BA, Pitchford, M, Malm, WC and Frank, NH (2012). Seasonal composition
of remote and urban fine particulate matter in the United States. Journal of Geophysical
Research: Atmospheres 117(D5).
Hand, JL, Schichtel, BA, Malm, WC and Frank, NH (2013). Spatial and Temporal Trends in
PM2.5 Organic and Elemental Carbon across the United States. Advances in Meteorology.
5-44

-------
Hauglustaine, DA, Balkanski, Y and Schulz, M (2014). A global model simulation of present
and future nitrate aerosols and their direct radiative forcing of climate. Atmos Chem Phys
14(20): 11031-11063.
Heald, CL, Ridley, DA, Kreidenweis, SM and Drury, EE (2010). Satellite observations cap the
atmospheric organic aerosol budget. Geophys Res Lett 37.
IPCC (2013). Climate change 2013: The physical science basis. Contribution of working group I
to the fifth assessment report of the Intergovernmental Panel on Climate Change. T. F.
Stocker, D. Qin, G. K. Plattner et al. Cambridge, UK, Cambridge University Press.
Jimenez, JL, Canagaratna, MR, Donahue, NM, Prevot, AS, Zhang, Q, Kroll, JH, Decarlo, PF,
Allan, JD, Coe, H, Ng, NL, Aiken, AC, Docherty, KS, Ulbrich, IM, Grieshop, AP,
Robinson, AL, Duplissy, J, Smith, JD, Wilson, KR, Lanz, VA, Hueglin, C, Sun, YL,
Tian, J, Laaksonen, A, Raatikainen, T, Rautiainen, J, Vaattovaara, P, Ehn, M, Kulmala,
M, Tomlinson, JM, Collins, DR, Cubison, MJ, Dunlea, EJ, Huffman, JA, Onasch, TB,
Alfarra, MR, Williams, PI, Bower, K, Kondo, Y, Schneider, J, Drewnick, F, Borrmann,
S, Weimer, S, Demeijian, K, Salcedo, D, Cottrell, L, Griffin, R, Takami, A, Miyoshi, T,
Hatakeyama, S, Shimono, A, Sun, JY, Zhang, YM, Dzepina, K, Kimmel, JR, Sueper, D,
Jayne, JT, Herndon, SC, Trimborn, AM, Williams, LR, Wood, EC, Middlebrook, AM,
Kolb, CE, Baltensperger, U and Worsnop, DR (2009). Evolution of organic aerosols in
the atmosphere. Science 326(5959): 1525-1529.
Kelly, J, Schmidt, M and Frank, N. (2012a). Memorandum to PM NAAQS Review Docket
(EPA-HQ-OAR-2007-0492). Updated comparison of 24-hour PM2.5 design values and
visibility index design values. December 14, 2012. Docket ID No. EPA-HQ-OAR-2007-
0492. Research Triangle Park, NC. Office of Air Quality Planning and Standards.
Available at: https://www3.epa.g0v/ttn/naaqs/standards/pm/data/2	dly.pdf.
Kelly, J, Schmidt, M, Frank, N, Timin, B, Solomon, D and Venkatesh, R. (2012b). Memorandum
to PM NAAQS Review Docket (EPA-HQ-OAR-2007-0492). Technical Analyses to
Support Surrogacy Policy for Proposed Secondary PM2.5 NAAQS under NSR/PSD
Programs. June 14, 2012. . Docket ID No. EPA-HQ-OAR-2007-0492. Research Triangle
Park, NC. Office of Air Quality Planning and Standards. Available at:
https://www3. epa. gov/ttn/naaq s/standards/pm/data/20120614Ketty.pdf.
Kloppmann, W, Bromblet, P, Vallet, JM, Verges-Belmin, V, Rolland, O, Guerrot, C and
Gosselin, C (2011). Building materials as intrinsic sources of sulphate: A hidden face of
salt weathering of historical monuments investigated through multi-isotope tracing (B, O,
S). Sci Total Environ 409(9): 1658-1669.
Kok, JF, Ridley, DA, Zhou, Q, Miller, RL, Zhao, C, Heald, CL, Ward, DS, Albani, S and
Haustein, K (2017). Smaller desert dust cooling effect estimated from analysis of dust
size and abundance. Nature Geoscience 10(4): 274-278.
Lanzon, M and Garcia-Ruiz, PA (2010). Deterioration and damage evaluation of rendering
mortars exposed to sulphuric acid. Mater Struct 43(3): 417-427.
5-45

-------
Lau, NT, Chan, CK, Chan, L and Fang, M (2008). A microscopic study of the effects of particle
size and composition of atmospheric aerosols on the corrosion of mild steel. Corros Sci
50(10): 2927-2933.
Lee, YH, Lamarque, JF, Flanner, MG, Jiao, C, Shindell, DT, Berntsen, T, Bisiaux, MM, Cao, J,
Collins, WJ, Curran, M, Edwards, R, Faluvegi, G, Ghan, S, Horowitz, LW, McConnell,
JR, Ming, J, Myhre, G, Nagashima, T, Naik, V, Rumbold, ST, Skeie, RB, Sudo, K,
Takemura, T, Thevenon, F, Xu, B and Yoon, JH (2013). Evaluation of preindustrial to
present-day black carbon and its albedo forcing from Atmospheric Chemistry and
Climate Model Intercomparison Project (ACCMIP). Atmos Chem Phys 13(5): 2607-
2634.
Leibensperger, EM, Mickley, LJ, Jacob, DJ, Chen, WT, Seinfeld, JH, Nenes, A, Adams, PJ,
Streets, DG, Kumar, N and Rind, D (2012). Climatic effects of 1950-2050 changes in US
anthropogenic aerosols - Part 2: Climate response. Atmos Chem Phys 12(7): 3349-3362.
Levy, H, Horowitz, LW, Schwarzkopf, MD, Ming, Yi, Golaz, JC, Naik, V and Ramaswamy, V
(2013). The roles of aerosol direct and indirect effects in past and future climate change.
Journal of Geophysical Research: Atmospheres 118(10): 4521-4532.
Liu, B, Wang, DW, Guo, H, Ling, ZH and Cheung, K (2015). Metallic corrosion in the polluted
urban atmosphere of Hong Kong. Environ Monit Assess 187(1): 4112.
Lombardo, T, Ionescu, A, Chabas, A, Lefevre, RA, Ausset, P and Candau, Y (2010). Dose-
response function for the soiling of silica-soda-lime glass due to dry deposition. Sci Total
Environ 408(4): 976-984.
Lowenthal, DH and Kumar, N (2004). Variation of mass scattering efficiencies in IMPROVE.
Journal of the Air and Waste Management Association (1990-1992) 54(8): 926-934.
Lowenthal, DH and Kumar, N (2016). Evaluation of the IMPROVE Equation for estimating
aerosol light extinction. J Air Waste Manage Assoc 66(7): 726-737.
Malm, WC, Sisler, JF, Huffman, D, Eldred, RA and Cahill, TA (1994). Spatial and seasonal
trends in particle concentration and optical extinction in the United States. J Geophys Res
99(D1): 1347-1370.
Malm, WC and Hand, JL (2007). An examination of the physical and optical properties of
aerosols collected in the IMPROVE program. Atmos Environ 41(16): 3407-3427.
McAlister, J, Smith, BJ and Torok, A (2008). Transition metals and water-soluble ions in
deposits on a building and their potential catalysis of stone decay. Atmos Environ 42(33):
7657-7668.
McNeill, VF, Woo, JL, Kim, DD, Schwier, AN, Wannell, NJ, Sumner, AJ and Barakat, JM
(2012). Aqueous-phase secondary organic aerosol and organosulfate formation in
atmospheric aerosols: a modeling study. Environ Sci Technol 46(15): 8075-8081.
5-46

-------
Mie, G (1908). Beitrage zur Optik truber Medien, speziell kolloidaler Metallosungen [Optics of
cloudy media, especially colloidal metal solutions], Annalen derPhysik25(3): 377-445.
Mooers, HD, Cota-Guertin, AR, Regal, RR, Sames, AR, Dekan, AJ and Henkels, LM (2016). A
120-year record of the spatial and temporal distribution of gravestone decay and acid
deposition. Atmos Environ 127: 139-154.
Myhre, G, Shindell, D, Breon, FM, Collins, W, Fuglestvedt, J, Huang, J, Koch, D, Lamarque, JF,
Lee, D, Mendoza, B, Nakajima, T, Robock, A, Stephens, G, Takemura, T and Zhang, H,
Eds. (2013). Anthropogenic and natural radiative forcing. Cambridge, UK, Cambridge
University Press.
Ozga, I, Bonazza, A, Bernardi, E, Tittarelli, F, Favoni, O, Ghedini, N, Morselli, L and Sabbioni,
C (2011). Diagnosis of surface damage induced by air pollution on 20th-century concrete
buildings. Atmos Environ 45(28): 4986-4995.
Pitchford, M, Maim, W, Schichtel, B, Kumar, N, Lowenthal, D and Hand, J (2007). Revised
algorithm for estimating light extinction from IMPROVE particle speciation data. J Air
Waste Manage Assoc 57(11): 1326-1336.
Pryor, SC (1996). Assessing public perception of visibility for standard setting exercises. Atmos
Environ 30(15): 2705-2716.
Radonjic, IS, Pavlovic, TM, Mirjanic, DLj, Radovic, MK, Milosavljevic, DD and Pantic, LS
(2017). Investigation of the impact of atmospheric pollutants on solar module energy
efficiency. Thermal Science 21(5): 2021-2030.
Rosso, F, Pisello, AL, Jin, WH, Ghandehari, M, Cotana, F and Ferrero, M (2016). Cool marble
building envelopes: The effect of aging on energy performance and aesthetics.
Sustainability 8(8): Article #753.
Ryan, PA, Lowenthal, D and Kumar, N (2005). Improved light extinction reconstruction in
interagency monitoring of protected visual environments. J Air Waste Manage Assoc
55(11): 1751-1759.
Samet, J. (2010). Letter from Jonathan Samet, Chair, Clean Air Scientific Advisory Committee,
to Administrator Lisa Jackson. Re: CAS AC Review of Policy Assessment for the Review
of the PM NAAQS - First External Review Draft (March 2010). May 17, 2010. EPA-
CASAC-10-011. U.S. EPA HQ, Washington DC. Office of the Administrator, Science
Advisory Board. Available at:
https://nepis.epa. gov/Exe/ZyPURL.cgi?Dockev:	xt.
Scott, CE, Rap, A, Spracklen, DV, Forster, PM, Carslaw, KS, Mann, GW, Pringle, KJ, Kivekas,
N, Kulmala, M, Lihavainen, H and Tunved, P (2014). The direct and indirect radiative
effects of biogenic secondary organic aerosol. Atmos Chem Phys 14(1): 447-470.
Shindell, DT, Lamarque, JF, Schulz, M, Flanner, M, Jiao, C, Chin, M, Young, PJ, Lee, YH,
Rotstayn, L, Mahowald, N, Milly, G, Faluvegi, G, Balkanski, Y, Collins, WJ, Conley,
5-47

-------
AJ, Dalsoren, S, Easter, R, Ghan, S, Horowitz, L, Liu, X, Myhre, G, Nagashima, T, Naik,
V, Rumbold, ST, Skeie, R, Sudo, K, Szopa, S, Takemura, T, Voulgarakis, A, Yoon, JH
and Lo, F (2013). Radiative forcing in the ACCMIP historical and future climate
simulations. Atmos Chem Phys 13(6): 2939-2974.
Sleiman, M, Kirchstetter, TW, Berdahl, P, Gilbert, HE, Quelen, S, Marlot, L, Preble, CV, Chen,
S, Montalbano, A, Rosseler, O, Akbari, H, Levinson, R and Destaillats, H (2014). Soiling
of building envelope surfaces and its effect on solar reflectance - Part II: Development of
an accelerated aging method for roofing materials. Sol Energy Mater Sol Cells 122: 271-
281.
Smith, AE and Howell, S (2009). An assessment of the robustness of visual air quality
preference study results. Washington, DC, CRA International.
Tai, APK, Mickley, LJ and Jacob, DJ (2010). Correlations between fine particulate matter
(PM2.5) and meteorological variables in the United States: Implications for the sensitivity
of PM2.5 to climate change. Atmos Environ 44(32): 3976-3984.
Takemura, T (2012). Distributions and climate effects of atmospheric aerosols from the
preindustrial era to 2100 along Representative Concentration Pathways (RCPs) simulated
using the global aerosol model SPRINTARS. Atmos Chem Phys 12(23): 11555-11572.
Twomey, S (1977). The influence of pollution on the shortwave albedo of clouds. Journal of the
Atmospheric Sciences 34(7): 1149-1152.
U.S. EPA. (2004). Air Quality Criteria for Particulate Matter. (Vol I of II). Research Triangle
Park, NC. Office of Research and Development. U.S. EPA. EPA-600/P-99-002aF.
October 2004. Available at:
https://nepis.epa. gov/Exe/ZyPURL.cgi?Dockey=P 100LFIQ.txt.
U.S. EPA. (2009). Integrated Science Assessment for Particulate Matter (Final Report). Research
Triangle Park, NC. Office of Research and Development, National Center for
Environmental Assessment. U.S. EPA. EPA-600/R-08-139F. December 2009. Available
at: https://cfpub.epa.gov/ncea/risk/recordisplay.cfm?de: 546.
U.S. EPA. (2010). Particulate Matter Urban-Focused Visibility Assessment (Final Document).
Research Triangle Park, NC. Office of Air Quality Planning and Standards, Health and
Environmental Impacts Division. U.S. EPA. EPA-452/R-10-004 July 2010. Available at:
https://nepis.epa. gov/Exe/ZyPURL.cgi?Dockey=P 100FQ5D.txt.
U.S. EPA. (2011). Policy Assessment for the Review of the Particulate Matter National Ambient
Air Quality Standards Research Triangle Park, NC. Office of Air Quality Planning and
Standards, Health and Environmental Impacts Division. U.S. EPA. EPA-452/R-11-003
April 2011. Available at:
https://nepis.epa. gov/Exe/ZyPURL.cgi?Dockey=P 100 AUMY.txt.
5-48

-------
U.S. EPA. (2012a). Report to Congress on Black Carbon. Washington, DC. U.S. Environmental
Protection Agency, Office of Air and Radition. U.S. EPA. EPA-450/R-12-001. March
2012. Availble at: http://www.epa.gov/blackcarbon/2012report/fullreport.pdf.
U.S. EPA. (2012b). Responses to Significant Comments on the 2012 Proposed Rule on the
National Ambient Air Quality Standards for Particulate Matter (June 29, 2012; 77 FR
38890). Research Triangle Park, NC. U.S. EPA. Docket ID No. EPA-HQ-OAR-2007-
0492. Available at: https://www3.epa.eov/ttn/naaqs/standards/pm/data/2C	.pdf.
U.S. EPA. (2015). Preamble to the integrated science assessments. Research Triangle Park, NC.
U.S. Environmental Protection Agency, Office of Research and Development, National
Center for Environmental Assessment, RTP Division. U.S. EPA. EPA/600/R-15/067.
November 2015. Available at:
https://cfpiib.epa.gov/ncea/isa/recordisplav.cfm7dei
. (2016). Integrated review plan for the national ambient air quality standards for particulate
matter. Research Triangle Park, NC. U.S. EPA. EPA-452/R-16-005. December 2016.
Available at: https://www3.epa.gOv/ttn./naaqs/standards/pm./data/2. final-integrated-
review-plan.pdf.
U.S. EPA. (2019). Integrated Science Assessment (ISA) for Particulate Matter (Final Report).
Washington, DC. U.S. Environmental Protection Agency, Office of Research and
Development, National Center for Environmental Assessment. U.S. EPA. EPA/600/R-
19/188. December 2019. Available at: https://www.epa.gov/naaqs/part.iculate-rnatter-prn-
standards-integrated-science-assessments-current-review.
Van de Hulst, H (1981). Light scattering by small particles. New York, Dover Publications, Inc.
Walwil, HM, Mukhaimer, A, Al-Sulaiman, FA and Said, SAM (2017). Comparative studies of
encapsulation and glass surface modification impacts on PV performance in a desert
climate. Solar Energy 142: 288-298.
Watt, J, Jarrett, D and Hamilton, R (2008). Dose-response functions for the soiling of heritage
materials due to air pollution exposure. Sci Total Environ 400(1-3): 415-424.
Zelinka, MD, Andrews, T, Forster, PM and Taylor, KE (2014). Quantifying components of
aerosol-cloud-radiation interactions in climate models. Journal of Geophysical Research:
Atmospheres 119(12): 7599-7615.
5-49

-------
APPENDIX A. SUPPELMENTAL INFORMATION ON
PM AIR QUALITY ANALYSES
This appendix provides supplemental information on the data sources and methods used
to generate the figures and table presented in Chapter 2 of this PA. Sections A. 1 to A.4 describe
the data sources and methods used to generate figures and tables in section 2.3.2. Section A.5
describes the data sources and methods used to generate figures and tables in section 2.3.3.
Section A.6 describes the data sources and methods used to generate figures and tables in section
2.4.
A.1 DATA SOURCES AND METHODS FOR GENERATING NATIONAL
PM2 5, PMio, PM10-2 s, AND PM2 5 SPECIATION FIGURES
•	PM2.5 annual average and 98th percentile mass concentrations: calculated from regulatory-
quality (Federal Reference Method or Federal Equivalent Method) 24-hour average
values from monitors with at least 75% completeness for each year. When a single site
has multiple monitors, the figure shows the average of the annual averages and 98th
percentiles from each monitor at the site. We downloaded the monitor-level
concentrations for all sites in the United States for all available days (including potential
exceptional events) for 2000-2017 from the EPA's Air Quality System (AQS,
https ://www. epa. gov/aq s)
•	PM10 annual average and 98th percentile mass concentrations: calculated from both
regulatory and non-regulatory methods using 24-hour average values from monitors with
at least 75% completeness for each year. When a single site has multiple monitors, the
figure shows the average of the annual averages and 98th percentiles from each monitor at
the site. We downloaded the monitor-level concentrations for all sites in the United
States for all available days (including potential exceptional events) for 2000-2017 from
the EPA's Air Quality System (AQS, https://www.epa.gov/aqs)
•	PMi0-2.5 annual average and 98th percentile mass concentrations: calculated from both
regulatory and non-regulatory methods using 24-hour average values from monitors with
at least 75% completeness for each year. When a single site has multiple monitors, the
figure shows the average of the annual averages and 98th percentiles from each monitor at
the site. We downloaded the monitor-level concentrations for all sites in the United
States for all available days (including potential exceptional events) for 2000-2017 from
the EPA's Air Quality System (AQS, https://www.epa.gov/aqs)
•	PM2.5 speciated annual average mass concentrations: calculated from filter-based, 24-hour
averages from monitors with at least 75% completeness for each year. We downloaded
data from monitors that are part of the Interagency Monitoring of Protected Visual
Environments (IMPROVE) network, Chemical Speciation Network (CSN), and the
NCore Multipollutant Monitoring Network for 2015-2017.
A-l

-------
•	The 2000-2017 trends are calculated from the Pearson correlation coefficient for monitors
having at least 75% of the available years with 75% completeness within each year.
When a single site has multiple monitors, the average of the annual averages and 98th
percentiles from each monitor at the site is taken prior to calculation of the Pearson
correlation coefficient.
A.2 DATA SOURCES AND METHODS FOR GENERATING NEAR-
ROAD PM2 5 DESIGN VALUE TABLE AND INCREMENT FIGURES
•	PM2.5 design values: calculated using the data handling described by 40 CFR Appendix N
to Part 50 - Interpretation of the National Ambient Air Quality Standards for PM2.5. We
downloaded the design values for all sites in the United States for all available days
(including potential exceptional events) for 2015-2017 from the EPA's Air Quality
System (AQS, https://www.epa.gov/aqs)
•	PM2.5 hourly, daily, and annual average mass concentrations: calculated from regulatory-
quality (Federal Reference Method or Federal Equivalent Method) monitors. When a
single site has multiple monitors, the figures shows the average from all monitors at the
site. We downloaded the monitor-level concentrations for all sites in the United States
for all available days (including potential exceptional events) for 2000-2017 from the
EPA's Air Quality System (AQS, https://www.epa.gov/aqs)
•	Near-road sites: designated from the list of near-road sites found at
https://www3 .epa.gov/ttnamti 1/files/nearroad/Near-
road%20Monitorin g%2.0Network%20 Site%20Li st%20~%20May%20 Isx.
•	The near-road PM2.5 increment is calculated by excluding the near-road site within a
CBS A, predict the interpolated concentration at the near-road site location using Inverse
Distance Weighting (IDW), and subtract the predicted concentration from the actual
concentration at the near-road site for each daily or hourly average. Only CBS As with at
least one non-near-road site within 5km of the near-road site are considered. For the
Elizabeth, NJ figure, the Elizabeth Lab site was considered a near-road site for the IDW
calculation.
A.3 DATA SOURCES FOR SUB-DAILY PM2.5 CONCENTRATION
FIGURE
•	PM2.5 hourly average mass concentrations: calculated from regulatory-quality Federal
Equivalent Method monitors. The 2-hour and 5-hour averages were calculated for periods
with each hourly average available. Only sites with a valid annual or 24-hour design
value for 2015-2017 are shown in the figure. The percentages of 2-hour average PM2.5
mass concentrations above 140 ng/in3 at individual sites are illustrated in Figure A-l.
Frequency distributions of 5-hour averages are presented in Figure A-2.
A-2

-------
Figure A-l. Percentages of 2015-2017 2-hour average PM2.5 mass concentrations above 140
Hg/m3.
Sites meeting both NAAQS
Sites violating either NAAQS
Percentiles (|jg m 3)
Percentiles (pg m
137.4
Percentiles (pg rrr)
Percentiles (ug m 3
168.5
Concentration (^g m )
Concentration (|ig m )
Figure A-2. Frequency distribution of 2015-2017 5-hour averages for sites meeting both or
violating either PM2.5 NAAQS for October to March (blue) and April to September
(red).
A-3

-------
A.4 DATA SOURCES FOR ULTRAFINE FRACTION OF PM2 5 MASS
FIGURE
• Annual average particle number and mass concentrations for Bondville. IL: calculated
from 24-hour average values for years with 66% data completion in 75% of the months
of the year from 2000-2017. We downloaded the mass concentrations from the EPA's
Air Quality System (AQS, https://www.epa.gov/aqs) and particle number concentrations
from NOAA's Earth System Research Laboratory's Global Monitoring Division
A.5 METHODS FOR PREDICTING AMBIENT PM2.5 BASED ON HYBRID
MODELING APPROACHES
A.5.1 Data Sources for 2011 PM2.5 Spatial Fields
•	The "HU2017" fields were provided by Professor Yang Liu of Emory University in the
form of comma-separated-values files (*.csv) of daily average PM2.5 on a national grid.
•	The "DI2016" fields were provided by Dr. Qian Di of Harvard in the form of MATLAB
files (*.mat) of daily average PM2.5 on a national grid.
•	The "VD2019" fields were provided by Dr. Aaron van Donkelaar in the form of netCDF
files (*.nc) of annual average concentration. These files are also available at:
http://fizz.ph.vs.dal.ca/~atm.os/martin/7page id=140.
•	The "downscaler" files were developed in terms of daily average Downscaler predictions
on a national grid following methods described in the risk assessment appendix.
A.5.2 Data Averaging and Coefficient of Variation
•	PM2.5 concentration fields were loaded into R version 3.4.4, and daily fields were
averaged to the annual period. Concentrations for each method at prediction points were
then averaged to the corresponding CMAQ grid cells to enable consistent comparisons
for Figure 2-24, 2-26, and Table 2-3.
•	The coefficient of variation (CoV) was calculated for each grid cell using the following
formula
• Data sources for Figure 2-30: Smoke and fire detections observed by MODIS in August
..gov/gmd ).
100 Yli-^Pi-P)2
CoV (%) = —		-
CoV(%) =
where P is the prediction for each of the four methods (i.e., N=4).
A.6 ANALYSES OF BACKGROUND PM
2017
Image was produced using the NASA Worldview platform
(https://worldview.earthdata.nasa.gov/). Layers selected were 1) Corrected
A-4

-------
Reflectance and 2) Fires and Thermal Anomalies, both from Aqua/MODIS. Day
selected was August 4, 2017.
•	Data sources for Figure 2-31: Fine PM mass time series during 2017 from North Cascades
IMPROVE site
-	Image was archived from the IMPROVE website
(http://vi ews.cira.colostate.edu/fed/SiteBrowser/Default.aspx?appkev=SBCFPm
HazeComp; hosted by C1RA/CSU and sponsored by NPS and USFS) for the
North Cascades (NOCA1) site in 2017.
•	Data sources for Figure 2-32: Speciated annual average fine PM mass from IMPROVE at
select remote monitors in 2004 and 2016
-	Speciated IMPROVE data from 2004 and 2016
(http://vi ews.cira.colostate.edu/fed/SiteBrowser/Default.aspx?appkev=SBCF Pm
HazeComp) were averaged annually for each monitor. Corresponding monitor
locations are shown in Figure 2-32.
A-5

-------
APPENDIX B. DATA INCLUSION CRITERIA AND
SENSITIVITY ANALYSES

-------
TABLE OF CONTENTS
B.l Forest Plots	B-l
B.2 Monitored PM2.5 Concentrations in Key Epidemiologic Studies	B-l
B.3 Hybrid model predicted PM2.5 Concentrations in Key Epidemiologic Studies	B-3
B.4 Design Value Box Plot Inclusion Criteria	B-7
B.4.1 Study Area Assignment	B-8
B.4.2 Study Population Assignment 	B-8
B.4.3 Air Quality Data Assignment by Study Area, by Study Period 	B-10
B.5 Percent of Study Area Population Captured in Design Value Plots	B-17
B.6 Sensitivity Analysis: Box Plots Using Counts of Health Events Versus Study Area
Population	B-20
B.7 Comparisons Between Annual and Daily Design Values	B-21
B.8 24-Hour Pseudo-Design Values and Distributions Across Study Areas	B-26
B.9 Pseudo-Design Value Distribution by Average County Pseudo-Design Values per 1
|ig/m3	B-28
B.10 Details of Key Epidemiologic Studies, Including Study Design, Exposure Metric, and
Statistical Analysis	B-33
References 	B-72
1

-------
This appendix presents supplemental information on the methods used to conduct the analyses
discussed in section 3.2.3.2 of this PA. It also presents information on additional sensitivity
analyses. Section B.l provides supplemental information on the forest plots presented in Figures
3-3 to 3-6. Sections B.2 and B.3 provide supplemental information on the study-reported PM2.5
concentrations presented in Figure 3-7 and Figure 3-8. Sections B.4 to B.6, and sections B.8 to
B.10, present supplemental information and sensitivity analyses related to the analyses of study
area pseudo-design values in section 3.2.3.2.2. Section B.7 presents comparisons between annual
and daily design values in CBS As.
B.l FOREST PLOTS
Forest Plots exhibiting effect estimates and 95% confidence intervals from epidemiologic
studies that have the potential to be most informative in reaching conclusions on the adequacy of
the current primary PM2.5 standards are shown in Figure 3-3 to Figure 3-6. Epidemiologic studies
included in these figures support "causal" or "likely to be causal" relationships with PM
exposures in the ISA U.S. EPA, 2019 and include mortality (all-cause mortality, CVD mortality,
respiratory mortality, lung cancer mortality), and morbidity (asthma incidence, lung cancer
incidence, lung function and lung development, CVD and respiratory emergency room visit or
hospital admission) health endpoints. Further, studies included in Figure 3-3 to Figure 3-6 were
restricted to multi-city studies in the United States or Canada. Multi-city studies within a single
State were not included, with the exception of respiratory morbidity endpoints, where multi-city
studies were limited. For some of the major cohort studies included in the previous ISA, like the
American Cancer Society (ACS) cohort, we included new studies that reanalyze epidemiologic
associations for multiple mortality endpoints (e.g. lung cancer mortality and IHD mortality) and
an extension of follow-up periods (e.g., Pope et al. (2015b), Turner et al. (2016), Jerrett et al.
(2016), and Thurston et al. (2016b)), as well as a reanalysis (Krewski et al. (2009) of the original
ACS dataset, including an extended follow-up period, that was evaluated in the previous ISA
(EPA, 2009). In total, 67 studies were included in Figure 3-3 to Figure 3-6.
B.2 MONITORED PM2 5 CONCENTRATIONS IN KEY EPIDEMIOLOGIC
STUDIES
Of the 67 key studies identified in Figure 3-3 to Figure 3-6, Figure 3-7 includes key
epidemiologic studies that report an overall study mean or median concentration of PM2.5 (as
opposed to a study mean/median range across study area locations) and based on ambient PM2.5
B-l

-------
monitored data. The plot includes studies that report significant effect estimates (22 studies) and
studies that only report non-significant effect estimates (5 studies). Further, to be included, only
key studies for which the years of air quality data used to estimate exposures overlap entirely
with the years during which health events are reported were included. The PM2.5 concentrations
reported by studies that estimate exposures from air quality corresponding to only part of the
study period, often including only the later years of the health data (e.g., Miller et al., 2007; Hart
et al., 2011; Thurston et al., 2013; Weichenthal et al., 2014; Weichenthal et al., 2016a; Pope et
al., 2015a; Villeneuve et al., 2015; Turner et al., 2016), are not likely to reflect the full ranges of
ambient PM2.5 concentrations that contributed to reported associations.1
Some of the included studies also provide city-specific study mean concentrations and
city-specific health events. Hence, PM2.5 exposure estimates corresponding to the 10th and 25th
percentiles of those events were determined in the following manner. City-specific cases and
PM2.5 concentrations were input in ascending order by PM2.5 concentration. The city-specific
percent of cases was calculated as a proportion of the total study cases and the cumulative
percent of cases was determined. The PM2.5 concentration associated with the cumulative percent
closest to the 10th and 25th percentiles were input in Figure 3-7 and the cumulative percent values
closest to the associated 10111 and 25th percentile inputs are shown in Table B-l2. Data for Bell et
al. (2008) and Zanobetti and Schwartz (2009) were previously provided by the study authors, as
described in Rajan (2011).
Table B-l. PM2.5 concentrations corresponding to the 25th and 10th percentiles of estimated
health events.

10th Percentile PM2.5
25th Percentile PM2.5
Citation
(|jg/m3) (Cumulative
(|jg/m3) (Cumulative

percent value closest)
percent value closest)
Bell et al. (2008)
9.8
11.5
Franklin et al. (2007)
10.4(11.1%)
12.9 (25.3%)
Stieb et al. (2009)
6.7(16.5%)
6.8 (20.5%)
Szyszkowicz (2009)
6.4(4.1%)
6.5(18.6%)
Zanobetti and Schwartz (2009)
10.3
12.5
1	This is an issue only for some studies of long-term PM2 5 exposures. While this approach can be reasonable in the
context of an epidemiologic study evaluating health effect associations with long-term PM2 5 exposures, under the
assumption that spatial patterns in PM2 5 concentrations are not appreciably different during time periods for
which air quality information is not available (e.g., Chen et al., 2016), our interest is in understanding the
distribution of ambient PM2 5 concentrations that could have contributed to reported health outcomes.
2	That is, 25% of the total health events occurred in study locations with mean PM2 5 concentrations (i.e., averaged
over the study period) below the 25th percentiles identified in Figure 3-7 and 10% of the total health events
occurred in study locations with mean PM2 5 concentrations below the 10th percentiles identified.
B-2

-------
B.3 HYBRID MODEL PREDICTED PM2 5 CONCENTRATIONS IN KEY
EPIDEMIOLOGIC STUDIES
Figure 3-8 focuses on multicity studies that are part of the evidence supporting "causal"
or "likely to be causal" determinations in the ISA and that use air quality data to estimate PM2.5
exposures for the entire range of years during which health events occurred. In addition, as
detailed in section 3.2.3.2.1, we also consider the approach used to validate model predictions,
and the studies included in Figure 3-8 are those for which relatively robust model validation
analyses are reported to have been conducted for the full range of years during which PM2.5
exposures are estimated in the health study.3 All studies that met the criteria for inclusion were
conducted in the U.S.
Figure 3-8 presents overall means of hybrid model-predicted PM2.5 concentrations for
key studies, and the concentrations corresponding to the 25th and 10th percentiles of estimated
exposures or health events, when available. For Di et al. (2017b), we present 25th and 10th
percentiles of annual PM2.5 concentrations by zip code corresponding to long-term exposure
estimates, while for Di et al. (2017a), we present daily air pollution concentrations (short-term
exposure estimates) corresponding to the 25th and 10th percentiles of deaths at the zip-code level.
These values, along with other percentiles, are illustrated in Figure B-l and Figure B-2 (Jenkins,
2019a, Jenkins, 2019b). The study authors for Di et al. (2017b) additionally provided
information on population weighted percentile values corresponding to long-term PM2.5 exposure
(Chan, 2019). These are presented in Table B-2. For other studies included in Figure 3-8 [Kloog
et al. (2012), Kloog et al. (2014), Shi et al. (2016), Wang et al. (2017)], 25th percentiles of
exposure estimates were derived from study manuscripts of air quality descriptive statistics and
can be found in Table B-3.
3 For example, due to lack of spatial field availability before 1998, Crouse et al. (2015) use median annual PM2 5
concentrations for the 1998-2006 time period (van Donkelaar et al., 2010; van Donkelaar et al., 2015a;van
Donkelaar et al., 2013) to predict exposures during the 1984-2006 period. Similarly, for Pinault et al. (2016),
model validation is for 2004 to 2008 (van Donkelaar et al., 2015b) while exposures are estimated for 1998 to
2012. Paciorek et al. (2009), which presents the model validation results for Puett et al. (2009) and Puett et al.
(2011), notes that PM2 5 monitoring was sparse prior to 1999, with many of the available PM25 monitors in rural
and protected areas. Therefore, Paciorek et al. (2009) conclude that coverage in the validation set for most of the
study period (1988-1998) is poor and that their model -strongly- underestimates uncertainty Paciorek et al.
(2009), p. 392 in published manuscript). Hystad et al. (2013) used exposure fields developed by calibrating
satellite-based PM2.5 surfaces from a recent period (van Donkelaar et al., 2010) to estimate exposure for the 1975
to 1994 (Hystad et al., 2012). Hystad et al. (2012) noted that a random effect model was used to estimate PM2 5
based on TSP measurements and metropolitan indicator variables because only small number of PM25
measurements were available, and no measurements were made prior to 1984. Thus, these studies are not
included in Figure 3-8.
B-3

-------
Percentiles of PM2.5 By Zip Code
Percentile of P#t2,5, Based on ZIP code
0%
70%
PMl5 Value
:¦ C2J:G25
f	ZC
8.0043245
i l o i:y:i
4:4490:,
9.8273901
10.1797192
TJ ;;7*b t!
:
I	27:, '0r.
11.6666604
12 07C7952
I! 4t<16170
II	:Gr<6_UC
f „ J j 3C
I* i765291
"T . i..' 1 «,.i> — *t
1= r.troo'r
Figure B-l. Percentiles of annual PM2.5 concentrations by zip code corresponding to long-
term exposure estimates in Di et al., 2017b.
B-4

-------
Table B-2. Population weighted percentiles of annual PM2.5 concentrations by zip code
corresponding to long-term exposure estimates in Di et al., 2017b.
Percentile
Population Weighted PM2.5

(Hg/m3)
0.0
0.0
5.0
7.1
10.0
7.9
15.0
8.6
20.0
9.1
25.0
9.5
30.0
9.9
35.0
10.3
40.0
10.6
45.0
11.0
50.0
11.4
55.0
11.7
60.0
12.1
65.0
12.5
70.0
12.9
75.0
13.4
80.0
13.9
85.0
14.4
90.0
15.1
95.0
16.1
100.0
32.6
B-5

-------
Percentiles of PM2.5 By Zip Code
Percentile of Daily PM2 Based on ZiP code
PM26 Value
•; 4:n:;3c
5 C~2"ti40
•5 >?r jCS'r?
100%
7.9031599
: C^;»2G50
3.1636408
3.S74G438
¦O 6124079
>' 41 r 614
>: l?ic:5i
ic.'dtc"
•4 4I~" .14
'5 1 ff%1?
'7 53d!4 d~i I
20.0959732
ii 47plKl:C
401 57'71237
Figure B-2. Daily air pollution concentrations (short-term exposure estimates)
corresponding to various percentiles of deaths at the zip-county level in Di et al., 2017a.
B-6

-------
Table B-3. PM2.5 concentrations corresponding to the 25th and 10th percentiles of estimated
exposures in Figure 3-8.
Citation
10th Percentile PM2.5 (|jg/m3)
25th Percentile PM2.5 (|jg/m3)
Di et al. (2017a)
4.7
6.7
Di et al. (2017b)
7.3
9.1
Kloog et al. (2012)

6.4
Kloog et al. (2014)

7.9
Shi et al. (2016)

4.6
Shi et al. (2016)

6.2
Wang et al. (2017)

9.1
B.4 DESIGN VALUE BOX PLOT INCLUSION CRITERIA
Studies selected from Figure 3-3 to Figure 3-6 for inclusion in Figure 3-9 and Figure B-9
(box plots of pseudo-design value distributions) are those studies that define the study area/s
(city or county) and study-specific populations or study area health events. Studies that provide
county/city-specific health counts across the study period include: Lepeule et al. (2012);
Kioumourtzoglou et al. (2016); Franklin et al. (2008); Zanobetti et al. (2014); Yap et al. (2013);
Ostro et al. (2016); and Weichenthal et al. (2016b). In U.S. studies for which health counts were
not provided, county-specific population data derived from the 2015 American Community
Survey data4 was used. For Canadian studies, city-specific population from 2016 Statistics
Canada5 was used.
In constructing the plots in Figure 3-9 and Figure B-9, several assumptions were made. In
studies that report mortality, hospital admissions data or emergency department visits, it was
assumed that the number of cases is directly proportional to the population of the area. To test
this assumption, census population data and case event data is used in a sensitivity analysis and
discussed in Section B.6. It was assumed that the population of a county did not change
substantially over time relative to other counties, and that the rank order is consistent over time
since only U.S. 2015 Census data and 2016 data from Statistics Canada was used. In studies that
state the study area is the entire U.S. {i.e. in Medicare studies), it was assumed that cases came
from each county of the U.S. (i.e., proportional to the county population 65 years or older for
Medicare studies) and therefore, air quality was used from all U.S. counties with data.
4Available from: https ://data.census, gov/cedsci/
5 Available from: https://www 12. statcan. gc.ca/census-recensement/2016/dp-pd/prof/index.cfm?Lang=E&TABID= 1
B-7

-------
Studies that had health data that started before 1999 in the U.S. and before 2000 in
Canada were excluded since U.S. and Canadian PM2.5 monitoring became more widespread
starting around these times. 29 studies met these criteria and are found in Figure 3-9 and Figure
B-9. Details on study-area assignment (Section B.4.1), population/health events assignment
(Section B.4.2), and air quality linkages (Section B.4.3) for studies included in the pseudo-design
value (DV) box plots are outlined below.
B.4.1 Study area assignment
The first step in developing Figure 3-9 and Figure B-9 was to identify the study area. The
U.S. based analysis is at the county-level and each U.S. county within the study area was
identified for each specific study. For the studies that provided city names, the U.S. cities were
used to identify all counties from the metropolitan area of that city, unless the entire city is
contained within a single county or unless otherwise noted. In cases of studies where the study
authors state that data was used for the entire U.S., all U.S. counties were included in the study
area assignment. For example, all counties were included in studies using Medicare or National
Center for Health Statistics (NCHS) data, unless the study identified a subset of cities or counties
included. For some studies, there are uncertainties related to how we chose counties to represent
study areas. Many studies identify the counties or cities used for the study; however, some only
said that they used HA or ED visit data from a specific state or region and didn't specify any
counties or cities. In those instances, we operated under the assumption that every county that
fell within the state or region identified contributed to the study population.6
For studies based in Canada, city was used as the geographic unit for the study area, since
Canadian air quality data is available at the city-level. In cases where a study notes that the study
is a national study, all cities for which air quality was available were included to define the study
area.
Studies were excluded from Figure 3-9 and Figure B-9 if the counties included are unclear
or not identified. Studies were also excluded in situations where the study population selection
criteria was not random and not likely to be proportional to the underlying population, or the
population selection criteria was not clearly specified (e.g., such as in cohort studies like the
American Cancer Society cohort (ACS), Nurses' Health Study cohort (NHS), and the Health
Professionals Follow-up Study (HPFS)).
B.4.2 Study population assignment
Based on the study areas identified in step 1, area-specific health events or populations
were then assigned to U.S. counties and Canadian cities. If the study reported health events for
6 As discussed below (section B.4.3), not all counties have PM2 5 monitor.
B-8

-------
U.S. counties or Canadian cities, we assigned those events to the specific counties or cities
identified. In the absence of reported health events at these geographic levels for studies where
hospital admissions or emergency department visits data, Medicare data, NCHS data, or other
national survey data was used, we assumed that study participants were randomly selected and
that the number of health events reported in the study was directly proportional to the population
of the area. For these studies, area-specific populations were assigned using U.S. 2015 American
Community Survey population data or 2016 Canadian population data (Statistics Canada).7 For
the remaining studies (i.e., for which the number of study participants or health events in each
location was not provided and for which the study population selection process appeared to not
be random or proportional to underlying populations), area-specific populations were not
assigned, and the studies were excluded from analysis.
In U.S. studies that evaluate cities, and for which some cities are associated with more
than one county, 2016 "City-to-County finder" data from Stats America8 was used to find the
proportional distribution of city population within each county, and the same proportional
distribution strategy was used to divide the reported health events between counties. An
example of the proportional distribution of city populations within counties is illustrated in Table
B-4, using a subset of cities reported in Zanobetti et al. (2014). Note, for cities not listed in Table
B-4, the city population was associated with one county and as a result, the health events for the
specific city were assigned to the corresponding county.
Table B-4. Percent of population by county associated with each city reported in the study
area.
City
Counties (% of population)
Atlanta, GA
Dekalb (6.7%), Fulton (93.3%)
Austin, TX
Travis (95.5%), Williamson (4.5%)
Columbus, OH
Franklin (97.9%), Fairfield (1.2%)
Dallas, TX
Dallas (93.9%), Collin (3.9%),
Denton (2.2%)
Fort Worth, TX
Tarrant (99%), Denton (1%)
Holland, Ml
Ottawa (78.8%), Allegan (21.2%)
Houston, TX
Harris (98%), Fort Bend (2%)
Lansing, Ml
Ingham (96%), Eaton (4%)
Middletown, OH
Butler (94.5%), Warren (5.5%)
7	While this approach contributes uncertainty to our analyses of pseudo-design values, we do not expect the rank
order of county population to substantially differ over the time periods of the studies and, therefore, we do not
expect this uncertainty to systematically bias our results.
8	Available from: http://www. statsamerica.org/Default.aspx
B-9

-------
New York, NY
Kings (30.6%), Queens (27.3%),
New York (19.4%), Bronx (16.9%),
Richmond (5.7%)
Oklahoma City, OK
Oklahoma (81.3%), Cleveland (11%),
Canadian (7.7%)
Tulsa, OK
Tulsa (98.4%), Osage (1.6%)
Charleston, SC
Charleston (93.3%), Berkeley (6.7%)
B.4.3 Air Quality data assignment by study area, by study period
The third step in developing Figure 3-9 and Figure B-9 was to assign air quality data by
study area, by study period. Ambient air quality data for PM2.5 in the United States and Canada
became more widely available across a broad proportion of the United States and Canada in the
late 1990s. To ensure a large proportion of air quality data points and subsequent 3-year design
values were available, the studies selected were those that examine air quality data starting in
1999 for U.S. studies and 2000 for Canadian studies. Construction of pseudo-design value box
plots (Figure 3-9 and Figure B-9) is described below. The air quality metric is termed a "pseudo-
design value", since both FRM/FEM monitors, as well as high quality non-FRM/FEM data, are
used to expand the number of areas with air quality data.9 Air quality data in the U.S. was
obtained from the EPA Air Quality System (AQS)10. For regulatory monitors, design values
were calculated using the data handling described by 40 CFR Appendix N to Part 50 -
Interpretation of the National Ambient Air Quality Standards for PM2.5. For non-regulatory data,
only monitors with 75% completeness for each of the 12 quarters in a 3-year design value period
were included. For Canadian air quality data, only sites with 75% completeness for each year of
the 3-year design value period were included.11 These criteria are slightly different than that of
actual design values, which have strict rounding conventions and substitution tests for sites with
less than 75% completeness for each quarter. For each given study and each previously identified
study area, each valid pseudo-DV was identified over each study period. For each county, or
city, the maximum PM2.5 pseudo-design value for each 3-year period of the study was identified.
Next, by county/city, the study-period average of the maximum pseudo-design value was
calculated ("average maximum pseudo-design value" or "average max pseudo-DV"). For each
study, locations were ordered by increasing average max pseudo-DVs and the corresponding
population or number of health events was used to calculate the cumulative percent of population
9 As noted in section B.5, sensitivity analyses using only regulatory FRM/FEM monitors gave similar results.
10Available from: https://www.epa.gov/aas
11 Available from: http://maps-cartes.ec. gc.ca/rnspa-naps/data.aspx?lang=en
B-10

-------
at or below each corresponding average max pseudo-DV. Next, the average max pseudo-DV
associated with the cumulative population closest to the 5th, 25th, 50th, 75111 and 95th percentiles
were identified. The actual cumulative percents that are closest to the 5th, 25, 50th, 75th, and 95th
percentiles, for all long- and short-term exposure studies and for annual and 24-hr PM2.5
concentrations, are illustrated in Figure B-3 and Figure B-4. The average max pseudo-DVs
associated with these percentiles in these studies are then presented in Table B-5 and Table B-6.
Counties that had no air quality monitors or no valid design values did not contribute to the
percentile calculation.
B-ll

-------
Percentile
Figure B-3. Cumulative population percentile closest to the 5th, 25th, 50, 75, and 95th
percentile: studies of long-term exposure and annual PM2.5 concentrations (top panel)
and studies of short-term exposure and annual PM2.5 concentrations (bottom panel).
B-12

-------
Percentile
Figure B-4. Cumulative population percentile closest to the 5th, 25th, 50, 75, and 95th
percentile: studies of long-term exposure and 24-hr PM2.5 concentrations (top panel)
and studies of short-term exposure and 24-hr PM2.5 concentrations (bottom panel).
B-13

-------
Table B-5. Annual average maximum pseudo-DVs corresponding to population or health
event percentiles in box-and-whisker plots in Figure 3-9.12
Citation
Pseudo DVs by percentiles
5th percentile
25th percentile
50th percentile
75th percentile
95th percentile
Baxter et al., 2017
7.53
11.86
14.63
16.70
21.95
Bell et al., 2008
8.55
11.35
13.72
15.94
23.05
Bell et al., 2014
12.43
12.43
13.30
13.40
16.47
Bell et al., 2015
8.18
10.81
12.81
15.31
20.95
Bravo et al., 2017
8.17
11.20
13.03
14.93
17.40
Dai et al., 2014
10.13
12.43
14.94
16.96
21.96
Di et al., 2017b
6.63
9.98
11.70
13.88
19.38
Di et al., 2017a
6.63
9.98
11.70
13.88
19.38
Dominici et al., 2006
9.15
12.05
14.10
17.00
24.70
Franklin et al., 2008
11.30
14.13
15.79
19.97
22.56
Kioumourtzoglou et al., 2016
8.49
10.86
13.36
15.70
20.50
Kloog et al., 2012
6.35
9.50
11.17
12.94
14.04
Kloog et al., 2014
11.10
12.44
13.77
15.22
16.96
Lee et al., 2015a
9.20
10.53
11.60
12.98
13.20
Lepeule et al., 2012
8.65
8.65
14.26
14.82
16.29
Malig et al., 2013
8.25
11.05
15.39
19.31
21.04
McConnell et al., 2010
10.50
16.30
16.30
20.56
24.11
Ostro et al., 2016
10.97
13.52
19.00
19.32
20.45
Peng et al., 2009
8.32
11.86
14.70
16.86
21.96
Pinaultet al., 2016
4.33
6.00
7.31
8.62
10.57
Shi et al., 2016
6.11
8.70
9.93
10.95
13.63
Urman et al., 2014
9.85
16.70
21.59
22.87
25.58
Wang et al., 2017
7.27
9.03
11.09
13.13
14.94
Weichenthal et al., 2016b
4.20
6.67
7.39
8.42
8.44
Weichenthal et al., 2016c
4.22
7.22
7.39
8.42
8.44
Yap et al., 2013
12.68
17.67
21.05
22.56
23.93
Zanobetti et al., 2009
11.60
14.15
16.90
22.30
24.00
12 As a sensitivity analysis, we also calculated study period averages of maximum design values using only
regulatory FRM/FEM monitors for Di et al. (2017a) and Di et al. (2017b) and Shi et al. (2016). Results were
similar to those based on the pseudo-design values using both regulatory and non-regulatory monitors. Using only
regulatory monitors for the studies by Di et al. (2017a) and Di et al. (2017b), 5th, 25th, 50th, 75th and 95th
percentiles of annual design values were 7.4, 9.7, 11.7, 13.9 and 17.6 |ig/m3, respectively. For these studies, 5th,
25th, 50th, 75th and 95th percentiles of 24-hour design values were 19, 26, 30, 36 and 49 |-ig/m3. respectively. For
Shi et al., 2016, 5th, 25th, 50th, 75th and 95th percentiles of annual design values were 7.7, 9.1, 10.4, 11.4 and 13.0
l_ig/m3, respectively while 5th, 25th, 50th, 75th and 95th percentiles of 24-hour design values were 21, 26. 29, 31 and
35 |-ig/m3. respectively.
B-14

-------
Zanobetti and Schwartz,
2009
9.72
12.18
14.43
17.30
23.05
Zanobetti et al., 2014
8.82
11.92
14.59
16.43
20.95
Table B-6. 24-hr average maximum pseudo-DVs corresponding to population or health
event percentiles in box-and-whisker plots in Figure B-9.
Citation
Pseudo DVs by percentiles
5th percentile
25th percentile
50th percentile
75th percentile
95th percentile
Baxter et al., 2017
22.00
31.00
38.67
45.50
58.33
Bell et al., 2008
19.20
30.34
36.40
42.67
62.20
Bell et al., 2014
34.67
34.67
37.67
40.00
40.33
Bell et al., 2015
21.23
28.10
33.56
39.57
55.78
Bravo et al., 2017
19.00
28.00
33.00
37.50
43.00
Dai et al., 2014
22.13
31.34
38.14
45.25
64.80
Di et al., 2017b
17.35
25.38
30.27
35.50
51.18
Di et al., 2017a
17.35
25.38
30.27
35.50
51.18
Dominici et al., 2006
22.00
31.00
37.50
44.50
68.00
Franklin et al., 2008
28.93
30.75
40.75
55.00
64.75
Kioumourtzoglou et al., 2016
20.22
29.72
34.38
40.07
54.05
Kloog et al., 2012
20.77
30.40
32.50
36.80
37.89
Kloog et al., 2014
30.00
34.00
37.20
39.50
45.60
Lee et al., 2015a
19.73
23.00
24.33
26.33
29.23
Lepeule et al., 2012
22.00
22.00
30.20
34.77
41.29
Malig et al., 2013
28.50
40.50
48.00
52.00
65.20
McConnell et al., 2010
23.00
47.00
47.00
56.00
65.00
Ostro et al., 2016
27.67
40.33
50.27
54.68
64.47
Peng et al., 2009
20.50
31.34
38.33
44.27
58.91
Pinaultet al., 2016
12.44
20.67
24.20
28.04
33.07
Shi et al., 2016
18.84
25.00
29.23
31.00
35.25
Urman et al., 2014
20.00
48.00
57.78
61.92
67.52
Wang et al., 2017
17.63
21.85
25.00
29.05
33.33
Weichenthal et al., 2016b
16.13
22.44
23.83
26.39
27.06
Weichenthal et al., 2016c
14.33
23.83
25.06
26.39
27.06
Yap et al., 2013
41.50
55.00
58.75
61.00
71.00
Zanobetti et al., 2009
28.00
38.50
43.50
63.00
72.50
Zanobetti and Schwartz,
2009
21.59
30.34
37.53
44.60
62.20
Zanobetti et al., 2014
22.67
31.11
37.91
41.25
55.78
B-15

-------
For each study in Figure 3-9 and Figure B-9, an assessment of the percent of 3-year
average pseudo-DVs available for each study area and study period is presented in Figure B-5.
For example, in a study with a study area of 5 counties that was completed for study a period
from 2000-2004, 3 possible 3-year average pseudo-DVs exist per county (i.e. 2000-2002, 2001-
2003, and 2002-2004), with a total of 15 possible pseudo-DVs. However, if one county only has
one valid 3-year average pseudo-DV, then the study would have 13 out of a possible 15 pseudo-
DVs. Figure B-5 displays a percent of 3-year average pseudo-DV data points available in each
study.
, 2017
Sen et si . 2022
Bel: et a!, 2214
Seis e: al . >015
Bravo et3> 201"
Carets! .2014
Di«: si . 201"
si ,201""
So^i-.-cietai ,2036
=rS--;;iir eta;., 200S
K;a-:C'j-ti03lou « ar , 20
K'ooj et si . 2012
-(iooa si , 2024
Us. et a:, 2022
LeoeUe et 3! , 2212
Mala et al., 2022
2'>cco—erl£t3* ,2012
Ost/e-et al., 2226
Peseta,, 200S
Pinault 2222
S>- J-si , 2225
, 2014
at , 2C2"
ttsicfte-thsletal.. 2022
vVercr-e-t^dle-ai , 221S
''•a;.etsr. 2213
2a'"osetti et a; , 222?
Is-iotieitte-t 2- , 2022
2s-KJ3stti et a: . 2021
¦ • ,42
100,00 •
83,02
68,54
8830 •
7 _
58,57
ft3,?6*
100,00 •
100,00 •
• 84,91
58.95
• 70,53
89.58 •
SO, 00 •
100,00 •
94,23 •
•	82,76
•	83,88
97
87,02 •
100,00 #
8S.48#
92 ..
95,25 •

97.00 •
30.34 •
• &4.3S
. :-Sl
85.71 <
13
97,06#
loom m
100.00 •
100.00
• 73,48
30.00 •
100,00 •
94.23 •
50,24 •
90,85 #
25 100
Figure B-5. Studies used in box-and-whisker plots (Figure 3-9 and Figure B-9) and the
percent of pseudo-DVs available by study.
B-16

-------
There are important uncertainties to consider when assigning air quality to a study area.
Pseudo-design values are based on individual monitors in each county included in study areas.
Counties may or may not reflect actual non-attainment areas, which can include multiple
counties or parts of counties. For studies conducted in Canada, this potential mismatch is of
greater concern. Pseudo-design values are not actual design values. Our analyses considered all
available monitoring data, even from monitors not meeting strict completeness requirements for
determining non-attainment. While we conclude this is a reasonable approach, as it allows the
consideration of ambient PM2.5 concentrations in a greater proportion of study areas than if the
analysis were restricted only to valid design values, it remains an uncertainty in our analyses.
Additional uncertainties are discussed above in section 3.2.3.2.2.
B.5 PERCENT OF STUDY AREA POPULATION CAPTURED IN DESIGN
VALUE PLOTS
Figure 3-9 and Figure B-9 include annual (Figure 3-9) and 24-hour (Figure B-9) pseudo-
design values corresponding to 5,25,50,75, and 95th percentiles of study populations or health
events for U.S. and Canadian studies of long-term or short-term exposures, and for studies of
mortality or morbidity outcomes. Further analyses were completed to determine the proportion
of the study area populations captured in these analyses. Within each study, the cumulative
population of counties with a valid 3-year average pseudo-DV was determined as a proportion of
the total population in counties included in the study. For example, if valid air quality data was
available in each county of the study area, then 100% of the study area population would be
captured within the design value box plots. For most studies included in Figure 3-9 and Figure
B-9, valid pseudo-DVs are available for counties accounting for at least about 70% of the total
study area population (Table B-7 and Table B-8).
When design values are calculated using only the regulatory monitors, as discussed in
section B.4.3 above, the total study area population captured in the calculation declines. For
example, for Di et al. (2017b) and Di et al. (2017a), when calculation of design values was
completed using air quality data only from regulatory monitors, the analyses captured 67.35%) of
population for annual design values (compared to 70.38%> of population for annual pseudo-
design values when data from all monitors was used). Similarly, analyses captured 67.43%> of
population for 24-hour design values from regulatory monitors alone, compared to 70.47%) of
population for pseudo-design values when data from all monitors was used. For Shi et al. (2016),
calculation of annual and 24-hour design values from regulatory monitors captured 11.31% of
population, compared to 11.22% of population when data from all the monitors was used.
B-17

-------
Table B-7. Percent population included in annual pseudo-DV boxplots (Figure 3-9).
Citation
Population Used
Study Area
Counties
Total Population
Population with
DV
Population
with DV
(%)
Baxter etal., 2017
US 2015
113
113,053,365
100,129,153
88.57
Bell et al., 2008
US 2015 (65+yrs)
202
23,206,934
21,974,015
94.69
Bell et al., 2014
US 2015 (65+yrs)
4
490,357
490,357
100.00
Bell et al., 2015
US 2015 (65+yrs)
202
23,206,934
22,529,386
97.08
Bravo et al., 2017
US 2015 (65+yrs)
807
31,056,109
21,909,224
70.55
Dai et al., 2014
US 2015
95
95,890,830
91,262,160
95.17
Diet al., 2017b
US 2015 (65+yrs)
3220
48,387,814
34,057,020
70.38
Di et al., 2017a
US 2015 (65+yrs)
3220
48,387,814
34,057,020
70.38
Dominici et al., 2006
US 2015 (65+yrs)
202
23,206,934
20,272,093
87.35
Franklin et al., 2008
Franklin 2008
25
1,313,983
1,313,983
100.00
Kioumourtzoglou et
al., 2016
Kiomourtzoglou 2016
222
11,391,912
11,050,835
97.01
Kloog et al., 2012
US 2015 (65+yrs)
67
2,361,375
1,588,345
67.26
Kloog et al., 2014
US 2015 (65+yrs)
366
9,099,500
6,471,367
71.12
Lee et al., 2015a
US 2015
305
25,153,808
14,033,573
55.79
Lepeule et al., 2012
Lepeule 2012
11
14,562
12,932
88.81
Malig et al., 2013
US 2015
35
36,607,640
36,533,148
99.80
McConnell et al.,
2010
US 2015 (18 and
under)
7
5,008,800
5,008,587
100.00
Ostro et al., 2016
Ostro Asthma 2016
8
43,904
43,904
100.00
Peng et al., 2009
US 2015 (65+yrs)
119
13,944,304
13,732,109
98.48
Pinault et al., 2016
Canada 2016
5162
35,151,728
18,242,308
51.90
Shi et al., 2016
US 2015 (65+yrs)
67
2,361,375
1,823,456
77.22
Urman et al., 2014
Urman 2014 5-7yrs
5
1,811
1,811
100.00
Wang et al., 2017
US 2015 (65+yrs)
616
9,779,426
6,336,200
64.79
Weichenthal et al.,
2016b
Weichenthal Ml 2016
16
30,101
30,101
100.00
Weichenthal et al.,
2016c
Canada 2016
15
4,673,938
4,673,938
100.00
Yap et al., 2013
Yap 2013 Asthma 1-
9yrs
12
146,224
146,224
100.00
Zanobetti et al., 2009
US 2015 (65+yrs)
35
6,630,577
5,974,387
90.10
Zanobetti and
Schwartz, 2009
US 2015
156
126,026,116
114,529,073
90.88
Zanobetti et al., 2014
Zanobetti 2014
126
6,828,055
6,703,284
98.17
B-18

-------
Table B-8. Percent population included in 24-hr pseudo-DV boxplots (Figure B-9).
Citation
Population Used
Study Area
Counties
Total Population
Population with
DV
Population
with DV (%)
Baxter etal., 2017
US 2015
113
113,053,365
97,125,414
85.91
Bell et al., 2008
US 2015 (65+yrs)
202
23,206,934
21,903,002
94.38
Bell et al., 2014
US 2015 (65+yrs)
4
490,357
490,357
100.00
Bell et al., 2015
US 2015 (65+yrs)
202
23,206,934
22,564,564
97.23
Bravo et al., 2017
US 2015 (65+yrs)
807
31,056,109
21,083,502
67.89
Dai et al., 2014
US 2015
95
95,890,830
91,262,160
95.17
Diet al., 2017b
US 2015 (65+yrs)
3220
48,387,814
34,097,655
70.47
Di et al., 2017a
US 2015 (65+yrs)
3220
48,387,814
34,097,655
70.47
Dominici et al., 2006
US 2015 (65+yrs)
202
23,206,934
20,097,018
86.60
Franklin et al., 2008
Franklin 2008
25
1,313,983
1,313,983
100.00
Kioumourtzoglou et
al., 2016
Kiomourtzoglou 2016
222
11,391,912
11,050,835
97.01
Kloog et al., 2012
US 2015 (65+yrs)
67
2,361,375
1,546,500
65.49
Kloog et al., 2014
US 2015 (65+yrs)
366
9,099,500
6,429,318
70.66
Lee et al., 2015a
US 2015
305
25,153,808
12,127,123
48.21
Lepeule et al., 2012
Lepeule 2012
11
14,562
12,932
88.81
Malig et al., 2013
US 2015
35
36,607,640
35,908,846
98.09
McConnell et al.,
2010
US 2015 (18 and
under)
7
5,008,800
5,008,587
100.00
Ostro et al., 2016
Ostro Asthma 2016
8
43,904
43,904
100.00
Peng et al., 2009
US 2015 (65+yrs)
119
13,944,304
13,596,370
97.50
Pinault et al., 2016
Canada 2016
5162
35,151,728
18,242,308
51.90
Shi et al., 2016
US 2015 (65+yrs)
67
2,361,375
1,823,456
77.22
Urman et al., 2014
Urman 2014 5-7yrs
5
1,811
1,811
100.00
Wang et al., 2017
US 2015 (65+yrs)
616
9,779,426
6,306,215
64.48
Weichenthal et al.,
2016b
Weichenthal Ml 2016
16
30,101
30,101
100.00
Weichenthal et al.,
2016c
Canada 2016
15
4,673,938
4,673,938
100.00
Yap et al., 2013
Yap 2013 Asthma 1-
9yrs
12
146,224
146,224
100.00
Zanobetti et al., 2009
US 2015 (65+yrs)
35
6,630,577
5,974,387
90.10
Zanobetti and
Schwartz, 2009
US 2015
156
126,026,116
114,529,073
90.88
Zanobetti et al., 2014
Zanobetti 2014
126
6,828,055
6,703,284
98.17
B-19

-------
B.6 SENSITIVITY ANALYSIS: BOX PLOTS USING COUNTS OF
HEALTH EVENTS VERSUS STUDY AREA POPULATION
As discussed in Section 3.2.3.2.2, Figure 3-9 and Figure B-9 present box-and-whisker
plots reflecting the PM2.5 3-year average maximum pseudo-design values that correspond to
various percentiles of the study area population or study area health events. When area-specific
health events are available, Figure 3-9 and Figure B-9 present percentiles of air quality and study
area health events. There is uncertainty regarding the extent to which the populations in counties
included in key studies reflect the true distribution of cases in those studies. Many studies used
registry data, or similar data sources that may be expected to capture the majority of cases within
a study location; however, these studies often didn't report the exact number of cases per area.
When the number of cases were not available, we instead used the underlying county-level
population obtained using 2015 U.S. census data. While this approach contributes uncertainty to
our analyses of pseudo-design values, for the limited number of studies with information on the
number of cases per county, the distributions of pseudo-design values relative to the number of
cases were similar to the distributions relative to the county population (particularly for annual
pseudo-design values). Figure B-6 provides a comparison of studies where health event data are
available, to assess the distribution of pseudo-design values when study area population is used
versus study area health events.
B-20

-------
Country Enc,P°'nt citation LT/STstuc1^ Geographic Areas Population
Group	• Years
; la t,
Canada
~ : -
i. a . , i.i.uC'
------ ~ : - -
- - - - ~ ~ -
ST 2005-2009 3 CaliforniaCour
. ... ----- " " =
U C Z01= 54
IJ S I01E 54
i	r::;:p	1
i—	1
i—-Hcztzi———I
i—	:—i
h:::;z;:ii::z;:::;i	1
h r i i—i
i-4-hzizzizj	1
i—j—fzzzrzi	1
i—j—i i i	1
i—!—	1
j i	tin	1
20 30 40 50 60 70
Country	Citation LT/ST	Geographic Areas Population
Group
- -- :
ears
' —
- - - - -I- -
et a:... 20i6
-=

Kiorpourtzoglou 2016
U.S. 20X5 >64
.Cities	U.S. 2015>64
Zanobetti 2014
Cities	Franklin ZOOS
USZOISAj
i'ai ST 20r?-I'jC? i lalrt:! i c lc jrt.e; C»t>4=th "a ZCle
i -"	:	-	
US 2015 Ali
. . .. : . - . -: -	;
-^	G-s: ro ¦!- di u 16
US 2015 Ah'
I -I _ I !	-i
!— f ; i s-	i
-i ¦ i i	.-i
i —i I I	— I
f I
I i
h
r —|
« I
9 I
» t
» I
IH
IH
¦!- - 4
I M
! h-1
5 1C 15 2l 25
Avg. Max PseudoDV
Figure B-6. PM2.5 pseudo-design values corresponding to various percentiles of study area
populations and health events for studies of 24-hour PM2.5 exposures and long-term
studies (top panel) and annual PM2.5 exposures and long-term studies (bottom panel).
B.7 COMPARISONS BETWEEN ANNUAL AND DAILY DESIGN
VALUES
As discussed above in section 3.2.3.2, for an area to meet the NAAQS, all valid design
values in that area, including the highest annual and 24-hour values, must be at or below the
levels of the standards. Because monitors are often required in locations with high PM2.5
concentrations (section 2.2.3), areas meeting an annual PM2.5 standard with a particular level
would be expected to have long-term average PM2.5 concentrations (i.e., averaged across space
and over time in the area) somewhat below that standard level. Figure B-7 and Table B-9
B-21

-------
indicate that, based on recent air quality in U.S. CBS As, maximum annual PM2.5 design values
are often 10% to 20% higher than annual average concentrations (i.e., averaged across multiple
monitors in the same CBSA). The difference between the maximum annual design value and
average concentration in an area can be smaller or larger than this range, likely depending on
factors such as the number of monitors, monitor siting characteristics, and the distribution of
ambient PM2.5 concentrations. Given that higher PM2.5 concentrations have been reported at
some near-road monitoring sites, relative to the surrounding area (section 2.3.2.2.2), recent
requirements for PM2.5 monitoring at near-road locations in large urban areas (section 2.2.3.3)
may increase the ratios of maximum annual design values to averaged concentrations in some
areas. Such ratios may also depend on how the average concentrations are calculated (i.e.,
averaged across monitors versus across modeled grid cells). Compared to annual design values,
Figure B-8 indicates a more variable relationship between maximum 24-hour PM2.5 design
values and annual average concentrations.
B-22

-------
•• •
• m/
• fc /
• V
/ 1:1 line
Other (AK, HI)
•	IndustMidwest
•	Northeast
•	Northwest
•	SoCal
•	Southeast
•	Southwest
•	UpperMidwest
l	i	i	I	i	I	i	I
1	I
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
CBSA Average 2015-2017 Annual Design Value (|jg nr3)
Figure B-7. Comparison of CBSA average annual design values and CBSA maximum
annual design values for 2015-2017. (Note: Includes all CBSAs with at least 3 valid annual
DVs.)
B-23

-------
Table B-9. National Averages of ratios of maximum annual design values to averaged
concentrations.
Year of
monitoring data
Number of monitors
per CBSA
Number
of CBSAs
Ratio of max Annual
DV to CBSA average
Ratio of max 24-hr
DV to CBSA average
2009-2011
3 or more
67
1.12
1.13
4 or more
33
1.14
1.16
5 or more
18
1.17
1.19
2012-2014
3 or more
60
1.15
1.15
4 or more
38
1.17
1.18
5 or more
23
1.19
1.21
2015-2017
3 or more
65
1.16
1.19
4 or more
38
1.19
1.21
5 or more
30
1.20
1.24
B-24

-------
E
01
D1 40 -
CT
V)
(U
Q
ru
O
O
r\i
ld
l—I
O
r\j
E
=5
E
x
03
<
CO
CO
u
60 -i
55 -
50

25 -
20 -
15 -
10 -
5 -
Other (AK, HI)
IndustMidwest
Northeast
Northwest
SoCal
Southeast
Southwest
UpperMidwest
35 	
30 -
• •
• •• •
•m
• •
• • •
• •
~r
4
~1	1	1	1	1	1	1	1	1	I	1
8 9 10 11 12 13 14 15 16 17 18
CBSA Average 2015-2017 Annual Design Value (|jg rrr3)
Figure B-8. Comparison of CBSA average annual design values and CBSA maximum daily
design values for 2015-2017. (Note: Dashed lines indicate the level of the current 24-hour
PM2.5 standard (35 ug m/3) and the current annual PM standard (12 ug m/3). Includes all
CBSAs with at least 3 valid daily and 3 valid annual DVs.)13.
13 The CBSA maximum 2015-2017 daily design value (y-axis) was cut off at 60 |ig/ml to improve the visualization
of data, but this removed the Fairbanks CBSA from the plot, which had a daily design value of 85 |ig/m3 and an
annual design value of 15.7 |ig/ml
B-25

-------
B.8 24-HOUR PSEUDO-DESIGN VALUES AND DISTRIBUTIONS
ACROSS STUDY AREAS
As described in section 3.2.3.2.2 of the PA, and section B.4 of this appendix, for
locations evaluated in key epidemiologic studies we identify annual and 24-hour PM2.5 pseudo-
design values and the number of people (or health events). Figure 3-9 in the PA presents box-
and-whisker plots summarizing those data for annual pseudo-design values. Figure B-9 (below)
presents box-and-whisker plots summarizing those data for 24-hour pseudo-design values.
B-26

-------
Long-term exposure studies
Country
Endpoint
Group
Citation
Study
Years
Geographic Areas




U.S.
Mortality
Kiomourtzoglou et a!2016*
2000-2010 207 U.S. Cities
1	1
1 1	
	1

Di et al., 2017b*
2000-2012 U.S. Nationwide
1	1 I

	1

Lepeule et a!., 2012*
Shi et al., 2016*
Wang et al., 2017*
2001-2009 6 U.S. Cities
2003-2008 6 NE U.S. States
2000-2013 7 SE U.S. States
1 1
I	1







Morbidity
Urman et al., 2014*
2002-2007 8 CA Counties
1	

—I I I-
—1
Mcconnell et al., 2010
2003-2005 13 CA Communities
1	


-1
Canada
Mortality
Pinault etal., 2016*
2000-2011 Multicity 1-
	1 1 1	
l


Short-term exposure studies
Endpoint
Country ^ Citation
Group
Study
Years
10
Geographic Areas
20 30
40
50 60
70
U.S.
Mortality
Morbidity
Franklin etal., 2008*
2000-2005
25 U.S. Cities
H 1 1	1
Baxter etal., 2017*
2001-2005
77 U.S. Cities
i	1 ; i i	1
Dai etal., 2014*
2000-2006
75 U.S. Cities
i	1 i i	1
Zanobetti et al., 2014*
1999-2010
121 U.S. Cities
i	1 i i	1
Zanobetti and Schwartz, 200
.. 1999-2005
112 U.S. Cities
i	1 ¦ i i	1
Di etal., 2017a*
Shi etal., 2016*
Lee etal., 2015b*
Yap et al, 2013*
2000-2012
2003-2008
2007-2011
2000-2005
U.S. Nationwide
6 NE U.S. States
3 SE U.S. States
CA (Central & Southern Counties)
i	1 r
HTH
i—
'l	
1	
—1
	[ I 1	
-1
Ostroetal., 2016*
2005-2009
8 CA Counties

1 1	1

Malig et al., 2013*
2005-2008
35 CA Counties
I	:—l I I	1
Zanobetti et al., 2009*
2000-2003
26 U.S. Cities
l	¦—l l I	1
Peng et al., 2009*
Bell etal., 2014*
2000-2006
2000-2004
119 U.S. Urban Counties
4 U.S. Counties, MA&CT
I	1 l l	1
to
Dominici etal., 2006*
Kloog et al., 2014*
1999-2002
2000-2006
204 U.S. Urban Counties
7 U.S. Mid-Atlantic States & D.C.
1	c
i	r
^ 7
	1

Bell etal., 2008*
1999-2005
202 U.S. Urban Counties
—i—
	1



Bell et al.r 2015'
Bravo etal., 2017*
1999-2010
2002-2006
213 U.S. Urban Counties
708 U.S. Counties
i—i i i	1
1	1 [ 1	!


Kloog et al., 2012*
2000-2006
6 NE U.S. States
1	~
CH


Canada
Morbidity
Weichenthal et al., 2016c*
2004-2011
15 Ontario Cities
1	un :


Weichenthai et al., 2016b
2004-2011
16 Ontario Cities
1	CO :
20 30 40 50 60 70
Avg. Max PseudoDV
Figure B-9. P1VI2.5 24-hour pseudo-design values corresponding to various percentiles1'1 of
study area populations or health events for studies of long-term and short-term PM2.5
exposures.^21
[1] Whiskers reflect PM2.5 pseudo-design values corresponding to 5th and 95lh percentiles of study area populations
(or health events), boxes correspond to the 25th and 75th percentiles, and the vertical lines inside the boxes
correspond to 50th percentiles. Asterisks next to study citations denote statistically significant effect estimates.
M For most of the studies included in Figure B-9, pseudo-design values are available for >70% of study area
populations (or health events). Exceptions are Kloog et al. (2012), Lee et al. (2015b), Pinault et al. (2016), Wang
et al. (2017), and Bravo et al. (2017), with pseudo-design values available for 65%, 48%, 51%, 68%, and 64% of
study area populations, respectively.
B-27

-------
B.9 PSEUDO-DESIGN VALUE DISTRIBUTION BY AVERAGE COUNTY
PSEUDO-DESIGN VALUES PER 1 jiG/M3
Figure 3-9 and Figure B-9 exhibit distributions of pseudo-DVs corresponding to study
areas within each study and based on averaging pseudo-DVs. That is, for each study location,
maximum 3-year pseudo-design values are averaged over study periods. Depending on the years
of air quality evaluated by the study, for some locations those averages could reflect air quality
that violated the current standards during part of the study period and met the current standards
during part of the study period. We have examined this issue in greater detail for the studies by
Di et al. (2017b) and Shi et al. (2016).
Figure B-10 and 0 present the relationship between annual pseudo-DVs averaged over the
study period and the individual 3-year pseudo-DVs that contribute to those study-period averages
for Di et al. (2017b). Of the 6,315 3-year pseudo-DVs available for this study, 3,915 (62%) are
less than or equal to 12.04 |ig/m3 (i.e., lower than the current annual standard). Of the counties
that have study-period average pseudo-DV's < 12.04 |ig/m3, 89.3% of individual 3-year pseudo-
DVs are < 12.04 |ig/m3 (i.e., 3,410 of 3,820 3-year pseudo-DVs).
B-28

-------
30
26
24
sr
E
20
¦S 18
16
12
B 10
II
I
I
II
lil I
11!

I
I

i
1
IIH
i
|i
I
i
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 IS 19 20 21 2;
Study period county average PM2.5 pseudo-design value (ng/m3)
Figure B-10. County average pseudo-DV by 1 jig/m3 and distribution of individual county
pseudo-DVs within each 1 jig/m3 interval for study counties in Di et al., 2017b. Note: X-
axis values of 11 correspond to county average pseudo-DVs from 11.0 to 12.0 jig/m3.
Thus, x-axis values of 11 or below correspond to pseudo-DVs at or below the level of the
current annual standard.
B-29

-------
Table B-10. County average pseudo-DV by 1 jig/m3 and distribution of county pseudo-DVs
within each 1 jig/m3 interval for study counties in Di et al., 2017b
County average pseudo-DV PM2.5
concentration (mq/ti3) interval
Count (percent) of pseudo-DV's <
12.04 |jg/m3
Count (percent) of pseudo-DV's >
12.04 |jg/m3
2.04 < PM2.5 ^ 3.04
93 (100.00)
0 (0.00)
3.04 < PM2.5 ^ 4.04
117 (100.00)
0 (0.00)
4.04 < PM2.5 ^ 5.04
198 (100.00)
0 (0.00)
5.04 < PM2.5 ^ 6.04
235 (100.00)
0 (0.00)
6.04 < PM2.5 ^ 7.04
293 (99.35)
2 (0.68)
7.04 < PM2.5 ^ 8.04
283 (100.00)
0 (0.00)
8.04 < PM2.5 ^ 9.04
501 (100.00)
0 (0.00)
9.04 < PM2.5 ^ 10.04
533 (99.84)
1 (0.19)
10.04 < PM2.5 2 11.04
619 (92.23)
61 (8.97)
11.04 < PM2.5 ^ 12.04
538 (66.03)
346 (39.14)
12.04 < PM2.5 ^ 13.04
332 (30.46)
635 (65.67)
13.04 < PM2.5 ^ 14.04
128 (13.19)
525 (80.40)
14.04 < PM2.5 ^ 15.04
38(5.14)
433(91.93)
15.04 < PM2.5 ^ 16.04
7(1.27)
228 (97.02)
16.04 < PM2.5 ^ 17.04
0 (0.47)
70 (100.00)
17.04 < PM2.5 ^ 18.04
0 (0.00)
21 (100.00)
18.04 < PM2.5 ^ 19.04
0 (0.00)
11 (100.00)
19.04 < PM2.5 ^ 20.04
0 (0.00)
33 (100.00)
20.04 < PM2.5 ^ 21.04
0 (0.00)
12 (100.00)
21.04 < PM2.5 ^ 22.04
0 (0.00)
11 (100.00)
22.04 < PM2.5 ^ 23.04
0 (0.00)
11 (100.00)
Total
3,915(62.0)
2,400 (38.0)
Figure B-l 1 and Table B-l 1 present the relationship between annual pseudo-DVs
averaged over the study period and the individual 3-year pseudo-DVs that contribute to those
study-period averages for Shi et al. (2016). Of the 116 3-year pseudo-DVs available for this
study, 102 (88%) are less than or equal to 12.04 |ig/m3. Of the counties that have study-period
average pseudo-DV's < 12.04 |ig/m3 98.1% of individual 3-year pseudo-DVs are < 12.04 |ig/m3
(i.e., 102 of 104 3-year pseudo-DVs).
B-30

-------
6	7	8	9	10	11
Study period county average PM2.5 pseudo-design value (pg/m3)
Figure B-ll. County average pseudo-DV by 1 jig/m3 and distribution of individual county
pseudo-DVs within each 1 jig/m3 interval for study counties in Shi et al., 2016. Note: X-
axis values of 11 correspond to county average pseudo-DVs from 11.0 to 12.0 jig/m3.
Thus, x-axis values of 11 or below correspond to pseudo-DVs at or below the level of the
current annual standard.
B-31

-------
Table B-ll. County average pseudo-DVs by 1 jig/m3 and distribution of county pseudo-
DVs within each 1 jig/m3 interval for study counties in Shi et al., 2016.
County average pseudo-DV PM2.5
concentration (mq/ti3) interval
Count (percent) of pseudo-DV's <
12.04 |jg/m3
Count (percent) of pseudo-DV's >
12.04 |jg/m3
4.04 < PM2.5 ^ 5.04
8 (100.00)
0 (0.00)
5.04 < PM2.5 ^ 6.04
5 (100.00)
0 (0.00)
6.04 < PM2.5 ^ 7.04
7 (100.00)
0 (0.00)
7.04 < PM2.5 ^ 8.04
16 (100.00)
0 (0.00)
8.04 < PM2.5 ^ 9.04
12 (100.00)
0 (0.00)
9.04 < PM2.5 ^ 10.04
26 (100.00)
0 (0.00)
10.04 < PM2.5 2 11.04
21 (95.45)
1 (0.00)
11.04 < PM2.5 ^ 12.04
7 (87.50)
1 (0.00)
12.04 < PM2.5 ^ 13.04
0 (0.00)
4 (0.00)
13.04 < PM2.5 ^ 14.04
0 (0.00)
8 (0.00)
Total
102 (88.0)
14(12.0)
B-32

-------
B.10 DETAILS OF KEY EPIDEMIOLOGIC STUDIES, INCLUDING STUDY DESIGN, EXPOSURE
METRIC, AND STATISTICAL ANALYSIS
Table B-12 below summarizes additional details related to the designs of the U.S. and Canadian epidemiologic studies
included in Figure 3-7, Figure 3-8, Figure 3-9, and Figure B-9 and the risk assessment (Table 3-4).
Table B-12. Study characteristics from key studies.
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Baxter etal.,
2017
ST
All-cause
mortality
77 US Cities
Time Series study
(NCHS data)
Average daily
monitored PM2.5
concentration in each
city. 2-day moving
average (lag 0-1 days)
of PM2.5 conc. Included
in the model.
Poisson regression model and
meta-regression
In stage 1, ran single city Poisson
time-series models; adjusted for
temperature and dew point
temperature, including variables for
previous day temperature, temporal
trends, and trends by age.
In stage 2, meta-regression with
cluster analysis (5 clusters) based
on characteristics of residential
infiltration.
B-33

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Bell et al., 2008
ST
CVD HA Age
65+
202 US Counties with
populations>200,000
Time Series study
(MEDICARE
enrollees)
Daily monitored PM2.5
concentrations. Used
lagO PM2.5 in the
model.
2-stage Bayesian hierarchical model
In stage 1, adjusted for temperature
and dew point temperature,
including variables for previous
day's conditions, day-of-the-week,
temporal trends, and differential
temporal trends by age. In stage 2,
county-specific estimates were
combined, accounting for their
statistical uncertainty.
B-34

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Bell et al., 2014
ST
CVD, Asthma,
and COPD HA
Age 65+
4 Counties in MA and
CT
Time-series study
(MEDICARE
enrollees)
PM2.5 Teflon filter
samples obtained from
CT and MA DEP and
used to measure PM2.5
total mass.
Fairfield County (2
monitors): Estimated
exposures using
population-weighted
averaging of values and
assigned exposure to
the nearest monitor.
Exposures were
averaged, weighted by
each tracts' 2000
census population. For
other counties, values
from the single monitor
within the county were
used. Explored various
lags and presented lagO
PM2.5 model.
Log-linear Poisson regression
analysis
Adjusted for temperature and dew
point temperature, including
previous day's temperature and dew
point temperature, day-of-the-week
temporal trends, and region.
B-35

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Bell et al., 2015
ST
HF HA 65+
213 U.S. Counties
Time-series study
(MEDICARE
enrollees)
For each county and
day, PM2.5
measurements for
monitors within a county
were averaged.
Explored various lags
and presented lagO
PM2.5 model.
2-stage Bayesian hierarchical model
The stage 1 model included county-
specific model adjusted for weather
(temperature, dew point, previous
days' temperature and dew point),
day-of-the-week, and temporal
trends. In stage 2 county-specific
effect estimates were pulled
together to present overall
association.
B-36

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Bravo et al.,
2017
ST
CVD HA Age
65+
418 U.S. Counties
Time-series study
(MEDICARE
enrollees)
Exposure estimated
from monitoring data
and monitors with
multiple measurements
for the same day and
county were averaged.
Explored various lags
and distributed lags of
PM2.5 exposure.
2-stage Bayesian hierarchical model
The stage 1 included log-linear
Poisson regression models with
over-dispersion fit at county-level.
Model adjusted for same-day
temperature and dew point
temperature, 3-day moving average
of temperature and dew point
temperature, temporal trends in
hospitalizations, day-of-the-week,
and age. Fitted distributed lag model
with multiple lags (0- to 7-day lags)
of PM2.5 cone simultaneously in the
county-specific model.
The stage 2 estimated the
association for the entire study area
using two-level normal independent
sampling estimation with priors thus
allowing to combine risk estimates
across counties while accounting for
within county SE and between-
county variability in the true RR.
B-37

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Bravo et al.,
2017
ST
CVD HA Age
65+
708 U.S. Counties
Time-series study
(MEDICARE
enrollees)
Daily PM2.5
concentrations
estimated at census
tracts using the
downscaler method. 24-
hr county-level PM2.5
estimates for counties
with population >
50,000 were calculated
from a population-
weighted average of
PM2. Concentrations
predicted by the
downscaler at census
tracts within each
county using 2000 U.S>
Census Data. Explored
various lags and
distributed lags of
PM2.5 exposure.
2-stage Bayesian hierarchical model
The stage 1 included log-linear
Poisson regression models with
over-dispersion fit at county-level.
Model adjusted for same-day
temperature and dew point
temperature, 3-day moving average
of temperature and dew point
temperature, temporal trends in
hospitalizations, day-of-the-week,
and age. Fitted distributed lag model
with multiple lags (0- to 7-day lags)
of PM2.5 cone simultaneously in the
county-specific model.
The stage 2 estimated the
association for the entire study area
using two-level normal independent
sampling estimation with priors thus
allowing to combine risk estimates
across counties while accounting for
within county SE and between-
county variability in the true RR.
B-38

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Burnett and
Goldberg, 2003
ST
All-cause
mortality
8 Canadian Cities
Time-series study
Monitored
measurements
Generalized additive model (GAM)
analysis to generate pooled
estimate of air pollution effect
among the eight cities.
The model adjusted for day-of-the-
week, temporal trends, and weather
variables (daily average
temperature, daily average relative
humidity, and barometric pressure
lagged 0 and 1 days).
Burnett et al.,
2004
ST
All-cause
mortality
12 Canadian Cities
Time-series study
(data from Statistics
Canada)
Daily summary pollution
exposure
measurements based
on averaging data over
all monitors within each
city. Various lags and
moving average
assessed and
presented data for lag 1
for PM2.5.
Random-effects regression model.
Adjusted for temporal trends in
mortality and effects of weather
using humidex index at lag 0 and lag
1 (a measure of combined effect of
temperature and humidity)
B-39

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Crouse et al.,
2012
LT
All-cause
mortality
11 Canadian Cities
National Cohort
study (Subset of
Canadian census
mortality follow-up
study; 43%)
Mean annual
concentration from
ground-based monitors
averaged from 1987-
2001. Participants were
assigned exposure
based on 11 census
divisions.
Another set of exposure
estimate was derived
from satellite remote
sensing for period
2001-2006. Estimates
at grid-level assigned to
the cohort members by
linking grid to the
enumeration area of
residence in 1991.
2 different modelling approach.
Approach 1: Cox proportional
hazards model, and Approach 2:
nested, spatial random-effects Cox
model with spatial clusters.
Models adjusted for individual-level
covariates, urban/rural indicator, and
ecological covariates (%
unemployed, % without high school
diploma, lowest income quintile, and
rural/urban indicator).
Dai et al., 2014
ST
All-cause, CVD,
and Respiratory
mortality
75 U.S. Cities (with
available daily
mortality data and
PM2.5 data for at
least 400 days
between 2000 and
2006)
Time-series study
(NCHS)
Mean daily monitored
PM2.5 concentrations.
For cities with more
than one sampling site,
concentration data were
averaged. Average of 2-
day lag (lag 0 and 1)
PM2.5 used.
Two stage: Stage 1. City-specific
season-stratified time-series
analysis using Poisson regression in
GAM
Model adjusted for 24-hr average
temperature from closest weather
station to the city center at lagO and
Iag1, temporal trends, and day-of-
the-week. Stage 2. Multivariate
random effects meta-analysis to
combined 300 (i.e. 75 cities * 4
seasons) effect estimates to obtain
overall association.
B-40

-------
Diet al.,2017b
LT
All-cause
mortality 65+
US Nationwide
Di et al., 2017b
(< 12 ug/m3)
Open Cohort
(MEDICARE
enrollees)
Artificial neural network
that incorporated
satellite-based
measurements,
simulation outputs from
a chemical transport
model, land-use terms,
meteorological data,
and other data to
predict daily
concentrations of PM2.5.
The neural network was
fit with monitored PM2.5
data and daily PM2.5
concentrations were
predicted for nationwide
grids that were 1x1 km.
For each calendar year
during which a person
was at risk of death the
annual average PM2.5
concentration was
assigned according to
the ZIP Code of the
person's residence. As
part of a sensitivity
analysis, monitored
PM2.5 data was
matched with each
person in the study
within a distance of 50
km of the nearest
monitoring site.
Analysis restricted to
persons-years with
Two-pollutant Cox proportional
hazards model with generalized
estimating equation to account for
correlation between ZIP codes.
Accounted for individual variables,
(sex, race, Medicaid eligibility, and
average age at study entry), zip
code-level variables (% Hispanic, %
Black, median household income,
median value of housing, % > 65
living below poverty level, % > 65
with less than high school
education, % of owner-occupied
housing units, and population
density), county-level variables
(county-level BMI and % ever
smokers), hospital service area-level
variables (% low-density lipoprotein
level measured, % glycated
hemoglobin level measured, and %
>1 ambulatory visits), 32 km2
gridded weather and 1 km2 gridded
pollution variables (annual average
PM2.5 concentration, annual average
temperature, and annual average
humidity), monitor level air pollution
variables (PM2.5 monitored data),
and a regional dummy variable.
B-41

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed





PM2.5 exposures lower
than 12 ug/m3

B-42

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Di et al., 2017a
ST
All-cause
mortality 65+
US Nationwide
Case-crossover
study (MEDICARE
enrollees)
Artificial neural network
that incorporated
satellite-based
measurements,
simulation outputs from
a chemical transport
model, land-use terms,
meteorological data,
and other data to
predict daily
concentrations of PM2.5.
The neural network was
fit with monitored PM2.5
data and daily PM2.5
concentrations were
predicted for nationwide
grids that were 1x1 km.
For each case day
(date of death) and its
control days, the 24-
hour PM2.5
concentrations were
assigned based on zip
code of residence of the
individual. As part of a
sensitivity analysis,
monitored PM2.5 data
was matched with each
person in the study
within a distance of 50
km of the nearest
monitoring site.
Conditional logistic regression.
"Case Day" defined as death. For
the same person, compared daily air
pollution exposure on the case day
vs. daily air pollution exposure on
"control days." Control days were
chosen (1) on the same day of the
week as the case day to control for
potential confounding effect by day
of week; (2) before and after the
case day to control for time trend;
and (3) only in the same month as
the case day to control for seasonal
and subseasonal patterns.
Individual-level covariates and zip
code-level covariates that did not
vary day to day (e.g., age, sex,
race/ethnicity, SES, smoking, and
other behavioral risk factors) were
not considered to be confounders as
they remain constant when
comparing case days vs control
days.
The regression model adjusted for
air and dew point temperature.
B-43

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Dominici et al.,
2006
ST
HF and COPD
HA 65+
204 Urban U.S.
counties
Time-series study
(MEDICARE
enrollees)
Monitored PM2.5
concentrations. Of the
204 counties, 90% had
daily PM2.5 data across
the study period and the
remaining counties had
PM2.5 data collected
once every 3 days for at
least 1 full year. Various
lags and distributed
lags assessed and
presented.
2-stage Bayesian hierarchical
models to estimate county-specific,
region-specific, and national-
average associations.
Stage 1 model included single lag
and distributed lag over-dispersed
Poisson regression models to
estimate county-specific risk.
Models adjusted for temperature
and dew point on the same day and
the 3 previous days, calendar time
to control for seasonality and other
time-varying influences, daily
numbers of individuals at risk, and
day-of-the-week. In Stage 2, to
produce a national average
estimate, Bayesian hierarchical
models were used to combine RRs
across counties and accounting for
within-county statistical error and for
between-county variability or
heterogeneity. To produce regional
estimates. The Stage 2 hierarchical
models described above was used
for 7 regions separately.
B-44

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Franklin et al.,
2007
ST
All-cause, CVD,
and Respiratory
mortality
27 U.S. communities
(with PM2.5
monitoring and daily
mortality data for at
least 2 years of 6-
year study period
1997-2000)
Case-crossover
study (NCHS)
Monitored PM2.5
concentrations with
data for at least 2 years
of a 6-year period.
Within a community,
any monitor that was
not well correlated with
others was excluded,
and values were
averaged to account for
true variability in
concentrations across
the days measured in
the county. Calculated
and presented various
lags and averages for
PM2.5.
2-stage time-stratified analysis: 1)
Conditional logistic regression
analysis to generate community
specific estimates; 2) Meta-
regression analysis to combined
community specific estimates to
generate overall pooled effect
estimate.
Stage 1 of the model adjusted for
day-of-the-week, as well as
apparent temperature at lagO and
lag! Cases were defined as
"deaths" and control days for a
particular subject were chosen to be
every third day within the same
month and year that death occurred.
Effect modification of age and
gender was examined using
interaction terms in stage 1, while
effect modification of community-
specific characteristics including
geographic location, annual PM2.5
concentration > 15 ug/m3 and
central AC prevalence was used in
stage 2.
B-45

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Franklin et al.,
2008
ST
All-cause, CVD,
and Respiratory
mortality
25 U.S communities
(with PM2.5
monitoring and daily
mortality data for at
least 4 years
between 2000-2005)
Case-crossover
study (NCHS)
Monitored PM2.5
concentrations with
data for at least 4 years
of a 6-year period.
Within a community,
any monitor that was
not well correlated with
others was excluded,
and values were
averaged to account for
true variability in
concentrations across
the days measured in
the county. Calculated
and presented various
lags and averages for
PM2.5.
2-stage time-stratified analysis: 1)
Conditional logistic regression
analysis to generate community
specific estimates; 2) Meta-
regression analysis to combined
community specific estimates to
generate overall pooled effect
estimate.
Stage 1 of the model adjusted for
day-of-the-week, as well as
apparent temperature at lagO and
lag! Cases were defined as
"deaths" and control days for a
particular subject were chosen to be
every third day within the same
month and year that death occurred.
Effect modification of age and
gender was examined using
interaction terms in stage 1.
B-46

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Gharibvand et
al., 2016
LT
Lung cancer
incidence
US Nationwide
Cohort study
(AHSMOG-2 study)
Using monitored PM2.5
data from 2000-2001,
inverse distance
weighted interpolations
methods, monthly
pollution surfaces for
PM2.5 were created.
Monthly exposure
averages were based
on daily PM2.5
measurements.
Participants were
assigned monthly
exposure based on their
baseline residential
address.
Cox proportional hazards model
Covariates included sex, race,
smoking status, years since
participant quit smoking, average
number of cigarettes per day during
all smoking years, and education
level. Additional covariates included
calendar time, alcohol consumption,
family income, BMI, physical activity,
and marital status. 3 variables
identified a priori as either as
confounders or effect modifiers:
hours/day spent outdoors, years of
pre-study residence length at
enrollment address, and moving
distance from enrollment address
during follow-up.
Hart et al., 2015
(monitored)
LT
Ail-cause
mortality
US Nationwide
Cohort study
(Nurses' Health
study)
Calculated monthly
average PIVh.sfrom the
nearest monitoring
location for all
addresses.
Nearest monitor
exposures were
validated against
personal exposures to
PM2.5 of ambient origin.
Cox proportional hazards model.
Information on potential confounders
was available every two years (4
years for diet information) and each
woman was assigned updated
covariate values for each
questionnaire cycle. Confounders
examined include age, race, region,
B-47

-------
Hart et al., 2015
(modeled)
LT
All-cause
mortality
US Nationwide
Cohort study
(Nurses' Health
study)
Spatio-temporal models
of PM2.5 were
developed to estimate
monthly PM2.5
exposures at each
geocoded questionnaire
mailing address. The
model was developed
using monitored data
and included
meteorological and
GIS-derived covariates,
such as urban land use
within 1 km, elevation,
tract- and county-level
population density,
distance to the nearest
road for road classes
A1-A3 and point-source
emission density within
7.5 km.
Modeled exposures
were validated against
personal exposures to
PM2.5 of ambient origin.
Previous 12-month
moving average of
exposure either from
nearest monitor or
spatio-temporal models
were assigned to study
participants.
season, physical activity, BMI,
hypercholesterolemia, family history
of Ml, smoking history, Current
smoking status, diet, SES
(education level, occupation of both
of the nurses' parents when she was
16, marital status, and husband's
education if applicable). Also
adjusted for area-level SES (census
tract level median income and
house value), and long-term
temporal trends.
Risk set regression calibration for
time-varying exposures was used to
correct for bias due to exposure
measurement error in the hazard
ratios of all-cause mortality using the
personal exposure validation data.
I to et al., 201314
ST
All-cause
mortality
150 U.S. cities
Time-series study
24-hr average PM2.5
mass data in a given
city, and when data
Poisson regression analysis
B-48

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed





from multiple monitors
were available in a
given city, computed
the average of the daily
values after
standardizing each
site's data using the
mean and standard
deviation of the sites
data.
Pollutant concentration
is expressed in the
model as a deviation
from the monthly mean
to reduce the influence
of the seasonal cycles
of the pollutants on the
overall associations and
help focus on the short-
term associations.
First city- and season-specific
Poisson regression was run, and
then city-specific estimates were
combined using random effects
approach
Adjusted for temporal trends (annual
cycles and influenza epidemics),
immediate and delayed
temperature, and day-of-week
pattern, for entire years (2001-2006)
and for warm (April-September) and
cold (October-March) seasons.
In second stage, assessed effect
modification using land-use
variables and average air pollution
levels.
14 This study is not referenced individually in the ISA, but is study 3 of the National Particle Component Toxicity (NPACT) Initiative published in HEI
(Lippmann et al., 2013).
B-49

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Jerrett et al.,
2016
LT
IHD mortality
30+
U.S. Nationwide
Cohort study (ACS
Cancer Prevention
Study II)
Multiple exposure
estimation approaches
evaluated within the
study - risk assessment
uses results based on
an ensemble approach
that incorporates
chemical transport
modeling, land use
data, satellite data, and
data from ground-based
monitors
Cox proportional hazards regression
Covariates included current and
former smoking status as well as
smoking duration, amount, age
started, second hand cigarette
smoke (hours/day exposed),
exposure to PM2.5 in the workplace
for each of the subject's major
lifetime occupation, self-reported
exposure to dust/fumes at work,
marital status, level of education,
BMI, alcohol consumption, dietary
vegetable/fruit/fiber index, dietary fat
index, missing nutrition information.
Ecologic characteristics included
median household income,
percentage of people with < 125%
of poverty-level income, percentage
of persons > 16 who are
unemployed, percentage of adults
with < 12th grade education, and
percentage of population who were
Black or Hispanic.
B-50

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Kioumourtzoglou
et al., 2016
LT
All-cause
mortality 65+
207 U.S.
communities
Open Cohort study
(MEDICARE
enrollees)
City-specific annual and
2-year PM2.5 averages
using data from all
available monitors in
each city using US EPA
monitors. Calculated
average annual,
summer and winter
temperatures for each
city using National
Climatic Data.
2-stage approach for modelling.
In Stage 1, Cox proportional
hazards model was fit for each city
stratified by age, gender, race and
follow-up time in study. Control for
slowly varying potential confounders
(e.g., SES) and confounders that
vary across subjects, city, and time.
City-characteristics for: proportion of
city population > 65, median
household income, proportion in
poverty, proportion of city families in
poverty, proportion of white, black,
and Asian residents, proportion of
residents with/without high-school
degrees and a college degree, and
city-specific smoking and obesity
rates. Population-weighted city
averages were developed based on
census data at the county level. Also
included average annual
temperature in the model.
In stage 2, combined the city-
specific estimates using a random
effects meta-analysis to generate
region-specific effects. Assessed
effect modification by annual
temperature levels, and population
and city characteristics (greenness,
poverty, racial composition, etc.).
B-51

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Klemm and
Mason, 2003
ST
All-cause
mortality
Harvard Six-City
study reanalysis
Time-series study
24-hour monitored
PM2.5 samples in 6
communities
Generalized additive and
Generalized linear models
Model adjusted for temporal trends,
day-of-the-week, weather (average
daily temperature and average daily
dew point temperature).
B-52

-------
Kloog et al.,
2012
ST
CVD HA Age
65+
New England Area
with 6 U.S. States
Mixed study design
(with time series
and cohort
components)
Spatiotemporal model:
Used day-specific
calibrations of aerosol
optical depth (AOD)
data, using ground
PM2.5 measurements.
Incorporated land use
regressions and
meteorological
variables (temperature,
wind speed, visibility,
elevation, distance to
major road, percent of
open space, point
emissions and area
emissions). Model
used to predict daily
PM2.5 concentrations at
a 10 x 10 km spatial
resolution.
Short-term exposure:
used the mean of PM2.5
on the day of admission
and day before
admission. Long-term
exposure: calculated as
the mean exposure in
each zip-code across
the study period. Short
term exposure was
defined as the
difference between the
two-day average and
the long-term average.
Equivalence between Poisson
regression and the piecewise
constant proportional hazard model
to model the time to a hospital
admission as a function of both
long-term and short-term exposure
simultaneously and enabling
simultaneously examination of short
term and long-term associations
with hospital admissions
(Hierarchical mixed Poisson
regression model).
The model adjusts for temperature,
age, percent minorities, median
income and percent of people with
no high school education.
B-53

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Kloog et al.,
2014
ST
CVD and COPD
HA Age 65+
7 U.S. Mid-Atlantic
States and D.C.
Case-crossover
design (MEDICARE
enrollees)
Spatiotemporal model:
Used day-specific
calibrations of aerosol
optical depth (AOD)
data, using ground
PM2.5 measurements.
Incorporated land use
regression (elevation,
distance to major roads,
percent of open space,
point emissions and
area emissions) and
meteorological
variables (temperature,
wind speed, relative
humidity and visibility).
Model used to predict
daily PM2.5
concentrations at a 10 x
10 km spatial
resolution.
Daily predicted PM2.5
exposure estimates
were matched to zip
codes.
Conditional logistic regression
analysis
Temperature with the same moving
average as PM2.5 was included in
the model as a potential confounder.
Study design samples only cases
and compares each subject's
exposure experience in a time
period just before a case-defining
event with the subject's exposure at
other times, eliminating confounding
(unmeasured or measured) that do
not vary over time. Cases were
matched on day of the week and
defined the relevant exposure time
window as the mean exposure of
the day of and day before the
patient's hospital admission. Effect
modification: 1) assessed whether
subject residence within 30 km of a
monitor or farther modified the PM2.5
association; 2) examined interaction
between exposure and income level
and gender.
B-54

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Krall et al., 2013
ST
All-cause
mortality
72 Urban U.S.
Communities
Time-series study
(NCHS)
Monitored daily
community-level
pollutant exposure as
the arithmetic mean of
daily monitor
observations within the
community. Used lag 1
PM2.5 in model.
Log-linear Poisson Regression
Model
Model adjusted for temperature and
previous day's temperature, long-
term and seasonal trends, age, and
day-of-the-week. Also included
interaction term for pollutant
concentration and seasons.
Lee et al., 2015a
ST
All-cause,
Cardiovascular,
respiratory
mortality
3 U.S. Southeast
States
Case-crossover
design (Dept. of
Pub Health data)
AOD data and predicted
data at 1 km2 resolution
aggregated into the zip
code level and assigned
to resident zip code.
Mean exposure was
calculated using lagO
and Iag1 value.
Monitored PM2.5
concentrations from the
nearest EPA and
IMPROVE monitors
from resident zip code
identified. 24-hr PM
measurement for lagO
and Iag1 were used.
Conditional logistic regression
Model adjusted for temperature and
day of the week
Also ran stratified analysis by age,
sex, race, education and primary
cause of death.
Analysis also restricted for zip codes
where annual average of PM2.5 <12
or daily average <35 separately.
Sensitivity analysis: potential non-
linear relationship between temp
and mortality modelled using natural
spline to the temperature term.
B-55

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Lepeule et al.,
2012
LT
All-cause,
Cardiovascular,
lung cancer
mortality
HARVARD 6 cities
Prospective
Cohort/Longitudinal
follow-up study
(HARVARD 6 cities
data)
PM2.5 data from
monitors in the
participant's city. PM2.5
data 1979-1986/1988
from monitors, end of
monitoring to 1998
estimated from PM10
using US EPA monitors,
1999-2009 direct PM2.5
measurement from US
EPA monitors. 1-yror 1-
3yr or 1-5 yr. moving
PM2.5 averages were
assigned to participants
based on city of
residence.
Cox proportional hazard models,
Poisson survival analysis
Stratified analysis by sex, age and
time in the study (1-yr interval).
Confounders included: Baseline
information on smoking status,
smoking pack-years, education,
linear and quadratic term for BMI.
Also explored effect modification of
PM2.5 on mortality by smoking
status at enrollment, as well as time
period in study.
B-56

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Malig et al.,
2013
ST
Respiratory
morbidity
(Asthma and
COPD ED and
HA)
35 CA counties
(9 counties included
for PM2.5 analysis)
Case-crossover
design (CA Office
of Statewide Health
Planning and
Development Data)
PM2.5 data obtained
from California Air
Resources Board.
Same day lag and
various days lags
average were
calculated for PM2.5.
Participants were
assigned exposure from
the closest monitor from
the residential
population-weighted zip
code centroid.
County-level conditional logistic
regression analysis. Overall
estimate was then calculated by
combining county-level estimates
using a random-effects meta-
analysis
Time-invariant confounders and
seasonal trends were controlled for
given the study design.
Other confounders included in the
models were: other gaseous
pollutants including ozone, linear
and squared term for daily average
temperature.
Stratified analysis also by distance
to monitor: within 10 km vs. 10-20
km
McConnell et al.,
2010
LT
Asthma
Incidence
13CA communities
Cohort Study
(CHS)
PM2.5 measured in
central site monitors in
each community and
assigned to study
participants.
Multi-level Cox proportional hazard
model accounting for residual
variation in time to asthma onset
and clustering of children around
schools and communities
Models adjusted for: secondhand
smoke, pets in home, race/ethnicity,
age at study entry, sex, and random
effects for community and school.
B-57

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Ostro et al.,
2016
ST
Asthma and
COPD ED
8 metropolitan
areas/counties in CA
Case-crossover
design (CA Office
of Statewide Health
Planning and
Development Data)
PM2.5 chemical
speciation data from
U.S. EPA provided by
California Air
Resources Board.
Participants were
assigned exposure from
the closest monitor from
the residential
population-weighted zip
code centroid. Only
participants living in zip
codes within 20 km of
PM2.5 constituents
monitors were included.
County-level conditional logistic
regression analysis. Overall
estimate was then calculated by
combining county-level estimates
using a random-effects meta-
analysis
Time-invariant confounders and
seasonal trends were controlled for
given the study design.
Other confounders included in the
models were: linear and squared
term for lagO temperature, day of the
week.
B-58

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Peng et al.,
2009
ST
CVD HA Age
65+
119 U.S. Urban
counties>150,000
populations
Time-series
analysis
(MEDICARE
enrollees)
PM2.5 data obtained
from US EPA'sAQS
and STN.
Log-linear Poisson Regression
analysis
Adjusted for potential confounders
like: weather, day of the week,
unobserved seasonal factors. In
county-specific regression model,
following indicators were included:
indicator for the day of the weeks, a
smooth function of time per calendar
year to control for seasonality and
long-term trends, a smooth function
of current-day temperature, a
smooth function of the 3-day running
mean temperature, a smooth
function of current-day dew-point
temperature, and a smooth function
of the 3-day running mean dew-
point temperature. To model smooth
functions we used a natural spline
basis.
B-59

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Pinaultet al.,
2016
LT
All-cause, CVD
and lung cancer
mortality
Multicity Canada
Prospective Cohort
Study (subset of
participants of the
Canadian
Community Health
Survey)
PM2.5 concentration
derived from MODIS.
Geographically
weighted regression
including monitoring
and land use data was
applied to the estimates
from MODIS to produce
average PM2.5
concentration at 1 km2
resolution. These model
estimates extended to
1998-2003 using inter-
annual variation of Boys
et al.
Participants were
assigned exposure
based on their postal
code of residence.
Cox proportional hazards models
Models were stratified by age (5-yr
interval) and sex. Models adjusted
for individual socioeconomic
covariates and behavioral (BMI,
smoking and alcohol consumption,
fruit and vegetable consumption)
covariates, ecological variables
including neighborhood
socioeconomic status (both social
and material deprivation).
B-60

-------
Pope et al.,
2015a
LT
All-cause, IHD
mortality (30+)
U.S. Nationwide
Cohort study (ACS
Cancer Prevention
Study II)
Exposure to PM2.5 was
estimated by linking
geocoded home
addresses
of the study participants
to ambient PM2.5
concentrations derived
using
a national-level hybrid
land use regression
(LUR) and Bayesian
Maximum Entropy
(BME) interpolation
model (LUR-BME) that
incorporated data from
ground-based monitors
B-61
Cox proportional hazards models
The individual-level covariates
incorporated in the models included
13
variables that characterized current
and former smoking habits
(including
smoking status of never, former, or
current smoker, linear and
squared terms for years smoked
and cigarettes smoked per day,
indicator
for starting smoking at aged <18
years, and pipe/cigar smoker);
1 continuous variable that assessed
exposure to second-hand cigarette
smoke (hours/d exposed); 7
variables that reflected workplace
PM2.5
exposure in each subject's main
lifetime occupation; a variable that
indicated self-reported exposure to
dust and fumes in the workplace;
variables that represented marital
status (separated/divorced/widowed
or single versus married); variables
that characterized the level
of education (high school, more than
high school versus less than
high school); 2 body mass index
variables (linear and squared terms
for body mass index); variables that
characterized the consumption
of alcohol (beer, missing beer, wine,
missing wine, liquor, and missing

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed






liquor); and variables that indicated
quartile ranges of dietary fat
index and quartile ranges of a
dietary vegetable/fruit/fiber index.
Ecological covariates included
median household income;
percentage
of people with <125% of poverty-
level income; percentage of
unemployed individual aged >16
years; percentage of adults with
<12th grade education; and
percentage of the population who
were
black or Hispanic. These ecological
covariates were included in the
models using both zip code level
data and zip code deviations from
the county means.
B-62

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Shi et al., 2016
ST and LT
Total mortality
(65+)
New England Area
with 6 U.S. States
Open Cohort study
(MEDICARE
enrollees)
Daily PM2.5 was
predicted at 1-km2
spatial resolution from
novel 3-stage statistical
models. Similar 3-stage
approach was used to
estimate daily
temperature.
Participants were
assigned 365-day
moving average (for
long-term exposure)
and average Iag0-1 (for
short-term exposure)
based on the ZIP codes
of residence.
Chronic effects of air pollution
assessed using Cox proportional
hazard models. Acute effects of air
pollution assessed using Poisson
log-linear models.
Both acute and chronic effects were
assessed using Poisson survival
analysis. Analysis performed in full-
cohort as well as low exposure
cohorts.
Poisson survival models were
adjusted for smooth function of time,
temporal covariates such as
temperatures and day of the week,
spatial covariates such as zip code-
level socio-economic variables.
Stieb et al.,
2009
ST
Cardiac and
Respiratory ED
visits
Seven Canadian
Cities
Time series study
(Hospital cases)
PM Data from National
Air Pollution
Surveillance (NAPS)
system. City averages
of the exposure were
calculated by averaging
stations within the city.
Calculated average
concentration for lagO-
2.
Generalized Linear Models with
natural spline functions of time to
adjust for seasonal cycles in air
pollution and health
Confounders included: Mean daily
temperature and relative humidity at
lag 0,1, and 2 days, day of the week
and holidays.
B-63

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Szyszkowicz,
2009
ST
Angina ED
Seven Canadian
Cities
Time series study
(Hospital cases)
PM Data from National
Air Pollution
Surveillance (NAPS)
system. City averages
of the exposure were
calculated by averaging
stations within the city.
Calculated average
concentration for lagO-
2.
Generalized Linear Mixed models
Models adjusted for meteorological
variables such as relative humidity,
temperature and atmospheric
pressure (a daily 24-hr average
measurements were calculated).
Temperature and relative humidity in
models were represented by natural
splines. Stratified analysis by
season as well as combined for the
whole period.
Thurston et al.,
2016a
LT
All-cause, CVD
and respiratory
mortality
6 U.S. States and 2
MS As
Cohort study
(NIH_AARP cohort)
PM Data from US EPA
AQS. Census-tract
estimates generated
using hybrid LUR and
BME models that were
combined to generate
monthly estimates of
PM2.5.
Participants exposure
was estimated at
census-tract of
residence and included
annual mean
concentration in the
year of mortality, and 1-
year lag average.
Cox proportional hazard models
Stratified analysis by age, sex,
regions (6 states and 2 MSAs).
Confounders adjusted included:
race, education, marital status, BMI,
alcohol consumption, smoking
history, contextual variables such as
median household income and %
pop with less than high school
education. Several interactions
between PM2.5 and socio-
demographics were also tested.
B-64

-------
Turner et al.,
2016
LT
Lung cancer
mortality (30+)
U.S. Nationwide
Cohort study (ACS
Cancer Prevention
Study II)
Estimated PM2.5
concentrations were
obtained using a
national-level hybrid
land use regression
(LUR) and Bayesian
maximum entropy
(BME) interpolation
model. Monthly PM2.5
monitoring data were
collected from 1,464
sites from 1999 through
2008, with 10%
reserved for cross-
validation. The base
LUR model that
predicted PM2.5
concentrations
included traffic within 1
km and green space
within 100 m3. Residual
spatiotemporal variation
in PM2.5 concentrations
was interpolated with
a BME interpolation
model. The two
estimates were then
combined. The cross
validation
R2 was approximately
0.79. Mean PM2.5
(1999-2004)
concentrations
were used here.
Cox proportional hazards model
Models were adjusted for education;
marital status; BMI and BMI
squared; cigarette smoking status;
cigarettes per day and
cigarettes per day squared; years
smoked and years smoked squared;
started smoking at younger than 18
years of age; passive smoking
(hours); vegetable, fruit, fiber, and
fat intake; beer, wine, and liquor
consumption; occupational
exposures; an occupational
dirtiness index; and six
sociodemographic
ecological covariates at both the
postal code and postal code minus
county-level mean derived from the
1990 U.S. Census (median
household income and percentage
of African American residents,
Hispanic residents, adults with
postsecondary education,
unemployment, and poverty).
Potential confounding examined by
elevation, MSA size, annual average
daily maximum air temperature,
mean county-level residential radon
concentrations, and 1980
percentage of air conditioning.
Urman et al.,
2014
LT
Lung-function
decline
8 Southern CA
communities/counties
Cohort study (CHS)
Central monitors in
each community
Linear Regression model
B-65

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed





provided data on air
pollutants. Each child
was assigned exposure
based on the child's
resident community.
Models were adjusted for
demographic, socio-economic and
anthropometric variables (BMI,
height), study community.
Wang et al.,
2017
LT
Total mortality
(65+)
7 U.S. Southeast
States
Open Cohort study
(MEDICARE
enrollees data)
Three stage Hybrid
model to predict daily
PM2.5 concentration at
1 km2 resolution. Air
temperature also
estimated at similar
scale using satellite
remote sensing and
land use variables.
Participants were
assigned annual
averages of PM2.5 by
averaging estimated for
all grid cells within the
zip code tabulation area
(ZCTA) of residence.
Cox Proportional hazard models
Models were stratified by age
groups, sex, race. Adjusted for
variables: year of enrollment,
previous admission due to CHF,
COPD, Ml and diabetes, numbers of
days spent in ICU and CCU, state,
ZCTA level socio-demographic
variables such as % pop below
poverty, urbanicity, lower education,
median income and median home
value, and behavioral variables such
as % smokers and obesity at county
level. Further model also included
yearly mean summer temperature at
ZCTA level.
B-66

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Weichenthal et
al., 2016c
ST
Asthma and
COPD ED
15 cities in Ontario
Case-crossover
Design (cases
extracted from
NACRS database)
Daily average
concentration of PM2.5
collected from fixed-
monitoring stations in
Ontario, part of
Canada's National Air
Pollution Data.
Participants were
assigned data based on
the city of residence.
Various lags assessed
including lagO, Iag1,
Iag2 and lagO-2.
Conditional logistic regression
models
Models adjusted for 3-day mean
temperature and relative humidity
using cubic splines.
Weichenthal et
al., 2016b
ST
Ml ED
16 cities in Ontario
Case-crossover
Design (cases
extracted from
NACRS database)
PM data obtained from
20 provincial monitoring
sites located in 16
cities. Exposure at
various lags: lagO Iag1,
lag 2 and mean lagO-2
were assigned to
participants based on
the city of residence.
Conditional logistic regression
models
Models adjusted for 3-day mean
temperature and relative humidity
using cubic splines.
B-67

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Yap et al., 2013
ST
Asthma HA
12 CA counties
Time Series study
(Hospital
admissions)
PM2.5 data was
obtained from California
Air Resources Board
that maintains
information from the
National Air Monitoring
Stations. 24-hr average
mass concentration
calculated for each
county by averaging
monitors within the
county.
Participants were
assigned exposure
based on their county of
residence. PM at
various lags Iag0-lag6
were assessed.
Generalized Additive Poisson
Regression analysis were run at
county-level
Models adjusted for: long-term time
trends and seasonality, day of the
week and smoothing splines within
different lags for temperature. Effect
modification by single or composite
area-based SES assessed.
Zanobetti et al.,
2009
ST
Heart Failure
and Ml HA 65+
26 US communities
Time Series study
(MEDICARE
enrollees data)
PM2 5data obtained
from US EPAAQS.
Daily PM2.5data
available for various
monitors were averaged
over the county.
Generated 2-day
moving average PM2.5
conc..
Poisson regression analysis
Models stratified by season.
Controlled for long-term trend with
natural cubic spline for each season
and year, day of the week, three-day
average temperature and dew point
temperature.
B-68

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Zanobetti and
Schwartz, 2009
ST
All-cause, CVD
and respiratory
mortality
112 US cities
Time Series study
(NCHS data)
PM2.5 data obtained
from US EPAAQS.
Daily PM2.5data
available for various
monitors were averaged
over the county.
Generated 2-day
moving average (lag 0
and 1) PM2.5 conc.
Poisson regression analysis
First city- and season-specific
Poisson regression was run, and
then city-specific estimates were
combined using random effects
approach in total by season and
region.
Controlled for long-term trend with
natural cubic spline for each season
and year, day of the week, same
day and previous day temperature.
B-69

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Zanobetti etal.,
2014
ST
All-cause
mortality 65+
121 US
communities/cities
Case-Crossover
Design
(MEDICARE
enrollees)
PM2.5 data obtained
from US EPAAQS.
Daily PM2.5data
available for various
monitors were averaged
over the communities.
Participants were
assigned 2-day moving
average (lag 0 and 1)
based on community of
residence.
Conditional logistic regression
models at community level. In a
second stage of analysis, the
community specific results were
combined using the multivariate
meta-analysis techniques
Conditional logistic regression
controlled for confounders such as
average temp for the same and
previous day. Temperature was
modelled using spline to account for
nonlinear relationship. Effect
modification tested for cause of prior
admission due to neurological
disorders or diabetes, primary or
secondary hospitalization for other
disease conditions. Stratified
analysis by sex, age or race.
B-70

-------
Citation
Long-term
(LT)/Short-
term (ST)
Health
Endpoint
Geographic Area
Study Design
Exposure Metric
Statistical Analysis Including
Confounding Variables
Addressed
Zeger et al.,
2008
LT
All-cause
mortality 65+
668 U.S Urban
counties
Retrospective
Cohort Study of
MEDICARE
enrollees (MCAPS)
PM2.5 data available
from US EPA monitors.
Spatially smoothed
levels of 6-year average
PM2.5
Participants living within
6 miles of the zip code
centroid to EPA
monitors were assigned
exposure based on the
ZIP code of residence.
Log-linear Regression model ran for
specific US regions separately
Models adjusted for individual socio-
demographic variables and ZIP
code level SES variables
(education, income, poverty etc.).
Also included standardized mortality
ratio for COPD as a surrogate
indicator of long-term smoking
pattern of its residents.
B-71

-------
REFERENCES
Baxter, LK, Crooks, JL and Sacks, JD (2017). Influence of exposure differences on city-to-city
heterogeneity in PM2.5-mortality associations in US cities. Environ Health 16(1): 1.
Bell, ML, Ebisu, K, Peng, RD, Walker, J, Samet, JM, Zeger, SL and Dominic, F (2008).
Seasonal and regional short-term effects of fine particles on hospital admissions in 202
U.S. counties, 1999-2005. Am J Epidemiol 168(11): 1301-1310.
Bell, ML, Ebisu, K, Leaderer, BP, Gent, JF, Lee, HJ, Koutrakis, P, Wang, Y, Dominici, F and
Peng, RD (2014). Associations of PM2.5 constituents and sources with hospital
admissions: analysis of four counties in Connecticut and Massachusetts (USA) for
persons > 65 years of age. Environ Health Perspect 122(2): 138-144.
Bell, ML, Son, JY, Peng, RD, Wang, Y and Dominici, F (2015). Brief report: Ambient PM2.5 and
risk of hospital admissions: do risks differ for men and women? Epidemiology 26(4):
575-579.
Bravo, MA, Ebisu, K, Dominici, F, Wang, Y, Peng, RD and Bell, ML (2017). Airborne fine
particles and risk of hospital admissions for understudied populations: Effects by
urbanicity and short-term cumulative exposures in 708 U.S. counties. Environ Health
Perspect 125(4): 594-601.
Burnett, RT and Goldberg, MS (2003). Size-fractionated particulate mass and daily mortality in
eight Canadian cities. Boston, MA, Health Effects Institute.
Burnett, RT, Stieb, D, Brook, JR, Cakmak, S, Dales, R, Raizenne, M, Vincent, R and Dann, T
(2004). Associations between short-term changes in nitrogen dioxide and mortality in
Canadian cities. Arch Environ Occup Health 59(5): 228-236.
Chan, E. (2019). Email and attachment from Elizabeth Chan, EPA, Question re: Request for
additional PM2.5 information from Di et al 2017. Available in PM NAAQS Review
Docket (EPA-HQ-OAR-2015-0072). December 3, 2019. Research Triangle Park, NC.
Office of Air Quality Planning and Standards.
Chen, H, Burnett, RT, Copes, R, Kwong, JC, Villeneuve, PJ, Goldberg, MS, Brook, RD, van
Donkelaar, A, Jerrett, M, Martin, RV, Brook, JR, Kopp, A and Tu, JV (2016). Ambient
fine particulate matter and mortality among survivors of myocardial infarction:
population-based cohort study. Environ Health Perspect 124(9): 1421-1428.
Crouse, DL, Peters, PA, van Donkelaar, A, Goldberg, MS, Villeneuve, PJ, Brion, O, Khan, S,
Atari, DO, Jerrett, M, Pope, CA, Brauer, M, Brook, JR, Martin, RV, Stieb, D and
Burnett, RT (2012). Risk of nonaccidental and cardiovascular mortality in relation to
long-term exposure to low concentrations of fine particulate matter: a Canadian national-
level cohort study. Environ Health Perspect 120(5): 708-714.
B-72

-------
Crouse, DL, Peters, PA, Hystad, P, Brook, JR, van Donkelaar, A, Martin, RV, Villeneuve, PJ,
Jerrett, M, Goldberg, MS, Pope, CA, Brauer, M, Brook, RD, Robichaud, A, Menard, R
and Burnett, RT (2015). Ambient PM 2.5, O 3, and NO 2 exposures and associations with
mortality over 16 years of follow-up in the Canadian Census Health and Environment
Cohort (CanCHEC). Environ Health Perspect 123(11): 1180-1186.
Dai, L, Zanobetti, A, Koutrakis, P and Schwartz, JD (2014). Associations of fine particulate
matter species with mortality in the United States: a multicity time-series analysis.
Environ Health Perspect 122(8): 837-842.
Di, Q, Dai, L, Wang, Y, Zanobetti, A, Choirat, C, Schwartz, JD and Dominici, F (2017a).
Association of short-term exposure to air pollution with mortality in older adults. J Am
Med Assoc 318(24): 2446-2456.
Di, Q, Wang, Y, Zanobetti, A, Wang, Y, Koutrakis, P, Choirat, C, Dominici, F and Schwartz, JD
(2017b). Air pollution and mortality in the Medicare population. New Engl J Med
376(26): 2513-2522.
Dominici, F, Peng, RD, Bell, ML, Pham, L, McDermott, A, Zeger, SL and Samet, JL (2006).
Fine particulate air pollution and hospital admission for cardiovascular and respiratory
diseases. J Am Med Assoc 295(10): 1127-1134.
Franklin, M, Zeka, A and Schwartz, J (2007). Association between PM2.5 and all-cause and
specific-cause mortality in 27 US communities. Journal of Exposure Science and
Environmental Epidemiology 17(3): 279-287.
Franklin, M, Koutrakis, P and Schwartz, J (2008). The role of particle composition on the
association between PM2.5 and mortality. Epidemiology 19(5): 680-689.
Gharibvand, L, Shavlik, D, Ghamsary, M, Beeson, WL, Soret, S, Knutsen, R and Knutsen, SF
(2016). The association between ambient fine particulate air pollution and lung cancer
incidence: Results from the AHSMOG-2 study. Environ Health Perspect 125(3): 378-
384.
Hart, JE, Garshick, E, Dockery, DW, Smith, TJ, Ryan, L and Laden, F (2011). Long-term
ambient multi-pollutant exposures and mortality. Am J Respir Crit Care Med 183(1): 73-
78.
Hart, JE, Liao, X, Hong, B, Puett, RC, Yanosky, JD, Suh, H, Kioumourtzoglou, MA,
Spiegelman, D and Laden, F (2015). The association of long-term exposure to PM2.5 on
all-cause mortality in the Nurses' Health Study and the impact of measurement-error
correction. Environ Health 14: 38.
Hystad, P, Demers, PA, Johnson, KC, Brook, J, van Donkelaar, A, Lamsal, L, Martin, R and
Brauer, M (2012). Spatiotemporal air pollution exposure assessment for a Canadian
population-based lung cancer case-control study. Environ Health 11: 22.
B-73

-------
Hystad, P, Demers, PA, Johnson, KC, Carpiano, RM and Brauer, M (2013). Long-term
residential exposure to air pollution and lung cancer risk. Epidemiology 24(5): 762-772.
Ito, K, Ross, Z, Zhou, J, Nadas, A, Lippmann, M and Thurston, GD (2013). National Particle
Component Toxicity (NPACT) initiative: Study 3. Time-series analysis of mortality,
hospitalizations, and ambient PM2.5 and its components. Boston, MA, Health Effects
Institute: 95-125.
Jenkins, S. (2019a). Email and attachment from Scott Jenkins, EPA, Question re: PM2.5 LT
estimates Di et al. (2017) Studies. Available in PM NAAQS Review Docket (EPA-HQ-
OAR-2015-0072-0022). May 8, 2019. Research Triangle Park, NC. Office of Air Quality
Planning and Standards.
Jenkins, S. (2019b). Email and attachment from Scott Jenkins, EPA, Question re: PM2.5 ST
estimates Di et al. (2017) Studies. Available in PM NAAQS Review Docket (EPA-HQ-
OAR-2015-0072-0025). May 9, 2019. Research Triangle Park, NC. Office of Air Quality
Planning and Standards.
Jerrett, M, Turner, MC, Beckerman, BS, Pope, CA, van Donkelaar, A, Martin, RV, Serre, M,
Crouse, D, Gapstur, SM, Krewski, D, Diver, WR, Coogan, PF, Thurston, GD and
Burnett, RT (2016). Comparing the health effects of ambient particulate matter estimated
using ground-based versus remote sensing exposure estimates. Environ Health Perspect
125(4): 552-559.
Kioumourtzoglou, MA, Schwartz, J, James, P, Dominici, F and Zanobetti, A (2016). PM2.5 and
mortality in 207 us cities: Modification by temperature and city characteristics.
Epidemiology 27(2): 221-227.
Klemm, RJ and Mason, R (2003). HEI Special Report: Revised Analyses of Time-Series Studies
of Air Pollution and Health, Part II Replication of reanalysis of Harvard Six-City
mortality study. Boston, MA, Health Effects Institute.
Kloog, I, Coull, BA, Zanobetti, A, Koutrakis, P and Schwartz, JD (2012). Acute and chronic
effects of particles on hospital admissions in New-England. PLoS ONE 7(4): e34664.
Kloog, I, Nordio, F, Zanobetti, A, Coull, BA, Koutrakis, P and Schwartz, JD (2014). Short term
effects of particle exposure on hospital admissions in the Mid-Atlantic states: a
population estimate. PLoS ONE 9(2): e88578.
Krall, JR, Anderson, GB, Domini ci, F, Bell, ML and Peng, RD (2013). Short-term exposure to
particulate matter constituents and mortality in a national study of U.S. urban
communities. Environ Health Perspect 121(10): 1148-1153.
Krewski, D, Jerrett, M, Burnett, RT, Ma, R, Hughes, E, Shi, Y, Turner, MC, Pope, CA, III,
Thurston, G, Calle, EE, Thun, MJ, Beckerman, B, Deluca, P, Finkelstein, N, Ito, K,
Moore, DK, Newbold, KB, Ramsay, T, Ross, Z, Shin, H and Tempalski, B (2009).
B-74

-------
Extended follow-up and spatial analysis of the American Cancer Society study linking
particulate air pollution and mortality. Boston, MA, Health Effects Institute. 140.
Lee, H, Honda, Y, Hashizume, M, Guo, YL, Wu, CF, Kan, H, Jung, K, Lim, YH, Yi, S and Kim,
H (2015a). Short-term exposure to fine and coarse particles and mortality: A multicity
time-series study in East Asia. Environ Pollut 207: 43-51.
Lee, M, Koutrakis, P, Coull, B, Kloog, I and Schwartz, J (2015b). Acute effect of fine particulate
matter on mortality in three Southeastern states from 2007-2011. Journal of Exposure
Science and Environmental Epidemiology 26(2): 173-179.
Lepeule, J, Laden, F, Dockery, D and Schwartz, J (2012). Chronic exposure to fine particles and
mortality: an extended follow-up of the Harvard Six Cities study from 1974 to 2009.
Environ Health Perspect 120(7): 965-970.
Lippmann, M, Chen, LC, Gordon, T, Ito, K and Thurston, GD (2013). National Particle
Component Toxicity (NPACT) Initiative: Integrated epidemiologic and toxicologic
studies of the health effects of particulate matter components. Boston, MA, Health
Effects Institute.
Malig, BJ, Green, S, Basu, R and Broadwin, R (2013). Coarse particles and respiratory
emergency department visits in California. Am J Epidemiol 178(1): 58-69.
McConnell, R, Islam, T, Shankardass, K, Jerrett, M, Lurmann, F, Gilliland, F, Gauderman, J,
Avol, E, Kiinzli, N, Yao, L, Peters, J and Berhane, K (2010). Childhood incident asthma
and traffic-related air pollution at home and school. Environ Health Perspect 118(7):
1021-1026.
Miller, KA, Siscovick, DS, Sheppard, L, Shepherd, K, Sullivan, JH, Anderson, GL and
Kaufman, JD (2007). Long-term exposure to air pollution and incidence of
cardiovascular events in women. New Engl J Med 356(5): 447-458.
Ostro, B, Malig, B, Hasheminassab, S, Berger, K, Chang, E and Sioutas, C (2016). Associations
of source-specific fine particulate matter with emergency department visits in California.
Am J Epidemiol 184(6): 450-459.
Paciorek, CJ, Yanosky, JD, Puett, RC, Laden, F and Suh, HH (2009). Practical large-scale
spatio-temporal modeling of particulate matter concentrations. Annals of Applied
Statistics 3(1): 370-397.
Peng, RD, Dominici, F and Welty, LJ (2009). A Bayesian hierarchical distributed lag model for
estimating the time course of risk of hospitalization associated with particulate matter air
pollution. J Roy Stat Soc Ser C (Appl Stat) 58: 3-24.
Pinault, L, Tjepkema, M, Crouse, DL, Weichenthal, S, van Donkelaar, A, Martin, RV, Brauer,
M, Chen, H and Burnett, RT (2016). Risk estimates of mortality attributed to low
concentrations of ambient fine particulate matter in the Canadian community health
survey cohort. Environ Health 15(1): 18.
B-75

-------
Pope, CA, Turner, MC, Burnett, R, Jerrett, M, Gapstur, SM, Diver, WR, Krewski, D and Brook,
RD (2015a). Relationships between fine particulate air pollution, cardiometabolic
disorders and cardiovascular mortality. Circul Res 116(1): 108-U258.
Pope, CA, III, Muhlestein, JB, Anderson, JL, Cannon, JB, Hales, NM, Meredith, KG, Viet, L
and Home, BD (2015b). Short-Term Exposure to Fine Particulate Matter Air Pollution Is
Preferentially Associated With the Risk of ST-Segment Elevation Acute Coronary
Events. Journal of the American Heart Association 4(12).
Puett, RC, Hart, JE, Yanosky, JD, Paciorek, C, Schwartz, J, Suh, H, Speizer, FE and Laden, F
(2009). Chronic fine and coarse particulate exposure, mortality, and coronary heart
disease in the Nurses' Health Study. Environ Health Perspect 117(11): 1697-1701.
Puett, RC, Hart, JE, Suh, H, Mittleman, M and Laden, F (2011). Particulate matter exposures,
mortality and cardiovascular disease in the health professionals follow-up study. Environ
Health Perspect 119(8): 1130-1135.
Rajan, P, Schmidt, M., Hassett-Sipple, B. (2011). Memorandum to PM NAAQS Review Docket
(EPA-HQ-OAR-2007-0492). PM2.5 Distributional Statistical Analyses. April 7, 2011.
Docket ID No. EPA-HQ-OAR-2007-0492. Research Triangle Park, NC. Office of Air
Quality Planning and Standards. Available at:
https://www3.epa.gov/ttn/naaqs/standards/pm/data/Raian040711.pdf.
Shi, L, Zanobetti, A, Kloog, I, Coull, BA, Koutrakis, P, Melly, SJ and Schwartz, JD (2016).
Low-concentration PM2.5 and mortality: estimating acute and chronic effects in a
population-based study. Environ Health Perspect 124(1): 46-52.
Stieb, DM, Szyszkowicz, M, Rowe, BH and Leech, JA (2009). Air pollution and emergency
department visits for cardiac and respiratory conditions: A multi-city time-series analysis.
Environ Health 8(25): 25.
Szyszkowicz, M (2009). Air pollution and ED visits for chest pain. American Journal of
Emergency Medicine 27(2): 165-168.
Thurston, GD, Ito, K, Lall, R, Burnett, RT, Turner, MC, Krewski, D, Shi, Y, Jerrett, M, Gapstur,
SM, Diver, WR and Pope, CA, III (2013). NPACT Study 4. Mortality and long-term
exposure to PM2.5 and its components in the American Cancer Society's Cancer
Prevention Study II Cohort. M. Lippmann, L. C. Chen, T. Gordon, K. Ito and G. D.
Thurston. Boston, MA, Health Effects Institute: 127-166.
Thurston, GD, Ahn, J, Cromar, KR, Shao, Y, Reynolds, HR, Jerrett, M, Lim, CC, Shanley, R,
Park, Y and Hayes, RB (2016a). Ambient particulate matter air pollution exposure and
mortality in the NIH-AARP Diet and Health Cohort. Environ Health Perspect 124(4):
484-490.
Thurston, GD, Burnett, RT, Turner, MC, Shi, Y, Krewski, D, Lall, R, Ito, K, Jerrett, M, Gapstur,
SM, Diver, WR and Pope, CA, III (2016b). Ischemic heart disease mortality and long-
B-76

-------
term exposure to source-related components of U.S. fine particle air pollution. Environ
Health Perspect 124(6): 785-794.
Turner, MC, Jerrett, M, Pope, A, III, Krewski, D, Gapstur, SM, Diver, WR, Beckerman, BS,
Marshall, JD, Su, J, Crouse, DL and Burnett, RT (2016). Long-term ozone exposure and
mortality in a large prospective study. Am J Respir Crit Care Med 193(10): 1134-1142.
U.S. EPA. (2019). Integrated Science Assessment (ISA) for Particulate Matter (Final Report).
Washington, DC. U.S. Environmental Protection Agency, Office of Research and
Development, National Center for Environmental Assessment. U.S. EPA. EPA/600/R-
19/188. December 2019. Available at: https://www.epa.gov/naaqs/particulate-matter-pm-
standards-integrated-science-assessments-current-review.
Urman, R, McConnell, R, Islam, T, Avol, EL, Lurmann, FW, Vora, H, Linn, WS, Rappaport,
EB, Gilliland, FD and Gauderman, WJ (2014). Associations of children's lung function
with ambient air pollution: Joint effects of regional and near-roadway pollutants. Thorax
69(6): 540-547.
van Donkelaar, A, Martin, RV, Brauer, M, Kahn, R, Levy, R, Verduzco, C and Villeneuve, PJ
(2010). Global estimates of ambient fine particulate matter concentrations from satellite-
based aerosol optical depth: development and application. Environ Health Perspect
118(6): 847-855.
van Donkelaar, A, Martin, RV, Spurr, RJD, Drury, E, Remer, LA, Levy, RC and Wang, J (2013).
Optimal estimation for global ground-level fine particulate matter concentrations. Journal
of Geophysical Research: Atmospheres 118(11): 5621-5636.
van Donkelaar, A, Martin, RV, Brauer, M and Boys, BL (2015a). Use of satellite observations
for long-term exposure assessment of global concentrations of fine particulate matter.
Environ Health Perspect 123(2): 135-143.
van Donkelaar, A, Martin, RV, Spurr, RJ and Burnett, RT (2015b). High-resolution satellite-
derived PM2.5 from optimal estimation and geographically weighted regression over
North America. Environ Sci Technol 49(17): 10482-10491.
Villeneuve, PJ, Weichenthal, SA, Crouse, D, Miller, AB, To, T, Martin, RV, van Donkelaar, A,
Wall, C and Burnett, RT (2015). Long-term exposure to fine particulate matter air
pollution and mortality among Canadian women. Epidemiology 26(4): 536-545.
Wang, Y, Shi, L, Lee, M, Liu, P, Di, Q, Zanobetti, A and Schwartz, JD (2017). Long-term
exposure to PM2.5 and mortality among older adults in the Southeastern US.
Epidemiology 28(2): 207-214.
Weichenthal, S, Hatzopoulou, M and Goldberg, MS (2014). Exposure to traffic-related air
pollution during physical activity and acute changes in blood pressure, autonomic and
micro-vascular function in women: A cross-over study. Particle and Fibre Toxicology
11(1): 70.
B-77

-------
Weichenthal, S, Crouse, DL, Pinault, L, Godri-Pollitt, K, Lavigne, E, Evans, G, van Donkelaar,
A, Martin, RV and Burnett, RT (2016a). Oxidative burden of fine particulate air pollution
and risk of cause-specific mortality in the Canadian Census Health and Environment
Cohort (CanCHEC). Environ Res 146: 92-99.
Weichenthal, S, Lavigne, E, Evans, G, Pollitt, K and Burnett, RT (2016b). Ambient PM2.5 and
risk of emergency room visits for myocardial infarction: Impact of regional PM2.5
oxidative potential: A case-crossover study. Environ Health 15: 46.
Weichenthal, S, Lavigne, E, Evans, GJ, Godri Pollitt, KJ and Burnett, RT (2016c). PM2.5 and
emergency room visits for respiratory illness: effect modification by oxidative potential.
Am J Respir Crit Care Med 194(5): 577-586.
Yap, P, Gilbreath, S, Garcia, C, Jareen, N and Goodrich, B (2013). The influence of
socioeconomic markers on the association between fine particulate matter and hospital
admissions for respiratory conditions among children. Am J Public Health 103(4): 695-
702.
Zanobetti, A, Franklin, M, Koutrakis, P and Schwartz, J (2009). Fine particulate air pollution and
its components in association with cause-specific emergency admissions. Environ Health
8: 58.
Zanobetti, A and Schwartz, J (2009). The effect of fine and coarse particulate air pollution on
mortality: A national analysis. Environ Health Perspect 117(6): 1-40.
Zanobetti, A, Dominici, F, Wang, Y and Schwartz, JD (2014). A national case-crossover analysis
of the short-term effect of PM2.5 on hospitalizations and mortality in subjects with
diabetes and neurological disorders. Environ Health 13(1): 38.
Zeger, S, Dominici, F, McDermott, A and Samet, J (2008). Mortality in the Medicare population
and chronic exposure to fine particulate air pollution in urban centers (2000-2005).
Environ Health Perspect 116(12): 1614-1619.
B-78

-------
B-79

-------
APPENDIX C. SUPPLEMENTAL INFORMATION
RELATED TO THE HUMAN HEALTH RISK
ASSESSMENT

-------
TABLE OF CONTENTS
C. 1 Additional Technical Detail on the Risk Assessment Approach	C-l
C. 1.1 Selection of Key Health Endpoints and Specification of Concentration-Response
Functions from Epidemiologic Studies	C-2
C. 1.2 Specification of Demographic and Baseline Incidence Data Inputs	C-12
C.1.3 Study Area Selection	C-12
C.1.4 Generation of Air Quality Inputs to the Risk Assessment	C-17
C. 1.5 Risk Modeling Approach	C-46
C.2 Supplemental Risk Results	C-47
C.2.1 Risk Summary Tables and Underlying CBSA-Level Risk Estimates	C-48
C.2.2 Impact of Alternative Standards on the Distribution of Risk Across Ambient PM2.5
Levels 	C-72
C.3 Characterizing Variability and Uncertainty in Risk Estimates	C-81
C.3.1 Quantitative Assessment of Uncertainty	C-83
C.3.2 Qualitative Uncertainty Analysis	C-84
C.3.3 Conclusion	C-92
C.4 PM2.5 Design Values for the Air Quality Projections	C-93
References	C-l 17
1

-------
This appendix provides supplemental information related to the risk assessment described
in section 3.3 of final particulate matter (PM) policy assessment (PA), including:
•	Additional technical detail on the risk assessment approach, including sources and
derivation of key inputs to the risk modeling process (section C. 1).
•	Supplemental risk results (section C.2) intended to provide additional context for the
summary risk estimates presented in the PA section 3.3.2, including:
•	The modeled risk estimates that underly summary tables presented in PA section
3.3.2 aggregated to the CBSA-level (i.e., the urban study area) (section C.2.1).
•	Additional graphics including line plots, maps and scatter plots illustrating the
distribution of the grid-level risk estimates (section C.2.2).
•	Characterization of variability and uncertainty related to the risk assessment (section C.3).
C.l ADDITIONAL TECHNICAL DETAIL ON THE RISK ASSESSMENT
APPROACH
As discussed in section 3.3 of the PM PA, our general approach to estimating PM2.5-
associated human health risks in this review utilizes concentration-response (CR) functions
obtained from epidemiology studies to link ambient PM2.5 exposure to risk in the form of
incidence (counts) of specific health effects. The derivation and use of this type of CR function
in modeling PM2.5-attributable risk is well documented both in previous PM NAAQS-related risk
assessments (section 3.1.2 of U.S. EPA, 2010) and in Section C.l.l of this appendix. Inputs
required to model risk using these CR functions are identified below (Figure C-l) and include (a)
the concentration-response (CR) functions themselves, which are obtained from epidemiologic
studies (section C.l.l), (b) baseline health incidence data and information on population
demographics (section C.1.2), and (c) modeled ambient PM2.5 concentrations corresponding to
air quality scenarios of interest (section C.1.5).
C-l

-------
Specification of
demographic
and baseline
incidence inputs
Selection of Health
Endpoints and
Specification of CR
functions (including
selection of epi studies)
Risk modeling
(including generation
of risk metrics)
Using BenMAP-CE
Model
Risk estimates
(metrics)
Study area
selection
T
Air quality
characterization
Figure C-l. Key inputs to the risk assessment
C.l.l Selection of Key Health Endpoints and Specification of Concentration-Response
Functions from Epidemiologic Studies
In selecting specific CR functions for the risk assessment, we focus on health outcomes
for which the PM ISA determines the evidence supports either a "causal" or a "likely to be
causal" relationship with short- or long-term PM2.5 exposures (U.S. EPA, 2019). As discussed in
Chapter 3 of this final PA (Table 3-1), these outcomes include the following:
•	mortality (resulting from long- and short-term exposure),
•	cardiovascular effects (resulting from long- and short-term exposure),
•	respiratory effects (resulting from long- and short-term exposure),
•	cancer (resulting from long-term exposure), and
•	nervous system effects (resulting from long-term exposure).
We have focused the analysis on short- and long-term PM exposure-related mortality,
reflecting its clear public health importance, the large number of epidemiologic studies available
for consideration, and the broad availability of baseline incidence data. The specific set of health
effect endpoints included in the risk assessment are:
•	Long-term PM exposure-related mortality, all-cause, ischemic heart disease (IHD)
related, lung-cancer related
C-2

-------
•	Short-term PM exposure-related mortality, all-cause/non-accidental
To identify specific epidemiologic studies for potential inclusion in the risk assessment,
we focus on U.S. multicity studies assessed in the ISA. These studies are identified in section
3.2.3.2.1 of this PA (Figures 3-3 to 3-6). Of these, we used the following criteria to identify the
specific set of studies for inclusion in the risk assessment:
•	National-scale coverage: We focus on epidemiology studies reporting national-level
effect estimates. Epidemiology studies that focus on individual cities or regions were
excluded. Focusing on national-level epidemiological studies has the benefit of
characterizing PIVh.s-associted risks broadly across the U.S. and in relatively large
populations (compared with single-city or regional studies), which tends to improve
precision in the effect estimated generated.
•	Evaluation of relatively lower ambient PM concentrations: In selecting epidemiology
studies, to the extent possible, we favored those studies which characterized the ambient
PM2.5-mortality relationship at levels at or near the current NAAQS, given that the risk
assessment would be focusing on evaluating risk associated with the current NAAQS.
•	Populations with available baseline incidence data: For some populations (e.g., diesel
truck drivers), it can be challenging to model risk at the national-level given uncertainties
associated with specifying key inputs for risk modeling (i.e., baseline incidence rates for
mortality endpoints and detailed national-level demographics). For that reason, we
favored those epidemiology studies providing effect estimates for populations readily
generalizable to the broader U.S. population (e.g., specific age groups not differentiated
by additional socio-economic, or employment attributes).
•	Estimates of long-term PM2.5 exposures based on hybrid modeling approaches: For long-
term PM2.5 exposures, we focus on epidemiologic studies that estimate exposures with
hybrid modeling approaches. The primary rationale for this decision is the agreement
between the design of these epidemiology studies (i.e., their use of hybrid-based
modeling approaches in characterizing ambient PM) and the hybrid air quality surfaces
we are using in this risk assessment. This general agreement between the air modeling
surfaces used in long-term mortality epidemiology studies and our air quality modeling
reduces uncertainty in the risk assessment.
•	Estimates of short-term PM2.5 exposures based on composite monitor data: Short-term
mortality epidemiology studies utilizing hybrid modeling approaches, which are fewer in
number compared with long-term mortality studies, tend to be regional in scope and
consequently, did not meet the criterion of providing national-scale effect estimates. For
that reason, in modeling short-term mortality, epidemiology studies utilizing composite-
monitor based exposure surrogates were used as the basis for deriving CR functions. We
recognize the uncertainty introduced into the modeling of short-term mortality due to the
use of effect estimated obtained from studies utilizing composite monitors. However, we
C-3

-------
felt these use of national-scale epidemiology studies was a more important criterion for
selection.1
•	Evaluation of potential confounders and effect modifiers: Preference was given, to the
extent possible, to those studies which more fully address potential confounders and
effect modifiers and to those studies which utilize individual- rather than ecological
measures in representing those confounders/effect modifiers. Recognizing that both
single- and multi-pollutant models have advantages and disadvantages in characterizing
the ambient PM-mortality relationship, to the extent possible, we include epidemiology
studies (and associated effect estimates) based on both single- and multi-pollutant
models.
•	Exploration of multiple approaches for estimating exposures: For studies that estimate
PM2.5 exposures using hybrid modeling approaches, preference was given to studies that
also explore additional methods for estimating exposures (i.e., multiple hybrid methods
or hybrid methods plus monitor-based methods) and compare health effect associations
across approaches.
Application of the criteria listed above resulted in the selection of the epidemiology
studies presented in Table C-l for inclusion in the risk assessment as sources of effect estimates.
Table C-l includes summary information on study design, details on the selection of effect
estimates, the derivation of beta values, and specification of CR functional form based on those
effect estimates for use in the risk assessment. The procedure used to derive CR functions
(including specification of the beta values and mathematical forms for those functions) is
described below.
The remainder of this section describes the method used in specifying the concentration-
response (CR) functions used in the PM NAAQS REA (information presented in this section is
drawn from BenMAP Manual, Appendix C with additional detail specific to the epidemiology
studies selected for use in this risk assessment).2 These CR functions translate changes in
ambient PM2.5 into changes in baseline incidence rates for specific disease endpoints utilizing
beta (P) values obtained from epidemiology studies studying the association between ambient
PM2.5 exposure and specific health endpoints. These beta values (and associated standard errors)
are based on effect estimates obtained from the underlying epidemiology studies (equation
below). In addition, the mathematical forms for the health impact functions specified for use in
1	After identifying studies for inclusion in the draft risk assessment and initiating analyses, we became aware that Di
et al., 2017a uses a hybrid model-based approach to estimate PM2 5 exposures. The primary effect estimate reported
for this study (which reflects copollutant modeling including ozone) is larger than effect estimates selected for this
risk assessment. Specifically, the copollutant model for Di et al., 2017a reports an increased daily mortality risk of
1.05% (95th CI: 0.95-1.15%) with this effect estimate being two to three times larger than similar effect estimates
used in this risk assessment and has a substantially tighter confidence interval (Table C-l). Given the approximate
linearity of the CR functions used, we anticipate that this difference in effect estimate would translate into a similar
magnitude of difference in modeled mortality incidence (i.e., 2-3 times higher had the Di et al., 2017a effect
estimate been used in the risk assessment).
2	https://www.epa.gov/ben.map/beninap-ce-mannal-and-appendices
C-4

-------
this risk assessment reflect the models used in the epidemiology studies providing those effect
estimates. Consequently, derivation of the beta values based on effect estimates from underlying
epidemiology studies (and specification of the form of the health impact functions) represents a
key step in the design of the REA.
The majority of the epidemiology studies providing effect estimates for this PM REA
utilized either Poisson or Cox proportional hazard models which result in exponential (or log-
linear) forms for the CR functions, where the natural logarithm of mortality incidence is a linear
function of PM2.5.3 If we let xo denote the baseline (starting) PM2.5 level, and xi denote the
control (ending) PM2.5 level, yo denote the baseline incidences rate of the health effect, and Pop
the underlying population count for the applicable demographic group in the spatial unit of
analysis4 we can derive the following CR function specifying the relationship between the
change in x, Ax= (xo- xi) and the corresponding change in y, Ay (mortality incidence):
Ay = y0[l — e~PAx] * Pop
Given that the epidemiology studies providing effect estimates for long-term exposure-
related mortality and short-term exposure-related mortality in the context of the current PM REA
(Table C-l) use different categories of models (Cox proportional hazard and Poisson/Logistic,
respectively) we describe the process of deriving the betas and specifying CR functional forms
separately for each of these endpoint categories. As noted earlier, the logit model utilized in
Zanobetti et al., 2014, is discussed at the end of the section covering short-term PIVh.s-related
mortality.
Derivation of betas for long-term PM2 5 exposure-related mortality
Cox proportional hazard models used to evaluate mortality associated with long-term
PM2.5 exposure are designed to model effects on population survival. This class of epidemiology
model is based on a hazard function, defined as the probability that an individual die at time t,
conditional on that individual having survived up to time t. As such, the hazard function
represents a time-specific snapshot of the rate of mortality (events per unit time) within a study
population. While the risk can vary over time, in the case of the Cox proportional hazard model,
it is assumed that the hazard ratio is constant. The proportional hazard model takes the form:
3	One study. Zanobetti et al., 2014, supporting the modeling of short-term PM2 5 exposure-related mortality provided
a logistic-based model form, which is discussed at the end of this section.
4	Spatial unit of analysis refers to the geographic scale at which the CR function is applied in generating a risk
(incidence) estimate (e.g., zip code, county, 12km grid cell). Typically, the spatial unit of analysis used in a REA is
based on the spatial scale reflected in the epidemiology study(s) supplying the effect estimates. For this REA, the
spatial unit of analysis is the 12km grid cell.
C-5

-------
h{X,t)=hQ{t)ex'P ,
Where X is a vector of explanatory variables, P is a vector of associated coefficients and
ho(t) is the baseline hazard (the risk when all covariates (X) are set to zero).
Epidemiology studies utilizing the Cox proportional hazard model in characterizing
ambient PM2.5-health effects typically report hazard ratios (HRs) as the effect estimate. HRs
represent the ratio of hazard functions for the baseline and control scenarios reflecting a specific
difference in ambient PM2.5 exposure (typically a 10 ug/m3 increment). The HR simplifies as
shown (with the baseline hazard ratio dropping out), allowing us to readily derive the Beta value
from this effect estimate:
J, r, _ h(X0,t) _ h0(t)ex°'P _
h(Xc,t) h0(t)exc'P
It is then possible to calculate the beta as follows:
„ _ InjHR)
" APM
As noted in Sutradhar and Austin, 2018, the HR associated with a Cox-proportional
hazard model may approximate the RR when the effect estimate (and consequently the beta) is
relatively small. This is the case with the effect on mortality modeled for long-term exposure to
ambient PM2.5 (i.e., the size of the effect estimate supports an assumed equivalency between HR
and RR). The near equivalency between the HR and RR, allows us to utilize the beta derived
above in a CR-function based on a log-linear functional form of the type presented earlier, to
model changes in mortality related to changes in ambient PM.
Derivation of betas for short-term PM2 5 exposure-related mortality
The epidemiology studies selected for use in modeling short-term PM2.5 exposure-related
mortality utilize both the Poisson (log-linear) model form (Baxter et al., 2017) and the logit
model form (Zanobetti et al., 2014).5 In both cases, the epidemiology studies provide effects in
terms of percent increase in mortality.
The log-linear (Poisson) model is used to evaluate effects associated with continuous
(count) events. With the log-linear (Poisson) model, the relative risk is simply the ratio of the
two risks:
5 Note that the Ito et al., 2013 study also utilizes a Poisson model. However, that study provides beta values
(including standard errors) and for that reason the results of this study are directly applicable in modeling changes in
mortality without any of the derivations presented here for the other studies.
C-6

-------
= Z° = eP'APM
yc
The derivation of the beta with a Poisson model specified RR is as follows. Taking the
natural log of both sides, the beta coefficient in the CR function underlying the relative risk can
be derived as:
_ InjRR)
" APM
The beta derived in this fashion can then be used with a log-linear functional form (as
presented earlier) to model changes in mortality related to changes in ambient PM.
The logistic model form is used to model dichotomous events. With the logistic model
form, when we are provided with a RR value, as is the case here, we can make a similar
assumption to that used above with the Cox proportional hazard function (i.e., that the OR and
RR approach equivalency under conditions of relatively small effect levels). That observation in
turn allows us to assume that
RR = yf = {1- yQ) x e~APM'P + yQ
yc
Then, assuming (based on the relatively small size of the baseline incidence) that:
e-APM.(3 ^ (1 _ yo) x e-APM.(3 + yo
=> RR = e~APM'P
It is then possible to calculate the underlying beta coefficient as follows:
mm) _ n
-APM ~ "
Since the derivation of the beta is based on the assumption of a log linear functional
form, we can apply the beta in a log-liner CR function of the form described earlier:
Ay = y0[l — e~PAx] * Pop
C-7

-------
Table C-l. Details regarding selection of epidemiology studies and specification of concentration-response functions for the
risk assessment.





Additional









Location of
notes
Epi-



Selected




study effect
regarding
demio-

Selected

beta
Reference and

Exposure Estimation
CR function
estimate(s) in
effect estimate
logic
Mortality
effect
Selected
standard
study title
Study description
Approach
model
iournal article
selection
statistic
endpoint
estimate
beta
error (SE)
Long-term exposure-related mortality studies
Di et al., 2017b
Exploring relationship
Exposures estimated at
Cox proportional-
Table 2
Using single
Hazard
All-
1.084
8.07E-
1.18E-04

between air pollution (ozone,
zip code of residence
hazards model
Risk of death
pollutant, full
ratio (95
cause
(1.081-
03

Air Pollution and
PM2.5) and mortality
based on a neural network
with a
associated with
PM range
percent CI)

1.086)


Mortality in the
Key details:
model that incorporates
generalized
an increase of
model (model





Medicare
- Medicare population (65+)
satellite data, chemical
estimating
10 |jg/m3 PM2.5
for <12 |jg/m3





Population
- ecological control for
transport modeling, land-
equation to
or an increase of
applicable to






confounders
use terms, meteorology
account for the
10 ppb in ozone
only low-ozone






- all-cause mortality only
data, monitoring data, and
correlation
concentration.
days)6






- provides CR function
other data
between ZIP
Uses single







slopes for areas above and

codes
pollutant model







below the current PM


for full analysis.







NAAQS level (but model for










areas below current










standard only done for low










ozone cells)









Jerrett et al., 2016
Compares mortality effect
Multiple exposure
Cox proportional
Table 4 IHD,
Used the
Hazard
IHD
1.15(1.11-
1.40E-
1.78E-03

estimates for PM2.5 modeled
estimation approaches
hazard model
fully adjusted
ensemble
ratio (95

1.19)
02

Comparing the
from remote sensing to
evaluated - risk

(1990 ecological
estimate (pools
percent CI)




Health Effects of
those for PM2.5 modeled
assessment uses results

confounders)
effect estimates





Ambient
using ground-level
based on an ensemble

ensemble
generated





Particulate Matter
information.
approach that incorporates

estimate
using different





Estimated Using
- ACS cohort (Ages 30+)
chemical transport


exposure





Ground-Based
- IHD and diseases of
modeling, land use data,


estimates)





Versus Remote
circulatory system
satellite data, and data








Sensing Exposure
- individual-level confounder
from ground-based








Estimates
control
monitors








6 We note that Di et al., 2017b does include a copollutant model-based effect estimate (HR 1.073, 95th%CI 1.071-1.075). Had this effect estimate been used in
risk modeling (which would translate into a beta value of 7.05E-3), we would anticipate the risk estimates for all-cause mortality to be slightly less f 13% lower
based on comparison of calculated betas) than those estimated based on the single-pollutant model used in this risk assessment.
C-8

-------
Reference and
study title
Study description
Exposure Estimation
Approach
CR function
model
Location of
study effect
estimate(s) in
journal article
Additional
notes
regarding
effect estimate
selection
Epi-
demio-
logic
statistic
Mortality
endpoint
Selected
effect
estimate
Selected
beta
Selected
beta
standard
error (SE)
Pope et al., 2015
Relationships
Between Fine
Particulate Air
Pollution,
Cardiometabolic
Disorders, and
Cardiovascular
Mortality
Evaluates the relationship
between long-term exposure
to ambient PM2.5 and CVD
and cardiometabolic
disease, including effect
modification of the
relationships by pre-existing
cardiometabolic risk factors
-	ACS (30+) (oversampled
affluent individuals)
-	individual-level covariates
Exposures estimated at
home addresses based on
a land use regression and
Bayesian maximum
entropy (LUR-BME)
interpolation model that
incorporated data from
ground-based monitors
Cox proportional
hazard model
Table 1. Cox
model with
individual-level
plus ecological
covariates;
exposure based
on LUR-BME
NA
Hazard
ratio (95
percent CI)
All-
cause
1.07 (1.06-
1.09)
6.77E-
03
7.12E-04

NA
Hazard
ratio (95
percent CI)
IHD
1.14(1.1-
1.18)
1.31E-
02
1.79E-03
Thurston etal.,
2016
Ambient
Particulate Matter
Air Pollution
Exposure and
Mortality in the
NIH-AARP Diet
and Health Cohort
Reevaluates the relationship
between long-term exposure
to ambient PM2.5 and
mortality given recent
decline in U.S. ambient PM
concentrations.
Differentiation of risk for
fossil fuel PM2.5 versus total
PM2.5
-	NIH-AARP Cohort (only
select states - CA, FL, LA,
NJ, NC, PA, GA Ml) (55-
85yrs)
-	CVD, all-cause
-	residential locations
matched to census tract-
level PM2.5 estimates
Exposures estimated at
census tract centroids
based on land use data
and ground-based
monitors
Cox proportional
hazard model
Table 2. NIH-
AARP cohort
time
independent Cox
model PM2.5
mortality hazard
ratios (and 95th
percentile CI)
per 10 |jg/m3, by
cause and
cohort subgroup.
Cohort: ALL
NA
Hazard
ratio (95
percent CI)
All-
cause
1.03(1-
1.05)
2.96E-
03
1.24E-03
C-9

-------





Additional









Location of
notes
Epi-



Selected




study effect
regarding
demio-

Selected

beta
Reference and

Exposure Estimation
CR function
estimate(s) in
effect estimate
logic
Mortality
effect

Selected
standard
study title
Study description
Approach
model
journal article
selection
statistic
endpoint
estimate

beta
error (SE)
Turner et al., 2016
Evaluates the relationship
Exposures estimated at
Cox proportional
Table E4.
Note that the
Hazard
Lung
1.09 (1.03-

8.62E-
3.03E-03

between long-term exposure
residential locations based
hazard model
Adjusted HRs
non-cancer
ratio (95
cancer
1.16)

03

Long-Term Ozone
to ambient PM2.5 and all-
on land use data and

(95th percentile
mortality
percent CI)




Exposure and
cause and cause-specific
ground-based monitors

CI) for all-cause
endpoints





Mortality in a Large
mortality. Also, estimated


and cause-
provided in





Prospective Study
the association between


specific mortality
table E4 appear






PM2.5, regional PM2.5, and


in relation to
to mirror those






near-source PM2.5 and


each 10 unit
provided in






mortality in single-pollutant,


increase in PM2.5
Table 1 of Pope






copollutant and


LUR-BME
et al., 2015 -so






multipollutant models.


concentrations,
will use long-






-ACS (30+)


follow-up 1982-
cancer effect






- Includes lung cancer


2004, CPS-II
estimate from






(otherwise similar results to


cohort, United
this study only.






Pope et al., 2015)


States (n =







- county-level assessment


669,046).






Short-term exposure-related mortality studies
Baxter et al., 2017
Uses cluster-based
Exposure estimates based
Poisson (log-
Obtained from

Percent
24-hr
0.33(0.13-

3.29E-
1.02E-04

approach to evaluate the
on data from ground-
linear) at city-
results section in

increase in
non-
0.53)

04

Influence of
impact of residential
based monitors
level then
the text. After

24-hr
accident



exposure
infiltration factors on inter-

aggregated
pooling the city-

mortality
al



differences in city-
city heterogeneity in short-


specific effect

(95 percent
mortality



to-city
term PM-mortality


estimates into an

CI)




heterogeneity in
associations.


overall effect






PM2 5-mortality
- Mortality data from NCHS -


estimate, short-






associations in
77 U.S. CBSAs (all ages)


term PM2.5
MA





U.S. cities
- non-accidental mortality


exposure was
|\IM






- CBSA-level assessment


found to










increase 24-hr










non-accidental










mortality by










0.33% (95% CI:










0.13,0.53).










Based on lag 2










(day 0-1)






C-10

-------





Additional









Location of
notes
Epi-



Selected




study effect
regarding
demio-

Selected

beta
Reference and

Exposure Estimation
CR function
estimate(s) in
effect estimate
logic
Mortality
effect
Selected
standard
study title
Study description
Approach
model
journal article
selection
statistic
endpoint
estimate
beta
error (SE)
ltoetal.,2013
Use factor analysis to
Exposure estimates based
Poisson GLM
Appendix G,
Utilized lag-1
Betas with
24-hr all-
Study
1.45E-
7.47E-05

characterize pollution
on data from ground-

Table G.6 for
(all year) beta
SE (no
cause
provided
04

NPACT study 3.
sources, assess the
based monitors

Figure 4 - use
because that
conversion
mortality
bete and


Time-series
association between PM2.5


all-year lag 1
had the
required)

SE


analysis of
and PM2.5 components with


Bete:
strongest effect





mortality,
morbidity and mortality


Regression
for CVD





hospitalizations,
outcomes. Also evaluates


coefficients
mortality and





and ambient PM2.5
pollution levels, land-use,


(beta) and their
wanted our all-





and its
and other variables as


SE for air
cause to reflect





components
modifiers that may explain


pollutants at lag
that stronger






inter-city variation in PM-


0 through 3 days
lag-association






mortality effect estimates.


used to compute
for the CVD






- Mortality data from NCHS -


percent excess
effect (even






150 and 64 U.S. cities (two


risks in figures
though focusing






analyses) (all ages)


shown in the
on all-cause)






- MSA-level assessment


main text and in










Appendices B










and G










(corresponding










figures are










noted).






Zanobetti etal.,
Estimates the effect of short-
Exposure estimates based
Logistic
Table 2. Percent

Percent
All
0.64 (0.42-
6.38E-
1.09E-04
2014
term exposure to PM2.5 on
on data from ground-
regression
increase for 10

increase
deaths
0.85)
04


all-cause mortality.
based monitors

|jg/m3 increase

(95 percent




A national case-
Additionally, assesses the


in the two days

CI)




crossover analysis
potential for pre-existing


average PM2.5:






of the short-term
diseases to modify the


Combined






effect of PM2.5 on
association between PM2.5


across the 121






hospitalizations
and mortality (neurological


communities






and mortality in
disorders and diabetes)



NA





subjects with
- Medicare cohort -121 U.S.









diabetes and
communities (65+)









neurological
- Community-level









disorders
assessment (community










defined as the county or










contiguous counties










encompassing a city's










population)









C-ll

-------
C.1.2 Specification of Demographic and Baseline Incidence Data Inputs
This risk analysis requires both demographic and baseline-incidence data for the mortality
endpoint categories evaluated. For our analyses, these data are projected to the year 2015 since
the hybrid surfaces included in the analyses are based on a 2015 model year7. The BenMAP-CE
model8 is used in this risk assessment and the relevant demographic and baseline incidence data
for the contiguous U.S., from the sources described below, is readily available within the current
version of BenMAP-CE:
•	Demographic data: BenMAP-CE includes 2010 U.S. Census block-level age, race,
ethnicity and gender-differentiated data which the program can aggregate to various grid-
level definitions selected by the user, including the 12 km grid coverage used for risk
modeling in this analysis. In addition, BenMAP-CE has the ability to project future
demographics using county-level projections provided by Woods & Poole (2015). See
BenMAP-CE manual Appendix J for additional detail.9
•	Baseline incidence data for mortality endpoints: County-level mortality and population
data from 2012-2014 for seven causes of death in the contiguous U.S. was obtained from
the Centers for Disease Control (CDC) WONDER database. To estimate values for 2015,
we applied annual adjustment factors, based on a series of Census Bureau projected
national mortality rates for all-cause mortality. See BenMAP-CE manual Appendix D for
additional detail.9
C.1.3 Study Area Selection
In selecting U.S. study areas for inclusion in the risk assessment, we focus on the
following characteristics:
•	Available ambient monitors: We focus on areas with relatively dense ambient monitoring
networks, where we have greater confidence in adjustments to modeled air quality
concentrations in order to simulate "just meeting" the current and alternative primary
PM2.5 standards (air quality adjustments are described below in section C.1.4).
•	Geographical Diversity. We focus on areas that represent a variety of regions across the
U.S. and that include a substantial portion of the U.S. population.
•	PM2.5 air quality concentrations: We balance the value of including a broad array of
study areas from across the U.S. against the larger uncertainty associated with air quality
adjustments in certain areas. For example, many areas have recent air quality that meets
the current primary PM2.5 standards. Inclusion of such areas in the risk assessment
necessitates an upward adjustment to PM2.5 air quality concentrations in order to simulate
7	The 2015 model year was the most recent CMAQ modeling platform available at the time of the design of the risk
assessment and represents the central year of the 2014-2016 design value (DV) period. A single modeling year
was used in the risk assessment, rather than modeling risk for the full three-year design value period, because
model inputs for the 2016 period were not available at the time of the study (section C. 1.4.3).
8	https://www.epa.gov/beninap
9	https://www.epa.gov/ben.map/ben.map-ce-manual-and-appendices
C-12

-------
just meeting the current standards. Given uncertainty in how such increases could
potentially occur, we select areas requiring either a downward adjustment to air quality or
a relatively modest upward adjustment (i.e., no more than 2.0 ug/nr for the annual
standard and 5 |ig/m3 for the 24-hour standard). In addition, as discussed further in
section C.l .4.2, we excluded several areas that appeared to be strongly influenced by
exceptional events.
Applying these criteria resulted in the inclusion of 47 core-based statistical areas
(CBSAs) as study areas. These 47 study areas are identified in Figure C-2, with colors indicating
whether they meet either or both the design value cutoffs. Green indicates areas that only exceed
a 24-hr design value of 30 |ag/m3, blue indicates areas that only exceed an annual design value of
10 jag/m3, and red indicates areas that exceed both cutoffs.
45

-------
Q?
£>
0%m w
o

vQ
%
2018 Population
¦	0 to 8.880
¦	8.880 to 18.600
18,600 to 36,800
E 36,800 to 93,800
'¦ 93,800 to 10,300,000
Figure C-3. Map of the 2018 U.S. population by CBSA, with the selected urban study areas
outlined.
C-14

-------
CBSA
New York-Newark-Jersey City, NY-NJ-PA
Los Angeles-Long Beach-Anaheim, CA
Chicago-Naperville-Elgin, IL-IN-WI
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
Houston-The Woodlands-Sugar Land, TX
Atlanta-Sandy Springs-Roswell, GA
Detroit-Warren-Dearborn, Ml
Riverside-San Bernardino-Ontario, CA
St. Louis, MO-IL
Pittsburgh, PA
Sacramento-Roseville-Arden-Arcade, CA
Cincinnati, OH-KY-IN
Las Vegas-Henderson-Paradise, NV
Indianapolis-Carmel-Anderson, IN
Cleveland-Elyria, OH
Louisville-Jefferson County, KY-IN
Birmingham-Hoover, AL
Salt Lake City, UT
Fresno, CA
Akron, OH
Bakersfield, CA
Little Rock-North Little Rock-Conway, AR
Stockton-Lodi, CA
McAllen-Edinburg-Mission, TX
Lancaster, PA
Ogden-Clearfield, UT
Modesto, CA
Visalia-Porterville, CA
Canton-Massillon, OH
Provo-Orem, UT
Evansville, IN-KY
South Bend-Mishawaka, IN-MI
San Luis Obispo-Paso Robles-Arroyo Grande, CA
Merced, CA
Macon, GA
Elkhart-Goshen, IN
Napa, CA
Madera, CA
El Centro, CA
Wheeling, WV-OH
Johnstown, PA
Hanford-Corcoran, CA
Altoona, PA
Lebanon, PA
Weirton-Steubenville, WV-OH
Logan, UT-ID
Prineville, OR
OM 1M 2M 3M 4M 5M 6M 7M 8M 9M 10M 11M 12M
Population
Figure C-4. Population counts for ages 30 and above from each of the 47 CBSAs included
in the risk assessment.
C-15

-------
Table C-2. Population of the 47 urban study areas stratified by age.
Population Age Range
(Years)
Study Area Groupings (Millions)
47
30 (Annual-Controlled)
11 (24-hr-Controlled)
0-99
98.5
82.5
7.2
30-99
58.4
49.5
3.9
65-99
13.2
11.1
0.8
55-85
23.5
19.9
1.5
As noted in section 3.3 of this final PA and illustrated in Figure C-5, the 47 urban study
areas include 30 study areas where just meeting the simulated standards is controlled by the
current annual standard (12 |ig/m3), 11 study areas where just meeting the simulated standards is
controlled by the current 24-hr standard (35 |ig/m3), and 6 study areas where just meeting the
simulated standards is controlled by either the annual or 24-hr standard, depending on the air
quality scenario and adjustment strategy (discussed more fully in section C.1.4).
C-16

-------
30	Annual (Blue)	~50M
11	Daily (Green)	~4M
6	Mixed (Grey)	~5M
Total: 47	~60M
Number of Urban Study Controlling Population (>30
Areas (CBSAs)	Standard	years old)
Figure C-5. Map of 47 Urban Study Areas Reflected in Risk Modeling Identifying Subsets
Reflected in Risk Modeling (population estimates in millions of people).
C.1.4 Generation of Air Quality Inputs to the Risk Assessment
As described in detail below, air quality modeling was used to develop gridded PM2.5
concentration fields for the risk assessment. A PM2.5 concentration field for 2015 was developed
using a Bayesian statistical model that calibrates chemical transport model (CTM) predictions of
PM2.5 to surface measurements (Chapter 2, section 2.3.3). The 2015 PM2.5 concentration field
was then adjusted to correspond to just meeting the existing and potential alternative standards
using response factors developed from CTM modeling with emission changes relative to 2015.
The modeling approach applies realistic spatial response patterns from CTM modeling to a
concentration field, similar to those used in a number of recent epidemiologic studies, to
characterize PM2.5 fields at 12 km resolution for study areas.
C-17

-------
The adjustments to simulate just meeting the current standards and alternative standards
are approximations of these air quality scenarios. In reality, changes in PM2.5 in an area will
depend on what emissions changes occur and the concentration gradients of PM2.5 will vary
across an area accordingly. For our analyses, two different adjustment approaches were applied
to provide two outcomes that could represent potential bounding scenarios of PM2.5
concentrations changes across the study area. The two adjustment approaches used to guide the
generation of these modeled surfaces were:
•	Primary PM-based modeling approach (Pri-PM)\ This modeling approach simulates air
quality scenarios of interest by preferentially adjusting direct (i.e., primary, directly-
emitted) PM emissions. As such, the changes in PM2.5 tend to be more localized near the
direct emissions sources of PM. In locations for which air quality scenarios cannot be
simulated by adjusting modeled primary emissions alone, SO2 and NOx precursor
emissions are additionally adjusted to simulate changes in secondarily formed PM2.5.
•	Secondary PM-based modeling approach (Sec-PM): This modeling approach simulates
air quality scenarios of interest by preferentially adjusting SO2 and NOx precursor
emissions to simulate changes in secondarily formed PM2.5. In this case, the reductions in
PM2.5 tend to be more evenly spread across a study area. In locations for which air quality
scenarios cannot be simulated by adjusting precursor emissions alone, a proportional
adjustment of air quality is subsequently applied.
The air quality surfaces generated using these two approaches are not additive. Rather, they
should be viewed as reflecting two different broad strategies for adjusting ambient PM2.5 levels.
In addition, we also employed linear interpolation and extrapolation to simulate air
quality under two additional alternative annual standard levels, 11.0 and 9.0 |ig/m3, respectively
(section 3.3.1 of the PA, Figure 3-11). Interpolation and extrapolation were only performed for
grid cells in the subset of 30 urban study areas where the annual standard was controlling in both
Pri-PM and Sec-PM simulated air quality scenarios of both 12/35 and 10/30 standard
combinations. The interpolation and extrapolation were completed at the grid-cell level based on
values simulated using hybrid air quality modeling to just meet the current annual standard of
12.0 ug/m3 and alternative annual standard of 10.0 ug/m3 (section 3.3.1 of the PA, Figure 3-11).
A similar linear extrapolation/interpolation was not conducted for additional 24-hr standards due
to the weaker relationship between the 98th percentile of 24-hr PM2.5 concentrations, which are
most relevant for simulating air quality that just meets the 24-hour standard, and the
concentrations comprising the middle portion of the PM2.5 air quality distribution, which are
most relevant for estimating risks based on information from epidemiologic studies (i.e.,
discussed further in sections 3.1.2 and 3.2.3.2 in the PA).
The sections below provide more detailed information on the air quality modeling
approach used to adjust air quality to simulate just meeting the current or alternative primary
C-18

-------
PM2.5 standards. Tables containing PM2.5 DVs for the air quality projections can be found in
section C.4.
C.l.4.1 Overview of the Air Quality Modeling Approach
To inform risk calculations, recent PM2.5 measurements were analyzed to characterize the
magnitude and spatial distribution of PM2.5 concentrations. These data were then coupled with
air quality modeling data to project ambient air quality levels corresponding to just meeting the
existing and alternative PM2.5 NAAQS10 in specific areas. An overview of the approach is
provided in Figure C-6. The process starts by acquiring PM2.5 monitoring data from EPA's Air
Quality System (AQS)11 and simulating PM2.5 concentrations with the Community Multiscale
Air Quality (CMAQ)12 model for base case and emission-sensitivity scenarios (Figure C-6, Box
1). The monitored and modeled data are then fused using the Downscaler model and the
Software for Model Attainment Test-Community Edition (SMAT-CE)13 to develop a baseline
spatial field of PM2.5 concentrations and relative response factors (RRFs) for projecting PM2.5
concentrations, respectively (Figure C-6, Box 2). PM2.5 concentrations are projected in two main
steps using output from Downscaler and SMAT-CE (Figure C-6, Box 3). First, the PM2.5
concentrations measured at monitoring sites in an area are iteratively projected using the RRFs to
identify the percent change in anthropogenic emissions required for the highest monitored DV in
the area to just meet the controlling standard. Second, gridded spatial fields of PM2.5
concentrations are projected using the area-specific percent emission change14 that corresponds
to just meeting the standard at the controlling ambient data site. Additional details on the method
are provided in (Kelly et al., 2019a; application of the method to the PM NAAQS risk
assessment is described in the remainder of this appendix.
10	The phrase, "just meeting the PM2 5 NAAQS" is defined as the conditions where the highest design value (DV) for
the controlling standard in the area equals the existing or alternative NAAQS level under consideration. DVs are
statistics used in judging attainment of the NAAQS (www.epa.gov/air-frends/air-qnalitv-ciesign-vaines').
11	www.epa.gov/aas
12	www.epa. gov/cmaci
13	www.epa.gov/scram/photochemicai-modeling-toois
14	Scenarios based on a statistical projection approach were also developed for certain cases as discussed below.
C-19

-------
4. Risk Assessment
• Measured total and
speciated PM25
AQS
Downscaler
• Gridded PM25 spatial field
based on CMAQ and AQS data
• Simulated PM25 for
baseline and emission
sensitivity cases
CMAQ
SMAT-CE
*	Modeled relative response
factors (RRFs)
•	Speciated PM2 5 at monitors
and grid cells
1. Primary Data
B. Project spatial field to correspond to
just meeting NAAQS at monitors
•	Subset area from national spatial field
•	Project PM25 field using the RRF field
at the required % emission change
•	Output the adjusted field
2. Data Fusion
A. Project monitors to just meet NAAQS
•	Select areas, NAAQS levels, and
emission case(s)
•	Iteratively project PM2 5 to meet target
standards for emission case(s)
•	Output projected PM25 design values,
controlling monitor and standard, and
required percent emission change
3. Projecting PM2 5 to Target Standards
Figure C-6. Overview of the system for projecting PM2.5 concentrations to correspond to
just meeting NAAQS. See section C. 1.4.6 and Kelly et al., 2019a for more details.
C.l.4.2 PM2.5 Monitoring Data and Area Selection
The 2014-2016 DV period was the most recent period having a complete set of total and
speciated PM2.5 observations available at the time of the study. PM2.5 concentrations from the
2014-2016 DV period were used in selecting study areas and as the starting point for air quality
projections (Figure C-6, Box 1, "AQS"). Total and speciated PM2.5 concentrations for the 2014-
2016 DV period were acquired from AQS. For sites in Los Angeles and Chicago, DVs were
invalid during the 2014-2016 period. Los Angeles and Chicago have large populations, recent
valid DVs for sites in Los Angeles are above existing standards, and Chicago is part of a CBSA
that includes sites with valid 2014-2016 DVs in Indiana. For these reasons, invalid data for sites
in these areas were replaced with valid data from other recent periods to enable DVs to be
approximated for inclusion in the assessment. Specifically, for sites in Los Angeles and Orange
Counties in California, observations from April - October 2014 were replaced with observations
from the same months in 2013. For sites in Cook, DuPage, Kane, McHenry, and Will Counties in
Illinois, observations from January to mid-July 2014 were replaced with observations from the
same months in 2015.
Of the 56 areas initially identified as above the 10/30 selection threshold15, DVs for seven
areas16 appeared to meet the threshold due to the influence of wildfires. The influence of
15	"10/30" indicates an annual standard level of 10 ng/ m3 and a 24-hr standard level of 3 ng m~3
16	Butte-Silver Bow, MT; Helena, MT; Kalispell, MT; Knoxville, TN; Medford, OR; Missoula, MT; and Yakima,
WA
C-20

-------
wildfires on DVs for these areas was estimated in part by recalculating 2014-2016 DVs with
days removed that were clearly associated with summertime wildfires in the northwest. Since
wildfire influence is often excluded when judging NAAQS attainment, these seven areas were
excluded from further consideration. Additionally, the Eugene, OR CBS A was excluded. One
monitor in the Eugene CBS A has a 24-hr 2014-2016 DV slightly above the 10/30 selection
threshold17, but the monitor is in a small valley in Oakridge with very local high concentrations
of PM2.5 in winter that are distinct from conditions in the broader CBSA. Finally, the Phoenix-
Mesa-Scottsdale, AZ CBSA was excluded. This CBSA had one monitor slightly above the 10/30
DV threshold18, but projecting concentrations for the CBSA was judged to be relatively uncertain
because the annual DV is invalid at the only site that exceeded the threshold and the 24-hr DV is
just above the threshold.
The remaining 47 CBSAs were selected for the risk assessment. These areas are shown in
Figure C-7. The maximum 2014-2016 DVs and associated sites for each CBSA are provided in
Table C-3, and the counties associated with the CBSAs are listed in Table C-4. DVs were
calculated to an extra digit of precision for the air quality projections compared with official
DVs. This approach is consistent with DV calculations in previous air quality projections (e.g.,
USEPA, 201219) and provides a precise target for the iterative projection calculations.
17	The 410392013 monitor in Oakridge has a 24-hr 2014-2016 DV of 31 ng m3
18	The 040213015 monitor in the Phoenix-Mesa-Scottsdale, AZ CBSA has 24-hr 2014-2016 DV of 31 |ig m 3
19	USEPA (2012) Regulatory Impact Analysis for the Final Revisions to the National Ambient Air Quality
Standards for Particulate Matter. Office of Air Quality Planning and Standards, Health and Environmental
Impacts Division, Research Triangle Park, NC 27711. EPA-452/R-12-005 Available:
https://www3.epa.gov/ttii/ecas/regdata/RIAs/finalria.pdf
C-21

-------
V*.
.	, qmgmu
¦o 40	Above 10 annual and 30 daily
Above 30 daily
Above 10 annual
^ 35-	m
-
30-
'% \ l , w . *	Gulf of
2^ Jdan dawi|S0t9 GGoofe. INEGl	^
-120	-110	-100	-90	-80	-70
Longitude
Figure C-7. CBSAs selected for the risk assessment. Colors indicate whether the maximum
2014-2016 DVs in the CBSA are above the annual (10 |ig/m3) and/or 24-hr (30 |ig/m3)
selection criteria.
C-22

-------
Table C-3. Maximum annual and 24-hr PM2.5 DVs for 2014-2016 and associated sites for
selected CBSAs.
CBSA Name
# of
Annual
Annual
Max 14-16
DV
24-hr Max
24-hr Max
Sites
Max Site
Site
14-16 DV
Akron, OH
2
391530017
10.99
391530017
23.7
Altoona, PA
1
420130801
10.11
420130801
23.8
Atlanta-Sandy Springs-Roswell, GA
6
131210039
10.38
131210039
19.7
Bakersfield, CA
5
060290016
18.45
060290010
70.0
Birmingham-Hoover, AL
4
010732059
11.25
010730023
22.8
Canton-Massillon, OH
2
391510017
10.81
391510017
23.7
Chicago-Naperviiie-Elgin, IL-IN-Wia
22
170313103
11.10
170310057
26.8
Cincinnati, OH-KY-IN
9
390610014
10.70
390170020
24.2
Cleveland-Elyria, OH
8
390350065
12.17
390350038
25.0
Detroit-Warren-Dearborn, Ml
11
261630033
11.30
261630033
26.8
El Centra, CA
3
060250005
12.63
060250005
33.5
Elkhart-Goshen, IN
1
180390008
10.24
180390008
28.6
Evansville, IN-KY
4
181630023
10.11
181630016
22.0
Fresno, CA
4
060195001
14.08
060190011
53.8
Hanford-Corcoran, CA
2
060310004
21.98
060310004
72.0
Houston-The Woodlands-Sugar Land, TX
4
482011035
11.19
482011035
22.4
Indianapolis-Carmel-Anderson, IN
7
180970087
11.44
180970043
26.0
Johnstown, PA
1
420210011
10.68
420210011
25.8
Lancaster, PA
2
420710012
12.83
420710012
32.7
Las Vegas-Henderson-Paradise, NV
4
320030561
10.28
320030561
24.5
Lebanon, PA
1
420750100
11.20
420750100
31.4
Little Rock-North Little Rock-Conway, AR
2
051191008
10.27
051191008
21.7
Logan, UT-ID
1
490050007
6.95
490050007
34.0
Los Angeles-Long Beach-Anaheim, CAa
9
060371103
12.38
060371103
32.8
Louisville/Jefferson County, KY-IN
7
180190006
10.64
180190006
23.9
Macon, GA
2
130210007
10.13
130210007
21.2
Madera, CA
1
060392010
13.30
060392010
45.1
McAllen-Edinburg-Mission, TX
1
482150043
10.09
482150043
25.0
Merced, CA
2
060470003
11.81
060472510
39.8
Modesto, CA
2
060990006
13.02
060990006
45.7
Napa, CA
1
060550003
10.36
060550003
25.1
New York-Newark-Jersey City, NY-NJ-PA
17
360610128
10.20
340030003
24.5
Ogden-Clearfield, UT
3
490570002
8.99
490110004
32.6
Philadelphia-Camden-Wilmington, PA-NJ-DE-
MD
10
420450002
11.46
421010055
27.5
Pittsburgh, PA
10
420030064
12.82
420030064
35.8
Prineville, OR
1
410130100
8.60
410130100
37.6
Provo-Orem, UT
3
490494001
7.74
490494001
30.9
Riverside-San Bernardino-Ontario, CA
2
060658005
14.48
060658005
43.2
Sacramento-Roseville-Arden-Arcade, CA
6
060670006
9.31
060670006
31.4
Salt Lake City, UT
3
490353006
7.62
490353010
41.5
San Luis Obispo-Paso Robles-Arroyo Grande,
CA
3
060792007
10.70
060792007
25.9
C-23

-------
CBSA Name
# of
Sites
Annual
Max Site
Annual
Max 14-16
DV
24-hr Max
Site
24-hr Max
14-16 DV
South Bend-Mishawaka, IN-MI
1
181410015
10.45
181410015
32.5
St. Louis, MO-IL
6
290990019
10.12
295100007
23.7
Stockton-Lodi, CA
2
060771002
12.23
060771002
38.7
Visalia-Porterville, CA
1
061072002
16.23
061072002
54.0
Weirton-Steubenville, WV-OH
4
390810017
11.75
390810017
27.2
Wheeling, WV-OH
2
540511002
10.24
540511002
22.5
a DVs for Chicago-Naperville-Elgin, IL-IN-WI and
described in section C. 1.4.2.
.os Angeles-Long Beach-Anaheim, CA were approximated as
Table C-4. Counties associated with selected CBSAs
CBSA Name
Associated Counties
Akron, OH
Portage, Summit
Altoona, PA
Blair
Atlanta-Sandy Springs-Roswell, GA
Barrow, Bartow, Butts, Carroll, Cherokee, Clayton, Cobb, Coweta,
Dawson, DeKalb, Douglas, Fayette, Forsyth, Fulton, Gwinnett,
Haralson, Heard, Henry, Jasper, Lamar, Meriwether, Morgan,
Newton, Paulding, Pickens, Pike, Rockdale, Spalding, and Walton
Bakersfield, CA
Kern
Birmingham-Hoover, AL
Bibb, Blount, Chilton, Jefferson, St. Clair, Shelby, and Walker
Canton-Massillon, OH
Carroll, Stark
Chicago-Naperville-Elgin, IL-IN-WI
Cook, DeKalb, DuPage, Grundy, Kane, Kendall, Lake, McHenry,
Will, Jasper, Lake, Newton, Porter, and Kenosha
Cincinnati, OH-KY-IN
Dearborn, Ohio, Union, Boone, Bracken, Campbell, Gallatin,
Grant, Kenton, Pendleton, Brown, Butler, Clermont, Hamilton, and
Warren
Cleveland-Elyria, OH
Cuyahoga, Geauga, Lake, Lorain, and Medina
Detroit-Warren-Dearborn, Ml
Lapeer, Livingston, Macomb, Oakland, St. Clair, and Wayne
El Centra, CA
Imperial
Elkhart-Goshen, IN
Elkhart
Evansville, IN-KY
Posey, Vanderburgh, Warrick, and Henderson
Fresno, CA
Fresno
Hanford-Corcoran, CA
Kings
Houston-The Woodlands-Sugar Land, TX
Austin, Brazoria, Chambers, Fort Bend, Galveston, Harris,
Liberty, Montgomery, and Waller
Indianapolis-Carmel-Anderson, IN
Boone, Brown, Hamilton, Hancock, Hendricks, Johnson, Madison,
Marion, Morgan, Putnam, and Shelby
Johnstown, PA
Cambria
Lancaster, PA
Lancaster
Las Vegas-Henderson-Paradise, NV
Clark
Lebanon, PA
Lebanon
Little Rock-North Little Rock-Conway, AR
Faulkner, Grant, Lonoke, Perry, Pulaski, and Saline
C-24

-------
CBSA Name
Associated Counties
Logan, UT-ID
Franklin, Cache
Los Angeles-Long Beach-Anaheim, CA
Los Angeles and Orange
Louisville/Jefferson County, KY-IN
Clark, Floyd, Harrison, Scott, Washington, Bullitt, Henry,
Jefferson, Oldham, Shelby, Spencer, and Trimble
Macon, GA
Bibb, Crawford, Jones, Monroe, and Twiggs
Madera, CA
Madera
McAllen-Edinburg-Mission, TX
Hidalgo
Merced, CA
Merced
Modesto, CA
Stanislaus
Napa, CA
Napa
New York-Newark-Jersey City, NY-NJ-PA
Bergen, Essex, Hudson, Hunterdon, Middlesex, Monmouth,
Morris, Ocean, Passaic, Somerset, Sussex, Union, Bronx,
Dutchess, Kings, Nassau, New York, Orange, Putnam, Queens,
Richmond, Rockland, Suffolk, Westchester, and Pike
Ogden-Clearfield, UT
Box Elder, Davis, Morgan, and Weber
Philadelphia-Camden-Wilmington, PA-NJ-
DE-MD
New Castle, Cecil, Burlington, Camden, Gloucester, Salem,
Bucks, Chester, Delaware, Montgomery, and Philadelphia
Pittsburgh, PA
Allegheny, Armstrong, Beaver, Butler, Fayette, Washington, and
Westmoreland
Prineville, OR
Crook
Provo-Orem, UT
Juab and Utah
Riverside-San Bernardino-Ontario, CA
Riverside and San Bernardino
Sacramento-Roseville-Arden-Arcade, CA
El Dorado, Placer, Sacramento, and Yolo
Salt Lake City, UT
Salt Lake, and Tooele
San Luis Obispo-Paso Robles-Arroyo
Grande, CA
San Luis Obispo
South Bend-Mishawaka, IN-MI
St. Joseph and Cass
St. Louis, MO-IL
Bond, Calhoun, Clinton, Jersey, Macoupin, Madison, Monroe, St.
Clair, Franklin, Jefferson, Lincoln, St. Charles, St. Louis, Warren,
and St. Louis city
Stockton-Lodi, CA
San Joaquin
Visalia-Porterville, CA
Tulare
Weirton-Steubenville, WV-OH
Jefferson, Brooke, and Hancock
Wheeling, WV-OH
Belmont, Marshall, and Ohio
C.l.4.3 Air Quality Modeling
Air quality modeling was conducted using version 5.2.1 of the CMAQ modeling system
(Appel, 2018, Pye et al., 2018) to develop a continuous national field of PM2.5 concentrations
and estimates of how concentrations would respond to changes in PM2.5 and PM2.5 precursor
emissions (Figure C-6, "CMAQ"). The CMAQ modeling domain (Figure C-9) covered the
contiguous U.S. with 12 km horizontal resolution and 35 vertical layers. Since 2015 was the
C-25

-------
most recent modeling platform available at the time of the study and represents the central year
of the 2014-2016 DV period, 2015 was selected as the baseline modeling year for the PM2.5
projections. A single modeling year was used due to the time and resources needed to conduct
photochemical grid modeling, and because model inputs for the 2016 period were not available
at the time of the study.
Information on the CMAQ model configuration for the 2015 modeling is provided in
Table C-5. The 2015 model simulation and its evaluation against network measurements of
speciated and total PM2.5 has been described in detail previously (Kelly et al., 2019b). Model
performance statistics for PM2.5 organic carbon, sulfate, and nitrate were generally similar to or
improved compared to the performance for other recent national 12 km model simulations. One
exception to the generally good model performance was identified for the Northwest region (OR,
WA, and ID). Model performance statistics for this region were generally not as good as in our
recent modeling due to issues related to unusually high fire influences in 2015, atmospheric
mixing over sites near the Puget Sound, and other factors. However, model performance issues
in the Northwest have minimal influence on the risk assessment, because only two of the 47
CBSAs are in the Northwest region (i.e., Prineville, OR and part of the Logan, UT-ID, CBSA).
Also, the analysis uses ratios of model predictions rather than absolute modeled concentrations,
and systematic biases associated with mixing height and fire impact estimates may largely cancel
in the ratios. Moreover, fusion of monitor data with model predictions in developing PM2.5 RRFs
and the baseline concentration field helps mitigate the influence of biases in model predictions
(as discussed below). Overall, the model performance evaluation (Kelly et al., 2019b) indicates
that the 2015 CMAQ simulation provides concentration estimates that are generally as good or
better than in other recent applications and are reliable for use in projecting PM2.5 in the risk
assessment. Model performance statistics for PM2.5 by U.S. climate region and season are
provided in Table C-6 and statistic definitions can be found in Table C-7.
C-26

-------
CMAQ modeling domain.
Table C-5. CMAQ model configuration.
Category
Description
Grid resolution
12 km horizontal; 35 vertical layers
Gas-phase chemistry
Carbon Bond 2006 (CB6r3)
Organic aerosol
Non-volatile treatment for primary organic aerosol; secondary organic
aerosol from anthropogenic and biogenic sources
Inorganic aerosol
ISORROPIAII
NH3 surface exchange
Bi-directional NH3 surface exchange
Windblown dust emissions
Simulated online
Sea-spray emissions
Simulated online
Meteorology
Version 3,8 of Weather Research & Forecasting (WRF) Skamarock et
al.. 2005 model
Table C-6. Model performance statistics20,21 for P1VI2.5 at AQS sites for the 2015 base case.
Region21
Season
N
< 0 ™
Avg.
Mod.
(Kg m-3)
MB20
(M9 m"3)
NMB20
(%)
RMSE20
(M9 m"3)
NME20
(%)
f20

Winter
13001
10.04
12.74
2.71
27.0
7.33
48.0
0.68

Spring
13538
7,97
8.83
0.86
10.8
5.19
44.0
0.59
Northeast
Summer
13660
8,38
8.02
-0.36
-4,3
4.06
35.2
0.67

Fall
13270
7.18
9.08
1.90
26,5
5.40
50,0
0.73

Annual
53469
8.38
9,64
1.26
15,0
5.60
44,2
0.67

Winter
11190
8.07
10.28
2.21
27,4
5.65
47,4
0.58
Southeast
Spring
11961
8,06
8,25
0.18
2.3
4,08
33.6
0,55
Summer
11641
9,78
8.45
-1.33
-13.6
4,86
35.3
0,47

Fall
11365
6,93
8.13
1.20
17.3
4.32
41.7
0.70
20	See Table C-7 for definition of statistics.
21	See Figure C-10 for definition of regions.
C-27

-------
Region21
Season
N
<0 a
Avg.
Mod.
(ug m 3)
MB20
(H9 m"3)
NMB20
(%)
RMSE20
(H9 m"3)
NME20
(%)
r20

Annual
46157
8.22
8.76
0.54
6.6
4.75
39.1
0.55

Winter
10323
9.49
11.60
2.10
22.1
5.75
43.2
0.63

Spring
10867
8.90
9.85
0.95
10.6
4.60
36.3
0.65
Ohio Valley
Summer
10714
10.95
10.56
-0.39
-3.6
5.55
34.3
0.55

Fall
10568
8.41
10.96
2.54
30.2
6.23
47.1
0.65

Annual
42472
9.44
10.73
HT29
n\3£
15^56
^308
0.59

Winter
6478
8.79
9.72
0.92
10.5
4.75
38.2
0.70

Spring
6643
7.32
8.27
0.96
13.1
4.30
41.9
0.67
Upper Midwest
Summer
6718
7.88
7.85
-0.03
-0.4
5.26
40.8
0.56

Fall
6664
6.81
9.14
2.33
34.2
4.92
49.3
0.75

Annual
26503
7.69
8.74
1.04
13.6
4.82
42.2
0.64

Winter
8041
7.53
10.13
2.60
34.5
11.81
56.6
0.36

Spring
8369
8.08
7.12
-0.96
-11.9
4.24
36.3
0.51
South
Summer
8440
10.80
8.31
-2.49
-23.0
6.04
40.3
0.34

Fall
8340
7.55
7.99
0.44
5.9
3.76
35.5
0.63

Annual
33190
8.50
8.37
-0.13
-1.6
7.15
41.8
0.34

Winter
4911
7.46
7.90
0.45
6.0
6.50
55.9
0.52

Spring
4998
4.88
5.88
1.00
20.6
3.60
48.4
0.44
Southwest
Summer
5069
6.12
4.85
-1.27
-20.8
4.15
43.1
0.59
Fall
5091
5.31
15^90
IOT59
11.1
I435
I5Z2
0.49

Annual
20069
5.93
6.12
0.19
3.2
4.77
50.2
0.52

Winter
4987
5.57
3.60
-1.98
-35.5
6.80
63.4
0.23
N. Rockies &
Plains
Spring
5380
4.57
5.00
0.44
9.6
29.58
61.6
0.20
Summer
5260
9.98
7.68
-2.30
-23.1
17.61
57.4
0.57
Fall
5010
5.57
5.42
-0.15
-2.7
5.65
56.4
0.44

Annual
20637
6.43
5.45
-0.99
-15.3
18.06
59.2
0.34

Winter
8994
7.90
7.82
-0.08
-1.0
10.20
80.9
0.25

Spring
9306
5.02
6.84
1.82
36.2
6.65
71.5
0.48
Northwest
Summer
9993
9.17
11.12
1.95
21.2
32.40
67.7
0.46

Fall
9868
7.03
9.39
2.37
33.7
15.33
78.3
0.31

Annual
38161
7.31
8.85
1.55
21.2
19.26
74.3
0.43

Winter
10462
11.67
9.58
-2.08
-17.8
8.09
43.3
0.68

Spring
10989
7.52
6.95
-0.57
-7.6
4.17
38.3
0.55
West
Summer
11065
8.95
8.53
-0.43
-4.8
6.36
43.5
0.51

Fall
10587
8.61
9.11
0.50
5.8
16.85
46.9
0.37

Annual
43103
9.16
8.52
-0.64
-7.0
10.02
43.1
0.44
C-28

-------
Table C-7. Definition of statistics used in the CMAQ model performance evaluation.
Statistic
Description
MB (® m-3)=-JJ\L1{Pi-Oi)
Mean bias (MB) is defined as the average difference between
predicted (P) and observed (0) concentrations for the total number
of samples (n)
RMSE(Dgm-3) = v/£f=1(^-O077i
Root mean-squared error (RMSE)
CD
0*^
II
M
M cj
0 i
0
X
0
0
The normalized mean bias (NMB) is defined as the sum of the
difference between predictions and observations divided by the
sum of observed values
NME(%)=S^2 x 100
Li °l
Normalized mean error (NME) is defined as the sum of the
absolute value of the difference between predictions and
observations divided by the sum of observed values
r _ S?=1(Pi-P)(Oi-o)
jzhCPi-PfJxl^Ot-d)*
Pearson correlation coefficient
73 40-
Gulf of
Movirn
-80
Northeast
Northern Rockies & Plains
Northwest
Ohio Valley
South
Southeast
Southwest
Upper Midwest
West
25 "Oooale (Mao data©2018 Gooa!e.: INEGI (
-120	-100
Longitude
Figure C-10. U.S. climate regions22 used in the CMAQ model performance evaluation.
In addition to the national model performance evaluation just described, CMAQ
predictions of PM2.5 concentrations were evaluated specifically for the CBS As considered in the
risk assessment. In Table C-8, model performance statistics are provided for predictions at
monitors in the 47 CBSAs in 2015. Predictions generally agree well with observations over the
full set of areas, with NMBs less than 10% in all seasons except Fall (NMB: 23.6%) and
correlation coefficients greater than 0.60 in all seasons except Summer (r: 0.56). Model
predictions are compared with observations by CBSA in Figure C-l 1, and NMBs at individual
sites in the CBSAs are shown in Figure C-12. Predictions generally agree well with observations
in the individual CBSAs, although underpredictions occurred in the Chicago-Naperville-Elgin
: https://www.ncdc.noaa.gov/monitoring-references/maps/us-climate-regions.php
C-29

-------
CBSA when observed PM2.5 concentrations were > 40 |Lxg m"3. The high observed values in
Chicago were associated with the 4th of July holiday, and the underpredictions on July 4th and 5th
have small influence on the annual PM2.5 projections in the risk assessment. The NMB is highest
for model predictions in the Birmingham-Hoover CBSA (NMB: 66%). As mentioned above, the
effects of model bias are mitigated in part by use of relative response factors (i.e., the ratio model
predictions from a base and emission control simulation is used in projecting PM2.5
concentrations, and some model bias likely cancels in the ratio). For the risk assessment
projections, the key aspect of the CMAQ modeling is the spatial of pattern of PM2.5 response to
changes in emissions. The spatial response pattern was examined in the 47 CBSAs and found to
be reasonable even in areas with relatively high bias, such as Birmingham. In Figure C-13, the
spatial response pattern associated with the 10/30 projection case for the Birmingham-Hoover
CBSA is compared for the proportional projection method and the primary PM projection case
based on CMAQ modeling. Relatively high PM2.5 responsiveness occurred in the urban part of
Birmingham and along arterial roads in the CMAQ-based approach. This spatial pattern is
consistent with the location of PM2.5 emission sources in Birmingham and provides a realistic
spatial response pattern despite the relatively high bias in the concentration predictions. Overall,
both the national model performance evaluation and the evaluation for the 47 CBSAs of the risk
assessment support use of the CMAQ modeling in this application.
To inform PM2.5 projections, annual CMAQ modeling was conducted using the same
configuration and inputs as the 2015 base case simulation but with anthropogenic emissions of
primary PM2.5 or NOx and SO2 scaled by fixed percentages. Specifically, seven simulations were
conducted with changes in anthropogenic NOx and SO2 emissions (i.e., combined NOx and SO2,
not separate NOx and SO2 simulations) of-100%, -75%, -50%, -25%, +25%, +50%, and +75.
Two simulations were conducted with changes in anthropogenic PM2.5 emissions of -50% and
+50%). The sensitivity simulations were based on emission changes applied to all anthropogenic
sources throughout the year. These "across-the-board" emission changes facilitate projecting the
baseline concentrations to just meet a relatively wide range of standards in areas throughout the
U.S. using a feasible number of national sensitivity simulations.
C-30

-------
Table C-8. Performance statistics for CMAQ predictions at monitoring sites in the 47
CBSAs considered in the risk assessment.
Season
Average
Observed
(ng m 3)
Average
Modeled
(ng m 3)
MB
(ng m-3)
NMB
(%)
RMSE
(ng m 3)
NME (%)
r
Winter
12.40
13.45
1.05
8.5
8.03
42.4
0.61
Spring
9.17
9.94
0.77
8.4
5.15
38.6
0.62
Summer
10.35
10.08
-0.27
-2.6
5.51
34.6
0.56
Fall
9.00
11.11
2.12
23.6
6.26
45.6
0.67
C-31

-------
O)
3
_cd
CD
"D
o
120
80
40
0
120
80
40
0
120-
NMB: 5%

MB: 0.47
80
RMSE: 4.2

r: 0.64
40

0'
*

AkronO
120'
NMB: -13%

MB:-1.59
80
RMSE: 4.96

n 0.71
40
_
0
if-


120
Chicag
NMB: 45 %

MB: 4.47 ~
80
RMSE: 8.69

r:0j54 .
40
M • •


0
W
Evansv
Lancas
NMB: 7 %
MB: 0.79
RMSE: 5.8?
r: 0.72
*
Louisv
NMB: 9%
MB: 0.94
RMSE: 4.89
r: 0.65
120
NapaCA
NMB: -34 %

MB: -3.61 ~
80
RMSE: 7.75

r: 0.39
40

0


120-
ProvoO
NMB: -31 %

MB: -2.2
80
RMSE: 5.42

r: 0.45
40
• ~
0'
A'.



St. Lou
120
NMB: 21 %

MB: 2.16 ~
80
RMSE: 6.15

r: 0.60
40
JLz.-.
0
.
Altoon
NMB: 32 %
MB: -3.5
RMSE: 4.89
r: 0.77
Cincin
NMB: 24 %
MB: 2.35
RMSE: 5.06
r: 0.72
Fresno
NMB:-13%
MB: -4.73
RMSE: 8.48
r: 0.€tf .
••«•••
0-
LasVeg
NMB: 2 %
MB: 0.17
RMSE: 3.66
r: 0.74
MaconG
NMB: 10%
MB: 0.88
RMSE: 4.66
r: 0.60
NewYor
NMB: 52 %
MB: 4.61
RMSf: 8.17,
'.
-------
50
45
40
35
30
CD 25
"D ^
3
Fall
ro
50
45
40
35
30
25
-120
-100
-80
Spring




m i



I	0/-5—
—V_L[ 1
Summer
J ©
V
\ 1
—1—( °Xr\s
-120
Longitude
-100
-80
Winter
&
v
i—
1 1 °7 1'
^L [ j o0\ y
I
I
>120
100
80
60
40
20
0
-20
-40
-60
-80
-100
-120
Figure C-12. NMB for CMAQ PM2.5 predictions at monitoring sites in the 47 CBSAs by
season in 2015.
34.0-
Proportional
National Forest Cullman
/ >—l-
Primary PM
Bankhead „ „	$ j
Tl—4	
Bankhead
National Forest Cullman
/ ©
2 33.5
Tallade
a
... Tallad|< ;aloosa
—j National F<
:aloosa
nsboro
Map data,®201ft Google.
33.0-


\:*i -kL
! j *£}

Tajlade
Trr

Tallade'i
j -r-m

National F(
ft
&
nsboro	I I •
Map dataffiMS Gooflle, INEGf" " ~ -
• Controlling
NotCntling
% Chg
-5
-6
-7
-8
-9
-10
/o \
I
-87.6 -87.2 -86.8 -86.4 -87.6 -87.2 -86.8 -86.4
Longitude
Figure C-13. Percent change in 2015 annual average PM2.5 over the Birmingham CBSA
associated with projecting 2014-2016 DVs at monitors to just meet an alternative
NAAQS of 10/30 using the proportional projection method and the primary PM2.5,
CMAQ-based projection method.
C-33

-------
The two emission sensitivity scenarios (primary PM2.5 and NOx and SO2) were selected
to span a wide range of possible PM2.5 spatial response patterns. NOx and SO2 emission changes
influence concentrations of ammonium nitrate and ammonium sulfate, which are secondary
pollutants that often have broad spatial distributions. Primary PM2.5 emission changes have the
greatest influence on PM2.5 concentrations close to emission sources. The two distinctly different
PM2.5 response patterns for primary PM2.5 and NOx and SO2 emission changes enable PM2.5 to
be projected for a wide range of conditions. Projecting PM2.5 for a wide range of conditions is
desirable in this study because many PM2.5 spatial response patterns can cause PM2.5
concentrations to just meet NAAQS.
C.l.4.4 Relative Response Factors for PM2.5 Projection
The 2015 base case and sensitivity modeling results were used to develop RRFs for
projecting PM2.5 concentrations to correspond to just meeting NAAQS (Figure C-6, Box 2,
"SMAT-CE"). Baseline PM2.5 concentrations are projected by multiplication with RRFs. The
RRF for a PM2.5 species is calculated as the ratio of the concentration in the sensitivity
simulation to that in the base case:
nr>r?		 Csensitivity,species
KKrspecies ~~ ^	(¦*¦)
^base,species
where Csensitivity,species is the concentration of the PM2.5 species in the sensitivity
simulation, and Cbase,species is the concentration of the PM2.5 species in the base case simulation.
RRFs were calculated for each monitor, grid cell, calendar quarter, standard (annual or 24-hr),
species, and sensitivity simulation using SMAT-CE version 1.2.1. RRFs are used in projecting
air quality to help mitigate the influence of systematic biases in model predictions (National
Resources Council, U.S. EPA, 2018a). More details on the RRF projection method are provided
in EPA's modeling guidance document (U.S. EPA, 2018a) and the user's guide for the
predecessor to the SMAT-CE software (Abt Associates, 2014).
To apply the RRF approach for the risk assessment projections, RRFs for total PM2.5
were calculated from RRFs for the individual PM2.5 species using observation-based estimates of
PM2.5 species concentrations in SMAT-CE output. Specifically, total PM2.5 RRFs (RRFrot,PM2.5)
were calculated as the weighted average of the speciated RRFs using the observation-based
species concentrations (C species) as weights:
nnr?		 ^RRFspecies^species
KKrrot, PM2.5 ~ ^7	U)
Z. uspecies
Total PM2.5 RRFs were used to project base-case PM2.5 concentrations as follows:
^^2.5, projected ~ ^^^Tot,PM2.5^'^2.5, base	(3)
The species concentrations used in calculating the total PM2.5 RRFs were generally based
on application of the Sulfate, Adjusted Nitrate, Derived Water, Inferred Carbonaceous material
balance approacH (SANDWICH) (Frank, 2006) to measurements of PM2.5 species
C-34

-------
concentrations from the Chemical Speciation Network (CSN)23 and the Interagency Monitoring
of Protected Visual Environments (IMPROVE)24 network. The SANDWICH method corrects for
different artifacts in the measurements for PM2.5 species and total PM2.5. An alternative approach
to calculating total PM2.5 RRFs was applied for monitors and grid cells in California due to
factors including missing data at the Bakersfield speciation monitor25 throughout 2014 and part
of 2015. For projections in California, RRFs were calculated directly from the ratio of CMAQ
PM2.5 concentration predictions in the sensitivity simulation to the base simulation.
By default, PM2.5 RRFs for the annual standard are calculated using average
concentrations over all modeled days in the quarter, and RRFs for the 24-hr standard are
calculated using average concentrations over days with the top 10% of modeled PM2.5
concentration in the quarter. The default approach was generally followed here, with exceptions
for counties in the San Joaquin Valley (SJV) of California and Utah. In these counties26, the
average concentration over all days in the quarter was used to calculate RRFs for both the 24-hr
and annual standards for sites with valid 24-hr and annual DVs. This approach was used to
provide stability in projections of annual fields due the variability in the 24-hr and annual
RRFs27. Also, RRFs were set to one28 in the third quarter (July-September) for select counties in
the San Joaquin Valley and Utah29 to better reflect the seasonal nature of PM2.5 in these areas
(i.e., PM2.5 concentrations are relatively high in winter).
RRFs were calculated for each combination of emission sensitivity simulation and the
2015 base case. RRFs corresponding to the percent change in emissions for each sensitivity
simulation were then interpolated across the range of emission changes from -100 to +100% to
facilitate iterative projections of PM2.5 concentrations to the nearest percent emission change.
PM2.5 RRFs are shown in Figure C-14 and Figure C-15 as a function of changes in anthropogenic
primary PM2.5 and NOx and SO2 emissions for monitors in the U.S. during the first and third
23	www.epa.gov/amtic/cheniical-speciaflon-network-csn
24	http://vista.cira.colostate.edu/Improve/
25	Site identification number: 060290014
26	SJV counties: Fresno, Stanislaus, Kern, Merced, Madera, Tulare, San Joaquin, and Kings; Utah counties: Cache,
Box Elder, Davis, Morgan, Weber, Juab, Utah, Salt Lake, and Tooele.
27	This variability is less of an issue in regional modeling applications where emission changes can be targeted to
time periods of elevated PM2 5 concentrations in the area.
28	When the RRF is 1, the projected concentration equals the base concentration (Equation 3).
29	SJV counties: Fresno, Stanislaus, Kern, Merced, and Madera; Utah counties: Cache, Box Elder, Davis, Morgan,
Weber, Juab, Utah, Salt Lake, and Tooele. This approach was not applied for Kings, Tulare, and San Joaquin
counties in SJV because the percent exceedance of the annual standard was within 10% of the exceedance of the
24-hr standard suggesting that relatively uniform PM2 5 concentrations occur throughout the year compared with
the other SJV counties.
C-35

-------
calendar quarters. Spatial fields of PM2.5 RRFs for 50% reductions in anthropogenic primary
PM2.5 and NOx and SO2 emissions are shown in Figure C-16.
(a) July-September	(b) January-March
1	1	1	1	11	1	1	1	1
-100 -50 0 50 100-100 -50 0 50 100
Emission Change (%)
Figure C-14. Annual standard P1VI2.5 RRFs for quarters 1 and 3 as a function of the percent
change in anthropogenic primary PM2.5 emissions for monitoring sites in the contiguous
U.S.
1.50-
u_ 1-25-
DC
01 1.00-
c\i
0.75-
0.50-
-1
Figure C-15. Annual standard P1VI2.5 RRFs for quarters 1 and 3 as a function of the percent
change in anthropogenic NOx and SO2 emissions for monitoring sites in the contiguous
U.S.
(a) July-September	(b) January-March
0 50 100-100 -50 0
Emission Change (%)
C-36

-------
(a) 50% NOx and S02 Reduction
Mhd dal^ ©20' 8 Goocile IfvjE
(b) 50% Primary PM25 Reduction
-120 -110 -100	-90	-80	-70 -120 -110 -100	-90	-80	-70
Longitude
Figure C-16. Annual average PM2.5 RRFs at CMAQ grid-cell centers for 50% reductions in
anthropogenic (a) NOx and SO2 and (b) primary PM2.5 emissions.
(D
"O 40
C. 1.4.5 2015 PM2.5 Concentration Fields
To develop a baseline gridded PM2.5 concentration field for projection with PM2.5 RRFs,
a Bayesian statistical model (i.e., Downscaler) was applied (Figure C-6, Box 2, "Downscaler")
(Berrocal et al., 2012). Downscaler makes predictions of PM2.5 concentrations to a spatial field
of receptor points using PM2.5 monitoring data and CMAQ model predictions as inputs.
Downscaler takes advantage of the accuracy of the monitoring data and the spatial coverage of
the CMAQ predictions to develop new predictions of PM2.5 concentration over the U.S.
The Downscaler model is routinely applied by U.S. EPA to predict 24-hr average PM2.5
concentrations at the centroids of census tracts in the contiguous U.S. (U.S. EPA, 2018b). The
model configuration used here is generally consistent with the previous applications, but here
predictions were made to the centers of the CMAQ model grid cells rather than to census-tract
centroids. Also, PM2.5 measurements from the IMPROVE monitoring network were used in
addition to measurements included in the AQS database. 24-hr average PM2.5 concentrations
were predicted for the 2015 period, and the 24-hr PM2.5 fields were averaged to the quarterly
periods of the PM2.5 RRFs for use in projection.
Annual average PM2.5 concentrations from the monitoring network and CMAQ
simulation that were used in model fitting are shown in Figure C-17 along with the resulting
Downscaler predictions. Cross-validation statistics are provided in Table C-9 based on
comparisons of Downscaler predictions against the 10% of the observations that were randomly
withheld from model fitting.
C-37

-------
CD
45-
"O
-J
1
0


c3
35-
_i


30-
Observed
CMAQ
Downscaler


Vy J
rvO
v*
\Pr

-120 -110 -100 -90 -80 -70 -120 -110 -100 -90 -80 -70 -120 -110 -100 -90 -80 -70
Longitude
ug/m3
I
>20

15

10

5
0
Figure C-17. Annual average of the 2015 PM2.5 observations and CMAQ predictions used
in the Downscaler model, and the annual average of the Downscaler PM2.5 predictions.
Table C-9. Cross-validation statistics associated with the 2015 Downscaler predictions.
Number of Monitors
1101
Mean Bias3
(U9 rn -)
0.37
Root Mean Squared Errorb
	(BJg m3)	
3.17
Mean Coverage0
0.95
aThe mean of all biases across the CV cases, where the bias of each prediction is the downscaler
prediction minus the observed value.
bThe bias is squared for each CV prediction, then the square root of the mean of all squared
biases across all CV predictions is obtained.
CA value of 1 is assigned if the measured value lies in the 95th percentile CI of the Downscaler
prediction (the Downscaler prediction ± the Downscaler standard error), and 0 otherwise. This
column is the mean of all those O's and 1's,
C. 1.4.6 Projecting PM2.5 to Just Meet the Standards
PM2.5 was projected from baseline concentrations to levels corresponding to just meeting
NAAQS using the monitoring data (section C. 1.4.2), RRFs (section C.l.4.4), and baseline
concentration fields (section C.l.4.5) described above. The projection was done in two steps as
shown in Box 3 of Figure C-6. Projections were performed for the existing (12/35)30 and
alternative (10/30)31 standards.
First, monitors in the CBSA of interest were identified, and concentrations from these
monitors were subset from the national monitoring dataset. The measured concentrations were
then projected using the corresponding PM2.5 RRF. PM2.5 DVs were calculated using the
projected concentrations, and the difference between the maximum projected DV and target
standard was determined. DV projections over the complete range of percent emission changes (-
100 to 100%) were performed using bisection iteration until the difference between the
36 Annual standard level of 12 fig nr3 and 24-hr standard level of 35 jig nr3
31 Annual standard level of 10 jig nr3 and 24-lir standard level of 30 |ig nr3
C-38

-------
maximum projected DV in the CBS A and the standard level was zero or within the difference
associated with a 1% emission change. Iterative projections of annual and 24-hr DVs were
performed separately, and the controlling standard was determined as the standard requiring the
greater percent emission change32. In cases where the emission change needed to just meet the
target annual or 24-hr standard was outside of the ± 100% range, the standard could not be met
using the modeled air quality scenarios. If neither the annual nor 24-hr standard could be just met
with emission changes within ± 100%, then an alternative projection approach was used
(discussed below).
Second, 2015 PM2.5 concentration fields developed with Downscaler were projected
according to the percent emission change required for the maximum projected DV to just meet
the controlling standard. The projection was done by multiplying the gridded spatial fields of
quarterly average PM2.5 concentrations based on Downscaler modeling with the gridded spatial
fields of quarterly PM2.5 RRFs corresponding to the percent emission change required to just
meet the controlling standard. The projected fields of quarterly average PM2.5 concentrations
were then averaged to produce the annual average projected field.
Since PM2.5 concentrations can be projected in multiple ways to just meet a standard,
projections were done for two scenarios that provide results for a range of PM2.5 conditions. The
first scenario is referred to as "Primary PM" or Pri-PM because projections were largely based
on RRFs developed using CMAQ sensitivity simulations with primary PM2.5 emission changes.
For three CBSAs33, standards could not be met using primary PM2.5 emission reductions alone.
PM2.5 concentrations were projected for these areas using a combination of primary PM2.5 and
NOx and SO2 emission reductions in the Primary PM scenario34 (Figure C-18).
32	Note that calculations are performed in terms of percent emission reduction. Therefore, in cases where DVs are
projected to just meet standards greater than the baseline DVs, the required percent emission reduction is negative
(i.e., an emission increase is required), and the smaller absolute percent emission change is selected as the
controlling case. For example, the annual standard would be selected as controlling in a case where a 10%
emission increase is needed to meet the annual standard and a 50% emission increase is needed to meet the 24-hr
standard (because -10 is greater than -50).
33	Bakersfield, Hanford-Corcoran, and Visalia-Porterville (all in California)
34	This approach was applied by using RRFs from the NOx and SO2 emission sensitivity simulations to eliminate a
fraction of the difference between the maximum base DV and the standard level and then using RRFs from the
primary PM2 5 emission sensitivity simulations to eliminate the remainder of the difference. The fraction of the
difference eliminated with NOx and SO2 emission reductions was as follows: 0.4 for Bakersfield, 0.5 for Visalia-
Porterville, and 0.6 for Hanford-Corcoran
C-39

-------
45-
m m
wm. #JI
r-
0 40-
"O
=3
.3 35-
30-
tl
Primary PM

**k
<2.
25 _Mandalp©2019Gooale. INEG^p.,
-120	-110
Gulf of
"f
-100	-90
Longitude
-80
-70
Primary N0xS02+Primary
Figure C-18. Projection method used for each CBSA in the "Primary PM" projection case.
See text for details.
The second scenario is referred to as "Secondary PM" or Sec-PM because projections
were largely based on RRFs developed using CMAQ modeling with NOx and SO2 emission
changes, which affect concentrations of secondary PM components such as ammonium nitrate
and ammonium sulfate. For 22 CBSAs35, standards could not be just met using NOx and SO2
emission changes alone. These areas were projected using the proportional scaling method36
(Figure C-19). The proportional method was selected to gap-fill the Secondary PM case because
35	Altoona, PA; Atlanta-Sandy Springs-Roswell, GA: Bakersfield, CA; Chicago-Naperville-Elgin IL-IN-WI; El
Centra, CA; Elkhart-Goshen, IN; Fresno, CA; Hanford-Corcoran, CA; Las Vegas-Henderson-Paradise, NV: Los
Angeles-Long Beach-Anaheim, CA; Macon GA; Madera, CA; McAllen-Edinburg-Mission TX: Modesto, CA;
Napa, CA; New York-Newark-Jersey City, NY-NJ-PA; Prineville. OR; Riverside-San Bernardino-Ontario, CA;
St. Louis, MO-IL; San Luis Obispo-Paso Robles-Arroyo Grande. CA; Visalia-Porterville, CA; Wheeling, WV-
OH
36	In the proportional method, the spatial field is uniformly scaled by a fixed percentage that corresponds to the
percent difference between the controlling standard level and maximum PM2.5 DV for the controlling standard.
The controlling standard (annual or 24-lir) is identified as the one with the greater percent difference between the
maximum DV and the standard level.
C-40

-------
it is based on a spatially uniform percent change in PM2.5 over the area that is like the
conceptually broad spatial response pattern of PM2.5 to changes in secondary PM2.5 components.
The proportional method has been used previously in the Risk and Exposure Assessment for the
2012 PM NAAQS review (U.S. EPA, 2010).
45
d) 40
T3
135
30
25
-120	-110	-100	-90	-80	-70
Longitude
NOxS02 Proportional
Figure C-19. Projection method used for each CBSA in the "Secondary PM" projection
case.
The baseline 2015 concentration in the 47 CBSAs is shown in Figure C-20. These
concentrations are the same as those in Figure C-17 but are shown only for the CBSAs included
in the projections. In Figure C-2L the difference in annual concentration projected for the 12/35
case and the 2015 baseline concentration is shown. The positive and negative differences reflect
areas where concentrations were projected to higher and lower levels to just meet the standard,
respectively. In Figure C-22, the difference between the annual concentration projected for the
10/30 case and the and 2015 baseline concentration. Negative values indicate that concentrations
fe}§r
M
M
Secondary PM


1d?

J,

%
_Map data £2019 Gooole INEGfo,
Gulf of
l
C-41

-------
were projected to lower levels in all cases for the areas. The difference in projected
concentrations for the 10/30 and 12/35 fields is shown in Figure C-23. Baseline and projected
PM2.5 DVs for monitors in the 47 CBSAs are provided in Table C-33, Table C-34, Table C-35,
and Table C-36 in section C.4.J/
2015 PM2.5
CD
-o 40 H
-4^
03
ug/m3
¦:
Q
Figure C-20. Annual average 2015 PM2.5 concentrations in the 47 CBSAs based on
Downscaler modeling.
-100	-90
Longitude
37 The tables report the percent emission reduction associated with just meeting standards in the current modeling.
These values should not be interpreted as the percent emission reductions that would be required to meet the
standards in other application (e.g., attainment demonstrations for state implementation plans). The modeling
done here was designed to quickly project FM^ fields throughout the U.S. with a broad range of model response
patterns, rather than to apply model configurations and emission scenarios specific to just meeting standards most
efficiently in particular regions.
C-42

-------
Primary PM
Secondary PM
Longitude
Figure C-21. Difference between the annual average projected PM2.5 concentrations and
the 2015 baseline concentrations for the 12/35 projection cases (i.e., 12/35 - baseline).
C-43

-------
Primary PM
Secondary PM
-100	-90
Longitude
Figure C-22. Difference between the annual average projected PM2.5 concentrations and
the 2015 baseline concentrations for the 10/30 projection cases (i.e., 10/30 - baseline).
C-44

-------
Primary PM
Secondary PM
Longitude
Figure C-23. Difference between the annual average projected PM2.5 concentrations in the
10/30 and 12/35 cases (i.e., 10/30 - 12/35) for the Primary PM and Secondary PM
projection cases.
C.l.4.7 Limitations
There are several limitations associated with the air quality projections. First, the baseline
and projected concentrations rely on model predictions. Although state-of-the-science modeling
methods were applied, and model performance was generally good, there is uncertainty
associated with the model predictions. Second, due to the national scale of the assessment, the
C-45

-------
modeling scenarios are based on "across-the-board" emission changes in which emissions of
primary PM2.5 or NOx and SO2 from all anthropogenic sources throughout the U.S. are scaled by
fixed percentages. Although this approach tends to target the key sources in each area, it does not
tailor emission changes to specific periods or sources. More refined emission scenarios could be
beneficial for projections in areas with relatively large seasonal and/or spatial variability in
PM2.5. Similarly, fine scale simulations (e.g., 4 km or less), which are not possible due to the
national scale of the assessment, would be beneficial in areas with complex terrain and relatively
large spatial gradients in PM2.5. A third limitation arises because many emission cases could be
applied to project PM2.5 concentrations to just meet standards. We applied two projection cases
that span a wide range of possible conditions, but these cases are necessarily a subset of the full
set of possible projection cases.
C.1.5 Risk Modeling Approach
Risk modeling for this assessment was completed using the EPA's Environmental
Benefits Mapping and Analysis Program - Community Edition (BenMAP-CE) version
1.4.14.1,38 BenMAP-CE was used to estimate risk at the 12 km grid cell level for grid cells
intersected by the 47 urban study area CBSAs included in risk modeling. BenMAP-CE is an
open-source computer program that calculates the number and economic value of air pollution-
related deaths and illnesses. The software incorporates a database that includes many of the
concentration-response relationships, population files, and health and economic data needed to
quantify these impacts. BenMAP-CE also allows the user to import customized datasets for any
of the inputs used in modeling risk. For this analysis, CR functions developed specifically for
this assessment were imported into BenMAP-CE (section C.l.l). The BenMAP-CE tool
estimates the number of health impacts resulting from changes in air quality—specifically,
ground-level ozone and fine particles. BenMAP-CE can also translate these incidence estimates
into monetized benefits, although that functionality was not employed for this risk assessment.
Inputs to BenMAP-CE used for this risk assessment are identified above in Figure C-l and
described in detail in sections C.l.l, C.1.2, C.1.3, and C.1.4.
An overall flow diagram of the risk assessment approach is provided in Figure C-24.
Application of this approach resulted in separate sets of risk estimates being generated for three
groupings of urban study areas including: (a) the full set of 47, (b) the 30 areas controlled by the
annual standard, and (c) the 11 areas controlled by the 24-hr standard. Risk estimates are
presented and discussed for each of these groupings in PA section 3.3.2, with greater emphasis
being placed on results generated for the full set of 47 urban study areas and 30 annual -
38 BenMAP-CE is a free program which can be downloaded from: https://www.epa.gov/ben.rnap.
C-46

-------
controlled study areas given interest in national representation and on those study areas where we
could also consider the alternative annual standards of 9 and 11 |ig/m3.
Selection of standards modeled in the risk assessment
Identified 47 urban study areas with annual and daily
design values >10 and 30 ug/m3, respectively, that include
—60 million people aged 30+
J
Modeled/simulated air quality surfaces of the 47 urban study areas for:
1.	2015 current conditions (CC)
2.	Current standard combination of annual-12 ug/m3 and daily-35 ug/m3 (12/35)
3.	Alternate standard combination of annual-10 ug/m3 and daily-30 ug/m3 (10/30)
I
Interpolated/extrapolated
additional alternate annual
standards of 11 and 9 ug/m3
Estimated risk in all
47 study areas for CC,
12/35, and 10/30 ug/m3
Estimated risk in 30 annual-
controlled study areas
(~50M people 30+) for CC,
12, 11, 10, and 9 ug/m3
Estimated risk in 11
daily-controlled study
areas (~4M people 30+)
for CC, 35, and 30 ug/m3
Figure C-24, Flow diagram of risk assessment technical approach.
C.2 SUPPLEMENTAL RISK RESULTS
As noted earlier, this appendix presents more granular risk information that supplements
the aggregated risk estimates presented and discussed in section 3.3.2 of the PA. This
supplemental information is intended to provide additional context for the interpretation of
summary risk estimates presented in section 3.3.2 above, and includes:
• Modeled risk estimates that underly summary tables presented in PA section 3.3.2
aggregated to the CBSA-level (i.e., the urban study area) (section C.2.1). Here we begin
by presenting the summary table for the full set of 47 study areas followed by the CBSA-
level data underlying each summary table. We then present the summary table for the 30-
annual-controlled study areas, followed by the CBSA-level data underlying those
summary tables.
C-47

-------
• Additional graphics including line plots, maps and scatter plots illustrating the
distribution of the grid-level risk estimates (section C.2.2). These graphics allow the
reader to consider different aspects of the grid-level data underlying the summary tables
presented in the PA (e.g., spatial distribution of risk across the cities included in the risk
assessment, how the distribution of grid-cell level risk estimates shifts as lower
alternative standards are considered).
Note that at the end of section C.2 we present key observations from consideration both
of the CBSA-level risk estimates presented in section C.2.1 and the graphics illustrating the
distribution of grid-level risk estimates in section C.2.2.
C.2.1 Risk Summary Tables and Underlying CBSA-Level Risk Estimates
This section presents the full results of the risk assessment conducted in support of this
review of the PM NAAQS. This includes aggregate results for all 47 urban study areas across
each of the endpoints modeled, as well as the underlying results for individual cities for each
endpoint. The aggregate results are consistent with those reported above in the summary tables in
Chapter 3 (section 3.3.2). The more refined results for each urban study area presented below
reflect the detailed 12 km grid-level risk estimates aggregated to the CBSA-level (i.e., the urban
study area).
The results are organized as follows: the summary tables for the full set of 47 urban study
areas, followed by tables of the associated CBSA-level risk estimates, are presented in section
C.2.1.1. Then, in section C.2.1.2, we break out the 30 annual-controlled study areas (both in
summary form and by the associated CBSA-level risk estimates) to show the results of
simulating alternative annual standard levels of 11.0 |ig/m3 and 9.0 |ig/m3. We do not report the
results for the 11 daily-controlled areas separately, as readers can find the CBSA-level results for
these areas within the tables presented for the full set of 47 study areas.39 In reviewing the
CBSA-level risk estimates, it is important to consider several details related to these tables
including:
•	In addition to the information on current and alternative standards presented in PA
section 3.3.2, the tables below include information on 2015 current conditions.
•	The CBSA tables are organized by health endpoint (i.e., each table presenting risk
estimates for a specific endpoint). Then within a given CBSA table, the columns
39 The set of 11 daily-controlled study areas is shown in Figure C-5 and includes the following study areas: Fresno,
CA, Logan, UT-ID, Madera, CA, Merced, CA, Modesto, CA, Ogden-Clearfield, UT, Prineville, OR, Provo-Orem,
UT, Sacramento-Ro Seville-Arden-Arcade, CA, Salt Lake City, UT, Visalia-Porterville, CA.
C-48

-------
present risk estimates for specific air quality scenarios (e.g., current conditions,
current standard and so on) with the rows presenting risks for individual CBSAs. To
aid cross-walk comparison between the summary tables and the CBSAs, the order of
the standards presented in the CBS A tables matches the order of standards presented
in the summary tables.
•	Each CBS A table includes a "total" as the last row in the table, which provides the
sum for that air quality scenario/health endpoint combination across all study areas.
This total value can be used as a cross-check with the matching value presented in the
summary table for a particular air quality scenario/health endpoint combination.
•	Given the national-scale of the effect estimates used in modeling mortality risks,
greater confidence is associated with aggregated (cross-city) risk estimates (as
presented in PA section 3.3) than with individual CBSA-level results.
C.2.1.1 CBSA-Level Results for the 47 Urban Study Areas
Here we begin by presenting the summary tables of absolute risk and risk reduction for
the full set of 47 study areas (Table C-10 and Table C-l 1). Then we provide tables of individual
endpoint- and study- specific CBSA-level risk estimates (Table C-12, Table C-13, Table C-14,
Table C-15, Table C-16, Table C-17, Table C-18, Table C-19, and Table C-20).
C-49

-------
Table C-10. Absolute risk summary table of the 47 urban study areas, including current
conditions (2015).
Endpoint
Study
Absolute Risk
Current
Conditions
Simulation
Method*
Current Annual
Standard (12 ug/m3)
Alternative Annual
Standard HO iw/mJl
Alternative 24-hr
Standard GO iw/mJl
Long-term exposure related mortality
IHD Jerrett2016
Pope 2015
15.800
(12.100-19.400)
Pri-PM
16.500 (12.600-20.300)
14.400(11.000-17.700)
16.400(12.500-20.000)
Sec-PM
16.800 (12.800-20.500)
14.200(10.900-17.500)
16.500(12.600-20.200)
14.900
(11.100-18.500)
Pri-PM
15.600(11.600-19.400)
13.600(10.100-17.000)
15.400(11.500-19.200)
Sec-PM
15.800(11.800-19.600)
13.400 (9.970-16.700)
15.600(11.600-19.400)
All-cause Di2017
Pope 2015
Thurston 2015
44.100
(42.900-45.300)
Pri-PM
46.200 (45.000-47.500)
40.300 (39.200-41.400)
45.700 (44.500-47.000)
Sec-PM
46.900 (45.600-48.200)
39.700 (38.600-40.800)
46.200 (44.900-47.500)
49.000
(39.200-58.700)
Pri-PM
51.300(41.000-61.400)
44.700 (35.700-53.500)
50.700 (40.500-60.700)
Sec-PM
52.100(41.600-62.300)
44.000 (35.100-52.700)
51.300 (41.000-61.400)
12.900
(2.250-23.100)
Pri-PM
13.500 (2.360-24.200)
11.700 (2.050-21.100)
13.300 (2.330-24.000)
Sec-PM
13.700 (2.400-24.600)
11.500 (2.010-20.700)
13.500 (2.360-24.200)
Lung Turner 2016
cancer
3.700
(1.180-6.060)
Pri-PM
3.890 (1.240-6.360)
3.390 (1.080-5.560)
3.850 (1.230-6.300)
Sec-PM
3.950 (1.260-6.460)
3.330 (1.060-5.470)
3.890 (1.240-6.370)
Short-term exposure related mortality
All cause Baxter 2017
Ito 2013
Zanobetti 20 U
2.380
(936-3.810)
Pri-PM
2.490 (983-4.000)
2.160 (850-3.460)
2.460 (970-3.950)
Sec-PM
2.530 (998-4.060)
2.120 (837-3.400)
2.490 (982-3.990)
1.120
(-15-2.260)
Pri-PM
1.180 (-16-2.370)
1.020 (-14-2.050)
1.160 (-16-2.340)
Sec-PM
1.200 (-16-2.400)
1.000 (-14-2.020)
1.180 (-16-2.370)
3.630
(2.410-4.840)
Pri-PM
3.810 (2.530-5.080)
3.300 (2.190-4.400)
3.760 (2.500-5.020)
Sec-PM
3.870(2.570-5.160)
3.250 (2.160-4.330)
3.810 (2.530-5.070)
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-50

-------
Table C-ll. Summary of risk reduction in the 47 urban study areas when simulating a
change in air quality from the current standards to an alternative suite of standards.



Risk Reduction (Relative to Current
% Risk Reduction (Relative to


Simulation
Method*
Standard)
Current Standard)
Endpoint
Study
Alternative Annual
Alternative 24-hr
Alternative Annual
Alternative 24-hr


Standard
Standard
Standard
Standard



(12-10 jig/itf)
(35-30 jig/m5)
(12-10 jig/itf)
(35-30 ng/irf)
Long-term exposure related mortality
IHD
Jerrett 2016
Pri-PM
2,390 (1,800-2,970)
200 (150-249)
12.6
1.1


Sec-PM
2,870 (2,160-3,570)
266 (200-331)
15.0
1.4

Pope 2015
Pri-PM
2,240 (1,640-2,830)
187 (137-237)
12.7
1.1


Sec-PM
2,690 (1,970-3,400)
250 (183-315)
15.1
1.4
All-cause
Di2017
Pri-PM
6,440 (6,260-6,630)
573 (557-589)
12.9
1.2


Sec-PM
7,800 (7,580-8,020)
772 (750-793)
15.4
1.5

Pope 2015
Pri-PM
7,100 (5,640-8,550)
644 (511-776)
13.0
1.2


Sec-PM
8,630 (6,860-10,400)
828 (658-997)
15.6
1.5

Thurston 2015
Pri-PM
1,830 (316-3,320)
168 (29-305)
13.2
1.2


Sec-PM
2,230 (387-4,060)
209 (36-381)
15.9
1.5
Lung
Turner 2016
Pri-PM
548 (170-921)
42 (13-70)
13.0
1.0
cancer

Sec-PM
670 (208-1,120)
61 (19-102)
15.6
1.4
Short-term exposure related mortality
All cause
Baxter 2017
Pri-PM
335 (132-537)
30 (12-48)
13.5
1.3


Sec-PM
408 (160-654)
39 (15-62)
16.1
1.6

Ito 2013
Pri-PM
158 (-2-317)
14 (0-29)
13.4
1.2


Sec-PM
192 (-3-386)
18 (0-37)
16.1
1.5

Zanobetti2014
Pri-PM
513 (341-684)
46 (30-61)
13.4
1.2


Sec-PM
622 (413-830)
62 (41-82)
16.0
1.6
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-51

-------
Table C-12. CBSA level results for the 47 urban study areas using the Jerrett et al., 2016 long-term IHD mortality CR
function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current Annual Standard (12
jag/m3)
Alternative An
(10
inual Standard
g/irf)
Alternative 24-hr (30 jug/irf)
Alternative Annual Standard
(12-10 jug/ntf)
Alternative 24-lir Standard
(35-30 ug/nf)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
161
173
177
150
147
173
177
27
35
0
0
Altoona, PA
31
36
36
31
31
36
36
6
6
0
0
Atlanta-Sandy Springs-Roswell, GA
414
462
475
403
401
462
475
68
84
0
0
Bakersfield, CA
137
83
89
69
89
83
78
15
0
0
13
Birmingham-Hoover, AL
163
171
177
150
142
171
177
24
41
0
0
Canton-Massillon, OH
90
99
101
85
84
99
101
17
19
0
0
Chicago-Naperville-Elgin, IL-IN-WI
1,330
1,420
1,430
1,220
1,210
1,420
1,430
226
255
0
0
Cincinnati, OH-KY-IN
332
365
373
315
312
365
373
57
71
0
0
Cleveland-Elyria, OH
436
433
431
379
347
433
431
62
95
0
0
Detroit-Warren-Dearborn, MI
1,030
1,090
1,110
926
892
1,090
1,110
183
242
0
0
El Centre, CA
21
20
20
17
17
20
20
4
4
0
0
Elkhart-Goshen, IN
42
49
49
41
41
49
49
9
9
0
0
Evans ville, IN-KY
61
70
72
60
60
70
72
12
13
0
0
Fresno, CA
182
141
139
141
139
123
127
0
0
21
14
Hanford-Corcoran, CA
22
12
11
10
11
12
10
3
0
0
2
Houston-The Woodlands-Sugar Land, TX
682
723
746
624
600
723
746
114
167
0
0
Indianapolis-C arme 1-Ande rson, IN
282
293
296
254
248
293
296
45
54
0
0
.Johnstown, PA
39
43
44
37
37
43
44
7
9
0
0
Lancaster, PA
109
103
101
87
83
103
101
18
22
0
0
Las Vegas-Henderson-Paradise, NV
163
186
189
159
159
186
189
30
33
0
0
Lebanon, PA
25
27
27
23
23
27
27
5
5
0
0
Little Rock-North Little Rock-Conway, AR
100
116
117
98
98
116
117
21
22
0
0
Logan, UT-ID
6
6
6
6
6
6
6
0
0
1
1
Los Angeles-Long Beach-Anaheim, CA
2,250
2,190
2,190
1,870
1,850
2,190
2,190
365
388
0
0
Louisville/Jefferson County, KY-IN
184
204
208
176
174
204
208
32
40
0
0
Macon, GA
41
48
48
41
41
48
48
8
9
0
0
Madera, CA
36
31
31
31
31
28
28
0
0
3
3
McAllen-Edinburg-Mission, TX
94
110
110
93
93
110
110
19
20
0
0
Merced, CA
44
41
41
41
41
37
37
0
0
5
4
Modesto, CA
117
99
99
99
99
90
90
0
0
11
10
Napa, CA
23
27
27
23
23
27
27
4
5
0
0
New York-Newarik-Jersey City, NY-NJ-PA
3,540
4,020
4,130
3,480
3,480
4,020
4,130
616
730
0
0
Ogden-Clearfield, UT
44
47
46
47
46
42
43
0
0
6
4
Philadelphia-Camden-Wihnington, PA-NJ-DE-MD
1,000
1,040
1,070
898
846
1,040
1,070
167
251
0
0
Pittsburgh, PA
622
587
584
502
584
587
449
96
0
0
151
Prineville, OR
3
3
3
3
3
3
2
0
0
0
0
Provo-Orem, UT
20
22
21
22
21
20
20
0
0
3
2
Riverside-San Bernardino-Ontario, CA
586
498
486
498
415
443
486
0
78
61
0
Sacra me nto-Roseville-Arden-Arcade, CA
327
359
352
359
352
319
321
0
0
46
35
Salt Lake City, UT
65
55
59
55
59
45
55
0
0
10
4
San Luis Obispo-Paso Robles-Arroyo Grande, CA
29
33
33
28
28
33
33
6
6
0
0
South Bend-Mishawaka, IN-MI
59
64
68
64
68
56
55
0
0
10
14
St. Louis, MO-IL
569
656
668
564
565
656
668
106
119
0
0
Stockton-Lodi, CA
118
111
110
111
96
99
110
0
16
14
0
Visalia-Porterville, CA
96
66
65
66
65
57
57
0
0
10
10
Weirton-Steubenville, WV-OH
44
44
45
38
37
44
45
7
9
0
0
Wheeling, WV-OH
48
56
56
47
47
56
56
10
10
0
0
Totals
15,800
16,500
16,800
14,400
14,200
16,400
16,500
2,390
2,870
200
266
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-52

-------
Table C-13. CBSA level results for the 47 urban study areas using the Pope et al., 2015 long-term IHD mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current Annual Standard (12
jag/itf)
Alternative An
(10 it
inual Standard
g/irf)
Alternative 24-hr (30 jug/in3)
Alternative Annual Standard
(12-10 jug/ntf)
Alternative 24-lir Standard
(35-30 jug/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
152
163
167
141
138
163
167
25
33
0
0
Altoona, PA
29
34
34
29
29
34
34
6
6
0
0
Atlanta-Sandy Springs-Roswell, GA
390
435
447
379
378
435
447
64
79
0
0
Bakersfield, CA
129
78
84
65
84
78
73
14
0
0
12
Birmingham-Hoover, AL
154
162
167
142
133
162
167
23
38
0
0
Canton-Massillon, OH
85
93
95
80
79
93
95
16
18
0
0
Chicago-Naperville-Elgin, IL-IN-WI
1,250
1,340
1,350
1,150
1,140
1,340
1,350
213
239
0
0
Cincinnati, OH-KY-IN
313
344
352
297
293
344
352
54
67
0
0
Cleveland-Elyria, OH
411
408
406
357
327
408
406
58
89
0
0
Detroit-Warren-Dearborn, MI
967
1,020
1,040
871
839
1,020
1,040
172
227
0
0
El Centre, CA
20
19
19
16
16
19
19
3
3
0
0
Elkhart-Goshen, IN
40
46
46
39
39
46
46
8
8
0
0
Evans ville, IN-KY
57
66
67
57
57
66
67
11
13
0
0
Fresno, CA
171
133
131
133
131
116
119
0
0
19
13
Hanford-Corcoran, CA
21
12
11
9
11
12
9
2
0
0
2
Houston-The Woodlands-Sugar Land, TX
642
682
703
588
564
682
703
107
157
0
0
Indianapolis-C arme 1-Ande rson, IN
266
276
279
239
234
276
279
42
51
0
0
.Johnstown, PA
37
40
42
35
34
40
42
6
8
0
0
Lancaster, PA
103
97
96
82
78
97
96
16
20
0
0
Las Vegas-Henderson-Paradise, NV
153
175
178
149
150
175
178
28
31
0
0
Lebanon, PA
24
26
26
22
22
26
26
4
5
0
0
Little Rock-North Little Rock-Conway, AR
94
109
110
92
92
109
110
19
20
0
0
Logan, UT-ID
6
6
6
6
6
5
5
0
0
1
0
Los Angeles-Long Beach-Anaheim, CA
2,120
2,070
2,060
1,760
1,740
2,070
2,060
342
364
0
0
Louisville/Jefferson County, KY-IN
174
192
196
165
163
192
196
30
37
0
0
Macon, GA
39
45
46
39
39
45
46
7
8
0
0
Madera, CA
34
29
29
29
29
27
26
0
0
3
3
McAllen-Edinburg-Mission, TX
88
103
104
88
88
103
104
18
18
0
0
Merced, CA
42
39
39
39
39
35
35
0
0
5
4
Modesto, CA
110
93
93
93
93
84
84
0
0
10
10
Napa, CA
22
25
25
21
21
25
25
4
4
0
0
New York-Newarik-Jersey City, NY-NJ-PA
3,330
3,790
3,890
3,280
3,280
3,790
3,890
578
685
0
0
Ogden-Clearfield, UT
42
45
43
45
43
39
40
0
0
6
3
Philadelphia-Camden-Wihnington, PA-NJ-DE-MD
944
984
1,010
845
796
984
1,010
156
236
0
0
Pittsburgh, PA
586
553
550
473
550
553
423
90
0
0
141
Prineville, OR
3
3
3
3
3
2
2
0
0
0
0
Provo-Orem, UT
19
21
20
21
20
19
19
0
0
2
1
Riverside-San Bernardino-Ontario, CA
551
468
457
468
390
416
457
0
74
57
0
Sacra me nto-Roseville-Arden-Arcade, CA
308
338
331
338
331
301
302
0
0
43
33
Salt Lake City, UT
61
51
55
51
55
42
52
0
0
10
3
San Luis Obispo-Paso Robles-Arroyo Grande, CA
28
31
31
26
26
31
31
5
5
0
0
South Bend-Mishawaka, IN-MI
56
60
64
60
64
52
52
0
0
9
14
St. Louis, MO-IL
536
618
629
531
532
618
629
99
112
0
0
Stockton-Lodi, CA
111
104
104
104
91
93
104
0
15
13
0
Visalia-Porterville, CA
91
62
62
62
62
54
53
0
0
9
9
Weirton-Steubenville, WV-OH
41
42
42
36
35
42
42
7
8
0
0
Wheeling, WV-OH
45
52
53
44
44
52
53
9
9
0
0
Totals
14,900
15,600
15,800
13,600
13,400
15,400
15,600
2,240
2,690
187
250
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-53

-------
Table C-14. CBSA level results for the 47 urban study areas using the Di et al., 2017b long-term all-cause mortality CR
function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current Annual Standard (12
jag/m3)
Alternative An
(10 ^
inual Standard
g/irf)
Alternative 24-hr (30 jug/m3)
Alternative Annual Standard
(12-10 jug/m?)
Alternative 24-lir Standard
(35-30 jug/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
547
589
602
507
496
589
602
90
117
0
0
Altoona, PA
104
123
123
103
104
123
123
21
21
0
0
Atlanta-Sandy Springs-Roswell, GA
1,940
2,180
2,240
1,890
1,880
2,180
2,240
314
387
0
0
Bakersfield, CA
333
199
214
166
214
199
186
35
0
0
30
Birmingham-Hoover, AL
709
745
770
649
613
745
770
104
170
0
0
Canton-Massillon, OH
300
329
335
281
278
329
335
53
63
0
0
Chicago-Naperville-Elgin, IL-IN-WI
4,220
4,520
4,570
3,870
3,840
4,520
4,570
698
789
0
0
Cincinnati, OH-KY-IN
1,160
1,280
1,300
1,100
1,080
1,280
1,300
196
240
0
0
Cleveland-Elyria, OH
1,290
1,280
1,280
1,120
1,020
1,280
1,280
178
274
0
0
Detroit-Wairen-D e arboni, MI
2,430
2,570
2,620
2,180
2,100
2,570
2,620
421
562
0
0
El Centre, CA
51
48
48
40
41
48
48
8
8
0
0
Elkhart-Goshen, IN
114
133
133
112
112
133
133
23
23
0
0
Evans ville, IN-KY
207
242
247
206
206
242
247
39
45
0
0
Fresno, CA
506
389
383
389
383
338
348
0
0
56
37
Hanford-Corcoran, CA
64
35
33
28
33
35
28
7
0
0
5
Houston-The Woodlands-Sugar Land, TX
2,130
2,260
2,340
1,940
1,870
2,260
2,340
347
510
0
0
Indianapolis-Carmel-Anderson, IN
950
989
997
852
832
989
997
148
178
0
0
.Johnstown, PA
120
133
136
114
112
133
136
21
26
0
0
Lancaster, PA
397
374
370
317
299
374
370
62
76
0
0
Las Vegas-Henderson-Paradise, NV
543
622
633
529
531
622
633
98
108
0
0
Lebanon, PA
95
102
102
86
86
102
102
17
18
0
0
Little Rock-North Little Rock-Conway, AR
354
411
415
345
346
411
415
71
75
0
0
Logan, UT-ID
26
27
27
27
27
25
25
0
0
3
2
Los Angeles-Long Beach-Anaheim, CA
5,280
5,150
5,140
4,380
4,320
5,150
5,140
832
887
0
0
Louisville/Jefferson County, KY-IN
731
813
829
695
688
813
829
127
152
0
0
Macon, GA
129
149
152
128
128
149
152
23
26
0
0
Madera, CA
88
76
75
76
75
69
68
0
0
7
8
McAllen-Edinburg-Mission, TX
213
251
252
212
212
251
252
42
44
0
0
Merced, CA
115
106
107
106
107
95
97
0
0
13
11
Modesto, CA
268
226
225
226
225
204
204
0
0
24
23
Napa, CA
87
99
100
84
84
99
100
16
17
0
0
New York-Newarik-Jersey City, NY-NJ-PA
7,690
8,770
9,020
7,570
7,580
8,770
9,020
1,290
1,560
0
0
Ogden-Clearfield, UT
178
191
186
191
186
168
173
0
0
24
14
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
3,260
3,400
3,480
2,910
2,740
3,400
3,480
530
798
0
0
Pittsburgh, PA
1,870
1,760
1,750
1,500
1,750
1,760
1,340
281
0
0
441
Prineville, OR
12
11
11
11
11
10
10
0
0
1
2
Provo-Orem, UT
97
107
103
107
103
96
96
0
0
12
7
Riverside-San Bernardino-Ontario, CA
1,510
1,280
1,250
1,280
1,060
1,140
1,250
0
198
153
0
Sacra me nto-Roseville-Arden-Arcade, CA
990
1,090
1,070
1,090
1,070
965
972
0
0
136
103
Salt Lake City, UT
304
256
276
256
276
210
260
0
0
48
17
San Luis Obispo-Paso Robles-Arroyo Grande, CA
108
120
121
101
101
120
121
20
21
0
0
South Bend-Mishawaka, IN-MI
197
213
226
213
226
184
183
0
0
31
47
St. Louis, MO-IL
1,590
1,840
1,870
1,570
1,580
1,840
1,870
287
325
0
0
Stockton-Lodi, CA
357
333
331
333
289
296
331
0
46
40
0
Visalia-Porterville, CA
247
166
166
166
166
144
143
0
0
24
24
Weirton-Steubemille, WV-OH
102
104
104
89
86
104
104
16
20
0
0
Wheeling, WV-OH
124
144
145
122
122
144
145
24
25
0
0
Totals
44,100
46,200
46,900
40,300
39,700
45,700
46,200
6,440
7,800
573
772
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-54

-------
Table C-15. CBSA level results for the 47 urban study areas using the Pope et al., 2015 long-term all-cause mortality CR
function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current Annual Standard (12
jag/m3)
Alternative An
(10 ^
inual Standard
g/irf)
Alternative 24-hr (30 jug/m3)
Alternative Annual Standard
(12-10 jug/m?)
Alternative 24-lir Standard
(35-30 jug/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
595
641
656
551
539
641
656
97
126
0
0
Altoona, PA
107
126
126
106
106
126
126
22
22
0
0
Atlanta-Sandy Springs-Roswell, GA
2,310
2,590
2,660
2,240
2,230
2,590
2,660
371
457
0
0
Bakersfield, CA
404
240
258
200
258
240
224
42
0
0
36
Birmingham-Hoover, AL
831
874
903
761
717
874
903
121
198
0
0
Canton-Massillon, OH
318
349
355
297
294
349
355
56
66
0
0
Chicago-Naperville-Elgin, IL-IN-WI
4,660
4,990
5,040
4,270
4,230
4,990
5,040
767
866
0
0
Cincinnati, OH-KY-IN
1,310
1,440
1,480
1,240
1,220
1,440
1,480
220
270
0
0
Cleveland-Elyria, OH
1,390
1,380
1,370
1,200
1,100
1,380
1,370
191
293
0
0
Detroit-Wairen-D e arboni, MI
2,720
2,880
2,940
2,440
2,350
2,880
2,940
469
625
0
0
El Centre, CA
59
56
56
47
47
56
56
10
10
0
0
Elkhart-Goshen, IN
125
146
146
123
123
146
146
25
25
0
0
Evans ville, IN-KY
229
268
273
228
228
268
273
43
49
0
0
Fresno, CA
573
441
432
441
432
382
393
0
0
62
42
Hanford-Corcoran, CA
78
43
39
35
39
43
34
9
0
0
6
Houston-The Woodlands-Sugar Land, TX
2,590
2,760
2,850
2,360
2,270
2,760
2,850
421
617
0
0
Indianapolis-Carmel-Anderson, IN
1,080
1,130
1,130
968
946
1,130
1,130
168
201
0
0
.Johnstown, PA
126
139
143
119
118
139
143
21
27
0
0
Lancaster, PA
402
378
373
320
301
378
373
62
77
0
0
Las Vegas-Henderson-Paradise, NV
631
723
737
615
617
723
737
113
125
0
0
Lebanon, PA
97
104
105
88
87
104
105
17
19
0
0
Little Rock-North Little Rock-Conway, AR
414
481
486
404
405
481
486
83
87
0
0
Logan, UT-ID
27
28
28
28
28
25
26
0
0
3
2
Los Angeles-Long Beach-Anaheim, CA
5,800
5,660
5,650
4,810
4,740
5,660
5,650
909
969
0
0
Louisville/Jefferson County, KY-IN
841
935
954
799
791
935
954
145
174
0
0
Macon, GA
153
177
180
151
151
177
180
27
31
0
0
Madera, CA
104
88
88
88
88
81
79
0
0
8
9
McAllen-Edinburg-Mission, TX
243
286
288
241
241
286
288
47
49
0
0
Merced, CA
135
124
125
124
125
110
113
0
0
15
13
Modesto, CA
307
258
257
258
257
233
233
0
0
27
26
Napa, CA
89
102
103
87
86
102
103
17
18
0
0
New York-Newarik-Jersey City, NY-NJ-PA
8,230
9,400
9,670
8,100
8,110
9,400
9,670
1,380
1,660
0
0
Ogden-Clearfield, UT
195
209
203
209
203
184
189
0
0
27
16
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
3,570
3,730
3,820
3,190
3,000
3,730
3,820
578
872
0
0
Pittsburgh, PA
1,950
1,830
1,820
1,560
1,820
1,830
1,390
291
0
0
457
Prineville, OR
12
12
11
12
11
11
10
0
0
1
2
Provo-Orem, UT
105
116
112
116
112
104
104
0
0
13
8
Riverside-San Bernardino-Ontario, CA
1,740
1,470
1,430
1,470
1,220
1,300
1,430
0
226
177
0
Sacra me nto-Roseville-Arden-Arcade, CA
1,090
1,210
1,180
1,210
1,180
1,070
1,070
0
0
149
114
Salt Lake City, UT
350
294
317
294
317
241
298
0
0
55
19
San Luis Obispo-Paso Robles-Arroyo Grande, CA
112
125
125
105
105
125
125
21
21
0
0
South Bend-Mishawaka, IN-MI
214
231
246
231
246
200
198
0
0
34
50
St. Louis, MO-IL
1,750
2,030
2,070
1,740
1,740
2,030
2,070
314
356
0
0
Stockton-Lodi, CA
413
385
382
385
333
342
382
0
52
46
0
Visalia-Porterville, CA
289
193
193
193
193
167
166
0
0
28
28
Weirton-Steubemille, WV-OH
112
114
115
98
94
114
115
17
22
0
0
Wheeling, WV-OH
129
150
151
127
127
150
151
25
26
0
0
Totals
49,000
51,300
52,100
44,700
44,000
50,700
51,300
7,100
8,630
644
828
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-55

-------
Table C-16. CBSA level results for the 47 urban study areas using the Thurston et al., 2016 long-term all-cause mortality CR
function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current Annual Standard (12
jag/m3)
Alternative An
(10 ^
inual Standard
g/irf)
Alternative 24-hr (30 jug/m3)
Alternative Annual Standard
(12-10 jug/m?)
Alternative 24-lir Standard
(35-30 jug/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
157
169
173
145
142
169
173
25
33
0
0
Altoona, PA
27
32
33
27
27
32
33
5
5
0
0
Atlanta-Sandy Springs-Roswell, GA
644
725
746
626
624
725
746
102
126
0
0
Bakersfield, CA
114
67
72
56
72
67
63
11
0
0
10
Birmingham-Hoover, AL
231
243
252
211
199
243
252
33
55
0
0
Canton-Massillon, OH
84
92
94
78
77
92
94
14
17
0
0
Chicago-Naperville-Elgin, IL-IN-WI
1,220
1,310
1,320
1,120
1,110
1,310
1,320
197
223
0
0
Cincinnati, OH-KY-IN
353
390
400
334
330
390
400
58
72
0
0
Cleveland-Elyria, OH
359
357
355
310
282
357
355
48
75
0
0
Detroit-Wairen-D e arboni, MI
717
761
776
643
618
761
776
121
162
0
0
El Centre, CA
16
16
16
13
13
16
16
3
3
0
0
Elkhart-Goshen, IN
33
39
39
33
33
39
39
6
7
0
0
Evans ville, IN-KY
62
72
74
61
61
72
74
11
13
0
0
Fresno, CA
150
114
112
114
112
99
102
0
0
16
11
Hanford-Corcoran, CA
22
12
11
9
11
12
9
2
0
0
2
Houston-The Woodlands-Sugar Land, TX
729
776
803
664
636
776
803
116
171
0
0
Indianapolis-Carmel-Anderson, IN
293
305
308
262
256
305
308
45
54
0
0
.Johnstown, PA
31
34
35
29
29
34
35
5
7
0
0
Lancaster, PA
97
91
90
77
72
91
90
15
18
0
0
Las Vegas-Henderson-Paradise, NV
186
214
218
181
182
214
218
33
37
0
0
Lebanon, PA
25
26
26
22
22
26
26
4
5
0
0
Little Rock-North Little Rock-Conway, AR
116
135
137
113
113
135
137
23
24
0
0
Logan, UT-ID
7
7
7
7
7
6
6
0
0
1
1
Los Angeles-Long Beach-Anaheim, CA
1,470
1,430
1,430
1,210
1,190
1,430
1,430
225
240
0
0
Louisville/Jefferson County, KY-IN
231
258
263
220
217
258
263
39
47
0
0
Macon, GA
43
51
52
43
43
51
52
8
9
0
0
Madera, CA
28
24
24
24
24
22
22
0
0
2
2
McAllen-Edinburg-Mission, TX
66
78
79
66
66
78
79
13
13
0
0
Merced, CA
36
33
33
33
33
29
30
0
0
4
3
Modesto, CA
84
70
70
70
70
63
63
0
0
7
7
Napa, CA
22
25
26
21
21
25
26
4
4
0
0
New York-Newarik-Jersey City, NY-NJ-PA
2,070
2,370
2,440
2,030
2,040
2,370
2,440
343
410
0
0
Ogden-Clearfield, UT
50
54
52
54
52
47
48
0
0
7
4
Philadelphia-Camden-Wihnington, PA-NJ-DE-MD
912
953
977
812
763
953
977
145
219
0
0
Pittsburgh, PA
490
461
458
391
458
461
348
72
0
0
113
Prineville, OR
4
3
3
3
3
3
3
0
0
0
0
Provo-Orem, UT
26
29
28
29
28
26
26
0
0
3
2
Riverside-San Bernardino-Ontario, CA
480
404
395
404
335
357
395
0
61
48
0
Sacra me nto-Roseville-Arden-Arcade, CA
288
318
311
318
311
281
282
0
0
38
30
Salt Lake City, UT
89
75
80
75
80
61
76
0
0
14
5
San Luis Obispo-Paso Robles-Arroyo Grande, CA
27
30
30
25
25
30
30
5
5
0
0
South Bend-Mishawaka, IN-MI
55
60
64
60
64
52
51
0
0
9
13
St. Louis, MO-IL
463
539
550
460
460
539
550
82
93
0
0
Stockton-Lodi, CA
111
103
102
103
89
91
102
0
14
12
0
Visalia-Porterville, CA
77
51
51
51
51
44
44
0
0
7
7
Weirton-Steubemille, WV-OH
31
32
32
27
26
32
32
5
6
0
0
Wheeling, WV-OH
34
40
40
34
34
40
40
7
7
0
0
Totals
12,900
13,500
13,700
11,700
11,500
13,300
13,500
1,830
2,230
168
209
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-56

-------
Table C-17. CBSA level results for the 47 urban study areas using the Turner et al., 2016 long-term lung cancer mortality CR
function.

Absolute Risk
Risk Reduction (Relative to Current Standard)
CBSA
Current
Conditions
Current Annual Standard (12
jag/m3)
Alternative An
(10 ^
inual Standard
g/irf)
Alternative 24-hr (30 jug/m3)
Alternative Annual Standard
(12-10 jug/m?)
Alternative 24-lir Standard
(35-30 jug/m3)

(2015)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
48
51
52
44
43
51
52
8
10
0
0
Altoona, PA
6
7
7
6
6
7
7
1
1
0
0
Atlanta-Sandy Springs-Roswell, GA
183
204
210
178
177
204
210
29
36
0
0
Bakersfield, CA
27
16
17
13
17
16
15
3
0
0
2
Birmingham-Hoover, AL
63
66
69
58
55
66
69
9
15
0
0
Canton-Massillon, OH
25
28
28
24
24
28
28
5
5
0
0
Chicago-Naperville-Elgin, IL-IN-WI
379
406
410
348
345
406
410
63
71
0
0
Cincinnati, OH-KY-IN
122
134
137
115
114
134
137
20
26
0
0
Cleveland-Elyria, OH
111
111
110
96
88
111
110
15
24
0
0
Detroit-Wairen-D e arboni, MI
220
233
237
198
190
233
237
38
51
0
0
El Centre, CA
4
4
4
3
3
4
4
1
1
0
0
Elkhart-Goshen, IN
10
11
11
9
9
11
11
2
2
0
0
Evans ville, IN-KY
19
22
23
19
19
22
23
4
4
0
0
Fresno, CA
35
27
26
27
26
23
24
0
0
4
3
Hanford-Corcoran, CA
5
3
2
2
2
3
2
1
0
0
0
Houston-The Woodlands-Sugar Land, TX
194
206
213
177
170
206
213
31
47
0
0
Indianapolis-Carmel-Anderson, IN
102
106
107
91
89
106
107
16
19
0
0
.Johnstown, PA
8
9
9
8
8
9
9
1
2
0
0
Lancaster, PA
28
26
26
22
21
26
26
4
5
0
0
Las Vegas-Henderson-Paradise, NV
55
63
64
53
53
63
64
10
11
0
0
Lebanon, PA
9
9
9
8
8
9
9
2
2
0
0
Little Rock-North Little Rock-Conway, AR
37
43
43
36
36
43
43
7
8
0
0
Logan, UT-ID
1
1
1
1
1
1
1
0
0
0
0
Los Angeles-Long Beach-Anaheim, CA
360
351
351
299
295
351
351
57
61
0
0
Louisville/Jefferson County, KY-IN
82
91
93
78
78
91
93
14
17
0
0
Macon, GA
13
15
15
13
13
15
15
2
3
0
0
Madera, CA
7
6
6
6
6
5
5
0
0
1
1
McAllen-Edinburg-Mission, TX
11
13
13
11
11
13
13
2
2
0
0
Merced, CA
9
9
9
9
9
8
8
0
0
1
1
Modesto, CA
21
18
17
18
17
16
16
0
0
2
2
Napa, CA
7
8
8
6
6
8
8
1
1
0
0
New York-Newarik-Jersey City, NY-NJ-PA
590
672
691
580
581
672
691
99
119
0
0
Ogden-Clearfield, UT
8
8
8
8
8
7
7
0
0
1
1
Philadelphia-Camden-Wihnington, PA-NJ-DE-MD
284
296
303
253
238
296
303
46
70
0
0
Pittsburgh, PA
153
145
144
123
144
145
110
23
0
0
36
Prineville, OR
1
1
1
1
1
1
1
0
0
0
0
Provo-Orem, UT
3
3
3
3
3
3
3
0
0
0
0
Riverside-San Bernardino-Ontario, CA
120
102
99
102
85
90
99
0
16
12
0
Sacra me nto-Roseville-Arden-Arcade, CA
79
87
86
87
86
77
78
0
0
11
8
Salt Lake City, UT
14
12
13
12
13
10
12
0
0
2
1
San Luis Obispo-Paso Robles-Arroyo Grande, CA
8
9
9
7
7
9
9
1
2
0
0
South Bend-Mishawaka, IN-MI
17
18
20
18
20
16
16
0
0
3
4
St. Louis, MO-IL
158
182
186
156
157
182
186
28
32
0
0
Stockton-Lodi, CA
29
27
27
27
23
24
27
0
4
3
0
Visalia-Porterville, CA
18
12
12
12
12
11
10
0
0
2
2
Weirton-Steubemille, WV-OH
9
10
10
8
8
10
10
1
2
0
0
Wheeling, WV-OH
11
12
12
10
10
12
12
2
2
0
0
Totals
3,700
3,890
3,950
3,390
3,330
3,850
3,890
548
670
42
61
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-57

-------
Table C-18. CBSA level results for the 47 urban study areas using the Baxter et al., 2017 all-cause short-term mortality CR
function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current Annual Standard (12
jug/m3)
Alternative An
(10 ^
inual Standard
g/irf)
Alternative 24-hr (30 jug/m3)
Alternative Annual Standard
(12-10 jug/m?)
Alternative 24-lir Standard
(35-30 jug/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
29
31
32
27
26
31
32
5
6
0
0
Altoona, PA
5
6
6
5
5
6
6
1
1
0
0
Atlanta-Sandy Springs-Roswell, GA
111
126
129
108
108
126
129
18
22
0
0
Bakersfield, CA
20
11
12
9
12
11
11
2
0
0
2
Birmingham-Hoover, AL
40
42
44
37
35
42
44
6
9
0
0
Canton-Massillon, OH
15
17
17
14
14
17
17
3
3
0
0
Chicago-Naperville-Elgin, IL-IN-WI
228
245
248
208
206
245
248
37
41
0
0
Cincinnati, OH-KY-IN
63
70
71
59
59
70
71
10
13
0
0
Cleveland-Elyria, OH
68
67
67
58
53
67
67
9
14
0
0
Detroit-Wairen-D e arboni, MI
132
140
143
118
113
140
143
22
30
0
0
El Centre, CA
3
3
3
2
2
3
3
0
0
0
0
Elkhart-Goshen, IN
6
7
7
6
6
7
7
1
1
0
0
Evans ville, IN-KY
11
13
13
11
11
13
13
2
2
0
0
Fresno, CA
28
22
21
22
21
19
19
0
0
3
2
Hanford-Corcoran, CA
4
2
2
2
2
2
2
0
0
0
0
Houston-The Woodlands-Sugar Land, TX
126
134
139
114
109
134
139
20
29
0
0
Indianapolis-Carmel-Anderson, IN
52
54
55
47
46
54
55
8
9
0
0
.Johnstown, PA
6
7
7
6
6
7
7
1
1
0
0
Lancaster, PA
20
18
18
16
15
18
18
3
4
0
0
Las Vegas-Henderson-Paradise, NV
30
34
35
29
29
34
35
5
6
0
0
Lebanon, PA
5
5
5
4
4
5
5
1
1
0
0
Little Rock-North Little Rock-Conway, AR
20
23
24
20
20
23
24
4
4
0
0
Logan, UT-ID
1
1
1
1
1
1
1
0
0
0
0
Los Angeles-Long Beach-Anaheim, CA
284
277
277
234
231
277
277
43
46
0
0
Louisville/Jefferson County, KY-IN
41
45
46
38
38
45
46
7
8
0
0
Macon, GA
7
9
9
7
7
9
9
1
1
0
0
Madera, CA
5
4
4
4
4
4
4
0
0
0
0
McAllen-Edinburg-Mission, TX
12
14
14
12
12
14
14
2
2
0
0
Merced, CA
6
6
6
6
6
5
5
0
0
1
1
Modesto, CA
15
13
13
13
13
11
11
0
0
1
1
Napa, CA
4
5
5
4
4
5
5
1
1
0
0
New York-Newarik-Jersey City, NY-NJ-PA
401
459
473
394
394
459
473
66
79
0
0
Ogden-Clearfield, UT
9
10
10
10
10
9
9
0
0
1
1
Philadelphia-Camden-Wihnington, PA-NJ-DE-MD
172
180
184
153
144
180
184
27
41
0
0
Pittsburgh, PA
94
88
88
74
88
88
66
14
0
0
21
Prineville, OR
1
1
1
1
1
0
0
0
0
0
0
Provo-Orem, UT
5
6
5
6
5
5
5
0
0
1
0
Riverside-San Bernardino-Ontario, CA
85
71
69
71
59
63
69
0
11
8
0
Sacra me nto-Roseville-Arden-Arcade, CA
52
58
57
58
57
51
51
0
0
7
5
Salt Lake City, UT
16
14
15
14
15
11
14
0
0
3
1
San Luis Obispo-Paso Robles-Arroyo Grande, CA
5
6
6
5
5
6
6
1
1
0
0
South Bend-Mishawaka, IN-MI
10
11
12
11
12
10
10
0
0
2
2
St. Louis, MO-IL
84
98
100
83
83
98
100
15
17
0
0
Stockton-Lodi, CA
20
19
19
19
16
17
19
0
2
2
0
Visalia-Porterville, CA
14
9
9
9
9
8
8
0
0
1
1
Weirton-Steubemille, WV-OH
5
5
6
5
4
5
6
1
1
0
0
Wheeling, WV-OH
6
7
7
6
6
7
7
1
1
0
0
Totals
2,380
2,490
2,530
2,160
2,120
2,460
2,490
335
408
30
39
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-58

-------
Table C-19. CBSA level results for the 47 urban study areas using the Ito et al., 2013 all-cause short-term mortality CR
function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current Annual Standard (12
jug/m3)
Alternative An
(10 ^
inual Standard
g/irf)
Alternative 24-hr (30 jug/m3)
Alternative Annual Standard
(12-10 jug/m?)
Alternative 24-lir Standard
(35-30 jug/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
14
15
15
13
12
15
15
2
3
0
0
Altoona, PA
2
3
3
2
2
3
3
0
0
0
0
Atlanta-Sandy Springs-Roswell, GA
53
60
62
52
52
60
62
8
10
0
0
Bakersfield, CA
10
6
6
5
6
6
5
1
0
0
1
Birmingham-Hoover, AL
19
20
21
18
17
20
21
3
4
0
0
Canton-Massillon, OH
7
8
8
7
7
8
8
1
1
0
0
Chicago-Naperville-Elgin, IL-IN-WI
107
115
116
98
97
115
116
17
19
0
0
Cincinnati, OH-KY-IN
30
33
34
28
28
33
34
5
6
0
0
Cleveland-Elyria, OH
32
31
31
27
25
31
31
4
7
0
0
Detroit-Wairen-D e arboni, MI
62
66
68
56
54
66
68
10
14
0
0
El Centre, CA
1
1
1
1
1
1
1
0
0
0
0
Elkhart-Goshen, IN
3
3
3
3
3
3
3
1
1
0
0
Evans ville, IN-KY
5
6
6
5
5
6
6
1
1
0
0
Fresno, CA
14
10
10
10
10
9
9
0
0
1
1
Hanford-Corcoran, CA
2
1
1
1
1
1
1
0
0
0
0
Houston-The Woodlands-Sugar Land, TX
61
65
67
55
53
65
67
10
14
0
0
Indianapolis-Carmel-Anderson, IN
25
26
26
22
22
26
26
4
5
0
0
.Johnstown, PA
3
3
3
3
3
3
3
0
1
0
0
Lancaster, PA
9
9
9
7
7
9
9
1
2
0
0
Las Vegas-Henderson-Paradise, NV
14
16
17
14
14
16
17
3
3
0
0
Lebanon, PA
2
2
2
2
2
2
2
0
0
0
0
Little Rock-North Little Rock-Conway, AR
10
11
11
9
9
11
11
2
2
0
0
Logan, UT-ID
1
1
1
1
1
1
1
0
0
0
0
Los Angeles-Long Beach-Anaheim, CA
133
130
129
109
108
130
129
20
22
0
0
Louisville/Jefferson County, KY-IN
19
22
22
18
18
22
22
3
4
0
0
Macon, GA
4
4
4
3
3
4
4
1
1
0
0
Madera, CA
2
2
2
2
2
2
2
0
0
0
0
McAllen-Edinburg-Mission, TX
6
7
7
6
6
7
7
1
1
0
0
Merced, CA
3
3
3
3
3
3
3
0
0
0
0
Modesto, CA
7
6
6
6
6
5
5
0
0
1
1
Napa, CA
2
2
2
2
2
2
2
0
0
0
0
New York-Newarik-Jersey City, NY-NJ-PA
187
214
220
184
184
214
220
31
37
0
0
Ogden-Clearfield, UT
5
5
5
5
5
4
4
0
0
1
0
Philadelphia-Camden-Wihnington, PA-NJ-DE-MD
82
86
88
73
68
86
88
13
19
0
0
Pittsburgh, PA
44
42
41
35
41
42
31
6
0
0
10
Prineville, OR
0
0
0
0
0
0
0
0
0
0
0
Provo-Orem, UT
2
3
3
3
3
2
2
0
0
0
0
Riverside-San Bernardino-Ontario, CA
40
34
33
34
28
30
33
0
5
4
0
Sacra me nto-Roseville-Arden-Arcade, CA
25
28
27
28
27
24
25
0
0
3
3
Salt Lake City, UT
8
7
7
7
7
6
7
0
0
1
0
San Luis Obispo-Paso Robles-Arroyo Grande, CA
3
3
3
2
2
3
3
0
0
0
0
South Bend-Mishawaka, IN-MI
5
5
6
5
6
5
5
0
0
1
1
St. Louis, MO-IL
40
47
48
40
40
47
48
7
8
0
0
Stockton-Lodi, CA
10
9
9
9
8
8
9
0
1
1
0
Visalia-Porterville, CA
7
4
4
4
4
4
4
0
0
1
1
Weirton-Steubemille, WV-OH
3
3
3
2
2
3
3
0
0
0
0
Wheeling, WV-OH
3
3
3
3
3
3
3
1
1
0
0
Totals
1,120
1,180
1,200
1,020
1,000
1,160
1,180
158
192
14
18
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-59

-------
Table C-20. CBSA level results for the 47 urban study areas using the Zanobetti et al., 2014 all-cause short-term mortality CR
function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current Annual Standard (12
jug/m3)
Alternative An
(10 ^
inual Standard
g/irf)
Alternative 24-hr (30 jug/m3)
Alternative Annual Standard
(12-10 jug/m?)
Alternative 24-lir Standard
(35-30 jug/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
45
49
50
42
41
49
50
7
9
0
0
Altoona, PA
9
10
10
9
9
10
10
2
2
0
0
Atlanta-Sandy Springs-Roswell, GA
159
180
185
155
154
180
185
25
31
0
0
Bakersfield, CA
28
16
17
13
17
16
15
3
0
0
2
Birmingham-Hoover, AL
58
62
64
53
50
62
64
8
14
0
0
Canton-Massillon, OH
25
27
28
23
23
27
28
4
5
0
0
Chicago-Naperville-Elgin, IL-IN-WI
348
373
377
318
315
373
377
56
63
0
0
Cincinnati, OH-KY-IN
95
105
108
90
89
105
108
16
19
0
0
Cleveland-Elyria, OH
106
106
105
92
83
106
105
14
22
0
0
Detroit-Wairen-D e arboni, MI
200
212
216
179
172
212
216
34
45
0
0
El Centre, CA
4
4
4
3
3
4
4
1
1
0
0
Elkhart-Goshen, IN
9
11
11
9
9
11
11
2
2
0
0
Evans ville, IN-KY
17
20
21
17
17
20
21
3
4
0
0
Fresno, CA
42
32
32
32
32
28
29
0
0
4
3
Hanford-Corcoran, CA
5
3
3
2
3
3
2
1
0
0
0
Houston-The Woodlands-Sugar Land, TX
175
187
193
160
153
187
193
28
41
0
0
Indianapolis-Carmel-Anderson, IN
78
82
82
70
68
82
82
12
14
0
0
.Johnstown, PA
10
11
11
9
9
11
11
2
2
0
0
Lancaster, PA
33
31
31
26
24
31
31
5
6
0
0
Las Vegas-Henderson-Paradise, NV
44
51
52
43
43
51
52
8
9
0
0
Lebanon, PA
8
8
8
7
7
8
8
1
1
0
0
Little Rock-North Little Rock-Conway, AR
29
34
34
28
28
34
34
6
6
0
0
Logan, UT-ID
2
2
2
2
2
2
2
0
0
0
0
Los Angeles-Long Beach-Anaheim, CA
435
425
424
359
354
425
424
66
71
0
0
Louisville/Jefferson County, KY-IN
60
67
69
57
57
67
69
10
12
0
0
Macon, GA
11
12
13
11
11
12
13
2
2
0
0
Madera, CA
7
6
6
6
6
6
6
0
0
1
1
McAllen-Edinburg-Mission, TX
17
21
21
17
17
21
21
3
3
0
0
Merced, CA
10
9
9
9
9
8
8
0
0
1
1
Modesto, CA
22
19
19
19
19
17
17
0
0
2
2
Napa, CA
7
8
8
7
7
8
8
1
1
0
0
New York-Newarik-Jersey City, NY-NJ-PA
630
722
743
619
620
722
743
103
124
0
0
Ogden-Clearfield, UT
15
16
15
16
15
14
14
0
0
2
1
Philadelphia-Camden-Wihnington, PA-NJ-DE-MD
268
280
287
238
224
280
287
42
64
0
0
Pittsburgh, PA
154
145
144
123
144
145
109
22
0
0
35
Prineville, OR
1
1
1
1
1
1
1
0
0
0
0
Provo-Orem, UT
8
9
8
9
8
8
8
0
0
1
1
Riverside-San Bernardino-Ontario, CA
124
104
102
104
86
92
102
0
16
12
0
Sacra me nto-Roseville-Arden-Arcade, CA
81
90
88
90
88
79
80
0
0
11
8
Salt Lake City, UT
25
21
22
21
22
17
21
0
0
4
1
San Luis Obispo-Paso Robles-Arroyo Grande, CA
9
10
10
8
8
10
10
2
2
0
0
South Bend-Mishawaka, IN-MI
16
18
19
18
19
15
15
0
0
2
4
St. Louis, MO-IL
131
152
155
129
130
152
155
23
26
0
0
Stockton-Lodi, CA
30
28
27
28
24
24
27
0
4
3
0
Visalia-Porterville, CA
21
14
14
14
14
12
12
0
0
2
2
Weirton-Steubemille, WV-OH
8
9
9
7
7
9
9
1
2
0
0
Wheeling, WV-OH
10
12
12
10
10
12
12
2
2
0
0
Totals
3,630
3,810
3,870
3,300
3,250
3,760
3,810
513
622
46
62
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-60

-------
C.2.1.2 CBSA-Level Results for the 30 Annual-Controlled Urban Study Areas
Here we begin by presenting the summary tables of absolute risk and risk reduction for
the 30 annual-controlled study areas (Table C-21 and Table C-22) where the annual standard was
controlling. Then we provide tables of individual endpoint- and study- specific CBSA-level risk
estimates (Table C-23, Table C-24, Table C-25, Table C-26, Table C-27, Table C-28, Table C-
29, Table C-30, and Table C-31).
Table C-21. Absolute risk summary table of the 30 urban study areas, including current
conditions (2015).
Endpoint
Study
Absolute Risk
Current
Conditions (2015)
Simulation
Method*
Current Annual
Standard (12 u2/m3)
Alternative Annual
Standard (11 u2/m3)
Alternative Annual
Standard (10 u2/m3)
Alternative Annual
Standard (9 u2/m3)
Long-term exposure related mortality
IHD Jerrett2016
Pope 2015
13.300
(10.200-16.300)
Pri-PM
14.300(10.900-17.500)
13.300(10.200-16.300)
12.300(9.400-15.100)
11.300(8.610-13.900)
Sec-PM
14.600(11.100-17.800)
13.300(10.200-16.400)
12.100(9.240-14.900)
10.900(8.280-13.400)
12.500
(9.340-15.600)
Pri-PM
13.500(10.100-16.800)
12.500 (9.340-15.600)
11.600(8.620-14.500)
10.600(7.900-13.300)
Sec-PM
13.700(10.200-17.000)
12.600 (9.360-15.600)
11.400(8.480-14.200)
10.200(7.590-12.800)
All- Di2017
cause
Pope 2015
Thurston 2015
37.000
(36.000-38.000)
Pri-PM
39.800 (38.700-40.900)
36.900 (35.900-38.000)
34.100 (33.200-35.000)
31.200(30.400-32.100)
Sec-PM
40.500 (39.400-41.600)
37.000 (36.000-38.000)
33.500 (32.600-34.400)
29.900 (29.100-30.800)
41.000
(32.800-49.100)
Pri-PM
44.200 (35.300-52.800)
41.000(32.800-49.100)
37.800 (30.200-45.300)
34.600 (27.600-41.500)
Sec-PM
45.000 (35.900-53.800)
41.000(32.800-49.100)
37.100 (29.600-44.500)
33.200 (26.500-39.700)
10.700
(1.880-19.300)
Pri-PM
11.600(2.030-20.800)
10.700(1.880-19.300)
9.900(1.730-17.800)
9.050(1.580-16.300)
Sec-PM
11.800(2.070-21.200)
10.800(1.880-19.400)
9.710(1.700-17.500)
8.650 (1.510-15.600)
Lung Turner 2016
cancer
3.150
(1.000-5.160)
Pri-PM
3.400 (1.080-5.550)
3.160(1.010-5.170)
2.920 (927-4.790)
2.670 (847-4.400)
Sec-PM
3.460(1.110-5.650)
3.160(1.010-5.180)
2.860 (908-4.700)
2.560 (809-4.210)
Short-term exposure related mortality
All- Baxter 2017
cause
Ito 2013
Zanobetti 2014
1.990
(784-3.190)
Pri-PM
2.150 (846-3.440)
1.990(784-3.190)
1.830 (721-2.930)
1.670 (658-2.680)
Sec-PM
2.190(862-3.510)
1.990(785-3.190)
1.790 (707-2.880)
1.600 (630-2.560)
940
(-13-1.890)
Pri-PM
1.010 (-14-2.040)
939 (-13-1.880)
864 (-12-1.730)
789 (-11-1.580)
Sec-PM
1.030 (-14-2.070)
940 (-13-1.890)
847 (-11-1.700)
754 (-10-1.510)
3.040
(2.020-4.050)
Pri-PM
3.280(2.180-4.370)
3.040 (2.020-4.050)
2.790(1.860-3.730)
2.550(1.700-3.400)
Sec-PM
3.340 (2.220-4.450)
3.040 (2.020-4.050)
2.740(1.820-3.650)
2.440(1.620-3.260)
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM -based modeling approach)
C-61

-------
Table C-22. Summary of risk reduction in the 30 urban study areas when simulating a
change in air quality from the current standards to alternative annual standards.
Endpoint
Study
Simulation
Method*
Risk Reduction
(Relative to Current Standard)
Percent Risk Reduction
(Relative to Current Standard)
Alternative Annual
Standard
(12-11 ug/m5)
Alternative Annual
Standard
(12-10 ug/m3)
Alternative Annual
Standard
(12-9 ug/m3)
Alternative
Annual Standard
(12-11 ug/m3)
Alternative
Annual Standard
(12-10 ug/m3)
Alternative Annual
Standard
(12-9 ug/m3)
Long-term exposure related mortality
IHD Jerrett 2016 Pri-PM
Sec-PM
Pope 2014 Pri-PM
Sec-PM
1,140 (859-1,420)
2,270 (1,710-2,830)
3,390 (2,550-4,210)
7%
14%
21%
1,400(1,050-1,740)
2,770 (2,090-3,450)
4,130(3,110-5,130)
8%
17%
25%
1,070 (785-1,360)
2,130 (1,560-2,690)
3,180(2,340-4,010)
7%
14%
21%
1,310 (960-1,660)
2,600 (1,910-3,280)
3,880 (2,850-4,890)
8%
17%
25%
All- Di2017 Pri-PM
cause Sec-PM
Pope 2014 Pri-PM
Sec-PM
Thurston 2015 Pri-PM
Sec-PM
3,070 (2,980-3,160)
6,120 (5,950-6,300)
9,150 (8,890-9,410)
7%
14%
21%
3,800 (3,690-3,900)
7,560 (7,340-7,770)
11,300(11,000-11,600)
9%
17%
26%
3,390 (2,690-4,080)
6,760 (5,370-8,140)
10,100 (8,030-12,200)
7%
14%
22%
4,190 (3,330-5,050)
8,350 (6,640-10,100)
12,500 (9,930-15,000)
9%
17%
26%
871 (151-1,590)
1,740 (301-3,170)
2,610 (452-4,740)
7%
15%
22%
1,080 (187-1,970)
2,160 (374-3,930)
3,230 (561-5,870)
9%
18%
27%
Lung Turner 2016 Pri-PM
cancer Sec-PM
262 (81-441)
522 (162-877)
780(243-1,310)
7%
14%
21%
327 (101-550)
651 (202-1,090)
972(303-1,630)
9%
17%
26%
Short-term exposure related mortality
All- Baxter 2017 Pri-PM
cause Sec-PM
Ito 2013 Pri-PM
Sec-PM
Zanobetti 2014 Pri-PM
Sec-PM
160(63-256)
319(126-512)
478(188-767)
7%
15%
22%
197(78-316)
394 (155-632)
592 (233-948)
9%
18%
27%
75 (-1-151)
150 (-2-302)
226 (-3-453)
7%
15%
22%
93 (-1-187)
186 (-2-374)
279 (-4-561)
9%
18%
27%
244 (162-325)
487 (324-650)
731 (486-975)
7%
15%
22%
301 (200-402)
603 (400-804)
904(600-1,210)
9%
18%
27%
* Pri-PM (primary PM -based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-62

-------
Table C-23. CBSA level results for the 30 annual-controlled urban study areas using the Jerrett et al., 2016 long-term IHD
mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12 fA
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
161
173
177
162
162
150
147
138
131
14
18
27
35
40
53
Altoona, PA
31
36
36
33
34
31
31
28
28
3
3
6
6
10
10
Atlanta-Sandy Springs-Roswell, GA
414
462
475
433
438
403
401
372
364
34
42
68
84
102
126
Birmingham-Hoover, AL
163
171
177
161
160
150
142
140
123
12
21
24
41
36
60
Canton-Massillon, OH
90
99
101
92
92
85
84
78
76
8
10
17
19
25
29
Chicago-Naperville-Elgin, IL-IN-WI
1,330
1,420
1,430
1,320
1,320
1,220
1,210
1,120
1,100
114
128
226
255
338
380
Cincinnati, OH-KY-IN
332
365
373
341
343
315
312
290
280
29
36
57
71
86
106
Cleveland-EIyria, OH
436
433
431
406
389
379
347
351
304
31
48
62
95
92
142
Detroit-Warren-Dearborn, MI
1,030
1,090
1,110
1,010
1,000
926
892
844
783
92
122
183
242
273
360
El Centra, CA
21
20
20
19
19
17
17
15
15
2
2
4
4
5
5
Elkhart-Goshen, IN
42
49
49
45
45
41
41
38
38
4
4
9
9
13
13
Evansville, IN-KY
61
70
72
65
66
60
60
55
54
6
7
12
13
18
20
Houston-The Woodlands-Sugar Land, TX
682
723
746
674
673
624
600
574
525
58
84
114
167
170
249
Indianapolis-Carmel-Anderson, IN
282
293
296
274
272
254
248
234
224
23
27
45
54
67
81
Johnstown, PA
39
43
44
40
40
37
37
34
33
3
4
7
9
10
13
Lancaster, PA
109
103
101
95
92
87
83
80
73
9
11
18
22
26
32
Las Vegas-Henderson-Paradise, NV
163
186
189
172
174
159
159
145
144
15
17
30
33
44
49
Lebanon, PA
25
27
27
25
25
23
23
21
21
2
3
5
5
7
7
Little Rock-North Little Rock-Conway, AR
100
116
117
107
107
98
98
89
88
10
11
21
22
31
32
Los Angeles-Long Beach-Anaheim, CA
2,250
2,190
2,190
2,030
2,020
1,870
1,850
1,710
1,680
184
195
365
388
544
578
Louisville/Jefferson County, KY-IN
184
204
208
190
191
176
174
161
156
16
20
32
40
48
59
Macon, GA
41
48
48
44
45
41
41
38
37
4
4
8
9
11
13
McAllen-Edinburg-Mission, TX
94
110
110
101
102
93
93
85
85
9
10
19
20
28
29
Napa, CA
23
27
27
25
25
23
23
21
20
2
2
4
5
7
7
New York-Newark-Jersey City, NY-NJ-PA
3,540
4,020
4,130
3,750
3,810
3,480
3,480
3,200
3,160
310
368
616
730
918
1,090
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
1,000
1,040
1,070
971
958
898
846
823
732
84
127
167
251
249
374
San Luis Obispo-Paso Robles-Arroyo Grande, CA
29
33
33
30
30
28
28
25
25
3
3
6
6
8
9
St Louis, MO-IL
569
656
668
610
617
564
565
518
512
53
60
106
119
158
178
Weirton-Steubenville, WV-OH
44
44
45
41
41
38
37
35
33
4
4
7
9
10
13
Wheeling, WV-OH
48
56
56
51
52
47
47
43
43
5
5
10
10
14
15
Totals
13,300
14,300
14,600
13,300
13,300
12,300
12,100
11,300
10,900
1,140
1,400
2,270
2,770
3,390
4,130
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-63

-------
Table C-24. CBSA level results for the 30 annual-controlled urban study areas using the Pope et al., 2015 long-term IHD
mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12 fA
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
152
163
167
152
153
141
138
130
123
13
17
25
33
38
50
Altoona, PA
29
34
34
31
32
29
29
26
26
3
3
6
6
9
9
Atlanta-Sandy Springs-Roswell, GA
390
435
447
407
413
379
378
350
342
32
40
64
79
96
118
Birmingham-Hoover, AL
154
162
167
152
150
142
133
131
116
12
19
23
38
34
57
Canton-Massillon, OH
85
93
95
87
87
80
79
73
71
8
9
16
18
23
27
Chicago-Naperville-Elgin, IL-IN-WI
1,250
1,340
1,350
1,240
1,250
1,150
1,140
1,050
1,030
107
120
213
239
317
356
Cincinnati, OH-KY-IN
313
344
352
321
323
297
293
273
263
27
34
54
67
80
99
Cleveland-EIyria, OH
411
408
406
382
367
357
327
331
286
29
45
58
89
87
133
Detroit-Warren-Dearborn, MI
967
1,020
1,040
947
941
871
839
794
736
86
115
172
227
256
338
El Centra, CA
20
19
19
18
18
16
16
14
15
2
2
3
3
5
5
Elkhart-Goshen, IN
40
46
46
42
43
39
39
35
35
4
4
8
8
12
12
Evansville, IN-KY
57
66
67
61
62
57
57
52
51
6
6
11
13
16
19
Houston-The Woodlands-Sugar Land, TX
642
682
703
635
634
588
564
540
494
54
79
107
157
160
234
Indianapolis-Carmel-Anderson, IN
266
276
279
258
256
239
234
220
211
21
26
42
51
63
76
Johnstown, PA
37
40
42
38
38
35
34
32
31
3
4
6
8
10
12
Lancaster, PA
103
97
96
90
87
82
78
75
69
8
10
16
20
25
30
Las Vegas-Henderson-Paradise, NV
153
175
178
162
164
149
150
136
135
14
16
28
31
42
46
Lebanon, PA
24
26
26
24
24
22
22
20
20
2
2
4
5
6
7
Little Rock-North Little Rock-Conway, AR
94
109
110
101
101
92
92
83
83
10
10
19
20
29
30
Los Angeles-Long Beach-Anaheim, CA
2,120
2,070
2,060
1,920
1,900
1,760
1,740
1,610
1,580
172
183
342
364
510
543
Louisville/Jefferson County, KY-IN
174
192
196
179
180
165
163
152
147
15
19
30
37
45
56
Macon, GA
39
45
46
42
42
39
39
35
35
4
4
7
8
11
12
McAllen-Edinburg-Mission, TX
88
103
104
96
96
88
88
80
80
9
9
18
18
26
27
Napa, CA
22
25
25
23
23
21
21
19
19
2
2
4
4
6
7
New York-Newark-Jersey City, NY-NJ-PA
3,330
3,790
3,890
3,530
3,590
3,280
3,280
3,020
2,970
290
345
578
685
862
1,020
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
944
984
1,010
915
902
845
796
775
688
79
119
156
236
233
351
San Luis Obispo-Paso Robles-Arroyo Grande, CA
28
31
31
28
28
26
26
24
23
3
3
5
5
8
8
St Louis, MO-IL
536
618
629
575
581
531
532
487
482
50
56
99
112
148
167
Weirton-Steubenville, WV-OH
41
42
42
39
38
36
35
33
31
3
4
7
8
10
12
Wheeling, WV-OH
45
52
53
48
49
44
44
40
40
5
5
9
9
13
14
Totals
12,500
13,500
13,700
12,500
12,600
11,600
11,400
10,600
10,200
1,070
1,310
2,130
2,600
3,180
3,880
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-64

-------
Table C-25. CBSA level results for the 30 annual-controlled urban study areas using the Di et al., 2017b long-term all-cause
mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12 fA
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
547
589
602
548
549
507
496
465
441
45
59
90
117
134
174
Altoona, PA
104
123
123
113
113
103
104
94
94
11
11
21
21
32
32
Atlanta-Sandy Springs-Roswell, GA
1,940
2,180
2,240
2,030
2,060
1,890
1,880
1,740
1,700
158
194
314
387
470
578
Birmingham-Hoover, AL
709
745
770
697
692
649
613
601
533
52
85
104
170
156
253
Canton-Massillon, OH
300
329
335
305
307
281
278
256
249
27
31
53
63
80
93
Chicago-Naperville-Elgin, IL-IN-WI
4,220
4,520
4,570
4,200
4,200
3,870
3,840
3,550
3,470
350
396
698
789
1,040
1,180
Cincinnati, OH-KY-IN
1,160
1,280
1,300
1,190
1,190
1,100
1,080
1,000
970
98
120
196
240
293
358
Cleveland-EIyria, OH
1,290
1,280
1,280
1,200
1,150
1,120
1,020
1,030
891
89
138
178
274
266
410
Detroit-Warren-Dearborn, MI
2,430
2,570
2,620
2,380
2,360
2,180
2,100
1,990
1,840
211
283
421
562
630
840
El Centra, CA
51
48
48
44
45
40
41
36
37
4
4
8
8
12
12
Elkhart-Goshen, IN
114
133
133
122
123
112
112
101
101
11
12
23
23
34
35
Evansville, IN-KY
207
242
247
224
226
206
206
188
185
20
22
39
45
59
66
Houston-The Woodlands-Sugar Land, TX
2,130
2,260
2,340
2,100
2,100
1,940
1,870
1,780
1,630
174
256
347
510
519
761
Indianapolis-Carmel-Anderson, IN
950
989
997
921
915
852
832
783
749
74
89
148
178
221
266
Johnstown, PA
120
133
136
123
124
114
112
104
100
10
13
21
26
31
39
Lancaster, PA
397
374
370
346
334
317
299
288
263
31
38
62
76
93
114
Las Vegas-Henderson-Paradise, NV
543
622
633
575
582
529
531
482
479
49
54
98
108
146
161
Lebanon, PA
95
102
102
94
94
86
86
78
77
8
9
17
18
25
27
Little Rock-North Little Rock-Conway, AR
354
411
415
378
381
345
346
312
311
36
37
71
75
107
111
Los Angeles-Long Beach-Anaheim, CA
5,280
5,150
5,140
4,770
4,730
4,380
4,320
3,990
3,900
418
445
832
887
1,240
1,330
Louisville/Jefferson County, KY-IN
731
813
829
754
759
695
688
636
617
64
77
127
152
190
228
Macon, GA
129
149
152
138
140
128
128
117
115
12
13
23
26
35
39
McAllen-Edinburg-Mission, TX
213
251
252
231
232
212
212
192
192
21
22
42
44
62
65
Napa, CA
87
99
100
92
92
84
84
77
76
8
9
16
17
24
26
New York-Newark-Jersey City, NY-NJ-PA
7,690
8,770
9,020
8,170
8,310
7,570
7,580
6,960
6,850
649
781
1,290
1,560
1,940
2,320
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
3,260
3,400
3,480
3,160
3,110
2,910
2,740
2,660
2,360
266
401
530
798
792
1,190
San Luis Obispo-Paso Robles-Arroyo Grande, CA
108
120
121
111
111
101
101
92
91
10
10
20
21
30
31
St Louis, MO-IL
1,590
1,840
1,870
1,710
1,730
1,570
1,580
1,440
1,420
144
163
287
325
429
485
Weirton-Steubenville, WV-OH
102
104
104
96
95
89
86
82
76
8
10
16
20
24
30
Wheeling, WV-OH
124
144
145
133
133
122
122
110
110
12
13
24
25
36
37
Totals
37,000
39,800
40,500
36,900
37,000
34,100
33,500
31,200
29,900
3,070
3,800
6,120
7,560
9,150
11,300
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-65

-------
Table C-26. CBSA level results for the 30 annual-controlled urban study areas using the Pope et al., 2015 long-term all-cause
mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12 fA
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
595
641
656
596
598
551
539
506
479
49
63
97
126
145
188
Altoona, PA
107
126
126
116
116
106
106
96
96
11
11
22
22
32
33
Atlanta-Sandy Springs-Roswell, GA
2,310
2,590
2,660
2,420
2,450
2,240
2,230
2,070
2,020
186
229
371
457
555
683
Birmingham-Hoover, AL
831
874
903
817
811
761
717
704
623
61
100
121
198
181
296
Canton-Massillon, OH
318
349
355
323
325
297
294
271
263
28
33
56
66
84
98
Chicago-Naperville-Elgin, IL-IN-WI
4,660
4,990
5,040
4,630
4,640
4,270
4,230
3,910
3,820
384
434
767
866
1,150
1,290
Cincinnati, OH-KY-IN
1,310
1,440
1,480
1,340
1,350
1,240
1,220
1,130
1,100
110
136
220
270
329
404
Cleveland-EIyria, OH
1,390
1,380
1,370
1,290
1,230
1,200
1,100
1,110
956
96
147
191
293
285
438
Detroit-Warren-Dearborn, MI
2,720
2,880
2,940
2,660
2,640
2,440
2,350
2,220
2,050
235
314
469
625
702
933
El Centra, CA
59
56
56
51
52
47
47
42
42
5
5
10
10
14
14
Elkhart-Goshen, IN
125
146
146
134
135
123
123
111
111
12
13
25
25
37
38
Evansville, IN-KY
229
268
273
248
250
228
228
207
205
22
25
43
49
65
73
Houston-The Woodlands-Sugar Land, TX
2,590
2,760
2,850
2,560
2,560
2,360
2,270
2,170
1,980
211
310
421
617
629
922
Indianapolis-Carmel-Anderson, IN
1,080
1,130
1,130
1,050
1,040
968
946
889
851
84
101
168
201
251
300
Johnstown, PA
126
139
143
129
130
119
118
109
105
11
14
21
27
32
40
Lancaster, PA
402
378
373
349
337
320
301
290
265
31
38
62
77
93
114
Las Vegas-Henderson-Paradise, NV
631
723
737
669
677
615
617
560
557
57
63
113
125
170
187
Lebanon, PA
97
104
105
96
96
88
87
80
79
9
9
17
19
26
28
Little Rock-North Little Rock-Conway, AR
414
481
486
443
446
404
405
365
364
42
44
83
87
124
130
Los Angeles-Long Beach-Anaheim, CA
5,800
5,660
5,650
5,230
5,200
4,810
4,740
4,380
4,280
456
486
909
969
1,360
1,450
Louisville/Jefferson County, KY-IN
841
935
954
867
872
799
791
730
708
73
88
145
174
217
261
Macon, GA
153
177
180
164
166
151
151
139
137
14
16
27
31
41
46
McAllen-Edinburg-Mission, TX
243
286
288
264
265
241
241
219
218
24
25
47
49
71
74
Napa, CA
89
102
103
94
95
87
86
79
78
8
9
17
18
25
26
New York-Newark-Jersey City, NY-NJ-PA
8,230
9,400
9,670
8,750
8,890
8,100
8,110
7,450
7,330
694
831
1,380
1,660
2,070
2,480
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
3,570
3,730
3,820
3,460
3,410
3,190
3,000
2,910
2,580
290
438
578
872
864
1,300
San Luis Obispo-Paso Robles-Arroyo Grande, CA
112
125
125
115
115
105
105
95
95
10
11
21
21
31
32
St Louis, MO-IL
1,750
2,030
2,070
1,880
1,900
1,740
1,740
1,590
1,570
158
179
314
356
470
532
Weirton-Steubenville, WV-OH
112
114
115
106
105
98
94
90
84
9
11
17
22
26
33
Wheeling, WV-OH
129
150
151
138
139
127
127
115
114
13
13
25
26
38
39
Totals
41,000
44,200
45,000
41,000
41,000
37,800
37,100
34,600
33,200
3,390
4,190
6,760
8,350
10,100
12,500
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-66

-------
Table C-27. CBSA level results for the 30 annual-controlled urban study areas using the Thurston et al., 2016 long-term all-
cause mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
157
169
173
157
157
145
142
133
126
13
16
25
33
37
49
Altoona, PA
27
32
33
30
30
27
27
25
25
3
3
5
5
8
8
Atlanta-Sandy Springs-Roswell, GA
644
725
746
676
685
626
624
577
562
51
63
102
126
152
188
Birmingham-Hoover, AL
231
243
252
227
225
211
199
195
172
16
27
33
55
49
82
Canton-Massillon, OH
84
92
94
85
86
78
77
71
69
7
9
14
17
22
26
Chicago-Naperville-Elgin, IL-IN-WI
1,220
1,310
1,320
1,210
1,210
1,120
1,110
1,020
996
99
112
197
223
295
334
Cincinnati, OH-KY-IN
353
390
400
362
365
334
330
306
294
29
36
58
72
87
108
Cleveland-EIyria, OH
359
357
355
333
319
310
282
286
246
24
37
48
75
73
112
Detroit-Warren-Dearborn, MI
717
761
776
702
697
643
618
583
538
61
81
121
162
182
243
El Centra, CA
16
16
16
14
14
13
13
12
12
1
1
3
3
4
4
Elkhart-Goshen, IN
33
39
39
36
36
33
33
29
29
3
3
6
7
10
10
Evansville, IN-KY
62
72
74
67
68
61
61
56
55
6
7
11
13
17
19
Houston-The Woodlands-Sugar Land, TX
729
776
803
720
720
664
636
607
552
58
86
116
171
174
256
Indianapolis-Carmel-Anderson, IN
293
305
308
284
282
262
256
240
230
22
27
45
54
67
80
Johnstown, PA
31
34
35
32
32
29
29
27
26
3
3
5
7
8
10
Lancaster, PA
97
91
90
84
81
77
72
69
63
7
9
15
18
22
27
Las Vegas-Henderson-Paradise, NV
186
214
218
197
200
181
182
165
164
17
18
33
37
50
55
Lebanon, PA
25
26
26
24
24
22
22
20
20
2
2
4
5
6
7
Little Rock-North Little Rock-Conway, AR
116
135
137
124
125
113
113
102
102
11
12
23
24
34
36
Los Angeles-Long Beach-Anaheim, CA
1,470
1,430
1,430
1,320
1,310
1,210
1,190
1,100
1,080
113
120
225
240
338
360
Louisville/Jefferson County, KY-IN
231
258
263
239
240
220
217
201
194
20
24
39
47
59
71
Macon, GA
43
51
52
47
47
43
43
39
39
4
4
8
9
11
13
McAllen-Edinburg-Mission, TX
66
78
79
72
72
66
66
59
59
6
7
13
13
19
20
Napa, CA
22
25
26
23
24
21
21
19
19
2
2
4
4
6
6
New York-Newark-Jersey City, NY-NJ-PA
2,070
2,370
2,440
2,200
2,240
2,030
2,040
1,870
1,840
172
205
343
410
514
615
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
912
953
977
883
870
812
763
741
655
73
110
145
219
217
328
San Luis Obispo-Paso Robles-Arroyo Grande, CA
27
30
30
27
27
25
25
23
23
2
3
5
5
7
8
St Louis, MO-IL
463
539
550
499
505
460
460
420
415
41
46
82
93
122
139
Weirton-Steubenville, WV-OH
31
32
32
30
29
27
26
25
23
2
3
5
6
7
9
Wheeling, WV-OH
34
40
40
37
37
34
34
30
30
3
3
7
7
10
10
Totals
10,700
11,600
11,800
10,700
10,800
9,900
9,710
9,050
8,650
871
1,080
1,740
2,160
2,610
3,230
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-67

-------
Table C-28. CBSA level results for the 30 annual-controlled urban study areas using the Turner et al., 2016 long-term lung
cancer mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12 fA
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
48
51
52
48
48
44
43
41
38
4
5
8
10
12
15
Altoona, PA
6
7
7
7
7
6
6
6
6
1
1
1
1
2
2
Atlanta-Sandy Springs-Roswell, GA
183
204
210
191
194
178
177
164
160
15
18
29
36
44
54
Birmingham-Hoover, AL
63
66
69
62
62
58
55
54
47
5
8
9
15
14
23
Canton-Massillon, OH
25
28
28
26
26
24
24
22
21
2
3
5
5
7
8
Chicago-Naperville-Elgin, IL-IN-WI
379
406
410
377
378
348
345
319
312
32
36
63
71
94
106
Cincinnati, OH-KY-IN
122
134
137
125
126
115
114
106
102
10
13
20
26
31
38
Cleveland-EIyria, OH
111
111
110
103
99
96
88
89
77
8
12
15
24
23
35
Detroit-Warren-Dearborn, MI
220
233
237
215
214
198
190
180
166
19
26
38
51
57
76
El Centra, CA
4
4
4
3
3
3
3
3
3
0
0
1
1
1
1
Elkhart-Goshen, IN
10
11
11
10
10
9
9
9
9
1
1
2
2
3
3
Evansville, IN-KY
19
22
23
21
21
19
19
17
17
2
2
4
4
5
6
Houston-The Woodlands-Sugar Land, TX
194
206
213
191
191
177
170
162
148
16
24
31
47
47
70
Indianapolis-Carmel-Anderson, IN
102
106
107
99
98
91
89
84
80
8
10
16
19
24
29
Johnstown, PA
8
9
9
9
9
8
8
7
7
1
1
1
2
2
3
Lancaster, PA
28
26
26
24
23
22
21
20
18
2
3
4
5
6
8
Las Vegas-Henderson-Paradise, NV
55
63
64
58
59
53
53
49
48
5
5
10
11
15
16
Lebanon, PA
9
9
9
8
8
8
8
7
7
1
1
2
2
2
2
Little Rock-North Little Rock-Conway, AR
37
43
43
39
40
36
36
33
33
4
4
7
8
11
12
Los Angeles-Long Beach-Anaheim, CA
360
351
351
325
323
299
295
272
266
29
30
57
61
85
91
Louisville/Jefferson County, KY-IN
82
91
93
85
85
78
78
72
69
7
9
14
17
21
26
Macon, GA
13
15
15
14
14
13
13
11
11
1
1
2
3
3
4
McAllen-Edinburg-Mission, TX
11
13
13
12
12
11
11
10
10
1
1
2
2
3
3
Napa, CA
7
8
8
7
7
6
6
6
6
1
1
1
1
2
2
New York-Newark-Jersey City, NY-NJ-PA
590
672
691
626
637
580
581
534
525
50
60
99
119
148
178
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
284
296
303
275
271
253
238
232
205
23
35
46
70
69
104
San Luis Obispo-Paso Robles-Arroyo Grande, CA
8
9
9
8
8
7
7
7
7
1
1
1
2
2
2
St Louis, MO-IL
158
182
186
169
171
156
157
143
142
14
16
28
32
42
48
Weirton-Steubenville, WV-OH
9
10
10
9
9
8
8
8
7
1
1
1
2
2
3
Wheeling, WV-OH
11
12
12
11
11
10
10
9
9
1
1
2
2
3
3
Totals
3,150
3,400
3,460
3,160
3,160
2,920
2,860
2,670
2,560
262
327
522
651
780
972
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-68

-------
Table C-29. CBSA level results for the 30 annual-controlled urban study areas using the Baxter et al., 2017 all-cause short-
term mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12 fA
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
29
31
32
29
29
27
26
25
23
2
3
5
6
7
9
Altoona, PA
5
6
6
6
6
5
5
5
5
1
1
1
1
2
2
Atlanta-Sandy Springs-Roswell, GA
111
126
129
117
119
108
108
100
97
9
11
18
22
26
32
Birmingham-Hoover, AL
40
42
44
40
39
37
35
34
30
3
5
6
9
9
14
Canton-Massillon, OH
15
17
17
16
16
14
14
13
13
1
2
3
3
4
5
Chicago-Naperville-Elgin, IL-IN-WI
228
245
248
227
227
208
206
190
186
18
21
37
41
55
62
Cincinnati, OH-KY-IN
63
70
71
64
65
59
59
54
52
5
6
10
13
15
19
Cleveland-EIyria, OH
68
67
67
63
60
58
53
54
46
5
7
9
14
14
21
Detroit-Warren-Dearborn, MI
132
140
143
129
128
118
113
107
99
11
15
22
30
33
44
El Centra, CA
3
3
3
2
2
2
2
2
2
0
0
0
0
1
1
Elkhart-Goshen, IN
6
7
7
7
7
6
6
5
5
1
1
1
1
2
2
Evansville, IN-KY
11
13
13
12
12
11
11
10
10
1
1
2
2
3
3
Houston-The Woodlands-Sugar Land, TX
126
134
139
124
124
114
109
104
95
10
15
20
29
30
44
Indianapolis-Carmel-Anderson, IN
52
54
55
51
50
47
46
43
41
4
5
8
9
12
14
Johnstown, PA
6
7
7
6
6
6
6
5
5
0
1
1
1
1
2
Lancaster, PA
20
18
18
17
16
16
15
14
13
1
2
3
4
4
5
Las Vegas-Henderson-Paradise, NV
30
34
35
32
32
29
29
26
26
3
3
5
6
8
9
Lebanon, PA
5
5
5
5
5
4
4
4
4
0
0
1
1
1
1
Little Rock-North Little Rock-Conway, AR
20
23
24
21
22
20
20
18
18
2
2
4
4
6
6
Los Angeles-Long Beach-Anaheim, CA
284
277
277
255
254
234
231
212
208
22
23
43
46
65
69
Louisville/Jefferson County, KY-IN
41
45
46
42
42
38
38
35
34
3
4
7
8
10
12
Macon, GA
7
9
9
8
8
7
7
7
7
1
1
1
1
2
2
McAllen-Edinburg-Mission, TX
12
14
14
13
13
12
12
11
11
1
1
2
2
3
4
Napa, CA
4
5
5
5
5
4
4
4
4
0
0
1
1
1
1
New York-Newark-Jersey City, NY-NJ-PA
401
459
473
427
434
394
394
361
355
33
39
66
79
99
118
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
172
180
184
166
164
153
144
139
123
14
21
27
41
41
61
San Luis Obispo-Paso Robles-Arroyo Grande, CA
5
6
6
5
5
5
5
5
4
0
0
1
1
1
1
St Louis, MO-IL
84
98
100
90
91
83
83
76
75
7
8
15
17
22
25
Weirton-Steubenville, WV-OH
5
5
6
5
5
5
4
4
4
0
1
1
1
1
2
Wheeling, WV-OH
6
7
7
7
7
6
6
6
6
1
1
1
1
2
2
Totals
1,990
2,150
2,190
1,990
1,990
1,830
1,790
1,670
1,600
160
197
319
394
478
592
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-69

-------
Table C-30. CBSA level results for the 30 annual-controlled urban study areas using the Ito et al., 2013 all-cause short-term
mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12 fA
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
14
15
15
14
14
13
12
12
11
1
1
2
3
3
4
Altoona, PA
2
3
3
3
3
2
2
2
2
0
0
0
0
1
1
Atlanta-Sandy Springs-Roswell, GA
53
60
62
56
57
52
52
48
46
4
5
8
10
13
15
Birmingham-Hoover, AL
19
20
21
19
19
18
17
16
14
1
2
3
4
4
7
Canton-Massillon, OH
7
8
8
7
7
7
7
6
6
1
1
1
1
2
2
Chicago-Naperville-Elgin, IL-IN-WI
107
115
116
106
106
98
97
89
87
9
10
17
19
26
29
Cincinnati, OH-KY-IN
30
33
34
31
31
28
28
26
25
2
3
5
6
7
9
Cleveland-EIyria, OH
32
31
31
29
28
27
25
25
22
2
3
4
7
6
10
Detroit-Warren-Dearborn, MI
62
66
68
61
61
56
54
51
47
5
7
10
14
16
21
El Centra, CA
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
Elkhart-Goshen, IN
3
3
3
3
3
3
3
3
3
0
0
1
1
1
1
Evansville, IN-KY
5
6
6
6
6
5
5
5
5
0
1
1
1
1
2
Houston-The Woodlands-Sugar Land, TX
61
65
67
60
60
55
53
50
46
5
7
10
14
14
21
Indianapolis-Carmel-Anderson, IN
25
26
26
24
24
22
22
20
20
2
2
4
5
6
7
Johnstown, PA
3
3
3
3
3
3
3
2
2
0
0
0
1
1
1
Lancaster, PA
9
9
9
8
8
7
7
7
6
1
1
1
2
2
3
Las Vegas-Henderson-Paradise, NV
14
16
17
15
15
14
14
13
13
1
1
3
3
4
4
Lebanon, PA
2
2
2
2
2
2
2
2
2
0
0
0
0
1
1
Little Rock-North Little Rock-Conway, AR
10
11
11
10
10
9
9
8
8
1
1
2
2
3
3
Los Angeles-Long Beach-Anaheim, CA
133
130
129
120
119
109
108
99
97
10
11
20
22
30
32
Louisville/Jefferson County, KY-IN
19
22
22
20
20
18
18
17
16
2
2
3
4
5
6
Macon, GA
4
4
4
4
4
3
3
3
3
0
0
1
1
1
1
McAllen-Edinburg-Mission, TX
6
7
7
6
6
6
6
5
5
1
1
1
1
2
2
Napa, CA
2
2
2
2
2
2
2
2
2
0
0
0
0
1
1
New York-Newark-Jersey City, NY-NJ-PA
187
214
220
199
202
184
184
168
165
15
18
31
37
46
55
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
82
86
88
79
78
73
68
66
59
6
10
13
19
19
29
San Luis Obispo-Paso Robles-Arroyo Grande, CA
3
3
3
3
3
2
2
2
2
0
0
0
0
1
1
St Louis, MO-IL
40
47
48
43
44
40
40
36
36
4
4
7
8
11
12
Weirton-Steubenville, WV-OH
3
3
3
2
2
2
2
2
2
0
0
0
0
1
1
Wheeling, WV-OH
3
3
3
3
3
3
3
3
3
0
0
1
1
1
1
Totals
940
1,010
1,030
939
940
864
847
789
754
75
93
150
186
226
279
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-70

-------
Table C-31. CBSA level results for the 30 annual-controlled urban study areas using the Zanobetti et al., 2014 all-cause short-
term mortality CR function.
CBSA
Absolute Risk
Risk Reduction (Relative to Current Standard)
Current
Conditions
(2015)
Current
(12
Standard
g/m3)
Alternative Annual
Standard (11 jiig/m3)
Alternative Annual
Standard (10 jiig/m3)
Alternative Annual
Standard (9 jug/m3)
Alternative Annual
Standard (12-11 jig/m3)
Alternative Annual
Standard (12-10 ^ig/m3)
Alternative Annual
Standard (12-9 jxg/m3)
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Pri-PM
Sec-PM
Akron, OH
45
49
50
45
45
42
41
38
36
4
5
7
9
11
14
Altoona, PA
9
10
10
9
9
9
9
8
8
1
1
2
2
3
3
Atlanta-Sandy Springs-Roswell, GA
159
180
185
167
170
155
154
142
139
13
15
25
31
38
46
Birmingham-Hoover, AL
58
62
64
57
57
53
50
49
44
4
7
8
14
12
20
Canton-Massillon, OH
25
27
28
25
25
23
23
21
20
2
3
4
5
6
7
Chicago-Naperville-Elgin, IL-IN-WI
348
373
377
345
346
318
315
290
284
28
32
56
63
83
94
Cincinnati, OH-KY-IN
95
105
108
98
98
90
89
82
79
8
10
16
19
23
29
Cleveland-EIyria, OH
106
106
105
99
94
92
83
85
73
7
11
14
22
21
33
Detroit-Warren-Dearborn, MI
200
212
216
196
194
179
172
162
149
17
22
34
45
50
67
El Centra, CA
4
4
4
4
4
3
3
3
3
0
0
1
1
1
1
Elkhart-Goshen, IN
9
11
11
10
10
9
9
8
8
1
1
2
2
3
3
Evansville, IN-KY
17
20
21
19
19
17
17
15
15
2
2
3
4
5
5
Houston-The Woodlands-Sugar Land, TX
175
187
193
173
173
160
153
146
133
14
20
28
41
41
61
Indianapolis-Carmel-Anderson, IN
78
82
82
76
75
70
68
64
61
6
7
12
14
18
21
Johnstown, PA
10
11
11
10
10
9
9
9
8
1
1
2
2
2
3
Lancaster, PA
33
31
31
28
28
26
24
24
21
2
3
5
6
7
9
Las Vegas-Henderson-Paradise, NV
44
51
52
47
47
43
43
39
39
4
4
8
9
12
13
Lebanon, PA
8
8
8
8
8
7
7
6
6
1
1
1
1
2
2
Little Rock-North Little Rock-Conway, AR
29
34
34
31
31
28
28
26
25
3
3
6
6
9
9
Los Angeles-Long Beach-Anaheim, CA
435
425
424
392
389
359
354
326
319
33
35
66
71
99
106
Louisville/Jefferson County, KY-IN
60
67
69
62
63
57
57
52
50
5
6
10
12
15
18
Macon, GA
11
12
13
11
12
11
11
10
9
1
1
2
2
3
3
McAllen-Edinburg-Mission, TX
17
21
21
19
19
17
17
16
16
2
2
3
3
5
5
Napa, CA
7
8
8
8
8
7
7
6
6
1
1
1
1
2
2
New York-Newark-Jersey City, NY-NJ-PA
630
722
743
671
682
619
620
568
559
52
62
103
124
154
186
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
268
280
287
259
255
238
224
217
192
21
32
42
64
63
96
San Luis Obispo-Paso Robles-Arroyo Grande, CA
9
10
10
9
9
8
8
7
7
1
1
2
2
2
2
St Louis, MO-IL
131
152
155
141
142
129
130
118
117
11
13
23
26
34
39
Weirton-Steubenville, WV-OH
8
9
9
8
8
7
7
7
6
1
1
1
2
2
2
Wheeling, WV-OH
10
12
12
11
11
10
10
9
9
1
1
2
2
3
3
Totals
3,040
3,280
3,340
3,040
3,040
2,790
2,740
2,550
2,440
244
301
487
603
731
904
* Pri-PM (primary PM-based modeling approach), Sec-PM (secondary PM-based modeling approach)
C-71

-------
C.2.2 Impact of Alternative Standards on the Distribution of Risk Across Ambient PM2.5
Levels
The tables of risk results presented in section C.2.1 illustrate the estimated risk of
premature death under current and alternative PM2.5 standards. As the city-specific results
indicate, both total risk and risk reductions estimated to occur under alternative standards can
vary substantially by urban area. This is due to differences in underlying demographics (e.g., size
and age of population), health status (e.g., underlying death rates) and exposure (air quality
conditions). Furthermore, each of these CBSA estimates represents an aggregation of underlying
12 km grid cell results, masking the underlying variability in the distribution of risk under
different scenarios. Thus, it can be challenging to understand how patterns of risk are changing
under air quality simulated to just meet the current or alternative standards.
To better illustrate the distribution of risk under the current standards, and how that
distribution changes under potential alternative standards, this section presents graphics
depicting these changes both in aggregate and at the grid-cell level. It would be possible to
illustrate these changes separately for each endpoint and CR function, as was done numerically
in the tables in section C.2.1. However, because the pattern of risk and risk reduction is similar
across endpoints, we have chosen to focus on a single endpoint to illustrate the changes
graphically. Consequently, as with the graphics presented in the PA section 3.3.2, the graphics
presented in this section are also based on long-term exposure-related IHD mortality modeled
using effect estimates obtained from Jerrett et al. (2016). The first set of graphics presented in
this section (Figure C-25, Figure C-26, Figure C-27, Figure C-28, and Figure C-29) include
results for the full set of 47 urban study areas and the second set (Figure C-30 and Figure C-31)
include results for the 30 annual-controlled study areas. These graphical plots include:
•	Line graphs showing the distribution of gridded risk estimates across annual-
averaged PM2.5 concentrations (Figure C-25 and Figure C-30). These figures
allow the reader to consider how the distribution of risk shifts when simulating air
quality that just meets the current standard (12/35 |ig/m3) relative to 2015 current
conditions and subsequently how that distribution of risk shifts downward when
simulating air quality that just meets alternative standards of 10/30 |ig/m3,
•	Maps showing the 12 km grid-level risk estimates associated with each of the 47
urban study areas. In these representative maps each grid cell is shown as a
square, with the color of the square going from green (lower risk estimates) to red
(higher risk estimate) colors. The center of the color scales (the beginning of
yellow) has been set to a risk estimate of two premature deaths. This means that
green squares represent grid cells where 0-1 premature deaths are estimated,
C-72

-------
yellow squares represent grid cells in which at least two premature deaths are
estimated, and as the color graduation approaches red the number of estimated
premature deaths increases. Separate maps are presented for (a) the unadjusted
2015 current conditions simulation (Figure C-26), (b) simulation of the current
standard (12/35 |ig/m3) (Figure C-27), and (c) simulation of the change (delta) in
risk between the current and alternative standards (10/30 |ig/m3) (Figure C-28).
These maps are not repeated for just the 30 area set, as those areas are included in
the 47 area maps.
• Scatter plots depicting the distribution of modeled risk by annual-average PM2.5
concentration (Figure C-29 and Figure C-31). While these scatter plots present
similar distributional information as the line graphs, the scatter plots allow for a
more detailed consideration of the nature of the risk distribution in relation to
ambient PM2.5 levels. In these figures, each grid cell is shown as a dot, with the
frequency of dots shown on a color scale from cool (green - lower frequency) to
hot (red - higher frequency) colors.40 Consequently, it is possible to consider
whether, for example, a shift in risk involves a change in the magnitude of risk
across higher-risk cells, or in a change in the density of lower risk cells.
Key observations resulting from review of these graphics as well as the CBSA tables
presented in section C.2.1 are presented below, following the graphics.
40 For adjusted air quality, a small amount of risk is estimated at concentrations higher than the level of the annual
standard (e.g., some risk is estimated at an average concentration of 13 |ig/m3 when air quality is adjusted to just
meet the current standard). This can result because risk estimates are for a single year (i.e., 2015) within the 3 -
year design value period (i.e., 2014 to 2016). While the three-year average design value is 12.0 ng/m3, a single
year can have grid cells with annual average concentrations above or below 12.0 |ig/m3.
C-73

-------
Modeled
Scenario
Current
Conditions
(2015)
Just meeting
the current
standards
(12/35 ng/m3;
Just meeting
the alternative
standards
(10/30 ng/m3)
ro oj
E °
•jz 0J
¦B £
ro oj
E Q
8,000
6,000
4,000
2.000
8,000
6,000
4,000
2,000
6.000
4,000
2,000
Simulation Method
¦	Pri-PM
¦	Sec-PM
¦	2015
6 7 8 9 10 11 12 13
Annual PM Concentration (1 ng/m3 bins)
Figure C-25. Distribution of estimated PiVl2.5-associated mortality for current conditions
(2015), current standards (12/35 fig/m3), and alternative standards (10/30 jig/m3)
simulated for all 47 urban study areas.41
41 Risk is rounded toward zero into whole PM2 5 concentration values (e.g., risk estimate at 10 (.ig/m3 includes risk
occurring at 10.0-10.9 iig/'m3). Blue lines represent the Pri-PM risk estimates, green lines represent the Sec-PM risk
estimates, and black lines represent the 2015 current conditions risk estimates.
C-74

-------
%
*
'"IP
#
1



Risk Estimate (Premature Deaths)
0	484
Figure C-26. Estimated number of premature deaths (by 12 km grid cell) under 2015
current conditions in all 47 study areas.
•H •*
s-ip
^ J
1


~
Risk Estimate (Premature Deaths)
Figure C-27. Estimated number of premature deaths (by 12 km grid cell) when just
meeting the current PM standards (12/35) in all 47 study areas (Pri-PM simulation).
C-75

-------

v #
Risk Estimate (Premature Deaths)
0	484
Figure C-28. Estimated reduction in the number of premature deaths (by 12 km grid cell)
when going from just meeting the current standards (12/35) to just meeting the
alternative standards (10/30) in all 47 study areas (Pri-PM simulation).
C-76

-------
Modeled
Scenario
Current
Conditions
(2015)
Simulation
Method
2015
Just meeting
the current
standards
(12/35 ug/m3)
Just meeting
the alternative
standards
(10/30 |ig/mB)
Pri-PM
Pri-PM
ra 
-- E
ce Jjj
ra a
E a

400
200
0
400
200
0
400
200-
0-
•	mS. •*% .
—+~

7 8 9 10 11 12 13
Annual PM Concentration (ng/m3)
14 15 16 17
Figure C-29. Distribution of estimated premature death (by 12 km grid cell) for the current
standards (12/35 jig/m3), alternative standards (10/30 jig/m3), and current conditions
(2015) for all 47 urban study areas (Pri-PM simulation).
C-77

-------
TO QJ
E o
Modeled
Scenario
Current
Conditions
(2015)
Just meeting the
current annual
standard 12
Hg/m3
o> "M
Just meeting the to &
E Q
alternative t; o>
i/)
annual standard hj 3
, ,	Ja: to
llng/m3	E
TO Q>
E Q
Just meeting the
alternative
annual standard
10 ng/m3
Just meeting the
alternative
annual standard
9 |ig/m3
TO 0>
E O
TO 0>
E Q
8,000
6,000-
4,000-
2,000-
0-
8,000"
6,000-
4,000-
2,000-
0-
8,000"
6,000-
4,000-
2,000-
0-
8,000'
6,000-
4,000-
2,000-
0-
8,000"
6,000-
4,000-
2,000-
0-
Simulation Method
¦	2015
¦	Pri PM
¦	Sec PM
5 6 7 8 9 10
Annual PM Concentration (1 ng/m3 bins)
11
12
13
Figure C-30. Distribution of estimated PM2.5-associated mortality for current conditions
(2015), the current annual standard (12/35 jig/m3), and alternative standards (9.0,10.0,
and 11.0 jig/m3) simulated for the 30 annual-controlled urban study areas.42
42 Risk is rounded toward zero into whole PM2 5 concentration values (e.g., risk estimate at 10 (.ig/m3 includes risk
occurring at 10.0-10.9 iig/'m3). Blue lines represent the Pri-PM risk estimates, green lines represent the Sec-PM risk
estimates, and black lines represent the 2015 current conditions risk estimates.
C-78

-------
simulation
Modeled Scenario Method
"Fo ?
0
M "ZJ
rt
n
0
13
CL
rt
o'
U
(/)
Risk Estimate
(Premature
Deaths)
r\> -P*
0 0
000
1	1 1

























•• ~m »



> .





S a> 400-
Just meeting the £ J »
current annual Pri PM « f ro
4- A j n / 3 ^50; 200-
standard 12 M-9/m "So a Q
2
0-









•











•












V.

2 m 400-
m
Just meeting the E J «
alternative annual Pri PM 5 c S
. i i.. / 3 uj a 200-
standard 11 ug/m j* s- o
J/) Q_
<2 w
0-

















•










• . *



S a> 400-
ro i_
Just meeting the E 2 ^
alternative annual Pri PM tS p ^
a j ^ n / 3 U 5 ID 200-
standard 10 ug/m jc >- a
' CO CL
5 ^
0-



















•





«> — ••



• V

•
• • \ 1
, • r
Jrr.


S 01 400-
nj ^
Just meeting the E 2 ^
alternative annual Pri PM S c S
4- A A C\ / 3 ^ £ 0) 200-
standard 9 ug/m ^ s- a
Q_
5 v~"'
0-






























' • * 5*
• • »
.«




3 4 5 6 7 8 9 10 11 12 13
Annual PM Concentration of Lower Standard (jig/m3)
Figure C-31. Distribution of estimated premature death (by 12 km grid cell) for current
conditions (2015), the current annual standard (12.0 j^g/m3), alternative annual
standards (9.0,10.0,11.0 jig/m3), and for all 47 urban study areas (Pri-PM simulation).
Review of the CBSA-level risk estimates presented in Section C.2.1 along with the
distributional risk estimates presented in Section C.2.2 further support the key observations
presented in PA section 3.2. Briefly, these observations include:
• Under simulation of the current PM2.5 standards, long-term annual mortality
ranges up to 52,100 premature deaths (all-cause, based on Pope et al., 2015),
including 16,800 IHD-related deaths (based on Jerrett et al., 2016) and 3,950 lung
cancer deaths (based on Turner et al., 2016) for the full set of 47 urban study
areas. Estimates of short-term all-cause annual mortality range up to 3,870 deaths
(based on Zanobetti et al., 2014) for the full set of 47 urban study areas (Table C-
10).
C-79

-------
In considering the alternative suite of standards (10/30 |ig/m3) modeled for the
full set of 47 urban study areas, we note that larger risk reductions are estimated
for those urban study areas controlled by annual standards, relative to those
controlled by the 24-hour standard (Table C-10 and Table C-l 1).
Across the full set of alternative annual standards modeled including 11,10 and 9
|ig/m3 (each evaluated for the 30 annually-controlled study areas), we see a
consistent reduction in mortality (Table C-21 and Table C-22). In addition, we
note that these risk reductions are associated with iteratively lower ambient PM2.5
concentrations, such that with the lowest annual standard considered (9 |ig/m3)
the majority of remaining risk occurs in grid cells with ambient PM2.5
concentrations between 7 and 9 |ig/m3. In contrast, most of the risk occurring
under the current standard occurs in grid cells with ambient concentrations in the
range of 10-12 |ig/m3 (Figure C-29).
Patterns of risk reduction seen in the summary (aggregated) risk results tables
presented both in PA section 3.3 and in section C.2.1 are driven by considerable
underlying variability across both CBSAs and across the 12km grid-level risk
estimates. Specifically, if we consider the detailed CBSA-level risk estimates
presented in section C.2.1, we observe significant variation in the magnitude of
modeled risk across the 47 urban study areas. Similarly, if we consider both the
maps and scatter plots presented in section C.2.2, we see considerable spread (i.e.,
variability) in the grid-level risk estimates. We note that this underlying
variability in risk (both across CBSAs and across underlying 12km grid cells)
reflects local patterns of population density, baseline incidence and modeled
ambient PM2.5 levels. However, it is important to also note that the underlying
variability does not result from differences in CR functions, since for all mortality
endpoints modeled in this analysis, national-level effect estimates were utilized.
When considering the shift in the distribution of risks for the alternative standards
(Figure C-29 and Figure C-31), we note that risk reductions are estimated in grid
cells encompassing a wide range of PM2.5 concentrations. This includes grid cells
with typical (i.e. frequently occurring) concentrations (as seen in red) as well as
cells with concentrations that occur relatively infrequently (as seen in green).
Furthermore, these shifts reflect reductions both in areas with relatively few
estimated premature deaths (as represented by points near the bottom of each of
the scatter plots) and in areas with much larger numbers of estimated deaths
(points higher on the y-axis in these scatter plots).
C-80

-------
C.3 CHARACTERIZING VARIABILITY AND UNCERTAINTY IN RISK
ESTIMATES
An important component of the risk assessment is the characterization of variability and
uncertainty. Variability refers to the heterogeneity of a variable of interest within a population or
across different populations. Variability is inherent and cannot be reduced through further
research. Hence, the design of a population-level risk assessment is often focused on effectively
characterizing variability in estimated risks across populations. Uncertainty refers to the lack of
knowledge regarding the actual values of inputs to an analysis. In contrast to variability,
uncertainty can be reduced through improved measurement of key variables and ongoing model
refinement. This section discusses our approaches to addressing key sources of variability and
uncertainty in the PM2.5 risk assessment.
Variability in the risk of PIVh.s-associated mortality could result from a number of factors.
These can include variation in PM2.5 exposures within and across populations (e.g., due to
differences in behavior patterns, building characteristics, air quality patterns etc.) and in the
health responses to those exposures (e.g., because some groups are at increased risk of PM-
related health effects). There is also variation over space and time in both PM2.5 itself (e.g.,
concentrations, air quality patterns) and in the ambient pollutants that co-occur with PM2.5. In the
PM2.5 risk assessment discussed in this PA, we account for these and other sources of variability,
in part, by estimating risks based on CR functions from a number of epidemiologic studies.
These studies evaluate PM2.5 health effect associations for either annual or daily PM2.5 exposures
across various time periods; in numerous geographic locations, encompassing much or all of the
U.S.; in various populations, including some with the potential to be at higher risk than the
general population (e.g., older adults); and using a variety of methods to estimate PM2.5
exposures (e.g., hybrid modeling approaches, monitors) and to control for potential confounders.
In selecting areas in which to estimate PIVh.s-associated risks, we include areas that cover
multiple regions of the U.S., with varying population demographics. Additionally, we use two
different strategies for adjusting PM2.5 air quality, reflecting the potential for changes in ambient
PM2.5 concentrations to be influenced by changes in primary PM2.5 emissions and by changes in
precursor emissions that contribute to secondary particle formation.
Beyond the reliance on information from multiple epidemiologic studies to account for
the variability in key risk assessment inputs, we use a combination of quantitative and qualitative
approaches to more explicitly characterize the remaining uncertainty in risk estimates. The
characterization of uncertainty associated with risk assessments is often addressed in the
regulatory context using a tiered approach in which progressively more sophisticated methods
are used to evaluate and characterize sources of uncertainty depending on the overall complexity
of the risk assessment (WHO, 2008). Guidance documents developed by EPA for assessing air
C-81

-------
toxics-related risk and Superfund Site risks (U.S. EPA, 2004 U.S. EPA, 2001) as well as recent
guidance from the World Health Organization (WHO, 2008) specify multitiered approaches for
addressing uncertainty. The WHO guidance presents a four-tiered approach, where the decision
to proceed to the next tier is based on the outcome of the previous tier's assessment. The four
tiers described in the WHO guidance include:
•	Tier 0 - recommended for routine screening assessments, uses default uncertainty factors
(rather than developing site-specific uncertainty characterizations);
•	Tier 1 - the lowest level of site-specific uncertainty characterization, involves qualitative
characterization of sources of uncertainty (e.g., a qualitative assessment of the general
magnitude and direction of the effect on risk results);
•	Tier 2 - site-specific deterministic quantitative analysis involving sensitivity analysis,
interval-based assessment, and possibly probability bound (high- and low-end)
assessment; and
•	Tier 3 - uses probabilistic methods to characterize the effects on risk estimates of sources
of uncertainty, individually and combined.
With this four-tiered approach, the WHO framework provides a means for systematically
linking the characterization of uncertainty to the sophistication of the underlying risk assessment.
Ultimately, the decision as to which tier of uncertainty characterization to include in a risk
assessment will depend both on the overall sophistication of the risk assessment and the
availability of information for characterizing the various sources of uncertainty. EPA staff used
the WHO guidance as a framework for developing the approach used for characterizing
uncertainty in this risk assessment. The overall analysis in the PM NAAQS risk assessment is
relatively complex, thereby warranting consideration of a full probabilistic (WHO Tier 3)
uncertainty analysis. However, limitations in available information prevent this level of analysis
from being completed at this time. In particular, the incorporation of uncertainty related to key
elements of CR functions (e.g., alternative functional forms, etc.) into a full probabilistic WHO
Tier 3 analysis would require that probabilities be assigned to each competing specification of a
given model element (with each probability reflecting a subjective assessment of the probability
that the given specification is the "correct" description of reality). However, for many model
elements there is insufficient information on which to base these probabilities. One approach that
has been taken in such cases is expert elicitation; however, this approach is resource- and time-
intensive and consequently, it was not feasible to use this technique in the current PM NAAQS
review to support a WHO Tier 3 analysis.
For most elements of this risk assessment, rather than conducting a full probabilistic
uncertainty analysis, we have included qualitative discussions of the potential impact of
uncertainty on risk results (WHO Tierl) and/or completed sensitivity analyses assessing the
potential impact of sources of uncertainty on risk results. The remainder of this section is
C-82

-------
organized as follows. Those sources of uncertainty addressed quantitively in the risk assessment
are discussed in section C.3.1. Those sources of uncertainty addressed qualitatively in the risk
assessment are discussed in section C.3.2. Below we summarize key findings from both the
qualitative and quantitative assessments of variability and uncertainty in the context of assessing
overall confidence in the risk assessment and its estimates.
C.3.1 Quantitative Assessment of Uncertainty
The risk assessment includes three components which allow us to quantitatively evaluate
the impact of potentially important sources of uncertainty on the risk estimates generated. Each
of these is discussed below including conclusions drawn from each assessment regarding the
potential importance of each source of uncertainty:
•	95 percent CIs around point estimates of mortality risk: Each of the point estimates
presented in the results section includes 95 percent CIs generated by BenMAP-CE,
reflecting the standard error (SE) associated with the underlying effect estimate (i.e., a
measure of the statistical precision of the effect estimate). There is considerable variation
in the range of 95 percent CIs associated with the point estimates generated for this
analysis, with some health endpoint/study combinations displaying substantially greater
variability than others (e.g., short-term PM2.5 exposure and all-cause mortality based on
effect estimates from Ito et al., 2013 versus long-term PM2.5 exposure IHD mortality
estimates based on Jerrett et al., 2016, respectively—see tables presenting risk estimates
in section 3.3.2 of this PA). There are a number of factors potentially responsible for the
varying degrees of statistical precision in effect estimates, including sample size,
exposure measurement error, degree of control for confounders/effect modifiers, and
variability in PM2.5 concentrations.
•	Inclusion of range of mortality estimates reflecting variation in effect estimates across
studies: For some mortality endpoints, we include a range of risk estimates reflecting
different epidemiology studies and associated study designs (e.g., age ranges, methods
for controlling potential confounders). In some instances, we find that the effect estimate
used has only a small impact on risk estimates (i.e., modeling of IHD mortality using
effect estimates from Jerrett et al., 2016 and Pope et al., 2015, Table 3-5 in PA section
3.3.2). By contrast, for other mortality endpoints, such as all-cause mortality associated
with long-term exposures (e.g., Di et al., 2017b versus Thurston et al., 2016), the use of
different effect estimates can have a larger impact (section 3.3.2, Table 3-5). The degree
to which different CR functions result in different risk estimates could reflect differences
in study design and/or study populations evaluated, as well as other factors. For example,
the examination of different cohorts in Di et al., 2017b) and Thurston et al., 2016) could
contribute to greater divergence in risk estimates. Details regarding the design of
epidemiology studies providing effect estimates for this risk assessment are presented in
Table C-l).
•	Evaluation of two different strategies for simulating air quality scenarios: As noted
above, we use two methods to adjust air quality in order to simulate just meeting the
current and alternative standards (i.e., the Pri-PM-based method and the Sec-PM based
method). Our evaluation of these methods reflects the fact that there is variability, and
C-83

-------
uncertainty, in how emissions in a particular area could change such that the area "just
meets" either the current or alternative standards. By modeling risks based on adjusted
primary PM2.5 emissions and based on adjusted precursor emissions that contribute to
secondary PM2.5 formation, the risk assessment provides insight into the potential
significance of this source of uncertainty. As discussed in section 3.3.2 of this PA, the
approach to adjusting air quality had relatively modest impacts on overall risk estimates.
Specifically, the difference between the absolute risk estimates from two air quality
modeling approach methods was generally less than 5% (Table 3-5 in PA section 3.3.2).
C.3.2 Qualitative Uncertainty Analysis
While the methods described above address some of the potentially important sources of
uncertainty and variability in the risk assessment, there are a range of additional sources that
cannot be analyzed quantitatively due to limitations in data, methods and/or resources. We have
addressed these additional sources of uncertainty qualitatively (Table C-32).
In describing each source of uncertainty, we attempt to characterize both the magnitude
and direction of impact on mortality risk estimates, including our rationale for these
characterizations. The categories used in describing the potential magnitude of impact (i.e., low,
medium, or high) reflect EPA staff judgments on the degree to which a particular source of
uncertainty could produce a sufficient impact on risk estimates to influence the interpretation of
those estimates in the context of the PM NAAQS review. Sources classified as having a low
impact would not be expected to influence conclusions from the risk assessment. Sources
classified as having a medium impact have the potential to affect such conclusions and sources
classified as high are likely to influence conclusions. Because this classification of the potential
magnitude of impact of sources of uncertainty is qualitative, it is not possible to place a
quantitative level of impact on each of the categories.
C-84

-------
Table C-32. Qualitative analysis of sources of uncertainty and assessment of potential impact on risk assessment.
Source of
Uncertainty
Description
Direction
Magnitude
Comments
a) Simulating just
meeting current and
alternative
standards using
model-based
(Downscaler)
methods
a)	The baseline and adjusted
concentration fields were
developed using modeling to fill
spatial and temporal gaps in
monitoring and to explore air
quality scenarios of policy interest.
State-of-the-science modeling
methods were used, but model-
related biases and errors can
introduce uncertainty into the PM2.5
concentration estimates.
b)	Due to the national scale of the
assessment, the modeling
scenarios are based on "across-
the-board" emission changes in
which emissions of primary PMzsor
NOx and SO2 from all
anthropogenic sources throughout
the U.S. are scaled by fixed
percentages. Although this
approach tends to target the key
sources in each area, it does not
tailor emission changes to specific
periods or sources.
c)	Two adjustment cases were
applied that span a wide range of
emission conditions, but these
cases are necessarily a subset of
the full set of possible emission
cases that could be used to adjust
PM2.5 concentrations to just meet
standards.
This source of
uncertainty could
bias results in
either direction.
Medium
Use of state-of-the-science modeling systems with the
relative response factor adjustment approach provides
confidence in the broad features of the simulated national
PM2.5 distributions and how the distributions shift with
changing standards levels. Due to challenges in modeling
local features in the national annual simulations, quantitative
results for individual areas or small subsets of grid cells are
relatively uncertain compared with broad features of the
national PM2 5 distributions.
C-85

-------
Source of
Uncertainty
Description
Direction
Magnitude
Comments
b) Simulating just
meeting alternative
annual standards
with levels of 9.0
and 11.0 ug/m3
using linear
extrapolation/
interpolation
The use of extrapolation/
interpolation in simulating just
meeting annual standards
introduces uncertainty into the risk
assessment since this approach
does not fully capture potential
non-linearities associated with the
formation of secondary PM2.5.
Both
Medium
Extrapolation to generate the surface for 9.0 pg/m3 is
subject to greater uncertainty than interpolation to 11.0
|jg/m3 (i.e., since the former estimates concentrations below
those in modeled surfaces, while the latter estimates a
surface between two sets of modeled results). In addition,
linear extrapolation/interpolation based on the primary-PM
modeled surfaces (for current standard and 10.0 pg/m3) is
likely subject to less uncertainty than
extrapolation/interpolation based on the secondary-PM
modeled surfaces since the latter focus on secondary
formation which could involve a higher degree of non-
linearity.
c) Exposure
measurement error
in epidemiologic
studies assessing
the relationship
between mortality
and exposure to
ambient PM2.5
Epidemiologic studies have
employed a variety of approaches
to estimate population-level PM2.5
exposures (e.g., stationary
monitors, hybrid modeling
approaches). These approaches
are based on using measured or
predicted ambient PM2.5
concentrations as surrogates for
population exposures. As such,
exposure estimates in
epidemiologic studies are subject
to exposure error. This error in the
underlying epidemiologic studies
contributes to uncertainty in the risk
estimates that are based on
concentration-response
relationships in those studies.
Both
Low
Available studies indicate that PM2.5 health effect
associations are robust across various approaches to
estimating PM25 exposures. This includes recent studies
that estimate exposures using ground-based monitors alone
and studies that estimate exposures using data from
multiple sources (e.g., satellites, land use information,
modeling), in addition to monitors. While none of these
approaches eliminates the potential for exposure error in
epidemiologic studies, such error does not call into question
the findings of key PM2.5 epidemiologic studies. The ISA
notes that, while bias in either direction can occur, exposure
error tends to result in underestimation of health effects in
epidemiologic studies of PM exposure (U.S. EPA, 2019,
section 3.5). Consistent with this, a recent study Hart et al.
(2015) reports that correction for PM2.5 exposure error using
personal exposure information results in a moderately larger
effect estimate for long-term PM2.5 exposure and mortality
(though with wider confidence intervals). While most PM2.5
epidemiologic studies have not employed similar corrections
for exposure error, several studies report that restricting
analyses to populations in close proximity to a monitor (i.e.,
in order to reduce exposure error) result in larger PM2.5
C-86

-------
Source of
Uncertainty
Description
Direction
Magnitude
Comments




effect estimates (e.g., Willis et al., 2003; Kloog et al., 2013).
Thus, to the extent key PM2.5 epidemiologic studies are
subject to exposure error, correction for that error would
likely result in larger effect estimates, and thus larger
estimates of PM2 5-associated mortality incidence in the risk
assessment.
d) Representing
population-level
exposure with 12
km grid cell spatial
framework (in
context of modeling
long-term exposure-
related mortality)
The risk assessment utilizes a 12
km grid structure in modeling risk.
A source of uncertainty associated
with this approach is the mismatch
between the 12 km grid cell
framework and the exposure
estimation approaches used in the
epidemiology studies providing
effect estimates for the risk
assessment. This mismatch can
introduce additional exposure error
to risk estimates, beyond the error
in the underlying epidemiologic
study itself.
Both
Medium
There are a variety of spatial templates used across the five
epidemiology studies providing effect estimates used in the
risk assessment and that none of them are an exact match
with the 12km grid cell template used in the risk
assessment. For example, the Jerrett et al. (2013) effect
estimate is an ensemble model which integrates results
from a range of spatial templates (e.g., 1 km, 9.8, 30 km
and 36 km grids) while Pope et al. (2015) utilized a county-
level design. Differences between the exposure metric used
in the risk assessment and those used in the underlying
epidemiologic studies introduce uncertainty into risk
estimates.
e) Representing
population-level
exposure with 12
km grid cell spatial
framework (in
context of modeling
short-term
exposure-related
mortality)
As with long-term exposure-related
mortality, short-term exposure-
related mortality endpoints were
also modeled using the same 12
km grid cell template. The
disconnect between the spatial
template used in the underlying
short-term epidemiology studies
and the 12 km grid template used
in the risk assessment introduces
uncertainty into risk estimates.
Both
Medium-High
The three studies providing effect estimates for short-term
exposure-related mortality in the risk assessment all utilized
some form of urban-level spatial unit in characterizing
exposure (e.g., Baxter et al. (2017) utilizes the CBSA, Ito et
al. (2013), utilizes the MSA), which are larger (less spatially
differentiated) in general than the 12 km grid cells used in
modeling risk. This means that we are generally modeling
short-term exposure-related mortality at a finer level of
spatial resolution in the risk assessment than reflected in
the epidemiology studies supplying the effect estimates,
which does introduce uncertainty into the analysis.
f) Temporal
mismatch between
Several of the epidemiology
studies for long-term exposure-
Both
Low
This approach can be reasonable in the context of an
epidemiologic study evaluating health effect associations
C-87

-------
Source of
Uncertainty
Description
Direction
Magnitude
Comments
ambient air quality
data characterizing
exposure and
mortality in long-
term exposure-
related
epidemiology
studies
related mortality have a mismatch
between the time period associated
with ambient PM2.5 concentrations
used to characterize population-
level exposure and mortality data
(i.e., the ambient PM2.5 data
reflects a period near the end of
the mortality period for Jerrett et al.
(2016) and Pope et al. (2015)).


with long-term PM2.5 exposures, under the assumption that
spatial patterns in PM2.5 concentrations are not appreciably
different during time periods for which air quality information
is not available (e.g., Chen et al. (2016)), Thus, as long as
the overall spatial pattern of ambient PM2.5 levels in relation
to population-level exposure and mortality rates has held
relatively stable over time, then a temporal disconnect
between the time-period associated with mortality and the
ambient PM2.5 level used in characterizing exposure would
not be expected to introduce significant uncertainty into the
epidemiology studies and associated effect estimates.
g) Shape and
corresponding
statistical
uncertainty around
the CR function for
long-term and short-
term exposure-
related mortality
(especially at lower
ambient PM levels)
Interpreting the shapes of
concentration-response
relationships, particularly at PM2.5
concentrations near the lower end
of the air quality distribution, can be
complicated by relatively low data
density in the lower concentration
range, the possible influence of
exposure measurement error, and
variability among individuals with
respect to air pollution health
effects. These sources of variability
and uncertainty tend to smooth and
"linearize" population-level
concentration-response functions,
and thus could obscure the
existence of a threshold or
nonlinear relationship (U.S. EPA,
2015, section 6.c).
Both
Medium-High
With regard to long-term exposure-related (nonaccidental)
mortality, the ISA concludes that the majority of evidence
supports a linear, no-threshold concentration-response
relationship, though there is initial evidence indicating that
the slope of the concentration-response curve may be
steeper at lower concentrations for cardiovascular mortality
(U.S. EPA, 2019, section 1.5.3.2). For long-term exposure-
related mortality, the ISA notes that there is less certainty in
the shape of the concentration-response curve at mean
annual PM2.5 concentrations generally below 8 |jg/m3
because data density is reduced below this concentration
(section 11.2.4). Given that a portion of risk modeling in the
risk assessment does involve locations with ambient PM2.5
concentrations below 8 ug/m3 (although most of the
population modeled is associated with level above this), we
note the potential for significant uncertainty being introduced
into the risk assessment (particularly for that portion of risk
modeled at or below 8 ug/m3). With regard to short-term
exposure-related mortality, the ISA concludes that, while
difficulties remain in assessing the shape of the PM2.5-
mortality concentration-response relationship, as identified
in the 2009 PM ISA, and studies have not conducted
systematic evaluations of alternatives to linearity, recent
C-88

-------
Source of
Uncertainty
Description
Direction
Magnitude
Comments




studies continue to provide evidence of a no-threshold linear
relationship, with less confidence at concentrations lower
than 5 |jg/m3.
h) Potential
confounding of the
PM2.5-mortalty effect
Factors are considered potential
confounders if demonstrated in the
scientific literature to be related to
health effects and correlated with
PM. Omitting potential confounders
from analyses could either increase
or decrease the magnitude of PM2.5
effect estimates (e.g., Di et al.,
2017b, Figure S2 in Supplementary
Materials). Thus, not accounting for
confounders can introduce
uncertainty into effect estimates
and, consequently, into the risk
estimates generated using those
effect estimates. Confounders vary
according to study design,
exposure duration, and health
effect. While a range of
approaches to control for potential
confounders have been adopted
across the studies used in the risk
assessment, and across the
broader body of PM2.5
epidemiologic studies assessed in
the ISA, no individual study adjusts
for all potential confounders.
Both
Medium
Long-term PM2.5 exposure and mortality studies: For studies
of long-term exposures, potential confounders are those
that vary spatially. These may include socioeconomic
status, race, age, medication use, smoking status, stress,
noise, occupational exposures, and copollutant
concentrations. Cohort studies used to characterize the
PM25-mortality relationship used a variety of approaches to
account for these and other potential confounders (e.g., see
Appendix B, Table B-12). Across studies, a variety of study
designs and statistical approaches have been used to
account for potential confounding in the PM2.5-mortality
relationship. The fact that across this diverse body of
evidence epidemiologic studies continue to report
consistently positive associations that are often similar in
magnitude, adds support the conclusion that the PM2.5-
mortality association is robust. Specifically regarding
copollutants, the final PM ISA notes that, overall,
associations remained relatively unchanged in copollutant
models for total (nonaccidental) mortality, cardiovascular,
and respiratory adjusted for ozone (Figure 11-20). Studies
focusing on copollutant models with NO2, PM10-2.5, SO2 and
benzene were examined in individual studies, and across
these studies the PM2.5-mortality association was relatively
unchanged (Figure 11-21).
Short-term PM2.5 exposure and mortality studies: For studies
of short-term exposures, potential confounders are those
that vary temporally. These may include meteorology (e.g.,
temperature, humidity), day of week, season, medication
use, allergen exposure, copollutant concentrations, and
long-term temporal trends. Some recent studies have
C-89

-------
Source of
Uncertainty
Description
Direction
Magnitude
Comments




expanded the examination of potential confounders,
including long-term temporal trends, weather, and
copollutants. Overall, the ISA concludes that alternative
approaches to controlling for long-term temporal trends and
for the potential confounding effects of weather may
influence the magnitude of the association between PM2.5
exposures and mortality, but have not been found to
influence the direction of the observed association (U.S.
EPA, 2019, section 11.1.5.1). With regard to copollutants,
recent studies conducted outside the U.S. provide additional
evidence that associations between short-term PM2.5
exposures and mortality remain positive and relatively
unchanged in copollutant models with both gaseous
pollutants and PM10-2.5 (U.S. EPA, 2019, Section 11.1.4).
i) Compositional
and source
differences in PM
The composition of PM2.5 can differ
across study areas reflecting
underlying differences in primary
and secondary PM2.5 sources (both
natural and anthropogenic). If
these compositional differences
lead to differences in public health
impacts (per unit concentration in
ambient air) for PM2.5, then
uncertainty may be introduced into
risk estimates that are based on
concentration-response
relationships for PM2.5 mass.
Both
Low
The Integrated Synthesis chapter of the final ISA (Chapter
1, U.S. EPA, 2019) states that, the assessment of PM
sources and components confirms and continues to support
the conclusion from the 2009 PM ISA: Many PM2.5
components and sources are associated with health effects,
and the evidence does not indicate that any one source or
component is more strongly related with health effects than
PM2.5 mass.
j) Lag structure in
short-term
exposure-related
mortality
It can be challenging to
characterize the timing associated
with specific PM2.5-related health
effects and consequently specify
the lag-structure that should be
Both
Low-Medium
Given the emphasis placed in the risk assessment on
mortality (and specifically, IHD mortality), we focus here on
lags associated with cardiovascular-related mortality. The
ISA notes that the immediate effect of PM2.5 on
cardiovascular morbidity outcomes, specifically those
C-90

-------
Source of
Uncertainty
Description
Direction
Magnitude
Comments
epidemiology
studies
used in modeling those health
effects. This can introduce
uncertainty into the modeling of risk
for short-term exposure-related
endpoints.


related to ischemic events, are consistent with the lag
structure of associations observed in studies of
cardiovascular mortality that report immediate effects (i.e.,
lag 0-1 day), (final PM ISA, section 1.5.2.2, U.S. EPA, 2019)
k) Use of
associations
reported in
epidemiologic
studies to estimate
how mortality
incidence may
change with
changing PM2.5 air
quality.
The ISA's determination that the
evidence supports a causal
relationship between PM2.5
exposure and mortality is based on
assessing a broad body of
evidence from epidemiologic and
experimental studies. Thus, the
use of the concentration-response
relationship from any individual
epidemiologic study to estimate
how mortality incidence may
change with changing PM2.5 air
quality is subject to uncertainty.
Both
Low
The ISA assesses a longstanding body of health evidence
supporting relationships between PM2.5 exposures (short-
and long-term) and mortality. Much of this evidence comes
from epidemiologic studies conducted in North America,
Europe, or Asia that demonstrate generally positive, and
often statistically significant, associations between PM2.5
exposures and total or cause-specific mortality. In addition,
recent experimental evidence, as well as evidence from
panel studies, strengthens support for potential biological
pathways through which PM2.5 exposures could lead to
serious health outcomes, including mortality. While this
broad body of evidence from across disciplines provides the
foundation for the ISA's conclusions, the risk assessment
necessarily focuses on a small number of individual studies.
Although the studies selected for the risk assessment are
part of the evidence base supporting the ISA's causality
determinations for mortality, the concentration-response
relationship in any given study reflects the particular time
period, locations, air quality distribution and populations
evaluated in that study. Thus, the use of the concentration-
response relationship from any individual epidemiologic
study to estimate mortality incidence across the U.S. for
populations, locations and PM2.5 air quality distributions
different from those present during the study period is
subject to uncertainty.
C-91

-------
C.3.3 Conclusion
To increase overall confidence in the risk assessment, a deliberative process has been
used in specifying each of the analytical elements comprising the risk model, including selection
of urban study areas as well as specification of other inputs such as CR functions. This
deliberative process involved rigorous review of available literature addressing both PM2.5
exposure and risk combined with the application of a formal set of criteria to guide development
of each of the key analytical elements in the risk assessment. In addition, the risk assessment
design reflects consideration of CASAC and public comments on the Integrated Review Plan
(IRP) for the PM NAAQS (U.S. EPA, 2016). The application of this deliberative process
increases overall confidence in the risk estimates by ensuring that the estimates are based on the
best available science and data characterizing PM2.5 exposure and risk, and that they reflect
consideration of input from experts on PM exposure and risk through CASAC and public
reviews.
C-92

-------
C.4 PM2 5 DESIGN VALUES FOR THE AIR QUALITY PROJECTIONS
Table C-33. PM2.5 DVs for the Primary PM projection case and 12/35 standard level.
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m 3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
AkronO
391530017
Annual
Yes
0
-18
10.99
11.99
23.7
25.4
AkronO
391530023
Annual
No
0
-18
9.16
9.90
20.2
21.4
Altoon
420130801
Annual
Yes
0
-41
10.11
12.02
23.8
29.5
Atlant
131210039
Annual
Yes
0
-27
10.38
11.99
19.7
22.6
Atlant
132230003
Annual
No
0
-27
7.82
8.62
16.2
17.5
Atlant
131350002
Annual
No
0
-27
8.84
10.05
17.9
20.2
Atlant
130890002
Annual
No
0
-27
9.34
10.63
19.2
21.7
Atlant
130670003
Annual
No
0
-27
9.51
10.79
18.6
21.0
Atlant
130630091
Annual
No
0
-27
9.86
11.19
19.1
21.6
Bakers
060290010
24-hr
Yes
79
77
16.52
10.23
70.0
35.4
Bakers
060290016
24-hr
No
79
77
18.45
11.45
61.3
31.7
Bakers
060290015
24-hr
No
79
77
5.15
3.97
15.8
13.6
Bakers
060290014
24-hr
No
79
77
16.53
9.81
61.4
31.7
Bakers
060290011
24-hr
No
79
77
6.06
4.84
19.6
16.6
Birmin
010732059
Annual
Yes
0
-10
11.25
12.00
22.3
23.9
Birmin
010732003
Annual
No
0
-10
10.08
10.70
19.0
20.1
Birmin
010731010
Annual
No
0
-10
9.78
10.30
19.2
20.1
Birmin
010730023
Annual
No
0
-10
10.94
11.66
22.8
24.2
Canton
391510017
Annual
Yes
0
-23
10.81
12.04
23.7
26.1
Canton
391510020
Annual
No
0
-23
9.91
10.96
22.0
23.6
Chicag
170313103
Annual
Yes
0
-15
11.10
12.00
22.6
24.2
Chicag
550590019
Annual
No
0
-15
8.04
8.56
20.4
21.5
Chicag
181270024
Annual
No
0
-15
9.51
10.30
22.4
24.1
Chicag
180892004
Annual
No
0
-15
9.84
10.71
24.7
26.7
Chicag
180890031
Annual
No
0
-15
10.12
11.01
23.6
25.6
Chicag
180890026
Annual
No
0
-15
-
-
25.2
27.1
Chicag
180890022
Annual
No
0
-15
-
-
22.7
24.8
Chicag
180890006
Annual
No
0
-15
10.03
10.93
23.1
25.2
Chicag
171971011
Annual
No
0
-15
8.36
8.85
18.4
19.3
Chicag
171971002
Annual
No
0
-15
7.69
8.23
20.0
21.2
Chicag
170890007
Annual
No
0
-15
8.94
9.55
19.2
20.5
Chicag
170890003
Annual
No
0
-15
-
-
19.2
20.0
Chicag
170434002
Annual
No
0
-15
8.87
9.48
19.9
20.7
Chicag
170316005
Annual
No
0
-15
10.79
11.66
24.1
26.1
C-93

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Chicag
170314201
Annual
No
0
-15
9.00
9.61
21.4
22.6
Chicag
170314007
Annual
No
0
-15
9.49
10.17
-
-
Chicag
170313301
Annual
No
0
-15
10.37
11.18
23.5
25.2
Chicag
170310076
Annual
No
0
-15
10.18
10.96
22.5
24.0
Chicag
170310057
Annual
No
0
-15
11.03
11.89
26.8
28.4
Chicag
170310052
Annual
No
0
-15
10.00
10.78
23.3
24.9
Chicag
170310022
Annual
No
0
-15
10.38
11.30
22.4
23.9
Chicag
170310001
Annual
No
0
-15
10.13
10.88
21.7
23.4
Cincin
390610014
Annual
Yes
0
-24
10.70
12.02
22.9
24.7
Cincin
390610042
Annual
No
0
-24
10.29
11.47
22.6
24.5
Cincin
390610040
Annual
No
0
-24
9.45
10.53
21.0
22.9
Cincin
390610010
Annual
No
0
-24
9.43
10.41
21.3
22.9
Cincin
390610006
Annual
No
0
-24
9.46
10.56
20.3
21.8
Cincin
390170020
Annual
No
0
-24
-
-
24.2
26.5
Cincin
390170019
Annual
No
0
-24
10.24
11.51
22.0
23.8
Cincin
390170016
Annual
No
0
-24
9.79
10.91
22.1
23.7
Cincin
210373002
Annual
No
0
-24
9.06
10.00
20.9
22.6
Clevel
390350065
Annual
Yes
0
2
12.17
12.03
24.9
24.6
Clevel
391030004
Annual
No
0
2
8.73
8.66
19.6
19.5
Clevel
390933002
Annual
No
0
2
8.10
8.03
20.2
20.1
Clevel
390850007
Annual
No
0
2
7.88
7.82
17.4
17.3
Clevel
390351002
Annual
No
0
2
8.86
8.78
19.5
19.4
Clevel
390350045
Annual
No
0
2
10.61
10.50
22.9
22.7
Clevel
390350038
Annual
No
0
2
11.38
11.25
25.0
24.8
Clevel
390350034
Annual
No
0
2
8.87
8.79
20.4
20.2
Detroi
261630033
Annual
Yes
0
-15
11.30
12.04
26.8
28.4
Detroi
261630039
Annual
No
0
-15
9.11
9.63
22.3
23.7
Detroi
261630036
Annual
No
0
-15
8.68
9.13
21.8
23.2
Detroi
261630025
Annual
No
0
-15
8.98
9.54
24.1
25.2
Detroi
261630019
Annual
No
0
-15
9.18
9.75
22.4
24.1
Detroi
261630016
Annual
No
0
-15
9.62
10.19
24.4
25.4
Detroi
261630015
Annual
No
0
-15
11.19
11.91
25.5
27.0
Detroi
261630001
Annual
No
0
-15
9.50
10.14
23.3
24.9
Detroi
261470005
Annual
No
0
-15
8.89
9.34
24.3
25.4
Detroi
261250001
Annual
No
0
-15
8.86
9.41
24.2
25.7
Detroi
260990009
Annual
No
0
-15
8.80
9.29
26.2
27.6
ElCent
060250005
Annual
Yes
0
12
12.63
12.00
33.5
31.3
ElCent
060251003
Annual
No
0
12
7.44
7.01
19.8
18.5
C-94

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
ElCent
060250007
Annual
No
0
12
8.37
7.99
21.5
20.8
Elkhar
180390008
Annual
Yes
0
-47
10.24
12.01
28.6
33.2
Evansv
181630023
Annual
Yes
0
-44
10.11
12.03
21.5
24.0
Evansv
211010014
Annual
No
0
-44
9.64
11.32
20.7
22.3
Evansv
181630021
Annual
No
0
-44
9.84
11.68
21.6
23.3
Evansv
181630016
Annual
No
0
-44
10.02
11.91
22.0
24.0
Fresno
060195001
24-hr
Yes
0
70
14.08
10.87
49.3
35.4
Fresno
060195025
24-hr
No
0
70
13.63
9.98
47.9
31.7
Fresno
060192009
24-hr
No
0
70
8.47
7.26
31.3
25.1
Fresno
060190011
24-hr
No
0
70
14.07
10.01
53.8
34.4
Hanfor
060310004
24-hr
Yes
65
79
21.98
11.79
72.0
35.4
Hanfor
060311004
24-hr
No
65
79
16.49
9.68
58.9
30.7
Housto
482011035
Annual
Yes
0
-14
11.19
12.04
22.4
24.0
Housto
482011039
Annual
No
0
-14
9.22
9.82
21.7
23.1
Housto
482010058
Annual
No
0
-14
9.67
10.37
22.3
23.8
Housto
481671034
Annual
No
0
-14
7.36
7.57
20.3
20.8
Indian
180970087
Annual
Yes
0
-10
11.44
12.01
25.9
26.8
Indian
180970083
Annual
No
0
-10
11.06
11.59
23.9
24.9
Indian
180970081
Annual
No
0
-10
11.07
11.61
25.0
26.0
Indian
180970078
Annual
No
0
-10
10.14
10.60
24.4
24.9
Indian
180970043
Annual
No
0
-10
-
-
26.0
26.4
Indian
180950011
Annual
No
0
-10
9.05
9.40
21.8
22.3
Indian
180570007
Annual
No
0
-10
9.02
9.39
21.4
22.1
Johnst
420210011
Annual
Yes
0
-25
10.68
12.03
25.8
30.3
Lancas
420710012
Annual
Yes
0
12
12.83
12.00
32.7
30.4
Lancas
420710007
Annual
No
0
12
10.57
9.88
29.8
27.4
LasVeg
320030561
Annual
Yes
0
-22
10.28
11.98
24.5
29.4
LasVeg
320032002
Annual
No
0
-22
9.79
11.38
19.8
23.4
LasVeg
320031019
Annual
No
0
-22
5.18
5.70
11.5
12.2
LasVeg
320030540
Annual
No
0
-22
8.80
10.21
21.7
25.9
Lebano
420750100
Annual
Yes
0
-15
11.20
12.02
31.4
33.9
Little
051191008
Annual
Yes
0
-41
10.27
12.03
21.7
24.7
Little
051190007
Annual
No
0
-41
9.78
11.76
20.5
24.0
Loganll
490050007
24-hr
Yes
0
-7
6.95
7.15
34.0
35.4
LosAng
060371103
Annual
Yes
0
5
12.38
12.03
32.8
32.1
LosAng
060592022
Annual
No
0
5
7.48
7.33
15.3
15.0
LosAng
060590007
Annual
No
0
5
9.63
9.37
-
-
LosAng
060374004
Annual
No
0
5
10.25
9.97
27.3
26.7
C-95

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
LosAng
060374002
Annual
No
0
5
11.06
10.76
29.2
28.6
LosAng
060371602
Annual
No
0
5
11.86
11.52
32.3
31.5
LosAng
060371302
Annual
No
0
5
11.99
11.64
31.5
30.8
LosAng
060371201
Annual
No
0
5
9.46
9.24
25.6
25.0
LosAng
060370002
Annual
No
0
5
10.52
10.27
29.2
28.6
Louisv
180190006
Annual
Yes
0
-27
10.64
12.04
23.9
26.2
Louisv
211110075
Annual
No
0
-27
10.42
11.84
22.3
24.3
Louisv
211110067
Annual
No
0
-27
9.55
10.78
21.4
23.6
Louisv
211110051
Annual
No
0
-27
10.29
11.48
21.8
23.7
Louisv
211110043
Annual
No
0
-27
10.37
11.72
22.0
24.1
Louisv
180431004
Annual
No
0
-27
9.96
11.20
22.0
24.2
Louisv
180190008
Annual
No
0
-27
8.72
9.69
20.1
21.5
MaconG
130210007
Annual
Yes
0
-39
10.13
12.01
21.2
24.8
MaconG
130210012
Annual
No
0
-39
7.68
8.90
16.6
18.6
Madera
060392010
24-hr
Yes
0
56
13.30
11.03
45.1
35.3
McAlle
482150043
Annual
Yes
0
-67
10.09
12.02
25.0
27.4
Merced
060470003
24-hr
Yes
0
28
11.81
10.97
39.0
35.4
Merced
060472510
24-hr
No
0
28
11.68
10.57
39.8
35.1
Modest
060990006
24-hr
Yes
0
51
13.02
10.70
45.7
35.3
Modest
060990005
24-hr
No
0
51
-
-
38.8
32.5
NapaCA
060550003
Annual
Yes
0
-47
10.36
12.03
25.1
29.1
NewYor
360610128
Annual
Yes
0
-26
10.20
12.00
23.9
27.8
NewYor
361030002
Annual
No
0
-26
7.18
8.10
18.8
21.0
NewYor
360810124
Annual
No
0
-26
7.52
8.65
19.5
22.4
NewYor
360710002
Annual
No
0
-26
6.95
7.81
17.5
19.6
NewYor
360610134
Annual
No
0
-26
9.70
11.38
21.6
25.0
NewYor
360610079
Annual
No
0
-26
8.42
9.82
22.8
25.6
NewYor
360470122
Annual
No
0
-26
8.66
10.10
20.5
23.7
NewYor
360050133
Annual
No
0
-26
9.05
10.53
24.0
28.0
NewYor
360050110
Annual
No
0
-26
7.39
8.56
19.4
22.8
NewYor
340392003
Annual
No
0
-26
8.59
9.87
23.6
26.3
NewYor
340390004
Annual
No
0
-26
9.87
11.40
24.2
27.3
NewYor
340310005
Annual
No
0
-26
8.42
9.63
22.2
24.7
NewYor
340292002
Annual
No
0
-26
7.23
8.04
18.1
19.8
NewYor
340273001
Annual
No
0
-26
6.78
7.56
17.1
18.8
NewYor
340171003
Annual
No
0
-26
8.79
10.15
23.4
26.9
NewYor
340130003
Annual
No
0
-26
8.89
10.21
23.8
27.3
NewYor
340030003
Annual
No
0
-26
8.90
10.22
24.5
27.4
C-96

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
OgdenC
490110004
24-hr
Yes
0
-18
7.28
7.77
32.6
35.4
OgdenC
490570002
24-hr
No
0
-18
8.99
9.73
-
-
OgdenC
490030003
24-hr
No
0
-18
6.35
6.76
-
-
Philad
420450002
Annual
Yes
0
-8
11.46
12.04
26.0
27.2
Philad
421010057
Annual
No
0
-8
10.86
11.37
27.0
28.4
Philad
421010055
Annual
No
0
-8
11.43
12.03
27.5
29.0
Philad
421010048
Annual
No
0
-8
10.27
10.77
25.6
27.0
Philad
420290100
Annual
No
0
-8
9.64
10.03
23.9
25.1
Philad
340150004
Annual
No
0
-8
8.33
8.69
20.6
21.5
Philad
340071007
Annual
No
0
-8
8.84
9.23
21.0
22.0
Philad
340070002
Annual
No
0
-8
10.19
10.61
23.5
24.6
Philad
240150003
Annual
No
0
-8
8.70
9.02
22.6
23.4
Philad
100031012
Annual
No
0
-8
9.04
9.40
23.0
23.8
Pittsb
420030064
Annual
Yes
0
13
12.82
12.00
35.8
32.8
Pittsb
421290008
Annual
No
0
13
8.65
8.15
19.6
18.9
Pittsb
421255001
Annual
No
0
13
8.35
7.89
17.8
17.2
Pittsb
421250200
Annual
No
0
13
8.95
8.44
19.3
18.2
Pittsb
421250005
Annual
No
0
13
11.02
10.38
22.7
21.2
Pittsb
420070014
Annual
No
0
13
10.11
9.48
21.9
20.5
Pittsb
420050001
Annual
No
0
13
11.03
10.30
21.9
20.5
Pittsb
420031301
Annual
No
0
13
11.00
10.30
24.8
23.0
Pittsb
420031008
Annual
No
0
13
9.78
9.16
20.5
19.3
Pittsb
420030008
Annual
No
0
13
9.50
8.85
20.5
19.0
Prinev
410130100
24-hr
Yes
0
10
8.60
8.17
37.6
35.3
ProvoO
490494001
24-hr
Yes
0
-30
7.74
8.57
30.9
35.3
ProvoO
490495010
24-hr
No
0
-30
6.73
7.52
-
-
ProvoO
490490002
24-hr
No
0
-30
7.41
8.31
28.9
33.2
Rivers
060658005
24-hr
Yes
0
36
14.48
11.51
43.2
35.3
Rivers
060658001
24-hr
No
0
36
-
-
36.5
29.6
Sacram
060670006
24-hr
Yes
0
-23
9.31
10.40
31.4
35.4
Sacram
061131003
24-hr
No
0
-23
6.62
7.19
15.8
17.3
Sacram
060670012
24-hr
No
0
-23
7.30
8.01
19.8
21.2
Sacram
060670010
24-hr
No
0
-23
8.67
9.65
26.5
29.9
Sacram
060610006
24-hr
No
0
-23
7.58
8.47
20.3
22.3
Sacram
060610003
24-hr
No
0
-23
6.71
7.26
19.3
20.2
SaltLa
490353010
24-hr
Yes
0
44
-
-
41.5
35.3
SaltLa
490353006
24-hr
No
0
44
7.62
6.19
36.8
30.2
SaltLa
490351001
24-hr
No
0
44
7.07
5.85
32.1
25.8
C-97

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
SanLui
060792007
Annual
Yes
0
-46
10.70
12.04
25.9
29.4
SanLui
060798002
Annual
No
0
-46
5.71
6.33
-
-
SanLui
060792004
Annual
No
0
-46
8.25
9.26
19.8
21.4
SouthB
181410015
24-hr
Yes
0
-23
10.45
11.37
32.5
35.4
St.Lou
290990019
Annual
Yes
0
-39
10.12
12.02
22.8
24.9
St.Lou
295100094
Annual
No
0
-39
9.57
11.38
23.3
25.9
St.Lou
295100093
Annual
No
0
-39
-
-
23.7
26.6
St.Lou
295100085
Annual
No
0
-39
10.10
12.01
23.6
26.2
St.Lou
295100007
Annual
No
0
-39
9.78
11.52
23.7
26.4
St.Lou
291893001
Annual
No
0
-39
9.85
11.72
22.4
25.2
Stockt
060771002
24-hr
Yes
0
17
12.23
11.30
38.7
35.4
Stockt
060772010
24-hr
No
0
17
10.74
9.96
37.3
34.3
Visali
061072002
24-hr
Yes
48
56
16.23
10.93
54.0
35.4
Weirto
390810017
Annual
Yes
0
-5
11.75
12.02
27.2
27.8
Weirto
540090011
Annual
No
0
-5
9.75
9.95
22.8
23.5
Weirto
540090005
Annual
No
0
-5
10.52
10.74
22.4
22.9
Weirto
390810021
Annual
No
0
-5
9.29
9.47
22.2
22.6
Wheeli
540511002
Annual
Yes
0
-44
10.24
12.02
22.5
25.4
Wheeli
540690010
Annual
No
0
-44
9.61
11.32
19.7
22.6
a CBSA names are the first six characters of the full CBSAs names in Table C-3.
b Percent reduction in NOx and SO2 emissions associated with just meeting the standard in this case.
c Percent reduction in Primary PM2.5 emissions associated with just meeting the standard in this case.
C-98

-------
Table C-34. PM2.5 DVs for the Secondary PM projection case and 12/35 standard level.
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m 3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
AkronO
391530017
Annual
Yes
-67
0
10.99
12.04
23.7
26.8
AkronO
391530023
Annual
No
-67
0
9.16
10.20
20.2
21.8
Altoon
420130801
Annual
Yes
N/A
N/A
10.11
12.04
23.8
28.3
Atlant
131210039
Annual
Yes
N/A
N/A
10.38
12.04
19.7
22.9
Atlant
132230003
Annual
No
N/A
N/A
7.82
9.07
16.2
18.8
Atlant
131350002
Annual
No
N/A
N/A
8.84
10.25
17.9
20.8
Atlant
130890002
Annual
No
N/A
N/A
9.34
10.83
19.2
22.3
Atlant
130670003
Annual
No
N/A
N/A
9.51
11.03
18.6
21.6
Atlant
130630091
Annual
No
N/A
N/A
9.86
11.44
19.1
22.2
Bakers
060290010
24-hr
Yes
N/A
N/A
16.52
10.40
70.0
35.4
Bakers
060290016
24-hr
No
N/A
N/A
18.45
11.61
61.3
31.0
Bakers
060290015
24-hr
No
N/A
N/A
5.15
3.24
15.8
8.0
Bakers
060290014
24-hr
No
N/A
N/A
16.53
10.40
61.4
31.1
Bakers
060290011
24-hr
No
N/A
N/A
6.06
3.81
19.6
9.9
Birmin
010732059
Annual
Yes
-56
0
11.25
12.03
22.3
24.2
Birmin
010732003
Annual
No
-56
0
10.08
10.86
19.0
21.5
Birmin
010731010
Annual
No
-56
0
9.78
10.68
19.2
21.4
Birmin
010730023
Annual
No
-56
0
10.94
11.73
22.8
25.3
Canton
391510017
Annual
Yes
-78
0
10.81
12.04
23.7
26.1
Canton
391510020
Annual
No
-78
0
9.91
11.14
22.0
24.8
Chicag
170313103
Annual
Yes
N/A
N/A
11.10
12.04
22.6
24.5
Chicag
550590019
Annual
No
N/A
N/A
8.04
8.72
20.4
22.1
Chicag
181270024
Annual
No
N/A
N/A
9.51
10.32
22.4
24.3
Chicag
180892004
Annual
No
N/A
N/A
9.84
10.67
24.7
26.8
Chicag
180890031
Annual
No
N/A
N/A
10.12
10.98
23.6
25.6
Chicag
180890026
Annual
No
N/A
N/A
-
-
25.2
27.3
Chicag
180890022
Annual
No
N/A
N/A
-
-
22.7
24.6
Chicag
180890006
Annual
No
N/A
N/A
10.03
10.88
23.1
25.1
Chicag
171971011
Annual
No
N/A
N/A
8.36
9.07
18.4
20.0
Chicag
171971002
Annual
No
N/A
N/A
7.69
8.34
20.0
21.7
Chicag
170890007
Annual
No
N/A
N/A
8.94
9.70
19.2
20.8
Chicag
170890003
Annual
No
N/A
N/A
-
-
19.2
20.8
Chicag
170434002
Annual
No
N/A
N/A
8.87
9.62
19.9
21.6
Chicag
170316005
Annual
No
N/A
N/A
10.79
11.70
24.1
26.1
Chicag
170314201
Annual
No
N/A
N/A
9.00
9.76
21.4
23.2
Chicag
170314007
Annual
No
N/A
N/A
9.49
10.29
-
-
Chicag
170313301
Annual
No
N/A
N/A
10.37
11.25
23.5
25.5
C-99

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Chicag
170310076
Annual
No
N/A
N/A
10.18
11.04
22.5
24.4
Chicag
170310057
Annual
No
N/A
N/A
11.03
11.96
26.8
29.1
Chicag
170310052
Annual
No
N/A
N/A
10.00
10.85
23.3
25.3
Chicag
170310022
Annual
No
N/A
N/A
10.38
11.26
22.4
24.3
Chicag
170310001
Annual
No
N/A
N/A
10.13
10.99
21.7
23.5
Cincin
390610014
Annual
Yes
-72
0
10.70
12.04
22.9
26.1
Cincin
390610042
Annual
No
-72
0
10.29
11.66
22.6
26.2
Cincin
390610040
Annual
No
-72
0
9.45
10.79
21.0
25.4
Cincin
390610010
Annual
No
-72
0
9.43
10.75
21.3
24.4
Cincin
390610006
Annual
No
-72
0
9.46
10.75
20.3
24.3
Cincin
390170020
Annual
No
-72
0
-
-
24.2
27.8
Cincin
390170019
Annual
No
-72
0
10.24
11.40
22.0
24.5
Cincin
390170016
Annual
No
-72
0
9.79
11.06
22.1
25.1
Cincin
210373002
Annual
No
-72
0
9.06
10.42
20.9
25.1
Clevel
390350065
Annual
Yes
6
0
12.17
12.04
24.9
24.7
Clevel
391030004
Annual
No
6
0
8.73
8.61
19.6
19.2
Clevel
390933002
Annual
No
6
0
8.10
7.99
20.2
19.9
Clevel
390850007
Annual
No
6
0
7.88
7.78
17.4
17.1
Clevel
390351002
Annual
No
6
0
8.86
8.74
19.5
19.2
Clevel
390350045
Annual
No
6
0
10.61
10.49
22.9
22.6
Clevel
390350038
Annual
No
6
0
11.38
11.26
25.0
24.7
Clevel
390350034
Annual
No
6
0
8.87
8.75
20.4
20.1
Detroi
261630033
Annual
Yes
-56
0
11.30
12.04
26.8
30.2
Detroi
261630039
Annual
No
-56
0
9.11
9.88
22.3
24.8
Detroi
261630036
Annual
No
-56
0
8.68
9.39
21.8
23.4
Detroi
261630025
Annual
No
-56
0
8.98
9.75
24.1
26.5
Detroi
261630019
Annual
No
-56
0
9.18
9.97
22.4
24.1
Detroi
261630016
Annual
No
-56
0
9.62
10.38
24.4
27.4
Detroi
261630015
Annual
No
-56
0
11.19
11.97
25.5
28.2
Detroi
261630001
Annual
No
-56
0
9.50
10.20
23.3
25.0
Detroi
261470005
Annual
No
-56
0
8.89
9.50
24.3
26.1
Detroi
261250001
Annual
No
-56
0
8.86
9.65
24.2
26.7
Detroi
260990009
Annual
No
-56
0
8.80
9.48
26.2
28.4
ElCent
060250005
Annual
Yes
N/A
N/A
12.63
12.04
33.5
31.9
ElCent
060251003
Annual
No
N/A
N/A
7.44
7.09
19.8
18.9
ElCent
060250007
Annual
No
N/A
N/A
8.37
7.98
21.5
20.5
Elkhar
180390008
Annual
Yes
N/A
N/A
10.24
12.04
28.6
33.6
Evansv
181630023
Annual
Yes
-89
0
10.11
12.03
21.5
32.5
C-100

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Evansv
211010014
Annual
No
-89
0
9.64
11.58
20.7
30.2
Evansv
181630021
Annual
No
-89
0
9.84
11.79
21.6
32.4
Evansv
181630016
Annual
No
-89
0
10.02
11.95
22.0
32.8
Fresno
060190011
24-hr
Yes
N/A
N/A
14.07
10.46
53.8
35.4
Fresno
060195025
24-hr
No
N/A
N/A
13.63
10.13
47.9
31.5
Fresno
060195001
24-hr
No
N/A
N/A
14.08
10.47
49.3
32.4
Fresno
060192009
24-hr
No
N/A
N/A
8.47
6.30
31.3
20.6
Hanfor
060310004
24-hr
Yes
N/A
N/A
21.98
10.81
72.0
35.4
Hanfor
060311004
24-hr
No
N/A
N/A
16.49
8.11
58.9
29.0
Housto
482011035
Annual
Yes
-91
0
11.19
12.04
22.4
25.2
Housto
482011039
Annual
No
-91
0
9.22
10.16
21.7
24.9
Housto
482010058
Annual
No
-91
0
9.67
10.52
22.3
24.8
Housto
481671034
Annual
No
-91
0
7.36
8.27
20.3
23.3
Indian
180970087
Annual
Yes
-24
0
11.44
12.02
25.9
27.5
Indian
180970083
Annual
No
-24
0
11.06
11.64
23.9
25.2
Indian
180970081
Annual
No
-24
0
11.07
11.65
25.0
26.7
Indian
180970078
Annual
No
-24
0
10.14
10.72
24.4
26.2
Indian
180970043
Annual
No
-24
0
-
-
26.0
27.6
Indian
180950011
Annual
No
-24
0
9.05
9.51
21.8
23.1
Indian
180570007
Annual
No
-24
0
9.02
9.52
21.4
22.8
Johnst
420210011
Annual
Yes
-86
0
10.68
12.04
25.8
27.9
Lancas
420710012
Annual
Yes
40
0
12.83
12.03
32.7
31.6
Lancas
420710007
Annual
No
40
0
10.57
9.78
29.8
28.5
LasVeg
320030561
Annual
Yes
N/A
N/A
10.28
12.04
24.5
28.7
LasVeg
320032002
Annual
No
N/A
N/A
9.79
11.47
19.8
23.2
LasVeg
320031019
Annual
No
N/A
N/A
5.18
6.07
11.5
13.5
LasVeg
320030540
Annual
No
N/A
N/A
8.80
10.31
21.7
25.4
Lebano
420750100
Annual
Yes
-61
0
11.20
12.04
31.4
32.4
Little
051191008
Annual
Yes
-98
0
10.27
12.04
21.7
26.7
Little
051190007
Annual
No
-98
0
9.78
11.40
20.5
25.5
Loganll
490050007
24-hr
Yes
-28
0
6.95
7.12
34.0
35.4
LosAng
060371103
Annual
Yes
N/A
N/A
12.38
12.04
32.8
31.9
LosAng
060592022
Annual
No
N/A
N/A
7.48
7.27
15.3
14.9
LosAng
060590007
Annual
No
N/A
N/A
9.63
9.37
-
-
LosAng
060374004
Annual
No
N/A
N/A
10.25
9.97
27.3
26.6
LosAng
060374002
Annual
No
N/A
N/A
11.06
10.76
29.2
28.4
LosAng
060371602
Annual
No
N/A
N/A
11.86
11.53
32.3
31.4
LosAng
060371302
Annual
No
N/A
N/A
11.99
11.66
31.5
30.6
C-101

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
LosAng
060371201
Annual
No
N/A
N/A
9.46
9.20
25.6
24.9
LosAng
060370002
Annual
No
N/A
N/A
10.52
10.23
29.2
28.4
Louisv
180190006
Annual
Yes
-65
0
10.64
12.04
23.9
28.4
Louisv
211110075
Annual
No
-65
0
10.42
11.76
22.3
26.4
Louisv
211110067
Annual
No
-65
0
9.55
10.84
21.4
25.4
Louisv
211110051
Annual
No
-65
0
10.29
11.67
21.8
25.9
Louisv
211110043
Annual
No
-65
0
10.37
11.71
22.0
26.1
Louisv
180431004
Annual
No
-65
0
9.96
11.32
22.0
25.8
Louisv
180190008
Annual
No
-65
0
8.72
10.07
20.1
24.3
MaconG
130210007
Annual
Yes
N/A
N/A
10.13
12.04
21.2
25.2
MaconG
130210012
Annual
No
N/A
N/A
7.68
9.13
16.6
19.7
Madera
060392010
24-hr
Yes
N/A
N/A
13.30
11.15
45.1
35.4
McAile
482150043
Annual
Yes
N/A
N/A
10.09
12.04
25.0
29.8
Merced
060472510
24-hr
Yes
32
0
11.68
10.79
39.8
35.4
Merced
060470003
24-hr
No
32
0
11.81
10.89
39.0
34.1
Modest
060990006
24-hr
Yes
N/A
N/A
13.02
10.82
45.7
35.4
Modest
060990005
24-hr
No
N/A
N/A
-
-
38.8
30.1
NapaCA
060550003
Annual
Yes
N/A
N/A
10.36
12.04
25.1
29.2
NewYor
360610128
Annual
Yes
N/A
N/A
10.20
12.04
23.9
28.2
NewYor
361030002
Annual
No
N/A
N/A
7.18
8.48
18.8
22.2
NewYor
360810124
Annual
No
N/A
N/A
7.52
8.88
19.5
23.0
NewYor
360710002
Annual
No
N/A
N/A
6.95
8.20
17.5
20.7
NewYor
360610134
Annual
No
N/A
N/A
9.70
11.45
21.6
25.5
NewYor
360610079
Annual
No
N/A
N/A
8.42
9.94
22.8
26.9
NewYor
360470122
Annual
No
N/A
N/A
8.66
10.22
20.5
24.2
NewYor
360050133
Annual
No
N/A
N/A
9.05
10.68
24.0
28.3
NewYor
360050110
Annual
No
N/A
N/A
7.39
8.72
19.4
22.9
NewYor
340392003
Annual
No
N/A
N/A
8.59
10.14
23.6
27.9
NewYor
340390004
Annual
No
N/A
N/A
9.87
11.65
24.2
28.6
NewYor
340310005
Annual
No
N/A
N/A
8.42
9.94
22.2
26.2
NewYor
340292002
Annual
No
N/A
N/A
7.23
8.53
18.1
21.4
NewYor
340273001
Annual
No
N/A
N/A
6.78
8.00
17.1
20.2
NewYor
340171003
Annual
No
N/A
N/A
8.79
10.38
23.4
27.6
NewYor
340130003
Annual
No
N/A
N/A
8.89
10.49
23.8
28.1
NewYor
340030003
Annual
No
N/A
N/A
8.90
10.51
24.5
28.9
OgdenC
490110004
24-hr
Yes
-53
0
7.28
7.65
32.6
35.4
OgdenC
490570002
24-hr
No
-53
0
8.99
9.37
-
-
OgdenC
490030003
24-hr
No
-53
0
6.35
6.70
-
-
C-102

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Philad
420450002
Annual
Yes
-75
0
11.46
12.04
26.0
27.4
Philad
421010057
Annual
No
-75
0
10.86
11.54
27.0
28.1
Philad
421010055
Annual
No
-75
0
11.43
12.03
27.5
28.8
Philad
421010048
Annual
No
-75
0
10.27
10.91
25.6
27.4
Philad
420290100
Annual
No
-75
0
9.64
10.38
23.9
25.2
Philad
340150004
Annual
No
-75
0
8.33
8.94
20.6
23.2
Philad
340071007
Annual
No
-75
0
8.84
9.51
21.0
21.9
Philad
340070002
Annual
No
-75
0
10.19
10.95
23.5
24.6
Philad
240150003
Annual
No
-75
0
8.70
9.47
22.6
23.7
Philad
100031012
Annual
No
-75
0
9.04
9.81
23.0
23.6
Pittsb
420030064
Annual
Yes
30
0
12.82
12.02
35.8
34.8
Pittsb
421290008
Annual
No
30
0
8.65
8.06
19.6
18.0
Pittsb
421255001
Annual
No
30
0
8.35
7.78
17.8
16.4
Pittsb
421250200
Annual
No
30
0
8.95
8.32
19.3
18.2
Pittsb
421250005
Annual
No
30
0
11.02
10.30
22.7
21.7
Pittsb
420070014
Annual
No
30
0
10.11
9.52
21.9
20.6
Pittsb
420050001
Annual
No
30
0
11.03
10.45
21.9
20.4
Pittsb
420031301
Annual
No
30
0
11.00
10.28
24.8
23.6
Pittsb
420031008
Annual
No
30
0
9.78
9.20
20.5
19.0
Pittsb
420030008
Annual
No
30
0
9.50
8.89
20.5
19.2
Prinev
410130100
24-hr
Yes
N/A
N/A
8.60
8.10
37.6
35.4
ProvoO
490494001
24-hr
Yes
-76
0
7.74
8.29
30.9
35.4
ProvoO
490495010
24-hr
No
-76
0
6.73
7.21
-
-
ProvoO
490490002
24-hr
No
-76
0
7.41
7.95
28.9
33.2
Rivers
060658005
24-hr
Yes
N/A
N/A
14.48
11.87
43.2
35.4
Rivers
060658001
24-hr
No
N/A
N/A
-
-
36.5
29.9
Sacram
060670006
24-hr
Yes
-99
0
9.31
10.04
31.4
35.3
Sacram
061131003
24-hr
No
-99
0
6.62
7.08
15.8
19.0
Sacram
060670012
24-hr
No
-99
0
7.30
7.85
19.8
21.3
Sacram
060670010
24-hr
No
-99
0
8.67
9.30
26.5
30.2
Sacram
060610006
24-hr
No
-99
0
7.58
8.08
20.3
22.2
Sacram
060610003
24-hr
No
-99
0
6.71
7.04
19.3
20.7
SaltLa
490353010
24-hr
Yes
58
0
-
-
41.5
35.4
SaltLa
490353006
24-hr
No
58
0
7.62
6.91
36.8
31.5
SaltLa
490351001
24-hr
No
58
0
7.07
6.30
32.1
25.8
SanLui
060792007
Annual
Yes
N/A
N/A
10.70
12.04
25.9
29.1
SanLui
060798002
Annual
No
N/A
N/A
5.71
6.43
-
-
SanLui
060792004
Annual
No
N/A
N/A
8.25
9.28
19.8
22.3
C-103

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
SouthB
181410015
Annual
Yes
-92
0
10.45
12.04
32.5
34.8
St. Lou
290990019
Annual
Yes
N/A
N/A
10.12
12.04
22.8
27.1
St. Lou
295100094
Annual
No
N/A
N/A
9.57
11.39
23.3
27.7
St. Lou
295100093
Annual
No
N/A
N/A
-
-
23.7
28.2
St. Lou
295100085
Annual
No
N/A
N/A
10.10
12.02
23.6
28.1
St. Lou
295100007
Annual
No
N/A
N/A
9.78
11.64
23.7
28.2
St. Lou
291893001
Annual
No
N/A
N/A
9.85
11.72
22.4
26.6
Stockt
060771002
24-hr
Yes
42
0
12.23
11.41
38.7
35.4
Stockt
060772010
24-hr
No
42
0
10.74
9.96
37.3
34.3
Visali
061072002
24-hr
Yes
N/A
N/A
16.23
10.64
54.0
35.4
Weirto
390810017
Annual
Yes
-14
0
11.75
12.03
27.2
27.5
Weirto
540090011
Annual
No
-14
0
9.75
10.02
22.8
23.6
Weirto
540090005
Annual
No
-14
0
10.52
10.80
22.4
23.1
Weirto
390810021
Annual
No
-14
0
9.29
9.55
22.2
22.8
Wheeli
540511002
Annual
Yes
N/A
N/A
10.24
12.04
22.5
26.5
Wheeli
540690010
Annual
No
N/A
N/A
9.61
11.30
19.7
23.2
a CBSA names are the first six characters of the full CBSAs names in Table C-3.
b Percent reduction in NOx and SO2 emissions associated with just meeting the standard in this case; N/A indicates 'not
applicable' where proportional projection was used.
c Percent reduction in Primary PM2.5 emissions associated with just meeting the standard in this case; N/A indicates 'not
applicable' where proportional projection was used.
C-104

-------
Table C-35. PM2.5 DVs for the Primary PM projection case and 10/30 standard level.
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m 3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
AkronO
391530017
Annual
Yes
0
17
10.99
10.03
23.7
22.6
AkronO
391530023
Annual
No
0
17
9.16
8.46
20.2
19.1
Altoon
420130801
Annual
Yes
0
2
10.11
10.02
23.8
23.5
Atlant
131210039
Annual
Yes
0
6
10.38
10.01
19.7
19.0
Atlant
132230003
Annual
No
0
6
7.82
7.64
16.2
15.9
Atlant
131350002
Annual
No
0
6
8.84
8.57
17.9
17.3
Atlant
130890002
Annual
No
0
6
9.34
9.04
19.2
18.7
Atlant
130670003
Annual
No
0
6
9.51
9.22
18.6
18.2
Atlant
130630091
Annual
No
0
6
9.86
9.56
19.1
18.5
Bakers
060290016
Annual
Yes
91
100
18.45
10.01
61.3
29.1
Bakers
060290015
Annual
No
91
100
5.15
3.66
15.8
13.6
Bakers
060290014
Annual
No
91
100
16.53
8.37
61.4
26.0
Bakers
060290011
Annual
No
91
100
6.06
4.58
19.6
15.9
Bakers
060290010
Annual
No
91
100
16.52
8.87
70.0
27.9
Birmin
010732059
Annual
Yes
0
16
11.25
10.03
22.3
19.8
Birmin
010732003
Annual
No
0
16
10.08
9.06
19.0
17.2
Birmin
010731010
Annual
No
0
16
9.78
8.94
19.2
17.7
Birmin
010730023
Annual
No
0
16
10.94
9.77
22.8
20.6
Canton
391510017
Annual
Yes
0
15
10.81
10.01
23.7
22.6
Canton
391510020
Annual
No
0
15
9.91
9.21
22.0
21.0
Chicag
170313103
Annual
Yes
0
18
11.10
10.01
22.6
21.0
Chicag
550590019
Annual
No
0
18
8.04
7.42
20.4
18.8
Chicag
181270024
Annual
No
0
18
9.51
8.55
22.4
20.4
Chicag
180892004
Annual
No
0
18
9.84
8.78
24.7
22.8
Chicag
180890031
Annual
No
0
18
10.12
9.05
23.6
21.1
Chicag
180890026
Annual
No
0
18
-
-
25.2
22.8
Chicag
180890022
Annual
No
0
18
-
-
22.7
20.4
Chicag
180890006
Annual
No
0
18
10.03
8.93
23.1
20.5
Chicag
171971011
Annual
No
0
18
8.36
7.78
18.4
17.4
Chicag
171971002
Annual
No
0
18
7.69
7.04
20.0
18.7
Chicag
170890007
Annual
No
0
18
8.94
8.21
19.2
17.8
Chicag
170890003
Annual
No
0
18
-
-
19.2
18.1
Chicag
170434002
Annual
No
0
18
8.87
8.13
19.9
18.9
Chicag
170316005
Annual
No
0
18
10.79
9.73
24.1
21.7
Chicag
170314201
Annual
No
0
18
9.00
8.25
21.4
19.9
Chicag
170314007
Annual
No
0
18
9.49
8.66
-
-
Chicag
170313301
Annual
No
0
18
10.37
9.38
23.5
21.3
C-105

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Chicag
170310076
Annual
No
0
18
10.18
9.24
22.5
20.7
Chicag
170310057
Annual
No
0
18
11.03
9.99
26.8
25.1
Chicag
170310052
Annual
No
0
18
10.00
9.06
23.3
21.4
Chicag
170310022
Annual
No
0
18
10.38
9.28
22.4
20.9
Chicag
170310001
Annual
No
0
18
10.13
9.22
21.7
19.7
Cincin
390610014
Annual
Yes
0
12
10.70
10.04
22.9
21.8
Cincin
390610042
Annual
No
0
12
10.29
9.69
22.6
21.6
Cincin
390610040
Annual
No
0
12
9.45
8.91
21.0
20.0
Cincin
390610010
Annual
No
0
12
9.43
8.93
21.3
20.5
Cincin
390610006
Annual
No
0
12
9.46
8.91
20.3
19.5
Cincin
390170020
Annual
No
0
12
-
-
24.2
23.3
Cincin
390170019
Annual
No
0
12
10.24
9.60
22.0
21.1
Cincin
390170016
Annual
No
0
12
9.79
9.22
22.1
21.2
Cincin
210373002
Annual
No
0
12
9.06
8.58
20.9
20.0
Clevel
390350065
Annual
Yes
0
33
12.17
10.00
24.9
21.3
Clevel
391030004
Annual
No
0
33
8.73
7.57
19.6
17.8
Clevel
390933002
Annual
No
0
33
8.10
6.95
20.2
18.7
Clevel
390850007
Annual
No
0
33
7.88
6.84
17.4
15.4
Clevel
390351002
Annual
No
0
33
8.86
7.64
19.5
17.5
Clevel
390350045
Annual
No
0
33
10.61
8.84
22.9
20.1
Clevel
390350038
Annual
No
0
33
11.38
9.37
25.0
22.0
Clevel
390350034
Annual
No
0
33
8.87
7.58
20.4
18.2
Detroi
261630033
Annual
Yes
0
26
11.30
10.00
26.8
24.9
Detroi
261630039
Annual
No
0
26
9.11
8.21
22.3
20.3
Detroi
261630036
Annual
No
0
26
8.68
7.88
21.8
19.8
Detroi
261630025
Annual
No
0
26
8.98
7.99
24.1
21.7
Detroi
261630019
Annual
No
0
26
9.18
8.18
22.4
19.7
Detroi
261630016
Annual
No
0
26
9.62
8.63
24.4
22.6
Detroi
261630015
Annual
No
0
26
11.19
9.94
25.5
22.8
Detroi
261630001
Annual
No
0
26
9.50
8.39
23.3
20.4
Detroi
261470005
Annual
No
0
26
8.89
8.11
24.3
22.4
Detroi
261250001
Annual
No
0
26
8.86
7.90
24.2
22.2
Detroi
260990009
Annual
No
0
26
8.80
7.94
26.2
23.8
ElCent
060250005
Annual
Yes
0
50
12.63
10.01
33.5
25.0
ElCent
060251003
Annual
No
0
50
7.44
5.67
19.8
14.6
ElCent
060250007
Annual
No
0
50
8.37
6.80
21.5
18.5
Elkhar
180390008
Annual
Yes
0
6
10.24
10.01
28.6
27.8
Evansv
181630023
Annual
Yes
0
2
10.11
10.02
21.5
21.5
C-106

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Evansv
211010014
Annual
No
0
2
9.64
9.56
20.7
20.7
Evansv
181630021
Annual
No
0
2
9.84
9.76
21.6
21.5
Evansv
181630016
Annual
No
0
2
10.02
9.94
22.0
21.9
Fresno
060195001
24-hr
Yes
0
100
14.08
9.49
49.3
30.3
Fresno
060195025
24-hr
No
0
100
13.63
8.41
47.9
26.4
Fresno
060192009
24-hr
No
0
100
8.47
6.74
31.3
22.2
Fresno
060190011
24-hr
No
0
100
14.07
8.27
53.8
27.1
Hanfor
060310004
Annual
Yes
82
98
21.98
10.00
72.0
29.5
Hanfor
060311004
Annual
No
82
98
16.49
8.36
58.9
25.2
Housto
482011035
Annual
Yes
0
19
11.19
10.01
22.4
20.2
Housto
482011039
Annual
No
0
19
9.22
8.40
21.7
19.6
Housto
482010058
Annual
No
0
19
9.67
8.70
22.3
20.3
Housto
481671034
Annual
No
0
19
7.36
7.07
20.3
19.6
Indian
180970087
Annual
Yes
0
25
11.44
10.01
25.9
24.2
Indian
180970083
Annual
No
0
25
11.06
9.72
23.9
22.5
Indian
180970081
Annual
No
0
25
11.07
9.71
25.0
23.4
Indian
180970078
Annual
No
0
25
10.14
8.97
24.4
22.8
Indian
180970043
Annual
No
0
25
-
-
26.0
24.6
Indian
180950011
Annual
No
0
25
9.05
8.17
21.8
20.7
Indian
180570007
Annual
No
0
25
9.02
8.07
21.4
20.0
Johnst
420210011
Annual
Yes
0
12
10.68
10.02
25.8
23.5
Lancas
420710012
Annual
Yes
0
41
12.83
9.98
32.7
25.5
Lancas
420710007
Annual
No
0
41
10.57
8.20
29.8
22.0
LasVeg
320030561
Annual
Yes
0
4
10.28
9.97
24.5
23.6
LasVeg
320032002
Annual
No
0
4
9.79
9.50
19.8
19.2
LasVeg
320031019
Annual
No
0
4
5.18
5.08
11.5
11.3
LasVeg
320030540
Annual
No
0
4
8.80
8.55
21.7
20.9
Lebano
420750100
Annual
Yes
0
21
11.20
10.04
31.4
28.0
Little
051191008
Annual
Yes
0
6
10.27
10.00
21.7
21.3
Little
051190007
Annual
No
0
6
9.78
9.48
20.5
20.1
Loganll
490050007
24-hr
Yes
0
19
6.95
6.40
34.0
30.3
LosAng
060371103
Annual
Yes
0
34
12.38
9.99
32.8
27.8
LosAng
060592022
Annual
No
0
34
7.48
6.43
15.3
13.3
LosAng
060590007
Annual
No
0
34
9.63
7.84
-
-
LosAng
060374004
Annual
No
0
34
10.25
8.36
27.3
23.7
LosAng
060374002
Annual
No
0
34
11.06
9.02
29.2
24.9
LosAng
060371602
Annual
No
0
34
11.86
9.55
32.3
26.5
LosAng
060371302
Annual
No
0
34
11.99
9.64
31.5
27.0
C-107

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
LosAng
060371201
Annual
No
0
34
9.46
7.93
25.6
21.6
LosAng
060370002
Annual
No
0
34
10.52
8.81
29.2
25.0
Louisv
180190006
Annual
Yes
0
12
10.64
10.01
23.9
22.8
Louisv
211110075
Annual
No
0
12
10.42
9.79
22.3
21.4
Louisv
211110067
Annual
No
0
12
9.55
8.99
21.4
20.5
Louisv
211110051
Annual
No
0
12
10.29
9.76
21.8
21.2
Louisv
211110043
Annual
No
0
12
10.37
9.77
22.0
21.2
Louisv
180431004
Annual
No
0
12
9.96
9.41
22.0
21.0
Louisv
180190008
Annual
No
0
12
8.72
8.29
20.1
19.5
MaconG
130210007
Annual
Yes
0
2
10.13
10.03
21.2
21.0
MaconG
130210012
Annual
No
0
2
7.68
7.61
16.6
16.5
Madera
060392010
24-hr
Yes
0
84
13.30
9.89
45.1
30.4
McAile
482150043
Annual
Yes
0
2
10.09
10.03
25.0
24.9
Merced
060470003
24-hr
Yes
0
65
11.81
9.87
39.0
30.4
Merced
060472510
24-hr
No
0
65
11.68
9.11
39.8
28.8
Modest
060990006
24-hr
Yes
0
77
13.02
9.52
45.7
30.3
Modest
060990005
24-hr
No
0
77
-
-
38.8
29.2
NapaCA
060550003
Annual
Yes
0
9
10.36
10.04
25.1
24.6
NewYor
360610128
Annual
Yes
0
3
10.20
9.99
23.9
23.5
NewYor
361030002
Annual
No
0
3
7.18
7.07
18.8
18.6
NewYor
360810124
Annual
No
0
3
7.52
7.39
19.5
19.1
NewYor
360710002
Annual
No
0
3
6.95
6.84
17.5
17.2
NewYor
360610134
Annual
No
0
3
9.70
9.51
21.6
21.2
NewYor
360610079
Annual
No
0
3
8.42
8.26
22.8
22.5
NewYor
360470122
Annual
No
0
3
8.66
8.49
20.5
20.2
NewYor
360050133
Annual
No
0
3
9.05
8.87
24.0
23.6
NewYor
360050110
Annual
No
0
3
7.39
7.25
19.4
19.1
NewYor
340392003
Annual
No
0
3
8.59
8.44
23.6
23.2
NewYor
340390004
Annual
No
0
3
9.87
9.69
24.2
23.8
NewYor
340310005
Annual
No
0
3
8.42
8.28
22.2
21.9
NewYor
340292002
Annual
No
0
3
7.23
7.13
18.1
17.9
NewYor
340273001
Annual
No
0
3
6.78
6.69
17.1
16.9
NewYor
340171003
Annual
No
0
3
8.79
8.64
23.4
22.9
NewYor
340130003
Annual
No
0
3
8.89
8.73
23.8
23.4
NewYor
340030003
Annual
No
0
3
8.90
8.75
24.5
24.1
OgdenC
490110004
24-hr
Yes
0
15
7.28
6.89
32.6
30.3
OgdenC
490570002
24-hr
No
0
15
8.99
8.39
-
-
OgdenC
490030003
24-hr
No
0
15
6.35
6.02
-
-
C-108

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Philad
420450002
Annual
Yes
0
20
11.46
9.99
26.0
22.9
Philad
421010057
Annual
No
0
20
10.86
9.56
27.0
23.4
Philad
421010055
Annual
No
0
20
11.43
9.94
27.5
24.2
Philad
421010048
Annual
No
0
20
10.27
9.00
25.6
22.7
Philad
420290100
Annual
No
0
20
9.64
8.66
23.9
21.2
Philad
340150004
Annual
No
0
20
8.33
7.43
20.6
18.2
Philad
340071007
Annual
No
0
20
8.84
7.86
21.0
18.8
Philad
340070002
Annual
No
0
20
10.19
9.11
23.5
20.6
Philad
240150003
Annual
No
0
20
8.70
7.90
22.6
20.5
Philad
100031012
Annual
No
0
20
9.04
8.15
23.0
21.1
Pittsb
420030064
Annual
Yes
0
44
12.82
10.04
35.8
26.2
Pittsb
421290008
Annual
No
0
44
8.65
6.96
19.6
16.9
Pittsb
421255001
Annual
No
0
44
8.35
6.78
17.8
15.7
Pittsb
421250200
Annual
No
0
44
8.95
7.22
19.3
15.7
Pittsb
421250005
Annual
No
0
44
11.02
8.85
22.7
18.0
Pittsb
420070014
Annual
No
0
44
10.11
7.98
21.9
17.5
Pittsb
420050001
Annual
No
0
44
11.03
8.58
21.9
17.8
Pittsb
420031301
Annual
No
0
44
11.00
8.64
24.8
18.7
Pittsb
420031008
Annual
No
0
44
9.78
7.68
20.5
16.1
Pittsb
420030008
Annual
No
0
44
9.50
7.30
20.5
16.3
Prinev
410130100
24-hr
Yes
0
33
8.60
7.19
37.6
30.4
ProvoO
490494001
24-hr
Yes
0
3
7.74
7.65
30.9
30.4
ProvoO
490495010
24-hr
No
0
3
6.73
6.65
-
-
ProvoO
490490002
24-hr
No
0
3
7.41
7.32
28.9
28.4
Rivers
060658005
24-hr
Yes
0
58
14.48
9.69
43.2
30.4
Rivers
060658001
24-hr
No
0
58
-
-
36.5
25.4
Sacram
060670006
24-hr
Yes
0
6
9.31
9.02
31.4
30.4
Sacram
061131003
24-hr
No
0
6
6.62
6.47
15.8
15.4
Sacram
060670012
24-hr
No
0
6
7.30
7.11
19.8
19.4
Sacram
060670010
24-hr
No
0
6
8.67
8.41
26.5
25.7
Sacram
060610006
24-hr
No
0
6
7.58
7.34
20.3
19.9
Sacram
060610003
24-hr
No
0
6
6.71
6.56
19.3
19.0
SaltLa
490353010
24-hr
Yes
0
85
-
-
41.5
30.4
SaltLa
490353006
24-hr
No
0
85
7.62
4.85
36.8
23.8
SaltLa
490351001
24-hr
No
0
85
7.07
4.72
32.1
21.0
SanLui
060792007
Annual
Yes
0
22
10.70
10.04
25.9
24.9
SanLui
060798002
Annual
No
0
22
5.71
5.42
-
-
SanLui
060792004
Annual
No
0
22
8.25
7.76
19.8
19.2
C-109

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
SouthB
181410015
24-hr
Yes
0
18
10.45
9.72
32.5
30.3
St. Lou
290990019
Annual
Yes
0
2
10.12
10.02
22.8
22.7
St. Lou
295100094
Annual
No
0
2
9.57
9.48
23.3
23.2
St. Lou
295100093
Annual
No
0
2
-
-
23.7
23.5
St. Lou
295100085
Annual
No
0
2
10.10
10.00
23.6
23.4
St. Lou
295100007
Annual
No
0
2
9.78
9.69
23.7
23.6
St. Lou
291893001
Annual
No
0
2
9.85
9.76
22.4
22.3
Stockt
060771002
24-hr
Yes
0
43
12.23
9.86
38.7
30.3
Stockt
060772010
24-hr
No
0
43
10.74
8.75
37.3
29.6
Visali
061072002
24-hr
Yes
58
74
16.23
9.67
54.0
30.4
Weirto
390810017
Annual
Yes
0
33
11.75
10.00
27.2
22.6
Weirto
540090011
Annual
No
0
33
9.75
8.42
22.8
19.8
Weirto
540090005
Annual
No
0
33
10.52
9.07
22.4
19.8
Weirto
390810021
Annual
No
0
33
9.29
8.06
22.2
19.3
Wheeli
540511002
Annual
Yes
0
5
10.24
10.03
22.5
22.1
Wheeli
540690010
Annual
No
0
5
9.61
9.42
19.7
19.4
a CBSA names are the first six characters of the full CBSAs names in Table C-3.
b Percent reduction in NOx and SO2 emissions associated with just meeting the standard in this case.
c Percent reduction in Primary PM2.5 emissions associated with just meeting the standard in this case.
C-110

-------
Table C-36. PM2.5 DVs for the Secondary PM projection case and 10/30 standard level.
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m 3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
AkronO
391530017
Annual
Yes
45
0
10.99
10.04
23.7
20.8
AkronO
391530023
Annual
No
45
0
9.16
8.24
20.2
17.7
Altoon
420130801
Annual
Yes
N/A
N/A
10.11
10.04
23.8
23.6
Atlant
131210039
Annual
Yes
N/A
N/A
10.38
10.04
19.7
19.1
Atlant
132230003
Annual
No
N/A
N/A
7.82
7.56
16.2
15.7
Atlant
131350002
Annual
No
N/A
N/A
8.84
8.55
17.9
17.3
Atlant
130890002
Annual
No
N/A
N/A
9.34
9.03
19.2
18.6
Atlant
130670003
Annual
No
N/A
N/A
9.51
9.20
18.6
18.0
Atlant
130630091
Annual
No
N/A
N/A
9.86
9.54
19.1
18.5
Bakers
060290010
24-hr
Yes
N/A
N/A
16.52
8.99
70.0
30.4
Bakers
060290016
24-hr
No
N/A
N/A
18.45
10.04
61.3
26.6
Bakers
060290015
24-hr
No
N/A
N/A
5.15
2.80
15.8
6.9
Bakers
060290014
24-hr
No
N/A
N/A
16.53
9.00
61.4
26.7
Bakers
060290011
24-hr
No
N/A
N/A
6.06
3.30
19.6
8.5
Birmin
010732059
Annual
Yes
71
0
11.25
10.04
22.3
20.2
Birmin
010732003
Annual
No
71
0
10.08
8.86
19.0
16.1
Birmin
010731010
Annual
No
71
0
9.78
8.39
19.2
16.6
Birmin
010730023
Annual
No
71
0
10.94
9.72
22.8
20.3
Canton
391510017
Annual
Yes
36
0
10.81
10.04
23.7
21.7
Canton
391510020
Annual
No
36
0
9.91
9.13
22.0
19.4
Chicag
170313103
Annual
Yes
N/A
N/A
11.10
10.04
22.6
20.4
Chicag
550590019
Annual
No
N/A
N/A
8.04
7.27
20.4
18.5
Chicag
181270024
Annual
No
N/A
N/A
9.51
8.60
22.4
20.3
Chicag
180892004
Annual
No
N/A
N/A
9.84
8.90
24.7
22.3
Chicag
180890031
Annual
No
N/A
N/A
10.12
9.15
23.6
21.3
Chicag
180890026
Annual
No
N/A
N/A
-
-
25.2
22.8
Chicag
180890022
Annual
No
N/A
N/A
-
-
22.7
20.5
Chicag
180890006
Annual
No
N/A
N/A
10.03
9.07
23.1
20.9
Chicag
171971011
Annual
No
N/A
N/A
8.36
7.56
18.4
16.6
Chicag
171971002
Annual
No
N/A
N/A
7.69
6.96
20.0
18.1
Chicag
170890007
Annual
No
N/A
N/A
8.94
8.09
19.2
17.4
Chicag
170890003
Annual
No
N/A
N/A
-
-
19.2
17.4
Chicag
170434002
Annual
No
N/A
N/A
8.87
8.02
19.9
18.0
Chicag
170316005
Annual
No
N/A
N/A
10.79
9.76
24.1
21.8
Chicag
170314201
Annual
No
N/A
N/A
9.00
8.14
21.4
19.4
Chicag
170314007
Annual
No
N/A
N/A
9.49
8.58
-
-
Chicag
170313301
Annual
No
N/A
N/A
10.37
9.38
23.5
21.3
C-lll

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Chicag
170310076
Annual
No
N/A
N/A
10.18
9.21
22.5
20.4
Chicag
170310057
Annual
No
N/A
N/A
11.03
9.98
26.8
24.2
Chicag
170310052
Annual
No
N/A
N/A
10.00
9.05
23.3
21.1
Chicag
170310022
Annual
No
N/A
N/A
10.38
9.39
22.4
20.3
Chicag
170310001
Annual
No
N/A
N/A
10.13
9.16
21.7
19.6
Cincin
390610014
Annual
Yes
28
0
10.70
10.03
22.9
21.2
Cincin
390610042
Annual
No
28
0
10.29
9.61
22.6
20.8
Cincin
390610040
Annual
No
28
0
9.45
8.78
21.0
19.0
Cincin
390610010
Annual
No
28
0
9.43
8.78
21.3
19.6
Cincin
390610006
Annual
No
28
0
9.46
8.82
20.3
18.4
Cincin
390170020
Annual
No
28
0
-
-
24.2
22.5
Cincin
390170019
Annual
No
28
0
10.24
9.66
22.0
20.6
Cincin
390170016
Annual
No
28
0
9.79
9.16
22.1
20.1
Cincin
210373002
Annual
No
28
0
9.06
8.38
20.9
18.9
Clevel
390350065
Annual
Yes
79
0
12.17
10.04
24.9
20.5
Clevel
391030004
Annual
No
79
0
8.73
6.75
19.6
13.9
Clevel
390933002
Annual
No
79
0
8.10
6.28
20.2
13.8
Clevel
390850007
Annual
No
79
0
7.88
6.10
17.4
12.9
Clevel
390351002
Annual
No
79
0
8.86
6.81
19.5
14.4
Clevel
390350045
Annual
No
79
0
10.61
8.50
22.9
17.0
Clevel
390350038
Annual
No
79
0
11.38
9.33
25.0
19.7
Clevel
390350034
Annual
No
79
0
8.87
6.90
20.4
15.4
Detroi
261630033
Annual
Yes
60
0
11.30
10.03
26.8
24.3
Detroi
261630039
Annual
No
60
0
9.11
7.82
22.3
18.8
Detroi
261630036
Annual
No
60
0
8.68
7.43
21.8
19.1
Detroi
261630025
Annual
No
60
0
8.98
7.63
24.1
19.1
Detroi
261630019
Annual
No
60
0
9.18
7.83
22.4
20.3
Detroi
261630016
Annual
No
60
0
9.62
8.33
24.4
21.3
Detroi
261630015
Annual
No
60
0
11.19
9.88
25.5
22.0
Detroi
261630001
Annual
No
60
0
9.50
8.26
23.3
20.1
Detroi
261470005
Annual
No
60
0
8.89
7.81
24.3
20.6
Detroi
261250001
Annual
No
60
0
8.86
7.49
24.2
20.5
Detroi
260990009
Annual
No
60
0
8.80
7.57
26.2
21.8
ElCent
060250005
Annual
Yes
N/A
N/A
12.63
10.04
33.5
26.6
ElCent
060251003
Annual
No
N/A
N/A
7.44
5.91
19.8
15.7
ElCent
060250007
Annual
No
N/A
N/A
8.37
6.65
21.5
17.1
Elkhar
180390008
Annual
Yes
N/A
N/A
10.24
10.04
28.6
28.0
Evansv
181630023
Annual
Yes
3
0
10.11
10.03
21.5
21.2
C-112

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Evansv
211010014
Annual
No
3
0
9.64
9.56
20.7
20.3
Evansv
181630021
Annual
No
3
0
9.84
9.76
21.6
21.2
Evansv
181630016
Annual
No
3
0
10.02
9.95
22.0
21.7
Fresno
060190011
24-hr
Yes
N/A
N/A
14.07
9.48
53.8
30.4
Fresno
060195025
24-hr
No
N/A
N/A
13.63
9.18
47.9
27.1
Fresno
060195001
24-hr
No
N/A
N/A
14.08
9.49
49.3
27.9
Fresno
060192009
24-hr
No
N/A
N/A
8.47
5.71
31.3
17.7
Hanfor
060310004
24-hr
Yes
N/A
N/A
21.98
9.28
72.0
30.4
Hanfor
060311004
24-hr
No
N/A
N/A
16.49
6.96
58.9
24.9
Housto
482011035
Annual
Yes
84
0
11.19
10.04
22.4
19.6
Housto
482011039
Annual
No
84
0
9.22
8.09
21.7
18.7
Housto
482010058
Annual
No
84
0
9.67
8.57
22.3
19.1
Housto
481671034
Annual
No
84
0
7.36
6.29
20.3
17.8
Indian
180970087
Annual
Yes
48
0
11.44
10.03
25.9
21.8
Indian
180970083
Annual
No
48
0
11.06
9.64
23.9
21.4
Indian
180970081
Annual
No
48
0
11.07
9.66
25.0
20.8
Indian
180970078
Annual
No
48
0
10.14
8.73
24.4
19.9
Indian
180970043
Annual
No
48
0
-
-
26.0
20.9
Indian
180950011
Annual
No
48
0
9.05
7.86
21.8
18.3
Indian
180570007
Annual
No
48
0
9.02
7.75
21.4
17.8
Johnst
420210011
Annual
Yes
31
0
10.68
10.04
25.8
25.1
Lancas
420710012
Annual
Yes
98
0
12.83
10.01
32.7
26.2
Lancas
420710007
Annual
No
98
0
10.57
7.81
29.8
23.4
LasVeg
320030561
Annual
Yes
N/A
N/A
10.28
10.04
24.5
23.9
LasVeg
320032002
Annual
No
N/A
N/A
9.79
9.56
19.8
19.3
LasVeg
320031019
Annual
No
N/A
N/A
5.18
5.06
11.5
11.2
LasVeg
320030540
Annual
No
N/A
N/A
8.80
8.59
21.7
21.2
Lebano
420750100
Annual
Yes
53
0
11.20
10.03
31.4
28.6
Little
051191008
Annual
Yes
11
0
10.27
10.04
21.7
21.1
Little
051190007
Annual
No
11
0
9.78
9.57
20.5
19.9
Loganll
490050007
24-hr
Yes
56
0
6.95
6.51
34.0
30.4
LosAng
060371103
Annual
Yes
N/A
N/A
12.38
10.04
32.8
26.6
LosAng
060592022
Annual
No
N/A
N/A
7.48
6.07
15.3
12.4
LosAng
060590007
Annual
No
N/A
N/A
9.63
7.81
-
-
LosAng
060374004
Annual
No
N/A
N/A
10.25
8.31
27.3
22.1
LosAng
060374002
Annual
No
N/A
N/A
11.06
8.97
29.2
23.7
LosAng
060371602
Annual
No
N/A
N/A
11.86
9.62
32.3
26.2
LosAng
060371302
Annual
No
N/A
N/A
11.99
9.72
31.5
25.5
C-113

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
LosAng
060371201
Annual
No
N/A
N/A
9.46
7.67
25.6
20.8
LosAng
060370002
Annual
No
N/A
N/A
10.52
8.53
29.2
23.7
Louisv
180190006
Annual
Yes
24
0
10.64
10.02
23.9
22.0
Louisv
211110075
Annual
No
24
0
10.42
9.83
22.3
20.3
Louisv
211110067
Annual
No
24
0
9.55
8.96
21.4
19.9
Louisv
211110051
Annual
No
24
0
10.29
9.68
21.8
20.2
Louisv
211110043
Annual
No
24
0
10.37
9.77
22.0
20.2
Louisv
180431004
Annual
No
24
0
9.96
9.37
22.0
20.4
Louisv
180190008
Annual
No
24
0
8.72
8.13
20.1
18.3
MaconG
130210007
Annual
Yes
N/A
N/A
10.13
10.04
21.2
21.0
MaconG
130210012
Annual
No
N/A
N/A
7.68
7.61
16.6
16.5
Madera
060392010
24-hr
Yes
N/A
N/A
13.30
10.04
45.1
30.4
McAile
482150043
Annual
Yes
N/A
N/A
10.09
10.04
25.0
24.9
Merced
060472510
24-hr
Yes
68
0
11.68
9.74
39.8
30.4
Merced
060470003
24-hr
No
68
0
11.81
9.82
39.0
29.8
Modest
060990006
24-hr
Yes
N/A
N/A
13.02
9.75
45.7
30.4
Modest
060990005
24-hr
No
N/A
N/A
-
-
38.8
25.8
NapaCA
060550003
Annual
Yes
N/A
N/A
10.36
10.04
25.1
24.3
NewYor
360610128
Annual
Yes
N/A
N/A
10.20
10.04
23.9
23.5
NewYor
361030002
Annual
No
N/A
N/A
7.18
7.07
18.8
18.5
NewYor
360810124
Annual
No
N/A
N/A
7.52
7.40
19.5
19.2
NewYor
360710002
Annual
No
N/A
N/A
6.95
6.84
17.5
17.2
NewYor
360610134
Annual
No
N/A
N/A
9.70
9.55
21.6
21.3
NewYor
360610079
Annual
No
N/A
N/A
8.42
8.29
22.8
22.4
NewYor
360470122
Annual
No
N/A
N/A
8.66
8.52
20.5
20.2
NewYor
360050133
Annual
No
N/A
N/A
9.05
8.91
24.0
23.6
NewYor
360050110
Annual
No
N/A
N/A
7.39
7.27
19.4
19.1
NewYor
340392003
Annual
No
N/A
N/A
8.59
8.46
23.6
23.2
NewYor
340390004
Annual
No
N/A
N/A
9.87
9.72
24.2
23.8
NewYor
340310005
Annual
No
N/A
N/A
8.42
8.29
22.2
21.9
NewYor
340292002
Annual
No
N/A
N/A
7.23
7.12
18.1
17.8
NewYor
340273001
Annual
No
N/A
N/A
6.78
6.67
17.1
16.8
NewYor
340171003
Annual
No
N/A
N/A
8.79
8.65
23.4
23.0
NewYor
340130003
Annual
No
N/A
N/A
8.89
8.75
23.8
23.4
NewYor
340030003
Annual
No
N/A
N/A
8.90
8.76
24.5
24.1
OgdenC
490110004
24-hr
Yes
29
0
7.28
7.01
32.6
30.4
OgdenC
490570002
24-hr
No
29
0
8.99
8.71
-
-
OgdenC
490030003
24-hr
No
29
0
6.35
6.10
-
-
C-114

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
Philad
420450002
Annual
Yes
86
0
11.46
10.04
26.0
22.3
Philad
421010057
Annual
No
86
0
10.86
9.12
27.0
22.5
Philad
421010055
Annual
No
86
0
11.43
9.95
27.5
23.9
Philad
421010048
Annual
No
86
0
10.27
8.70
25.6
21.1
Philad
420290100
Annual
No
86
0
9.64
7.87
23.9
19.5
Philad
340150004
Annual
No
86
0
8.33
6.99
20.6
16.9
Philad
340071007
Annual
No
86
0
8.84
7.23
21.0
17.1
Philad
340070002
Annual
No
86
0
10.19
8.40
23.5
20.2
Philad
240150003
Annual
No
86
0
8.70
6.90
22.6
17.5
Philad
100031012
Annual
No
86
0
9.04
7.21
23.0
17.7
Pittsb
420030064
24-hr
Yes
100
0
12.82
9.22
35.8
30.4
Pittsb
421290008
24-hr
No
100
0
8.65
6.04
19.6
12.9
Pittsb
421255001
24-hr
No
100
0
8.35
5.90
17.8
11.1
Pittsb
421250200
24-hr
No
100
0
8.95
6.10
19.3
13.7
Pittsb
421250005
24-hr
No
100
0
11.02
7.78
22.7
18.1
Pittsb
420070014
24-hr
No
100
0
10.11
7.38
21.9
15.2
Pittsb
420050001
24-hr
No
100
0
11.03
8.39
21.9
15.5
Pittsb
420031301
24-hr
No
100
0
11.00
7.79
24.8
19.7
Pittsb
420031008
24-hr
No
100
0
9.78
7.11
20.5
14.7
Pittsb
420030008
24-hr
No
100
0
9.50
6.81
20.5
14.2
Prinev
410130100
24-hr
Yes
N/A
N/A
8.60
6.95
37.6
30.4
ProvoO
490494001
24-hr
Yes
6
0
7.74
7.68
30.9
30.4
ProvoO
490495010
24-hr
No
6
0
6.73
6.68
-
-
ProvoO
490490002
24-hr
No
6
0
7.41
7.36
28.9
28.4
Rivers
060658005
Annual
Yes
N/A
N/A
14.48
10.04
43.2
30.0
Rivers
060658001
Annual
No
N/A
N/A
-
-
36.5
25.3
Sacram
060670006
24-hr
Yes
18
0
9.31
9.11
31.4
30.4
Sacram
061131003
24-hr
No
18
0
6.62
6.50
15.8
15.1
Sacram
060670012
24-hr
No
18
0
7.30
7.17
19.8
19.3
Sacram
060670010
24-hr
No
18
0
8.67
8.50
26.5
25.5
Sacram
060610006
24-hr
No
18
0
7.58
7.45
20.3
19.9
Sacram
060610003
24-hr
No
18
0
6.71
6.63
19.3
18.9
SaltLa
490353010
24-hr
Yes
79
0
-
-
41.5
30.3
SaltLa
490353006
24-hr
No
79
0
7.62
6.46
36.8
29.3
SaltLa
490351001
24-hr
No
79
0
7.07
5.88
32.1
23.2
SanLui
060792007
Annual
Yes
N/A
N/A
10.70
10.04
25.9
24.3
SanLui
060798002
Annual
No
N/A
N/A
5.71
5.36
-
-
SanLui
060792004
Annual
No
N/A
N/A
8.25
7.74
19.8
18.6
C-115

-------
CBSAa
Site
Controlling
Standard
Controlling
Site?
NOx &
S02
Reduction
(%)b
Primary
PM2.5
Reduction
(%)c
Base
Annual
DV
(ug m-3)
Projected
Annual
DV
(ug m-3)
Base 24-
hr DV
(H9 m"3)
Projected
24-hr DV
(H9 m"3)
SouthB
181410015
24-hr
Yes
30
0
10.45
9.68
32.5
30.4
St. Lou
290990019
Annual
Yes
N/A
N/A
10.12
10.04
22.8
22.6
St. Lou
295100094
Annual
No
N/A
N/A
9.57
9.49
23.3
23.1
St. Lou
295100093
Annual
No
N/A
N/A
-
-
23.7
23.5
St. Lou
295100085
Annual
No
N/A
N/A
10.10
10.02
23.6
23.4
St. Lou
295100007
Annual
No
N/A
N/A
9.78
9.70
23.7
23.5
St. Lou
291893001
Annual
No
N/A
N/A
9.85
9.77
22.4
22.2
Stockt
060771002
Annual
Yes
97
0
12.23
10.04
38.7
29.7
Stockt
060772010
Annual
No
97
0
10.74
8.69
37.3
28.4
Visali
061072002
24-hr
Yes
N/A
N/A
16.23
9.14
54.0
30.4
Weirto
390810017
Annual
Yes
62
0
11.75
10.02
27.2
23.8
Weirto
540090011
Annual
No
62
0
9.75
8.14
22.8
19.9
Weirto
540090005
Annual
No
62
0
10.52
8.82
22.4
18.8
Weirto
390810021
Annual
No
62
0
9.29
7.68
22.2
18.5
Wheeli
540511002
Annual
Yes
N/A
N/A
10.24
10.04
22.5
22.1
Wheeli
540690010
Annual
No
N/A
N/A
9.61
9.42
19.7
19.3
a CBSA names are the first six characters of the full CBSAs names in Table C-3.
b Percent reduction in NOx and SO2 emissions associated with just meeting the standard in this case; N/A indicates 'not
applicable' where proportional projection was used.
c Percent reduction in Primary PM2.5 emissions associated with just meeting the standard in this case; N/A indicates 'not
applicable' where proportional projection was used.
C-116

-------
REFERENCES
Abt Associates, Inc. (2014). User's guide: modeled attainment test software. Research Triangle
Park, NC, U.S. Environmental Protection Agency.
Appel, KW, Napelenok, S., Hogrefe, C., Pouliot, G., Foley, K.M., Roselle, S.J., Pleim, J.E.,
Bash, J., Pye, H.O.T., Heath, N., Murphy, B., Mathur, R. (2018). Overview and
evaluation of the Community Multiscale Air Quality (CMAQ) Modeling System Version
5.2. Air Pollution Modeling and Its Application XXV. C. Mensink, Kallos, G.
Switzerland, Springer International Publishing, Cham: 69-73.
Baxter, LK, Crooks, JL and Sacks, JD (2017). Influence of exposure differences on city-to-city
heterogeneity in PM2.5-mortality associations in US cities. Environ Health 16(1): 1.
Berrocal, VJ, Gelfand, AE and Holland, DM (2012). Space-time data fusion under error in
computer model output: an application to modeling air quality. Biometrics 68(3): 837-
848.
Chen, H, Burnett, RT, Copes, R, Kwong, JC, Villeneuve, PJ, Goldberg, MS, Brook, RD, van
Donkelaar, A, Jerrett, M, Martin, RV, Brook, JR, Kopp, A and Tu, JV (2016). Ambient
fine particulate matter and mortality among survivors of myocardial infarction:
population-based cohort study. Environ Health Perspect 124(9): 1421-1428.
Di, Q, Dai, L, Wang, Y, Zanobetti, A, Choirat, C, Schwartz, JD and Dominici, F (2017a).
Association of short-term exposure to air pollution with mortality in older adults. J Am
Med Assoc 318(24): 2446-2456.
Di, Q, Wang, Y, Zanobetti, A, Wang, Y, Koutrakis, P, Choirat, C, Dominici, F and Schwartz, JD
(2017b). Air pollution and mortality in the Medicare population. New Engl J Med
376(26): 2513-2522.
Frank, NH (2006). Retained nitrate, hydrated sulfates, and carbonaceous mass in federal
reference method fine particulate matter for six eastern US cities. Journal of the Air
Waste Management Association 56(4): 500-511.
Hart, JE, Liao, X, Hong, B, Puett, RC, Yanosky, JD, Suh, H, Kioumourtzoglou, MA,
Spiegelman, D and Laden, F (2015). The association of long-term exposure to PM2.5 on
all-cause mortality in the Nurses' Health Study and the impact of measurement-error
correction. Environ Health 14: 38.
Ito, K, Ross, Z, Zhou, J, Nadas, A, Lippmann, M and Thurston, GD (2013). National Particle
Component Toxicity (NPACT) initiative: Study 3. Time-series analysis of mortality,
hospitalizations, and ambient PM2.5 and its components. Boston, MA, Health Effects
Institute: 95-125.
Jerrett, M, Burnett, RT, Beckerman, BS, Turner, MC, Krewski, D, Thurston, G, Martin, RV, van
Donkelaar, A, Hughes, E, Shi, Y, Gapstur, SM, Thun, MJ and Pope, CA, 3rd (2013).
C-117

-------
Spatial analysis of air pollution and mortality in California. Am J Respir Crit Care Med
188(5): 593-599.
Jerrett, M, Turner, MC, Beckerman, BS, Pope, CA, van Donkelaar, A, Martin, RV, Serre, M,
Crouse, D, Gapstur, SM, Krewski, D, Diver, WR, Coogan, PF, Thurston, GD and
Burnett, RT (2016). Comparing the health effects of ambient particulate matter estimated
using ground-based versus remote sensing exposure estimates. Environ Health Perspect
125(4): 552-559.
Kelly, JT, Jang, CJ, Timin, B, Gantt, B, Reff, A, Zhu, Y, Long, S and Hanna, A (2019a). A
system for developing and projecting PM2. 5 spatial fields to correspond to just meeting
national ambient air quality standards. Atmos Environ 2: 100019.
Kelly, JT, Koplitz, SN, Baker, KR, Holder, AL, Pye, HOT, Murphy, BN, Bash, JO, Henderson,
BH, Possiel, NC, Simon, H, Eyth, AM, Jang, CJ, Phillips, S and Timin, B (2019b).
Assessing PM2.5 model performance for the conterminous U.S. with comparison to model
performance statistics from 2007-2015. Atmos Environ 214: 116872.
Kloog, I, Ridgway, B, Koutrakis, P, Coull, BA and Schwartz, JD (2013). Long- and short-term
exposure to PM2.5 and mortality: Using novel exposure models. Epidemiology 24(4):
555-561.
NRC. (2004). Air Quality Management in the United States. Washington, DC. The National
Academies Press. National Resources Council.
Pope, CA, Turner, MC, Burnett, R, Jerrett, M, Gapstur, SM, Diver, WR, Krewski, D and Brook,
RD (2015). Relationships between fine particulate air pollution, cardiometabolic
disorders and cardiovascular mortality. Circul Res 116(1): 108-U258.
Pye, HOT, Zuend, A, Fry, JL, Isaacman-Vanwertz, G, Capps, SL, Appel, KW, Foroutan, H, Xu,
Lu, Ng, N and Goldstein, AH (2018). Coupling of organic and inorganic aerosol systems
and the effect on gas-particle partitioning in the southeastern US. Atmos Chem Phys
18(1): 357-370.
Skamarock, WC, Klemp, JB, Dudhia, J, Gill, DO, Barker, DM, Wang, W and Powers, JG
(2005). A description of the advanced research WRF version 2. Boulder CO, National
Center For Atmospheric Research.
Sutradhar, R and Austin, PC (2018). Relative rates not relative risks: addressing a widespread
misinterpretation of hazard ratios. Ann Epidemiol 28(1): 54-57.
Thurston, GD, Ahn, J, Cromar, KR, Shao, Y, Reynolds, HR, Jerrett, M, Lim, CC, Shanley, R,
Park, Y and Hayes, RB (2016). Ambient particulate matter air pollution exposure and
mortality in the NIH-AARP Diet and Health Cohort. Environ Health Perspect 124(4):
484-490.
C-118

-------
Turner, MC, Jerrett, M, Pope, A, III, Krewski, D, Gapstur, SM, Diver, WR, Beckerman, BS,
Marshall, JD, Su, J, Crouse, DL and Burnett, RT (2016). Long-term ozone exposure and
mortality in a large prospective study. Am J Respir Crit Care Med 193(10): 1134-1142.
U.S. EPA. (2001). Risk Assessment Guidance for Superfund: Volume III - Part A, Process for
Conducting Probabilistic Risk Assessment. Washington, D.C. Office of Emergency and
Remedial Response. U.S. EPA. EPA 540-R-02-002. Available at:
https://www.epa.eov/sites/prodiiction/files/2015-09/documents/rags3adt complete.pdf.
U.S. EPA. (2004). An Examination of EPA Risk Assessment Principles and Practices.
Washington, D.C. Office of Research and Development. U.S. EPA. EPA/100/B-04/001.
Available at: https://semspub.epa.gov/work/10/500006305.pdf.
U.S. EPA. (2010). Quantitative Health Risk Assessment for Particulate Matter (Final Report).
Research Triangle Park, NC. Office of Air Quality Planning and Standards, Health and
Environmental Impacts Division. U.S. EPA. EPA-452/R-10-005. June 2010. Available
at: https://nepis.epa.eov/Exe/ZvPIJRL.cei?Dockey=	t.
U.S. EPA. (2015). Preamble to the integrated science assessments. Research Triangle Park, NC.
U.S. Environmental Protection Agency, Office of Research and Development, National
Center for Environmental Assessment, RTP Division. U.S. EPA. EPA/600/R-15/067.
November 2015. Available at:
https://cfpub.epa. gov/ncea/isa/recordisplav.cfm?deid=310244.
U.S. EPA. (2016). Integrated review plan for the national ambient air quality standards for
particulate matter. Research Triangle Park, NC. Office of Air Quality Planning and
Standards. U.S. EPA. EPA-452/R-16-005. December 2016. Available at:
https://www3.epa.eov/ttn/naaqs/standards/pm/data/	al-integrated-review-
plan.pdf.
U.S. EPA. (2018a). Modeling Guidance for Demonstrating Attainment of Air Quality Goals for
Ozone, PM2.5, and Regional Haze. Research Triangle Park, NC. Office of Air Quality
Planning and Standards. U.S. EPA. EPA 454/R-l8-009. Available at:
https://www3.epa.eov/ttn/scram/eiiidance/euide/03-PM-RH-Modeline Guidance-
f.
U.S. EPA. (2018b). Bayesian space-time downscaling fusion model (downscaler) -Derived
Estimates of Air Quality for 2014. Research Triangle Park, NC. Office of Air Quality
Planning and Standards. U.S. EPA. EPA-454/R-18-008. Available at:
https://www.epa.eov/hesc/rsie-related-downloadable-data-files.
U.S. EPA. (2019). Integrated Science Assessment (ISA) for Particulate Matter (Final Report).
Washington, DC. U.S. Environmental Protection Agency, Office of Research and
Development, National Center for Environmental Assessment. U.S. EPA. EPA/600/R-
19/188. December 2019. Available at: https://www.epa.gov/naaqs/particulate-matter-pm'
standards-inteerated-science-assessments-current-review.
C-119

-------
WHO. (2008). Uncertainty and Data Quality in Exposure Assessment. Geneva. The
Internaltional Programme on Chemical Safety. WHO.
https://www.who.int/ipcs/piiblications/methods/harmonization/exposure assessment.pdf.
Willis, A, Jerrett, M, Burnett, RT and Krewski, D (2003). The association between sulfate air
pollution and mortality at the county scale: an exploration of the impact of scale on a
long-term exposure study. J Toxicol Environ Health, A: Curr Iss 66(16-19): 1605-1624.
Zanobetti, A, Dominici, F, Wang, Y and Schwartz, JD (2014). A national case-crossover analysis
of the short-term effect of PM2.5 on hospitalizations and mortality in subjects with
diabetes and neurological disorders. Environ Health 13(1): 38.
C-120

-------
APPENDIX D. QUANTITATIVE ANALYSES FOR
VISIBILITY IMPAIRMENT
D.l BACKGROUND
To inform the EPA's decision in the last review on the adequacy of protection provided
by the secondary PM standards the EPA conducted a technical analysis of the relationships
between a 3-year average daily visibility metric and the 24-hour PM2.5 mass-based standard
(Kelly et al., 2012). The 3-year visibility metric was calculated as the 3-year average of the 90th
percentile of daily visibility index values.1 Light extinction coefficient (be**) values for the
visibility index were calculated using the original IMPROVE equation (Equation D-l in section
D.2.2 below), which at the time of the last review, the EPA considered to be better suited to
urban sites that were the focus of the analysis than other versions of the IMPROVE equation,
with a few modifications to the equation: excluding the coarse mass2 and sea salt3 terms in the
equation and using a multiplier of 1.6 for converting OC to OM.4
1	The visibility index is a logarithmic transformation of the light extinction coefficient, tw, the use of which ensures
that increases or decreases in light extinction coefficient always produce, respectively, increases or decreases in
visibility index (Kelly et al., 2012).
2	PM25 is the size fraction of PM responsible for most of the visibility impairment in urban areas (U.S. EPA, 2009,
section 9.2.2.2). Data available at the time of the last review suggested that, generally, PM10-2.5 was a minor
contributor to visibility impairment most of the time (U.S. EPA, 2010) although the coarse fraction may be a
major contributor in some areas in the desert southwestern region of the country. Moreover, at the time of the last
review, there were few data available from continuous PMi 0-2.5 monitors to quantify the contribution of coarse
PM to calculated light extinction.
3	In estimating light extinction in the last review, the EPA did not consider it appropriate to include the term for
hygroscopic sea salt in evaluating urban light extinction, given that sea salt is not a major contributor to light
extinction in urban areas compared with more remote coastal locations. In particular, Pitchford (2010) estimated
that the contribution of sea salt to PM2 5 light extinction was generally well below 5% for PM2 5 light extinction
greater than 24 dv (U.S. EPA, 2010, p. 3-22; U.S. EPA, 2012, p. IV-5).
4	At the time of the last review, the EPA considered the multiplier of 1.8 recommended by Pitchford et al. (2007) to
convert OC to OM for use in the revised IMPROVE equation (Equation D -2 below) to be too high for urban
environments. The composition of, and the mix of emission sources contributing to, PM2 5 differ between urban
and remote areas, and consequently, the light extinction may differ between urban and remote areas. Organic
mass in urban areas is often from local and regional sources and would have a greater percentage of fresh
emissions compared with aged emissions, which tend to be more prominent in rural areas, and a different PM
mass to OC ratio than in urban areas. The EPA also considered the multiplier of 1.4 used with the original
IMPROVE equation to be too low to adequately account for the contribution of OM to visibility impairment,
particularly in urban areas where OM concentrations tend to be higher. Based on these considerations, along with
an evaluation of the OC to OM relationship at CSN sites (2011 PA, Appendix F, section F.6), the EPA chose to
use a multiplier of 1.6 to convert OC to OM in the light extinction calculations used in the last review (U.S. EPA,
2012, pages IV-5-IV-8).
D-l

-------
Using 2008-2010 air quality data for 102 CSN network sites,5 the 2012 analysis explored
the relationship between the 3-year design values for the existing 24-hour PM2.5 standard and
values of the 3-year visibility metric.6 The analysis indicated that increases in 24-hour PM2.5
design values generally correspond to increases in the 3-year visibility metric values, and vice-
versa (78 FR 3201, January 15, 2013).The analysis also found linear correlations between the 24-
hour PM2.5 design values and the 3-year visibility metric with an average r2 value of 0.75 across
all of the sites (Kelly et al., 2012). A key implication of this analysis was that for the level
proposed by the EPA for a visibility index-based standard, the 24-hour PM2.5 standard of 35
|ig/m3 would be controlling in almost all or all instances (78 FR 3202, January 15, 2013).
D.2 ANALYSIS: METHODS AND INPUTS
Consistent with the analyses conducted in the last review described above, we have
conducted analyses examining the relationship between PM mass concentrations and estimated
light extinction in terms of a PM visibility metric. These analyses are intended to inform our
understanding of visibility impairment in the U.S. under recent air quality conditions,
particularly those conditions that meet the current standards, and our understanding of the
relative influence of various factors on light extinction. These analyses were conducted using
three versions of the IMPROVE equation (Equations D-l through D-3 below) to estimate light
extinction to better understand the influence of variability in inputs across the three equations.
This analysis included 67 monitoring sites that are geographically distributed across the U.S. in
both urban and rural areas (see Figure D-l). The data set is comprised of sites with data for the
2015-2017 period that supported a valid 24-hour PM2.5 design value7 and met strict criteria for
PM species. Light extinction at these 67 monitoring sites was calculated without the coarse
fraction in the IMPROVE equations, consistent with the analyses conducted in the last review.
For a subset of 20 of the 67 monitoring sites where PM10 data were available and met
completeness criteria, the coarse fraction was included when calculating light extinction to better
characterize the influence of coarse PM on light extinction. Results for these two sets of analyses
are presented in Figures 5-3 and 5-4 and discussed in section 5.2.1.2 of Chapter 5 and presented
in Table D-7 and Table D-8 and Figure D-2 in section D.3 below.
5	The 102 sites included in the Kelly et al. (2012) analysis were those sites that met the data completeness criteria
used for that analysis (Kelly et al., 2012, p. 15).
6	The EPA used monthly average relative humidity values rather than shorter-term (e.g., hourly) values to estimate
light extinction in the last review in order to capture seasonal variability of relative humidity and its effects on
visibility impairment. This was intended to focus more on the underlying aerosol contributions to visibility
impairment and less on the day-to-day variations in humidity (U.S. EPA, 2012, p. IV-10).
7	The design value (DV) for the standard is the metric used to determine whether areas meet or exceed the NAAQS.
A design value is a statistic that describes the air quality status of a given area relative to the NAAQS.
D-2

-------
Northeast (n = 19) T
Southeast (n=9) H
IndustMidwest (n = 13)
•	UpperMidwest (n=10)
•	Southwest (n=4)
•	Northwest (n=7)
•	SoCal (n=4)
•	Alaska (n = l)
"^¦sites violating NAAQS
Figure D-l. Locations of monitoring sites with data for 2015-2017 with a valid PM2.5 design
value and meeting completeness criteria for PM species.
D.2.1 Data Sources for Inputs to Estimate Light Extinction
D.2.1.1 Relative Humidity
Relative humidity data were downloaded from the North American Regional Reanalysis
(NARR). NARR is the National Centers for Environmental Prediction's (NCEP) high resolution
combined model and assimilated meteorological dataset. NARR is an extension of the NCEP
Global Reanalysis which is run over North American using the Eta Model (32 km) together with
the Regional Data Assimilation System. Files for 3-hour average 10 m relative humidity data for
2015-2017 are available at https://esrl.noaa.gov/psd/data/gridded/data.narr.html.
Using NARR latitudes, relative humidity data were reassigned to each grid cell from
coordinated universal time (UTC) to their closest time zone and the 3-hour relative humidity data
were then averaged to 24-hour local time averages in order to approximate the 24-hour averaging
D-3

-------
time (midnight-midnight) of the daily PM2.5 measurements. The PM2.5 and PM2.5 component
daily mass data (described in subsequent sections) were temporally and spatially matched with
the closest 24-hour average relative humidity grid cell.
D.2.1.2 PM2.5 Concentrations
The raw data for PM2.5 site-level daily mass concentrations came from an Air Quality
System (AQS)8 query of the daily site-level concentrations. Data files used were for 24-hour
average values from regulatory monitors for all sites in the U.S. for all available days (including
potential exceptional events) for 2015-2017. When a single site had multiple monitors, the
previously-determined primary monitor concentration was used. If the primary monitor value
was missing, the average of the collocated monitors was used. These data were screened so that
all days either had a valid filter-based 24-hour concentration measurement9 or at least 18 valid
hourly concentrations measurements.
D.2.1.3 Coarse PM Concentrations
The raw data for PM10-2.5 monitor-level daily mass concentrations came from an AQS
query of the daily monitor-level concentrations. Data files used were for 24-hour average
concentrations from monitors mainly in the Interagency Monitoring of Protected Visual
Environments (IMPROVE) network and NCore Multipollutant Monitoring Network. Data were
included for sites with >11 valid days for each quarter of 2015-2017.
D.2.1.4 PM2.5 Component Concentrations
The raw data for PM2.5 component concentrations for the components listed in Table D-l
came from an AQS query of the daily monitor-level concentrations. Data files used were for
filter-based, 24-hour average concentrations from monitors in the Interagency Monitoring of
Protected Visual Environments (IMPROVE) network, Chemical Speciation Network (CSN), and
NCore Multipollutant Monitoring Network. Data were included for days with valid data for all
chemical components listed in Table D-l below and for sites with >11 valid days for each
quarter of 2015-2017.
8	The Air Quality System is an EPA database of ambient air quality monitoring data (https://www.epa. gov/aas).
9	A valid filter-based 24-hour concentration measurement is one collected via FRM, and that has undergone
laboratory equilibration (at least 24 hours at standardized conditions of 20-23°C and 30-40% relative humidity)
prior to analysis (see Appendix L of 40 CFR Part 50 for the 2012 NAAQS for PM).
D-4

-------
Table D-l. PM2.5 components from AQS used in IMPROVE equations.
PM2.5 Component Drawn from AQS
AQS Parameter Code
Sulfate
88403
Nitrate
88306
OC (TORs)
88320, 88370
EC (TORs)
88321,88380
Aluminum (Al), Silica (Si), Calcium (Ca), Iron
(Fe), Titanium (Ti)
88104 (Al), 88165 (Si), 88111 (Ca), 88126
(Fe), 88161 (Ti)
Chloride, Chlorine
88115 (Chlorine), 88203 (Chloride)
a OC and EC values are based on the thermal optical reflectance (TOR) analytical method,
which replaced the NIOSH 5040-like thermal optical transmittance (TOT) method in the CSN
network after 2009 (Spada and Hyslop, 2018).
D.2.1.5 24-Hour PM2.5 Design Values
Files for 24-hour PM2.5 design values for 2015-2017 are located at
https://www.epa.gov/air-trends/air-qualitv-design-values. Data handling of the 2015-2017 PM2.5
design values is described in Appendix N of 40 CFR Part 50 for the 2012 National Ambient Air
Quality Standards (NAAQS) for Particulate Matter (PM).
D.2.1.6 24-Hour PM10 Design Values
Files for 24-hour PM10 design values for 2015-2017 are located at
https://www.epa.gov/air-trends/air-qualitv-design-values. Data handling of the 2015-2017 PM10
design values is described in Appendix K of 40 CFR Part 50.
D.2.1.7 Annual PM2.5 Design Values
Files for annual PM2.5 design values for 2015-2017 are located at
https://www.epa.gov/air-trends/air-qualitv-design-values. Data handling of the 2015-2017 PM2.5
design values is described in Appendix N of 40 CFR Part 50 for the 2012 National Ambient Air
Quality Standards (NAAQS) for Particulate Matter (PM).
D.2.2 Calculating Light Extinction for Visibility Impairment Analyses
For all days with a valid relative humidity value, PM2.5 mass concentration, and all
chemical components listed in Table D-l, daily light extinction was calculated using three
versions of the IMPROVE equation, as shown below. Formulas for derivation of the equation
variables from the AQS parameters are presented in Table D-6.
D-5

-------
Original IMPROVE Equation (Malm et al., 1994):
bext = 3f(RH)([AS] + [AN]) + 4 [OM] + 10 [EC] + 1[F5] + 0.6 [CM] + 10
Equation D-l
where:
[AS] is concentration in |ig/m3 of ammonium sulfate,
[AN] is concentration in |ig/m3 of ammonium nitrate,
[OM] is concentration in |ig/m3 of organic matter,
[EC] is concentration in |ig/m3 of elemental carbon,
[FS] is concentration in |ig/m3 of fine soil,
[CM] is concentrations in |ig/m3 of coarse mass, and
f(RH) is the relative-humidity-dependent water growth function, assigned values as shown
in Table D-2:
Table D-2. Relatively-humidity-dependent water growth function for use in the original
IMPROVE equation.
RH (%)
1-36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
f(RH)
1
1.02
1.04
1.06
1.08
1.1
1.13
1.15
1.18
1.2
1.23
1.26
1.28
1.31
1.34
1.37
1.41
1.44
1.47
1.51
1.54






















RH (%)
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
f(RH)
1.58
1.62
1.66
1.7
1.74
1.79
1.83
1.88
1.93
1.98
2.03
2.08
2.14
2.19
2.25
2.31
2.37
2.43
2.5
2.56
2.63






















RH (%)
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98a
f(RH)
2.7
2.78
2.86
2.94
3.03
3.12
3.22
3.33
3.45
3.58
3.74
3.93
4.16
4.45
4.84
5.37
6.16
7.4
9.59
14.1
26.4
Note: See fRHOriainallMPROVE.csv file from http://vista.cira.colostate.edu/lmprove/the-improve-alaorithm/ (Maim et al., 1994).
a For our application, any relative humidity values greater than 98% were assigned the f(RH) value associated with 98%, the highest
value available for the relative humidity function.
D-6

-------
The various coefficients are the empirically derived extinction efficiency (mass scattering and
absorption) coefficients, as originally specified by Malm et al. (1994).
Revised IMPROVE Equation (Pitchford et al., 2007):
bext = 2.2fs(RH)[small sulfate] + 4.8/L(RH)[large sulfate] + 2.4fs(RH)[small nitrate]
+ 5.IfL(RH) [large nitrate] + 2.8[small OM] + 6.1 [large OM] + 10 [EC]
+ 1[FS] + 1.7 fss(RH)[SS] + 0.6 [CM] + 10
Equation D-2
where:
[small sulfate], [large sulfate], [small nitrate], [large nitrate], [small OM] and [large OM]
are defined as follows in Table D-3:
Table D-3. Values for use in the revised IMPROVE equation for small and large sulfate,
nitrate, and organic matter concentrations.

If [ ] >20
If [ ] <20
Large sulfate
[AS1
[AS1+20
Small sulfate
0
[AS1 - (fASl+20)
Large nitrate
[ANl
[AN1-20
Small nitrate
0
[ANl - (fANl+20)
Large OM
fOMl
[OM1-20
Small OM
0
[OM1 - ([OM1-20)
Note: [AS], [AN] and [OM] are defined as for Equation D-1.
[SS] is sea salt; and,
fss(RH), fs(RH), and fiXRH) are defined as shown in Table D-4:
D-7

-------
Table D-4. Relatively-humidity-dependent water growth function for sea salt, small
particles, and large particles for use in the revised IMPROVE equation.
RH (%)
1-36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
fss(RH)
1
1
1
1
1
1
1
1
1
1
1
2.3584
2.3799
2.4204
2.4488
fs(RH)
1
1.38
1.4
1.42
1.44
1.46
1.48
1.49
1.51
1.53
1.55
1.57
1.59
1.62
1.64
fi_(RH)
1
1.31
1.32
1.34
1.35
1.36
1.38
1.39
1.41
1.42
1.44
1.45
1.47
1.49
1.5
















RH (%)
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
fss(RH)
2.4848
2.5006
2.5052
2.5279
2.5614
2.5848
2.5888
2.616
2.6581
2.6866
2.7341
2.7834
2.8272
2.8287
2.8594
fs(RH)
1.66
1.68
1.71
1.73
1.76
1.78
1.81
1.83
1.86
1.89
1.92
1.95
1.99
2.02
2.06
fi_(RH)
1.52
1.54
1.55
1.57
1.59
1.61
1.63
1.65
1.67
1.69
1.71
1.73
1.75
1.78
1.8
















RH (%)
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
fss(RH)
2.8943
2.9105
2.9451
3.0105
3.0485
3.1269
3.1729
3.2055
3.2459
3.2673
3.3478
3.4174
3.5202
3.5744
3.6329
fs(RH)
2.09
2.13
2.17
2.22
2.26
2.31
2.36
2.41
2.47
2.54
2.6
2.67
2.75
2.84
2.93
fi_(RH)
1.83
1.86
1.89
1.92
1.95
1.98
2.01
2.05
2.09
2.13
2.18
2.22
2.27
2.33
2.39
















RH (%)
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95 *
fss(RH)
3.6905
3.808
3.9505
4.0398
4.1127
4.2824
4.494
4.6078
4.8573
5.1165
5.3844
5.7457
6.1704
6.7178
7.3492
fs(RH)
3.03
3.15
3.27
3.42
3.58
3.76
3.98
4.23
4.53
4.9
5.35
5.93
6.71
7.78
9.34
fi_(RH)
2.45
2.52
2.6
2.69
2.79
2.9
3.02
3.16
3.33
3.53
3.77
4.06
4.43
4.92
5.57
Note: See fRHRevisedlMPROVE.csv file from htto://vista.cira.colostate.edu/lmprove/the-improve-alaorithm/ (Pitchford et al„
2007).
a For our application, any relative humidity values greater than 95% were assigned the f(RH) value associated with 95%, the
highest value available for the relative humidity function.
and
[EC], [FS] and [CM] are defined as for Equation D-l.
This equation is generally dividing PM components into small and large particle sizes10 with
separate mass scattering efficiencies and hygroscopic growth functions for each size (included in
the equation as fs(RH) for small particles, fiXRH) for large particles, and fss(RH) for sea salt).
111 The large mode for sulfate, nitrate, and OM represents aged and/or cloud processed particles, whereas the small
mode represents freshly formed particles. These size modes are described by log-normal mass size distributions
with geometric mean diameters and geometric standard deviations of 0.2 |im and 2.2 for small mode and 0.5 |im
and 1.5 for the large mode, respectively.
D-8

-------
Lowenthal and Kumar (2016) Equation:
bext = 2.2 fs(RH) [small sulfate] + 4.8/L(RH)[large sulfate] + 2.4fs(RH)[small nitrate]
+ 5.IfL(RH) [large nitrate] + 2.8fs(RH)0M [small OM]
+ 6.lfL(RH)0M [large OM] + 10 [EC] + 1[F5] + 1.7fss(RH) [55] + 0.6[CM]
+ 10
Equation D-3
where:
fs(RH)oM and fiXRH)oM are the relative-humidity-dependent water growth function for small and
large organic matter, respectively, as defined in Table D-5 below.
Table D-5. Relatively-humidity-dependent water growth function for small organic matter
and large organic matter for use in the original IMPROVE equation.
RH (%)
0-29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
fs(RH)oM
1.000
1.321
1.325
1.329
1.333
1.337
1.340
1.343
1.346
1.349
1.352
1.354
1.356
1.358
1.360
1.362
1.364
Fl(RH)om
1.000
1.267
1.271
1.274
1.278
1.280
1.283
1.286
1.288
1.290
1.292
1.294
1.296
1.297
1.299
1.300
1.302


















RH (%)
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
fs(RH)oM
1.366
1.368
1.369
1.371
1.373
1.75
1.377
1.379
1.382
1.384
1.387
1.390
1.393
1.397
1.400
1.404
1.409
fs(RH)oM
1.303
1.305
1.306
1.308
1.309
1.311
1.306
1.308
1.309
1.311
1.313
1.314
1.316
1.318
1.320
1.323
1.325


















RH (%)
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
fs(RH)oM
1.413
1.419
1.424
1.430
1.437
1.444
1.452
1.460
1.469
1.478
1.489
1.500
1.511
1.524
1.537
1.51
1.566
fs(RH)oM
1.328
1.331
1.334
1.338
1.342
1.346
1.350
1.355
1.385
1.393
1.401
1.409
1.418
1.428
1.438
1.449
1.461


















RH (%)
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95 a

fs(RH)oM
1.582
1.599
1.617
1.637
1.657
1.679
1.703
1.727
1.754
1.782
1.812
1.843
1.877
1.912
1.950
1.989

fs(RH)oM
1.473
1.486
1.500
1.515
1.531
1.548
1.566
1.585
1.605
1.626
1.648
1.672
1.696
1.722
1.750
1.779

Note: See Table 1 in Lowenthal and Kumar (2016).
a For our application, any relative humidity values greater than 95% were assigned the f(RH) value associated with 95%, the highest
value available for the relative humidity function.
and
[small sulfate], [large sulfate], [small nitrate], [large nitrate], [small OM], [large OM], [EC],
[FS], [SS], [CM], fs(RH), fiXRH) and fss(RH) are defined as above for Equation D-2.
This equation updates the multiplier for estimating the concentration organic matter, [OM], from
the concentration of organic carbon to 2.1 and incorporates fs(RH)oM and fL(RH)oM representing
water absorption by soluble organic matter as a function of relative humidity for small and large
organic matter, respectively.
D-9

-------
Based on each equation, site-specific visibility metrics were derived for each site as
follows. Daily light extinction values were derived for 2015, 2016, and 2017, the 90th percentile
of daily values for each year was calculated, and the three years of values were averaged. The 3-
year averages of the 90th percentiles of daily light extinction values were paired with the 2015-
2017 PM2.5 24-hour design values for each site having valid data for both statistics.
Table D-6. Derivation of equation variables from AQS PM2.5 component concentrations.
Equation Variable
How Calculated from AQS Parameter Values
Ammonium Sulfate
All three equations: 1.375x[Sulfate]A
Ammonium Nitrate
All three equations: 1.29x[Nitrate]B
Organic Matter
Original IMPROVE equation: 1.6x[OC]c
Revised IMPROVE equation: 1.6x[OC]c
Lowenthal and Kumar (2016) equation: 2.1x[OC]
Elemental Carbon
[ECl
Fine Soil
All three equations:D
2.2x[All+2.49x[Sil+1,63x[Cal+2.42x[Fel+1,94x[Til
Sea Salt
Revised IMPROVE and Lowenthal and Kumar, 2016 equations:0
1.8x[Chloride]
1.8x[Chlorinej (if chloride is missing)
A This formula is based on molar molecular weights of ammonium sulfate and sulfate (Malm et al., 1994).
B This formula is based on molar molecular weights of ammonium nitrate and nitrate (Malm et al., 1994).
c See footnote 4 earlier in this appendix.
D This formula is documented in Malm et al. (1994).
D.3 SUMMARY OF RESULTS
Results for the visibility impairment analyses are discussed in section 5.2.1.2 of Chapter
5. Table D-7 presents the 24-hour PM2.5, 24-hour PM10 design values, and 3-year visibility
metrics based on light extinction calculations using the three versions of the IMPROVE equation
with the coarse mass fraction excluded for the 67 monitoring sites included in the analyses. Table
D-8 presents the 24-hour PM2.5 and 24-hour PM10 design values, along with the 3-year visibility
metrics based on light extinction calculations using the three versions of the IMPROVE equation
with and without the coarse mass fraction for the subset of 20 monitoring sites with coarse PM
monitoring data that meet the completeness criteria as described above. Figure 5-3 and 5-4 in
Chapter 5 show a comparison of the 3-year visibility metric and the 24-hour PM2.5 design values
for the 67 monitoring sites in the analyses where light extinction was calculated using the
D-10

-------
original IMPROVE equation11 and the Lowenthal and Kumar IMPROVE equation.12 Figure D-2
below presents the 3-year visibility metric and the 24-hour PM2.5 design values for the 67
monitoring sites with light extinction calculated using the revised IMPROVE equation.13
11	For this analysis, the original IMPROVE equation in Equation D-l was modified to use a 1.6 multiplier to convert
OC to OM and to remove the coarse mass fraction from the light extinction calculation, consistent with the
modifications in the last review.
12	For this analysis, the Lowenthal and Kumar IMPROVE equation in Equation D -3 was modified to remove the
coarse mass fraction from the light extinction calculation.
13	For this analysis, the revised IMPROVE equation in Equation D-2 was modified to use a 1.6 multiplier to convert
OC to OM and to remove the coarse mass fraction from the light extinction calculation, consistent with the
modifications in the last review.
D-l 1

-------
Table D-7. Summary of 24-hour PM2.5, 24-hour PM10, and annual PM2.5 design values, and 3-year visibility metrics at 67
monitoring sites (2015-2017).
Monitor ID
State
Region
24-hour
PM2.5
Design
Value
(Hg/m3)A
24-hour PM10
Design Value
(number of
exceedances)B
c
Annual PM25
Design Value
(Hg/m3)D
3-year Visibility Metric (deciviews)E
Original
IMPROVE
Equation F
Revised
IMPROVE
Equation G
Lowenthal &
Kumar
IMPROVE
Equation H
010730023
Alabama
Southeast
22
0
10.4
21
21
26
020900034
Alaska
Alaska
35
0
9.5
27
27
31
040139997
Arizona
Southwest
21
0.3
7.1
18
18
21
040191028
Arizona
Southwest
12

5.5
13
13
15
051190007
Arkansas
Southeast
19
0
9.4
20
20
24
060190011
California
SoCal
54
0.3
14
25
27
31
060371103
California
SoCal
32
0
12.1
24
25
27
060658001
California
SoCal
34
0
12.3
23
25
28
060670006
California
Northwest
34
0
9.6
24
25
30
060850005
California
Northwest
27
0
9.3
22
22
26
090050005
Connecticut
Northeast
13
0
4.6
17
16
18
110010043
District of
Columbia
Northeast
21
0
9.2
23
22
25
120573002
Florida
Southeast
17
0
7.4
18
17
20
130890002
Georgia
Southeast
19
0
9.0
20
19
24
160010010
Idaho
Northwest
31

7.6
23
23
26
170191001
Illinois
IndustrialMidwest
17

7.6
21
20
21
170314201
Illinois
IndustrialMidwest
21
0
8.4
23
23
25
180970078
Indiana
IndustrialMidwest
21
0
9.1
23
23
26
191370002
Iowa
UpperMidwest
16

6.5
18
17
19
191630015
Iowa
IndustrialMidwest
20
0
8.2
22
21
23
191770006
Iowa
UpperMidwest
18
0
6.9
21
20
22
D-12

-------
202090021
Kansas
UpperMidwest
21

00
CO
21
21
24
211110067
Kentucky
IndustrialMidwest
19

8.6
22
21
24
220330009
Louisiana
Southeast
20
0
9.0
21
20
24
230090103
Maine
Northeast
12
0
4.1
18
16
19
240053001
Maryland
Northeast
23

8.9
23
23
26
240230002
Maryland
IndustrialMidwest
14

5.5
17
17
18
240330030
Maryland
Northeast
18
0
8.4
21
20
24
250130008
Massachusetts
Northeast
14

5.7
20
19
23
250250042
Massachusetts
Northeast
16
0
7.0
20
19
22
260810020
Michigan
IndustrialMidwest
23
0
8.5
23
23
25
261630001
Michigan
IndustrialMidwest
22
0
8.9
24
24
26
270031002
Minnesota
UpperMidwest
18
0
6.7
20
20
23
270530963
Minnesota
UpperMidwest
18

7.2
22
22
24
270750005
Minnesota
IndustrialMidwest
12

4.0
15
15
17
295100085
Missouri
IndustrialMidwest
20
0
8.9
22
21
24
300490004
Montana
Northwest
33

4.1
15
15
20
310550019
Nebraska
UpperMidwest
20
0
8.9
19
18
20
320030540
Nevada
SoCal
23
0.7
8.2
19
19
22
320310016
Nevada
Northwest
20
0
7.2
18
18
22
330115001
New Hampshire
Northeast
12

4.6
14
13
15
330150018
New Hampshire
Northeast
14

5.1
18
17
19
340010006
New Jersey
Northeast
15

6.8
19
19
20
340130003
New Jersey
Northeast
20
0
8.6
23
23
26
340390004
New Jersey
Northeast
23

9.7
24
24
27
350010023
New Mexico
Southwest
18
0
5.8
15
15
18
360050110
New York
Northeast
19

6.9
23
23
25
360551007
New York
Northeast
16

6.5
21
21
23
360610134
New York
Northeast
21

9.3
24
24
27
360810124
New York
Northeast
19

7.3
22
21
24
361010003
New York
Northeast
12

5.0
18
17
19
371190041
North Carolina
Southeast
17

8.5
19
19
23
D-13

-------
371830014
North Carolina
Southeast
18

8.8
19
18
22
380070002
North Dakota
UpperMidwest
18
0
4.1
14
13
15
380130004
North Dakota
UpperMidwest
24
0
4.3
18
18
18
390610040
Ohio
IndustrialMidwest
20
0
8.9
23
22
24
391351001
Ohio
IndustrialMidwest
17

7.7
22
21
23
460330132
South Dakota
UpperMidwest
16
0
3.7
12
11
14
460710001
South Dakota
UpperMidwest
15
0
3.5
12
11
14
471570075
Tennessee
Southeast
15

7.6
19
18
21
481410044
Texas
Southwest
23

8.9
17
17
20
482011039
Texas
Southeast
20
0
8.6
21
21
24
500070007
Vermont
Northeast
10

3.2
16
15
17
510870014
Virginia
Northeast
16
0
7.4
20
19
24
530330080
Washington
Northwest
20

6.4
20
20
23
550270001
Wisconsin
IndustrialMidwest
18
0
6.8
22
22
24
560210100
Wyoming
Northwest
14

4.1
13
12
15
A The 24-hour PM2.5 design value is the 3-year average of the 98th percentile of daily PM2.5 mass concentrations. The current 24-hour PM2.5 NAAQS is set at a level of 35
|jg/m3.
B The 24-hour PM10 design value is not to be exceeded more than once per year on average over three years. The current 24-hour PM10 NAAQS is set at a level of 150
|jg/m3.
c For some monitoring locations, PM10 design values are not available because of a lack of collocated PM10 monitoring at the site or insufficient data after applying
completeness criteria for calculating PM10 design values.
D The annual PM2.5 design value is the annual mean, averaged over three years. The current secondary annual PM2.5 NAAQS is set at a level of 15.0 |jg/m3.
E The 3-year visibility metric is the 3-year average of the 90th percentile of daily light extinction. In the last review, the target level of protection identified for the 3-year visibility
metric was 30 deciviews.
F The original IMPROVE equation in Equation D-1 was modified to use a 1.6 multiplier to convert OC to OM and to remove the coarse mass fraction from the light extinction
calculation, consistent with the modifications in the last review.
G The revised IMPROVE equation in Equation D-2 was modified to use a 1.6 multiplier to convert OC to OM and to remove the coarse mass fraction from the light extinction
calculation, consistent with the modifications in the last review.
H The Lowenthal and Kumar IMPROVE equation in Equation D-3 was modified to remove the coarse mass fraction from the light extinction calculation.
D-14

-------
Table D-8. Summary of 24-hour PM2.5, 24-hour PM10 and annual PM2.5 design values, and 3-year visibility metrics at 20
monitoring sites with collocated PM2.5 and PM10 monitoring data (2015-2017).
Monitor ID
State
Region
24-hour
PM2.5
Design
Value
(Mg/m3)A
24-hour PM10
Design Value
(number of
exceedances)
BC
Annual PM2.5
Design Value
(Mg/m3)D
3-year Visibility Metric (deciviews)E
Original IMPROVE
Equation F
Revised IMPROVE
Equation G
Lowenthal & Kumar
IMPROVE Equation
Without
[CM]H
With
[CM] i
Without
[CM]H
With
[CM] i
Without
[CM]H
With
[CM] i
051190007
Arkansas
Southeast
19
0
9.4
20
21
20
21
24
24
060670006
California
Northwest
34
0
9.6
24
25
25
25
30
29
060850005
California
Northwest
27
0
9.3
22
23
22
23
26
27
120573002
Florida
Southeast
17
0
7.4
18
19
17
18
20
20
160010010
Idaho
Northwest
31

7.6
23
22
23
23
26
25
180970078
Indiana
Industrial Midwest
21
0
9.1
23
24
23
23
26
26
191630015
Iowa
Industrial Midwest
20
0
8.2
22
22
21
22
23
24
211110067
Kentucky
Industrial Midwest
19

8.6
22
22
21
22
24
24
230090103
Maine
Northeast
12
0
4.1
18
19
16
17
19
19
250250042
Massachusetts
Northeast
16
0
7.0
20
20
19
20
22
22
260810020
Michigan
Industrial Midwest
23
0
8.5
23
23
23
23
25
26
261630001
Michigan
Industrial Midwest
22
0
8.9
24
25
24
25
26
27
320310016
Nevada
Northwest
20
0
7.2
18
19
18
19
22
23
340130003
New Jersey
Northeast
20
0
8.6
23
24
23
24
22
26
390610040
Ohio
Industrial Midwest
20
0
8.9
23
24
22
23
24
25
391351001
Ohio
Industrial Midwest
17

7.7
22
22
21
21
23
23
471570075
Tennessee
Southeast
15

7.6
19
20
18
19
21
22
500070007
Vermont
Northeast
10

3.2
16
16
15
15
17
17
510870014
Virginia
Northeast
16
0
7.4
20
20
19
20
24
24
530330080
Washington
Northwest
20

6.4
20
21
20
20
23
25
A The 24-hour PM2.5 design value is the 3-year average of the 98th percentile of daily PM2.5 mass concentrations. The current secondary 24-hour PM2.5 NAAQS is set at a level of 35
|jg/m3.
B The 24-hour PM10 design value is not to be exceeded more than once per year on average over three years. The current secondary 24-hour PM10 NAAQS is set at a level of 150
|jg/m3.
D-15

-------
c For some monitoring locations, PM10 design values are not available because of a lack of collocated PM10 monitoring at the site or insufficient data after applying completeness
criteria for calculating PM10 design values.
D The annual PM2.5 design value is the annual mean, averaged over three years. The current secondary annual PM2.5 NAAQS is set at a level of 15.0 |jg/m3.
E The 3-year visibility metric is the 3-year average of the 90th percentile of daily light extinction. In the last review, the target level of protection identified for the 3-year visibility metric
was 30 deciviews.
F The original IMPROVE equation in Equation D-1 was modified to use a 1.6 multiplier to convert OC to OM, consistent with the modifications in the last review.
G The revised IMPROVE equation in Equation D-2 was modified to use a 1.6 multiplier to convert OC to OM, consistent with the modifications in the last review.
H Light extinction was calculated with the coarse mass fraction removed from the equation.
1 Although the addition of coarse mass increases the daily extinction calculation, it is possible for the 90th percentile value to decrease due to a different set of days having valid
measurements of both PM2.5 chemical composition and PM10-2.5.	
D-16

-------
u
c
4->
x
cu
c
° ^
§
CL>
*>
"u
cu
.2» ,	ro
=	cu
fO	>*
"a	(V)
*¦*¦—	|
°	cu
CU	>
=	O
4->
C TJ
cu cu
u cn
u 2
a. cu
b*
as
40
35
£ 30
25
20
15
10
0
• •
• •
• •
• •,
•	Northeast (n=19)
•	Southeast (n=9)
•	IndustMidwest (n = 13)
•	UpperMidwest (n=10)
Southwest (n=4)
Northwest (n = 7)
•	SoCal (n=4)
•	Alaska (n = l)
0 5 10 15 20 25 30 35 40 45 50 55 60
98th percentile of daily PM25 concentration,
averaged over 3 years (|jg m 3)
Figure D-2. Comparison of 90th percentile of daily light extinction, averaged over three
years, and 98th percentile of daily PM2.5 concentrations, averaged over three years, for
2015-2017 using the revised IMPROVE equation. (Note: Dashed lines indicate the level of
current 24-hour PM2.5 standard (35 |ig/m3) and the target level of protection identified for the
3-year visibility metric (30 dv).)
D-17

-------
REFERENCES
Abt Associates, Inc. (2001). Assessing public opinions on visibility impairment due to air
pollution: Summary report. Research Triangle Park, NC, U.S. Environmental Protection
Agency.
Kelly, J, Schmidt, M, Frank, N, Timin, B, Solomon, D and Venkatesh, R. (2012). Memorandum
to PM NAAQS Review Docket (EPA-HQ-OAR-2007-0492). Technical Analyses to
Support Surrogacy Policy for Proposed Secondary PM2.5 NAAQS under NSR/PSD
Programs. June 14, 2012. . Docket ID No. EPA-HQ-OAR-2007-0492. Research Triangle
Park, NC. Office of Air Quality Planning and Standards. Available at:
https://www3.epa.gov/ttn/naaqs/standards/pm/data/20120614Kellv.pdf.
Lowenthal, DH and Kumar, N (2016). Evaluation of the IMPROVE Equation for estimating
aerosol light extinction. J Air Waste Manage Assoc 66(7): 726-737.
Malm, WC, Sisler, JF, Huffman, D, Eldred, RA and Cahill, TA (1994). Spatial and seasonal
trends in particle concentration and optical extinction in the United States. J Geophys Res
99(D1): 1347-1370.
Pitchford, M, Maim, W, Schichtel, B, Kumar, N, Lowenthal, D and Hand, J (2007). Revised
algorithm for estimating light extinction from IMPROVE particle speciation data. J Air
Waste Manage Assoc 57(11): 1326-1336.
Pitchford, M. (2010). Memorandum to PM NAAQS Review Docket (EPA-HQ-OAR-2007-
0492). Assessment of the Use of Speciated PM2.5 Mass-Calculated Light Extinction as a
Secondary PM NAAQS Indicator of Visibility. . November 17, 2010. Docket ID No.
EPA-HQ-OAR-2007-0492. Research Triangle Park, NC. Office of Air Quality Planning
and Standards. Available at:
https://www3.epa. gov/ttn/naaq s/standards/pm/data/Pitchford 11172010.pdf.
Spada, NJ and Hyslop, NP (2018). Comparison of elemental and organic carbon measurements
between IMPROVE and CSN before and after method transitions. Atmos Environ 178:
173-180.
U.S. EPA. (2009). Integrated Science Assessment for Particulate Matter (Final Report). Research
Triangle Park, NC. Office of Research and Development, National Center for
Environmental Assessment. U.S. EPA. EPA-600/R-08-139F. December 2009. Available
at: https://cfpub.epa.gov/ncea/risk/recordisplav.cfm?deid=216546.
U.S. EPA. (2010). Particulate Matter Urban-Focused Visibility Assessment (Final Document).
Research Triangle Park, NC. Office of Air Quality Planning and Standards, Health and
Environmental Impacts Division. U.S. EPA. EPA-452/R-10-004 July 2010. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100FQ5D.txt.
U.S. EPA. (2012). Responses to Significant Comments on the 2012 Proposed Rule on the
National Ambient Air Quality Standards for Particulate Matter (June 29, 2012; 77 FR
D-18

-------
38890). Research Triangle Park, NC. U.S. EPA. Docket ID No. EPA-HQ-OAR-2007-
0492. Available at: https://www3.epa.gov/ttn/naaqs/standards/pm/data/20121214rtc.pdf.
D-19

-------
ATTACHMENT: SUMMARY OF VISIBILITY PREFERENCE
STUDIES
The preference studies available at the time of the last review were conducted in four
urban areas. Three western preference studies were available, including one in Denver, Colorado
(Ely et al., 1991), one in the lower Fraser River valley near Vancouver, British Columbia,
Canada (Pryor, 1996), and one in Phoenix, Arizona (BBC Research & Consulting, 2003). A pilot
focus group study was also conducted for Washington, DC (Abt Associates, 2001), and a
replicate study with 26 participants was also conducted for Washington, DC (Smith and Howell,
2009).14 Study specific details for these preference studies are shown in Table D-9.
14 The replicate study with 26 participants was one test group of three included in Smith and Howell (2009). This
study also included two additional test groups to assess varying light extinction conditions using the same scene
as was used in the first test group. Study details in Table D-9 reflect all three test groups included in the study.
However, for reasons described in section 2.5.2 of U.S. EPA (2010), results from the other two test groups were
not included in the EPA's evaluation of levels of acceptable visibility impairment from the preference studies.
D-20

-------
Table D-9. Summary of visibility preference studies. (Adapted from Table 9-2 in U.S. EPA,
2009).

Denver, CO
Phoenix, AZ
Vancouver, British
Columbia
Washington, DC
Washington, DC
Report Date
1991
2003
1996
2001
2009
Duration of

45 minutes
50 minutes
2 hours

session





Compensation
None
$50
None
$50
None
# focus group
16a
27 b
4
1
3 tests
sessions





# participants
214
385
180
9
64
Age range
Adults
18-65+
University students
27-58
Adults
Annual or
Wintertime
Annual
Summertime
Annual
Annual
seasonal





# and type of
scene
Single scene of
downtown
Single scene of
downtown
Single scene from
each of two suburbs in
Single scene of
Potomac River,
Single scene of
DC Mall and
presented
Denver with the
Phoenix with the
the lower Fraser River
Washington Mall
downtown, 8 km

mountains in the
Estrella
valley - Chilliwack and
and downtown
maximum sight

south in the
Mountains in the
Abbotsfordc
Washington, DC,


background
background, 42
km max. distance

8 km max. sight

# total visibility
20 conditions (+
21 conditions (+
20 conditions (10 from
20 conditions (+
22 conditions
conditions
presented
5 duplicates)
4 duplicates)
each city)
5 duplicates)

Source of
Actual photos
Win Haze
Actual photos taken at
Win Haze
WinHaze
slides
taken between
9am and 3pm

1pm or 4pm


Medium of
presentation
Slide projection
Slide projection
Slide projection
Slide projection
Slide projection
Ranking scale
used
7 point scale
7 point scale
7 point scale
7 point scale
7 point scale
Visibility range
presented (dv)
11-40
15-35
Chilliwack: 13-25
Abbotsford: 13.5-31.5
9-38
9-45
Health issue
directions
Ignore potential
health impacts;
visibility only
Judge solely on
visibility, do not
consider health
Judge solely on
visibility, do not
consider health
Health never
mentioned,
"Focus only on
visibility"
Health never
mentioned,
"Focus only on
visibility"
Key questions
•Rank VAQ (1-7
•Rank VAQ (1-7
•Rank VAQ (1-7
•Rank VAQ (1-7
•Rank VAQ (1-7
asked
scale)
•Is each slide
scale)
•Is each slide
scale)
•Is each slide
scale)
•Is each slide
scale)
•Is each slide

"acceptable"
"acceptable"
"acceptable"
"acceptable"
"acceptable"

•"How much
haze is too
much?"
•How many days
a year would this
picture be
"acceptable"

•If this hazy, how
many hours
would it be
acceptable (3
slides only)
•Valuation
question

Mean dv found
20.3
23-25
Chilliwack: -23
-20
O
CO
I
"acceptable"


Abbotsford: -19
(range 20-25)

a No preference data were collected at a 17th focus group session due ot a slide projector malfunction.
b The 27 focus groups were conducted in 6 neighborhood locations in Phoenix, with 3 focus groups held in Spanish.
c Chilliwack scene includes downtown buildings in the foreground with mountains in the background up to 65 km away. Abbotsford scene
has fewer manmade objects in the foreground and is primarily a more rural scene with mountains in the background up to 55 km away.
D-21

-------
REFERENCES
Abt Associates, Inc. (2005). Particulate matter health risk assessment for selected urban areas:
Draft report. Research Triangle Park, NC, U.S. Environmental Protection Agency: 164.
BBC Research & Consulting (2003). Phoenix area visibility survey. Denver, CO.
Ely, DW, Leary, JT, Stewart, TR and Ross, DM (1991). The establishment of the Denver
Visibility Standard. Denver, Colorado, Colorado Department of Health.
Pryor, SC (1996). Assessing public perception of visibility for standard setting exercises. Atmos
Environ 30(15): 2705-2716.
U.S. EPA. (2009). Integrated Science Assessment for Particulate Matter (Final Report). Research
Triangle Park, NC. Office of Research and Development, National Center for
Environmental Assessment. U.S. EPA. EPA-600/R-08-139F. December 2009. Available
at: https://cfpub.epa.gov/ncea/risk/recordisplav.cfm?deid=216546.
U.S. EPA. (2010). Particulate Matter Urban-Focused Visibility Assessment (Final Document).
Research Triangle Park, NC. Office of Air Quality Planning and Standards, Health and
Environmental Impacts Division. U.S. EPA. EPA-452/R-10-004 July 2010. Available at:
https://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=P 100FQ5D.txt.
D-22

-------
United States	Office of Air Quality Planning and Standards	Publication No. EPA-452/R-20-002
Environmental Protection	Health and Environmental Impacts Division	January 2020
Agency	Research Triangle Park, NC

-------