-------
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
-------